Licensing of Vowel Length in Czech
POTSDAM LINGUISTIC INVESTIGATIONS
POTSDAMER LINGUISTISCHE UNTERSUCHUNGEN
RECHERCHES LINGUISTIQUES À POTSDAM
Edited by / Herausgegeben von / Edité par
Peter Kosta, Gerda Haßler, Teodora Radeva-Bork, Lilia Schürcks,
Nadine Thielemann and/und/et Vladislava Maria Warditz
Editorial Board:
Tilman Berger (University of Tübingen, Germany)
Željko Bošković (University of Connecticut, USA)
Sarah Dessì Schmid (University of Tübingen, Germany)
Anna Maria di Sciullo (UQAM / Université du Québec à Montreál, Montreal, Canada)
Steven Franks (Indiana University, Bloomington, USA)
Atle Grønn (University of Oslo, Norway)
Holger Kuße (Dresden University of Technology, Germany)
Hans-Georg Wolf (University of Potsdam, Germany)
Ghil'ad Zuckermann (University of Adelaide, Australia)
Vol./Bd 22
Zu Qualitätssicherung und Peer Review
der vorliegenden Publikation
Notes on the quality assurance and peer
review of this publication
Die Qualität der in dieser Reihe
erscheinenden Arbeiten wird vor der
Publikation durch einen externen, von der
Herausgeberschaft benannten Gutachter
im Double Blind Verfahren geprüft. Dabei
ist der Autor der Arbeit dem Gutachter
während der Prüfung namentlich nicht
bekannt; der Gutachter bleibt anonym.
Prior to publication, the quality of the
work published in this series is double
blind reviewed by an external referee
appointed by the editorship. The
referee is not aware of the author's
name when performing the review;
the referee's name is not disclosed.
Markéta Ziková
Licensing of Vowel Length
in Czech
The Syntax–Phonology Interface
Bibliografische Information der Deutschen Nationalbibliothek
The Deutsche Nationalbibliothek lists this publication in the
Deutsche Nationalbibliografie; detailed bibliographic data is
available online at http://dnb.d-nb.de.
Library of Congress Cataloging-in-Publication Data
A CIP catalog record for this book has been applied for at the
Library of Congress.
Cover Design: © Olaf Gloeckler, Atelier Platen, Friedberg
Printed by CPI books GmbH, Leck
ISSN 1862-524X
ISBN 978-3-631-71705-9 (Print)
E-ISBN 978-3-631-74623-3 (E-PDF)
E-ISBN 978-3-631-74624-0 (EPUB)
E-ISBN 978-3-631-74625-7 (MOBI)
DOI 10.3726/b13276
© Peter Lang GmbH
Internationaler Verlag der Wissenschaften
Berlin 2108
All rights reserved.
Peter Lang – Berlin ∙ Bern ∙ Bruxelles ∙
New York ∙ Oxford ∙ Warszawa ∙ Wien
All parts of this publication are protected by copyright.
Any utilisation outside the strict limits of the copyright law,
without the permission of the publisher, is forbidden
and liable to prosecution. This applies in particular to
reproductions, translations, microfilming, and storage
and processing in electronic retrieval systems.
This publication has been peer reviewed.
www.peterlang.com
5
Acknowledgments
I am indebted to Pavel Caha for his valuable comments and suggestions.
I am also grateful to Lucie Taraldsen Medová for improving the manuscript.
Above all, I would like to thank Tom, Eli and Veru for their patience.
The research for this book was supported by a three-year grant (no. 14-04215S)
from the Czech Science Foundation (GA CR).
Motto: Une théorie inexacte amène une rectification, tandis que l’absence de théorie
n’amène rien.
Maurice Grammont
9
Content
1.	Introduction.................................................................................................11
2.	 Vowels Autosegmentalized....................................................................15
2.1	 Vowel System of Czech.........................................................................15
2.2	 Decomposing Vowels: Element Theory..............................................17
2.2.1	 The Basic Pattern /I/-/U/-/A/ and Its Phonetic Spell-Out...... 20
2.3	 Vowel Length Alternations...................................................................23
2.3.1	 Segmental Homonymy................................................................ 23
2.3.2	 Pure vs. Qualitative Lengthening............................................... 27
2.3.3	 Mid-Vowel Length Alternations................................................ 32
2.3.4	 Diphthongal Length Alternations.............................................. 37
2.4	 Morphemic Length Alternations: Prosodic Affixes..........................39
3.	 Length in Hypocoristics..........................................................................45
3.1	 The Basic Phonological Pattern...........................................................45
3.2	 Bethin’s (2003) Templatic Analysis.....................................................47
3.3	 Decomposing the Hypocoristics.........................................................51
3.3.1	 The Templatic Morpheme: Rewriting of
Prosodic Structure....................................................................... 55
3.3.2	 The Morphosyntactic Hierarchy: Closeness
over Familiarity............................................................................ 58
3.4	 Closeness-Expressing Hypocoristics..................................................59
3.4.1	 The Gender Pattern..................................................................... 60
3.4.2	 Hypocoristics as Syntactic Compounds................................... 63
3.4.3	 Lengthened Forms....................................................................... 66
3.4.4	 Decomposing Gender: Gender-Ambiguous Names............... 68
3.4.5	 Feminine Hypocoristics.............................................................. 70
3.4.6	 Interim Summary........................................................................ 72
10
3.5	 Familiarity-Expressing Hypocoristics................................................72
3.6	 Suppletive Forms...................................................................................74
3.7	Summary................................................................................................76
4.	 Length in Diminutives.............................................................................79
4.1	 Scheer’s (2003, 2004) Templatic Analysis...........................................80
4.2	 Phonology of Diminutive Suffixes and Diminutive Stems...............84
4.2.1	 Evidence That -ek Starts with [e] Lexically............................... 84
4.2.2	 Evidence That -ík Is Not Just Lengthened -ek.......................... 85
4.2.3	 Evidence That -ík Starts with [i] Lexically................................ 87
4.2.4	 Length Alternations in Bases: The Gender Asymmetry......... 88
4.3	 Decomposing the Diminutives............................................................91
4.3.1	 Long vs. Short Diminutive Stems.............................................. 91
4.3.2	 Against Shortened Stems............................................................ 95
4.3.3	 The Suffix -ek Is a Fully-Fledged Noun................................... 102
4.3.4	 The Morphosyntactic Hierarchy: Size over Affection........... 105
4.3.5	 Decomposing the Suffix -ík...................................................... 110
4.4	 Double Diminutives........................................................................... 114
4.5	Summary............................................................................................. 118
5.	Conclusion................................................................................................. 119
References.............................................................................................................. 121
11
1. Introduction
Czech has a simple vowel system and each of the vowels has a distinctive
length—it is either short or long, as illustrated in the minimal pairs in table (1).1
(1) [i] – [iː] vir [vir] vír [viːr] ‘virus, eddy’
[u] – [uː] kur [kur] kůr [kuːr] ‘fowl, choir’
[e] – [eː] pero [pero] péro [peːro] ‘pen, (spiral) spring’
[o] – [oː] lože [loʒe] lóže [loːʒe] ‘bed, loge’
[a] – [aː] lak [lak] lák [laːk] ‘lacquer, brine’
In the literature, e.g. Kučera (1961), Petr et al. (1986), Palková (1994), inter alia,
Czech short and long vowels are analyzed as having free distribution. This is
usually supported by the following claims: (i) vocalic length does not depend on
a word stress: long vowels can appear in both stressed and unstressed syllables,
(ii) long vowels can appear in any position; in particular, adjacent syllables with
long vowels are possible.
As for the relationship between length and stress, it is a widespread phenomenon
that long vowels tend to appear under stress, while the appearance of short
vowels is stress-independent. In Czech, however, word stress always falls on the
first syllable, which can be either long or short. This leads Hayes (1995:102) to
claim that Czech is a “quantity-disrespecting syllabic trochee language” (together
with, for example, Finno-Ugric languages like Hungarian, Finnish or Estonian).
Either short or long vowels are legitimate in unstressed syllables as well. The
stress-free distribution of vocalic length can be illustrated by disyllables sharing
the same stress profile, i.e., (ˈσσ)PW
(PW stands for “prosodic word”). In table
1	 In Czech orthography, vowel length is marked by an acute accent or by a superscript
circle over the vowel. The relevant sound-grapheme correspondences are as follows:
[aː] = <á>, [iː] = <í,ý>, [uː] = <ú,ů>, [eː] = <é>, [oː] = <ó>.
	 Throughout this book, examples are presented as follows: orthographic forms are in italics,
[square brackets] and /slash brackets/ denote [phonetic] and /phonological/ form
respectively; morpheme boundaries are marked by hyphen-s and glosses are enclosed
by ‘single quotation marks’. Should the need arise to distinguish between lexically stored
underlying forms and those arising from phonological computation, the former are
marked by //double slashes//. For phonetic transcription, I use the standard IPA symbols:
segmental length is marked by a triangular colon [ː], primary stress by a high vertical
line before the syllable [ˈ] and inter-syllable boundaries are indicated by dots [.].
12
(2), I summarize the four logically possible combinations of long (VV) and short
(V) vowels within prosodic words of the type (ˈσσ)PW
. To show that morphology
plays absolutely no role in the distribution of short/long vowels, the patterns are
further divided into two groups according to whether the particular combination
is tautomorphemic or heteromorphemic. What we see is that all of the four
patterns are readily attested in Czech.2
(2) tautomorphemic heteromorphemic
(ˈVV.V)pw rákos [ˈraː.kos] rán-a [ˈraː.na] ‘reed, wound’
(ˈV.VV)pw talár [ˈta.laːr] ran-á [ˈra.naː] ‘robe, early (fem.)’
(ˈV.V)pw jasan [ˈja.san] pat-a [ˈpa.ta] ‘ash tree, heel’
(ˈVV.VV)pw sárí [ˈsaː.riː] pát-á [ˈpaː.taː] ‘saree, fifth (fem.)’
The last row in table (2) illustrates another property traditionally attributed
to long vowels in Czech: not only are they stress-independent, they are also
position-independent. As a consequence, there are words with two (and even
three or four) adjacent long syllables. Some examples of such patterns are shown
below.3
(3) 3 VV b[iː]d[aː]k[uː]v pl[aː]šť[iː]k[uː]
‘scoundrel’s’ ‘of mantelets’
4 VV zař[iː]k[aː]v[aː]n[iː] vypoč[iː]t[aː]v[aː]n[iː]
‘invocation’ ‘calculation’
As a matter of fact, stress-independence and position-independence are indeed two
separate issues, as illustrated in Slovak. In Slovak—like in Czech—stress always falls
on the first syllable, which can be either long or short. But unlike Czech, Slovak
restricts the distribution of long vowels in adjacent syllables; this vocalic-lengthrestriction
is known as Rhythmic Law; see e.g. Peciar (1946). Table (4) shows the
effects of this rule. There are four third-person singular present forms, each containing
a root followed by the so-called theme suffix. We can see that the quantity
2	 Note that the distribution of individual vowels across the patterns varies considerably;
see Bičan (2015) for discussion on frequency of vocalic segments.
3	 In fact, the pattern with two tautomorphemic long vowels is restricted to a small group
of borrowed nouns including vádí [vaːdiː] ‘wadi’, vúdú [vuːduː] ‘voodoo’, týpí [tiːpiː]
‘tepee’ or propria, as Rádžív [raːdʒiːf] or Kádár [kaːdaːr]. The heteromorphemic patterns,
on the other hand, are relatively frequent, because long vowels are productive
inflectional markers.
13
of the suffix depends on the quantity of the root: if the root is short, the theme is
long—and vice versa: long roots are followed by short variants of the themes.
(4) short root – long theme long root – short theme
hľ[a]d-[aː] ‘s/he finds’ v[iː]t-[a] ‘s/he invites’
k[a]l-[iː] ‘s/he muddies’ kr[aː]t-[i] ‘s/he finds’
Yet another restriction on vocalic length (known from many languages) is related
to syllable structure: vowel length often correlates with the syllable being open.
Closed syllables, on the other hand, tend to be short; see e.g. Odden (2011:466).
In Czech, however, no such restriction exists: table (5) demonstrates that closed
syllables can be long, both word-finally and word-internally. Moreover, the examples
in the second and in the fourth row demonstrate that long vowels appear
even in super-heavy syllables, i.e., those closed by two consonants.
(5) VVC# dr[aː]p ‘claw’ l[eː]k ‘medicine’
VVCC# n[aː]rt ‘instep’ dvan[aː]ct ‘twelve’
VVC.C b[oː]j.ka ‘small buoy’ sc[eː]n.ka ‘small scene’
VVCC.C č[aː]st.ka ‘amount’ ř[iː]ms.ka ‘small cornice’
Finally, long vowels are known to be sensitive to segmental context. For a wide
range of languages, vowel length is licensed only before voiced consonants; see
e.g. Maddieson (1997). Czech—again—displays no sensitivity to consonantal
voice, as demonstrated in the table below.
(6) VVC[+voiced]
m[iːz]a ‘lymph’ k[oːd]a ‘coda’ s[aːg]a ‘saga’
VVC[-voiced]
m[iːs]a ‘bowl’ k[oːt]a ‘dimension’ p[aːk]a ‘lever’
Summing up: in Czech, vocalic length is distinctive and does not follow any phonological
restrictions described in the literature cross-linguistically.
Even more strikingly, Czech sometimes seems to behave against all the phonologically
plausible patterns. An example of what we might call an “antiphonological”
distribution of length is shown in table (7). The table contains three
monosyllabic roots, i.e., ‘frost’, ‘salt’, ‘house’, whose vowels alternate in length.
(7) V.CV VVC# VC#
mraz-u [mra.zu] mráz [mraːs] mraz [mras]
sol-i [so.li] sůl [suːl] sol [sol]
dom-u [do.ma] dům [duːm] dom [dom]
14
From the phonological point of view, the distribution of length across this paradigm
is “weird” for two reasons. First, long vowels occur only in closed syllables.
Second, both long and short vowels appear in the same phonological context,
yielding minimal pairs mr[aː]z ‘a frost’ – mr[a]z ‘freeze!’, s[uː]l ‘a salt’ – s[o]l ‘salt!’
and d[uː]m ‘house’ – d[o]m ‘home’. Given these facts, we may be tempted to consider
the appearance of the long vowels in the second column of table (7) as purely
accidental. It is quite likely that if we were to consider just these minimal pairs
from a purely phonological perspective, we might get that picture. However, that
would be an oversimplification. We cannot conclude that just because the length
in these roots is not controlled by pure phonology, it is random. If we consider the
distribution of long and short vowels with morphosyntactic information in mind,
the emerging picture is far from accidental. For the cases under discussion, for instance,
the forms with long vowels are all nominative/accusative singular forms—
and that, as I am going to argue, is ultimately the reason why they are long.
The main goal of this book is thus to develop a morphosyntactic account of
vowel length in contemporary Czech. I will build on work by Scheer (2001, 2003,
2004a), Bethin (2003), Caha & Scheer (2008), Ziková (2012, 2016), and Caha &
Ziková (2016) who all propose that there is a connection between the morphosyntax
and the phonology, such that certain morphosyntactic categories regulate
the distribution of vowel length in Czech. In particular, I provide a thorough
analysis of two morphological categories, hypocoristics and diminutives, which
connects their morphosyntactic and phonological properties.
The book is organized as follows. Chapter 2 introduces the framework for distinguishing
different phonological types of vowel length alternations in contemporary
Czech. Using analytical tools of two autosegmental theories, Strict CV
(Lowenstamm 1996, Scheer 2004) and Element Theory (Harris & Lindsey 1995,
Backley 2011), I argue that length alternations can be modelled as arising from
prosodic affixes of various phonological shapes. The next two chapters provide
case studies on hypocoristics and diminutives. Following the central ideas of the
theory of Nanosyntax (Caha 2009, Starke 2009), I decompose these two morphological
categories into hierarchically ordered syntactic projections. Then, I
argue that some of these projections are spelled out by prosodic affixes which are
responsible for lengthening in hypocoristic and diminutive stems.
15
2.  Vowels Autosegmentalized
2.1  Vowel System of Czech4
Czech vocalic inventory is traditionally described as consisting of five short vowels,
five long vowels and three diphthongs (e.g. Kučera 1961, Petr et al. 1986,
Palková 1994, among many others):
(1) short monophthongs [i u a e o]
long monophthongs [iː uː aː eː oː]
diphthongs [ou au eu]
At a first sight, the system seems to be symmetrical: (i) each of the short vowels
has a long counterpart, (ii) all three diphthongs are of the falling type and exhibit
an identical final segment, i.e., [u]. However, if we focus on the distribution of
particular segments, we soon discover a clear unbalance in the system.
Starting with the diphthongs, there is a dividing line between [au] and [eu]
on the one hand and [ou] on the other. The diphthong [ou] is involved in length
alternations: it alternates with a short [u] (as, for example, in a root ‘buy’: k[ou]pit
‘to buy’ – k[u]p-ovat ‘to buy repeatedly’). By contrast, diphthongs [au] and [eu]
never alternate with any short monophthongs, i.e., neither with [u] nor with [a]
or [e]. Moreover, the non-alternating diphthongs are limited to loans (especially
to loans of the Latin and Greek origin such as k[au]ce ‘bail’ or n[eu]ron ‘neuron’),
as opposed to [ou]; the diphthong [ou] is distributed across both the native (e.g.
k[ou]t ‘corner’ of Common Slavic origin) and non-native vocabulary (e.g. k[ou]č
‘coach, trainer’, an English borrowing).
Among the monophthongs, it is the long mid back [oː] that has a special status.
First of all, it is noticeably less frequent than the other vowels. This might be
easily ascribed to the fact that [oː] is not a part of any inflectional marker. In this,
it contrasts with its short cousin [o] on the one hand—and the rest of the long
monophthongs on the other: they are all involved in inflectional endings. Roots
with [oː] are typically loans (e.g. t[oː]n ‘tone’ or šapit[oː] ‘circus tent’) and, in
4	 The present-day Czech has two main sociolinguistic varieties, Literary Czech (“spisovná
čeština”) and Common Czech (“obecná čeština”), and also several territorial
dialects with slightly different vowel systems. In this book, I focus mainly on Literary
Czech (henceforth Czech) which is an established norm.
16
addition, they tend to occur in expressive reduplicative structures such as č[oː]ro
m[oː]ro ‘mess-up’ and hal[oː] hal[oː] ‘attention please, attention please!’. However,
for us, the most important point is that [oː] shows a peculiar behavior with
respect to alternations, a fact I turn to immediately below.
In Czech, long monophthongs often alternate with their short counterparts.
Length alternations occur, for example, in bisyllabic truncations, including hypocoristics
and expressive common nouns (typical for Colloquial Czech), as illustrated
in table (2). Here, we have a set of female first names (in the left part of
the table) and a set of common nouns (in the right part). Each set then consists
of five pairs of nouns, in which the first member is always a full form of the
noun and the second is its truncated form. What we see is that the truncation is
typically accompanied by a lengthening of the root vowel. And with respect to
the lengthening, all monophthongs—including the mid back one—are treated
alike. Specifically, lengthening under truncation does not involve a vowel-quality
change: [o] simply lengthens to [oː]—just as [e] lengthens to [eː], [a] to [aː] and
so on.
(2) truncated first names truncated common nouns
i~iː J[i]tka – J[iː]ťa v[i]svědčení – v[iː]zo ‘certificate’
u~uː L[u]cie – L[uː]ca [u]čitelka – [uː]ča ‘female teacher’
e~eː P[e]tra – P[eː]ťa v[e]čeře – v[eː]ča ‘dinner’
o~oː B[o]žena – B[oː]ža k[o]čka – k[oː]ča ‘cutie’
a~aː M[a]rie – M[aː]ja sv[a]čina – sv[aː]ča ‘snack’
Alternations between short and long vowels of the same quality occur also in
contexts other than truncations. Two of them are shown in (3). In (3a), masculine
roots alternate in length in the diminutive context. Verbal pairs in (3b)
illustrate an alternation between short and long root forms depending on their
aspectual context.
(3) a. hř[i]b hř[iː]b-ek ‘boletus, small boletus’
hr[a]d hr[aː]d-ek ‘castle, small castle’
b. vy-kl[i]d-it vy-kl[iː]z-et ‘to clean out (once), to clean out (repeatedly)’
vy-tr[a]t-it vy-tr[aː]c-et ‘to fade out (once), to fade out (repeatedly)’
The fact that the alternation [o]~[oː] is missing from the table above, however,
does not mean that the mid back vowel fails to alternate in these morphosyntactic
contexts altogether: it does alternate between the short and long variant,
17
analogically to [a] and [i], as illustrated in (4). In this case, however, the length
alternation involves also a change in the vowel quality: in the pair pl[o]t – pl[uː]tek
[o] alternates with a long [uː] and in the aspectual pair vy-t[o]č-it – vy-t[aː]č-et,
it alternates with a long [aː].
(4) a. pl[o]t pl[uː]t-ek ‘fence, small fence’
b. vy-t[o]č-it vy-t[aː]č-et ‘to dial (once), to dial (repeatedly)’
To sum up: [oː] functions as a long counterpart to [o] only in truncations. In
other contexts (in which length alternations are triggered), [o] is replaced by either
[uː] or [aː]. In what follows, I am going to argue that this three-way behavior
of [o] with respect to length alternations can be best captured by using the tools
of Element Theory based on monovalent melodic primes.
2.2  Decomposing Vowels: Element Theory
In Element Theory (ET), segments are represented by privative melodic primes,
called elements (Kaye et al. 1985, Harris 1990, Harris & Lindsey 1995, Backley
2011, inter alia).5
Elements, as autosegmental entities, are structurally independent
which entails at least two things. First, any element can be, in principle, embedded
in any prosodic structure: a single element can thus be dominated either
by a V-slot or by a C-slot which constitute the prosodic structure. As a consequence,
vocalic and consonantal segments share pieces of their melodic structure.
This is shown in table (5): each of the so-called resonance elements—I, U
and A—defines classes of both vocalic and consonantal segments (according to
their aperture/color features and place of articulation, respectively).6
5	 Historically, ET is connected with Government Phonology (Kaye et al. 1985, 1990)
and its radical version called Strict CV (Lowenstamm 1996, Scheer 2004). However,
as Backley (2011:xii) points out “there is no necessary connection between Government
Phonology and elements. In fact, elements can be traced back to similar units
that were being used in other theories including Dependency Phonology and Particle
Phonology, both of which predate Government Phonology.”
6	 In fact, in different versions of the theory, the number of elements varies. The three
resonance elements are, however, shared among all of them. Other elements—employed
to define phonological segments—are, for example, L or N: they represent
vocalic tonality and nasality, respectively. Since these non-resonance elements are not
relevant for our discussion at the moment, they are not listed in table (5).
18
(5) Resonance elements and classes of segments (adapted from Nasukawa 2014)
element V-dominated C-dominated
I front vowels dentals, palatals
U rounded vowels labials, velars
A non-high vowels uvulars, pharyngelars
The independence of melodic elements and prosodic positions is usually illustrated
by the example in (6). There is an element I which is associated with a
V-slot or a C-slot: the V-dominated structure corresponds to a high front vowel
[i] (6a) and the C-dominated structure to a palatal glide [j] (6b). The symbol ●
in the figures below marks a root node which serves as an “umbrella point” of
melodic elements belonging to a single segment. I come back to it later.
(6) a. V ↔ [i] b. C ↔ [j]
| |
● ●
| |
I I
Yet another thing results from the fact that elements as autosegmental objects
live their own lives: they can freely combine to create complex segments. However,
this possibility is not driven by their need to be spelled out, i.e., to be linked
to a prosodic slot. From this perspective, vocalic segments (as defined by the
three resonance elements listed in table (5) above), can be arranged on a scale
according to their complexity. On this scale, one-element vowels are followed by
two-element ones. All possible types of simplex and complex vocalic segments
are summarized in (7). I also include corresponding sounds in the table. Note,
however, that the mapping between phonological segments and sounds is not as
straightforward as the table suggests. I will come back to this issue later. (Threeelement
structures are not involved in (7): they typically represent rounded nonback
vowels which are missing in Czech.)7
7	 The scale illustrates one of the main principles of ET, that is, non-recursiveness: a
particular element can occur only once per segment; but see Nasukawa (2014) who
argues for recursive elements. I should also mention that simple counting of elements
is only one of the possible ways to measure a segmental complexity—though the most
widespread one. In Pöchtrager’s (2006) version of the theory, complexity is measured
in terms of melodic projections inside a given segment. For example, fricatives are
19
(7) simplex vowels > complex vowels
/I/ ↔ [i] /I U/ ↔ [y]
/U/ ↔ [u] /I A/ ↔ [e]
/A/ ↔ [a] /U A/ ↔ [o]
The notion of segmental complexity is an effective analytical tool for capturing
typological generalizations about phonological processes; see e.g. Harris (1997).
For example, there exists a bunch of non-assimilatory changes of consonants,
such as spirantization (when stops change to fricatives) or glidization (when
consonants turn to semivowels), covered by the term lenition. The word lenition
means “weakening”: lenited consonants weaken their articulation. If classes of
segments differ in the degree of complexity, as is proposed in ET, then lenition
processes can be treated in a uniform, even iconic, way: weakening of consonants
is based on weakening of their internal structure, that is, outputs of lenition
are less complex than the inputs.8
In a vocalic space, there is a parallel to lenition: reduction of vowels. In the
Russian examples in (8), taken from Iosad (2012), a mid back vowel [o] is reduced
to a low [ɐ] (8a) and a mid front vowel [e] is reduced to a high [ɪ] in an
unstressed position. Given that mid vowels are complex segments, then both
reduction processes, i.e., the o-to-ɐ lowering and the e-to-ɪ raising, can be represented
as involving a decrease in complexity, as is illustrated in the right-most
column of table (8).
(8) a. [ˈkot] – [kɐˈta] ‘cat, Nsg, Gsg’ UA [o] > A [ɐ]
b. [ˈlj
es] – [lj
ɪsˈɲik] ‘forest, forester’ IA [e] > I [ɪ]
generally taken to be less complex than stops: which means that they either lack an
occlusive element ʔ contained in stops or, according to Pöchtrager (2006:63), they are
one-projection-layer smaller than stops.
8	 Moreover, it is well-known that lenition is related to the syllabic structure: consonants
in codas and intervocalic onsets are typically prone to undergo weakening, as opposed
to consonants in postcodas and initial onsets (both non-weak positions being covered
by the term Coda Mirror); see Ségéral & Scheer (2005), Scheer & Ziková (2010). In
Government Phonology, whose part ET is, the relationship between the segmental
weakening and the syllabic structure is expressed via licensing: put it simply, both intervocalic
onsets and codas appear under certain licensing conditions that differ from
that for consonants in the Coda Mirror. And this is the reason why they are targets of
different types of segmental changes.
20
However, it is not always the case that the decrease in vowel complexity is lenition.
The raising of mid vowels—which involves element reduction—typically
occurs under lengthening, which is not taken to be an instance of lenition. I will
discuss the relevant examples from Czech in section 2.3.3.
2.2.1  The Basic Pattern /I/-/U/-/A/ and Its Phonetic Spell-Out
Furthermore, the term segmental complexity is going to be useful, as it expresses
generalizations not only about the grammar, i.e., about the phonological processes,
but also generalizations about the lexicon, i.e., about the segmental inventories.
As for vowel systems—the prime interest of this book—the complexity
scale in (7) can be assumed to be strongly one-way implicational. Given that the
existence of complex vowels implies the existence of simplex ones, two things
concerning typology of vocalic inventories are predicted. First, in languages with
just three phonologically contrastive vowels, all of them will be simplex. Second,
in systems with four vowels and more, three of the vowels will always be simplex.
The first prediction says that three-vowel inventories can be reduced to a single
underlying pattern /I/-/U/-/A/ when each vowel is represented by just one of
the three resonance elements. Provided that there are correspondences between
the elements on the one hand and the classes of vocalic segments on the other
(such as those proposed in table (5) above), then there should not exist threevowel
languages with either two front or two rounded vowels. This prediction
seems to be borne out, as shown by the data in table (9). There are nine different
three-vowel systems reported in Hyman (2008) and Backley (2011)—and none
of them shows more than one rounded and one front vowel.9
9	 In fact, Hyman (2008) reports on two three-vowel inventories such that they lack either
a front or a rounded vowel: the former case is represented by a triad /ə o ɐ/ found
in Qawasqar, the latter by a triplet /i ɯ a/ appearing in Jaquaru. Both inventories are
compatible with our prediction: none of the three-vowel systems includes two (or even
three) rounded or front vowels, respectively. Provided that all three-vowel systems follow
the underlying pattern /I/-/U/-/A/, then [ə] realizes the I-segment in Qawasqar
and [ɯ] spells out the U-segment in Jaquaru, even if the former is not a front vowel
and the latter is not a rounded vowel. However, such “clashes” between phonological
features, i.e., elements, on the one hand—and corresponding sounds on the other are
expected if phonology and phonetics are, indeed, two separate language modules. The
fact that such phonology–phonetics conflicts are nevertheless not as common as one
might have expected is discussed in Scheer (2014); Scheer’s interpretation is presented
below. Moreover, Hyman (2008) himself notes that the systems found in Qawasqar and
Jaquaru are quite unique, especially as compared to the systems listed in table (9).
21
(9) Three-vowel inventories (Hyman 2008, Backley 2011)
/i u a/ Aleut
/ɪ ʊ a/ Bella Coola
/i ʊ a/ Haida
/ɪ ʊ ɐ/ Quechua
/i u ɐ/ Greenlandic
/i u æ/ Shilha
/i o a/ Pirahã
/i u ɜ/ Gadsup
/e o ɐ/ Amuesha
Furthermore, in the representative sample of three-vowel languages shown
above, each of the I-segment and U-segment is realized by three different phonetic
objects, i.e., [i ɪ e] and [u ʊ o], respectively. The A-segment has even four
corresponding vocalic spell-outs: [a ɐ ɜ æ]. Such variation is, however, not surprising:
although elements are assumed to be associated with certain phonetic
features, primarily, they are cognitive categories and their principal function is
to first, encode lexical contrasts among segments and, second, to identify natural
classes of segments with respect to phonological processes (see Harris & Lindsey
1995).
From this perspective, phonology and phonetics are thus two separate linguistic
modules and each operates with their own units. Phonology operates
with segments made up from elements and phonetics acts on sounds associated
with instructions to the auditory–articulatory system. A consequence of
this view is—as pointed out by Scheer (2014)—the arbitrariness of the phonetic
spell-out. Scheer draws a parallel between two interfaces phonology is involved
in: the syntax–phonology interface and the phonology–phonetics interface. For
both interfaces, there are lexicons and for their entries, we assume arbitrary
pairings between syntactic and phonological objects (in the former case) – and
phonological and phonetic objects, for phonology–phonetics interface. Scheer
(2014:268f.) further proposes that the interface lexicons differ in the degree of
arbitrariness: while “the relationship between morphosyntactic structure and its
exponent phonological material is 100% arbitrary”, the phonetic and phonological
identities of lexical items at the phonology–phonetics interface are—more
or less—closely related in “the overwhelming majority of cases”. Scheer explains
this situation via grammaticalization: “The output of the grammaticalization
process that turns phonetic into phonological items is akin to the phonetic input,
22
and also uses the same broad categories. By contrast, the relationship between
the items related by the upper spell-out is not one of grammaticalization: tense,
person, number, etc., are not the grammaticalized versions of labial, occlusion,
etc. Therefore, there is no way to even imagine any similarity.”10
Of course, since the phonetic spell-out is arbitrary, phonological and phonetic
features may be in conflict: in principle, each of the simple segments can be realized
by whatever vocalic sound. And Scheer (2014:266), referring to Uffmann
(2010), brings one example of such a “phonology–phonetics clash”: in a SouthEast
British sociolect of English, a front, unrounded sound [i] spells out the Usegment,
which is realized by a high rounded vowel [u] in Standard English; for
example the verb do is spelled out as [diː] in this sociolect. According to Scheer,
the underlying U-identity of the [i]-sound is uncovered by observing its behavior,
for example, with respect to external sandhi gliding: the [i] alternates with a
labial glide [w] and, crucially, not with a palatal [j]—as would be expected, if [i]
was really /I/ underlyingly. Thus, the verbal phrase do it is spelled out with [w]
both in Standard English ([duː w it]) and the South-East British sociolect ([diː
w it]). However, Scheer points out that the phonology-to-phonetics mapping of
this type, i.e., /U/ ↔ [i] in this particular case, is rare and he attributes the rarity
of such a mapping to the grammaticalization reasons mentioned above: U-segments
are typically realized by rounded vowels, while a front vowel [i] typically
spells out an I-segment.11
Summing up, the modular view on the phonology-to-phonetics mapping
leads ultimately to the conjecture that homonymy might not be unique to the
lexicon of the syntax–phonology interface, but it is also possible to observe it
among the lexical items of the phonology–phonetics interface. In this interface,
10	 Scheer’s unified account to the phonology interfaces opens a door to a myriad of interesting
questions on the other possible parallels between the interfaces. Could it be,
for instance, that a matching procedure from syntax–phonology interface (known as
Subset/Superset Principle) has a phonology-to-phonetics parallel? As fascinating as
these questions might be (leave alone their consequences), I have to leave them open
for further research.
