Bracketing Paradoxes Resolved
Heather Newell
Université du Québec à Montréal
newell.heather@uqam.ca
Abstract
Bracketing Paradoxes (BPs) have been the subject of many different analyses since the 1970s.
Each of these analyses have included BP-specific machinery to account for the apparent mismatch
between the syntactico-semantic and morpho-phonological structures argued to be necessary for a
complete analysis of this phenomenon. This article proposes that independently necessary
operations and structures in the morpho-syntactic and phonological modules allow for an analysis
of BPs that avoids postulating ad-hoc tools. Specifically, a system that includes cyclic (phasal)
interpretation of the morpho-syntax (Chomsky 2001; Marantz 2007) in combination with a flat
(CVCV) phonological framework (Lowenstamm 1996, 1999; Scheer 2004) avoids the emergence
of paradoxical structures altogether. The discussion therefore includes both current morphosyntactic
and phonological analyses of each construction proposed to give rise to a BP;
comparatives (unhappier), Level-ordering BPs (ungrammaticality), Phrasal BPs (modular
grammarian), Compound BPs (particle physicist), Particle-verbs (podžëg ‘set fire’ [Russian]), and
Reduplicated BPs (kwíita-kwíita ‘to pour a bit’ [Kihehe]). The proposal that a flat phonological
framework is key in avoiding the paradoxical nature of BPs has implications for the correct
structure of phonological representations generally.
Keywords
Bracketing Paradox, CVCV Phonology, Cyclic Derivation (by Phase)
“Morphological/syntactic structure and phonological structure are independent levels
of analysis subject to independent constraints and principles. If this conclusion is
correct, “lexical phonology”, in which phonological rules and morphological
affixation work in tandem, is deeply wrong.” (Marantz 1987:203)
1 The problem with Bracketing Paradoxes
This paper has as its central concern the proposition in Marantz (1987) that Bracketing Paradoxes
(BPs) have deep implications for the (in)correctness of certain proposals within the domain of
generative phonology. Where it differs from Marantz’ account, and from every previous account
of BPs in the literature, is in the absence of an appeal to ad-hoc tools to eliminate these paradoxical
derivations. It is argued that Bracketing Paradoxes cease to emerge if we adopt a theory wherein
phonological representations are limited to (operations over) linear strings, and wherein
* Many thanks for comments and discussion go to Ricardo Bermúdez-Otero, to the audiences at GLOW, MfM, CLA,
and MoMOT, to the members of the Tromsø (CASTLFish) and Montreal (WSRG) research groups where this work
was presented, and to two anonymous reviewers.
- 2 phonological
operations must be triggered via phonological means (no Level-specific
morphological diacritics or operations).
The problem posed by Bracketing Paradoxes for an appropriate analysis of both morphosyntax
and morpho-phonology is well known. Assuming a compositional semantic module, the
morpho-syntactic structure of a BP must conform with its attested semantic interpretation (1a).
Assuming that the phonological proximity or distance of an affix to/from its base is due to its
interpretation in the same or in a different morpho-phonological cycle (as in Lexical Phonology’s
Level 1/2 distinction [Kiparsky 1982a; Mohanan 1982]), phonological stratum (Kiparsky 2000;
Bermúdez-Otero 2017), or phase (Chomsky 2001; Marantz 2007) the structure of a BP must also
conform to its surface phonological representation (1b).
(1) a. 4 b. 4
4 ity un 4
un grammatical grammatical ity
[[un grammatical]-ity] [[un][grammatical-ity]]
The structure in (1a) conforms to the compositional semantic interpretation of ungrammaticality;
‘the state of being ungrammatical’, and the structure in (1b) conforms to the phonological
requirements that -ity be within the phonological domain of grammatical (as evidenced by stress
shift) and that un be outside of the phonological domain of grammaticality (as evidenced by the
lack of nasal place assimilation).
Since the 1970s, Bracketing Paradoxes have been subject to numerous analyses (among
which are Allen 1979; Pesetsky 1979, 1985; Lieber 1980; Nash 1980; Williams 1981; Strauss
1982; Kiparsky 1982b; Selkirk 1982; Speas 1984; Pesetsky 1985; Marantz 1984a/b, 1987, 1989;
Sproat 1984, 1985, 1988; Nespor and Vogel 1986; Beard 1991; Carnie 1991; Lieber 1992; Booij
and Lieber 1993; Merchant 1995; Newell 2005a/b, 2008, 2018; Haugen and Siddiqi 2016; and
Bermúdez-Otéro 2017, 2019). See Newell (2019) for a historical overview of the research on BPs.
Each previous analysis of the BP problem has required the proposal of a BP-specific mechanism,
such as special LF rules for prefixes (Pesetsky 1979; Williams 1981), rebracketing at PF (Williams
1981; Nespor and Vogel 1986; Marantz 1984a/b, 1987, 1989; Sproat 1984, 1985, 1988), quantifier
raising of non-quantifiers (Pesetsky 1985), or suspension of Bracket Erasure (Kiparsky 1982b), to
account for derivations like that in (1). Another commonality of all previous analyses is that they
assume hierarchical (bracketed) structure in both the syntax and the phonology. In this article I
take it as given that hierarchical structure is necessary for syntactic analyses, but it is proposed
here to be unnecessary in the phonology (see also Lowenstamm 1996, 1999; Scheer 2004; Newell
2017a, 2017b, 2017c; Newell and Scheer 2017). The current, standardly assumed, hierarchical
theory of phonological structure, Prosodic Phonology (Selkirk (1981 [1978]), 1982, 2011; Nespor
and Vogel 1986, and seq.) proposes that phonological domain formation is regulated by the
Prosodic Hierarchy (PH). The PH was proposed as a replacement for the undesirable boundary
symbols found in SPE (Chomsky and Halle 1968), whose function was to explain the application
or non-application of phonological rules within a particular portion of a phonological string (see
Scheer 2011 for an overview of arguments against these boundary symbols). Interestingly, a
consequence of this modification of phonological theory (from linear to hierarchical phonological
structure) evidenced the birth of Bracketing Paradoxes. It is clear that BPs are only possible in a
system wherein phonological representations are hierarchical.
- 3 (2)
##modul+ar##grammar+ian## à (mɔ́dʒə)<ləɹ> gɹə(méɹi)<ən>
In the SPE-style representation in (2) stress rules ignore the phonological segment + but are
blocked by ##, ensuring that a phonological division emerges only between the two words.
Morpho-syntactic constituency in SPE was translated into these phonological [-segmental] primes
inserted into the linear string. As no hierarchical structure was built in representations such as (2),
constituency was not at issue. Consequently, these representations were incapable of introducing
Bracketing Paradoxes. Although the arguments in the literature against boundary symbols are clear
and convincing (Pyle 1972; Rotenberg 1978; Devine and Stephens 1976; Kenstowicz and
Kisseberth 1977; Hyman 1978, among others), the BPs introduced by Prosodic Phonology are
argued here to indicate that such a theory is not the correct replacement for an SPE-style linear
phonological system. The emergence of BPs is directly due to the hierarchical representations of
the Prosodic Hierarchy, and therefore indicative of problems inherent to these representations. In
the following sections I will discuss the syntactic derivations of all classes of word-level BP in the
literature (and one phrasal example) and will demonstrate how their syntactic derivations, in
combination with a non-diacritic linear phonological system (following Scheer 2008), capture the
data in a uniquely satisfying way. This analysis highlights modifications that I will argue must be
made to our theories of phonological representations.
Specifically, I will argue that the timing of spell-out of each morpheme in a derivation,
along with the edge-marking of phonological domains by an empty CV (following Lowenstamm
1999, Scheer 2009a) can account for derivations traditionally labeled BPs without ever giving rise
to phono-syntactic paradoxes.
2 The phonological and morphosyntactic frameworks assumed herein
Newell (2016a,b, 2017c) introduces a liaison account of English Level 1/Level 2 morphophonology
(3,4). Level 1/Level 2 will henceforth be referred to as cohering/non-cohering,
following Rafelsiefen (1999), in order to abstract away from the theoretical implications of the
former labels. The latter imply that an affix is treated as interior/exterior to the domain of footing
and stress assignment of its base. The liaison analysis is argued to be a better account of
morphological class-membership than the classic analyses in Lexical Morphology and Phonology
type frameworks where affixes are either assigned to specific levels, or subcategorize for certain
types of bases (Classic LMP: Mohanan 1982; Kiparsky 1982a; Giegerich 1999, LMP-OT:
Kiparsky 2000, Stratal OT: Bermúdez-Otero 2017), Affix-specific phonologies (Benua 1995;
Orgun 1996; Inkelas 1998; Raffelsiefen 1999, 2015; Plag 1999; Steriade 2000; Pater 2000, among
others), and Output-Output/Paradigm-based correspondence models (McCarthy 1995; Steriade
2000; Downing et. al 2005; Kiparsky, 2005) for the following reasons. First, it is fully modular
(Fodor 1983, 1985. See e.g. Reiss 2007; Scheer 2011 for discussion); the phonological structures
proposed make no reference to either morphological classes or to morpho-phonological constraints
like Alignment or Correspondence. A modular account of phonology (and of the morpho-syntactic
module) is preferable in that it restricts possible phonological operations to those that are triggered
via strictly phonological means. Second, it accounts for the fact that all cohering affixes in English
are vowel-initial (following Raffelsiefen 1999, 2015, van Oostendoorp 1994). It is generally stated
that lexical morphological classes cannot be uniformly distinguished based on their phonological
shape. Although this may be true for some languages (necessitating further research to incorporate
them into the modular theory argued for here), it is not strictly correct for English; there is, on the
- 4 surface,
a one-way correlation between the first segment of a cohering affix and its phonological
behavior (vowel-initial). Newell (2017c) argues that this one-way correlation can actually be
represented as a two-way correlation; all and only cohering affixes begin with a floating vowel.
Third, it has been shown clearly that many English affixes may have both cohering and noncohering
variants (Giegerich 1999; Bermúdez-Otero 2011). The liaison proposal accounts for the
fact that affixes that are variably cohering and non-cohering do not display cohering behaviour
when merged outside of another affix. Among the affixes that switch classes, their cohering variant
is always root-attaching, while their non-cohering variant is always affixed to a complex base.
Given that cohering affixes are not restricted to root-attachment (ex. -al in govern-ment2-al1) this
pattern is unexpected in any system where class-membership is determined lexically.
The analysis in Newell (2017c) is represented in (3) and (4). Floating vowels link to a final
empty V slot on the CV tier of their base (3). This analysis presupposes a Strict-CV (CVCV) linear
phonology (Scheer 2004), where the timing tier is a uniform sequence of Cs and Vs. Cs and Vs
may remain unpronounced; these are indicated with a Æ. Final Empty Nuclei (FEN) are
parametrically licenced. As English allows consonant-final words, it is clear that this parameter is
active. Affixes normally proposed to be non-cohering have a fully linked melodic tier under this
analysis, as in (4).
(3) C V C V C V C V C V C V
‘grammarian’| | | | | | | | | | |
g Æ ɹ ə m e ɹ i j ə n Æ
(4) C V C V C V C V C V C V
‘grammarless’| | | | | | | | | | | |
g Æ ɹ æ m e ɹ Æ l ə s Æ
The association of the floating [i] in (3) forces a cohering analysis of the syllabification and stress
of grammarian as the phonological system sees a single unified string of timing slots on the CV
tier after linking is effected. Affixes with an initial floating vowel must merge inside of the domain
to their left, as the vowel needs to link to the CV tier in order to be pronounced. This analysis is
directly akin to the generally-accepted phenomenon of liaison in French (ex. peti[t] garcon vs.
peti[t] ami) (Encrevé 1983). In (4), however, grammar and -less are not syllabified as a single
uninterrupted string. -less leans on the phonological string to its left (it is suffixal/clitic-like),
explaining its lack of independent stress. Importantly for the discussion of BPs here is the fact that,
under Newell’s analysis, whether an item will be an independent word or an affix (either cohering
or non-cohering) is not predictable based on syntactic structure. Both -less and -ian are affixes,
regardless of how much morpho-syntactic structure separates them from grammar.
