Three Levels of Insertion: English nominalizing suffixes
Joseph Emonds, Palacký University
Among English "nominaj^erived from Verbs, based on suffixes such as -ing and -ment, there is a pervasive contrast between "event nominate" and "result nominals." Thus:
(1) - Nouns ending in -ing (writing, building, cooking, etc.) can be events/ processes or their results.
- Similarly development, destruction, maintenance, etc. can be processes or results of processes.
- However, many such nouns with -ing (eating, refusing, growing, etc.) can only be events.
- Still others (intention, pronouncement, restriction, etc.) can only be results.
1. DEEP INSERTION OF N SUFFIXES IN RESULT NOMINALS
Result nominals like writing(s), building(s), developments) combine grammatical suffixes with open class roots (write, develop) that have "purely semantic features" /. Such/are not used ii syntax, and occur only with open classes of the lexical head categories N, V, A, P.
Combinations of open class roots with suffixes can occur with further specific f, and are bold in (2).
(2) Result Nominals:
a. [ Those two ancient writings on parchment ] are worth millions.
b. [ Some smudged hand writing about the politics of that time ] didn't seem worth saving.
c. The citizens won't appreciate [ another treeless highway development of unknown cost ].
Result nominals are typical open-class lexical nouns because they have item-particular, non-compositional meanings. Such items are listed in lexicons and inserted in trees using (3):
(3) Deep Lexicalization. Lexical heads associated with purely semantic features / satisfy^exical insertion conditions before transformations apply in a phrasal domain containing them.
(4) RESULT NOMINALS have the following regular properties, which are all illustrated in (2).
(i) They describe a physical reality,
(ii) can be quantified or pluralized, if they are count nouns,
(iii) take modifying adjectives expressing physical properties,
(iv) form productive N-N compounds, and
(v) can occur with q/~-phrases expressing time and other adjective-like properties.
The regular properties (i)-(v) are typical syntactic properties of nouns, even though the lexical roots of these nominals are verbs. That is, result nominals are syntactically pure nouns. - i.e. the combination V+N in (6) enters .derivation as a Noun (Right hand head ruleh., Heading a NP.
Other properties of RN:
Idiomatic interpretations and lack of productivity confirm the open class character of result nominals:
(5)     a. develop, resent. contain
b. development, resentment, containment
c. gallop, present, constrain, maintain, retain
d. *gallopment, ^presentment, *constrainment,    *maintainment, detainment
e. gallop, presentation, constraint, maintenance, retention
1
(6) "Compositional" characteristics of V+ing
'your writ-ing'
(7)     Semantic features
a) pure semenatic features " f" (a wide variety, open class)
b) "grammticalized" semantic features " F " (language specific, limited ammount)
Result nominal suffixes such as -ing and -ment have semantic/' when they combine with roots. So, due to (3), they are inserted as "deep lexical heads" of words, as in (8).
Note an extra f on N, in addition to the f on V.
(8)      " another treeless highway \ievelop+meni{ of unknown cost "
[V°,+/] + [N°, +FJ
another    treeless highway'-..
develop V\+/
-mt'iu N° +V
of unknown cost
With result nominals, the nominalising morpheme -ing is "deep" inserted, i.e. it becomes a head, and unit enters derivation as a N from the very beginning.
2. SYNTACTIC INSERTION IN EVENT NOMINALS_
A second type of propositional nominal: John's refusing of the offer, John's proving of the theorem, the growing of tomatoes. Grimshaw's (1990) thorough study names them "complex event nominals."
(9) Event Nominals (some examples from Grimshaw's study):
a. [ The assignment of unsolvable problems to novices } should be avoided.
b. We are forced to witness [ the constant felling of ancient forests ].
c. [ The frequent expression of one's feelings ] can get on your friends' nerves.
d. I was watching [ the teacher's deliberate examination of the papers ].
e. [ Construction of a hotel without first obtaining permits ] can take several years.
(10) The following Event Nominals contrast with the Result Nominals in (2).
a. [ The writing of love letters on parchment ] was a common practice.
b. [ Your quick and willful penning of a political tract Monday ] was dangerous.
c. They are protesting [ any rapid development of new roads into the hills to attract industry ].
In contrast to result nominals, the suffixal N heads of event nominals seem grammatically inert. Event nominals act rather as if the stem verbs are their heads, since they have many verbal properties.
Thus, event nominals are incompatible with plurals, cardinal numerals, and adjectives describing physical attributes, etc. Chomsky (1970) notes that noun heads are incompatible with purpose and other adverbial clauses. But such clauses are acceptable in event nominals (9)-(10).