11	 Moreover, it seems to be rather uncontroversial that the most widespread triad among
three-vowel inventories is the /i u a/ one; in Hyman’s (2008) and Backley’s (2011) surveys,
thirteen languages in total have this inventory; the second most ‘popular’ one is
the triad /ɪ ʊ a/ (in four languages). This indicates that the vowels [i u a]—the corners
of the classical vocalic triangle—can be viewed as default spell-outs of a simplex I-,
U- and A-segments.
23
homonymy arises among items that have a distinct elemental makeup—but they
are spelled out by identical sounds.
2.3  Vowel Length Alternations
Having introduced the basic tools of the framework, let me now go back to the
Czech length alternations. Highlighting of what there is to come, I will now investigate
various options of capturing the facts using the tools at our disposal. I
start with the discussion of the options that appear theoretically simplest, but I
end up rejecting several of them for empirical reasons, before I arrive at a satisfactory
solution.
2.3.1  Segmental Homonymy
The table below repeats the pattern of the length alternations with a mid rounded
vowel [o]: it either simply lengthens to [oː] or the alternation involves raising
[o]~[uː] or lowering [o]~[aː].
(10) o~oː B[o]žen-a B[oː]ž-a
o~uː pl[o]t pl[uː]t-ek ‘fence, small fence’
o~aː vy-t[o]č-it vy-t[aː]č-et ‘to dial (once), to dial (repeatedly)’
By applying the same logic as in the previous cases, we might be tempted to analyze
these alternations in terms of homonymy. In other words, we might claim
that the three-way behavior of the vowel [o] with respect to the length alternations
indicates that this sound has a three-fold underlying identity. That is, the
phonology-phonetics interface lexicon of Czech contains three homonymous
items, but in each case, the mid rounded vowel is paired with a different phonological
structure: /X/ ↔ [o], /Y/ ↔ [o] and /Z/ ↔ [o]. The evidence for the child
to postulate these three distinct objects would be provided by the fact that each
of them behaves differently under lengthening.12
As to the actual identity of X, Y and Z, the first possibility that comes to mind
is that the phonological structure of the particular [o] is identical to the phonological
structure of its long alternant. Thus, the [o] that alternates with a long
[uː] is a spell-out of a simplex U-segment, the [o] alternating with a long [aː] is
12	 The same logic is followed by, for example, Gussmann (2007) who proposes [ɛ]homonymy
for Polish. His homonymy analysis aims to capture the different behavior
of a front mid vowel with respect to the two phonological processes generally associated
with front vowels, i.e., palatalization and glidization.
24
a simplex A-segment. By contrast, mid vowels involved in [o]~[oː] alternations
realize a complex segment made up from both U and A. In that case, three homonymous
lexical items would be listed in the phonology–phonetics interface
lexicon of Czech, as depicted in (11).
(11) a. [o] alternating with
[aː]
b. [o] alternating with
[uː]
c. [o] alternating with
[oː]
/V/ ↔ [o] /V/ ↔ [o] /V/ ↔ [o]
| | |
● ● ●
| | |
A U UA
In (11), the three segments associated with the sound [o] differ both with respect
to their complexity (the segments in (11a–b) are simplex, the one in (11c)
is complex) and their elemental make-up: each of them is made up from different
elements. This analysis, however, is inadequate. The reason is that there exist
also sounds [a] and [u] in Czech that also realize the simplex A-segments and
U-segments. This thus means that the phonology–phonetics interface would
contain two pairs of lexical items, i.e., /A/ ↔ [o] and /A/ ↔ [a] on the one hand
and /U/ ↔ [o] and /U/ ↔ [u] on the other, each involving a single segment
that is spelled out by two distinct sounds. However, the spell-out is not predictable
from phonotactics alone. To see this, consider, for instance, the roots b[o]k
‘flank’ and b[u]k ‘beech’: they are identical except for their vowel, which is either
[o] or [u]. Since the mid vowel [o] of the root ‘flank’ alternates with the high
vowel [uː] (as is proved by the diminutive b[uː]č-ek), it should correspond to the
U-segment underlyingly (according to the proposal in (11b) above). From this,
it follows that the roots ‘flank’ and ‘beech’ being both stored with the U-segment
would have to be diacritically marked in order to be paired with the appropriate
vocalic sounds.
Now, let me try to follow the same logic (postulating three different underlying
objects), but in a way that will allow us to dispense with diacritic marking.
So far, I have been considering segments to be unordered sets of elements. However,
following Dependency Phonology (Anderson & Ewen 1987), ET assumes
that segments are internally organized. Without going into details, the shared
idea among different approaches to segment-internal structure (represented by
Kaye 2001, Pöchtrager 2006 and Nasukawa 2014, inter alia) is that a single set of
25
elements can be arranged in different ways to produce distinct segments of the
same elemental make-up.
In this book, the internal structure of segments is going to be expressed in
terms of structural dependency. From the dependency perspective, two elements
X and Y can be merged in three different ways (and thus, they produce three different
segments): either X is structurally dependent on Y (or vice versa), or both
X and Y are structurally equal, i.e., mutually independent. Thus, if we take U and
A to be the X and Y, we can derive three distinct complex segments from the elements
U and A, as depicted in (12). The [o] that alternates with [aː] corresponds
to an element U dependent on the element A (cf. U-dependent segment) (12a),
and the [o] which alternates with [uː] is an A-dependent segment as in (12b).
Finally, (12c) shows a structure of the [o] involved in a pure length alternation:
in that case, neither of the elements A and U is dependent.
(12) a. [o] alternating with
[aː]
b. [o] alternating with
[uː]
c. [o] alternating with
[oː]
/V/ ↔ [o] /V/ ↔ [o] /V/ ↔ [o]
| | |
● ● ●
| | | |
A U U A
| |
U A
The homonymy scenario in (12) thus dispenses with diacritic marking: all AUsegments—and
only them—are realized by the sound [o]. This is a desirable
result, but, unfortunately, still not empirically correct. (12) still fails to capture
the distribution of the long alternants. If the underlying structure of the short
alternant determines the form of its long cousin, we should not find “combined”
length alternations. In other words, since each o-containing morpheme
is stored with a particular o-segment in (12), we do not expect to see morphemes
with more than one type of the length alternation. But that is what we
find, as illustrated in table (13) below. In (13a), there are six triplets of words;
each triplet is based on the same root which appears with both a long [uː] and
long [aː]. The short root in (13b) also has two long forms, one with [oː] and the
other with [aː].
26
(13) a. kr[uː]č-ek kr[aː]č-et kr[o]k
‘small step’ ‘to walk’ ‘a step’
s-ch[uː]d-ek s-ch[aː]z-et s-ch[o]d
‘small stair’ ‘to step down’ ‘a step’
po-kl[uː]p-ek vy-kl[aː]p-ět po-kl[o]p
‘small cover’ ‘to dump’ ‘a cover’
tv[uː]r-ce vy-tv[aː]ř-et vý-tv[o]r
‘creator’ ‘to create’ ‘creation’
str[uː]j-ce pře-str[aː]j-et str[o]j-it
‘originator’ ‘to dress up (repeatedly)’ ‘to dress up’
kr[uː]p-ěj s-kr[aː]p-ět kr[o]p-it
‘drop’ ‘to spray (repeatedly)’ ‘to spray’
b. t[oː]č-o o-t[aː]č-et t[o]k
‘topsy-turvy’ ‘to turn’ ‘a flow’
The examples above thus show that the form of the long alternant is not predictable
from the underlying structure of the short vowel. Hence: the combined
alternations—when a single morpheme has more than one long form—must be
treated suppletively, and that, in turns, means that both long forms of the roots
in (13) must be lexically stored. But if we were to subscribe to this analysis, we
might have missed an important point. The thing is that the suppletive scenario
misses a generalization: the long vowels in the roots are not distributed randomly,
there is a clear pattern they follow. First, all root forms with a long [aː] appear
in verbal forms with an iterative reading. Second, the word t[oː]čo built on the
root with a long [oː] has an expressive meaning (and, as such, it is labeled as
colloquial), literally ‘a situation when things flow around in a chaotic manner’.
It thus patterns with expressive truncations such as k[oː]ča (< k[o]čka) ‘cutie’ or
B[oː]ža (< B[o]žena) ‘first-name hypocoristic’ mentioned in section 2.1.
To sum up, a short mid round vowel alternates with three distinct long vowels
in Czech. There exist morphemes that are involved in more alternation patterns.
This fact disqualifies the scenario according to which these patterns are associated
with lexically different o-vowels. A key to understanding the three-way behavior
of [o] is therefore the morphosyntactic distribution of long alternants. In
other words, I propose that there is only a single o-segment underlyingly and it
undergoes different types of lengthening because of its merger with phonologically
different prosodic affixes that realize particular morphosyntactic features.
The next sections elaborate on the phonological details of this proposal.
27
2.3.2  Pure vs. Qualitative Lengthening
Having established the ingredients of the melodic and the prosodic structures,
let me proceed to vowel length and its representation. In the strict CV model,
where the prosodic structure is made up from a linear sequence of C- and Vslots,
the contrast between short and long vowels is represented by the longvowel
spanning over two V-slots (separated by an empty C-slot). Since I assume
that melodic units are linked to prosodic slots via root nodes, it then follows that
different subunits of vowel structure can be doubly-linked. In other words, either
a root node, i.e., an entire segment can be doubly-linked—or just an individual
element. These two possibilities are depicted in (14). In (14a), an element X is
dominated by a single root node, which, in turn, is associated to two V-slots. In
(14b), by contrast, there are two root nodes (and hence two V-slots) to which the
element X is doubly-linked.
(14) a. doubly-linked root node b. doubly-linked element
V C V ↔ [αː] V C V ↔ [αː]
| | | |
● ● ●
| | |
X X
Suppose now that the representations in (14) arise from lengthening of underlyingly
short vowels. Under the lengthening scenario, either the entire root node
spreads to the adjacent V-slot, as in (15a), or an individual element does, as in
(15b); spreading is marked by an arrow in (15) and throughout. Which type of
spreading is chosen is dependent on the adjacent V-slot. In (15a), the absence of
a root node in V2
triggers the linking of the node from the V1
. By contrast, (15b)
depicts the situation when the V2
is provided with the root node. In this case,
then, just an individual element spreads from the V1
to the V2
.
(15) a. root-node spreading b. element spreading
V2
C V1
V2
C V1
| | |
● ● ●
| |
X X
28
These two types of empty prosodic categories, i.e., with a root node and without
it, are proposed by John (2014). He argues that each type is involved in different
classes of local spreading processes such as vowel harmony, glide formation and
lengthening of both consonants and vowels. John (2014:37) argues as follows:
since the function of root nodes is to harbor melody, only the empty prosodic
categories provided with root nodes can be targets for spreading of individual
elements; “otherwise the individual element that spreads would have nowhere
to dock.”13
Furthermore, the figures in (15) indicate that spreading goes leftward. The
reason is that licensing as a melody-supporting relation between two prosodic
objects is leftward as well. Generally, there are two lateral relations that operate
over C- and V-slots, i.e., licensing and government, and they define particular
syllabic units and their configurations. For example, an intervocalic consonant,
i.e., a syllable onset occurring after an open syllable, is identified as a segment
dominated by a C-slot that is governed by the following V-slot. Both relations,
then, have the same source and both operate regressively: only a V-slot can govern
or license either a (preceding) V-slot or a (preceding) C-slot. The difference
between the relations lies in the way they link to a melody: while licensing is a
melody supporter, government is a melody inhibitor. From this it follows, among
other things, that empty V-slots must always be governed to remain silenced, i.e.,
unassociated with any melody.14
With this as a background, let me now look at the lengthening scenarios proposed
in (15) above. Both are repeated in (16) below together with the government–licensing
relations that hold among the relevant prosodic positions. In
both cases, the underlying short vowel (V1
) serves as a self-licensor: it licenses
the preceding empty V2
that thus accommodates its melodic features; the result
13	 Furthermore, John (2014) argues that the difference between these two types of empty
categories is manifested not only in melody spreading, but also in melody insertion.
According to him, insertion of what he calls default segments, i.e., insertion of schwas
in empty nuclei and glottal stops in empty onsets or codas, is incumbent on the presence
of a root node: only those empty categories that contain root nodes are capable of
generating default melody. Notice, however, that it is a last-resort scenario: the empty
root node receives a default melody only when spreading from the adjacent segments
is, for some reasons, impossible.
14	 The regressiveness issue is thoroughly discussed in Scheer (2004). He points to a contrast
between relations holding among prosodic units, i.e., government and licensing,
and those operating on the melodic level, such as melody spreading: the former are
always regressive, while the latter can work in both directions. I probe this distinction
further on when discussing syllable-sensitive alternations in vowel length.
29
is a doubly-linked vowel. Since V1
as a full vowel is not only a proper licensor, but
also a proper governor, it governs the preceding empty C-slot.
(16) a. L G
b. L G
V2
C V1
V2
C V1
| | |
● ● ●
| |
X X
In both diagrams in (16), the short vowel (V1
)is preceded by two empty prosodic
positions, hence, both the C and the V2
can be—in principle—targeted by either
licensing or government. In other words, if nothing was said about how both
relations are distributed with respect to each other, licensing and government
could work the other way around than it is proposed in (16). In that case, the V1
would license its melody in the C-slot and the V2
would be a target of government,
which then means that it would have been silenced. As a consequence, we
would not derive an alternation between a short and a long vowel, but between
a short vowel and a consonant-vowel sequence. Thus, to get a long vowel, the
scenario depicted in (16) is the only option.
Jubilation would be misplaced, though, as “We get what we wanted” is still but a
claim, not an explanation. A question we should answer is whether the configurations
in (16) follow from any general principle of the theory. And, lo and behold,
there is a general principle: a government-over-licensing hierarchy proposed by
Scheer & Ziková (2010) in order to distinguish among different types of strong
and weak syllable positions. Applied to structures in (16), this principle says that if
a full vowel is preceded by two empty prosodic positions, then the closest one, i.e.,
the C-slot, is governed and the more distant one, i.e., the V2
, is licensed.
Summing up, lengthening of short vowels involves licensing either of their
entire melodic content, or just of its part, depending on whether the preceding
empty V-slot lacks its own root node or whether it is provided with it. Let me
now explore the consequences of this proposal.
First of all, I have to note that the contrast between the root-node spreading
in (16a) and the individual-element spreading in (16b) is visible only in complex
vowels. If the source of the spreading is a simplex vowel (made up from
a single element), then both scenarios will yield the same effect; see also John
(2014:37). In other words, if the V1
is, for example, a short low vowel [a] (that is,
a simplex A-segment), then its long cousin will be [aː] according to both possible
30
configurations in (16). Lengthening of two-element vowels, on the other hand,
will produce differential results: pure lengthening is a result of the scenario in
(16a) only; that is, the root node spreading induces pure lengthening, i.e., licensing
of both elements of the underlying short vowel. But if the very same vowel
is involved in the derivation depicted in (16b), only one of its elements can be
licensed, then. The result is what I call qualitative lengthening.
Let me now elaborate on the individual-element spreading scenario in (16b).
Suppose that we deal with a complex vowel made up from elements X and Y,
these elements enter the derivation. What we want to know is whether there
are good reasons to decide which of these two elements will be licensed in the
preceding empty V-slot. A common practice in the literature on Element Theory
is to regulate melody-spreading via language-specific parameters such as “in a
language L, an element E never spreads”. For example, Charette & Göksel (1996)
and Kaye (2001) analyze vowel harmony in Turkish (and other languages) along
these lines. They assume that segments are internally organized, but, crucially,
their behavior in melody spreading is not predictable from their internal structure
alone. Instead, they propose the so-called licensing constraints that regulate
element linking (and delinking) in the particular languages. A consequence of
this account of segment internal structure is that heads and operators (i.e., cover
terms for elements occupying different structural positions within segments) are
both equally active in spreading.15
By contrast, under the dependency account of melodic level (that I use and
develop in this book), behavior of complex segments is directly derivable from
their internal structure. In particular, I assume that only structurally prominent
melodic features, i.e., non-dependent elements in the proposed model, can
spread individually to adjacent empty root nodes. What are the predictions of
this proposal?
Figures in (17) repeat what I have already mentioned: two elements X and Y
can be combined in three different ways (and thus deriving up to three different
segments of the same melodic content). Either Y is structurally dependent on
X (17a), or vice versa, X is embedded under Y (17b). Finally, the figure in (17c)
shows the situation when neither of the elements is dependent on the other: both
X and Y are associated directly with a root node.
15	 Moreover, elements can even change their roles: a head element can turn into an operator
when spreading, and vice versa, operators can switch and become heads; cf. Kaye’s
(2001:260f.) analysis of vowel harmony in Finnish.
31
(17) a. V b. V c. V
| | |
● ● ●
| | | |
X Y X Y
| |
Y X
If the segment is represented as an unordered set of two elements (as in (17c)),
then each of them can in principle spread into the preceding empty V-slot and
associate with its root node. In that case, we expect to see a particular short vowel
alternating freely with its two different long cousins. In other words, lengthening
of the short vowel of the type (17c) can—in principle—create structures in both
(18a) and (18b), where either X or Y, respectively, is doubly-linked.
(18) a. V2
C V1
b. V2
C V1
| | | |
● ● ● ●
| | | |
X Y Y X
The configurations in (18) represent diphthongs whose members share a part of
their melodic structure: the activated element is interpreted in both segments
involved in a diphthong. For example, if the structure in (18) represents a mid
rounded vowel [o], i.e., the UA-segment, then its lengthening can yield two
o-final diphthongs: [uo] or [ao].
If the presence of a root node in an empty V-position triggers the spreading
of the non-dependent element from the following vowel, the only possible scenarios
for lengthening of the complex XY-segments in (17a) and (17b) are those
depicted in (19a) and (19b), respectively.
(19) a. V2
C V1
b. V2
C V1
| | | |
● ● ● ●
| |
X Y
Y X
32
In fact, the derivations in (19) involve two reverse processes: both linking and
delinking as well. Disconnecting of the bottom elements in (19) (indicated by
dashed lines) follows from the dependence relation: since spreading of nondependent
elements entails spreading of those that are dependent on them, the
embedded elements in (19) must be disconnected. As a consequence, lengthening
of complex vowels goes hand in hand with a simplification of their melodic
structure: thus, a change of their quality is expected.16
Summing up: differences among length alternations result from the underlying
structures of particular vowels and/or the structure of the empty V-slots to
which these vowels spread.
2.3.3  Mid-Vowel Length Alternations
How does the dependency model account for the morphemes that are involved
in three different types of length alternations? From what has been said so far it
follows that the pure lengthening of mid vowels (which are complex segments
themselves) always includes the root-node spreading. Thus, in isolation, the pure
lengthening of [o] to [oː] can be modelled in three different ways depicted in
(20).
(20) a. V1
→ V2
C V1
b. V1
→ V2
C V1
c. V1
→ V2
C V1
| | | | | |
● ● ● ● ● ●
| | | | | | | |
U U A A A U A U
| | | |
A A U U
All possible scenarios for the o-to-ó lengthening in (20) are based on the same
mechanism: the root node—a complex of two elements A and U—spreads to
the preceding empty V-slot (V2
). The difference is in how these two elements are
organized. Thus, it is either the A-dependent (20a) or the U-dependent segment
that spreads (20b), or, finally, what spreads is a segment in which neither of the
elements is dependent on the other, as in (20c). The result is always the same:
regardless of a lexical structure associated with the mid vowel [o], the lexical
16	 Note that there are also other approaches to melody dependency; see Ewen (1995) for
an overview.
33
structure gets always preserved under the root-node spreading and the result
is [oː]. In sum, the pure lengthening does neither indicate, nor exclude any underlying
structure of the short vowel. The actual structure is only revealed in
that qualitative lengthening, which involves spreading of individual, i.e., nondependent
elements.
Given the dependency relations between its elements A and U, there are two
types of qualitative lengthening of [o] and both of them are shown in (21): the
short mid back vowel [o] either raises to [uː] in (21a), or it lowers to [aː], as
shown in(21b).
(21) a. [o] > [uː] b. [o] > [aː]
V1
→ V2
C V1
V1
→ V2
C V1
| | | | | |
● ● ● ● ● ●
| | | |
U U A A
| |
A A U U
(21) suggests a direct correspondence between the type of qualitative lengthening
and the lexical structure of a particular vowel: the A-dependent mid vowel
[o] lengthens to [uː], while the U-dependent mid vowel lengthens to [aː]. Moreover,
both of them can lengthen to [oː]: in case the V-slot (the target of the melody-spreading)
lacks its own root node; see (20a) and (20b) above. In sum, the
proposed model derives the contrast between morphemes showing the alternations
[o]~[uː] and [o]~[oː] on the one hand and those showing the alternations
[o]~[aː] and [o]~[oː] on the other.
Let me now turn back to the examples introduced in (13a) (repeated below in
(22)): a single morpheme—the root, in this case—is involved in both types of the
qualitative length alternations with [o].
(22) kr[uː]č-ek kr[aː]č-et kr[o]k
‘small step’ ‘to walk’ ‘a step’
s-ch[uː]d-ek s-ch[aː]z-et s-ch[o]d
‘small stair’ ‘to step down’ ‘a step’
34
po-kl[uː]p-ek vy-kl[aː]p-ět po-kl[o]p
‘small hatch’ ‘to dump’ ‘hatch’
tv[uː]r-ce vy-tv[aː]ř-et vý-tv[o]r
‘creator’ ‘to create’ ‘creation’
str[uː]j-ce pře-str[aː]j-et str[o]j-it
‘originator’ ‘to dress up
(repeatedly)’
‘to dress up’
kr[uː]p-ěj s-kr[aː]p-ět kr[o]p-it
‘drop’ ‘to spray’ ‘to spray’
(repeatedly)
So far I have been assuming that the qualitative lengthening yields different results
depending on the lexical structure of the particular mid vowel. From this
perspective, however, only one of the two long variants of the roots in (22) can
be a result of phonological computation; the other one must be lexically stored.
How to decide which of the two alternants, i.e., the raised [uː] or the lowered [aː],
is lexically stored and which is derived?
A strong argument for the derivational account of the alternation [o]~[uː]
is its distribution across different types of morphological contexts: the [o]~[uː]
alternation is seen in roots (23a), in suffixes (23b) and in prefixes as well (23c).
Thus, we can conclude that the underlying identity of a short mid back vowel [o]
in Czech is a complex segment, in which A depends on U.
(23) a. chob[o]t chob[uː]t-ek ‘trunk, small trunk’
b. muž-[o]v-a muž-[uː]v ‘man’s, Gsg, Nsg’
c. pr[o]-nik-nout pr[uː]-nik ‘to penetrate, a penetration’
Compared to [o]~[uː], the alternation [o]~[aː] is restricted only to a closed set of
monosyllabic roots. This by itself might be an indicator of the lack of active phonological
process in Czech that lowers (and lengthens) [o] to [aː]. Therefore, the
long variants of the roots that show a low vowel [aː]—as in (22)—must thus be
stored as independent lexical items. This conclusion complies precisely with our
suggestion that all o’s in Czech are A-dependent segments. It, however, does not
account for the fact that the low-vowel alternants are not distributed randomly,
but follow a clear pattern. In particular, the appearance of the lengthened and
lowered vowel [aː] in the root correlates with the iterative reading of that root.
It is, therefore, possible to assume that both the length alternations, i.e., [o]~[uː]
and [o]~[aː], result from phonological computation. Since the short vowel [o]
35
is an A-dependent segment, an additional mechanism needs to be invoked to
derive the long [aː], a simplex A-segment. In what follows I propose that this additional
mechanism might be a prosodic template.
In fact, to spell-out the iterative morphosyntactic structure as a result of a
prosodic template has already been proposed by Scheer (2001, 2003, 2004a).
Without going into details, the core idea of Scheer’s proposal is that there is an
iterative template comprising two morphological units, i.e., the verbal root and
the so-called thematic suffix. This template, in turn, regulates the distribution of
vowel length. In particular, monomoraic thematic suffixes, i.e., those consisting
of a single short vowel, need to merge with bimoraic roots to satisfy the iterativetemplate
restriction. Hence: if a verbal root such as /krok/ (and the others listed
in (22)) contains a single short vowel, it has to lengthen when it merges with
a short thematic vowel in the iterative context (i.e., with the -e in the case at
hand). Suppose now, that the iterative template cares not only about the length of
the vowels (as proposed by Scheer), but also about their quality, which, in turn,
means that the o-to-á lowering is triggered by the insertion of the template into
the morphosyntactic structure. The claim that the iterative template is responsible
also for the vowel-quality change is supported by the fact that both vowels
included in the templatic domain, i.e., the root vowel and the thematic vowel,
change their articulation. This is illustrated by aspectual pairs in (24). The perfective
verbs in the left-hand column contain roots showing a short mid vowel
[o] followed by a high-vowel theme [i]. In their iterative cousins (on the right),
long low vowels [aː] are followed by a short mid vowels [e].
(24) perfective iterative
po-kl[o]p-[i]-t po-kl[aː]p-[e]-t ‘to tilt’
vy-tv[o]ř-[i]-t vy-tv[aː]ř-[e]-t ‘to create’
o-t[o]č-[i]-t o-t[aː]č-[e]-t ‘to turn’
s-kr[o]p-[i]-t s-kr[aː]p-[e]-t ‘to spray’
pře-str[o]j-[i]-t pře-str[aː]j-[e]-t ‘to dress up’
Both vowels in the iterative domain share a piece of their internal structure, i.e.,
an A-element. From this perspective, the perfectives differ from the iteratives: the
root vowel [o], corresponding to a complex UA-segment, and the theme vowel
[i], corresponding to a simplex I-segment, do not share any element. The iterative
template thus might be seen as a vowel-harmony domain with an extension
of the A-element. In particular, the contrast between UA-segment in the perfective
root and A-segment in its iterative cousin might be taken as strengthening
36
of the A-element in the iterative context. And, similarly, the contrast between the
perfective and the iterative theme, i.e., I-segment and IA-segment, respectively,
might be interpreted as the addition of an A-segment.17
I leave the implementation
of this idea for future research. For our purposes, it is important to note
that there is a systematic connection between particular types of vowel length
alternations and morphosyntactic contexts, in which they participate. And that
these cases of length alternations can arise from a phonological computation.
I established that the mid back vowel [o] alternates with a long [uː] and this
alternation is a default. This, in turn, means that [o] corresponds to a complex
segment, in which A is dependent on U. Since the mid front vowel [e] alternates
typically with [iː], as illustrated in (25), we might want to apply the same logic
as in the previous case and claim that [e] corresponds to a complex segment, in
which A is dependent on I. The e-to-í lengthening is depicted in (26).18
(25) a. parap[e]t parap[iː]t-ek ‘a sill, a small sill’
b. uči-t[e]l uči-t[iː]l-ek ‘a teacher, a small teacher’
(26) [e] > [iː]
V1
→ V2
C V1
| | |
● ● ●
| |
I I
|
A A
To conclude: I align differences in length alternations to three different sources
(and their combinations): (i) the underlying structures of particular vowels
(only non-dependent melodic elements can spread individually), (ii) the structure
of the empty V-slots, to which these vowels spread (the individual-element
17	 Note that Cyran (2010) identifies a similar kind of interaction between root vowels
and thematic vowels in Polish. In particular, Cyran shows that a short mid vowel [o]
alternates with a short low vowel [a], if a verbal root is merged with a low-vowel theme.
Also he explains this pattern as arising from a prosodic template.
18	 The assumption that the mid front vowel is an A-dependent segment is further supported
by the fact that lexically long e’s often change to long i’s in the colloquial speech.
For example, ml[eː]ko ‘milk’ is spelled as ml[iː]ko and so on.
37
spreading targets only root nodes), (iii) the phonological structure of the lexical
item, which the particular morpheme is merged with (the o-to-á lowering is triggered
by the merger with the iterative template). Before discussing morphemic
length alternations in more detail, we turn to u-to-ou diphthongization that we
identify as a special subtype of the qualitative lengthening.
2.3.4  Diphthongal Length Alternations
As I have already mentioned, the high back vowel [u] has two long counterparts:
[uː] and [ou]. The long [uː] is a simplex vowel, similarly as to its short cousin. It
might be represented either as in (27a) or as in (27b). The falling diphthong [ou],
on the other hand, is more complex melodically. Its structure is shown in (27c):
the element U is shared by both segments of the diphthong and, in addition, the
first segment has an extra A-element. As a consequence, the U-element is interpreted
twice in (27c): in the V1
—where it corresponds to [u]—and in the V2
(that
precedes it), in which it is merged with the A-element and the merger of U and
A corresponds to a mid vowel [o].
(27) a. V2
C V1
↔ [uː] b. V2
C V1
↔ [uː] c. V2
C V1
↔ [ou]
| | | | | |
● ● ● ● ●
| | | | | |
U U A U
If the diphthong [ou] has the structure as in (27c), then the u-to-ou lengthening,
seen for example in the diminutive context (cf. kl[u]k – kl[ou]č-ek ‘boy, small
boy’), must involve linking of two elements, i.e., U and A. There are two possibilities
of how to interpret the linking of the A-element: it is either a part of
the short vowel [u] or it belongs to the V-slot, which this vowel spreads to. The
first scenario is depicted in (28a): the short vowel [u] is a complex UA-segment
lexically, but only one of the two elements (i.e., U) is lexically associated with
the root node. The lengthening, then, involves a spreading of both the lexically
associated U-element and the lexically floating A-element. The second possible
scenario in (28b) suggests that the A-element is a part of the V-slot, to which the
simplex U-segment spreads.
38
(28) a. V1
→ V2
C V1
b. V1
→ V2
C V1
| | | | | |
● ● ● ● ● ●
| | | | |
A U A U U A U
To decide which of the two approaches to the u-to-ou diphthongization is more
plausible, we have to consider the identity of the V-slot, to which the short vowel
spreads. In what follows, I argue that lengthening of base vowels in diminutives
is induced by a prosodic affix. Under this assumption, the target of the melody
spreading, i.e., the V2
in (28), and the short vowel (V1
) belong to two distinct
lexical items. If the u-to-ou lengthening follows the derivation in (28b), the diminutive
prosodic affix must be an A-segment underlyingly. In (29), I apply the
proposal to a high front vowel [i] and mid vowels [e] and [o].
(29) a. V1
→ V2
C V1
b. V1
→ V2
C V1
c. V1
→ V2
C V1
| | | | | | | | |
● ● ● ● ● ● ● ● ●
| | | | | | | | |
I A I I A I U A U
| |
A A A A
Triggered by the insertion of the diminutive A-segment, the lengthening of the
front vowels, i.e., the high [i] and the mid [e], yields a falling diphthong [ei] in
(29a) and (29b). In (29c), the mid back vowel [o] undergoes a diphthongization
as well: the result is [ou]. However, only bases containing the high back [u]
diphthongize in the diminutive context: bases with [i] undergo pure lengthening
and those with mid vowels undergo raising in diminutives, as illustrated in the
table below.