This non-isomorphism between morpho-syntactic and phonological domains was one of
the motivations for Prosodic Phonology, but this divergence is not taken to be the norm. Typically,
non-isomorphism is seen as an occasional deviation from complete isomorphism, as in Selkirk’s
Match Theory (2011), and that words are generally the phonological interpretation of (complex)
X0
s. Therefore, within the framework of the Prosodic Hierarchy the difference between (3) and (4)
must be that the former is a complex X0
interpreted as a single Prosodic Word (PWd), and the
latter is a complex X0
interpreted as a nested PWd structure or a Composite Group (Vogel 2009).
Unfortunately for such a theory, it is evident from the cross-linguistic variation that words are not
restricted to the phonological interpretation of X0
s (Julien 2002; Haspelmath 2011; Svenonius
- 5 -
2016; Newell 2017a; Newell et al. 2017). Regardless, the difference between the phonological
structures of (3) and (4) is clear. In (3) the phonology is sensitive to the procedure of liaison. In
(4) liaison does not occur. Why -less is an affix and not a separate word may be due to distributional
factors (ex. Julien 2002), or lexico-syntactic marking of affix-status (ex. Svenonius 2016), but a
coherent theory of wordhood vs. affixhood is beyond the scope of this paper and will not be crucial
to the analysis of BPs below. What is important in the discussion to follow is that the linearization
of phonological objects, and the phonological operations effected over these strings, are sensitive
to the underlying lexical representations of each morpheme.
An additional tool to be employed in the following pages is one proposed within the theory
of Strict-CV phonology to delimit phonological domains; the cycle-initial empty CV proposed by
Lowenstamm (1999) and elaborated on by (Scheer 2009a and subsequent work).1
Scheer proposes
an empty initial CV as a modular replacement for SPE’s # and for the PWd of the Prosodic
Hierarchy. He argues that the CV, unlike the PWd, is a native phonological object, making
predictions in the phonology. An initial CV is (i) an interface-visible object (both phonological
rules and rules of allomorphy must have access to the CV tier), and (ii) a non-diacritic boundary
marker (see Scheer 2009b). We will see that this initial CV (in grey throughout) can block
phonological rules from treating (for example) modifiers as part of a single phonological domain
with their base (5).
(5) C V C V C V C V C V C V C V C V C V C V
| | | | | | | | | | | | | | | | | | | |
Ø Ø m ɑ dʒ ə l ə ɹ Ø Ø Ø g Ø ɹ æ m ə ɹ Ø
The empty CVs in (5) are inserted upon interpretation of modular and grammar, assuming a phasebased
morpho-syntax as in much work in Distributed Morphology (Arad 2003; Marantz 2007;
2013; Marvin 2002, 2013; Newell 2008; Embick 2010 etc.). In this type of framework,
categorizing adjectival, nominal, and verbal heads trigger spell out ([[modul-ar]a [grammar-Ø]n]n),
in addition to the ‘phrase-sized’ vP, CP, PP, and DP phases (as in Chomsky 2001 and subsequent
work in the field). In this way, cycles for the application of phonological rules are determined by
the syntax. The insertion of phase-initial empty CVs is restricted phonologically, as we will discuss
further in Section 3.5 when we examine particle verbs in Slavic languages.
We have in this paper, therefore, a framework where spell-out domains are determined in
the morpho-syntax (phases). In the phonology, strings computed in the same phase will be treated
as single domains for the application of phonological operations. Inter-cyclic phonological
communication is blocked by the non-interaction of melodic structure in separate cycles,
sometimes with the aid of an initial empty syllable/CV. Inter-cyclic communication will, however,
be forced by phonological means in the case of liaison, or any similar case of Phonological Merger
(Newell and Piggott 2014) where a phonological operation is triggered, forcing the insertion of a
phonological object (ex. floating feature, floating segment, unfooted syllable, tone) into a
previously computed domain and triggering resyllabification (among other processes) if necessary
(as in (3)). We will see below that such an analysis, when combined with the independently
1
Note that the grammatical insertion of syllabic space is not a proposal specific to CVCV phonology. Chierchia (1986)
and Larsen (1998), for example, have proposed the insertion of syllables or morae for similar reasons. The form of
the syllabic space inserted, a CV sequence, is specific to the theory assumed here.
- 6 supported
morpho-syntax of the constructions at issue, eliminates both Bracketing Paradoxes and
the need to appeal to the Prosodic Hierarchy.2
3. Kinds of Bracketing Paradoxes
Newell (2019) summarizes the literature on Bracketing Paradoxes, and notes that crosslinguistically
BPs have been split into 5 types. The comparative BP is specific to English (6a) and
gets its own entry due to the amount of publicity it has received, including two LI squibs: Sproat’s
(1992) Unhappier is not a “Bracketing Paradox” followed by Kang’s (1993) Unhappier really is
a “Bracketing Paradox”) (treated in Section 3.1). An example of Level-Ordered paradoxes, where
a cohering affix is merged outside of a non-cohering affix can be seen in (6b) (see Section 3.2).
Compound and Phrasal paradoxes are generally considered sub-categories of Level-Ordered
paradoxes (6c) but in these the ‘prefix’ is the left-hand member of a phrase or compound. The
distinct behaviours of the sub-types of (6c) will be teased apart in Sections 3.3 and 3.4.
(6) Comparative Level-Ordered Compound/Phrasal
a. Degree
4
A Deg
4 er
un A
happy
b. N
4
A N
4 ity
un A
grammatical
c. N
4
N N
4 ist
A N
particle/ physic(s)
nuclear
Prefixed Verb Reduplication
d. Tense
4
V T
4 mi
P(re) V
tirl pardi
e. Inf
4
ku V
4
RED V
íita íita (adapted, Newell 2019)
Additional categories of BP are prefixed verbs (ex. Warlpiri, Nash 1980 in (6d)) (see Section 3.5),
and reduplicative structures (ex. Kihehe, Marantz 1987 in (6e)) (see Section 3.6).
The reader may note, as Kiparsky (1982b), Newell (2005a,b, 2008) and others have, that
all BPs contain one component (affix or modifier) that does not project in the syntax. While this
characteristic is crucial in the accounts to follow, it will be shown that the non-projection of
(prefixal) elements in the morpho-syntax may come about via different means.
2
This analysis also represents a linear phonological account of the cross-linguistic pattern whereby prefixes are much
more likely to be phonologically independent from their base than suffixes (ex. Newell and Scheer 2017).
- 7 -
3.1 Level-ordered Bracketing Paradoxes
What the above phonological analysis in Newell (2017c) entails for Level-Ordered Bracketing
Paradoxes in particular is that it is only linear order, and not the Prosodic Hierarchy or spell-out
within a particular phase/cycle, that determines the phonological behaviour of the relevant
cohering affixes. Just like infixation (absobloodylutely) and Phonological Merger of unfootable
material (see Newell and Piggott 2014 and Newell and Scheer 2017 on Ojibwe), liaison obscures
the boundaries between phases/cycles for the simple reason that the phonology does not contain,
and therefore cannot be sensitive to, a representation of hierarchical structure. In BP derivations
like ungrammaticality there is no issue with the phonological output being non-isomorphic with
the morpho-syntactic representation. The morpho-syntax is sent to spell-out in phases, and in line
with the requirements of a compositional semantic module the interpretation of ungrammaticality
is ‘the property of being ungrammatical’ (7); an interpretation that is consistent with the morphosyntactic
structure required by the selectional restrictions of the respective affixes. The output of
phonological computation is strictly linear as in (8a-d) where the (PF) phases of interpretation in
this derivation are represented. These phases follow the proposals that (i) category-defining heads
(little a, n, and v) are phase heads that are spelled out with, or are phonologically visible to, their
complements (see Arad 2003; Embick 2010; Marantz 2013 and previous work; Marvin 2002,
2013; Newell 2008), and (ii) that left-branches (adjuncts and specifiers) are interpreted (both
phonologically and semantically) separately from the larger tree into which they are merged
(Uriagereka 1999; Johnson 2004; Svenonius 2016; Starke 2018, among others).
(7) n
4
a n
4 ity
un a
4
Ö a
grammatic al
c. C V C V C V C V C V C V C V C V C V C V
| | | | | | | | | | | | | | | | | | | |
… Ø Ø Ø ʌ n Ø Ø Ø g Ø ɹ ə m æ t ɪ k ə l Ø
d. C V C V C V C V C V C V C V C V C V C V C V
| | | | | | | | | | | | | | | | | | | | |
Ø Ø Ø ʌ n Ø Ø Ø g Ø ɹ ə m æ t ɪ k æ l ɪ ɾ i
(8) a. C V C V C V C V C V C V C V
… | | | | | | | | | | | | | |
Ø Ø g Ø ɹ ə m æ t ɪ k ə l Ø
(10) b. C V C V C V
| | | | | |
… Ø Ø Ø ʌ n Ø
- 8 Beginning
with the phases in (8a,b), we must note that they are derived in parallel, in that neither
one syntactically contains the other, and therefore neither can be said to be processed prior to the
other. The prefix un- has been analysed in the literature as either a morphological adjunct (Newell
2005a, 2005b, 2008) or as the specifier of NegP (ex. De Clercq 2013).3
Studies of adjunct/specifier
left-branches in the syntax contend with the fact that these are islands for movement, and that they
evidence prosodic and semantic separateness from their bases of attachment. I will assume in the
representations here that un- is an adjunct, although nothing crucial hinges on this. Uriagereka
(1999) proposes to account for the particular nature of syntactic left-branches by proposing that
they are interpreted in a separate but parallel syntactic and phonological computation, and then
merged post-spell out. Similar analyses are also assumed in a Spanning framework (Svenonius
2016), or a Nanosyntactic framework (Starke 2018). According to such proposals, both un- and
grammatical (a phase itself, triggered by the category-defining head -al) are interpreted separately
before merger to one another. This independent interpretation will endow each of these domains
with an initial CV in English, separating them phonologically even under affixation (8c). This
derivation offers two potential motivations for the non-assimilation of the nasal consonant in the
prefix to the place of articulation of the following consonant; (i) the nasal is interpreted before
affixation to the base, and therefore default place is assigned to it, bleeding assimilation, or (ii) the
intervening empty CV causes the nasal to be non-local to the following consonant, blocking the
application of assimilation.
Affixation of -ity leads to the spell-out of n in (7). As the head n c-commands the head a,
grammatical will be visible to the floating vowel of -ity at phonological interpretation.4
Its floating
vowel will link to the final empty vowel position to its left. This linking creates a phonological
structure identical to one that would have been created were grammatical and -ity to have been
spelled-out in the same phase.5
Resyllabification (or the re-analysis of Government and Licensing
relations in Government Phonology (GP, including CVCV)) and stress shift therefore occur, but
do not affect un-, as it remains insulated by empty syllabic space from the domain of syllabification
of grammatical. What is clear in this proposed derivation of ungrammaticality, and of similar BPs,
is that there is no paradoxical output at any point. The phonology never contains any hierarchical
representation. No matter the morpho-syntactic distance of the liaison affix from its base, it must
be syllabified inside the domain to its left, and the prefix in these constructions can be shown to
3
Newell (2005a/b, 2008) argues that un- (i) does not project, (ii) merges with multiple syntactic categories, and (iii)
adds only compositional meaning to its base of attachment, and therefore it can be concluded that un- is an adjunct.
Note that this analysis of verbal (undo), nominal (unBirthday), and adjectival (unhappy) un- as a single object has no
bearing on the analysis of Bracketing Paradoxes herein, as each un- displays adjunct-like behaviour independently.
This is also consistent with an analysis of un- as the Specifier of NegP, as in De Clercq (2013) and subsequent work,
if NegP can select for verbal, adjectival, and nominal complements.
Work such as Bruening (2014) proposes that un- is a head. It is difficult to align such an analysis with the
phonological behaviour of un- in a non-diacritic manner.