(30) a. hř[i]b hř[iː]b-ek ‘boletus’
b. žl[e]b žl[iː]b-ek ‘valley’
c. hr[o]b hr[uː]b-ek ‘grave’
39
In sum, the V-slot involved in the diminutive prosodic affix is melodically empty.
It is provided with a root node and the root node then accommodates non-dependent
elements of a particular short vowel. Since the short high back vowel [u]
contains two such elements, i.e., the lexically associated U and floating A, both of
them are linked to the preceding V-slot. As a consequence, the falling diphthong
[ou] is derived.
Furthermore, the length alternations with the high back vowel, i.e., [u]~[ou]
and [u]~[uː], do not overlap (as opposed to the alternations with [o] discussed
above): I am not aware of any morpheme that shows both of the alternations.
The contrast between u’s that alternate with [uː] and that alternating with the
diphthong thus can be encoded lexically: the former are stored as simplex Usegments,
the latter are stored with a floating A-element in addition.
The next section elaborates on the idea that vowel-length alternations can be
induced by prosodic affixes and that these prosodic affixes are parts of the morphosyntactic
structure.
2.4  Morphemic Length Alternations: Prosodic Affixes
As has been argued in the literature, e.g. Davis & Ueda (2002), Bye & Svenonius
(2012), Trommer & Zimmermann (2014), inter alia, morphosyntactically
determined manipulation with prosodic features such as vowel length or tone
can be effectively reduced to affixation of phonologically incomplete lexical
items; that is items whose phonological representation consists of floating segmental
features or empty pieces of prosodic structure. Then, phonetic interpretability
of such items depends on their integration into the phonological structure
of morphemes to which they concatenate. Building on this general idea, I argue
that also some instances of vowel-length manipulation in Czech arise from the
prosodic affixation when an empty V-slot is inserted next to a lexically short
vowel in order to spell out a part of the morphosyntactic structure. Before developing
an analysis of particular cases of the prosodic affixation in chapters 3
and 4, let me first show how the lexical shape of the prosodic affix determines its
merger with the underlying short vowel.
Up to now, I have been assuming that derivation of long vowels involves leftward
spreading of either individual elements or entire root nodes. However, as I
have already mentioned, processes operating on the melodic level can in general
work in both directions (as opposed to government and licensing, i.e., relations
holding among prosodic units, that work only regressively). The null hypothesis
is that this also holds for melody spreading. As a consequence, right-branching
long vowels are expected to exist, where targets of spreading follow the source.
40
Given that lexically short vowels can lengthen to both directions, prosodic affixes
of two shapes, i.e., VC and CV, can exist. This prediction is depicted in (31).
There is a root in (31a) whose phonological structure is made of three segments:
a short vowel and two consonants and these segments are arranged into
the CVC-shape. In (31b), the root is merged with a VC-affix (shaded in the diagram),
which actually means that the prosodic structure in the left-hand side
of root’s vowel is extended. By contrast, the merger of a CV-affix in (31c) leads
to the prosodic extension to the right of the root vowel. In sum, the VC-affix
in (31a) induces leftward spreading of the short vowel (VX
) and the CV-affix
its rightward spreading (31b). (Note that the prosodic affixes of both prosodic
shapes can further differ in their content: they are either provided with root
nodes or not, and/or they can contain floating melodic features.)
(31) a. Cx
Vx
Cy
b. Cx
V C Vx
Cy
c. Cx
Vx
C V Cy
| | | | | | | | |
● ● ● ● ● ● ● ● ●
| | | | | | | | |
C V C C V C C V C
Let me now explore two predictions of the proposal above. The first prediction
concerns the way how prosodic affixes are integrated into the phonological
structure of morphemes which they merge with. Since prosodic structure
is represented as a sequence of alternating C-slots and V-slots, then the lexical
shape of the prosodic affix determines unambiguously its position with respect
to the short vowel: VC-affixes are “prefixed” to it, CV-affixes, on the other hand,
are “suffixed”. This is an advantage of the Strict CV approach to vowel lengthening
compared to a moraic one (see e.g. Hayes 1989): a mora as a prosodic unit
has no underlying structure, and, thus, moraic affixes that trigger lengthening of
short vowels, are in general ambiguous between suffixes and prefixes. To avoid
the prefix—suffix ambiguity of moraic morphemes, diacritic marks are typically
used in their representation; for example Trommer & Zimmermann (2014) represent
suffixal and prefixal moras as /-μ/ and /μ-/ respectively, using traditional
hyphens as diacritics, thus, without a phonological content per se.
In the proposed model, the position of prosodic affixes with respect to affected
short vowels is fully predictable from their phonological structure. At the same
time, the phonological difference between VC- and CV-affixes itself does not
indicate their position within polysyllabic strings. In other words, if a morpheme
41
to which a lengthening affix is concatenated contains several short vowels, we
need to specify to which of them it should be prefixed or suffixed.
The idea that position of affixes with respect to their morphological bases can
be phonologically unpredictable, is elaborated in Yu (2003). He analyzes infixation
from a cross-linguistic perspective and develops a typology of phonological
positions to which infixes can be associated. These prosodic anchors typically
include peripheral or prosodically prominent segments of morphological bases
that undergo infixation. However, Yu (2003) shows rather convincingly that in
many cases, a particular connection between an affix and its anchor simply cannot
be predicted from the general phonological properties of a given language,
and, thus, it must be lexically encoded.
The existence of affix-specific prosodic anchors can be illustrated in an illuminating
way in the following example from Czech. There is a bisyllabic root
‘stone’ and both its vowels alternate in length. The first vowel of the root is long
only in the nominative/accusative singular context, as shown in the upper part of
the leftmost column in table (32); in all other appearances, including the rest of
the declensional paradigm and the derivatives of various types (the third column
of the table), the first vowel is a short [a]. The second vowel of the root is a short
[e], except in the diminutive form, in which it is replaced by its long cousin [iː],
i.e., kam[iː]n-ek.
(32) Sg Pl
N k[aː]men k[a]men-y k[a]men-í ‘mass of stones’
A k[aː]men k[a]men-y k[a]men-ný ‘stony’
G k[a]men-e k[a]men-ů k[a]men-ík ‘stonemason’
D k[a]men-i k[a]men-ům k[a]men-olom ‘quarry’
L k[a]men-i k[a]men-ech
I k[a]men-em k[a]men-y k[a]m[iː]n-ek ‘small stone’
Given the morphosyntactic distribution of the long variants of the root, it is
plausible to assume that they arise from the merger of two lexically distinct prosodic
affixes, each being associated with a particular piece of the morphosyntactic
structure. Further arguments supporting this assumption are provided in
chapter 4, which primarily focuses on length alternations in diminutives. What
is relevant at this point is that the “inflectional” affix and the “diminutive” affix
must be lexically specified for different prosodic anchors to ensure that the first
one induces lengthening of the first syllable, while the second one the second
syllable of the root.
42
The second prediction of the contrast between VC-affixes on the one hand
and CV-affixes on the other concerns the phonological behavior of long vowels
derived by them. In particular, lengthening induced by the merger of CV-affixes
is predicted to be sensitive to the phonological context to the right of the given
short vowel. By contrast, insertion of the VC-affix provokes lengthening of a
short vowel, irrespective of its phonological context. This prediction stems from
the claim that short vowels (or to be more precise, their individual elements or
entire root nodes) can spread only to licensed V-slots.
When a VC-affix is merged, a short vowel itself licenses the inserted V-slot,
which actually means that the vowel automatically lengthens; this scenario is
depicted in (33a). The merger of a CV-affix, on the other hand, calls for external
licensing (because it operates only regressively). In (33b), the licensing comes
from the V-slot that immediately follows the inserted prosodic affix and that is
associated with a melodic level. If there is no such a V-slot available, as in (33c),
affix’s empty V-position is not licensed and the short vowel therefore does not
lengthen. Put it in more conventional terms, insertion of the CV-affix leads to
vowel lengthening in open syllables (33b), but not in closed syllables (33c).
(33) a. L
b. L
c.
V C V V C V C V V C V C (V C)
| | | | | | |
● ● ● ● ● ● ●
| | | | | | |
V V C V V C C
Both types of long vowels, i.e., left-branching and right-branching, have already
been proposed by Scheer (2004:267f.).19
In Scheer’s model, however, the direction
of branching distinguishes between non-alternating lexically long vowels
and those that are involved in length alternations: left-branching vowels, being
self-licensors, are invariably long—as opposed to their right-branching cousins
that alternate between long and short, depending whether they are followed by
proper licensors or not.
At the first sight, Scheer’s account to vowel length seems to contradict the
model proposed here when also left-branching long vowels can arise from
19	 Note that in Scheer’s terminology left-branching and right-branching vowels are called
head-final and head-initial, respectively.
43
phonological computation, i.e., from lengthening of lexically short vowels, based
on melody-spreading to the licensed position. By contrast, in Scheer’s model,
left-branching structures represent long vowels that have been already stored in
the lexicon. However, Scheer in fact also assumes that left-branching representations
involve melody-spreading. The only difference between Scheer’s proposal
and my proposal is the origin of the V-slot targeted by the leftward spreading.
In other words, if both the licensor and the licensee are part of the same lexical
entry, then leftward melody-spreading is established already in the lexicon—and
that actually means that the morpheme in hand shows a long vowel when inserted
into a syntactic structure. On the other hand, if the licensor and the licensee
belong to two separate morphemes, then the leftward melody-spreading is
triggered only after the merger of the two lexical items and then, the licensor is a
lexically short vowel, while the licensee is an empty V-slot.
It should be noted that Scheer (2004:251) explicitly states that he does not deal
with those cases, in which the vowel length has a morphemic status. Nevertheless,
he suggests a solution somewhat similar to my proposal: the right-hand
V-slot can either be lexically present—or not. The presence of the right-hand
licensee in the lexical representation of a given morpheme differentiates between
two length-manipulating processes: closed syllable shortening and tonic lengthening
(the latter targets stressed vowels in open syllables). Thus, in Scheer’s model,
there is a single mechanism underlying both of these (rather opposite) processes—a
spreading of vowel melody onto the licensed V-slot. The only difference
is the origin of the licensee: the empty V-slot either originates due to stress (in
tonic lengthening) or it is already present in the lexical entry (in closed syllable
shortening). In both cases, the lexically associated vowel spreads to an empty
V-slot. This empty prosodic position, then, either materializes as stress—in this
case, its presence arises from an insertion in the phonological component—or
the empty prosodic position lexically encodes a vowel that alternates in length
depending on the syllabic context to its right. (Note that viewed from this perspective,
the term closed syllable shortening is rather misleading because shortening
of long vowels normally involves delinking of one of their two prosodic
positions: however, as we can see in (33c), derivation of the short vowel alternant
does not involve melody delinking, but rather “melody non-spreading”).
Summing up, I propose that vowel lengthening can work in two directions
which distinguish between syllable-sensitive length alternations and those that
are independent of the syllable context: the latter involve leftward lengthening,
the former the rightward one. I furthermore propose that both types can arise
from merger of prosodic affixes. This, in turn, means that the lexical shape of the
44
prosodic affix itself, i.e., VC vs. CV, determines whether the short vowel, which
the particular affix is concatenated to, would lengthen automatically or depending
on its syllable environment. In next two chapters I argue that both types of
prosodic affixes, hence both types of morphemic lengthening, exist in Czech.
45
3.  Length in Hypocoristics
This chapter is devoted to derivation of hypocoristics. There are two formal
processes involved in their derivation: truncation of a base to one syllable and
lengthening of its vowel. I argue that both processes have morphemic status:
each realizes a certain part of the hypocoristic morphosyntactic structure. Following
the central ideas of the theory of Nanosyntax (Caha 2009, Starke 2009), I
propose that the two components of the hypocoristic meaning identified in the
literature, i.e., familiarity and closeness, are encoded syntactically as two hierarchically
ordered projections. Each of them has its own spell-out: the familiarity
projection is spelled out as a prosodic template that triggers truncation when
applied to a root; the closeness projection, on the other hand, is spelled out by
a prosodic affix inserted into the root, so that its insertion lengthens the root’s
vowel. Using Strict CV phonological representations (introduced in the previous
chapters), I argue that the prosodic affix is of the VC-shape, and that moreover,
the VC-affix is empty of the root node: that then means that the whole basevowel
melody spreads. As a consequence, hypocoristics expressing closeness
have lengthened vowels whose quality does not differ from their short cousins.
3.1  The Basic Phonological Pattern
Let us start with data in table (1). The table compares hypocoristic forms of masculine
first names (in the right part of the table) with their basic, i.e., “pragmatically-neutral”,
forms (on the left). Ignoring the ending of the hypocoristics for
now, two processes manipulating phonological structure can be identified: truncation
and—last but not least—the process which is of prime interest to us, i.e.
vowel lengthening. The hypocoristics have no special suffix, since the final vowel
[a] found across all hypocoristics is an inflectional marker expressing case and
number features. I will return to this morpheme later when discussing Bethin’s
(2003) templatic analysis of a-final hypocoristics in Czech. Because of the lack of
a suffix, I will call this group suffix-less hypocoristics.
(1) first name hypocoristic form
i > iː F[i]lip F[iː]l-a
u > uː D[u]šan D[uː]š-a
e > eː Kl[e]ment Kl[eː]m-a
o > oː Br[o]nislav Br[oː]ň-a
a > aː J[a]roslav J[aː]r-a
46
From the perspective of the proposed account of vowel length, in which lengthening
is induced by extension of the prosodic structure in close proximity of the
affected vowel, we need to find answers to the three following questions: 1. What
type of an empty V-slot is inserted in hypocoristics? 2. Is it inserted to the left of
the root vowel or to its right? 3. What is the origin of the inserted material? In
other words, does the vowel length in hypocoristics have a morphemic status?
The answer to the first question comes out clearly from the first column in table
(1): the mid vowels do not undergo raising when lengthened. More examples
illustrating this are provided in (2).
(2) first name hypocoristic form
a. P[e]tr P[eː]ť-a
Bř[e]tislav Bř[eː]ť-a
M[e]toděj M[eː]ť-a
b. B[o]humil B[oː]ž-a
Ant[o]nín T[oː]ň-a
M[o]nik-a M[oː]ň-a
Given the examples above, we can conclude that the lengthening in the hypocoristics
involves the spreading of the entire root node while the whole internal
structure of the lengthened segment is preserved, i.e., all its elements and also
dependency relations still hold among them. As a consequence, neither raising
nor other vowel-quality changing processes are triggered. To sum up: the Vslot,
inserted during hypocoristic derivation, which is a target of the root-node
spreading, is empty of any prosodic structure lexically.20
Let me now turn to the other two abovementioned questions concerning
vowel length in hypocoristics. These can only be answered on basis of a closer
examination of their morphological properties.
Recall that if an empty V-slot were to be inserted to the right of the given vowel,
the lengthening should be syllable-sensitive. At the first sight, this is exactly
20	 Originally, mid vowels rose to high even in hypocoristics; see e.g. Svoboda (1964).
Synchronically, the raised forms like T[uː]m-a (from T[o]máš) or Kl[iː]m-a (from
Kl[e]ment) are not hypocoristics, but serve as official surnames. In the proposed model,
the diachronic change from qualitative change to pure lengthening is the change in
the lexical representation of the prosodic morpheme which realizes a given part of
the hypocoristic structure. In particular, the change consists of simplification of the
prosodic structure which loses the root-node in the empty V-slot.
47
what we see in (3), where a new type of hypocoristic is introduced. This new type
is given in the last column and it exhibits the suffix -d. In the middle column,
there is the old type of suffix-less hypocoristics, and in the first column the bases
are listed. There are thus five pairs of hypocoristics, each derived from the same
first name.
(3) full name lengthened hypocoristic d-suffixed hypocoristic
M[i]lan M[iː]l-a M[i]l-d-a
J[u]lius J[uː]l-a J[u]l-d-a
Př[e]mysl Př[eː]m-a Př[e]m-d-a
Ant[o]nín T[oː]ň-a T[o]n-d-a
J[a]roslav J[aː]r-a J[a]r-d-a
The table reveals a striking pattern: lengthening is triggered only when the case
marking vowel [a] is adjacent to the truncated root; if these two morphemes
are separated by a consonantal suffix -d, the root vowel remains short. In other
words, quantity of the root vowel seems to be dependent on a number of consonants
that follow it: presence of one consonant correlates with vowel length,
presence of two consonants with its shortness.
In sum, lengthening seems to be triggered only in open syllables (Já.ra); in
closed syllables, i.e., when the suffix -d appears after the root, it is blocked (Jar.da).
3.2  Bethin’s (2003) Templatic Analysis
Bethin (2003) analyzes the patterns shown above within an Optimality Theory
framework. Her main claim links vowel length alternation in hypocoristics to
a set of size-restricting prosodic constraints associated with hypocoristics as a
particular morphosyntactic category. Hence, these constraints derive (what we
might call) a “hypocoristic template”.
In hypocoristics (still following Bethin 2003), a prosodic word must completely
coincide with a metrical foot, which, by force, is bisyllabic. This is a reason
why polysyllabic bases such as Jaroslav are truncated to a single syllable.
The second syllable, required by the hypocoristic template, is added by the vocalic
suffix -a. Furthermore, to capture the complementary distribution of vowel
length and the consonantal suffix, Bethin (2003) postulates a weight-restricting
constraint: one of the two syllables must be heavy, i.e., bimoraic. In the d-suffixed
forms, the first syllable is heavy due to a weight-by-position: the first syllable in
Jar.da (and other hypocoristics of this type, as shown in the right-most column
in table (3)) is closed by a coda consonant contributing to the syllable weight,
48
hence, the first syllable counts as moraic. By contrast, in forms like Já.ra (those
without the consonantal suffix), both syllables are open: thus, by lengthening the
stem vowel, a required heavy syllable is created (under the generally accepted
assumption that long vowels are prosodically heavy by default). In other words,
the first syllable of both the suffix-less and the d-suffixed hypocoristic is heavy: it
is heavy due to the vowel length (Jáμμ.ra) or due to the coda (Jaμrμ.da).21
Hence, Czech follows the cross-linguistically strong trend for hypocoristics: a
fixed prosodic size. Therefore, truncation of morphological bases seems to be a
rather common way languages use to fit a given prosodic template, as illustrated
in table (4) by examples from English and Japanese.22
(4) Hypocoristic truncation in English and Japanese (Akasaka & Tateishi 2001)
base hypocoristic
English Pamela Pam
Elisabeth Beth
Jeffrey Jeff
Japanese Saburo Saa-chan, Sabu-chan
Haruka Haa-chan, Haru-chan
Hiroshi Hii-san, Hiro-san
The Japanese examples show that templatic restrictions can control also prosodic
size of proper morphological subparts of hypocoristics. As pointed out by Akasaka
& Tateishi (2001:12), the vocative suffixes -chan and -san stand outside the
templatic domain; hence it is only the prosodic size of the root that is restricted.
As we can see in (4), the full first names are truncated either to one syllable—or
two; in the former case, vowel lengthening is triggered in order to satisfy the
bimoraic template.
In sum, hypocoristic formation in Japanese involves the same phonological
processes that we have seen in Czech: truncation and vowel lengthening. Both
of these processes have been interpreted as a consequence of the weight-restricting
constraints associated with this particular morphosyntactic category. The
21	 One can, of course, ask why it is the first syllable—and not the second one—that gets
always lengthened. Bethin (2003) explains it by pointing to a general Czech tendency
to trochaic feet.
22	 As pointed out by Alber & Arndt-Lappe (2012), this output-oriented truncation must
be distinguished from what is usually called subtractive morphology: it refers to a phenomenon,
when a deleted phonological material is of the fixed prosodic structure.
49
variable, then, is the size of the morphological domain to which these constrains
should be applied. According to Akasaka & Tateishi (2001), in Japanese, this
domain is just the root and its weight must be exactly two moras. By contrast,
Bethin’s (2003) analysis of Czech considers the templatic domain to be the root
plus all suffixes that follow it.
What remains rather problematic is that the root-following suffixes include
inflectional markers, specifically portmanteau morphemes of case and number.
For instance, the suffix -a—seen both in the suffix-less forms (Jár-a) and
the d-suffixed forms (Jard-a)—is an inflectional marker for features denoting
nominative case and singular number. If it contributes to the templatic domain
(Bethin’s crucial assumption), then root-vowel lengthening required to create a
bimoraic syllable, should be triggered only in the context of prosodically light
endings, i.e., only before short vowels. In other words, under Bethin’s analysis,
root length alternations are expected to occur when the suffix is -a (the nominative
singular suffix [a] is light, hence, the root vowel lengthens), but not when the
suffix is -ů (the genitive plural suffix [uː] is heavy, hence, the short root vowel).
But as is illustrated in table (5), this prediction is wrong: all inflectional forms
of hypocoristics have lengthened roots—regardless of the prosodic size of the
ending. Notice, in particular, the two middle columns: the lengthening occurs
even before diphthongal and long-vowel endings, both of which are bimoraic.
Moreover, nominative plural forms (in the right-most column, ending in -ov-é),
violate systematically everything Bethin (2003) postulates: not only is the oneheavy-syllable
constraint violated, but so is the bisyllabic one.
(5) VV-[a] (Nsg) VV-[ou] (Isg) VV-[uː] (Gpl) VV-ov-[eː] (Npl)
M[iː]l-a M[iː]l-ou M[iː]l-ů M[iː]l-ov-é
J[uː]l-a J[uː]l-ou J[uː]l-ů J[uː]l-ov-é
Př[eː]m-a Př[eː]m-ou Př[eː]m-ů Př[eː]m-ov-é
T[oː]ň-a T[oː]ň-ou T[oː]ň-ů T[oː]ň-ov-é
J[aː]r-a J[aː]r-ou J[aː]r-ů J[aː]r-ov-é
To conclude: Bethin’s analysis of hypocoristics based on the prosodic constraints
that count both the number of syllables and their weight, works only for inflected
forms derived by one-short-vowel endings. But in reality, only five out of the
twelve paradigm cells (i.e., combinations of six cases and two numbers) contain
a single short vowel; in other words, most of the hypocoristic forms predicted as
ungrammatical by Bethin (2003) are in fact perfectly natural (and clearly gram-
matical).
50
Furthermore, Bethin (2003) ascribes the absence of lengthening in d-suffixed
forms to the presence of root-final (moraic) codas. It is well-known from the
literature that weight-by-position has a parametric character (cf. Hayes 1989,
Rosenthall & van der Hulst 1999, inter alia). However, even according to Bethin
(2003:64), treating codas as weight-bearing is “something which is otherwise not
characteristics of Czech phonology”. Moreover, the moraicity of codas is claimed
to be sensitive to phonological properties rather than to morphological categories:
the relevant factors cross-linguistically are the quality of the coda consonant
(sonorant vs. obstruent) or its position within a string (internal vs. final coda);
on the other hand, information about the codas’ morphological context does not
seem to play any role at all; it is of no consequence whether the codas appear in
a hypocoristic or, for example, a present tense context.
However, even if we admit that the weight-by-position is applied selectively
to hypocoristics (as claimed by Bethin 2003), the crucial problem is that codas
should be moraic not only in the d-suffixed forms, but also in suffix-less hypocoristics
with roots ending in a cluster. In other words, the weight-based approach
predicts that lengthening does not take place in such forms, and consequently, it
cannot explain the existence of forms such as those listed in (6): here, the lengthening
is triggered before root-internal codas and, thus, trimoraic initial syllables
are produced.23
(6) full name hypocoristic
[a]rtur [aː]μμrμ.ť-a
B[a]rtoloměj B[aː]μμrμ.ť-a
M[a]rtin M[aː]μμrμ.ť-a
Summing up: the data presented in table (5) suggest that the templatic domain
does not cover the whole hypocoristic morphological structure—but only a subpart
of it (similarly to Japanese). I elaborate on this idea in the next section: specifically,
I argue that the hypocoristic template scopes just over the root and its
main goal is to make it monosyllabic. Furthermore, I argue that the root-vowel
length is not templatic, as suggested by the data in (6). Rather, I claim that length
arises from insertion of a prosodic VC-affix, a part of the hypocoristic meaning.
23	 It is fair to say that some suffix-less hypocoristics may exhibit length variation. For
instance, Martin has two hypocoristic forms: a long M[aː]rťa and a short M[a]rťa.
Such variation is discussed further on.
51
Thus, the complementary distribution between lengthening and segmental suffixation
is a consequence of their distinct morphosyntactic structures.
3.3  Decomposing the Hypocoristics
Alber & Arndt-Lappe (2012), following Schneider (2003), identify two main
pragmatic functions of hypocoristics: one is to mark the speaker’s familiarity
with the referent, the other is to signal a close relationship between the speechact
participants. They argue that in English, these two pragmatic functions are
marked by different means: monosyllabic truncations such as Tom (from Thomas)
express just familiarity, bisyllabic [i]-ending truncations such as Tommy, on
the other hand, are signal a particularly close relationship.24
Following a cartographic approach (Cinque & Rizzi 2009), I assume that the
two formally identifiable components of the hypocoristic meaning correspond
to two syntactic features, i.e., Fam (= familiarity) and Cls (= closeness), which
(rather than bundled together) are organized in an implicational hierarchy. Specifically,
there is a containment relation between the two hypocoristic forms: the
bisyllabic truncation Tomm-y contains the monosyllabic one Tom. By the logic
of the cartographic approach, I take the containment of the forms to indicate an
actual containment in terms of the syntactic structure, which leads to the hierarchy
Cls>Fam. This hierarchy, a part of the hypocoristic functional sequence, says
that the familiarity meaning (encoded syntactically in the Fam-head) is present
in both the familiarity-expressing hypocoristics and those expressing closeness.
The proposal is depicted in (7) below. Here I show the syntactic structures
of both types of hypocoristics, i.e., the Tom-type in (7a) and the Tommy-type in
(7b). In describing the structures, I am ignoring the arrows for the time being,
and I come back to their meaning later. Both structures contain an XP, which is
a complement of the Fam-head in (7). The XP stands for a sequence of projections
included in the syntactic structure of the first name, all of it bottoming out
in a nominal-root phrase (NP). The internal structure of the XP (or, to be more
precise, its highest part encoding gender features), is discussed later on, when
we focus on the morphosyntactic properties of different types of hypocoristics
in Czech. Both forms also contain a Fam-head, and the hypocoristic expressing
a close relationship has in addition (the closeness-expressing) Cls-head on top
of the Fam-head.
24	 Note that the doubling of the root-final consonant is of no phonological relevance here.
52
(7) a. Tom-type hypocoristic b. Tommy-type hypocoristic
FamP =˃ truncation
  | |
Fam XP =˃ root
    | |
X NP
ClsP =˃ suffix /-i/
   | |
Cls FamP =˃ truncation
     | |
Fam XP =˃ root
      | |
X NP
The double arrows in the figures in (7) represent spell-out. Both the familiarityexpressing
form Tom and the closeness-expressing Tommy involve truncation of
the full root to a single syllable. This, then, indicates that the truncation is triggered
during the spell-out of the FamP, the projection of the “familiarity” feature.
The arrow is placed at the phrasal node (rather than at the terminal), which indicates
that the truncation applies to the material contained in this phrasal node
(i.e., Thomas); I come back to this in some detail later. The “closeness” feature,
present in the hypocoristic Tommy, is spelled out by the suffix /-i/, then.
Summing up, the proposal in (7) says that both of the formal processes involved
in derivation of hypocoristics in English, i.e., truncation and suffixation,
have morphemic status—each spells out a particular part of the morphosyntactic
structure.
Let us now take a closer look at the spell-out mechanism, only roughly
sketched in (7). Technically speaking, spell-out is nothing else than insertion of
an appropriate lexical material into the syntactic structure. I use the term “appropriate”
in the sense of the theory of Nanosyntax: a given lexical item can spell
out the syntactic structure equal to the size of that item—or smaller than it; the
procedure that “matches” the sizes of the syntactic structure stored in lexical
items and those generated by the grammar is known as the Superset Principle
(Caha 2009, Starke 2009, 2014). In other words, lexical items realize syntactic
constituents of various sizes provided that the syntactic tree is contained inside
the lexical item. From this perspective, the particular double arrows in (7) indicate
spell-out by the following three lexical entries.
(8) a. </tɒmǝs/, [XP X ...[NP N]]>
b. </CVC/, [FamP Fam [XP X ...[NP N]]]>
c. </-i/, [ClsP Cls]>
The lexical entry for the first-name root /tɒmǝs/ in (8a) is an ordered pair that
puts together the relevant phonology with the sequence of projections that start
53
from the nominal root projection, i.e., the NP, and end in the XP; recall that XP
stands for a sequence of projections included in the first-name syntactic structure.
(According to the Superset Principle, the form /tɒmǝs/ can spell out either
the NP alone, or the entire structure comprised by the NP and the XP—and, also,
any syntactic subconstituent of the XP.) (8c) is a lexical entry for the suffix /-i/
that marks the closeness feature; this morpheme is added to the truncated root.
The root will have to extract to the left from within the ClsP, movement that is
ignored in (7b). After the movement, ClsP contains only the feature ClsP, and
so there is a perfect match between that structure and (8c).
The entry in (8b) differs phonologically from the two others. It represents a
morpheme which is responsible for the truncation of the bisyllabic form /tɒmǝs/
to the monosyllabic form /tom/. This “truncating” morpheme is lexically associated
with the syntactic constituent of the FamP (and all the other projections
dominated by it) which actually means that the truncation is applied to the form
which spells out a sister of the Fam-node. On the phonological side, the entry
in (8b) contains an empty CVC-string. This representation expresses the fact
that the hypocoristic truncation is an output-oriented phonological process: hypocoristic
roots fit the CVC-template. Such “materialization” of morphologically
triggered phonological processes is one of the main contributions of autosegmental
phonology since McCarthy’s (1979) dissertation on Semitic templates.
In this particular case, the truncation is “materialized” as an empty prosodic
template which expresses the familiarity meaning. Furthermore, the lexical entry
in (8b) is in accordance with the hypothesis that morphological units relevant to
templatic domains correspond to particular constituents of the morphosyntactic
structure; see e.g. Hyman & Inkelas (1997) or Inkelas & Zoll (2005).
Let us now put the pieces in (7) and (8) together. Before doing it, we must emphasize
one important aspect of Nanosyntax: spell-out proceeds both cyclically
and from the bottom up. Cyclicity means that each syntactic Merge triggers a
search in the lexicon for an appropriate lexical item (in the sense of the Superset
Principle). If such an item is found, the particular phrasal node is spelled out,
i.e., it receives the phonological information encoded in the given lexical entry.
In the next step, Merge creates a one-feature bigger syntactic constituent and,
subsequently, the lexicon is consulted again. If an appropriate item is found (i.e.,
a lexical item whose syntactic structure includes the structure of this newly created
constituent), the (new) phonological material is inserted and this newly inserted
phonological material then overwrites the phonological material inserted
in the previous Merge-and-spell-out step.
54
To illustrate the cyclic nature of the spell-out, let us go back to the two possible
syntactic structures of hypocoristic in (7) and their corresponding lexical
entries in (8). After the Merge of the NP, the spell-out is triggered and /tɒmǝs/
is inserted, as depicted in (9a).25
In the next step, the XP matching the syntactic
structure in the entry in (8a) is created by Merge; the form /tɒmǝs/ is thus reinserted,
as in (9b) (technically speaking, the form /tɒmǝs/ inserted in the NP is
overwritten by the form /tɒmǝs/ inserted during the spell-out of the XP).
(9) a. NP =˃ /tɒmǝs/
   | |
N ...
b. XP =˃ /tɒmǝs/
    | |
X NP /tɒmǝs/
Once the XP has been built in (9b), FamP is built in the next step. Since the lexicon
contains the entry (8b) and since the syntactic structure of that entry matches
the structure of the FamP, the previous spell-out is overwritten: the form /tɒmǝs/
inserted in the preceding step is replaced by the relevant form of the lexical entry
in (8b). But, once it is replaced, we are facing a rather strange result: the familiarity
expressing hypocoristic of the first name Thomas is spelled out as a CVC-template,
which is, moreover, empty of any melodic content. This scenario is depicted in (10).
(10) FamP =˃ /CVC/
   | |
Fam XP /tɒmǝs/
     | |
X NP
Taking the “overwriting” proposal seriously, the lexical entries in (8a) and (8b)
apparently derive a non-existing alternation between [tɒmǝs]—the full form of
the first name—and phonetically “nothing” which, however, should correspond
to the hypocoristic form of the very name [tɒmǝs]. This is, however, a wrong
result, so we better think again about the representations both on the phonological
and the syntactic side.
25	 Generally speaking, any nominal root can be inserted into the NP. The idea is that an
insertion of a particular root depends fully on the speaker’s choice. However, once the
choice is made, the root—or, to be more precise, its semantic and syntactic features—is
kept through the derivation until another root node is merged.
55
3.3.1  The Templatic Morpheme: Rewriting of Prosodic Structure
On the phonological side, the contrast between the root morpheme and the templatic
morpheme is that the former is a fully-fledged phonological object, but the
latter is phonologically incomplete. In Strict CV, a phonological model used in
this book, fully-fledged phonological representations include both prosodic and
melodic structure and they are interconnected via association lines. The phonological
part of the lexical entry for the root /tɒmǝs/ thus looks as in (11a). The
figure in (11b) shows the representation of the templatic morpheme.
(11) a. // C V C V C // b. // C V C //
| | | | |
● ● ● ● ●
| | | | |
t ɒ m ǝ s
From the perspective of the representations above, the spell-out of the FamP
involves what we might call a “partial overwriting”: only a particular part of the
phonological representation of the earlier spell-out is overwritten. The proposal
is given in (12). The insertion of the templatic morpheme triggers deletion of the
prosodic structure of the root, which is then replaced by the template prosodic
structure. Since the root prosodic structure (i.e., CVCVC) is bigger than the template
prosodic structure (CVC), two of the root’s segments remain unassociated
prosodically, which actually means that they do not receive any phonetic interpretation:
hence, the form /tɒmǝs/ is truncated to the /tom/ as a consequence.
(12) C V C template
↓ ↓ ↓
C V C V C root
| | | | |
● ● ● ● ●
| | | | |
t o m a s
Note, that the matching procedure between the template and the root is only
sketched here rather roughly—and so are also the lexical representations in (11).
First of all, the melodic part of the lexical entry in (11a) should be further decomposed
into hierarchically organized elements (as we proposed in chapter 2).
56
Second, to ensure that the root /tɒmǝs/ is truncated to /tɒm/ rather than to /mǝs/
(or even /tɒs/ or /tǝs/), the information that the CVC-template is aligned with
the left edge of the root must also be involved in the spell-out.26
Note at this point that the overwriting mechanism actually provides an answer
to the question why the morpheme in (11b) is not simply affixed to the
root. Its phonological incompleteness, i.e., a fact that it does not contain any melodic
features, cannot be the reason: as we already saw, affixes which are empty
prosodic strings also exist. In other words, we need to tell apart two types of
prosodic morphemes: prosodic affixes whose insertion leads to the extension of
the prosodic structure of adjacent morphemes and prosodic templates whose
insertion triggers its overwriting. Since morphemes are pairings of phonological
and syntactic representations, the contrast between the two types of prosodic
morphemes can logically be encoded either in their phonological part or in their
syntactic part.
Up to now, we have been looking at the picture in (12) from above, i.e., from
the perspective of the template which overwrites root’s prosodic structure. Taken
from the opposite perspective, what we see in (12) is that the root provides the
templatic morpheme with melodic features and, in this way, it makes it phonetically
interpretable. More generally, (12) depicts a situation when one lexical item
feeds phonologically the other.
In the Nanosyntactic literature, such “feeding” among lexical items has already
been identified in idioms (Starke 2009, 2014). For example, an idiom for
‘die’ kick the bucket is made up from three lexical items: a verb /kɪk/, a determiner
/ðə/, and a noun /bʌkɪt/. To capture a fact that these forms exist independently
of the idiom, Starke (2009, 2014) proposes that idiom’s lexical entry does not
contain any phonological information and it acquires its form from other lexical
26	 The left-to-right association algorithm between segments and prosodic positions provided
by the template is proposed to be universal (Goldsmith 1976). It can explain the
truncation of /tɒmǝs/ to /tɒm/, but it cannot explain that in other cases, it is either
final (Isaac → Zac) or middle parts (Elisabeth → Liz) of the first-name roots that are
retained in the truncated forms; see Lappe (2003) for a detailed overview. In an Optimality
Theory framework, such a variation is derived by a set of violable constraints
(McCarthy & Prince 1999). However, we face the problem of how to account for this
fact within the framework, in which prosodic templates are assumed to be lexical items
similar to “ordinary” roots or affixes, in other words, we want to propose a sustainable
theory of pairing phonological and syntactic representations. A template allomorphy
seems like the only tenable answer at this point. I leave this issue open for further research,
but see Faust & Torres-Tamarit (2017) for combining Strict CV representations
and Optimality Theory violable constraints.
57
entries. Technically speaking, the lexical entry for the idiom kick the bucket includes
so-called pointers which occupy particular syntactic nodes of a lexical entry
and refer to other lexical entries: concretely, there is a pointer in the VP-node
that refers to the lexical entry for the verb /kɪk/, the pointers in the DP-node
and the NP-node “point” to the entries for the determiner /ðə/ and the noun
/bʌkɪt/, respectively. These three entries thus provide idiom’s entry with appropriate
forms (though their semantics is overwritten).
From this perspective, the templatic morpheme discussed above (recall, it
is the one responsible for truncation in the FamP-node in hypocoristics), has
certain aspects in common with multiword idioms of the kick the bucket type:
the CVC-template points to the first-name root which provides it with melodic
features. Despite this similarity with idioms, the templatic morpheme is special
in two ways.
First, it is not completely phonologically empty. It contains a piece of the prosodic
structure which triggers what we called partial overwriting of the phonological
structure of the pointed item. Second, there is the truncation of the first
name and the first names can be of various types: in general, truncated hypocoristic
can be derived from virtually any first name, that is, first names of various
morphological properties; for example, Tom is a truncated masculine name,
while Liz is a truncated feminine. Moreover, the same type of truncation, i.e.,
truncation expressing speaker’s familiarity with the denotee, is applied productively
also to other nouns than to the first names; cf. e.g. brother > bro or moment
> mo. To describe this observation properly, we have to say that pointers can
“point” not only to particular lexical entries (as in the case of a multiword idiom
kick the bucket), but also to their sets. In this particular case, the FamP-node—a
part of the template syntactic tree—points to a set of structurally defined items,
or, to be more precise, to a set of subtrees shared among more lexical items. I
leave an exact implementation of this idea for further research, but let me note,
that variables in lexical entries might be used.
Summing up our proposal: hypocoristic truncation is triggered by the spellout
of a particular node of the syntactic tree shared between both types of hypocoristics,
i.e., familiarity-expressing ones as well as the ones expressing closeness.
Truncating morpheme is lexically represented as a prosodic template which
points to other lexical items. It is the pointer what makes difference between
prosodic templates on the one hand and prosodic affixes on the other. Phonologically,
they are both represented as empty pieces of prosodic structure, but by
pointing to other forms, the hypocoristic template causes overwriting of their
prosodic structure.
58
3.3.2  The Morphosyntactic Hierarchy: Closeness over Familiarity
Now, let us look at one-projection bigger hypocoristics, those expressing closeness:
there is going to be an extra spell-out step in their derivation. Recall that
closeness-expressing hypocoristics of the Tommy-type are derived from the
truncated roots by the /i/-suffix. Both pieces relevant to their spell-out are repeated
below: (13a) repeats the lexical entry for the closeness-marking suffix,
(13b) reproduces the syntactic structure of a closeness-expressing hypocoristic.
After ClsP has been created in syntax, the lexicon is searched through to find
an appropriate lexical entry. At the first sight, the lexical entry in (13a) seems
to be a relevant candidate for insertion—it contains the ClsP projection after
all. However, the ClsP in the tree in (13b) contains other subconstituents, i.e.,
the FamP and the other projections dominated by it—and these are not present
in the lexical entry of the suffix /-i/ in (13a). In other words, the entry in (13a)
does not match the structure of the closeness phrase in (13b). Should the ClsP
be—indeed—spelled out by the suffix /-i/, its structure must match the syntactic
structure in (13b)—and the way to achieve it is to move its complement, i.e.,
the FamP away. This, then, leads to the so-called spell-out-driven-movement depicted
in (13c): the FamP moves to the left of the ClsP and, as a consequence,
the ClsP matches the syntactic tree of the lexical entry of the suffix /-i/ in (13a).
Thus: the ClsP is spelled out as /-i/.27
27	 Starke (to appear) defines spell-out-driven movement as a procedure that is not strictly
speaking driven by a particular shape of lexical entries. Rather, it is defined as a series
of ordered operations that happen automatically every time there is no matching entry.
The definitions are as follows:
	 (i)  Insert feature and spell out.
	 (ii) If fail, try a cyclic (spec-to-spec) movement of the node inserted at the previous
cycle and spell out.
	 (iii) If fail, try a snowball movement of the complement of the newly inserted feature
and spell out.
	 I will ignore here the mechanical steps required by the procedure, and show instead
directly how the structure is transformed so that it matches the lexical entries as proposed.
For instance, in (13c), it is only the movement type (iii) what delivers the correct
configuration, but I will not discuss here the failed attempts (as this is not my primary
interest here) and jump directly to the option that works.
59
(13) a. </-i/, [ClsP
Cls]> b. ClsP
      | |
Cls FamP
     | |
Fam XP
      | |
X NP
c. ClsP
     | |
FamP ClsP =˃ /-i/
    | |      |
Fam XP Cls
     | |
X NP
In this section, we have established that hypocoristics in English fall into two
main types: monosyllabic ones and bisyllabic ones derived by the /i/-suffix. Both
types are based on truncated roots and I propose that their formal similarity
stems from the shared syntactic structure. This idea was implemented using
the Nanosyntactic framework: I argue that truncation is triggered by the templatic
morpheme that spells out the FamP, a syntactic projection shared among
all types of hypocoristics. The /i/-suffixed hypocoristics are syntactically bigger:
the suffix spells out the ClsP and the ClsP projection is the consequence of the
merge of Cls-feature with the FamP. This analysis is in line with the claims by
Schneider (2003) and Alber & Arndt-Lappe (2012): they notice that hypocoristic
meaning has two components, namely the familiarity component and the
closeness component. I, in addition, propose that these two components are in
a hierarchical relation and I added the way to derive them systematically. From
this perspective, let us now examine phonological and morphosyntactic properties
of hypocoristics in Czech.
3.4  Closeness-Expressing Hypocoristics
As in English, the derivation of hypocoristics in Czech involves (as an ingredient)
the truncation of polysyllabic roots to a single syllable. Table (14) shows
(again) the data from section 3.1; they illustrate that truncated roots are the base
for both suffix-less as well as suffixed hypocoristics. (Recall, that the final a is an
inflectional ending which is marked by enclosing it in brackets.) As in English,
I will analyze the templatic morpheme (responsible for the truncation) as the
spell-out of the FamP, which is shared across all types of hypocoristics, both in
English and in Czech.
60
(14) full name suffix-less form d-suffixed form
M[i]loslav M[iː]l(-a) M[i]l-d(-a)
J[u]lius J[uː]l(-a) J[u]l-d(-a)
Př[e]mysl Př[eː]m(-a) Př[e]m-d(-a)
Ant[o]nín T[oː]ň(-a) T[o]n-d(-a)
J[a]roslav J[aː]r(-a) J[a]r-d(-a)
In Czech, the truncated roots further undergo lengthening and suffixation. As
the examples in (14) show, these two processes are in complementary distribution;
notice, however, that it is the very same (shortened) root that is either
lengthened, or receives the suffix -d. I take this complementarity of lengthening
and suffixation—and the fact that the root is always the same, regardless of the
type of the morphological process—to indicate that lengthening and suffixation
are not allomorphs, but rather realizations of different morphosyntactic structures.
What is then the difference between the two types?
3.4.1  The Gender Pattern
As is illustrated in (15), lengthened suffix-less hypocoristics show different morphosyntactic
behavior than the d-suffixed ones. The lengthened hypocoristics
(in the leftmost column) are ambiguous between masculine (M) and feminine
(F) gender; in this, they pattern with the corresponding full-names (in the middle
column), which are gender-ambiguous as well. By contrast, the d-suffixed
forms (in the rightmost column) are gender-specific: these are only masculine.
(15) lengthened form full name suffixed form
M[iː]l(-a) (M, F)
M[i]loslav (M) M[i]l-d(-a) (M)
M[i]loslav(-a) (F)
J[aː]r(-a) (M, F)
J[a]roslav (M) J[a]r-d(-a) (M)
J[a]roslav(-a) (F)
D[aː]ň(-a) (M, F)
D[a]niel (M) D[a]n-d(-a) (M)
D[a]niel(-a) (F)
Notice that the only difference between the gender-ambiguous full names and
the gender-ambiguous lengthened forms is their declension pattern, as illustrated
in (16). (I.e., there is no apparent derivational suffix in either the masculine
or the feminine form.) In the left-hand part of the table, the singular case markers
for full names follow the typical masculine and feminine case paradigms.
61
Hypocoristic paradigms, on the other hand, as shown in the right-hand part of
the table, are mostly syncretic for both genders (though notice the exceptions in
the dative and locative cases). Stating the obvious: it seems that the root itself is
genuinely ambiguous between masculine and feminine gender, and the suffixes
reflect this ambiguity both in the full form and in the hypocoristic form.
(16) full name (M) full name (F) hypocoristic (M) hypocoristic (F)
N Miloslav-Ø Miloslav-a Míl-a Míl-a
A Miloslav-a Miloslav-u Míl-u Míl-u
G Miloslav-a Miloslav-y Míl-y Míl-y
D/L Miloslav-ov-i Miloslav-ě Míl-ov-i Míl-e
I Miloslav-em Miloslav-ou Míl-ou Míl-ou
To deal with the gender specific denotation of the d-suffixed form described above,
we might simply assume that the first-name roots in (15) are not lexically specified
for any gender features, while the suffix -d is specified for spelling out the masculine
feature. Being gender-less, the roots can thus appear in both masculine and feminine
nouns, depending on their syntactic context. Simply speaking, the root can be
syntactically merged with a masculine feature yielding a masculine name or with
a feminine feature yielding a feminine name. This scenario would be consistent
with the hypothesis that in the lexicon, roots are totally underspecified for syntactic
features, which is one of the core ideas of Distributed Morphology (Harley 2014).
A problem is, however, that there exist first names that are actually gender-specific:
names in (17a) are only masculine and those in (17b) are only feminine. If we adopt
the idea that all roots lack any syntactic features, including the gender ones, the
three-way gender behavior of first names seen in Czech cannot be explained in a
systematic way. In other words, if the first-name roots /miroslav/, /mojmiːr/ or /barbor/
are lexically equal, as is assumed in the Distributed Morphology framework,
we cannot explain why only the first one derives names of both genders, while the
latter two can be involved only in masculine and feminine names, respectively.
(17) full name d-suffixed hypocoristic
a. Přemysl (M) Přem-d(-a)
Cyril (M) Cyr-d(-a)
Mojmír (M) Moj-d(-a)
b. Monik(-a) (F) *Mon-d(-a)
Barbor(-a) (F) *Bar-d(-a)
Nel(-a) (F) *Nel-d(-a)
62
Furthermore, the gender-less approach to nominal roots cannot capture even the
behavior of the suffix -d. As the data in tables (15) and (17) demonstrate, the suffix
-d concatenates with either gender-ambiguous or masculine names, but not
with feminine names. Again, if all roots are unspecified for the gender context
(Distributed Morphology’s crucial assumption), d-suffixed forms derived from
feminine names should not be ruled out. However, as shown in (18), the nonexistence
of hypocoristics like *Mon-d(-a) and the others in (18b) is far from
being accidental.
Instead, hypocoristic feminines are derived by a suffix -č [ʧ]: it combines with
either feminine names (18b) or gender-ambiguous names (18c), but not with
those that are masculine (18a). (Note as well that when the -č attaches to such
ambiguous names, the resulting hypocoristic refers to females only.)
(18) full name d-suffixed hypocoristic (M) č-suffixed hypocoristic (F)
a. Přemysl (M) Přem-d(-a) *Přem-č(-a)
Cyril (M) Cyr-d(-a) *Cyr-č(-a)
Mojmír (M) Moj-d(-a) *Moj-č(-a)
b. Monik(-a) (F) *Mon-d(-a) Mon-č(-a)
Barbor(-a) (F) *Bar-d(-a) Bar-č(-a)
Nel(-a) (F) *Nel-d(-a) Nel-č(-a)
c. Miloslav(-ØM
/-aF
) Mil-d(-a) Mil-č(-a)
Jaroslav(-ØM
/-aF
) Jar-d(-a) Jar-č(-a)
Daniel(-ØM
/-aF
) Dan-d(-a) Dan-č(-a)
To complete the pattern, I must add that gender-specific names produce also
gender-specific lengthened hypocoristics which show no overt suffix. A couple
of relevant examples is given in the table below.
(19) full name lengthened form
a. Př[e]mysl (M) Př[eː]m(-a)
C[i]ril (M) C[iː]r(-a)
b. M[o]nik(-a) (F) M[oː]ň(-a)
B[a]rbor(-a) (F) B[aː]r(-a)
To sum up the gender pattern: lengthened hypocoristics without an overt suffix
can be either gender-ambiguous (M[iː]l(-a)M/F
) or gender-specific (C[iː]r(-a)M
,
M[oː]ň(-a)F
), depending on the gender features of their corresponding full
63
names. Hypocoristic forms derived by the suffixes, on the other hand, are only
gender-specific: d-suffixed forms are masculine, č-suffixed ones are feminine.
The gender of the hypocoristic suffixes matches the gender of names they concatenate
with: the masculine suffix -d concatenates with masculines (C[i]rilM
>
C[i]r-d(-a)M
), the feminine -č attaches to feminines (M[o]nikaF
> M[o]n-č(-a)F
),
and not vice versa. If the first name is gender-ambiguous, both of the suffixes can
be attached yielding either a d-suffixed masculine (M[i]l-d(-a)M
) or a č-suffixed
feminine (M[i]l-č(-a)F
).
In what follows I argue, that the described pattern can be explained only under
the assumption that both root morphemes and suffixes are specified for gender
features already in the lexicon (which does not prevent gender from being a
separate projection in syntax). This assumption is consistent with the Nanosyntactic
approach to spell-out which, as we already saw, involves matching of syntactic
trees stored in lexical entries with syntactic trees generated by the syntax.
In particular, I propose that in syntax, there is a part of the functional sequence
which is dedicated to gender. The specific gender subtree I am proposing here is
[FemP
Fem [MascP
Masc]], in which FemP dominates MascP. The syntactic structure
of masculine nouns contains only the lower gender projection MascP, while
feminines are one-projection bigger—they involve the whole gender subtree.
The different behavior of particular roots—some of them occur simultaneously
in both masculine and feminine names, some of them appear only in masculines
or only in feminines—arises from the fact that the roots are lexically stored with
gender subtrees of different sizes.28
And the similar proposal will be made for the
hypocoristic suffixes -d and -č.
3.4.2  Hypocoristics as Syntactic Compounds
In a hypocoristic form C[i]r-d(-a), the masculine suffix -d is attached to a truncated
masculine noun; recall that the first name C[i]ril does not have a feminine
counterpart C[i]ril(-a). From this, it inevitably follows that the masculine feature
is present twice in the structure of the noun C[i]r-d(-a) (once in the root, and
once in the suffix).
A similar reasoning leads to the conclusion that also other features are present
twice in the structure. Consider, for instance, the fact that the suffix -d appears
28	 Note that also some versions of Distributed Morphology try to accommodate that
gender features are a prototypical root-dependent property of nouns. For example,
Acquaviva (2009) or Kramer (2015) propose that gender is encoded in a categorial
head n which merges directly with a root.
64
specifically in hypocoristics and only in them: there are no other morphological
types of masculine nouns derived by -d. This suffix thus spells out the hypocoristic
meaning which, I proposed, is encoded in two syntactic projections: a lower
FamP and a higher ClsP. However, the same two projections must also be spelled
out by the root. To see that, recall the proposal that the familiarity-encoding
projection (FamP) is spelled out by a templatic morpheme which is responsible
for the truncation of polysyllabic roots. Since C[i]r-d(-a) and the other d-suffixed
forms in (18) are built on truncated roots, we can conclude that not only the gender
feature, but also the familiarity feature is placed twice in these hypocoristic
nouns.
To account for this systematic “feature doubling”, I propose that d-suffixed
hypocoristics are complex structures made up from two fully-fledged nominal
trees that are joined together in a structure that is similar to compounding. Having
two fully-fledged nominal projections is a logical consequence of the cartographic
approach to the grammar where the functional sequence, i.e., the set
of hierarchically organized syntactic features, cannot be recursive—not even in
principle, if we assume that X cannot dominate X, see e.g. Starke (2004).
The binominal proposal is depicted in (20). Here, the suffix -d spells out the
fully-fledged nominal tree with a NP phrase at the bottom; higher up, there is a
masculine gender phrase (MascP) dominating the NP and these two are dominated
by the hypocoristic projections. The d-suffix tree is adjoined to the truncated
root yielding a construction dominated by the higher hypocoristic projection,
i.e., ClsP. As a consequence, there are two ClsP-nodes in the structure from
which only the lower one receives spell-out (it is realized by the suffix). The reason
is that these two ClsP are created by different means: the lower ClsP arises
from featural Merge, when a closeness feature is merged with FamP. By contrast,
the higher ClsP in (20) results from phrasal Merge which does not trigger cyclic
spell-out. To sum up: the structure in (20) captures the observation that the
d-suffixed forms like C[i]r-d(-a) are masculine hypocoristics derived from truncated
masculine nouns.
(20) ClsP
         | |template <= FamP ClsP =˃ /-d/
     | |             | |
Fam MascP =˃ root Cls FamP
      | |           | |
Masc NP Fam MascP
                  | |
Masc NP
65
The only difference between the two nouns in (20) is their “morphological” status:
while the left-hand noun is realized by a root morpheme, the right-hand one
by a morpheme that would traditionally be classified as an affix. The proposal
that affixes can realize fully-fledged nominal structures (which include a root
projection at the bottom) is not completely new—it has already appeared in the
literature; see e.g. Bachrach & Wagner (2007), De Belder (2011) or Lowenstamm
(2014) who have proposed something similar within a Distributed Morphology
framework. In principle, both separate noun projections in (20) can be realized
by (what a traditional morphologist would analyze as) a root morpheme, so that
we obtain so-called coordinative compounds; these (two-rooted) hypocoristic
compounds are not found in Czech, but they appear in, for example, English, e.g.
Billy Boy, Eddie Baby or pussycat. The Czech forms receive here an analysis that
is analogous in terms of the syntactic structure attributed to them, but differs in
that the second noun is phonologically reduced and patterns with affixes in this
regard. (To rule out the third logical possibility (not encounter in any language,
as far as I know), i.e., that both nouns are realized by affixes, an additional spellout
filtering mechanism must be invoked.)
To conclude this section, let me summarize the lexical entries for the three
morphemes involved in the derivation of the masculine hypocoristic C[i]r-d(-a).
The entry for the root /ciril/ in (21a) contains the masculine-gender projection.
Such a representation explains why this root produces only a masculine
first name C[i]ril and not a feminine one *C[i]ril(-a). The entry for the prosodic
template, which is responsible for the truncation of /ciril/ to /cir/, is in (21b). As
I proposed in section 3.3.1, it spells out the familiarity meaning. This prosodic
morpheme contains a pointer, marked as , which refers to a set of structurally
defined nouns whose prosodic structure is overwritten by the prosodic template.
The entry for the masculine suffix -d is represented in (21c): the suffix is lexically
associated with the fully-fledged nominal tree which comprises the masculine
gender feature and both hypocoristic projections. Simply speaking, the
representation in (21c) says that the d-suffixed forms are closeness-expressing
masculines.
66
(21) a. </ciril/, [MascP
Masc [NP
N]]>
b. </CVC/, [FamP
Fam [XP
X ...[NP
N]]]>
c. </-d/, [ClsP
Cls [FamP
Fam [MascP
Masc [NP
N]]]]>
3.4.3  Lengthened Forms
Truncated masculine names undergo either d-suffixation or lengthening: it holds
for the name C[i]ril which has two hypocoristic forms C[i]r-d(-a) and C[iː]r(-a).
The fact that these two processes are complementary to each other might lead
us to consider them to be allomorphs. In that case, the d-suffixed hypocoristic
C[i]r-d(-a) and the lengthened hypocoristic C[iː]r(-a) would be structurally
equal and both hypocoristic affixes, i.e., the prosodic VC-affix which triggers
lengthening of the root vowel and the segmental suffix /-d/, would be lexically
specified for spelling out the same syntactic tree.
This proposal is, however, unable to explain the observation made above
that the d-suffixed forms are always masculine, while the lengthened forms can
be potentially both masculine and feminine (simply copying the gender of the
base). This latter fact indicates that (unlike the d-suffix) the prosodic affix itself
does not carry any gender features. To express this gender pattern, I thus propose
that the closeness-expressing hypocoristics are of two different syntactic types.
The d-suffixed forms are binominal structures with structures as given in (20),
where the gender is encoded twice: both in the root (which spells out the lefthand
nominal tree) and in the suffix (realizing the right-hand noun). A crucial
point is that since this is a coordinative compound, where a given individual satisfies
the denotation of each part, these two genders have to match: the suffix -d
derives masculine hypocoristics only from masculine names. In the lengthened
forms in (22a), on the other hand, there is only one gender-encoding subtree.
(22)
a.
ClsP
     | |
Cls FamP
       | |
Fam MascP
         | |
Masc NP
b. ClsP
         | |template <= FamP ClsP =˃ length
       | |        |
Fam MascP Cls
           | |
Masc NP
The closeness-hypocoristic structure in (22a) is the same as that proposed for
the /i/-ending hypocoristics in English: the FamP is spelled out by the templatic
morpheme in both languages. The only difference is that the closeness meaning
(encoded syntactically in ClsP) is realized by a segmental suffix in English, but
67
in Czech it is spelled out by a prosodic affix which is responsible for lengthening
of the root vowel. In both cases, however, the spell-out-driven movement of the
FamP to the left must be postulated, which is depicted in (22b). This enables the
ClsP to be realized either segmentally or prosodically, depending on the available
lexical items: the first option is instantiated in English because its lexicon
contains the entry </-i/, [CLSP
Cls]>. The Czech lexicon stores the prosodic entry
</VC/, [CLSP
Cls]>, hence the closeness feature is realized as vowel length in this
language.
The three morphemes involved in the derivation of the closeness-expressing
hypocoristic C[i:]r(-a) are summarized in (23); two of them are identical to the
morphemes participating in the derivation of the d-suffixed form C[i]r-d(-a)
discussed in the previous section.
(23) a. </ciril/, [MascP
Masc [NP
N]]>
b. </CVC/, [FamP
Fam [XP
X ...[NP
N]]]>
c. </VC/, [ClsP
Cls]>
Taken from the phonological perspective, both hypocoristic morphemes in
(23b) and (23c) differ substantially from the root morpheme in (23a): only the
latter is a fully-fledged phonological object, which is lexically represented on
both a prosodic and a melodic level. The two hypocoristic morphemes are prosodic
affixes which, however, have different phonological impact on the root: the
templatic morpheme overwrites root’s prosodic structure (so the root fits the
CVC-shape), while the prosodic affix expands the prosodic space of the root (so
the root-vowel spreads its melody). As I have already proposed, this difference
follows from presence of the pointer in the templatic-morpheme lexical entry in
(23b): by pointing to other lexical items, the CVC-template overwrites their prosodic
structure. But the prosodic affix in (23c) must also be stored together with
additional phonological information. In particular, the entry for the prosodic
affix must contain information about the specific point of insertion in the CVC
tier, that is, the fact that it is inserted next to the root vowel—and, say, not suffixed
to the root-final consonant. The insertion to the left of the vowel, when the
vowel serves as a melody self-licensor, follows automatically from the prosodic
morpheme’s VC-shape.29
29	 There is yet another phonological process involved in hypocoristic formation: palatalization.
In particular, alveolars [d t n] turn to palatals [ɟ c ɲ] (R[ad]oslav > R[aːɟ](-a), P[et]r
> P[eːc](-a), M[on]ik(-a) > M[oːɲ](-a)) and [x h] change to [ʃ ʒ] (R[ix]ard > R[iːʃ](-a),
B[oh]dan > B[oːʒ](-a)). Kochetov & Alderete (2011) analyze the palatalization patterns
68
The phonological effects of these prosodic morphemes are illustrated in (24)
below. The figure in (24a) depicts truncation of a bi-syllabic root /ciril/ to /cir/
induced by the template (marked by shadowing). (24b), then, illustrates lengthening
of this truncated form to /ciːr/ triggered by the empty VC-affix being inserted
into the phonological structure of the truncated root.
(24) a. C V C b.
↓ ↓ ↓ L        
|
C V C V C C V C V C
| | | | | | | |
● ● ● ● ● ● ● ●
| | | | | | | |
c i r i l c i r
3.4.4  Decomposing Gender: Gender-Ambiguous Names
Let us now turn to closeness-expressing hypocoristics of the M[iː]l(-a) type
that are ambiguous between a masculine and a feminine. I follow the idea that
phi-features are decomposable into privative units (e.g. Harley & Ritter 2002)
and propose that masculine and feminine gender features are encoded as two
in hypocoristics from the cross-linguistic perspective and argue that this so-called
“expressive palatalization” is a “phonologically unmotivated process”, i.e., a process
without (a clear) phonological trigger. Czech data, however, go against this assumption.
If palatalization were associated with hypocoristics as an “expressive” morphological
category (as proposed by Kochetov & Alderete 2011), then we might expect that it
would be applied across-the-board. In reality, however, only root-final consonants—
and not consonants involved in hypocoristic suffixes—are targets of palatalization: in
hypocoristics like [aːɟ](-a), R[aːɟ](-a) and B[eːɟ](-a), that are derived from [ad]am,
R[ad]oslav and B[ed]řich, [d] palatalizes to [ɟ]; in suffixed forms like Ton-[d](-a), on the
other hand, there is no palatalization. Thus, it seems that palatalization in hypocoristics
goes hand in hand with lengthening (but not with segmental suffixation). Strikingly,
if the root is lexically monosyllabic, these two processes (lengthening and palatalization)
are the only hypocoristic markers; see e.g. the full name-hypocoristic pairs like
D[it](-a) – D[iːc](-a), H[an](-a) – H[aːɲ](-a) or J[an](-a) – J[aːɲ](-a). To account for
this, we have to assume that the lexical entry for the lengthening affix includes also a
palatal feature that triggers palatalization. Derivation of the palatalization patterns in
Czech (even in hypocoristics) is beyond the scope of this book; let us only notice that
the palatalization trigger might be represented as a lexically floating I-element, which
docks in the root-final consonant.
69
separate syntactic heads: a Masc-head and a Fem-head. Furthermore, there is
a cross-linguistic observation: in many languages, feminine animate nouns are
built formally on masculine ones. This holds also for Czech: female-denoting
nouns are productively derived by attaching a suffix to masculine nouns, as illustrated
by gender pairs like žák – žák-yn(-ě) ‘pupilMASC
– pupilFEM
’ or soused
– soused-k(-a) ‘neighborMASC
– neighborFEM
’. Following the same logic as in the
case of truncated roots that are shared among hypocoristics of various types, I
take the containment of the gender forms mentioned above to indicate a gender
hierarchy: Fem>Masc.