4
There is evidence that the phonological visibility of a base is influenced by c-command relations between an affix
and said base. See Kaisse (1985), Odden (1990), Newell (2008), Newell and Piggott (2014) and Kalivoda (2018) for
discussion. Note that this restriction on visibility does not entail that the phonology makes direct reference to syntactic
structure.
5
See Embick (2014) and Newell (2017b,c,d) for discussions of how Phase Impenetrability does not apply to the
relevant phonological structure here, allowing -ity to affect a previously interpreted cycle (the aP containing
grammatical).
- 9 have
the characteristics of any other syntactic left-branch, ensuring its interpretation in an
independent phonological cycle.6
3.2 Comparative Bracketing Paradoxes
The derivation of comparative BPs is analogous to the above, but here the left-branch nature of
un- insulates it from conditioning the allomorphy of the comparative (or superlative) suffix, as
well as ensuring that it is not in the domain of main stress assignment. The base of attachment of
the comparative morpheme that conditions its allomorphy as synthetic (-er/-est) or analytic (more,
most) is the phonological domain to its left.7
As un- is never syllabified within the domain of the
its base of affixation, it is invisible to (or unimportant to) the phonologically-conditioned
Vocabulary Insertion of the comparative suffix. The structure in (9), necessary for the correct
semantic interpretation of unhappier (‘more unhappy’, rather than ‘not more happy’) is compatible
with a linear phonological representation where the comparative is sensitive to the size of its
domain of attachment, a domain that excludes un-.8
(9) Degree
4
a Deg
4 er
un a
4
Ö a
happy Ø
(10)
…
C V C V C V C V C V C V C V C V
| | | | | | | | | | | | | | | |
Ø Ø Ø ʌ n Ø Ø Ø h æ p i j ə ɹ Ø
The stages of interpretation of (9) and (10) are analogous to those of ungrammaticality. un- and
happy will be interpreted separately and then linearized after they have been merged. The degree
head is then sensitive to the domain to its left; a domain that excludes un-. Scheer (2016)
demonstrates that allomorphy may be sensitive to the CV/skeletal structure it merges with (but not
to melody). The invisibility of un- to -er/more allomorphy is consequently unsurprising.
Note that this analysis does not imply that -er/-est need be cohering affixes; they never
induce stress shift and therefore plausibly do not contain initial floating vowels. Bermúdez-Otero
(2013) demonstrates that the comparative suffix is outside of the first phase/phonological cycle in
Belfast English except when it conditions allomorphy of its base of attachment. He shows that in
6
To elaborate on the statement in fn 5, this is not a direct-access account of the interface. I am not proposing that
phases or syntactic structure have any influence within the phonological module. Phases determine which chunks of
syntactic structure will undergo spell-out. After Vocabulary Insertion (Halle and Marantz 1994) the phonological form
of each morpheme and any empty phonological structure (either morphological, e.g. a reduplicative morpheme, or
grammatically/cyclically determined, e.g. a left-edge CV) are the only objects that are considered during the
phonological derivation.
7
More complex degree derivations may evidence the effects of syntactic locality on this allomorphy in way that is not
relevant to the analysis here (see Svenonius 2016 for a recent discussion).
8
A reviewer brings up examples like *punch-drunker or *house-prouder, originally discussed in (Spencer 1988). The
left-members of these compounds bleed the insertion of -er, in opposition to un-. These examples will be discussed in
Section 3.4.
- 10 this
dialect coronal stops are dentalised before /(ə)ɹ/ (11a), except when that /(ə)ɹ/ is the exponent
of a non-cohering affix (11b). Importantly, in suppletive comparatives /(ə)ɹ/ triggers dentalization
(11c).
(11) a. Peter [ˈpit̪əɹ
], ladder [ˈlad̪ ə], dinner [ˈdɪn̪ əɹ
], pillar [ˈpɪl̪əɹ]
b. fa[t]er, lou[d]er, fi[n]er, coo[l]er
c. better [ˈbɛt̪əɹ
] ‘good.comparative’ (cf. better [ˈbɛtəɹ
] ‘one who bets’)
(Bermúdez-Otero 2011: 2022)
The distinct behaviour of better ‘good-comparative’ is open to more than one explanation. First,
though the adjectival head (a phase head) between the root and Deg in (10) will normally induce
spell-out of its complement separately from -er/-est, when the complement contains a listed
allomorph (good/bet) Domain Suspension (Bobaljik and Wurmbrand 2013) will prevent the aP
from undergoing spell-out until the Deg head is merged. Bobaljik and Wurmbrand argue that
morphemes which evidence outward-in allomorphy (allomorphy conditioned by an outer
morpheme) will delay Vocabulary Insertion until a subsequent head is merged into the tree,
allowing for the suspension of spell-out just in case allomorphic selection will be triggered by this
outer head. In this case, the comparative morpheme will be included in the spell-out domain of the
root, allowing for dentalization to apply. DegP will undergo spell-out in the same cycle as aP (still
excluding the independently spelled-out un-). A second option, pointed out by a reviewer, is that
better is simply monomorphemic. Bruening (2017) argues this to be the case due the unique
behaviour of better/best as achievement verbs (to better/best X); an environment where
comparative/superlative affixes are generally not permitted (*to wider(en)/*to widest(en)). In
either case, the data in (11) show that regular degree affixes are outside of the phonological domain
to which they attach, but may still be sensitive to the phonological properties of their base; a pattern
that is completely in keeping with a phonologically realizational analysis of the morpho-syntax
(ex. Bobaljik 2000). Importantly for our purposes, this type of derivation is compatible with a
linear representation of the phonology, avoiding the need to appeal to readjustment rules to account
for the phonological separateness of un-. As in Section 3.1, an appeal to hierarchical structure to
account for the phonological domains in the comparative derivation is unnecessary, and its
inclusion would therefore be more complicated than the alternative that assumes only linear
phonological representations.
3.3 Phrasal Bracketing Paradoxes
This section presents an analysis of Phrasal BPs, such as nuclear physicist, which obviates the
need for any discussion of these BPs as BPs, the syntactico-semantic derivation of these
constructions being isomorphic with the phonological interpretation under any account. In Section
3.4 I will argue that the difference in the position of stress in Compound BPs (particle physicist)
and Phrasal BPs argues for distinctions in the analysis of these constructions. The conclusion of
this section aligns with the conclusions presented for Comparative and Level-Ordering BPs above;
if the correct syntax for each construction is paired with a linear phonological representation, no
paradoxes emerge.
- 11 The
structure normally assumed for Phrasal BPs is as in (12), where the modifier is merged
below the nominalizing suffix. Here we assume a structure where the modifier, following Steddy
(2019), is merged directly to the root of its modifyee.
(12) n
4
Ö n
5 ian
a Ö
4 grammar
Ö a
modul ar
This is proposed to be the necessary structure for such constructions as the semantics of the
modifier and its base (modular grammar) has idiomatic properties that must be negotiated prior to
the interpretation of the suffix. In other words, a modular grammarian is ‘a practitioner of modular
grammar’ and not ‘a modular practitioner of grammar’; the latter being the proposed
compositional semantics for a structure where modular is merged outside of -ian, as in (13).
(13) n
5
a n
4 4
Ö a Ö n
modul ar grammar ian
There are at least two reasons to believe, however, that the structure in (13) is more appropriate
for both interpretations of modular grammarian. Bermúdez-Otero (2017, 2019) notes that the
structure in (13) is not justified by the semantics of these constructions. This becomes clear when
one examines how the nominal suffix may compose with its base. He points to constructions where
it is not only the combination of modifier and base that triggers idiomatic interpretation, but also
the combination of the base with the suffix. Consider the following pair:
(14) a. nuclear physicist
b. nuclear physician (Bermudéz-Otero 2019:16)
If the semantics of [nuclear physic] in both of the cases above were to be negotiated before the
interpretation of the suffixes -ian/-ist we would expect both to have an identical core meaning.
This is not the case. (14a) refers to a practitioner of nuclear physics and (14b) refers to a
practitioner of nuclear medicine. The semantics of physicist and physician therefore must be
negotiated independently of the interpretation of nuclear, as expected from a structure such as
(13).
How then, is the idiomatic reading of nuclear physics or modular grammar derived if the
adjectival modifiers are merged outside of the nominalizing head? It must be the case that this type
of interpretation is negotiated later in the derivation (as is the case for phrasal idioms such as kick
the bucket)9
. In support of this proposition, consider that Fábregas (in press) argues convincingly
9
Note that this negotiation of special meaning with a sub-part of the domain it modifies is not specific to
morphologically complex constructions like nuclear physicist. Nominal modifiers may negotiate an idiomatic
interpretation with a sub-part of the semantics of the base noun to which they are merged even in cases where the part
- 12 that
qualitative and relational adjectives inhabit distinct morpho-syntactic positions, but that both
are merged outside of the nominalizing head, as in (15).10
He argues that (among other distinctions)
relational adjectives must be merged as specifiers of a functional head (F1) in the nominal domain
that is lower than the functional head (F2) that hosts qualitative adjectives. Relational adjectives
are argued to be KPs (case phrases) and to therefore be too small to exhibit scalar modification
(e.g. they cannot be modified by very). Qualitative adjectives will be interpreted as predicational
modifiers and may be scalar (e.g. modifiable by very). (15) is the structure of a (very) nuclear
nuclear physicist; ‘a nuclear physicist who is (very) central a particular project’.
(15)
(modified from Fábregas, in press)
This type of analysis also easily allows for cross-linguistic variation of morpheme order in these
constructions. Consider the French translation of FP1; physicien(ne) nucléaire (16). Here the
nominalizing suffix (and its concomitant gender suffix) intervenes linearly between the base
(physic) and its modifier (nucléaire). This would be unexpected if nuclear were to modify the root
directly, as linearization of a higher suffix (either n or gender) should not target the left-edge of
physic in such a structure. In a Fábregas-style analysis, however, physicien(ne) in French raises up
over its modifier, as in (16).
(16) a. physicien(ne) nucléaire ‘nuclear physicist(F)’
physic-ien-(ne) nuclé-aire
physic-ist-(F) nucle-ar
of the semantics of the noun that is crucial to the idiomatic interpretation is not morphologically distinguishable (and
hence not the locus of adjunction of the modifier in the syntax). This is exemplified by constructions such as beautiful
chef, which can mean either ‘someone who cooks beautifully’ or ‘a chef who is beautiful’. In this type of example
there is no possible locus for low adjunction of the modifier directly to a sub-part of the monomorphemic chef, but the
two readings (relational and qualitative) are nonetheless available (Postal 1969). Thank you to Gillian Ramchand and
an anonymous reviewer for this example.
10
See Fábregas (in press) and references therein for detailed discussions of the syntactic properties and positions of
different classes of adjectives. An in-depth discussion of these data would take us too far afield from the discussion at
hand.
- 13 -
b.
This type of evidence suggests that all adjectival modifiers are merged outside of the domain of
the nominalizing suffixes in purported Phrasal BPs. There is therefore no expectation for
paradoxes to arise in these derivations. The nominalizing suffix (e.g. -ian) in a Phrasal BP is
interpreted with its base before the modifier is merged. Relational adjectives are interpreted
distinctly from qualitative adjectives due to their position of merger, but this position is nonetheless
exterior to the derived nominal, explaining their phonological independence from the noun they
modify.
3.4 Compound Bracketing Paradoxes
Although the phrasal BPs above do not call for a derivation wherein the modifiers are merged to
the root, compound BPs (such as particle physicist) do argue for a low-attachment of their lefthand
members, following Harley (2009) and Steddy (2019). Steddy follows the analysis of
compounding proposed in Harley (2009) and argues that the allomorphy evidenced in physicist is
unproblematic if both adjectival modifiers like nuclear and nominal modifiers like particle are the
first items merged to the root PHYSIC. First-merged items then incorporate into the head of the
compound, giving the structures in (17a).
(17) a. n
4
Ö n
5 ist
n Ö
4 phycis
Ö n
part icle11
*b. n
4
Ö n
5 ist
a Ö
4 phycis
Ö a
nucle ar
(Steddy 2019)
A reviewer correctly points out, however, that there is a well-known distinction in stress
patterns between modifier-noun derivations and compound-noun derivations. In nuclear physicist,
the dominant stress falls on physicist, while in particle physicist (a primary compound in the
terminology of Harley 2009) the dominant stress falls on particle, just as it does in synthetic
11
Whether the root and affix in nuclear or particle are synchronically separable is not pertinent to the discussion here.