From the perspective of this hierarchy, roots that produce gender-ambiguous
names are those that are lexically specified for the whole gender subtree. This
proposal is shown in (25). Specifically, the table (25) compares two lexical entries
for two roots that show different gender behavior. The entry for the root /ciril/ in
(25a) contains only a lower gender projection, i.e., MascP, which expresses the
fact that it can only be used as a masculine name (recall that there is no feminine
first name *Cyril(-a)). The root /miloslav/, on the other hand, appears in both
masculine (Miloslav) and feminine names (Miloslav(-a)): this is the reason why
its lexical representation in (25b) contains both gender projections.
(25) a. </ciril/, [MascP
Masc [NP
N]]>
b. </miloslav/, [FemP
Fem [MascP
Masc [NP
N]]]>
Given the Superset Principle, the entry in (25b) can spell out both a feminine
noun (syntactic structure of which matches the lexical tree exactly) and a masculine
noun (whose syntactic structure lacks FemP, so it matches the lexical subconstituent
[MascP
Masc [NP
N]]). This is illustrated in (26) which depicts syntactic
structures of a masculine (26a) and a feminine (26b) version of the closenessexpressing
hypocoristic M[iː]l(-a).
(26)
a.
ClsP =˃ length
    | |
Cls FamP =˃ template
        | |
Fam MascP =˃ /miloslav/
            | |
Masc NP
b. ClsP =˃ length
  | |
Cls FamP =˃ template
     | |
Fam FemP =˃ /miloslav/
       | |
Fem MascP
           | |
Masc NP
From the perspective of the analysis presented above, labelling full names
like M[i]loslav(-Ø/-a) and their lengthened hypocoristic forms M[iː]l(-a) as
70
gender-ambiguous is a bit misleading. Syntactically, there is no ambiguity: masculine
and feminine versions of these nouns have different syntactic structures.
In fact, the masculine M[iː]l(-a) and the feminine M[iː]l(-a) are two syncretic
hypocoristic forms. This, in turn, means that they are derived from a single
lexical representation, i.e., from a single root /miloslav/, which is stored with
the full gender subtree. In other words, gender-ambiguous can be roots in the
lexicon, but not syntactically derived nouns; these can be gender-syncretic.
Finally, note that this approach allows us to neatly distinguish two different
types of homophony. One type of homophony (presented above) arises due to
lexical relatedness (i.e., syncretism) where a gender-ambiguous base (spelling
out different sizes of structure) leads to ambiguous hypocoristics (derived from
bases of various sizes). In addition, there are also homonymous forms, which,
crucially, are derived from two separate lexical items and end up homophonous
because of how truncation works. For example, Áď(-a) is a lengthened hypocoristic
both for the masculine name Adam and for the feminine name Adél(-a).
3.4.5  Feminine Hypocoristics
In the previous two sections I proposed that the contrast between gender-syncretic
nouns (Miloslav(-Ø/-a)) and gender-specific masculines (Cyril) arises from
lexical representations of their roots. Roots deriving gender-syncretic nouns are
lexically stored with both gender features: this enable them to spell out either
the full gender subtree [FemP
Fem [MascP
Masc]] which syntactically identifies a
feminine noun, or just its proper constituent [MascP
Masc] (in masculine nouns).
If a root appears only in masculine nouns, it is because its lexical entry does
not contain the FemP. The same logic can be applied to the suffix -d: it lacks the
FemP, hence it can derive only masculine hypocoristics.
Let us now consider gender-specific feminines. Since feminine nouns are
syntactically defined as [FemP
Fem [MascP
Masc]], roots such as /monik/ in (27c),
producing exclusively feminines (recall that a feminine first name Monik(-a)
has no masculine counterpart *Monik) must be lexically specified for the whole
subtree.
(27) a. </ciril/, [MascP
Masc [NP
N]]>
b. </miloslav/, [FemP
Fem [MascP
Masc [NP
N]]]>
c. </monik/, [FemP
Fem [MascP
Masc [NP
N]]]>
Comparing the entries in (27b) and (27c) we can see that there is no lexical difference
between roots producing gender-syncretic names and those producing
71
only feminines: both roots /miloslav/ and /monik/ are specified for the full gender
subtree. Within Nanosyntax, the problem is technically that feminine roots
would be expected to “shrink” to express the masculine, but they never do. In
other words, while /miloslav/ can actually spell out either gender-feature structure
(Fem and Masc together—or just the lower Masc-feature), the form /monik/
always spells out both the gender features together. In other words, it behaves
like a strict portmanteau morpheme. Such “unshrinkable” lexical items have
already been identified in the Nanosyntactic literature, see Starke (2009:9) on
French indefinites and Dékány (2011:117) on prepositions in Hungarian. Note,
however, that the exact representation of “unshrinkability” of a structure of a
lexical item is still far from clear (and subject to further research).
However, there exist not only gender-specific feminine roots but also a feminine
hypocoristic suffix -č. Recall that it attaches either to feminine truncated
forms (M[o]nik(-a) > M[o]n-č(-a)) or gender-syncretic forms (M[i]loslav(-a) >
M[i]l-č(-a)), in which case the result is always a feminine hypocoristic. To explain
this distribution, I propose that č-suffixed hypocoristics (like d-suffixed
ones) are binominal structures (a coordinative compound), where the two nominal
projections match in gender, very much like in (20), but with the Fem-feature
in both parts of the compound. The right-hand extended NP is spelled out by
the suffix. Since this tree contains the Fem-feature, the suffix -č must be lexically
specified for it. Now recall that -č only derives feminine hypocoristics, so on
the face of it, it looks like we must simply draw a parallel between this genderspecific
suffix and the gender-specific root /monik/ and propose that both are
lexically stored as portmanteau morphemes associated with the full gender tree
[femp Fem [mascp Masc]]. However, in this case, a more attractive possibility
is to take the feminine suffix -č to be a feminine version of the masculine suffix
-d. This relationship can be expressed by using a pointer: in the entry for the
feminine suffix -č in (28b), the pointer directs to the entry for the masculine
suffix -d in (28a).
(28) a. </-d/, [ClsP
Cls [FamP
Fam [MascP
Masc [NP
N]]]]>
b. </-č/, [ClsP
Cls [FamP
Fam [Fem
Fem </-d/>]]]>
The representation in (28b) says that the suffix -č is just a suppletive form of the
suffix -d. This representation captures the intuition that these two suffixes differ
only in their gender. In simple terms, the suffix -č is just a feminine version of
the suffix -d. Crucially, this pair of lexical entries delivers also the fact that -č can
never derive masculine hypocoristics, for in that context, the suffix -d is a better
match. As a result, no problem with shrinking arises: the entry can shrink, but
72
when it does, its pronunciation is [d]. Whether this approach can be extended to
the case of “unshrinkable” first-name roots remains to be seen.
3.4.6  Interim Summary
Both lengthened and suffixed hypocoristics contain the full hypocoristic subtree
ClsP>FamP. Yet, they differ in how this subtree is integrated into the syntactic
structure (and this, in turn, impacts the spell-out). In the lengthened forms, the
hypocoristic subtree dominates the gender subtree and each node is spelled out
by a separate lexical item: the FamP is spelled out by a templatic morpheme
(which truncates polysyllabic roots) and the ClsP is realized by a prosodic VCaffix
(which triggers root-vowel lengthening). In the suffixed forms (which are
binominal structures morphosyntactically), the FamP is present twice: on the
one hand, it dominates the gender subtree of the full-name root, and, on the
other, it is a part of the full hypocoristic subtree. In the former case, it is spelled
out by a templatic morpheme (applied to the root), in the latter, it is realized
together with the dominating ClsP by the hypocoristic suffix (either -d or -č,
depending on the gender features). All the four lexical items relevant for hypocoristic
formation are summarized in (29).
(29) a. <CVC, [FamP
Fam [XP
X ...[NP
N]]]>
b. <VC, [ClsP
Cls]>
c. </-d/, [ClsP
Cls [FamP
Fam [MascP
Masc [NP
N]]]]>
d. </-č/, [ClsP
Cls [FamP
Fam [Fem
Fem </-d/>]]]>
3.5  Familiarity-Expressing Hypocoristics
So far, we have only seen cases where hypocoristics always have both hypocoristic
projections, i.e., FamP and ClsP. However, given the implicational hierarchy
ClsP>FamP, we should not be surprised to find hypocoristics without the (higher)
ClsP level; in that case, they should have the following structure.
(30) FamP
        | |
Fam (FemP)
           | |
(Fem) MascP
               | |
Masc NP
73
Given the structure in (30), what are the expected phonological and morphological
properties of such hypocoristics? First, we expect that the FamP is going
to be spelled out by the templatic morpheme, but no lengthening (and palatalization)
will occur, since these only apply at the spell-out of ClsP. In sum, we expect
that familiarity expressing hypocoristics (based on the bare FamP) should be
bare truncations, they should have short vowels, and they should not palatalize
the root-final consonants (as these two features are linked to the ClsP).
Note as well that nothing in the structure (30) prevents gender-ambiguous
roots, hence, gender-syncretic hypocoristic forms are expected.
The examples in (31) below show that all of these predictions are, indeed,
correct. For instance, let us start from the gender-ambiguous root /daniel/, seen
the first column in (31). Its simple truncation produces the form in the middle
column. This truncated form can be further lengthened and yields the form in
the last column.
These two truncated forms differ in the pragmatic context in which they can
be used. Specifically, the contexts in which the lengthened (and palatalized)
truncated form D[aː]ň(-a) can be used form a subset of contexts, in which the
non-lengthened (and non-palatalized) form D[a]n(-Ø/-a) is used. Quite informally,
the lengthened variant is used only when the hearer/referee is a child or
an old friend. The non-lengthened variant, on the other hand, signals that the
speaker is familiar with the hearer/referee. Moreover, it is true that many of the
non-lengthened (and non-palatalized) truncations became regular first names;
see e.g. Max (< Maxmilián), Ed(-a) (< Eduard), these are masculine, or feminines
Dor(-a) (< Dorot(-a)), El(-a) (< Eleonor(-a)).30
30	 It should be noted that are also truncations with palatalization, but, crucially, without
lengthening, as, for example, P[e]ť(-a) (< P[e]tr), P[a]ť(-a) (< P[a]trik), St[a]ň(-a)
(< St[a]nislav) or L[a]ď(-a) (< L[a]dislav). At the first blush, it might seem as if these
examples violate our claim that lengthening and palatalization are triggered by a single
lexical entry. However, these short forms also have lengthened cousins, i.e., P[eː]ť(-a),
P[aː]ť(-a), St[aː]ň(-a), L[aː]ď(-a). Quite generally, such variation in length is strictly
dialectal and, moreover, it strongly correlates with vowel length variation in other
morphological categories. So, a speaker using a short hypocoristic P[e]ť(-a) would also
typically use a short version of the monosyllabic infinitives like sp[a]t ‘to sleep’ and/or a
short version of a feminine noun of a-declension like ž[a]ba ‘frog’. The dialectal variation
goes in the other direction as well, so speakers that use the lengthened hypocoristic
P[eː]ť-a tend to use long versions of both the infinitive sp[aː]t and the feminine noun
ž[aː]ba.
74
(31) full form familiarity form closeness form
D[a]niel(-ØM
/-aF
) D[a]n(-ØM
/-aF
) D[aː]ň(-aM/F
)
To state the obvious: the middle form in (31) represents an expected case where
only the lower hypocoristic projection is present. Both the meaning (familiarity)
and the phonological properties (simple truncation) clearly confirm the existence
of an intermediate step in the derivation of the closeness hypocoristics. This
intermediate step in the derivation of the closeness hypocoristic is a constitutive
feature of the present account, and contrasts with Bethin’s (2003) approach,
where closeness-expressing hypocoristics are derived in a single step.
3.6  Suppletive Forms
Hypocoristics are also involved in suppletion. For example, a masculine hypocoristic
Pep(-a) corresponds to a masculine first name Jozef. This, then, means that
the lexical entry with the form /pep/ (32a) points to the lexical entry with the
form /jozef/ in (32b). The entries in (32c) and (32d) derive a suppletive pair
Ludmil(-a) – Líd(-a) where both names are feminines.
(32) a. </pep/, [ClsP
Cls [FamP
Fam </jozef/>]]>
b. </jozef/, [MascP
Masc [NP
N]]>
c. </liːd/, [ClsP
Cls [FamP
Fam </ludmil/>]]>
d. </ludmil/, [FemP
Fem [MascP
Masc [NP
N]]]>
Finally, the forms in (33)—examples of a hypocoristic suppletion—provide
an argument for splitting hypocoristic and gender features syntactically. The
gender-ambiguous root /jan/ has a suppletive form /jen/; and it is the form
/jen/ that appears in both the lengthened and the suffixed hypocoristic of the
masculine gender. However, the feminine hypocoristics of both types show no
suppletion at all: they are built on the form /jan/, as can be seen in the right-most
column of the table in (33).
(33) first name masculine hypocoristics feminine hypocoristics
J[a]n(-ØM
/-aF
) lengthened: J[eː]ň(-a) lengthened: J[aː]ň(-a)
suffixed: J[e]n-d(-a) suffixed: J[a]n-č(-a)
The lexical entries for both root forms are given in (34). The entry for /jan/ in
(34a) contains the full gender tree and this way, its gender-ambiguity is encoded
75
(and preserved). (34b) shows the entry for /jen/: FamP dominates MascP and
the whole structure points to the lexical entry for the root /jan/.
(34) a. </jan/, [FemP
Fem [MascP
Masc [NP
N]]]>
b. </jen/, [FamP
Fam [MascP
Masc </jan/>]]>
The figure in (35) illustrates the derivation of the feminine form J[aː]ň(-a), the
lexical items used in the derivation are shown in (36). The spell-out proceeds
in the manner described above for other bi-gender forms. So, after the FemP
was created, the item that matches the full gender tree overwrites the previous
spell-out. In the next step, the FamP is spelled out by the one-syllable template;
its association with the one-syllable root has no overt effect. Finally, after the
merger of ClsP, its complement undergoes leftward movement, so that the ClsP
matches the entry in (36c) and it can thus spell it out.
(35) ClsP
              | |template <= FamP ClsP =˃ length
         | |          |
Fam FemP =˃ /jan/ Cls
         | |
Fem MascP /jan/
               | |
Masc NP
(36) a. </jan/, [FemP
Fem [MascP
Masc [NP
N]]]>
b. <CVC, [FamP
Fam [XP
X ...[NP
N]]]>
c. <VC, [ClsP
Cls]>
Let us proceed to derivation of the masculine form J[eː]ň(-a), built, recall, on the
suppletive root. In (37), there are two lexical items that can, in principle, spell
out FamP: both the item for the template in (36b) and the item for the suppletive
root in (34b) match the tree dominated by the FamP. In this competition, the
suppletive root wins—and, consequently, the previous spell-out /jan/ is overwritten
by the form /jen/. The derivation in (37) illustrates the well-known Elsewhere
Principle (also sometimes called Minimize Junk in Nanosyntax; see Starke
2009); according to the Elsewhere Principle, from a set of spell-out candidates,
the most specific candidate “wins”. In the case discussed here, the lexical entry
for the suppletive root in (34b) contains two syntactic nodes (FamP and MascP)
and the pointer to the entry in (34a), it is thus more specific then the entry for
76
the template in (36b), as this includes just one node (FamP) and the pointer to
a structurally defined set of entries. Hence, the suppletive root is more specific
than the template (it can be used in less syntactic contexts) and, thus, it “wins”.31
(37) ClsP
             | |/jen/ <= FamP ClsP =˃ length
        | |               |
Fam MascP /jan/ Cls
          | |
Masc NP
It is worth pointing out that once again, these suppletive forms nicely illustrate
the need for the intermediate step in the derivation of closeness hypocoristics.
Specifically, what we see is that both of the forms J[e:]ň(-a) and J[e]n-d(-a) show
the regular ingredients of a hypocoristic: lengthening in the first case, and d-suffixation
in the second case. So, this part of the formation is fully regular. The only
irregular thing is that the base is not the expected Jan, but instead a suppletive
base Jen. What that shows is that there is a need for the notion of a “base” that
is different from the actual name, and which (as I argue) corresponds to FamP.
3.7 Summary
This chapter was devoted to hypocoristics. I proposed that there are two components
of the hypocoristic meaning, as identified in the literature, i.e., familiarity
and closeness, and that these components correspond to two hierarchically ordered
syntactic projections. In Czech, these two projections, in turn, are spelled
out by different means, namely by the truncation of the root to one syllable, and
by pure lengthening of (the root’s) vowel or a consonantal suffix. I argued that
the truncation (shared by all types of hypocoristics) is induced by a prosodic
template; it spells out the lower hypocoristic projection, the one encoding the
familiarity meaning. The complementary distribution between the two other
formal processes (lengthening and suffixation) follows from the rather different
31	 There exists yet another hypocoristic form associated to the first name Jan, i.e., Honz(-a)
(borrowed from German Hans). Since it is a familiarity-expressing masculine, its
lexical entry should look like that for the root /jen/ in (31b). In that case, the roots
/jen/ and /honz/ would be absolute synonyms, which means that they should be interchangeable
in all syntactic contexts. This is, however, not the case: only the root
/jen/ occurs in the context of the ClsP. I leave the solution of this synonymy problem
for further research.
77
syntactic structures I proposed for lengthened and suffixed hypocoristics. In
particular, I proposed that suffixed hypocoristics are binominal constructions
in which the left-hand noun corresponds to a truncated first-name root and the
right-hand one is realized by a hypocoristic suffix (which is either -d or -č depending
on the gender).
79
4.  Length in Diminutives
This chapter is devoted to diminutives whose formation is based on the two formal
processes: segmental suffixation and lengthening of base-final vowels. Both
processes are illustrated in (1). The table compares diminutive forms of masculine
first names with their basic forms: the diminutives (in the right part of the
table) all end in the suffix -ek which is concatenated with the lengthened base.
(1) base ek-diminutive
Fil[i]p Fil[iː]p-ek
Jak[u]b Jak[ou]b-ek
Joz[e]f Joz[iː]f-ek
Vikt[o]r Vikt[uː]r-ek
Ad[a]m Ad[aː]m-ek
In the previous chapter, I discussed hypocoristic forms in which suffixation and
lengthening were complementary. Recall that affixation of the masculine hypocoristic
suffix -d is never accompanied by lengthening of the preceding vowel
(F[i]lip > F[i]l-d(-a) / *F[i:]l-d-a); base vowels, on the other hand, lengthen only
in suffix-less forms (F[i]lip > F[iː]l(-a)). I explained this pattern as resulting from
different morphosyntactic structures of these two types of hypocoristics. In particular,
I proposed that they share a single head spelled out either by a prosodic
affix (inducing base-vowel lengthening)—or by the suffix -d, and that the spellout
really depends on the position of this head in the morphosyntactic tree.
From this perspective, the co-occurrence of lengthening and suffixation—as
seen in diminutives in (1) above—can be interpreted in two ways. First, we can
consider lengthening to be a by-product of suffixation (i.e., a kind of readjustment
rule). In the spell-out model used in this book, it translates to the requirement
that both the formal processes must be introduced in a single spell-out
step. Under this scenario, the prosodic affix (responsible for lengthening) and the
segmental suffix -ek are nothing else but two parts of a single lexical entry. The
second possibility is to follow the same logic as applied to hypocoristics and thus
claim that each of the affixes represents a separate lexical item. In such a case, diminutives
are by necessity derived in two steps, and that, in turn, means that they
span over several syntactic projections (as do hypocoristics). In what follows, I
argue in favor of the latter scenario. Before developing it in the Nanosyntactic
80
framework, the next section reviews an alternative, a template-based approach
to the diminutive length, proposed by Scheer (2003, 2004).
4.1  Scheer’s (2003, 2004) Templatic Analysis
Table (1) lists diminutives derived exclusively by the suffix -ek. However, there
are also those ending in -ík. Moreover, sometimes can either of these suffixes attach
to the same root, as illustrated by the three pairs of diminutive forms in (2).
(2) base VV-ek V-ík
sl[o]n sl[uː]n-ek sl[o]n-ík ‘elephant’
kl[u]k kl[ou]č-ek kl[u]č-ík ‘boy’
sr[a]b sr[aː]b-ek sr[a]b-ík ‘piker’
The examples above follow a clear distributional pattern: base-vowel lengthening
occurs only before the short diminutive suffix -ek. If a diminutive is formed by
the long suffix -ík, the base-vowel remains short.
Such a vowel-length pattern (i.e., lengthening is triggered only before a short
suffix) led Scheer (2003, 2004) to claim that the diminutive formation is of a
templatic nature. In his view, then, there is a diminutive template, which involves
the base and the diminutive suffix. Crucially, vowels in this templatic domain
must weigh exactly three moras. (Recall that short vowels are monomoraic,
while long vowels and diphthongs are bimoraic.) This is the reason why a short,
(i.e., monomoraic) suffix -ek triggers the lengthening of a short, i.e., monomoraic
base: sraμ
b > sráμμ
b-eμ
k. Merger of a short base and a long suffix -ík produces
a required, trimoraic result, hence, no lengthening needs to occur in ík-forms:
sraμ
b > sraμ
b-íμμ
k.
Moreover, Scheer (2003:100, 2004:229) strives to show that the base-vowel
length is manipulated not only by the ek-diminutives, but by the ík-diminutives
as well. To illustrate the point, he cites the following six examples of ík-formed
diminutives apparently accompanied by base-vowel shortening: hl[iː]n(-a) –
hl[i]n-ík ‘earth, aluminium’, vr[aː]n(-a) – vr[a]n-ík ‘crow, black horse’, chl[eː]b
– chl[e]b-ík ‘bread, small bread’, citr[oː]n – citr[o]n-ík ‘lemon, lemon tree’, k[iː]bl
– k[i]bl-ík ‘bucket, small bucket’, t[aː]t(-a) – t[a]t-ík ‘father, fogey’. These examples—according
to Scheer—prove existence of the diminutive template; in this
instance, bimoraic bases shorten their vowels before the bimoraic suffix -ík to
satisfy the trimoraic template.
However, I now set to show that Scheer’s templatic analysis of the length pattern
illustrated in (2) is problematic—and is such for a number of reasons. First
81
of all, not all diminutive forms fit the trimoraic template, as is illustrated in (3).
The ík-forms such as pláμμ
št-íμμ
k (and the others in the left part of the table) are
one-mora too big. On the other hand, the ek-forms such as svaμ
l-eμ
k (and the
others on the right), are one-mora too small—they are only bimoraic. And let
me also add that such quadrimoraic and/or bimoraic forms are not rare at all.
(3) base VV-ík base V-ek
pl[aː]šť pl[aː]št-ík ‘coat’ sv[a]l sv[a]l-ek ‘muscle’
š[eː]f š[eː]f-ík ‘boss’ bl[o]k bl[o]č-ek ‘notepad’
čm[ou]d čm[ou]d-ík ‘smoke’ čl[u]n čl[u]n-ek ‘boat’
Furthermore, the evidence for diminutive shortening (Scheer’s main argument
for the trimoraic template), is rather questionable. Since the six bases mentioned
above alternate in length also in other contexts than just diminutives, it is natural
to ask whether the ík-forms really involve shortening of lexically long vowels.
In other words, there is no clear evidence that the roots that—according to
Scheer—involve “shortening” are lexically stored with long vowels; let us consider
the facts below.
Looking at the roots hl[iː]n(-a) ‘earth’ and vr[aː]n(-a) ‘crow’ first, notice that
these roots appear short not only in the ík-diminutives, but also in ek-derivatives
(hl[i]n-k(-a) ‘loess’, vr[a]n-k(-a) ‘small crow’) and in adjectives (hl[i]n-ěn(-ý)
‘earthen’, vr[a]n(-ý) ‘black’). In fact, these two roots show long vowels only
when concatenated with case markers, e.g. with the nominative singular marker
-a (vr[aː]n-a, hl[iː]n-a) or the instrumental singular marker -ou (vr[aː]n-ou,
hl[iː]n-ou). Given this peculiar distribution, it seems difficult to maintain that
the derivation of the ík-forms involves shortening of the lexically long roots.
Rather, it is not unthinkable to claim that the length alternants of the roots are
suppletive and stored separately, rather than derived by a regular process.32
32	 Moreover, it is doubtful that these forms are diminutives to begin with. To see that,
consider the common observation that Czech belongs to languages where diminutive
formation preserves the gender of the base; Czech, thus, is in stark contrast to,
for example, German where both masculine and feminine bases produce diminutive
of neuter gender (see e.g. Wiltschko & Steriopolo 2007). From this perspective, it is
striking that the masculines hlin-ík and vran-ík are derived by the suffix -ík from the
feminines hlín(-a) and vrán(-a). Therefore, it is doubtful that they can be considered
as genuine diminutive forms (as Scheer assumes). By contrast, the ek-forms hlin-k(-a)
and vran-k(-a) preserve the feminine gender of the bases.
82
Also in the other pair reported by Scheer, i.e., chl[eː]b – chl[e]b-ík, the long
variant of the root has restricted distribution. It occurs only in a nominative/
accusative singular for, that is in forms that lack an overt ending. In all other environments,
including, for example, the nominative plural form chl[e]b-y or the
adjectival form chl[e]b-ov(-ý), the root is short. The highly restricted distribution
of the long alternant thus indicates that the root is lexically short and it lengthens
only in certain case-number contexts—and not the other way around.33
As for the pair citr[oː]n – citr[o]n-ík, a long <ó> alternates with a short <o> in
both the base (citr<ó/o>n) and the ík-form (citr<ó/o>ník), these two forms are
orthographic (and orthoepic) variants. Furthermore, an ek-form shows orthographic
doublets as well: citr<ó>n-ek and citr<o>n-ek. Such a variation suggests
that the root ‘lemon’ (borrowed from Italian), can be stored with either a long
or a short vowel. Hence, there is no evidence supporting the claim that the íkderivation
involves shortening of [oː] to [o].
Another root cited by Scheer is ‘bucket’ (this time borrowed from German).
Also in this case we cannot unequivocally say that this root is lexically stored as
long. First, there is an orthographic variation between k<y>bl-ík and k<ý>bl-ík.
Moreover, the root has a variant kbel, with an initial consonant cluster followed
by a short epenthetic vowel, which also produces the ík-form, i.e., kbel-ík.
The final member of the “shortening” set not only has an ík-form t[a]t-ík,
there is also an ek-form t[a]ť-k(-a) (not mentioned by Scheer). Since both forms
contain a short variant of the root, one can conclude that the short form is selected
precisely by the evaluative context. This assumption is further supported by
an endearment form t[a]t-i, with a short variant of the root: it is used in vocative
speech acts where it stands in contrast to the pragmatically-neutral form t[aː]t-o
built on the long root.
To conclude, none of the six examples offered by Scheer (2003, 2004) shows
convincingly that ík-suffixation involves shortening of lexically long roots and,
thus, the existence of the diminutive template fails to be proven.
There is an additional—and complicating—issue for the templatic analysis:
vowel-zero alternations. The initial vowel of the short diminutive suffix -ek alternates
with zero. The alternation is predictable, it happens whenever the suffix is
33	 There are other nominal roots with the same alternating pattern as chléb, incidentally
not mentioned by Scheer (2003, 2004): ‘horse’ k[uː]ň – k[o]n-ě – k[o]n-ík ‘Nsg, Npl,
dimin.’, ‘tallow’ l[uː]j – l[o]j-e – l[o]j-ík ‘Nsg, Npl, dimin.’, ‘carriage’ v[uː]z – v[o]z-y –
v[o]z-ík ‘Nsg, Npl, dimin.’, ‘rain’ d[eː]šť – d[e]št-ě – d[e]št-ík ‘Nsg, Npl, dimin.’, ‘frost’
mr[aː]z – mr[a]z-y – mr[a]z-ík ‘Nsg, Npl, dimin.’. Such (allegedly) shortening roots are
analyzed in section 4.3.2.
83
followed by another vowel. However, any such form with the vowel-less version
of the suffix violates the trimoraic templatic restriction. For example, a short root
vt[i]p ‘joke’ lengthens to vt[iː]p in the diminutive context, producing the nominative
singular form vt[iː]p-ek; this form is trimoraic, as predicted by Scheer’s
analysis: (vtíμμ
p-eμ
k). However, things work differently in, for instance, the nominative
plural form. The nominative plural is marked by a vocalic ending, thus, as
a consequence, the diminutive suffix gets the vowel-less -k. Assuming that only
vowels are moraic in Czech (as Scheer does), the templatic domain in the nominative
plural form is thus bimoraic: (vtíμμ
p-k)-y. At the first sight, this problem
can be solved by including inflectional markers in the template; in that case, the
nominative plural form fits the postulated trimoraic requirement: (vtíμμ
p-k-yμ
).
This templatic-domain extension works for those inflectional forms that include
short-vowel endings. However, things get worse once we take into consideration
also long-vowel case endings: should they be included in the template, the produced
forms are quadrimoraic, as shown by the following forms: genitive plural
(vtíμμ
p-k-ůμμ
), dative plural (vtíμμ
p-k-ůμμ
-m) and locative plural (vtíμμ
p-k-áμμ
-ch).
Moreover, there are also polysyllabic bases for diminutives in Czech—and
they pose a problem for defining the template boundaries (and, as it happen,
pose another problem for Scheer’s templatic analysis of diminutives). Consider
the following trisyllabic base učit[e]l ‘teacher’ and its lengthened diminutive
form učit[iː]l-ek. The base is composed of three morphemes: the verbal root uč,
the thematic suffix -i and the agentive suffix -tel, i.e., uč-i-tel. Thus, in principle,
one might want to see the templatic domain as composed from two morphological
units: the agentive suffix -tel and the diminutive suffix -ek, i.e., uč-i(-týμμ
l-ekμ
).
There are, however, monomorphemic bases consisting of more than one syllable—and
they lengthen as well. For example, in a bisyllabic base jet[e]l ‘clover’,
the second e lengthens in the diminutive form jet[iː]l-ek, even though the string
tel does not have a morphemic status (as opposed to the tel-string in učitel). If
lengthening is triggered by the trimoraic template, as proposed by Scheer (2003,
2004), we have to claim that in diminutives like jetýlek, based on monomorphemic
bases, the templatic domain involves the diminutive suffix and only one
phonological part of the root, i.e., its last syllable: je(týμμ
l-ekμ
). This scenario must
be ruled out: it allows morphological templates (and the diminutive template in
Scheer’s rendering is certainly such) not to correspond to morphological units.
That is clearly not what we want.
Summing up: there is no base-shortening in ík-diminutives, ek-diminutives,
on the other hand, very often show lengthening of the base last syllable. Once we
84
see the data in this way, suddenly there are only few counterexamples (like svalek
in table (3) above)—and I come back to them in section 4.3.1.
4.2  Phonology of Diminutive Suffixes and Diminutive Stems
Let us start with recapitulating the phonological properties of diminutives. We
have identified two formal processes involved in their formation: (i) lengthening
base-final vowels, (ii) suffixation by [ek] or [iːk]. In this chapter, I examine these
formal processes in greater detail.
4.2.1  Evidence That -ek Starts with [e] Lexically
We have already seen the suffix -ek alternations: it is either vowel-initial or consonant-initial
depending on the context to its right. The alternating pattern is
summarized in (4).
(4) a. -k/ _-V b. -ek/ _# c. -ek/ _-ek d. -ek/ _-k
dár-[k]-ov-ý dár-[ek] dár-[eʧ ]-[ek] dár-[eʧ ]-[k]-ov-ý
In the adjective dár-k-ov-ý ‘gift-related’ in (4a), a vowel-initial (adjectival) suffix
-ov follows the diminutive suffix and it then surfaces as vowel-less. Crucially,
notice that the suffix -ov itself is not alternating: its initial vowel does not alternate
with zero. If the diminutive suffix is followed by another alternating suffix,
it always appears with an initial [e], as shown in (4c,d). Both (4c,d) contain a
double diminutive with two suffixes -ek following one another. The first diminutive
suffix always surfaces as vowel-initial—regardless whether the second suffix
is actually vocalized (as in the nominative singular form dár-eč-ek in (4c)) or not
(as in the adjectival form dár-eč-k-ov-ý in (4d)). Finally, the nominative singular
forms of both a simple diminutive dár-ek in (4b) and its double-diminutive
cousin dár-eč-ek in (4c) illustrate that the diminutive suffix appears vowel-initial
also word-finally, i.e., when there is no other suffix following it.34
The vowel-zero alternating pattern (found in all Slavic languages) has been
analyzed intensively in the literature (for an overview, see Scheer & Ziková
(2010)). Generally, there are two approaches: vowels alternating with zero are
either lexically present or they are epenthetic. The main argument in favor of the
34	 Note that the examples in (4) also illustrate the alternation of the consonant of the
diminutive suffix consonant shows: it alternates between a velar [k] and a postalveolar
[ʧ], the latter one appearing in the double diminutive form. This (perfectly regular)
consonant alternation is discussed further on in section 4.2.3
85
former account, i.e., an alternating vowel is lexically stored, is the fact that the
alternating vowel is not predictable from phonotactics alone. This unpredictability
is well illustrated by the (near) minimal pairs in (5): there are three identical
consonant clusters, but this cluster is sometimes perfectly fine as a word-final
cluster, but sometimes it just must be broken up by the alternating vowel [e].
Thus: the contexts with the vowel-alternant must be lexically distinguished from
those context in which it does not appear—and, clearly, the most straightforward
option is to include the vowel in the lexical representation of those roots and affixes
where we do see it (even if it is not always the vowel surfaces).
(5) Cx
Cy
# Cx
[e]Cy
# Cx
Cy
-V
pa[rk] ‘park, Nsg’ dá[rek] dá[rk]-u ‘small gift, Nsg, Gsg’
zmá[ʧ k] ‘he pushed’ ptá[ʧek] ptá[ʧk]-a ‘small bird, Nsg, Gsg’
su[lʦ] ‘aspic, Nsg’ pa[leʦ] pa[lʦ]-e ‘thumb, Nsg, Gsg’
Notice, however, that the alternating vowel [e] has to be underlyingly distinguished
from its non-alternating cousin, as both types of e’s can occur in the
same phonotactic environment. Table (6) provides a couple of examples illustrating
this surface ambiguity between the alternating and non-alternating [e].
(6) Cx
[e]Cy
# Cx
[e]Cy
-V Cx
[e]Cy
# Cx
Cy
-V
úl[lek] úl[lek]-u úl[lek] úl[lk]-u ‘scare, small hive; Nsg, Gsg’
je[tel] je[tel]-e da[tel] da[tl]-a ‘clover, woodpecker; Nsg, Gsg’
ba[ret] ba[ret]-u [ret] [rt]-u ‘beret, lip; Nsg, Gsg’
In the Strict CV framework—the phonological framework we use in this book—
it is relatively easy to distinguish between vowels alternating with zero on the one
hand and stable vowels on the other: the former are lexically floating segments
while the non-alternating vowels are segments that are associated with their Vslots
already in the lexicon. Thus, the diminutive suffix -ek involves a lexically
floating vowel [e] on its left. Its exact lexical representation is discussed further
on in section 4.3.5.
4.2.2  Evidence That -ík Is Not Just Lengthened -ek
Since lengthening in diminutives involves raising of mid vowels, as I illustrated
by the examples in (1) above, one might be tempted to claim that the lexical
identity of the long high vowel in -ík is a short mid vowel [e]—only lengthened.
In other words, it is not unthinkable to propose an analysis that would have it
86
so that there is a single diminutive suffix which surfaces either as e-initial or
í-initial, depending on its position with respect to the prosodic affix that causes
lengthening in diminutives. If the prosodic affix were attached higher than the
segmental suffix -ek, the prosodic suffix would trigger lengthening—and raising
of its initial [e] to [iː]. If the prosodic affix were to attach lower than -ek, the
lengthening would rather affect the root and the segmental suffix would thus
surface in its lexical form, i.e., with an initial e.
This scenario predicts that both shapes of the diminutive suffix should behave
in a similar way phonologically: if -ek and -ík shared a single underlying form,
they would be are expected to share phonological properties. This is, however,
not the case: these two suffixes differ phonologically—and in fact, they are different
in two different ways. First, the initial vowel alternates with zero in -ek, but
not in -ík; cf. the contrast between dár-[ek], dár-[k]-u ‘small gift, Nsg, Gsg’ and
oltář-[iːk], oltář-[iːk]-u ‘small altar, Nsg, Gsg’. Second, each of the suffixes triggers
different palatalization patterns, as is illustrated in table (7). The comparison of
three pairs of ek-diminutives (in the left-hand part) and ík-diminutives (on the
right) shows that root-final dentals [d t n] are preserved before the -ek, but they
change to palatals [ɟ c ɲ] before the suffix -ík.
(7) base ek-diminutive base ík-diminutive
scho[d] schů[d]-ek ‘step’ bo[d] bo[ɟ]-ík ‘point’
plo[t] plů[t]-ek ‘fence’ chr[t] chr[c]-ík ‘geyhound’
župa[n] župá[n]-ek ‘bath robe’ po[n](-y) po[ɲ]-ík ‘pony’
One might want to claim that the palatalization of dentals in ík-diminutives is
due to the raising of a mid [e] to a high [iː]. However, this explanation is untenable
as the data in table (8) show. There are three nominal roots in (8) whose final
syllable starts with a dental followed by the mid vowel [e]. And these dentals are
preserved even in ek-diminutives in which the base internal [e] has raised (and
lengthened) to [iː]. These data thus clearly show that the e-to-í raising—in fact—
does not lead to dental palatalization.
(8) base ek-diminutive
mo[de]l mo[diː]l-ek ‘model’
prs[te]n prs[tiː]n-ek ‘ring’
pa[ne]l pa[niː]l-ek ‘panel’
87
In sum, there is phonological evidence that the suffix -ík is not just a long version
of the suffix -ek. There is an additional piece of evidence pointing in the same
direction: feminine nouns in which the suffix appears short and the short vowel
is [i]. I discuss them presently.
4.2.3  Evidence That -ík Starts with [i] Lexically
Up to now, I have been discussing diminutive forms of masculine gender. Both
diminutive suffixes (-ek and -ík), however, derive also feminine diminutives (and
neuter ones that I leave aside for the moment). The suffix -ek behaves phonologically
the same way in both masculines and feminines: if it surfaces as vowelinitial,
the vowel is a short [e]. The table below illustrates that it holds for simple
and double diminutives of both genders, i.e., masculines in (9a) and feminines
in (9b). Note that for the feminine diminutive, I opt for genitive plural form, as it
is that form in the feminine paradigm that lacks an overt case marker—just like
the nominative singular of the masculine gender. This adjustment then allows us
to see the parallel behavior of the suffixes clearly.
(9) simple diminutive double diminutive
a. dár-[e]k dár-[e]č-[e]k
gift-dim.nom.sg gift-dim-dim.nom.sg
b. strán-[e]k strán-[e]č-[e]k
page-dim.gen.pl page-dim-dim.gen.pl
Phonological behavior of the suffix -ík, on the other hand, is gender-sensitive:
it is long in masculines, but short in feminines. The pattern is illustrated in the
table in (10). The table lists double diminutive forms, in which the i-initial suffix
is followed by the e-initial one. (Note that the velar of the first suffix is regularly
palatalized to [ʧ ].) The masculine forms in (10a) display a long version of the
i-initial suffix—while the feminines in (10b) show a short version. (A phonological
independence of the alternation is demonstrated by comparing the forms in
the left-hand column, where the vowel in question appears in an open syllable,
i.e., ko.n[iː].ček, with the forms on the right showing the vowel in a closed syllable,
i.e., ko.n[iː]č.ka. Once again, this leads to comparing the masculine nominative
singular with the feminine genitive plural in the first column, and the
masculine genitive singular with the feminine nominative plural in the second
column.)
88
(10) a. kon-[iː]č-ek kon-[iː]č-k-a
horse-dim-dim.nom.sg horse-dim-dim-gen.sg
kluč-[iː]č-ek kluč-[iː]č-k-a
boy-dim-dim.nom.sg boy-dim-dim-gen.sg
les-[iː]č-ek les-[iː]č-k-a
forest-dim-dim.nom.sg forest-dim-dim-gen.sg
b. ryb-[i]č-ek ryb-[i]č-k-a
fish-dim-dim.gen.pl fish-dim-dim-nom.sg
ruč-[i]č-ek ruč-[i]č-k-a
hand-dim-dim.gen.pl hand-dim-dim-nom.sg
věž-[i]č-ek věž-[i]č-k-a
tower-dim-dim.gen.pl tower-dim-dim-nom.sg
Generally, the length alternation in the i-initial suffix can be represented either
derivationally or suppletively. In the first case, there are further two scenarios:
the suffix either starts with a short [i] lexically and it lengthens in the masculine
context, or, the second possibility is to see the long [iː] stored lexically and the
vowel shortens in the context of feminine nouns. Under the suppletive scenario,
both versions of the suffix (i.e., both the long and short version) would be stored
in the lexicon (together with their gender specification). In what follows, I argue
for the derivational approach and within that, I favor the lengthening scenario.
Before elaborating on it, let me explore the length pattern in ek-diminutives.
4.2.4  Length Alternations in Bases: The Gender Asymmetry
It is also in ek-diminutives we can observe a robust correlation between vowellength
alternations and the gender features. The gender-contrast is well illustrated
by the diminutive forms of gender-ambiguous first names, as shown in (11). As we
have discussed already in chapter 3, a lot of first names (or, at least the first-name
bases stripped off their inflectional morphology) are ambiguous between masculine
and feminine gender. Recall that such gender-ambiguous nouns actually differ
only in the declension patterns. The most significant difference regards the nominative
singular (which is also the citation form used in metalinguistic contexts): in
the feminine paradigm, it is marked by -a, while the masculine paradigm has no
overt marker in the nominative singular form. However, there is another ending
-a in the masculine declension: the genitive singular marker. To complete the symmetry,
there is a paradigm cell in the feminine declension that lacks an overt case
89
marker: genitive plural. Thus, if we take the bare-root form Ivan and the a-final
form Ivan-a, each form can be either feminine or masculine, thus, both Ivan and
Ivan-a are gender-ambiguous. Importantly, it is only non-diminutive forms that
are ambiguous—diminutive forms (even though they follow the same declensional
patterns as the first names) do not show such ambiguity. And, not surprisingly,
it is the length of the base-final vowel that gives the (otherwise syncretic) forms
out. While the masculine diminutive forms systematically show length, the corresponding
feminine forms are short (i.e., lack lengthening).
To see this, consider, for instance, the name Ivan. As it should be apparent
from the discussion, the form Ivan is ambiguous between masculine nominative
singular and feminine genitive plural. When the diminutive suffix is attached,
the ambiguity disappears and a minimal diminutive pair is born: Iv[aː]n-ek
(masculine, lengthened) vs. Iv[a]n-ek (feminine, not lengthened). Analogously,
the form Ivana is ambiguous between masculine genitive singular and feminine
nominative plural. Under diminutivization, the ambiguity is resolved by length:
the lengthened form Iv[aː]n-k-a is unambiguously masculine, while the short
form Iv[a]n-k-a is feminine. The table below illustrates the contrast for a variety
of first names. All of the names are—like Ivan—ambiguous without the diminutive
suffix, but once they get diminutivized, the ambiguity disappears.
(11) Masc (Nsg) Fem (Gpl) Masc (Gsg) Fem (Nsg)
Iv[aː]n-ek Iv[a]n-ek Iv[aː]n-k-a Iv[a]n-k-a
Rom[aː]n-ek Rom[a]n-ek Rom[aː]n-k-a Rom[a]n-k-a
Vlad[aː]n-ek Vlad[a]n-ek Vlad[aː]n-k-a Vlad[a]n-k-a
Bohd[aː]n-ek Bohd[a]n-ek Bohd[aː]n-k-a Bohd[a]n-k-a
Mart[iː]n-ek Mart[i]n-ek Mart[iː]n-k-a Mart[i]n-k-a
Vikt[uː]r-ek Vikt[o]r-ek Vikt[uː]r-k-a Vikt[o]r-k-a
To summarize so far: ek-diminutives alternate in vowel length in a way similar to
ík-diminutives. In both cases, the alternation is gender-sensitive and in both cases,
the long alternants occur in masculines. The difference is a locus of the alternation:
it is the suffix in ík-diminutives, but the base in ek-diminutives. And then
there is the question of productivity (which poses another difference between
the diminutives). The alternation of the i-initial suffix is regular and productive:
no masculine diminutive ever shows a short version of the suffix, thus, there are
no forms like *kon-[i]k or *kon-[i]č-ek. Masculine ek-diminutives, by contrast,
do not behave uniformly at all: they show lengthened bases (see (11) above)—or
there is variation. For example, there is both Jindř[iː]š-ek and Jindř[i]š-ek or both
90
Šim[uː]n-ek and Šim[o]n-ek with either a lengthened or a non-lengthened variant
of the first-name base.
Despite the variation, the gender-sensitive pattern is still detectable even in
ek-diminutives (summarized in (12)): if a base alternates in length, the long alternant
always appears in the masculine ek-form. The complementary pattern of
an alternation, cf. a short masculine and a long feminine, is not attested.
(12) base ek-masculine ek-feminine
V V/VV V
Jarm[i]l(-a) Jarm[i/iː]l-ek Jarm[i]l-k(-a)
V VV V
Mart[i]n(-a) Mart[iː]n-ek Mart[i]n-k(-a)
To conclude, the behavior of gender-ambiguous nouns indicates that lengthening
of base-final vowels is activated only in masculine diminutives. Moreover,
gender-specific nouns support this conjecture as well, as illustrated in table (13):
masculine-only names lengthen their vowels in ek-diminutives (in the left part
of the table), while feminine-only names (on the right) do not. (Notice also that
all bases below have a similar phonological structure—all end in /an/—, which
suggests that the alternation is really not sensitive to the phonotactics.)
(13) base ek-masculine base ek-feminine
Mil[a]n Mil[aː]n-ek Zuz[a]n(-a) Zuz[a]n-k(-a)
Duš[a]n Duš[aː]n-ek H[a]n(-a) H[a]n-k(-a)
Al[a]n Al[aː]n-ek Jol[a]n(-a) Jol[a]n-k(-a)
The length pattern is followed not only by gender-specific first names, but also
by gender-specific common nouns, as illustrated in (14). Each row of the table
compares two ek-diminutives built on phonologically similar bases. (Note that
the roots in the second row are even homonymous.) And again: only masculine
bases undergo lengthening. Moreover, these bases are all loans, which suggests
that the diminutive lengthening is really a productive process synchronically.
(14) base ek-masculine base ek-feminine
muff[i]n muff[iː]n-ek ‘muffin’ mik[i]n(-a) mik[i]n-k(-a) ‘hoodie’
kab[e]l kab[iː]l-ek ‘cable’ kab[e]l(-a) kab[e]l-k(-a) ‘bag’
trakt[o]r trakt[uː]r-ek ‘tractor’ amf[o]r(-a) amf[o]r-k(-a) ‘amphora’
festiv[a]l festiv[aː]l-ek ‘festival’ mand[a]l(-a) mand[a]l-k(-a) ‘mandala’
91
It is fair to add that there exist also lengthened feminine forms of both first
names (e.g. Mil[a]d(-a) – Mil[aː]d-k(-a)) and common nouns (e.g. str[a]n(-a)
– str[aː]n-k(-a) ‘page’). However, such lengthened ek-feminines are generally
much less frequent then lengthened ek-masculines. To explain this contrast between
masculines and feminines, I propose that lengthening is an active process
only in ek-masculines. In particular, I claim that length arises from the insertion
of a prosodic VC-affix that, in turn, is a realization of (a part of) the diminutive
meaning. In ek-feminines, this part can be realized by the base itself and that is
why they only have short vowels (i.e., they do not have to lengthen the base).
From this perspective, lengthened feminine bases are suppletive, i.e., stored in
the lexicon (as opposed to their lengthened masculine cousins whose length results
from phonological derivation). This proposal is developed in the next sec-
tions.
4.3  Decomposing the Diminutives
This section presents the way the findings about phonological properties of diminutives,
presented in the previous chapters, can be derived from their morphosyntactic
structure.
4.3.1  Long vs. Short Diminutive Stems
When analyzing hypocoristics, I have established that the category of gender
is decomposed into two syntactic projections, hierarchically organized as
[femp Fem [mascp Masc]]. The idea was that morphosyntactic structure of
masculine nouns involves only the lower projection of the gender subtree, while
feminines involve both of them. Let me now introduce a new feature X: X encodes
(a part of) the diminutive meaning. From now on, the tree dominated by
XP is a diminutive stem. Since the X-head is placed on top of the gender subtree,
there are two syntactically different types of diminutive stems. (15a) shows a
structure of the masculine diminutive stem in which the diminutive feature X is
merged with MascP. The feminine stem in (15b) is then one-projection bigger:
the X is merged with the whole gender subtree.
92
(15) a. XP
    | |
X MascP
        | |
Masc NP
b. XP
     | |
X FemP
        | |
Fem MascP
            | |
Masc NP
In the previous section I showed that lengthening is an active process only in
masculine diminutives. To capture this generalization, I propose that there is a
prosodic affix and once it is merged in the tree, it triggers lengthening of basefinal
vowels. Moreover, this affix is lexically paired with both the masculine and
the diminutive features. The lexical entry for this affix is shown in (16).35
(16) </VC/, [XP
X [MascP
Masc]]>
The contrast between lengthened masculine ek-diminutives (e.g. kr[o]k – kr[uː]č-ek
‘step’) and those that do not lengthen (and are, in fact, exceptional, as e.g.
br[o]k – br[o]č-ek ‘slug’) then arises from the different spell-out of their stems.
In particular, the diminutive-stem tree in (15a) is spelled out either by a single
item, i.e., by the root itself (in non-lengthened forms such as br[o]č-ek), or it is
spelled out by two items, i.e., the root and the prosodic affix (in lengthened forms
such as kr[uː]č-ek). From this perspective, the contrast between masculine roots
maintaining short vowels and roots whose vowels lengthen in the diminutive
stem boils down to their lexical storage, as is illustrated in (17).
35	 Note, that the empty V-slot of the diminutive prosodic affix differs phonologically
from the empty V-slot of the hypocoristic prosodic affix discussed in chapter 3. Specifically,
the V-slot of the diminutive prosodic affix is provided with the root node and
this root node is absent in the hypocoristic affix. This difference derives the contrast
between qualitative lengthening (in diminutives) and pure lengthening (in hypocoristics).
Note further, that the VC-shape of the prosodic affix in (16) ensures that it is
inserted before the base-vowel, which means that its melody automatically spreads to
the inserted empty V-position. The lexical form of the prosodic affix thus determines
that manipulation with base-vowel length is independent of the diminutive syllable
structure; the reader is directed to section 2.5 for more technical details on melody
spreading. However, the VC-shape of the prosodic affix itself does not indicate which
of the base vowels should be lengthened. Since it is always the last vowel in the string
that gets the length, as clearly shown by polysyllabic bases in (14), the prosodic affix
must be lexically specified for the given prosodic anchor.
93
(17) a. </krok/, [MASCP
Masc [NP
N]]>
b. </brok/, [XP
X [MASCP
Masc [NP
N]]]>
The root /brok/ stored as a structure in (17b), is able to spell out the whole
diminutive-stem tree in (15a): the trees, i.e., the lexical one in (17b) and the one
created in syntax (shown in (15a)), fully match. This one-item-spell-out scenario
is depicted in (18b). The two item-scenario, involving insertion of the root /krok/
and the VC-affix (which triggers o-to-ů lengthening), is shown in (18a). Technically,
under this scenario, the NP moves to spec of the XP—as this movement
is the only possibility for the XP to be spelled out. Even more technically, the
spell-out procedure in (18a) involves the so-called backtracking (Starke 2018):
a portion of a syntactic structure spelled out by the root /krok/ first “increases”
and then “shrinks” in the subsequent steps of the cyclic derivation. In particular,
the form /krok/—lexically paired with [MASCP
Masc [NP
N]]—is inserted into the
structure after the merge of NP. It is reinserted after the next merge-step, in which
the MascP is created. In the next step, syntax creates the XP which, however, can
no longer be spelled out by the form /krok/. Thus, as the last option, the NPto-specXP
movement is applied leading to the constituent [XP
X [MASCP
masc]].
And this structure, then, matches the syntactic structure of the prosodic affix
in (16). As a consequence, the form /krok/ spells out just the NP, as it had done
previously, i.e., before the MascP was merged.
(18) a. XP
         | |
/krok/ <= NP XP =˃ /VC/
            | |
X MascP
      |
Masc
b. XP =˃ /brok/
    | |
X MascP
        | |
Masc NP
Let me now turn to the feminine stems; recall that they do not lengthen productively
in diminutives. I propose that the absence of lengthening is in fact a consequence
of the syntactic structure of feminines. Recall that for the feminine stems,
I have proposed the structure in (19a): the diminutive feature X is merged with
the whole gender tree. Notice also that there is no way the prosodic affix (the
trigger of lengthening) can enter this structure: the VC-affix is lexically specified
for [XP
X [MASCP
Masc]] (see the entry in (16) above). However, no such constituent
is present in the feminine stem in (19a) and, in addition, no thinkable
type of spell-out driven movement can derive it from the structure in (19a). (For
the typology of spell-out operations, see footnote 8 in section 3.3.1). Thus, the
94
diminutive feature X can be spelled out under one (and only) scenario, namely
the one-item scenario: a single lexical form spells out the whole diminutive-stem
tree in (19a). This proposal, moreover, draws a parallel between feminines and
non-lengthening masculines: both are built on bases that are lexically specified
for spelling out the diminutive feature X. The parallel becomes obvious once
we compare the lexical entry for the feminine root /bot/ ‘shoe’ in (19b) (which
produces a non-lengthened diminutive bot-k(-a)) with the lexical entry for the
masculine root /brok/ ‘slug’ depicted in (17b).
(19) a. XP
    | |
X FemP
        | |
Fem MascP
            | |
Masc NP
b. </bot/, [XP
X [FEMP
Fem [MASCP
Masc [NP
N]]]]>
In section 3.4.5, I looked at gender-ambiguous first-name roots, such as /ivan/
that produce both a masculine name Ivan(-Ø) and a feminine name Ivan(-a).
I claim that these roots involve the whole gender subtree lexically. Since their
diminutive forms follow the general pattern, whereby masculines are lengthened
(Iv[aː]n-ek) and feminines are not (Iv[a]n-k(-a)), the gender-ambiguous roots
must be lexically stored with the diminutive feature X. To illustrate this, consider
a lexical entry for the root ‘Ivan’: </ivan/, [XP
X [FemP
Fem [MascP
Masc [NP
N]]]]>.
In the feminine diminutive stem, the root behaves exactly like any other feminine-only
root, cf. /bot/ from above: it spells out the whole diminutive structure
shown in (19a). In the masculine stem, on the other hand, the spell-out scenario
for the masculine-only root /krok/ in (18a) is repeated. Concretely, the form
/ivan/ spells out just the NP and the rest of the masculine diminutive tree is spellout
by the prosodic affix; as a consequence /ivan/ lengthens to /ivaːn/.
Finally we can move to the last exception I have already mentioned: a couple of
lengthened feminine stems, e.g. str[a]n(-a) – str[aː]n-k(-a) ‘page’ or jah[o]d(-a)
– jah[uː]d-k(-a) ‘strawberry’. The same strategy used for masculines, cf. the prosodic
affix, which induces diminutive lengthening, cannot be used in feminines
at all: the feminines include the whole gender subtree. As a consequence, long
feminine stems must be lexically stored. This proposal is depicted in (20a). Literally,
this lexical entry says that the long form /jahuːd/ is “a diminutive version of
the root /jahod/”: it is specified for spelling out the diminutive feature X and it
points out to the lexical entry in (20b), which stores a short version of the root.
To express the lexical relatedness between the long and the short form of the
95
root, the old tool, i.e., a pointer , is used in (20a) (see section 3.3.1 for more
technical details on pointers).
(20) a. </jahuːd/, [XP
X </jahod/>]
b. </jahod/, [FemP
Fem [MascP
Masc [NP
N]]]>
Summing up, I proposed that lengthened masculine diminutive stems arise from
concatenation of the prosodic affix while lengthened feminine stems are lexically
stored. The suppletive approach to length alternations in feminines fits well
with our finding that length alternations are much less widespread in feminine
diminutives than in masculine ones.36
4.3.2  Against Shortened Stems
As we already seen, diminutive lengthening is not the only process recognized
in the literature—there is also a reverse process, i.e., diminutive shortening. It
is illustrated by pairs like d[uː]m – d[o]m-ek ‘house’ or ž[aː]b(-a) – ž[a]b-k(-a)
‘frog’; see e.g. Anderson & Browne (1973:460), Petr et al. (1986:302), Scheer
(2004:621), among others. In what follows, I argue that the only length-manipulation
process involved in diminutives is lengthening and it is induced by the
prosodic affix, as discussed above. In particular, I claim that there is no diminutive
shortening at all. The main argument supporting this hypothesis is the fact
that shortening analyses equate citation forms (i.e., nominative singular forms)
with lexically stored forms that serve as bases for diminutive stems—and that
this alignment is inaccurate.
Let us start with the following generalization regarding “shortening” masculines:
all masculine roots that undergo diminutive shortening show length
alternation throughout the declensional paradigm. The actual correlation between
vowel-length behavior of roots in the masculine declension and in ekdiminutives
is illustrated in (21). The table in (21a) shows the root ‘house’, its
short diminutive form d[o]mek is claimed to involve ů-to-o shortening. The first
two columns of the table demonstrate that the root alternates in length also in
the declensional paradigm. By contrast, roots that do not alternate in length in
the declension, never “shorten” their vowels in ek-diminutives, as is illustrated
36	 In the model proposed here, length variations seen in some diminutive stems can be
explained as arising from variations in lexical storage. For example, some speakers use
a non-lengthened version of the ek-diminutive rob[o]t-ek ‘robot’ while others use its
lengthened form rob[uː]t-ek; that translates in our system to the claim that while the
former store the root /robot/ with the diminutive feature X, the latter without it.
96
by the root ‘throne’ in (21b). This root has a long vowel [uː] in all the forms in
the declensional paradigm (as shown in the first two columns in (21b)), in the
ek-diminutive tr[uː]nek, and in the other derivatives as well (in the third column).
(21) a. Sg Pl
N d[uː]m d[o]m-y d[o]m-ek ‘small house’
A d[uː]m d[o]m-y d[o]m-a ‘at home’
G d[o]m-u d[o]m-ů d[o]m-ovník ‘caretaker’
D d[o]m-u d[o]m-ům d[o]m-ácí ‘homemade’
L d[o]m-ě d[o]m-ech d[o]m-obrana ‘militia’
I d[o]m-em d[o]m-y
b. Sg Pl
N tr[uː]n tr[uː]n-y tr[uː]n-ek ‘small throne’
A tr[uː]n tr[uː]n-y tr[uː]n-ní (sál) ‘throne (room)’
G tr[uː]n-u tr[uː]n-ů tr[uː]n-it ‘to be on the throne’
D tr[uː]n-u tr[uː]n-ům
L tr[uː]n-u tr[uː]n-ech
I tr[uː]n-em tr[uː]n-y
Even a glance at the table in (21a) reveals how disproportionately the short and
long forms of the root are distributed: the long form appears only in two (syncretic)
paradigm cells, namely in the nominative and accusative singular; the
short form, on the other hand, appears elsewhere, i.e., in the rest of the declensional
paradigm and also in all types of derivatives, including the ek-diminutive
(as is shown in the third column in (21a)). Given the rather limited, case-sensitive
distribution of the length, the length alternation d[uː]m ‘house’ – d[o]m-ek
‘small house’ can hardly be interpreted as involving ů-to-o shortening. On the
contrary, a much more plausible analysis is such, according to which the root is
lexically stored as short and it lengthens in the given case/number context.
If we take on the hypothesis that the length in the root /duːm/ is due to its
merger with the nominative/accusative case marking prosodic affix, we are then
in a position to unify three distributional observations that would have otherwise
remained unrelated. In particular, assuming that the length marks the nominative/accusative
case, it is no coincidence that: (i) there is correlation between
the long variant of the root and the absence of an overt case marker (the explanation:
the case is marked by the length itself); (ii) existence of a directional adverb
97
dom (e.g. Petr šel dom ‘Petr went home’) is not surprising any more (the adverb,
made up from a short bare root, does not carry the nominative feature, hence, it
does not merge with the prosodic affix, hence, it does not lengthen); (iii) there
is only one vowel-length alternating pattern in the masculine paradigm and the
contrast is between the nominative/accusative singular on the one hand and the
rest of the paradigm cells on the other (and not between, say, dative singular and
the other cases).
As far as I can see, the shortening in the masculine diminutive was “invented”
solely based on the length contrast between a particular inflectional form of
the root and its diminutive form. However, if distribution of length variants of
“shortening” roots is considered from a broader perspective, then, in fact, the
contrast lies between the nominative/accusative singular form and the rest of
the inflectional and derivational forms. Thus, there is no difference between the
short diminutive domek and the short diminutive broček ‘slug’ (discussed in the
previous section): in both, there is a lexically short root that spells out the whole
diminutive-stem tree (with the concatenated diminutive suffix -ek). The only distinction
is that the root ‘house’ and the other roots listed in (22) merge with a
prosodic affix in the nominative singular context, while the root ‘slug’ does not.
(22) ek-diminutive Gsg Nsg
dv[o]r-ek dv[o]r-a dv[uː]r ‘court’
hn[o]j-ek hn[o]j-e hn[uː]j ‘dung’
l[o]j-ek l[o]j-e l[uː]j ‘tallow’
st[o]l-ek st[o]l-u st[uː]l ‘table’
v[o]l-ek v[o]l-a v[uː]l ‘ox’
To complete the discussion of “shortening” masculines, there is yet another alternating
pattern illustrated in (23). The root ‘god’ alternates in length in the
denclension paradigm the same way as the roots above; however, it—contrary to
the stems above—has a long vowel also in the ek-diminutive, cf. b[uː]ž-ek.
(23) Sg Pl
N b[uː]h b[o]h-ové b[uː]ž-ek ‘little god’
A b[o]h-a b[o]h-y
G b[o]h-a b[o]h-ů b[o]h-yně ‘goddess’
D b[o]h-u b[o]h-ům b[o]ž-í ‘godly’
L b[o]h-u b[o]z-ích b[o]ž-stvo ‘deity’
I b[o]h-em b[o]h-y
98
The third column in the table above demonstrates that all derivatives include
a short root—except the diminutive. To explain this length contrast, I draw a
parallel between the diminutive bůžek and the diminutive krůček ‘step’, analyzed
previously. I assume that both are built on lexically short roots that merge with
the prosodic affix, and from this merger, a diminutive stem is created. Under this
analysis, the co-distribution of the length across the nominative singular and the
ek-diminutive in tables (23) and (24) is coincidental: it arises from the merger of
two lexically distinct prosodic affixes.
(24) ek-diminutive Gsg Nsg
d[uː]l-ek d[o]l-u d[uː]l ‘pit’
v[uː]z-ek v[o]z-u v[uː]z ‘cart’
hr[aː]š-ek hr[a]ch-u hr[aː]ch ‘pea’
mr[aː]z-ek mr[a]z-u mr[aː]z ‘frost’
l[iː]ž-ek l[i]h-u l[iː]h ‘alcohol’
sn[iː]ž-ek sn[e]h-u sn[iː]h ‘snow’
To conclude, the contrast between the bůžek-type roots in (24) and the domektype
roots in (22) boils down to the old contrast between roots that undergo
diminutive lengthening and those that do not: only the latter are capable to spell
out the whole diminutive stem, the former need help of a prosodic affix.