- 14 compounds
like truck-driver).12
If we take this distinction to be due to the position of Nuclear
Stress, and following Cinque (1993), propose that Nuclear Stress is assigned to the most embedded
word in the above constructions, then the stress in (17a) is consistent with the left-hand member
being more embedded than its head noun, but the stress of (17b) is not. Phonological evidence
therefore supports the analysis of Phrasal BPs discussed in the previous section; the correct
derivation for nuclear physicist is therefore not (17b), but (15). In (15) neither nuclear nor physicist
contain one another structurally. Each will therefore be sent to PF upon merger of their categorydefining
heads, and each will be endowed with a Left-Edge CV. Nuclear Stress will be assigned
to each word. After nuclear has been merged into the specifier of the functional structure above
physicist the two words will be linearized, but phrasal Nuclear Stress will fall on physicist, as it is
structurally lower than its modifier. The final linear phonological structure of nuclear physicist is
therefore as in (18).
(18) C V C V C V C V C V C V C V C V C V C V C V C V
| | | | | | | | | | | | | | | | | | | | | | | |
… Ø Ø n u k Ø l i j ə ɹ Ø Ø Ø f ɪ z ɪ s ɪ s Ø t Ø
In the derivation of particle physicist, the distinct determination of stress is predicted by
the derivation where the left member of the compound is incorporated into the head noun, as in
(17a). In this derivation, particle is contained within the structure of physicist and therefore will
be sent to spell-out prior to interpretation of the head noun. In this cycle, particle will receive word
and Nuclear Stress, as well as a Left-edge CV. In the subsequent cycle of interpretation, physicist
will be interpreted. As particle is c-commanded by physicist, the latter will neither be the most
embedded element at spellout, nor the leftmost in its spell-out domain. It will therefore receive
word stress (it will be footed), but will receive neither Nuclear Stress nor a Left-Edge CV. The
linear representation of particle physicist is therefore as in (19).13
(19) C V C V C V C V C V C V C V C V C V C V C V
| | | | | | | | | | | | | | | | | | | | | |
… ø ø p a ɹ Ø t i k ə l Ø f ɪ z ɪ s ɪ s Ø t Ø
This distinction in phonological structure receives support from the behaviour of degree
allomorphy in compound structures. A reviewer brings up examples like *punch-drunker or
*house-prouder, originally discussed in (Spencer 1988). The left-members of these compounds
bleed the insertion of the synthetic degree allomorphs -er/-est. This can be directly compared to
the non-bleeding of synthetic degree morphology in derivations prefixed with un- (e.g. unhappier),
as discussed in §3.2.14
This distinction is easily explained in terms of phonological domain
12
Some compounds have unexpected stress on the rightmost member (c.f. APPLE cake vs. apple PIE). These exceptions
will not be treated here. Main stress falling on the left member of a compound is assumed here to be the default pattern.
13
The maintenance of word stress on each member of the compound is predicted as the procedural (as opposed to
representational) derivation stresses first particle then the newly-inserted morphemes in physicist. As physicist
contains no floating elements/segment, interactions between the two words is not predicted regardless of the absence
of an intervening empty CV. How the position of stress in such a derivation is distinguished from the position in a
suffixed form such as governmental is left to future work.
14
It may also be the case that allomorphy of the degree head is sensitive to the position of Nuclear Stress, which would
be too far from the right edge in a compound to trigger selection of the synthetic suffixes. Note that -er suffixation is
also complicated by non-compound modification (e.g. ??
incredibly whiter). See Newell (2008) and references therein
for a discussion of these constructions.
- 15 distinctions.
The two words in a compound like house-proud are not separated by an empty CV,
and therefore at the point of Vocabulary Insertion for the degree head, the entire phonological
string house-proud is visible and available to condition allomorphy. The -er/-est allomorphs do
not select for bases that are bi-syllabic which end in stops (c.f. exceptions such as politer) and
therefore the output of such a derivation will be more/most house-proud.
The conclusions of Sections 3.1-3.4 are as follows. First, procedural cyclicity is important
for determining the position of stress in phrasal, compound, and prefixed structures. The most
embedded element in each receives Nuclear Stress. Default nasal place will be assigned to un- in
its first cycle of interpretation, bleeding place assimilation. In cases of
incorporation/compounding, the procedural account of stress placement argues for a distinct
representational output from phrasal compounds; notably, the two members of compounds are not
separated by an empty CV. This falls out of the predictions of the system but is then confirmed by
the behaviour of compounds in degree derivations. All in all, none of these derivations,
traditionally thought to give rise to variations of Level-Ordering-type BPs, give rise to bracketing
paradoxes.
3.5 Particle Verb Bracketing Paradoxes
The first full description of a BP was Pesetsky (1979)’s analysis of Russian Yer deletion in
Prefixed verbs.15
Pesetsky follows the traditional analysis of Yer vowels, wherein they are
underlying high vowels (realized as mid-vowels) or unpronounced. These options are derived via
the two following rules:
(20) Yer lowering: Lower a Yer in a syllable preceding another Yer (I/U à e/o / __C0 I/U)
(21) Yer deletion: Delete any non-lowered yers.
When multiple Yers are evidenced in a row in Russian, we can see that the rule of Yer-Lowering
applies from left-to-right, as in the following derivation of denëček ‘day-diminutive-diminutive-
nominative’.
(22) Underlying [[[[dIn] Ik] Ik] U]
Cycle 1 ---
Cycle 2 e
Cycle 3 e
Cycle 4 e
Yer-Deletion Ø
Other rules denëček
(Pesetsky 1979:7)
This pattern has been reanalyzed within the CVCV framework promoted here. In Scheer (2004),
Scheer and Ziková (2010), and Ziková (2008) it is proposed that Yers are floating vowels. The
15
Here when I say prefixed verbs or particle verbs, I refer to lexical (not supralexical) constructions in Slavic, and to
separable (not fused) constructions in Germanic. The literature on different kinds of particle verbs is large and the
variation in the behaviour of distinct types of particles is beyond the scope of this paper.
- 16 melody
of these alternating vowels will link to the CV tier only if ungoverned (if they are not
followed by an overt vowel in the cycle/phase in which they are interpreted). Non-alternating
vowels, such and the NOM.SG suffix in (22), are proposed to contain no melodic content, while
alternating vowels are proposed to be mid-vowels underlyingly. Unlinked vowels will not be
pronounced. In (23), underlyingly unlinked vowels are grey.
(23) Underlying [[[[den] ek] ek] Ø]
Cycle 1 den
Cycle 2 ek
Cycle 3 ek
Cycle 4 Ø
Other rules denëček
If vowel linking were applied from Right-to-Left in the fully formed noun, the predicted form
would be the ungrammatical *denček (*denkk in Pesetsky’s lowering analysis).16
The relevant
paradox comes into view when considering prefixed verbs. There, the realization of the vowel in
the prefix varies depending on whether the following syllable contains a pronounced vowel (see
also Matushansky 2002 and Gribanova 2012 for discussions of the morpho-phonology of particle
verbs in Russian).
(24) a. podo-žeg-l-Ø à podžëg ‘set fire’
under-burn-past-masc.
b. podo- žeg-l-a à podožgla ‘set fire’
under-burn-past-fem.
Note that the output of (24a) is distinctly not the pattern that occurs in (22/23) where all
unlinked/alternating vowels (except the last) are realized. The attested pattern is predicted if the
prefix is merged after (structurally outside of) the masculine suffix, and the derivation proceeds
cyclically as in (25)17
.
(25) Underlying [podo[[žeg] l-Ø]]
Cycle 1 žeg
Cycle 2 žeg-l-Ø
Cycle 3 podožeg-l-Ø
Other rules podžëg
16
See Ziková (2008) and Scheer and Ziková (2010) for a discussion of the Havlík pattern of Yer realization where the
Old Czech form of the double diminutive shows the pattern predicted by right-to-left government (domček ‘house-
DIM-DIM’).
17
Matushansky (2002) proposes that the prefix is marked as non-cyclic, causing it to be ignored by cyclic rules like
Yer-Lowering. As argued in the previous sections, this analysis is diacritic and non-modular and therefore cannot be
adopted here. We will see that there is a fully modular explanation for the behavior of these morphemes. Note also
that Cycles 1 and 2 in (25) may be collapsed.
- 17 The
problem here is that the prefix is decidedly not merged outside/after the final -Ø ‘MASC’.
Firstly, the semantic interpretation of particle verbs, including this one, may be idiomatic. Here
the meaning ‘set fire’ is not predictable from the simple composition of the meanings of the particle
and the verb. It is generally held that all members of an idiom must be found within a certain
domain, at the phrasal level, proposed to be the vP (Marantz 1984b, 1997). Also, the syntactic
literature on particle verbs has converged upon the conclusion that the particle heads a PP/Small
Clause complement to the verb, as in (26). For the detailed syntactic motivations behind this type
of analysis see Wurmbrand (1998), Taraldsen (2000), Ramchand and Svenonius (2002), Svenonius
(2004), Caha and Ziková (2016) and references cited within.
(26) VP
3
V0
RP
g 3
throw R0
pP
3
DP p
# 3
the dog p0
PP
3
P0
DP
g #
out (the door)
(Svenonius 2004: 222)
It is clear in many languages that the particle does not remain in-situ (hence the possibility of
reordering within the VP in (26); throw out the dog). In Slavic languages these particles are always
preverbal and cannot be separated from the verb. There have been two different types of proposal
in the literature explaining how the Slavic particles come to be pronounced in this position.18
The
first is that they raise via head-movement, as in (27).
(27) a. Samoljot pere-letajet granicu
plane across-flies border
‘The plane is flying across the border’ (Russian)
(Svenonius 2004:220)
18
The motivation for movement of the particle is proposed to be the necessity of the null ground operator (perhaps
incorporated into the particle itself, to scope over and bind a variable in Asp(ect)P, a phrase that sits above VP (see
30). This is proposed by Svenonius (2004) to explain the perfectivizing effect of particles. Note that in Germanic
languages the particle may be separated from the verb by the object (English : threw the dog out) by morphology
(German: Part-zu-Verb) or by V2 movement (German: AUF hat Peter die Tür gemacht (nicht zu) ‘Peter has opened
the door’ (Wurmbrand 1998:272)). This behaviour is consistent with an XP movement account of particles, to be
supported in this section.
- 18 b.
VP
3
V0
RP
2 2
R0
V0
tfigure R
g g 2
pere- let- tR DP
‘across’ ‘fly’ @
granicu
‘border’
(Svenonius 2004:223 (analysis not adopted by Svenonius, but see Rojina 2004))
The second proposal is that particles come to precede the verb via phrasal movement, as in (28).
(28) AspP
3
PP AspP
3
vP Asp
3
VP v
3
PP V
(Caha and Ziková 2016: line 442)
Wurmbrand (1998) for German, Svenonius (2004) for Russian, Taraldsen (2000) for Norwegian,
and Caha and Ziková (2016) for Czech give arguments that the particle, even in Slavic languages
where it is never separated from the verb, moves to the left of the verb via XP movement. The
evidence for this for Czech in Caha and Ziková (2016) is especially striking.19
Caha and Ziková (2016), following (Ziková 2012; Scheer 2001), note that there is a
difference in the vowel length of the prefix in Czech depending on whether it finds itself within
the scope of an aspectual head. If it is base-generated below Asp, the particle must raise (to scope
over and bind the perfectivity-inducing variable in Asp0
(following Svenonius 2004)). If there is
no Asp0
in the structure there is no trigger for raising and the prefix stays low (ex. in root VP
nominalizations). If the particle remains in the VP, its vowel is spelled-out as long (length is
indicated by an acute accent), and if it raises its vowel is short.