Let us now turn to “shortening” feminines: in certain respects, they resemble
“shortening” masculines—and yet, they differ from them. Again, the “shortening”
analysis is postulated based on comparison between a short diminutive
and a long nominative singular. However, as with masculines, short versions of
feminine roots are not restricted to the diminutive context, as illustrated in the
third column of table (25). As a matter of fact, the short form of the root ‘frog’ is
distributed across derivatives of various semantic types.
(25) Sg Pl
N ž[aː]b-a ž[aː]b-y ž[a]b-ka ‘small frog’
A ž[aː]b-u ž[aː]b-y ž[a]b-ák ‘male frog’
G ž[aː]b-y ž[a]b ž[a]b-inec ‘chickweed’
D ž[aː]b-ě ž[aː]b-ám ž[a]b-í ‘froggy’
L ž[aː]b-ě ž[aː]b-ách
I ž[aː]b-ou ž[aː]b-ami
99
To capture the distribution of the short forms—really wide-spread outside the
declension—I assume that it is the short version of the root ‘frog’ that is stored
in the lexicon, and that is why it can spell out the whole diminutive stem. Hence,
there is no diminutive shortening at all, neither in masculines discussed previously,
nor in the feminines like žába – žabka. But: despite their similarity, there
is still a crucial distinction between alternating masculines on the one hand and
alternating feminines on the other regarding length distribution in the declension.
Long-root variants are in the minority in the masculine declensional paradigm,
but they are prevalent in the feminine paradigm, as illustrated in the first
two columns in (25). To accommodate this generalization, I propose that all case
markers concatenate with a declensional stem, and it is this stem that involves a
lengthening morpheme for feminine nouns like ‘frog’.
The lengthening proposal builds on the idea introduced by Bittner & Hale
(1996) and Bayer, Bader & Meng (2001): they argue for a functional head K that
syntactically unifies various different morphological cases. Expressed in morphological
terms, the K-head defines a declensional stem to which particular
case markers are attached. Since the case forms in (25) share the long variant of
the root (which, moreover, does not occur outside the declensional paradigm), I
claim that the K-head is spelled out by a prosodic affix that triggers a root-vowel
lengthening. But: not all declensional forms in (25) have the lengthened root. In
particular, the root is short in the genitive plural form. Notice, incidentally (or
not), that the absence of root-vowel lengthening in genitive plural correlates with
the absence of an overt case marker. Viewed from the syllable-perspective, the
only declensional form with the short root in (25) is the one where the vowel appears
in a closed syllable; lengthened root vowels, on the other hand, are always
in open syllables. It thus indicates that the length alternation in the feminine
declensional paradigm is syllable-sensitive.
In a Strict CV model, used in this book, the contrast beween syllable-sensitive
and syllable-independent vowel lengthening is derived from the direction
of melody spreading. As I have already shown in chapter 2 (in sections 2.6 and
2.7, in particular), leftward spreading entails syllable-independent lengthening
and that actually means that the lengthened vowel can appear in both open and
closed syllables. The rightward spreading, on the other hand, derives long vowels
in open syllables. Technically speaking, a lexically short vowel can spread its
melody only to a licensed empty V-slot. If this slot precedes the particular vowel,
it is always licensed by it; hence, vowel lengthening is automatically triggered.
But in case the empty V-slot follows the short vowel, its licensing depends on its
100
right syllable context, i.e., whether there is another vowel to the right of the given
empty V-slot that could license it.
From what has been said it inevitably follows that the prosodic affix that, as
I claim, spells out the K-head shared by all cases (including the genitive plural),
must be of the CV-shape: only in this case, the empty V-slot (i.e., a target of
melody spreading), is inserted to the right of the root vowel (and not to its left).
The entry for the prosodic affix, which defines what I called declensional stem,
looks therefore as in (26).37
(26) </CV/, [KP
K]>
The figures in (27) show the derivation of the long and short versions of the declensional
stem; both cases involve the prosodic affix (marked by shading). The
derivation proceeds as follows. The declensional stem involves a merger of the
prosodic affix with the lexically short root (whose lexical form is in (27a)). The
CV-shape ensures that the prosodic affix is inserted after the root vowel (and not
before it). In the next step, the declensional stem is merged with a particular case
ending. Since all endings in the paradigm in (25) are vowel-initial, the empty
slot of the declensional stem is always licensed. As a consequence, root-vowel
lengthening is triggered, as illustrated by the nominative singular form ž[aː]b-a
in (27b). The genitive plural form in (27c) is expressed by a bare stem: it means
that its empty V-slot lacks a licensor. Being unlicensed, the root vowel therefore
does not spread and it surfaces as short.
37	 Note that only a closed subset of feminine roots shows the length contrast between
the genitive plural and the rest of the declensional forms. In the proposed model, the
difference between alternating (e.g. ž[aː]b-a – ž[a]b) and non-alternating roots (e.g.
sl[aː]v-a – sl[aː]v ‘glory’) is that the former need to merge with the prosodic affix to spell
out a declensional stem, while the latter are capable to spell it out themselves. From
this perspective, the diachronic change that makes alternating roots non-alternating
can be interpreted as the change in the lexicon. Specifically, the root accommodates
the K-feature into its lexical entry, hence it becomes capable to spell out the whole
declensional stem, i.e., without help of the prosodic affix. An example of such lexical
restoring might be a root ‘cart-load’ which actually shows a long vowel [uː] throughout
the whole paradigm, i.e., f[uː]r-a ‘Nsg’ – f[uː]r ‘Gpl’, but originally it had a short vowel
[o] in the genitive plural, i.e., f[uː]r-a – f[o]r.
101
(27) a. b. L
c.
C V C C V C V C - V C V C V C
● ● ● ● ● ●
ž a b
● ● ● ●
ž a b a ž a b
In sum, the proposed model describes the co-distribution of the short form /žab/
across the declension and the ek-diminutive, as not having a single source. While
it is true that in both cases the root surfaces with a short vowel, the reason is different.
Moreover, even the underlying structure of the short version of the root
is not the same: in derivatives, the short root matches the lexical root-form, in
the genitive plural form, the root is prosodically bigger; compare the underlying
forms of the lexical structure in (27a) and the genitive plural structure in (27c).
Yet, both surface with a short vowel [a].
The table below provides a couple of alternating feminine roots that follow the
same pattern as the root ‘frog’: they exhibit a short vowel in the genitive plural
and in the diminutive ek-form.38
(28) Nsg Gpl ek-diminutive
j[aː]m-a j[a]m j[a]m-k(-a) ‘hole’
kr[aː]v-a kr[a]v kr[a]v-k(-a) ‘cow’
r[aː]n-a r[a]n r[a]n-k(-a) ‘wound’
sk[aː]l-a sk[a]l sk[a]l-k(-a) ‘rock’
chv[iː]l-e chv[i]l chv[i]l-k(-a) ‘moment’
l[iː]p-a l[i]p l[i]p-k(-a) ‘linden’
ž[iː]l-a ž[i]l ž[i]l-k(-a) ‘vein’
38	 Note that some of these roots can appear short also in paradigm cells other than the genitive
plural. However, these non-genitive short forms always co-exist with long forms.
For example, the root ‘cow’ appears invariably short in the genitive plural (kr[a]v),
but it shows variation in length in other plural cells, i.e., in the dative (kr[aː]/[a]v-ám),
locative (kr[aː]/[a]v-ách) and instrumental (kr[aː]/[a]v-ami). I follow Scheer (2011),
Bethin (2003a) and Sukač (2013) who argue that the optional appearance of short
forms outside the genitive plural is a relict of diachrony and assume that these forms
are suppletive.
102
The pattern above has already been analyzed by Scheer (2004:621) in the Strict
CV framework. He assumes that the co-distribution of short root alternants in
genitive plural forms and ek-diminutives has a single source, namely the closed
syllable context. If we were to look only at the triplets in (28), we might be inclined
to share Scheer’s claim. However, if we take other forms of these roots into
account, it is clear that shortness of the root vowel does not always come from a
closed syllable context. First of all, the diminutive suffix itself is vowel-initial and
once it combines with the genitive plural form, the root vowel appears in an open
syllable. And yet, the vowel is still short (e.g. j[a].m-ek, kr[a].v-ek, r[a].n-ek, and
so on). Furthermore, short root-vowels appear in open syllables also in double
diminutives (e.g. kr[a].v-ič-k(-a) and chv[i].l-ič-k(-a)) or other types of derivatives
(e.g. kr[a].v-inec ‘cowpat’ or sk[a]l-isk(-o) ‘a piece of rock’).
Moreover, Scheer’s analysis, while it apparently unifies the behavior of length
in the genitive plural and the diminutive, does not account for the pattern illustrated
in (29). Here, only the genitive plural form is short and the ek-diminutive
is long. This pattern thus provides a strong piece of evidence for our proposal
that sees the vowel quantity in ek-diminutives and in the genitive plural as inde-
pendent.
(29) Nsg Gpl ek-diminutive
č[aː]r-a č[a]r č[aː]r-k(-a) ‘line’
dr[aː]h-a dr[a]h dr[aː]ž-k(-a) ‘track’
chv[aː]l-a chv[a]l chv[aː]l-k(-a) ‘praise’
v[aː]h-a v[a]h v[aː]ž-k(-a) ‘weight’
s[iː]l-a s[i]l s[iː]l-k(-a) ‘power’
Summing up, nouns showing a length contrast between the long nominative singular
and the short ek-diminutive are built on lexically short roots. Thus, there
is no productive diminutive shortening at all, instead, the only length-manipulating
process involved in diminutives is lengthening. The length is induced by
the insertion of the CV-affix which spells out a particular syntactic part of the
diminutive stem.
4.3.3  The Suffix -ek Is a Fully-Fledged Noun
The syntactic structure of the diminutive stem thus involves gender projections
dominated by the diminutive phrase XP. Let us now turn to the suffix -ek that
attaches to the diminutive stem. To define its syntactic structure, consider the
examples in (30). There are three ek-derivatives and neither of them carries the
103
diminutive meaning. Rather, the ek-suffix serves as a pure nominalizer: it turns
non-nominal bases (on the left) to fully-fledged nouns (on the right). Specifically,
it may turn a numeral to a number-denoting noun in (30a), a directional
adverb to a place noun (30b) or a preposition to a place-denoting noun (30c).
(30) a. pět ‘five’ pět-k(-a) ‘five’
(numeral) (number-denoting noun)
b. ven ‘out’ ven-ek ‘outside’
(adverb) (place noun)
c. před ‘in front of’ před-ek ‘the front’
(preposition) (place noun)
Furthermore, the suffix -ek often replaces nouns in multiword fixed expressions,
a process sometimes called univerbization (a trait typically colloquial). An example
of such an “univerbizing” function of the -ek is given below.
(31) průmysl-ov-á škol-a průmysl-ov-k-a ‘technical school’
industry-adj-agr school-nom.sg industry-adj-n-nom.sg
To capture the observations above, I propose that the suffix -ek is syntactically a
fully-fledged noun with a general meaning ‘thing’; that, in turn, means that its
syntactic structure involves a nominal-root phrase (NP) at the bottom. While it
might sound strange, if you think about it, the only irregularity is that this noun
is affixal in nature. In addition, recall that the idea that affixes can spell out even
the root projections is not completely new: it has already been proposed in the
literature (see section 3.4.2 for references) and I also used it when analyzing the
structure of hypocoristics. There is, however, a difference between the “root” -ek
on the one hand and the hypocorictic “root” -d on the other: the suffix -d derives
only masculine hypocoristics, as we have already seen, while ek-derivatives are
of different semantic types (and different genders as well). This contrast suggests
that the root phrase associated with -ek is semantically more general than the
one involved in the suffix -d.
Despite the difference between their suffixes, d-hypocoristics and ek-diminutives
are both bi-nominal syntactic constructions (i.e., compounds). The structure of
the ek-diminutive is depicted in (32). The suffix spells out a fully-fledged nominal
tree with the nominal-root phrase NP at the bottom and the diminutive phrase XP
104
on the top. The suffix tree is then adjoined to the diminutive-stem tree, yielding the
bi-nominal construction.39
(32)
               | |dim. stem <= XP XP =˃ dim. suffix
     | |       | |
X (FemP) X NP
           | |
(Fem) MascP
              | |
Masc NP
The bi-nominal analysis proposes that the suffix -ek itself is not associated with
any gender features—and that, in turn, captures the generalization that diminutives
“inherit” gender of their bases.40
Furthermore, from (32) it follows that the
diminutive meaning—encoded in the X-head—is present twice in the structure,
i.e., in both nominal trees. In the right-hand tree, it is spelled out by the suffix
-ek, while in the left-hand tree it is spelled out by a prosodic affix, which triggers
lengthening of the stem. This explains another generalization, namely that
ek-suffixation is accompanied by vowel lengthening only in diminutives. If -ek is
suffixed to a non-diminutive stem (and that actually means that a nominal tree
to its left is not dominated by the XP), no lengthening is triggred. To illustrate
the contrast between vowel-length behavior in ek-diminutives on the one hand
and other types of ek-derivatives on the other, consider the examples in (33). The
table contains three pairs of words, each pair sharing both the root and the suffix
-ek. All these roots alternate in length following the same pattern: they appear
39	 Bachrach & Wagner (2007) propose something similar for diminutives in Brazilian
Portuguese (within a Distributed Morphology framework). They claim that BP diminutives
are syntactic bi-nominals in which one of the nouns is spelled out by the root
morpheme and the other by the diminutive suffix /-ziɲ/.
40	 From this perspective, Czech differs from, say, German, in which diminutive suffixes
-chen and -lein turn masculines and feminines into neuters. Using a Distributed Morphology
framework, Wiltschko & Steriopolo (2007) interpret this contrast between
gender-changing and gender-transparent diminutive suffixes as stemming from their
position in the syntactic structure: the former are assumed to be syntactic heads, while
the latter are assumed to be modifiers of a categorizing head n (which carries gender
features). In the system proposed in this book, the difference between the two types
of diminutive affixes is encoded syntactically as well. In particular, syntactic structure
spelled out by the suffix either involves gender projections—or it does not.
105
long only in those ek-forms that carry diminutive meaning (in the right-hand
column of the table), otherwise they are short.
(33) ek-nominalization ek-diminutive
pod-hř[i]b-ek hř[iː]b-ek ‘species of boletus; small boletus’
ná-hr[o]b-ek hr[uː]b-ek ‘gravestone; small grave’
pod-v[o]z-ek v[uː]z-ek ‘undercarriage; small carriage’
4.3.4  The Morphosyntactic Hierarchy: Size over Affection
Since there is a single diminutive projection shared by the diminutive stem and
the diminutive suffix, all diminutives are expected to have the same properties.
However, this is obviously not true—neither in Czech, nor cross-linguistically.
However, there seems to be a kind of consensus in the literature (e.g. Dressler
& Merlini Barbaresi 1994, Juraffsky 1996, Bauer 1997, Grandi & Körtvélyessy
2015, among many others), that there are two core meanings associated with
diminutive forms, i.e., size and affection. Hence, there are at least two types of
diminutives.
Cinque (2015), using a cartographic approach, proposes that these two core
functions of diminutives are encoded syntactically as two separate projections.
I take on Cinque’s proposal and in what follows, I use the labels Aff(ective)P
and SmallP for the two projections. Of course, languages differ in how these
two projections are spelled out. Cinque (2015) shows that they are spelled out
by two formally distinct suffixes in Italian. For example, a diminutive cas-ett(-a)
expresses speaker’s affection toward a denoted house and cas-in(-a) refers to its
size. According to Cinque (2015), this semantic contrast between the two diminutive
forms—that share the same root—suggests that the suffix -in is associated
with SmallP syntactically, while -ett with AffP.
Czech, similarly to Italian, has two formally distinct diminutive suffixes, i.e.,
e-initial and i-initial. But unlike their Italian cousins, each of these two affixes
derives semantically ambiguous diminutives. In order to show that, let me start
from the fact that mass nouns are incompatible with size modifiers (e.g., *a big
air); for a mass noun to get modified by a size modifier, it would have to be first
packaged into units (of some sort), so that we can talk about size (of the packaging,
clearly). In this light, consider the sentences in (34) and (35). In (34), there
are two diminutives of mass nouns, i.e., vzdouš-ek ‘air’ in (34a) and rum-ík ‘rum’
in (34b), that are modified by a quantifier ‘a lot of’. The grammaticality of the
quantifier phrases indicates that these are mass nouns—and, thus, they should be
106
incompatible with size reading of the diminutive. Therefore, we must conclude
that neither the ek-form nor the ík-form in (34) refer to the size, but, rather, both
are used affectively.
(34) a. Je tady spousta čerstvého vzdoušku.
is here a_lot_of fresh air.dim
‘There is a lot of fesh air here.’
b. V láhvi je pořád spousta rumíku.
in bottle is still a_lot_of rum.dim
‘There is still a lot of rum in the bottle.’
To illustrate a size use of ek- and ík-forms, consider the sentence in (35). The
meaning indicates that we are dealing with the size reading of the diminutive.
Once again, this environment is grammatical with both an ek-form nehýtek and
ík-form nehtík—thus, both of them refer to a (small) size.
(35) Nehýtek/nehtík je nehet, který je malý.
nail.dim is nail that is little
‘Nail (diminutive) is a nail that is little.’
The examples above demonstrate that neither of the two diminutive suffixes is
restricted to spell out just AffP (as opposed to Italian, in which the suffix -ett
realizes AffP, but not SmallP). This, however, does not mean that all diminutive
forms in Czech are ambiguous between a size reading and an affection reading.
The generalization is this: diminutives (of either form) are either ambiguous
or they are affective-only. In other words, those forms that can refer to a size
can always be used affectively as well, but the opposite is not true. To show the
contrast between affective-only and ambiguous diminutive forms, compare the
examples above with those in (36) and (37). Here, the same diminutive forms as
above are used, but in the opposite contexts. In particular, the diminutive forms
of the noun ‘nail’ are modified by the adjective ‘big’ in (36) which enforces their
affective meaning.
(36) Musíme ustřihnout ten velký nehýtek/nehtík.
we.must cut_away that big nail.dim
‘We must cut that big nail (diminutive).’
By contrast, the diminutives vzdouš-ek and rum-ík do not express a size meaning,
as is testified by the ungrammaticality of the sentence below.
107
(37) *Vzdoušek/rumík je vzduch/rum, který je malý.
air.dim/rum.dim is air/rum that is small
‘*Air/rum (diminutive) is an air/rum that is small.’
Summing up, the diminutives nehýt-ek and neht-ík can express both a small size
(35) and affection (36), but the diminutives vzdouš-ek and rum-ík are affectiveonly
(see the contrast between (34) and (36)). And, in Czech, there are no diminutives
that express just size; affectiveness is always present.41
In a cartographic approach, the generalization above (the meaning of diminutive
forms is either size- or affection-related) can be explained by assuming
that the diminutive projections are organized in an implicational hierarchy:
SmallP>AffP. This hierarchy predicts that there are exactly two types of diminutive
forms: affective-only and ambiguous. The affective-only forms involve only
a lower diminutive projection, i.e., AffP. The ambiguity reading of diminutives,
then, follows from the spell-out: given the Superset Principle, the form that can
spell out the SmallP is also able to spell out the AffP.
The diminutive hierarchy is the same as proposed by Cinque (2015); Cinque,
however,does not consider it to be implicational. Cinque establishes the ordering
of the diminutive features on the basis of the ordering of the diminutive suffixes
in Italian (and in other languages as well). In particular, he claims that the affective
suffix -ett (mentioned above) always precedes the size suffix -in. In Italian,
thus, there are double diminutives like cas-ett-in(-a) ‘small cosy house’—but not
like *cas-in-ett(-a). For Cinque, this ordering constitutes his main argument for
the feature hierarchy Small>Aff. In Cinque’s system, each simple diminutive
41	 The both affective-only diminutives discussed here are derived from mass nouns, denoting
homogenous substances. This, however, does not automatically mean that all diminutive
forms of mass nouns are affective-only. On the contrary, De Belder (2011) and
De Belder et al. (2014) show that in many languages diminutive forms of mass nouns
are typically count, which means that they refer to a size. It holds also for Czech: for
example, diminutives like chleb-ík or čokolád-k(-a), derived from mass nouns ‘bread’
and ‘chocolate’, may denote a small piece of bread or chocolate. However, despite being
able to denote a small-size units, the diminutives chleb-ík and čokolád-k(-a) can also
be modified by a quantifier ‘a lot of’, i.e., spousta chlebíku/čokoládky, which indicates
their affective use (and the mass nature as well). The behavior of these forms thus corresponds
to the generalization above: diminutives derived from mass nouns are either
affective-only or ambiguous. The former retain mass properties of their bases while
the latter alternate between count (in the size use) and mass (in the affective use).
108
(i.e., cas-ett(-a) and cas-in(-a)) involves a single projection encoding the given
diminutive meaning, i.e., either SmallP or AffP.
However, Cinque (2015:71) also notes that there is a formal ambiguity between
the size and affection readings. In particular, he points out that simple
diminutives derived by the suffix -in refer not only to the size, but they can be
used affectively as well. There is thus a contrast between the suffix -ett on the one
hand: it derives only affective diminutives—as opposed to the suffix -in: it derives
both affective and size diminutives. This generalization follows automatically, if
the hierarchy Small>Aff is assumed to be implicational (as I do here). The implicational
relationship between the SmallP and AffP predicts the existence of
ambiguous suffixes such as -in and affective-only suffixes such as -ett.42
Having established the diminutive feature hierarchy let me turn back to the
bi-nominal diminutive proposal introduced in the previous section. Recall that
I have established a single diminutive projection, i.e., XP, shared between the
diminutive stem and the diminutive suffix. In what follows, I propose that the
projection dominating the diminutive stem is in fact distinct from the projections
involved in the diminutive suffix. Specifically, I assume that SmallP and
AffP replace the old XP only in the diminutive-suffix syntactic tree—but not in
the diminutive-stem tree. The proposal builds on the idea introduced in section
4.3.2: I argued for existence of a declensional stem, to which particular case endings
are concatenated. The crucial point now is that the declensional stem is defined
syntactically, concretely by the projection labelled as KP. KP thus serves as
a syntactic “umbrella point” of all projections encoding particular cases. Applying
the similar logic to diminutives, I assume that the XP defining the diminutive
stem, is the “umbrella point” of the projections encoding particular diminutive
meanings. The proposal is depicted in (38).43
The figure in (38a) represents a syntactic structure of a diminutive with the
affective meaning: the suffix tree (on the right in (38a)) involves just AffP. In
(38b), there is a syntactic structure of a diminutive referring to the size of the
denotee: it involves the whole diminutive subtree, i.e., [SmallP
Small [AffP
Aff]].
42	 Steriopolo (2008) examines semantic and syntactic properties of more than 30 expressive
suffixes in Russian and she identifies exactly those two types predicted by the
proposed hierarchy, i.e., suffixes expressing only speaker’s attitude and suffixes that
“convey an attitude and refer to the size of the referent” (p. 3).
43	 I label the stem projection simply as XP, leaving aside its exact semantic content. However,
if we adopt the idea that diminutives always involve speaker’s judgment—the
speaker considers the denotee either small or endearing—, then this just might be the
semantic content of the XP. I leave this issue open for further research.
109
(38) a.          | |
stem <= XP AffP =˃ suffix
     | |          | |
X (FemP) Aff NP
       | |
(Fem) MascP
          | |
Masc NP
b.          | |
stem <= XP SmallP =˃ suffix
     | |           | |
X (FemP) Small AffP
       | |              | |
(Fem) MascP Aff NP
          | |
Masc NP
From the perspective of the structures above, ambiguous diminutives in fact differ
syntactically. In other words, if a particular form expresses affection, its syntactic
structure is as in (38a). And if the same form is used to denote the small
size of the denotee, it has the structure in (38b). The ambiguity of diminutives
like nehýt-ek thus arises from the spell-out, concretely from the Superset Principle.
Given this principle, the suffix -ek (lexical entry for which is in (39)), can
spell out both right-hand trees of the binominal structures above: the lexical tree
of the suffix in (39) matches the right-hand tree in (38b) and it involves the tree
in (38a). As a consequence, syncretic ek-diminutives are derived.44
(39) </-ek/, [SMALLP
Small [AFFP
Aff [NP
NP]]]>
The proposed analysis makes a prediction concerning the distribution of vowel
length. Since both types of diminutives involve the stem dominated by the XP,
they should not differ with respect to the vowel length. The reason is that the
length pattern is defined within the XP, i.e., independently of the suffix, which
spells out the particular diminutive meaning. This prediction is confirmed by
diminutives that are syncretic between both meanings. In fact, there exist no
minimal pairs like l[a]b-ek – l[aː]b-ek in which the vowel length would differentiate
between the size meaning and the affective meaning. (Recall that there
exist ek-diminutives that show the length variation, e.g. a noun ‘robot’ has either
a lengthened (rob[uː]t-ek) or a non-lengthened diminutive form (rob[o]t-ek).
However, as far as I know, the length variation does not correlate with the meaning
distinction. Hence, rob[uː]t-ek can denote both ‘a small robot’ and ‘a dear
robot’ and the same holds for the short form rob[o]t-ek.)
44	 Given the Superset Principle, the suffix -ek can potentially spell out just the NP, i.e., the
most embedded sub-constituent in the lexical tree in (39). And it really does: remember
that -ek serves also as a pure nominalizer which indicates that it is used to spell out just
a nominal-root projection.
110
4.3.5  Decomposing the Suffix -ík
In the previous section, I proposed that the size-affection syncretism of ek-diminutives
follows from the spell-out, concretely from the fact that the suffix -ek
is able to spell out the whole diminutive subtree. Since the same type of syncretism
appears also in ík-diminutives, I argue that both ek-forms and ík-forms
involve a single diminutive suffix, i.e., the -ek.
The idea that there is only one suffix shared by formally distinct diminutives
might seem to be in conflict with my previous claim that the suffix -ík cannot be
just a lengthened version of the suffix -ek; the arguments were presented in sections
4.2.2 and 4.2.3. However, the two claims are not conflicting if we assume
that the suffix surfacing as [iːk] is not a morphological atom. In other words, I
propose that there is not a single lexical item corresponding to the form [iːk];
rather, there are three independent items that put together phonologically give
the form [iːk]. Two of them we have already established: the VC-affix and the eksuffix.
The new ingredient is a high-vowel suffix -i.
The main benefit of the three-item proposal is that it unifies both semantic and
phonological properties of both types of diminutives. The proposed decomposition
of the suffix -ík enables us to explain length alternations in ek-diminutives
and ík-diminutives in a uniform way. Recall that the contrast between masculine
ek-forms that productively lengthen their bases and feminine ek-forms that do
not, results from the different spell-out of their stems: feminines are able to spell
out the whole diminutive stem dominated by the XP, masculines, on the other
hand, need help of the prosodic affix (whose lexical entry is repeated in (40)
below). Since also ík-diminutives differ in length between the masculine and the
feminine gender, it is reasonable to assume that both length alternations have
the same logic.
(40) </VC/, [XP
X [MascP
Masc]]>
The difference between ek-deminutives and ík-deminutives is that in ík-diminutives,
it is the high-vowel suffix itself which alternates in length: it is long in masculines
(e.g. les-[iː]ček ‘forest’) but short in feminines (e.g. kas-[i]čk(-a) ‘money
box’). This, thus, suggests that the i-suffix spells out a projection between the
nominal root and the masculine feature that is realized by the prosodic affix. In
(41), this newly introduced projection is labelled as YP and it is spelled out by the
suffix /-i/ in diminutive stems of both genders. The difference is that in a masculine
stem in (41a), the i-suffix realizes just the YP; the rest of the projections,
i.e., the masculine gender projection MascP dominated by the diminutive-stem
projection XP, are then realized by the prosodic affix. As a consequence, the
111
prosodic affix merges with the /-i/ and causes its lengthening. The prosodic affix
cannot realize the stem projection in feminines because in this case the XP
dominates FemP, as shown in (41b). This explains why neither ek-feminines nor
ík-feminines exhibit lengthening. However, the XP needs to be spelled out even
in feminines. Since in ek-feminines it is spelled out by the root itself, in their
ík-cousins, in which roots are merged with the -i, the XP must be spelled out by
this high-vowel suffix.
(41) a. XP =˃ /VC/
     | |
X MascP
        | |
Masc YP =˃ /-i/
           | |
Y NP =˃ root
b. XP =˃ /-i/
     | |
X FemP
        | |
Fem MascP
           | |
Masc YP
              | |
Y NP =˃ root
The lexical entry for the i-suffix thus looks as in (42): its lexical tree bottoms
at YP and ends with XP. Viewed from the spell-out perspective, the difference
between masculines and feminines is that the latter use the maximal and the
former the minimal spell-out extent of the item in (42).
(42) </-i/, [XP
X [FemP
Fem [MascP
Masc [YP
Y]]]]>
The third formal ingredient of ík-diminutives is the suffix -ek, whose lexical entry
is repeated in (43).
(43) </-ek/, [SmallP
Small [AffP
Aff [NP
NP]]]>
Summing up, both formal types of diminutives are bi-nominal structures, in
which the right-hand noun is spelled out by the suffix -ek. The difference is in
the left-hand noun, i.e., in the stem. The stem of ík-diminutives is syntactically
more complex than that of ek-diminutives. In particular, there is a feature Y that
is merged with the nominal-root projection in ík-diminutives and it is realized
by a high-vowel suffix -i.
Let me now show that the proposed decomposition of the suffix -ík is plausible
phonologically. How does it happen that the three lexical forms, i.e., /-i/,
/VC/ and /-ek/, merge into a single form, surfacing as [iːk]? The phonological
merger of the first two forms—which occurs in the stem—is depicted in (44).
First, the high-vowel suffix /-i/ concatenates with the nominal root. This first step
112
of the phonological derivation is shown in (44a): technically speaking, the V-slot
associated lexically with an i-segment is simply aligned after the last C-slot of
the root. In the next step, the VC-affix is inserted to the left of the i-suffix, which
is indicated by shadowing in (44b). The empty V-slot of the prosodic affix then
automatically accommodates the vowel of the suffix—and that vowel, in turn, is
spelled out long. To illustrate the two just described steps, there is the root /les/
‘forest’ and it enters the derivation: in the first step, the form /lesi/ is derived and
it turns into /lesiː/ in the second step.
(44) a. ... C - V b. ... C V C V
| | | | |
● ● ● ● ●
| | | |
C i C i
Subsequently, the lengthened i-stem is merged with the diminutive suffix -ek.
I have already established that this suffix involves a floating vowel (see section
4.2.1 for the relevant arguments): this, then, means that the e is not provided
with its own prosodic slot. When the suffix is concatenated with a vowel-final
morpheme, there is no empty slot for the floating e. Since being associated with
a V-slot is a necessary condition for the floating vowel to be phonetically realized,
the [e] is not heard after the lengthened i-stem. This is the reason why
the phonological merger of /lesiː/ and /-ek/ produces /lesiːk/. The phonological
merger of the i-stem and the ek-suffix is illustrated in (45).
(45) ... C V C V - C
| | | |
● ● ● ●
| | |
C i e k
Finally, the scenario in (46) illustrates what happens when the ek-suffix is
merged with a consonant-final stem such as, for example, /daːr/ ‘gift’. In this
case, the merger produces a linear string of two adjacent C-slots: the first
one belongs to the stem and the second one to the suffix. However, prosodic
structure is well-formed only if C-slots alternate regularly with V-slots. The
ill-formed CC-string (the result of a merger of the C-final stem and the diminutive
suffix) thus has to be repaired. As shown in (46), it is repaired by
113
insertion of an empty V-slot—and it is this empty V-slot that provides a potential
landing site for the floating vowel of the suffix. At this moment, I leave
aside the exact mechanism how this newly inserted V-slot accommodates a
melody of the suffix. At this point, the only relevant thing is that there is a phonological
contrast between vowel-final diminutive stems (i.e., those involving
i) and consonant-final stems: it is only in the latter that the floating vowel of the
diminutive suffix has an opportunity to surface.45
(46) ... V C V C - V C
| | | |
● ● ● ●
| | |
V C e k
I have just shown that the form [iːk] is created from three independent lexical
pieces. Let me provide further arguments for the proposed analysis. The first
argument is that ík-forms can express the same diminutive meanings as ek-forms.
As we have already seen, both the ek-form nehýtek and the ík-form nehtík (derived
from the noun ‘nail’) can each be either size- or affection-type diminutive.
Since both involve the suffix -ek (and, remember, -ek realizes certain diminutive
features), existence of such diminutive synonyms is predicted—hence, expected.
Another argument for the particular decomposition of the suffix -ík comes
from the data in (47). The table shows four pairs of diminutives, in each of them,
the high vowel is separated from the k-final suffix by another morpheme, in
this case by a consonantal suffix -n. Given the separation, the data thus support
the proposal that the high vowel and the [k] belong to two independent lexical
items. Moreover, notice that the high vowel alternates in length (it is long in
masculines, but short in feminines) even though it is not adjacent to the k-segment.
The fact that the alternation follows the same pattern as in ík-diminutives
suggests that there is a single suffix -i shared by ík-diminutives (like tatík) and
ínek-diminutives (like tatínek).
45	 The proposed scenario, according to which phonology inserts an empty V-slot for the
sake of repairing the prosodic structure, is not in line with the mainstream of the Strict
CV theory—but I do not address this issue in this book more than this brief note. For
an alternative analysis of the phonological structure of the suffix -ek, see Scheer &
Ziková (2010a).
114
(47) tat-[iː]-n-ek mam-[i]-n-k(-a) ‘father, mother’
strýč-[iː]-n-ek tet-[i]-n-k(-a) ‘uncle, aunt’
brouč-[iː]-n-ek tyč-[i]-n-k(-a) ‘beetle, lath’
ptáč-[iː]-n-ek sleč-[i]-n-k(-a) ‘bird, young lady’
Summing up: ík-diminutives and ek-diminutives are both syntactic bi nominals,
in which the right-hand noun is spelled out by the diminutive suffix. This suffix
is the same for both formal types. The left-hand noun represents a diminutive
stem with the feature Y spelled by a high-vowel suffix in ík-diminutives. For now,
I remain agnostic as for the precise semantic content of this feature; what is important
is that it distinguishes between the diminutive forms that are otherwise
synonymous.
4.4  Double Diminutives
We have already seen double diminutives in Czech. They end in eček and íček for
masculines and ečk(-a) and ičk(-a) for feminines. The table below summarizes all
the possible types both for masculines (48a-b) and feminines (48c-d).46
(48) a. dár-eček b. les-íček c. čár-ečk(-a) d. ruč-ičk(-a)
‘gift’ ‘forest’ ‘line’ ‘arm’
Given that the diminutive meaning is decomposed into two syntactic projections,
i.e., SmallP and AffP, each with a slightly different function, the null hypothesis
is that each of them is spelled out by a separate suffix, leading to the
double-diminutive forms. This is what is proposed by Cinque (2015): in double
diminutives, like in the Italian cas-ett-in(-a), the inner diminutive suffix, i.e., -ett,
spells out the lower projection AffP and the outer suffix, i.e., -in, spells out the
higher SmallP. In Cinque’s system, it means that double diminutives have an intersective
reading; thus, in the case above, the double diminutive casettina refers
to a house that is both endearing and small.
The intersective interpretation, however, does not work for Czech. In (49)
and (50), I repeat the sentences used previously to demonstrate that there are
mass diminutives like vzdouš-ek ‘air’ or rum-ík ‘rum’ that do not refer to a size.
46	 Since there are ík-diminutives—like tatínek ‘father’ and maminka ‘mother’, which involve
the suffix -n intervening between the high vowel and the diminutive suffix -ek
(as I have demonstrated at the end of the previous section), double diminutives can
actually also end in íneček or inečk(-a), i.e., tat-íneček or mam-inečka.
115
The use of double diminutive forms such as vzdouš-eček or rum-íček ‘rum’ does
not change the (un)grammaticality of the sentences below: thus, neither they do
have the size reading.
(49) a. Je tady spousta čerstvého vzdoušku/vzdoušečku.
is here a_lot_of fresh air.dim/air.dim.dim
‘There is a lot of fresh air here.’
b. V láhvi je pořád spousta rumíku/rumíčku.
in bottle is still a_lot_of rum.dim/rum.dim.dim
‘There is still a lot of rum in the bottle.’
(50) a. *Vzdoušek/rumík je vzduch/rum, který je malý.
air.dim/rum.dim is air/rum that is small
‘*Air/rum (diminutive) is an air/rum that is small.’
b. *Vzdoušeček/rumíček je vzduch/rum, který je malý.
air.dim.dim/rum.dim.dim is air/rum that is small
‘*Air/rum (double diminutive) is an air/rum that is small.’
Summing up, double diminutives in Czech are not interpreted intersectively—
rather, they have a degree reading: vzdoušeček or rumíček express a high degree
of speaker’s affection. And similarly, a simple ek-diminutive dárek ‘gift’, which
is syncretic between the size and affection, has a double diminutive cousin
dáreček—and that denotes either ‘a very small gift’ or ‘a very endearing gift’. And
the same holds for syncretic double diminutives ending in íček (rather than eček):
bratříček denotes either ‘a very little brother’ or ‘a very endearing brother’.
Consider now the non-diminutive forms in (51) which also have degree
interpretation. As we can see, what these forms have in common is that they
involve reduplication of a portion of their parts (the reduplicated strings are underlined).
In particular, adjectives in (51a) involve reduplication of the suffix,
nouns in (51b) reduplication of the prefix, and in (51c,d) the root morphemes
are reduplicated. The conclusion thus might be that the degree meaning can be
expressed by reduplication in Czech.
(51) a. Velikanánský dlouhatanánský vysokanánský
‘very big’ ‘very long’ ‘very tall’
b. prapravnuk praprababička praprarodič
‘great-great-grandson’ ‘great-great-grandmother’ ‘great-great-parent’
116
c. hrůza hrůz chyba chyb
‘absolute horror’ ‘absolute error’
d. div divoucí tma tmoucí leta letoucí
‘great miracle’ ‘absolute darkness’ ‘very long time’
In the light of the examples above, I assume that the double diminutives (that
carry the degree meaning as well) involve reduplication—exactly as the examples
in (51). In particular, I propose a Deg(ree)-head that is present in the structures
in (51) – and in double diminutives; this head is realized by reduplication.
Traditionally, the term reduplication refers to the doubling of a phonological
structure, where the reduplicants have a fixed prosodic shape. From this
perspective, however, the doubling of the degree-expressing forms in (51) can
hardly be interpreted as arising from such a reduplication: it is obvious that the
reduplicated pieces of the phonological structure do not have any uniform prosodic
shape. Fortunately, Inkelas & Zoll (2005) argue that there are two types of
reduplication and that these two types differ fundamentally regarding the nature
of the linguistic material that is copied in the process. Phonological reduplication
typically copies either syllables or feet. Consequently, such reduplication
follows prosodic constituency and that, in turn, means that reduplicants have
a fixed prosodic shape. The second type of reduplication is morphological, in
which morphosyntactic constituents are copied, irrespective of their prosodic
structure.
The morphological reduplication, as proposed in Inkelas & Zoll (2005), explains
the formal variability of the examples in (51) on the one hand and their
uniform semantic interpretation on the other. Put simply, we can claim that the
merger of the Deg-head, shared by the forms above, triggers copying of its sister
node (together with all other nodes dominated by it). In that case, both the sister
of the Deg-head and its syntactic copy can be spelled out by the same lexical form.
Figure (52) shows an implementation of this reduplicative scenario in the diminutive
context. Since reduplication targets the sister node of the Deg-head, a
whole subtree dominated either by SmallP or by AffP is copied into spec DegP
(depending on whether the given diminutive expresses size or affection).
117
(52)            | |
XP DegP
      | |      | |
(SmallP) DegP
                    | |
(Small) AffP Deg (SmallP)
                       | |
Aff NP (Small) AffP
                           | |
Aff NP
Having the same syntactic structure, both the source tree, i.e., a complement of
DegP, and its syntactic copy in the specifier of DegP (marked by dashed lines
in (52)) are expected to be spelled out by the same lexical item. How does this
expectation correspond to the generalization on the form of double diminutives
mentioned at the start of this section and repeated in (53)?
(53) Double diminutives end in eček or íček, but never in *ečík or *íčík.
At the first sight, the reduplicative structure in (52) excludes the existence of diminutives
like les-íček ‘forest’. That is because the traditional decomposition postulates
the inner diminutive suffix -ík, which is not a copy of the outer suffix -ek.
However, this problem disappears once we adopt the proposal that ík involves
the diminutive suffix -ek, which surfaces as [iːk] due to its merger with the i-final
stem (as I have established in the previous sections). Under this analysis, the
reduplicative scenario in (52) correctly predicts that all double diminutives are
ek-final. In other words, there is only one diminutive suffix -ek and it is this
unique suffix which is doubled when dominated by DegP. When the reduplicated
string /-ek-ek/, underlying all double diminutives, is merged with the i-initial
stem, it surfaces with the initial high vowel, as in the masculine les-íček ‘forest’
or the feminine ruč-ičk(-a) ‘arm’. The very same phonological string otherwise
surfaces as e-initial, cf. dár-eček ‘gift’ or čár-ečk(-a) ‘line’.
Moreover, the reduplicative account of double diminutives, assuming the decomposition
of the “suffix” ík, brings a new perspective on the so-called “diminutive
gaps.” The “diminutive gaps” occur when double-diminutive forms lack their
simple-diminutive cousins. A couple of examples is given in the table below. (54a)
shows the contrast between the masculines ‘gift’ and ‘eden’: the former produces
both diminutive forms, while the latter has only the double one, ending in eček. In
(54b), the same pattern is repeated with the masculine ‘friend’ which only has the
double diminutive form in íček. The most wide-spread type of this pattern is in
(54c), in which the only diminutive form of the feminine noun ‘bath’ ends ičk(-a).
118
(54) a. dar dár-ek dár-eček ‘gift’
ráj *ráj-ek ráj-eček ‘eden’
b. kostel kostel-ík kostel-íček ‘church’
přítel *přítel-ík přítel-íček ‘friend’
c. van(-a) *van-ik(-a) van-ičk(-a) ‘bath’
The pattern above provides the support for the claim that double-diminutive are
reduplicative structures of the type in (52). Since double-diminutives arise from
reduplication triggered by the merger of the DegP with the diminutive-suffix
tree, which, in turn, concatenates to the diminutive stem, then the contrast between
nouns like ‘gift’ and ‘eden’ is the contrast between diminutive stems that
combine with diminutive-suffix trees of different sizes and those that combine
only with trees dominated by DegP. Of course, one could ask why some stems
require DegP to create a diminutive. I leave the answer to this question open for
further research. Note, however, that the “diminutive gaps” such as those illustrated
in (54) pose a problem for traditional analyses in which double diminutives
are built incrementally from simple diminutives.
4.5 Summary
I argued that formation of diminutives in Czech involves a prosodic affix (of
the VC-shape) which is responsible for vowel-length alternations in both
ek-diminutives and ík-diminutives as well. To capture a robust correlation between
vowel-length alternations and the gender features, I proposed that the
VC-affix is involved only in masculine diminutive stems; that is, lengthening is
an active process only in masculine diminutives.
119
5. Conclusion
In this book, I have presented an approach where vowel length is connected with
the syntactic structure. The main idea was that there are nominal functional projections
in contemporary Czech which are spelled out by vowel length, or to be
more precise, by various types of prosodic affixes which give rise to different types
of vowel lengthening. I have argued that such prosodic affixes are the realizations
of syntactic parts of hypocoristics and diminutives. Furthermore, I have proposed
that length alternations in declensional paradigms such as dům – dom-u
or žáb-a – žab, that are traditionally taken to be a relic of diachrony, can be
analyzed as resulting from prosodic affixes that spell out particular inflectional
projections.
The present approach is strictly decompositional on both the phonological
and the syntactic side. In particular, I have proposed that length is to be represented
separately from the melodic content of vowels. Such a dissociation of
length from melody allows for the existence of lengthening morphemes that lack
any melodic features. I have used such lengthening morphemes extensively in
my analysis of various types of phenomena where lengthening is triggered by a
particular morphosyntactic environment.
Furthermore, if short vowels are made up from privative elements, which are
hierarchically organized, then these elements can be manipulated individually
under lengthening and as a consequence, length alternations of different types
can be derived.
Once lengthening is understood in this particular way, namely as an autosegmental
marking of a particular morphosyntactic category, it provides us with a
new tool that can be used to probe deeper into the morphological structure of
various categories where it is found. Specifically, following this approach, I was
led to decompose morphological categories such as diminutives and hypocoristics
into several hierarchically ordered syntactic projections. Some of these
projections are realized by “regular” affixes like a diminutive -ek or a hypocoristic
-d, while others are only revealed by prosodic affixes whose merger triggers the
lengthening of adjacent morphemes.
I believe that such an integrated approach which takes into account both phonological
and syntactic side of vowel-length alternations, enables us to understand
better some of the vowel length patterns we find in contemporary Czech.
121
References
Acquaviva, Paolo. 2009. Roots, Categories, and Nominal Concepts. Lingue e linguaggio
8(1), 25–51.
Akasaka, Yukiko & Tateishi, Koichi. 2001. Heaviness in Interfaces. In J. van der
Weijer & T. Nishihara (eds.), Issues in Japanese Phonology and Morphology.
Berlin – New York: Mouton de Gruyter. 3–46.
Alber, Bridgit & Arndt-Lappe, Sabine. 2012. Templatic and Subtractive Truncation.
In J. Trommer (ed.), The Morphology and Phonology of Exponence. Oxford:
Oxford University Press. 289–325.
Anderson, John & Ewen, Colin. 1987. Principles of Dependency Phonology. Cambridge:
Cambridge University Press.
Anderson, Stephen & Browne, Wayles. 1973. On Keeping Exchange Rules in
Czech. Papers in Linguistics 6, 445–482.
Backley, Phillip. 2011. An Introduction to Element Theory. Edinburgh: Edinburgh
University Press.
Bachrach, Asaf & Wagner, Michael. 2007. Syntactically Driven Cyclicity vs. Output-Output
Correspondence: The Case of Adjunction in Diminutive Morphology.
UPenn Working Papers in Linguistics 10(1), 157–173.
Bauer, Laurie. 1997. Evaluative Morphology: A Search for Universals. Studies in
Language 21, 533–575.
Bayer, Josef, Bader, Markus & Meng, Michael. 2001. Morphological Underspecification
Meets Oblique Case: Syntactic and Processing Effects in German.
Lingua 111, 465–514.
Bethin, Christina. 2003. Metrical Quantity in Czech: Evidence from Hypocoristics.
In W. Browne et al. (eds.), Formal Approaches to Slavic Linguistics 11: The
Amherst Meeting. Ann Arbor: MSP. 63–82.
Bethin, Chritina. 2003a. Prosodic Effects in Czech Morphology. In R. Maquire &
A. Timberlake (eds.), American Contributions to the Thirteenth International
Congress of Slavists. Ljubljana: Slavica Publications. 9–22.
Bičan, Aleš. 2015. Distribution of vocalic quantity in Czech. Grazer Linguistische
Studien 83, 133–138.
Bittner, Maria & Hale, Ken. 1996. The Structural Determantion of Case and
Agreement. Linguistic Inquiry 27, 1–68.
Bye, Patrick & Svenonius, Peter. 2012. Non-Concatenative Morphology as Epiphenomenon.
In J. Trommer (ed.), The Morphology and Phonology of Exponence.
Oxford: Oxford University Press. 427–495.
122
Caha, Pavel & Scheer, Tobias. 2008. The Syntax and Phonology of Czech Templatic
Morphology. In A. Antonenko, J. F. Bailyn & Ch. Bethin (eds.), Proceedings
of the Annual Workshop on Formal Approaches to Slavic Linguistics 16.
Ann Arbor: Michigan Slavic Publications. 68–83.
Caha, Pavel & Ziková, Markéta. 2016. Vowel Length as Evidence for a Distinction
Between Free and Bound Prefixes in Czech. Acta Linguistica Hungarica
63(3), 331–377.
Caha, Pavel. 2009. The Nanosyntax of Case. Doctoral dissertation, CASTL,
Tromsø.
Charette, Monik & Göksel, Asli. 1996. Licensing Constraints and Vowel Harmony
in Turkic Languages. SOAS Working Papers in Linguistics and Phonetics
6, 1–25.
Cinque, Guglielmo & Rizzi, Luigi. 2009. The Cartography of Syntactic Structures.
In B. Heine & H. Narrong (eds.), The Oxford Handbook of Linguistic
Analysis. Oxford: Oxford University Press. 65–78.
Cinque, Guglielmo. 2015. Augmentative, Pejorative, Diminutive and Endearing
Heads. In E. Di Domenico et al. (eds.), Structures, Strategies and Beyond: Studies
in Honour of Adriana Belletti. Amsterdam: John Benjamins. 67–82.
Cyran, Eugeniusz. 2010. The Phonetics, Phonology and Morphophonology of
Secondary Imperfectives in Polish. In E. Cyran & B. Szymanek (eds.), Verb
Structures. Between Phonology and Morphosyntax, Studies in Linguistics and
Methodology 2. Lublin: Wydawnictwo KUL. 111–138.
Davis, Stuart & Ueda, Isao. 2002. Mora Augmentation Processes in Japanese.
Journal of Japanese Linguistics 18, 1–23.
De Belder, Marijke, Noam, Faust & Lampitelli, Nicola. 2014. On a Low and a
High Diminutive. In A. Alexiadou et al. (eds.), The Syntax of Roots and the
Roots of Syntax. Oxford: Oxford University Press. 149–163.
De Belder, Marijke. 2011. Roots and Affixes, Eliminating Lexical Categories from
Syntax. Utrecht: LOT.
Dékány, Éva. 2011. A Profile of the Hungarian DP. The Interaction of Lexicalization,
Agreement and Linearization with the Functional Sequence. Doctoral
dissertation, CASTL, Tromsø.
Dressler, Wolfgang & Merlini Barbaresi, Lavinia. 1994. Morphopragmatics: Diminutives
and Intesifiers in Italian, German, and Other Languages. Berlin:
Mouton de Gruyter.
Ewen, Colin. 1995. Dependency Relations in Phonology. In J. A. Goldsmith
(ed.), A Handbook of Phonological Theory. Oxford: Blackwell. 570–585.
Faust, Noam & Torres-Tamarit, Francesc. 2017. Stress and Final /n/ Deletion in
Catalan: Combining Strict CV and OT. Glossa 2(1), 63.
123
Goldsmith, John. 1976. Autosegmental Phonology. Doctoral dissertation, MIT,
Cambridge.
Grandi, Nicola & Körtvélyessy, Lívia. 2015. Edinburgh Hanbook of Evaluative
Morphology. Edinburgh: Edinburgh University Press.
Gussmann, Edmund. 2007. The Phonology of Polish. Oxford: Oxford University
Press.
Harley, Heidi & Ritter, Elisabeth. 2002. Person and Number in Pronouns: A Feature-Geometric
Analysis. Language 78(3), 482–526.
Harley, Heidi. 2014. On the Identity of Roots. Theoretical Linguistics 40(3), 225–
276.
Harris, John & Lindsey, Geoff. 1995. The Elements of Phonological Representation.
In J. Durand & F. Katamba (eds.), Frontiers of Phonology: Atoms, Structures,
Derivations. Harlow: Longman. 34–79.
Harris, John. 1990. Segmental Complexity and Phonological Government. Phonology
7, 255–300.
Harris, John. 1997. Licensing Inheritance: An Integrated Theory of Neutralisation.
Phonology 14, 315–370.
Hayes, Bruce. 1989. Compensatory Lengthening in Moraic Phonology. Linguistic
Inquiry 20, 253–306.
Hayes, Bruce. 1995. Metrical Stress Theory: Principles and Case Studies. Chicago:
Chicago University Press.
Hyman, Larry. 2008. Universals in Phonology. The Linguistic Review 25, 83–137.
Hyman, Larry & Inkelas, Sharon. 1997. Emergent Templates: The Unusual Case
of Tiene. University of Maryland Working Papers in Linguisticy 5, 92–116.
Inkelas, Sharon & Zoll, Cheryl. 2005. Reduplication: Doubling in Morphology.
Cambridge: Cambridge University Press.
Iosad, Pavel. 2012. Vowel Reduction in Russian. Journal of Linguistics 48(3),
521–571.
John, Paul. 2014. The Phonology of Empty Categories. Doctoral dissertation,
University of Montreal, Montreal.
Juraffsky, Daniel. 1996. Universal Tendencies in the Semantics of the Diminutive.
Language 72, 533–578.
Kaye, Jonathan, Lowenstamm, Jean & Vergnaud, Jean-Roger. 1985. The Internal
Structure of Phonological Representations: A Theory of Charm and Government.
Phonology Yearbook 2, 305–328.
Kaye, Jonathan, Lowenstamm, Jean & Vergnaud, Jean-Roger. 1990. Constituent
Structure and Government in Phonology. Phonology Yearbook 7, 193–231.
124
Kaye, Jonathan. 2001. Working with Licensing Constraints. In K. DziubalskaKolaczyk
(ed.), Constraints and Preferences. Berlin & New York: Mouton de
Gruyter. 251–268.
Kochetov, Alexei & Alderete, John. 2011. Patterns and Scales of Expressive Palatalization.
Canadian Journal of Linguistics 56, 345–376.
Kramer, Ruth. 2015. The Morphosyntax of Gender. Oxford: Oxford University
Press.
Kučera, Henry. 1961. The Phonology of Czech. ‘s-Gravenhage: Mouton.
Lappe, Sabine. 2003. Monosyllabicity in Prosodic Morphology: The Case of
Truncated Personal Names in English. In G. Booij & J. van Marle (eds.), Yearbook
of Morphology 2002. Dordrecht: Kluwer. 135–186.
Lowenstamm, Jean. 1996. CV as the Only Syllable Type. In J. Durand & B. Laks
(eds.), Current Trends in Phonology. Models and Methods. Salford – Manchester:
ESRI. 419–441.
Lowenstamm, Jean. 2014. Derivational Affixes as Roots. In A. Alexiadou et al.
(eds.), The Syntax of Roots and the Roots of Syntax. Oxford: Oxford University
Press. 230–259.
Maddieson, Ian. 1997. Phonetic Universals. In W. J. Hardcastle & J. Laver (eds.),
The Handbook of Phonetic Sciences. Oxford: Blackwell. 619–639.
McCarthy, John. 1979. Formal Problems in Semitic Phonology and Morphology.
Doctoral dissertation, MIT, Cambridge.
Nasukawa, Kuniya. 2014. Features and Recursive Structure. Nordlyd 41(1), 1–19.
McCarthy, John & Prince, Alan. 1999. Faithfulness and Identity in Prosodic
Morphology. In R. Kager, H. van der Hulst & W. Zonneveld (eds.), The Prosody-Morphology
Interface. Cambridge: Cambridge University Press. 218–309.
Odden, David. 2011. The Representation of Vowel Length. In M. van Oostendorp
et al. (eds.), The Blackwell Companion to Phonology. Oxford: Blackwell.
465–490.
Palková, Zdena. 1994. Fonetika a fonologie češtiny. Praha: Univerzita Karlova.
Peciar, Štefan. 1946. Slovenská kvantita a rytmický zákon. Slovenská reč 7, 137–
152, 217–224.
Petr, Jan et al. 1986. Mluvnice češtiny I. Praha: Academia.
Pöchtrager, Markus. 2006. The Structure of Length. Doctoral dissertation, University
of Vienna, Vienna.
Rosenthall, Sam & van der Hulst, Harry. 1999. Weight-by-Position by Position.
Natural Language & Linguistic Theory 17(3), 499–540.
Ségéral, Philippe & Scheer, Tobias. 2005. What Lenition and Fortition Tells Us
About Gallo-Romance Muta cum Liquida. In T. Geerts, H. Jacobs & I. van
125
Ginneken (eds.), Romance Languages and Linguistic Theory. Amsterdam:
Benjamins. 235–267.
Scheer, Tobias & Ziková, Markéta. 2010. The Coda Mirror v.2. Acta Linguistica
Hungarica 57(4), 411–431.
Scheer, Tobias & Ziková, Markéta. 2010a. The Havlík Pattern and Directional
Lower. In W. Browne et al. (eds.), Formal Approaches to Slavic Linguistics.
The Second Cornell Meeting 2009. Ann Arbor: Michigan Slavic Publications.
470–485.
Scheer, Tobias. 2001. The Rhytmic Law in Czech: Vowel-final Prefixes. In G. Zybatow
et al. (eds.), Current Issues in Formal Slavic Linguistics. Frankfurt am
Main: Peter Lang. 37–48.
Scheer, Tobias. 2003. The Key to Czech Vowel Length: Templates. In P. Kosta
et al. (eds.), Investigations into Formal Slavic Linguistics. Frankfurt am Main:
Peter Lang. 97–118.
Scheer, Tobias. 2004. A Lateral Theory of Phonology. Vol. 1: What is CVCV, and
Why Should It Be? Berlin: Mouton de Gruyter.
Scheer, Tobias. 2004a. O samohláskové délce při derivaci v češtině. In Z. Hladká
& P. Karlík (eds.), Čeština – univerzália a specifika 5. Praha: NLN. 224–239.
Scheer, Tobias. 2011. Home-Made Western Slavic Vowel length. In R. Sukač (ed.)
From Present to Past and Back. Papers on Baltic and Slavic Accentology. Frankfurt
am Main: Peter Lang. 165–187.
Scheer, Tobias 2014. Spell-Out, Post-Phonological. In E. Cyran & J. SzpyraKozlowska,
Crossing Phonetics-Phonology Lines. Newcastle upon Tyne: Cambridge
Scholars. 255–275.
Schneider, Klaus-Peter. 2003. Diminutives in English. Tübingen: Niemeyer.
Starke, Michal. 2004. On the Inexistence of Specifiers and the Nature of Heads.
In A. Belletti (ed.), Structures and Beyond: The Cartography of Syntactic Structures.
Oxford: Oxford University Press. 252–268.
Starke, Michal. 2009. Nanosyntax. A Short Primer to a New Approach to Language.
Nordlyd 36, 1–6.
Starke, Michal. 2014. Towards Elegant Parameters: Language Variation Reduces
to the Size of Lexically Stored Trees. In C. M. Picallo (ed.), Linguistic Variation
in the Minimalist Framework. Oxford: Oxford University Press. 140–154.
Starke, Michal. 2018. Complex Left Branches, Spellout, and Prefixes. In Baunaz,
Lena et al. (eds.), Exploring Nanosyntax. Oxford: Oxford University Press.
349–363.
126
Steriopolo, Olga. 2008. Form and Funtion of Expressive Morphology: A Case
Study of Russian. Doctoral dissertation, University of British Columbia, Van-
couver.
Sukač, Roman. 2013. Fish and its Fisherman. Paradigmatic and Derivative
Length in Czech. Zeitschrift für Slavistik 58(1), 72–101.
Svoboda, Jan. 1964. Staročeská osobní jména a naše příjmení. Praha: Nakladatelství
ČSAV.
Trommer, Jochen & Zimmermann, Eva. 2014. Generalised Mora Affixation and
Quantity-Manipulating Morphology. Phonology 31, 463–510.
Uffmann, Christian. 2010. The Non-Trivialness of Segmental Representations.
Talk presented at the OCP 7, January 28–30, Nice.
Wiltschko, Martina & Steriopolo, Olga. 2007. Parameters of Variation in the Syntax
of Diminutives. In M. Radišič (ed.), Proceedings of the 2007 Annual Conference
of the CLA. Toronto: Canadian Linguistic Association. 1–12.
Yu, Alan. 2003. The Morphology and Phonology of Infixation. Doctoral dissertation,
University of California, Berkeley.
Ziková, Markéta. 2012. Lexical Prefixes and Templatic Domains: Prefix Lengthening
in Czech. In M. Ziková & M. Dočekal (eds.), Slavic Languages in Formal
Grammar. Proceedings of FDSL 8.5, Brno 2010. Frankfurt am Main: Peter
Lang. 325–338.
Ziková, Markéta. 2016. When Prosody Follows Syntax: Verbal Stems in Czech.
Linguistica Brunensia 64(1), 163–185.
Potsdam Linguistic Investigations
Potsdamer Linguistische Untersuchungen
Recherches Linguistique à Potsdam
Edited by / Herausgegeben von / Edité par
Peter Kosta, Gerda Haßler, Teodora Radeva-Bork, Lilia Schürcks,
Nadine Thielemann and / und / et Vladislava Maria Warditz
Band 1 Peter Kosta / Lilia Schürcks (eds.): Linguistics Investigations into Formal Description of
Slavic Languages. Contributions of the Sixth European Conference held at Potsdam University,
November 30–December 02, 2005. 2007.
Band 2 Lilia Schürcks: Binding and Discourse. Where Syntax and Pragmatics Meet. 2008.
Band 3 Christiane Hümmer: Synonymie bei phraseologischen Einheiten. Eine korpusbasierte
Untersuchung. 2009.
Band 4 Svetlana Friedrich: Definitheit im Russischen. 2009.
Band 5 Matthias Guttke: Strategien der Persuasion in der schriftkonstituierten politischen
Kommunikation. Dargestellt an Parteiprogrammen der Neuen Rechten in Polen. 2010.
Band 6 Peter Kosta / Lilia Schürcks (eds.): Formalization of Grammar in Slavic Languages. Contributions
of the Eighth International Conference on Formal Description of Slavic Languages
– FDSL VIII 2009. University of Potsdam, December 2–5, 2009. 2011.
Band 7 Roman Sukač (ed.): From Present to Past and Back. Papers on Baltic and Slavic Accentology.
2011.
Band 8 Diego Gabriel Krivochen: The Syntax and Semantics of Nominal Construction. A Radically
Minimalist Perspective. 2012.
Band 9 Teodora Radeva-Bork: Single and Double Clitics in Adult and Child Grammar. 2012.
Band 10 Anja Hennemann: A Context-sensitive and Functional Approach to Evidentiality in Spanish
or Why Evidentiality needs a Superordinate Category. 2013.
Band 11 Diego Gabriel Krivochen / Peter Kosta: Eliminating Empty Categories. A Radically Minimalist
View on Their Ontology and Justification. 2013.
Band 12 Christina Behme: Evaluating Cartesian Linguistics. From Historical Antecedents to Computational
Modeling. 2014.
Band 13 Kathleen Plötner: Raum und Zeit im Kontext der Metapher. Korpuslinguistische Studien zu
französischen und spanischen Raum-Zeit-Lexemen und Raum-Zeit-Lokutionen. 2014.
Band 14 Marion Eva Ernst: Produktnamen der Lebensmittelindustrie. Eine empirisch-strukturelle
Untersuchung. 2014.
Band 15 Stefanie Wagner: Eine „unbekannte“ Sprache lesen oder Von der Entdeckung des Nissart
durch Interkomprehension. 2015.
Band 16 Nataša Todorović: The Indicative and Subjunctive da-complements in Serbian: A SyntacticSemantic
Approach. 2015.
Band 17 Vladislava Warditz / Beatrix Kreß (eds.): Multilingualism and Translation. Studies on Slavonic
and Non-Slavonic Languages in Contact. 2015.
Band 18 Nadia Varley: Optionality and overgeneralization patterns in second language acquisition:
Where has the expletive ensconced itself? 2015.
Band 19 Verónica Böhm: La imperfectividad en la prensa española y su relación con las categorías
semánticas de modalidad y evidencialidad. 2016.
Band 20
Band 21
Band 22
Roland Wagner: Reflexivität im tschechisch-deutschen Sprachvergleich. Möglichkeiten und
Grenzen einer Prognose. 2016.
Ray C. H. Leung: Institutional Construction of Gamblers’ Identities. A Critical Multi-method
Discourse Study. 2017.
Markéta Ziková: Licensing of Vowel Length in Czech. The Syntax–Phonology Interface.
2018.
www.peterlang.com