(29) a. Petr na-psal dopis b. ná-pis
Petr on-wrote letter on-write
‘Peter wrote the letter’ ‘a sign’
(Caha and Ziková 2016: line 76)
19
Note that the object of the particle must evacuate the Particle Phrase before XP movement or the object would
always intervene between the particle and the verb (Ora Matushansky p.c.). The XP movement of the Particle Phrase
is therefore an instance of Remnant Movement as argued for in (e.g.) Nkemnji (1995), Müller (1996), and Koopman
(1996), among others.
- 19 (30)
a. AspP b. n/aP
3 3
PP AspP VP n/a
@ 3 3
na vP Asp PP V
3 g g
VP v P pis
3 g write
PP V ná
@ g on
ná pis
(Caha and Ziková 2016: line 329)
Caha and Ziková demonstrate this correlation very clearly with reference to verbal, participial, and
‘high-nominalization’ (outside of vP) constructions (short vowel) versus adjectives and ‘low
nominalization’ (inside vP) constructions (long vowel). All of the constructions with short vowels
are shown to be perfective, hence containing an AspP, while the constructions with long vowels
are not perfective, indicating the absence of AspP; the trigger for PP raising. The motivation for
vowel lengthening is proposed to be templatic.20
Inside a PWd (inside nP/aP) the vowel will be
long and outside the PWd (in AspP) the vowel will be short.21
Importantly for the discussion of
the syntactic structure, Caha and Ziková argue that this spell-out distinction is only possible if the
particle has escaped spellout in the domain of vP (as in 30a) or remained in the spell-out domain
of an nP (as in 30b). Caha and Ziková’s analysis aligns with the fact that the particle is not in the
same phonological domain (standardly considered the PWd) as the verb in constructions
containing AspP, aligning it with the analysis of Svenonius (2004), where XP-movement of the
PP to Spec,AspP is responsible for the aspectual effects of lexical particles in Slavic languages. If
the particle were to move via head-movement, this distinction would not be predicted. Morphemes
that are co-members of a complex head are almost uniformly considered to be spelled out in the
same cycle/PWd.
3.5.1 Particle Verb Bracketing Paradoxes resolved
Based on the evidence from Caha and Ziková’s analysis of Czech, from Svenonius’ arguments for
Russian, from the separability of particles and verbs in Germanic languages, and from a host of
other data that indicate a cross-linguistic pattern whereby particles are interpreted separately from
the verbs with which they combine, we can argue that XP movement is the means by which
particles are spelled out to the left of the verb. This will allow for a uniform analysis of the crosslinguistic
particle verb BPs exemplified (but by no means exhausted) by the following data.
First let us consider the Russian BP. Here the particle will project a PP in the complement of vP.
20
Scheer (2001) proposes that the prefix vowel alternations are due to lengthening, while Ziková (2012) and Caha and
Ziková (2016) argue that the alternation must be due to vowel shortening. Regardless of which analysis is correct, the
two distinct behaviours are clearly linked to the presence or absence of perfective aspect.
21
Another option for explaining the vowel-length difference in these examples might be that the vowel is spelled out
as short if interpreted alone (when moved, following Johnson 2004’s theory of numerphology), and long if spelled out
in combination with another morpheme. A correlation has been noted in the literature between the licensing of long
vowels and diphthongs and the presence of a following vowel (see Lowenstamm 1996, Kaye 1990). This distinction,
however, could not be generalized to all long and short vowels in Czech.
- 20 (31)
CP
3
AgrP22
C
3
TP Agr
3 Ø
AspP T
3 l
PP AspP
@ 3
podo vP Asp
3				žeg-Ø-Ø
VP v
3 žeg-Ø
PP V
@ g
podo žeg
In the derivation above, the particle will be spelled out in PP, assuming PP is a phase. Even if PP
were not a phase, it would undergo spell-out upon movement to Spec,AspP. Remember that leftbranches
(adjuncts, as well as specifiers) must undergo interpretation before (re)merger into the
tree according to Uriagereka (1999) and Johnson (2004). The output of the first cycle of
interpretation is therefore (32). In (32), the final Yer is unlinked/floating as it is final in the string.
Recall that in Section 2 we saw that Final Empty Nuclei (FEN) are parametrically licenced. As
Russian allows consonant-final words, it is clear that this parameter is active in the language.
The following cycle will see the spell-out of the verb and its suffixes. Gribanova (2013) argues
that the verb in Russian moves to a position above vP and below TP; AspP in (31). As the verb has
raised out of vP it will undergo spell-out in the CP phase with its suffixes (the first and second
cycles in (25) are thus collapsed). Here the output of the CP phase will be as in (33).
In (33) the final Yer remains unpronounced for the reasons stated above, and the l deletes (deletion
is indicated in grey) for independent reasons. As the root vowel is ungoverned, it must be linked
and pronounced. Subsequent to this cycle of spell-out the particle and verb are linearized (34).
22
AgrP is included here for expository purposes. Should agreement not project in the tree, but rather emerge due to
checking relations in the syntax, nothing will change about the analysis herein. (see Embick (1997) on dissociated
morphemes).
(32) C V C V
… | | |
p o d o
(33) C V C V C V
… | | | |
ž e g Ø l Ø
(34) C V C V C V C V C V
… | | | | | | | |
p o d o ž e g Ø l Ø
- 21 In
(34) the root-Yer has been linked to the CV tier prior to the linearization of the particle and the
verb (33). The full vowel of the root therefore governs the final vowel of the particle, allowing it
to remain unlinked and unpronounced.23
Of note here is the distinction in edge-marking between Russian and the English examples
we saw above. Russian phase edges will not be marked by an empty CV; there is no empty CV
between the particle and the verb. This lack of marking is independently supported in Lowenstamm
(1999) and Scheer (2009a). Initial CVs are supported only in languages whose onsets uniformly
conform to the Sonority-Sequencing Principle. In onsets that rise in sonority, the relationship of
Infrasegmental Government (IG) applies (Scheer 2004, Scheer and Cyran 2018). IG allows for a
sonorant to govern a preceding obstruent consonant (indicated by the straight, grey arrow in (35)).
This permits the cluster-internal vowel to remain unpronounced. This unpronounced vowel will
be licensed by the full vowel to its right (indicated by the dashed arrow in (35)), and this, in turn,
allows the cluster-internal vowel to govern (and suppress) the vowel of the initial empty CV. For
more details on the workings of IG, see the sources cited.
In languages where sonorants are not uniformly in a position to Infrasegmentally Govern an
obstruent in complex onsets (e.g. Russian), the vowel separating the two consonants must be
governed by the following overt vowel (and not by IG). The cluster-internal vowel is therefore
unlicensed, leaving no governor for an empty vowel in an initial CV. As ungoverned vowels must
be pronounced, and as no word-initial vowels are epenthesized, there can be no initial CV in such
languages.
Therefore, when the particle and verb are linearized in (34), the root vowel of the verb is local to
the final vowel of the particle, allowing for Government to operate between them. This explains
why the realization of the prefix Yer is dependent on the pronunciation of the following vowel. In
derivations where the following vowel is not pronounced, as in the feminine podožgla the particle’s
Yer is ungoverned (as well as non-final) after linearization, and therefore must be pronounced.
23
A reviewer notes that on purely phonological grounds, the prefix may spell-out for the first time after the CP-cycle
(žeg-l-Ø) has undergone phonological interpretation. Although this derivation also works, the examination of similar
cross-cyclic phonological behaviour (ex. liaison in French), as well as the theoretical proposals in work such as
Uriagereka (1999) and Johnson (2004), point towards derivations where the spell-out of syntactic left-branches occurs
before the structure linearized to its right is visible.
(35) C V C V C V
… | | | | | |
Ø Ø p Ø l e
(36) * C V C V C V
… | | | | | |
Ø Ø p Ø s e
(37) C V C V C V C V C V
… | | | | | | | | | |
p o d o ž e g Ø l a
- 22 This
configuration also explains why phonological operations are not blocked in general across
the particle-verb boundary in Russian. Particle-final consonants in Slavic languages are not
devoiced; they are not domain-final. Vowels in particles can affect stress placement/resist
unstressed vowel reduction in pretonic position; they are stressed during spell-out prior to
linearization with the verb and a stressed vowel is not a target for pretonic vowel reduction. It is
also the case that prefix-final consonants in Russian trigger retraction of following front vowels
instead of undergoing palatalization. Retraction applies to C[+back]V[-back] sequences that are
linearized post-cyclically. The members of a C-V sequence spelled out within a cycle will be
adjacent on the CV tier, while a C-V sequence that is linearized post-cyclically will be separated
by a VC sequence; in CVCV phonology each domain must begin with a C and end with a V on
the CV tier. Prefix-verb C-V sequences are therefore actually C[+back]VCV[-back] sequences, while
verb-suffix C-V sequences, spelled out in a single cycle, are simply C[+back]V[-back]. Exactly how
the empty CV positions lead to retraction rather than palatalization is beyond the scope of this
paper, but it is clear that the predictions of the system argued for here lead to two distinct
phonological derivations and structures depending on the timing of spell-out of these strings. The
same pattern will also explain why hiatus is not resolved across a particle-verb boundary (/po +
obedatj /à [poobedatj] (*pobedatj) ‘to have lunch’ (Gribanova 2008:224), as hiatus resolution is
demonstrably a cyclic rule. This means that, like for the discussion of C[+back]V[-back] sequences, a
segmental sequence of two vowels interpreted within a cycle will have a different phonological
structure than the same segmental sequence that is linearized after each vowel has been interpreted
in a separate cycle. One concrete proposal, mentioned above, is that (some or all) suffix-initial
vowels in Slavic languages are floating, as argued in Scheer and Ziková (2010), Ziková (2008,
2009), and for English in Section 2 and in Newell (2017c). Importantly, in none of these instances
do we need to appeal to hierarchical phonological structure to distinguish distinct domains of
application for these phonological operations.
The paradox related to particle verbs can now be resolved by stating, following Ramchand
and Svenonius (2002), that the idiomatic interpretation of particle verbs comes about in the same
way that other vP idioms do (see Marantz 1984b, 2007; Biskup 2019 for vP domain restrictions on
clausal idioms like He kicked the bucket). The phonological separateness of particles and verbs in
many languages is due to the need for the PP to raise into Spec,AspP. Some languages allow for
further syntactic operations that separate the particle and the verb (ex. Germanic) and others do
not (ex. Slavic). Derivations like the preceding can easily be extended to explain other particleverb
phonological patterns cross-linguistically, including but not limited to, Hungarian vowel
harmony (which does not cross the particle-verb domain (38)), Warlpiri vowel-harmony, stressassignment,
and verb conjugation class which is sensitive to the domain of the verb and not to the
particle (39), and compound stress patterns and the separability of the particle verb in German
(40). In these languages, onsets obey the sonority sequencing principle, and therefore cyclic spellout
ensures that there is an empty CV sequence separating the phonological domains of the particle
and the verb.24
(38) a. át-lép-és b. le-tartóztat-ás
across-step-dev down-hold-dev
‘transgression’ ‘arrest’
(front/back harmony: Kenesai 1995/1996:158)
24
See also Biskup, Putnam and Smith (2011) for a discussion of verb-particle spell-out.
- 23 (39)
a. pirri-kuju-rnu b. pirri-kiji-rni
preverb-throw-PAST preverb-throw-nonPAST
‘to scatter ' ‘to scatter '
(i/u vowel harmony: Nash 1980:140)
(40) a. áuf-gèben c. ver-gében
part-give prfx-give
‘give up’ ‘forgive’
(particle vs. prefix stress: Wurmbrand 1998)
To conclude this section, particles are interpreted separately from the verb with which they
combine due to the fact that they undergo phrasal movement out of the vP. This movement ensures
that particles undergo spell-out separately from the verbs that select for them. Subsequent to
movement, particles will be treated as separate phonological domains in languages that mark the
left edge of a phase with an empty CV (ex. German, Warlpiri), but may interact with the
phonological domain of the verb in languages that do not have the ability to CV-mark their
phonological domains (ex. Russian, Czech), although this interaction will be sensitive to
operations that have taken place in the independent phonological cycles that occur prior to the
linearization of the prefix and verb. Left branches (adjuncts and specifiers) will be spelled-out
separately from the structures they combine with. This generalization over spell-out domains, the
independently supported proposals that particles undergo XP movement (Wurmbrand 1998,
Taraldsen 2000, Svenonius 2004, Caha and Ziková 2016) and a flat theory of phonology combine
to erase any paradoxical nature of particle verb constructions. The question of why there is a
general tendency for particles to cliticize to verbs can be subsumed under the question of why
phonological clitics behave as dependents in general; a question that is beyond the scope of this
article (See Svenonius 2016 for a recent proposal).
3.6 Reduplication Bracketing Paradoxes
Finally, let us turn to the discussion of a reduplication paradox presented in Marantz (1987), taken
from Odden and Odden (1985). This section will demonstrate how the analyses of BPs in the
sections above can be easily extended to cover the Kihehe-type domain mismatches in Full
Reduplication as exemplified in (41) below.25
(41) a. ku-tova-RED à ku-tova-tova ‘to beat a bit’
Inf-beat-RED
b. ku-íita-RED à kwíita-kwíita ‘to pour a bit’
Inf-pour-RED
c. kú-lu-íita-RED à kú-lwiita-lwíita ‘to pour it a bit’
Inf-it-pour-RED
25
C.f. the unreduplicated forms found in Odden and Odden (1985):
i) kú-mu-tóva ‘to hit him’ (499) ii) kw-íita ‘to spill’ (499)
iii) kú-lw-iíta ‘to pour it’ (499) iv) neléke ‘may I cook’ (498)
- 24 d.
n-teléka-RED à neleka-neleka ‘I will cook a bit’
1SG-cook-RED
(Marantz 1987)
In (41a) we see the output of a reduplicated consonant-initial verb. It is clear that the reduplicative
marker scopes outside the domain of the verb (copied), and under the infinitive (not copied). Yet,
as pointed out by Marantz, there are two environments where an outer affix will apparently ‘tuck
in’ under the reduplicative morpheme, even though this causes a mismatch between the
phonological and morpho-syntactic domains. In (41b,c) we see constructions with V-initial verbs
following prefixes that end in a high round vowel. This vowel (together with a preceding
consonant) will syllabify in the onset of a following V-initial morpheme (as a C-glide onset). In
(41d) the 1SG prefix is underlyingly a floating nasal feature. This feature associates to the
segmental structure of the consonant to its right. In cases where gliding or featural association
combine with reduplication the affix that syntactically scopes over the reduplicative morpheme
behaves as though it sits within the reduplicative domain. Note that the reduplicated domain is the
stem, not the verb root (final vowels and other suffixes are reduplicated if present).26
In (42) I
suggest that the RED affix sits in AspP and the Infinitival prefix in TP. These labels are not crucial.
It is also not crucial whether the 1SG marker scopes above or below RED, as unlike the u-final
prefixes, the behaviour of the nasal feature is not phonologically variable. The nasal associates
with the onset position to its right regardless of the featural content of that position. What is crucial
for an explanation of the variable behaviour of the u-final prefixes is that RED scopes over the vP,
and under the position of the Infinitival marker.
(42) CP
3
TP
3
T AspP
ku 3
vP Asp
#							RED(moderative)
tova/íita
Marantz argues that “The Kihehe reduplication involves a morpheme unit, not some unit that is
phonologically definable independent of the stem.” (204). He proposes (along with Sproat 1985)
that BPs are caused by a phonologically-induced rebracketing, as in (43b)
(43) a. Phonological structure isomorphic with (42) b. Rebracketed structure
3 3
ku					3 3 RED
					tova/íita					RED ku íita
	
26
One example of many in Odden and Odden (1985) is the following.
i) sa-kúu-novagi-nov-ag-i 'he never hits me' (root -tov-)
neg.3sgSubj.-asp-redupl(lsg obj)-hit-habit-FV (502)
- 25 Marantz
proposes that adjacency at phonological structure is associative. This entails that [ku[íita
RED]] is equal to [[ku íita] RED]. The requirement in Kihehe that high-round segments syllabify
as onsets if possible is what triggers this re-bracketing in the case of vowel-initial stems. Therefore,
at spell-out, RED will scope over the relevant prefixes, explaining their inclusion in the
reduplicative domain.
The translation of Marantz’ solution into a derivation that functions in the linear system
proposed here is not difficult. First, we must consider the evolution of syntactic theory, and the
subsequent effect of phases on phono-syntactic relations. Assuming the verb stem is the spell-out
of vP, we can see that Marantz was mistaken in his statement to the effect that the domain of
reduplication is not definable in the phonology. The reduplicated domain is the entire phonological
string that is the output of the first phase.
(44) vP
#								à		 tova/íita
tova/íita
Next, we have to define how the RED morpheme targets this domain for doubling. In a
modular system, phonological reduplication cannot target the morpho-syntactic node that
dominates tova or íita, as phonological outputs cannot contain morphological information. RED
can only target a phonological object, as its copy-function is operative over the phonological output
of the previous phase. But, Marantz is also correct that the RED morpheme cannot be of a specific
phonological size (syllable, foot), as the size of the reduplicated domain tracks the size of the stem.
This type of pattern is easily captured within the system of reduplication found in Raimy (2000).
According to Raimy, reduplication adds a loop to the linearization algorithm of a string of
segments. Licit targets for the beginning and end of a loop include first and last segments,
consonants, vowels, etc. Full reduplication, as seen in Kihehe, is the insertion of a loop that
originates at the final segment and terminates at the first segment of a phonological domain.
According to this type of system, the output of the vP phase will be a strictly linearized string as
in (45), where # signals the beginning and % the end of the string.
(45) #àtàoàvàaà% or #àíàiàtàaà%27
If we translate this into a CVCV framework and insert an empty initial CV to mark the edge of the
domain, we get the following outputs of vP at PF
(46) a. # à C à V à C à V à C à V à %
| | | | | |
Ø Ø t o v a
b. # à C à V à C à V à C à V à C à V à %
| | | | |
Ø Ø Ø i t a
27
Note that Raimy’s # and % are not elements in the string like # and + were in SPE. Here they are included solely
to clarify the beginnings and endpoints of strings.
- 26 In
the CP phase in (42), the RED and infinitival affixes will be spelled out. RED, as stated above,
will insert a loop from the last to the first segment of the domain in its scope (here to its left, as
RED in Kihehe is suffixal). Note that this loop targets positions on the CV-tier (and not melodic
nodes), and therefore includes the initial CV inserted at the spellout of vP.
(47) a. # à C à V à C à V à C à V à %
| | | | | |
Ø Ø t o v a
b. # à C à V à C à V à C à V à C à V à %
| | | | |
Ø Ø Ø i t a
The infinitival (or person agreement/object clitic) affix will then be inserted. The u of the prefix
(or the [nasal] feature) will scan the melodic tier (48a). If there is a vowel to its right, the prefixal
melodic material will merge into the appropriate domain (48b). In a case where the environment
for merger is not met, the prefix will sit outside of the domain to its right (49). The predicted
phonological outputs are obtained. The prefix is included in the previously-determined domain of
the RED loop iff the output of the vP phase begins with a vowel on the melodic tier.28
No change
is made in the phonological representation of the domain of reduplication. The output of the
phonological computation at the CP phase will, however, reflect whether the prefix is internal or
external to the domain of RED.
(48) a. # C à V à C à V à C à V à C à V à C à V à C à V à % #
| | | | | | | | |
Æ Æ k u Æ Æ Æ i t a
à ku-íita-íita
b. # C à V à C à V à C à V à C à V à C à V à C à V à % #
| | | | | | | | |
Æ Æ Æ Æ k Æ w i t a
à kwíita-kwíita
28
Extra empty CVs (as in (48b) may be deleted in the final representation, but they are maintained below for ease of
exposition.
(49) # C à V à C à V à # à C à V à C à V à C à V à % #
| | | | | | | | | |
Æ Æ k u Æ Æ t o v a
à ku-tova-tova
- 27 In
this way, the phonological output of the CP phase is not inconsistent with the morpho-syntactic
bracketing. As seen in the previous sections, it is in fact impossible for a bracketing mismatch to
occur in a linear phonological system.29
4. Conclusion
In this paper I have argued for a fully modular derivational system, where only strictly
phonological objects and operations are present in the phonological module of each derivation.
Hierarchical phonological structures, if included in the derivation, only serve to create Bracketing
Paradoxes that would otherwise not arise. Once these BPs are admitted into the system, then they
must be repaired. Over the last 40 years, linguists have been proposing ad-hoc operations in various
attempts to rid the system of BPs. I have demonstrated here that, using proposals that have been
independently motivated; phases, the independent spell-out of adjuncts and specifiers,
linearization, lateral relations between segments, CVCV phonology, liaison, and modularity, we
have a system in which all the types of Bracketing Paradoxes that have troubled morphophonologists
over the last decades do not arise, and therefore do not need to be repaired. A system
that doesn’t produce impossible structures is clearly preferable to one that does, and this distinction
should inform our global evaluation of possible phonological representations.
References
Allen, Margaret R. 1979. Morphological investigations. University of Connecticut dissertation.
Arad, Maya. 2003. Locality constraints on the interpretation of roots: The case of Hebrew
denominal verbs. Natural Language and Linguistic Theory 21(4). 737-778.
Beard, Robert. 1991. Decompositional composition: The semantics of scope ambiguities and
‘bracketing paradoxes’. Natural Language and Linguistic Theory 9(2). 195-229.
Benua, Laura. 1995. Identity effects in morphological truncation. University of Massachusetts
Occasional Papers in Linguistics 18: Papers in Optimality Theory. Graduate Linguistic
Student Association, Amherst, Massachusetts. 77-136.
Bermúdez-Otero, Ricardo 2011. Cyclicity. In Marc van Oostendorp, Colin Ewen, Elizabeth
Hume and Keren Rice (eds.) The Blackwell Companion to Phonology, 2019-2048. Malden,
MA: Wiley-Blackwell.
Bermúdez-Otero, Ricardo. 2013. The stem-level syndrome. Paper presented as the UPenn
Linguistics Department Speaker Series. University of Pennsylvania.
Bermúdez-Otero, Ricardo. 2017. Stratal phonology. In S.J. Hannahs and Anna Bosch (eds.) The
Routledge Handbook of Phonological Theory, 114-148. Routledge.
29
Marantz discusses two other examples of reduplication BPs. The first is a Tagalog reduplication pattern that interacts
with an operation of root truncation. It is clear from Carrier (1979) that these examples are cases of allomorphy, and
not phonological derivation. The second is a case from Mende where initial consonant mutation occurs in certain
syntactic environment (generally when the word in question follows its complement). There is a debate in the literature
over whether the Mende case is triggered by a morpheme or by a strictly phonological environment (Conteh, Cowper
and Rice 1986; Cowper and Rice 1985, 1987; Lieber 1983, 1987; Rice and Cowper 1984; Seidl 2001; Iosad 2008,
2010). In either case, there is a requirement that the trigger of mutation be directly to the left of the relevant consonant,
indicating that a linear phonological analysis is feasible. Evaluating the various analyses of Mende mutation cannot
be accomplished herein.
- 28 Bermúdez-Otero,
Ricardo. 2019. Challenges to Stratal Phonology. Paper presented as part of a
Brugmann Fellow course. Session 2 handout., IGRA, Leipzig, July.
Biskup, Petr. 2019. Prepositions and Verbal Prefixes: The case of Slavic. John Benjamins.
Biskup, Petr, Michael Putnam, and Laura C. Smith. 2011. German particle and prefix verbs at the
syntax-phonology interface. Leuvense Bijdragen 97. 106-135.
Bobaljik, Jonathan D. 2000. The ins and outs of contextual allomorphy. University of Maryland
working papers in linguistics 10. 35-71.
Bobaljik, Jonathan D. and Susi Wurmbrand. 2013. Suspension across domains. In Ora
Matushansky and Alec Marantz (eds.) Distributed Morphology today: Morphemes for
Morris Halle. MIT Press. 185-198.
Booij, Geert and Rochelle Lieber. 1993. On the simultaneity of morphological and prosodic
structure. In Sharon Hargus and Ellen M. Kaisse (guest eds.) Phonetics and
Phonology: Studies in lexical phonology 4. 23-44.
Bruening, Benjamin. 2014. Word formation is syntactic: adjectival passives in English. Natural
Language and Linguistic Theory 32. 363-422.
Bruening, Benjamin. 2017. Consolidated morphology: A non-distributed, purely syntactic theory
of morphology. Newark: University of Delaware. ms. Online: http://udel. edu/~
bruening/Downloads/ConsMorphosyntax1. pdf.
Caha, Pavel and Markéta Ziková. 2016. Vowel length as evidence for a distinction between free
and bound prefixes in Czech. Acta Linguistica Hungarica 63(3). 331-377.
Carnie, Andrew. 1991. A domain based phonology: The evidence from Modern Irish.
In Proceedings of the Leiden Conference for Junior Linguists 3. 59-76.
Carrier, Jill Louise. 1979. The interaction of morphological and phonological rules in Tagalog: a
study in the relationship between rule components in grammar. Massachusetts Institute of
Technology dissertation.
Chierchia, Gennaro. 1986. Length, syllabification and the phonological cycle in Italian. Journal
of Italian Linguistics 8. 5-34.
Chomsky, Noam. 2001. Derivation by Phase. In Michael Kenstowicz (ed.) Ken Hale: A Life in
Language. 1-52. MIT Press.
Chomsky, Noam and Morris Halle. 1968. The Sound Pattern of English. New York: Harper and
Row.
Cinque, Guglielmo. 1993. A null theory of phrase and compound stress. Linguistic Inquiry 24(2).
239-297.
Conteh, Patrick, Elizabeth Cowper, and Keren Rice. 1986. The environment for consonant
mutation in Mende. In Gerrit Dimmendaal (ed.) Current issues in African linguistics 3.
107–116. Dordrecht: Foris.
Cowper, Elizabeth and Keren Rice. 1985. The destruction of tonal structure in Mende. Studies in
African Linguistics: Précis from the 15th
conference on African Linguistics, supplement
9. 57-62.
Cowper, Elizabeth and Keren Rice. 1987. Are morphosyntactic rules necessary? Phonology
Yearbook 4. 185-194.
De Clercq, Karen, 2013. A unified syntax of negation. Ghent University dissertation.
Devine, Andrew and Laurence Stephens. 1976. The Function and Status of Boundaries in
Phonology. In Alphonse Juilland (ed.) Linguistic Studies offered to Joseph Greenberg on
the occasion of his sixtieth birthday. 285-312. Saratoga, Cal.: Anma Libri.
- 29 Downing,
Laura. J., T. Alan Hall, and Renate Raffelsiefen. (eds.). 2005. Paradigms in
phonological theory. Oxford University Press on Demand.
Embick, David. 2014. Phase cycles, φ-cycles, and phonological (in) activity. In Sabrina
Bendjaballah, Noam Faust, Mohamed Lahrouchi, and Nicola Lampitelli (eds.) The form
of structure, the structure of forms: essays in honor of Jean Lowenstamm. 271-86. John
Benjamins Publishing Company.
Embick, David. 2010. Localism versus globalism in morphology and phonology. Linguistic
Inquiry Monograph 60. Cambridge. MIT Press.
Embick, David. 1997. Voice and the interfaces of syntax. University of Pennsylvania
dissertation.
Encrevé, P. 1983. La liaison sans enchaînement. Actes de la recherche en sciences sociales,
46(1). 39-66.
Fábregas, Antonio. In press. Morphologically Derived Adjectives in Spanish. Amsterdam: John
Benjamins.
Fodor, Jerry A. 1983. The Modularity of Mind. Cambridge Mass. MIT Press.
Fodor, Jerry A. 1985. Precis of the modularity of mind. Behavioral and brain sciences, 8(1). 1-5.
Giegerich, Heinz J. 1999. Lexical strata in English: Morphological causes, phonological effects
Cambridge Studies in Linguistics 89. Cambridge University Press.
Gribanova, Vera. 2008. Russian prefixes, particles and palatalization in Stratal OT. In Charles B.
Chang and Hannah J. Haynie (eds.) Proceedings of the 26th West Coast Conference on
Formal Linguistics. 217-225. Somerville, MA: Cascadilla Proceedings Project.
Gribanova, Vera. 2012. Connecting cyclic morphosyntax and morphophonology: Russian
Bracketing Paradoxes. Paper presented at Exploring the Interfaces 1: word structure.
McGill University.
Gribanova, Vera., 2013. Verb-stranding verb phrase ellipsis and the structure of the Russian verbal
complex. Natural Language and Linguistic Theory 31(1). 91-136.
Halle, Morris., and Marantz, Alec. 1994. Some key features of Distributed Morphology. In
Andrew Carnie and Heidi Harley, with Tony Bures (eds.) MIT working papers in
linguistics: Papers on Phonology and Morphology 21. 275-288.
Harley, Heidi. 2009. Compounding in distributed morphology. In Rochelle Lieber & Pavol
Štekauer (eds.), The Oxford handbook of compounding, 129–144. Oxford/New York:
Oxford University Press.
Haspelmath, Martin. 2011. The indeterminacy of word segmentation and the nature of
morphology and syntax. Folia Linguistica 45(1). 31-80.
Haugen, Jason D. and Daniel Siddiqi. 2016. Towards a Restricted Realization Theory. In Daniel
Siddiqi and Heidi Harley (eds.) Morphological metatheory: Linguistics Today 229. 343-
386. John Benjamins.
Hyman, Larry. 1978. Word Demarcation. In Joseph Greenberg (ed.) Universals of Human
Language 2. 443-470. Stanford: Stanford University Press.
Inkelas, Sharon. 1998. The theoretical status of morphologically conditioned phonology: a case
study of dominance effects. Yearbook of Morphology 1997. 121-155.
Iosad, Pavel. 2010. Right at the left edge: Initial consonant mutations in the languages of the
world. In Jan Wohlgemuth and Michael Cysouw (eds.) Rethinking Universals: How
rarities affect linguistic theory. 105-137. De Gruyter.
Iosad, Pavel. 2008. Initial consonant mutation and information flow in Mende. Paper presented
at the Old World Conference in Phonology 5, Université de Toulouse–Le Mirail.
- 30 Johnson,
Kyle. 2004. Towards an etiology of adjunct islands. Nordlyd, 31(1). 187-215.
Julien, Marit. 2002. Syntactic heads and word formation. Oxford University Press.
Kaisse, Ellen. 1985. Connected speech: the interaction of syntax and phonology. New York:
Academic Press.
Kalivoda, Nicholas. 2018. Syntax-prosody mismatches in Optimality Theory. UC Santa Cruz
dissertation.
Kang, Beom-mo. 1993. Unhappier Is Really a" Bracketing Paradox". Linguistic inquiry 24(4).
788-794.
Kaye, Jonathan. 1990. Coda licensing. Phonology, 7(1). 301-330.
Kenesei, István. 1995/1996. On Bracketing Paradoxes in Hungarian. Acta Linguistica Hungaria
43(1/2). 153-173.
Kenstowicz, Michael and Charles Kisseberth. 1977. Topics in Phonological Theory. New York:
Academic Press.
Kiparsky, Paul. 1982a. Lexical morphology and phonology. Linguistics in the morning calm:
Selected papers from SICOL-1981. 3-91. Hanshin Pub. Co.
Kiparsky, Paul. 1982b. Word-formation and the lexicon. Paper presented at the Mid-America
Linguistics Conference. University of Kansas.
Kiparsky, Paul. 2000. Opacity and cyclicity. The Linguistic Review 17(2-4). 351-367.
Kiparsky, Paul. 2005. Paradigm uniformity constraints. Ms. Stanford University.
www.stanford.edu/~kiparsky/Papers/LexConservatism.pdf .
Koopman, Hilda. 1996. The spec-head configuration. In Edward Garrett and Felicia Lee (eds.)
Syntax at Sunset 1: UCLA Working Papers in Syntax and Semantics. 37-65.
Larsen, Bergeton Uffe. 1998. Vowel length, Raddoppiamento Sintattico and the selection of the
definite article in Italian. In Patrick Sauzet (ed.) Langues et Grammaire II-III,
Phonologie. 87-102. Paris: Université Paris 8.
Lieber, Rochelle. 1980. On the organization of the lexicon. MIT dissertation.
Lieber, Rochelle. 1983. New developments in autosegmental phonology: Consonant mutation. In
Proceedings of the 2nd
West Coast Conference on Formal Linguistics. Stanford.
Lieber, Rochelle. 1987. An Integrated Theory of Autosegmental Processes . State University of
New York Press.
Lieber, Rochelle. 1992. Deconstructing morphology: Word formation in syntactic theory.
University of Chicago Press.
Lowenstamm, Jean. 1996. CV as the only syllable type. In Jacques Durand and Bernard Laks (eds.)
Current trends in phonology: Models and methods 2. 419-441. European Studies Research
Institute (ESRI), University of Salford: Manchester UK.
Lowenstamm, Jean. 1999. The beginning of the word. In John R. Rennison & Klaus
Kühnhammer (eds.) Phonologica 1996: Syllables !?. 153-166. The Hague: Thesus.
Marantz, Alec. 1984a. Tagalog Reduplication is Affixation, Too. Paper presented at the Linguistic
Society of America. Baltimore.
Marantz, Alec. 1984b. On the Nature of Grammatical Relations. Linguistic Inquiry Monographs
10. MIT Press.
Marantz, Alec. 1987. Phonologically induced bracketing paradoxes in full morpheme
reduplication. In Megan Crowhurst (ed.) Proceedings of the Sixth West Coast Conference
on Formal Linguistics. 203-212. CSLI Publications.
Marantz, Alec. 1989. Clitics and phrase structure. In Mark R. Baltin and Anthony S. Kroch (eds.)
Alternative conceptions of phrase structure. 99-116. University of Chicago Press.
- 31 Marantz,
A., 1997. No escape from syntax: Don't try morphological analysis in the privacy of your
own lexicon. University of Pennsylvania working papers in linguistics, 4(2). 201-225.
Marantz, A., 2007. Phases and words. In Sook-Hee Choe (ed.) Phases in the theory of grammar.
191-222. Dong-In Publishing Co.
Marantz, Alec. 2013. Locality domains for contextual allomorphy across the interfaces. In Ora
Matushansky and Alec Marantz (eds.) Distributed Morphology today: Morphemes for
Morris Halle. 95-115. MIT Press.
Marvin, Tatiana. 2002. Topics in the stress and syntax of words. MIT dissertation.
Marvin, Tatiana. 2013. Is word structure relevant for stress assignment. In Ora Matushansky and
Alec Marantz (eds.) Distributed Morphology today: Morphemes for Morris Halle. 79-93.
MIT Press.
Matushansky, Ora. 2002. On formal identity of Russian prefixes and prepositions. In Aniko
Csirmaz, Zhiqiang Li, Andrew Nevins, Olga Vaysman, and Michael Wagner (eds.) MIT
working papers in linguistics 42. 217-253.
McCarthy, John. 1995. Faithfulness in Prosodic Morphology and Phonology: Rotuman
Revisited. Ms. University of Massachusets.
Merchant, Jason. 1995. An Alignment Solution to Bracketing Paradoxes. In Rachel Walker, Ove
Lorentz, and Haruo Kubozono (eds.) Phonology at Santa Cruz, 4. 57-71.
Mohannan, Karuvannur P. 1982. Lexical phonology. MIT dissertation.
Müller, Gereon. 1996. Incomplete category fronting. SfS report 01-96, Seminar für
Sprachwissenschaft, Universität Tübingen
Nash, David. G. 1980. Topics in Warlpiri grammar. MIT dissertation.
Nespor, Marina and Irene Vogel. 1986. Prosodic phonology (studies in generative grammar
28). Dordrecht: Foris.
Newell, Heather. 2005a. A late adjunction solution to bracketing paradoxes. In Leah Bateman
and Cherlon Ussery (eds.) The proceedings of NELS Vol. 35(2).
Newell, Heather. 2005b. Bracketing paradoxes and particle verbs: a late adjunction
analysis. In Sylvia Blaho, Luis Vicente & Erik Schoorlemmer (eds.). Proceedings of
ConSOLE XIII. 249-272.
Newell, Heather. 2008. Aspects of the morphology and phonology of phases. McGill University
dissertation.
Newell, Heather. 2016a. The Ontology of English morpho-phonology. Paper presented at the 24th
Manchester Phonology Meeting (MfM). University of Manchester. 26-28 May.
Newell, Heather. 2016b. The Pathology of Level-Specific Morpho-Phonology. Paper presented at
the Montreal-Ottawa-Laval-Toronto Phonology Workshop (MOLT). Carleton University.
19 March.
Newell, Heather. 2017a. There is no word. Paper presented at GLOW 40. Leiden.
Newell, Heather. 2017b. The ontology of English morpho-phonology. Paper presented at the 24th
Manchester Phonology Meeting. Manchester.
Newell, Heather. 2017c. English Lexical Levels are not Lexical, but Phonological. Ms. UQAM
Newell, Heather. 2017d. Nested Phase Interpretation and the PIC. In Heather Newell, Máire
Noonan, Glyne Piggott, and Lisa Travis (eds.) The Structure of Words at the Interfaces.
OUP.
Newell, Heather. 2018. There are no Bracketing Paradoxes, or How to be a Modular Grammarian.
Poster presented at GLOW 41. Budapest.
- 32 Newell,
Heather. 2019. Bracketing Paradoxes in Morphology. In Rochelle Lieber (ed). Oxford
Encyclopedia of Morphology. OUP.
Newell, Heather and Tobias Scheer. 2017. Paper presented at the 25th
Manchester Phonology
Meeting University of Manchester. May 25-27.
Newell, Heather, Máire Noonan, Glyne Piggott, and Lisa Travis (eds). 2017. The structure of
words at the interfaces. OUP.
Newell, Heather and Glyne Piggott. 2014. Interactions at the syntax–phonology interface:
Evidence from Ojibwe. Lingua 150. 332-362.
Nkemnji, Michael Akamin. 1995. Heavy Pied-Piping in Nweh. University of California, Los
Angeles dissertation.
Odden, David. 1990. C-command or edges in Makonde. Phonology 7(1). 163-169.
Odden, David. and Mary Odden. 1985. Ordered reduplication in Kihehe. Linguistic Inquiry 16(3).
497-503.
van Oostendorp, Marc. 1994. Affixation and integrity of syllable structure in Dutch. Linguistics in
the Netherlands, 11(1). 151-162.
Orgun, C. Orhan. 1996. Sign-based morphology: a declarative theory of phonology-morphology
interleaving. University of California, Berkeley dissertation.
Pater, Joe. 2000. Non-uniformity in English secondary stress: the role of ranked and
lexically specific constraints. Phonology, 17(2). 237-274.
Pesetsky, David. 1979. Russian morphology and lexical theory. Ms. MIT.
Pesetsky, David. 1985. Morphology and logical form. Linguistic Inquiry, 16(2).193-246.
Plag, Ingo. 1999. Morphological productivity: Structural constraints in English derivation. Topics
in English Linguistics 28. Walter de Gruyter.
Postal, Paul. 1969. Anaphoric islands. In Robert I. Binnick (ed.) Papers from the Fifth Regional
Meeting of the Chicago Linguistic Society. 205–239. Chicago IL: CLS.
Pyle, Charles. 1972. On Eliminating BM's. In Paul Peranteau, Judith Levi and Gloria Phares (eds.)
Papers from the eighth regional meeting of the Chicago Linguistic Society. 516-532.
Chicago: Chicago Linguistic Society.
Raffelsiefen, Renate. 1999. Phonological constraints on English word formation. In Greville G.
Corbett, Geert Booij, and Jaap van Marle (eds.) Yearbook of Morphology 1998. 225-287.
Springer Netherlands.
Raffelsiefen, Renate. 2015. Phonological restrictions on English word-formation. In Peter
Müller, Ingeborg Ohnheiser, Susan Olsen, and Franz Rainer (eds.) Word-Formation. An
international handbook of the languages of Europe Vol. 2. 894-917. De Gruyter.
Raimy, Eric. 2000. The phonology and morphology of reduplication. Studies in Generative
Grammar Vol. 52. Walter de Gruyter. Vancouver.
Ramchand, Gillian and Peter Svenonius. 2002. The Lexical Syntax and Lexical Semantics of the
Verb-Particle Construction. In Line Mikkelsen and Christopher Potts (eds.) Proceedings
of WCCFL 21. 387–400. Somerville, MA: Cascadilla Press.
Reiss, Charles. 2007. Modularity in the “Sound” Domain: Implications for the Purview of
Universal Grammar. In Gillian Ramchand and Charles Reiss (eds.) The Oxford handbook
of linguistic interfaces. OUP.
Rice, Keren and Elizabeth Cowper. 1984. Consonant mutation and autosegmental morphology. In
Joseph Drogo, Veena Mishra and David Testen (eds.) Proceedings of CLS20. 309–320.
Chicago: Chicago University Press.
- 33 Rojina,
Nina. 2004. English particles, Russian prefixes, and prepositional phrases. Master’s
Thesis. UiT.
Rotenberg, Joel. 1978. The Syntax of Phonology. MIT dissertation.
Scheer, Tobias. 2001. The Rhythmic Law in Czech: Vowel-final Prefixes. In Gerhild Zybatow,
Uwe Junghans, Grit Mehlhorn & Luka Szucsich, (eds.) Current Issues in Formal Slavic
Linguistics. Frankfurt am Main: Lang. 37–48.
Scheer, Tobias. 2004. A lateral theory of phonology. Vol 1: What is CVCV, and why should it be?
Mouton de Gruyter, Berlin.
Scheer, Tobias. 2008. Why the Prosodic Hierarchy is a diacritic and why the Interface must be
Direct. In Jutta Hartmann, Veronika Hegedüs and Henk van Riemsdijk (eds.) Sounds of
Silence: Empty Elements in Syntax and Phonology. 145-192. Amsterdam: Elsevier.
Scheer, Tobias. 2009a. External sandhi: what the initial CV is initial of. Studi e Saggi Linguistici
47. 43-82.
Scheer, Tobias. 2009b. Representational and procedural sandhi killers: diagnostics, distribution,
behaviour. In Mojmír Dočekal and Markéta Ziková (eds.) Czech in Formal Grammar.
155-174. Lincom Europa.
Scheer, Tobias. 2011. A guide to morphosyntax-phonology interface theories: how extraphonological
information is treated in phonology since Trubetzkoy’s Grenzsignale.
Walter de Gruyter.
Scheer, Tobas. 2016. Melody-free syntax and phonologically conditioned allomorphy.
Morphology 26(3-4). 341-378.
Scheer, Tobias and Eugeniusz Cyran. 2018. Syllable Structure in Government Phonology. In S.J.
Hannahs & Anna Bosch (eds.) The Routledge Handbook of Phonological Theory. 262-
292. Oxford: Routledge.
Scheer, Tobias and Markéta Ziková. 2010. The Havlík Pattern and Directional Lower. In Draga
Zec & Wayles Browne (eds.) Formal Approaches to Slavic Linguistics. The Second
Cornell Meeting. 471-486. Ann Arbor: Michigan Slavic Publications.
Seidl, Amanda. 2001. Minimal indirect reference: A theory of the syntax-phonology interface.
London/ New York: Routledge.
Selkirk, Elizabeth. 1981 [1978]. On prosodic structure and its relation to syntactic structure. In
Thorstein Fretheim (ed.) Nordic Prosody II. 111-140. Trondheim:TAPIR.
Selkirk, Elizabeth. 1982. The syntax of words. Linguistic Inquiry Monograph 7. Cambridge, MA.
Selkirk, Elizabeth. 2011. The syntax-phonology interface. In John Goldsmith, Jason Riggle, and
Alan C. L. Yu (eds.) The handbook of phonological theory, Second edition. 435-483.
Wiley.
Speas, Margaret. 1984. Navajo Prefixes and Word Structure Typology. In Margaret Speas and
Richard Sproat (eds.) MIT Working Papers in Linguistics 7: Papers from the January
1984 MIT Workshop in Phonology. 86-109.
Spencer, Andrew. 1988. Bracketing paradoxes and the English lexicon. Language 64(4). 663-
682.
Sproat, Richard. 1984. On bracketing paradoxes. In Margaret Speas and Richard Sproat (eds.)
MIT Working Papers in Linguistics 7: Papers from the January 1984 MIT Workshop in
Phonology. 110-130.
Sproat, Richard. 1985. On deriving the lexicon. MIT dissertation.
- 34 Sproat,
Richard. 1988. Bracketing paradoxes, cliticization and other topics: the mapping between
syntactic and phonological structure. In Everaert, Martin and H. Schultink (eds.)
Morphology and modularity: in honour of Henk Schultink. 339-360. Dordrecht: Foris.
Sproat, Richard. 1992. Unhappier Is Not a" Bracketing Paradox". Linguistic Inquiry 23(2). 347-
352.
Steriade, Donca. 2000. Paradigm uniformity and the phonetics-phonology boundary. In:
Michael B. Broe and Janet B. Pierrehumbert (eds.) Papers in Laboratory Phonology V:
Language Acquisition and the lexicon. 313-334. Cambridge: Cambridge University Press.
Starke, Michal. 2018. Complex left branches, spellout, and prefixes. In Lena Baunaz, Liliane
Haegeman, Karen De Clercq, and Eric Lander (eds.). Exploring nanosyntax. 239-249.
OUP.
Steddy, Sam. 2019. Compounds, composability, and morphological idiosyncrasy. The Linguistic
Review 36(3). 453-483.
Strauss, Steven. L. 1982. On “relatedness paradoxes” and related paradoxes. Linguistic
Inquiry 13(4). 694-700.
Svenonius, Peter. 2004. Russian prefixes are phrasal. Ms. University of Tromsø.,
Svenonius, Peter. 2016. Spans and words. In Daniel Siddiqi and Heidi Harley (eds.)
Morphological metatheory: Linguistics Today 229. 201-222. John Benjamins.
Taraldsen, Tarald. 2000. V-movement and VP-movement in derivations leading to VO-order. In
Peter Svenonius (ed.) The Derivation of VO and OV. 97-122. Amsterdam: John Benjamins.
Uriagereka, Juan. 1999. Multiple Spell-Out. In Samuel David Epstein and Norbert Hornstein (eds.)
Working minimalism. 251-282. Cambridge, Mass.: MIT Press.
Vogel, Irene. 2009. The status of the Clitic Group. In Janet Grijzenhout and Baris Kabak (eds.)
Phonological domains: universals and deviations. 15-46. Walter de Gruyter.
Williams, Edwin. 1981. On the notions “Lexically related” and “Head of a word”. Linguistic
Inquiry 12(2). 245-274.
Wurmbrand, Susi. 1998. Heads or phrases? Particles in particular. In Wolfgang Kehrein and
Richard Wiese (eds.) Phonology and morphology of the Germanic languages.
Linguistische Arbeiten 386. 267-295. Tübingen: Niemeyer.
Ziková, Markéta. 2008. Alternace e-nula v současné češtině. Autosegmentální laterální analýza.
Masarykova univerzita v Brně dissertation.
Ziková, Markéta. 2009. Vowel-zero alternations in Czech double diminutives: from the Havlík to
the Lower pattern. In Mojmír Dočekal and Markéta Ziková (eds.) Czech in Formal
Grammar. 237-250. Lincom Europa.
Ziková, Markéta. 2012. Lexical Prefixes and Templatic Domains: Prefix Lengthening in Czech.
In Markéta Ziková and Mojmír Dočekal (eds.) Slavic Languages in Formal Grammar:
Proceedings of FDSL 8.5, Brno 2010. 325-338. Frankfurt am Main: Peter Lang.