(11) EVENT NOMINALS have the following syntactic properties of verbs, which are illustrated in (9)-(10):
(i) Event Nominals describe events rather than physical reality,
(ii) - accept quantification or plurals with difficulty,
(iii) - can be modified by adjectives with the meaning of temporal and volitional adverbs,
(iv) - occur with ^-phrase counterparts to direct objects (but not with temporal o/-phrases), and
(v) - don't easily form N-N compounds._
(12) "Later/Syntactic" characteristics of the event nominal "-ing "
NP 'your writ-ing'
With event nominals, the nominalising morpheme -ing is inserted "later/as late as in syntax"., i.e. the unit enters the beginning of the derivation as a V - it becomes a Noun only after the -ing morpheme is added.
3
An empty node in a head position, e.g. N, doesn't act like a head until a lexical item is inserted. Heads have the following properties, the second going back to the structuralist Zellig Harris:
(13)   Headedness. The head of a construction is the lexical item in the construction which satisfies selection relations with other elements (i) inside and (ii) outside the construction.
In Event Nominals, the absence of N heads in underlying syntax can account for their verbal patterns (a)-(e). By juxtaposing properties of result nominals (4i-v) and event nominals (lla-e), one can construct a wide range of unacceptable combinations:
(14)    a. *Those two ancient writings of love letters are worth millions.
b. They suggested some rapid French cooking Monday for the board of directors.
c. * Much writing of love letters on parchment was a common practice.
d. * The student scribbling of equations just confused me.
e. *We are protesting three constant developments of no beauty into the hills.
f. *No one wants to see another treeless development to attract industry.
g. * Constant warfare resulted from the White House containments of Communism.
Event nominal have no separate f on the N. So their verbal co-occurrence properties can be accounted for if their head Ns are absent at the outset of derivations, as represented in (15).
In the tree (15), open class verbs satisfy selection restrictions by Deep Lexicalization (3), and the empty
nodes are ignored. The V develop thus selects a direct object DP new roads and a PP into the hills.
All the paradigms in (9a-e) can be accounted for if the heads -ment and -ing don't enter the underlying
empty position [n 0] in (15) until the end of the syntactic derivation of the Noun Phrase.
Note that (15) is structuraly identical with (8). The only distinction is the level of insertion of -ing. And
therefore there are NO extra f on N, in addition to the f on V.
(15) any rapid development of new roads into the hills to atttract industry
Note on empty P. P generally signals space/ time LOCATION, but not with of, for, without, etc. The feature +<P on X (here P) indicates that Xdoes not receive its usual interpretation in Logical Form.
Feature +EVENT
Nominalizers such as -ing and -ment have very simple lexical specifications, such as N, +EVENT,
+V_, where EVENT is a syntactic feature suggested by Grimshaw. Whatever the exact nature of the
EVENT feature, it cannot combine with stems of "stative" verbs:
(16) a.       * The sudden knowing of the entire periodic table would be quite useful.
b. *He would disapprove of any owing of more money.
c. * Continual possessing of too many material goods can worsen obsessive behavior.
I claim that since the head Ns -ment and -ing in event nominals have no purely semantic feature /', they are inserted into a tree (15) only during syntactic processing of the NP domain.
However, they still must enter the tree prior to interpreting EVENT in Logical Form, and hence prior to later syntactic operations on any domain containing this NP. Contrast (17) with Deep Insertion (3).
(17) Syntactic Lexicalization. Items with interpreted syntactic F (but no purely semantic /) satisfy insertion conditions as part of transformations applying in a phrasal domain containing them.
From the point of lexical insertion onward, an event nominal has an N head. Therefore in larger syntactic domains it is selected and distributed like a NP, and no longer like a V-headed construction.
Productivity: Unlike result nominals, event nominals are productive in English (Anderson 1982). Event nominals for activity (= non-stative) verbs are just as productive as, say, past tense is for English verbs. (Note on what "productive" means. Irregular forms are irrelevant for deciding whether a formation is productive; what makes something non-productive is absence of combinatorial options.)
Nominals using -ing are thus default event nominals, e.g. eating, growing, surrounding, refusing, writing, whenever a verb lacks a lexically specified variant, such as construction, development, etc.
(18) a,       The surrounding of the enemy camp took place at night, b.      None of us witnessed the building of the Charles Bridge.
3. LATE OR PHONOLOGICAL INSERTION OF UNINTERPRETED GRAMMATICAL MORPHEMES
The tree (15) contains an empty P which is filled in PF with an uninterpreted morpheme, the case-assigning preposition of Emonds (2000) and several other works justify such "late insertion."
(19) Phonological Lexicalization. Vocabulary items specified solely in terms of contextual and other uninterpretable features are inserted subsequent to any operation contributing to LF.
The same nominating suffix -ing is used for both result nominals (Deep Lexicalization) and event nominals (Syntactic Lexicalization). Moreover, this -ing also has a third much studied grammatical use, which results from the only interpretable syntactic feature in its lexical entry being optional:
(20) Lexical entry: ing, N, +V___, (+EVENT)
5
If -ing is selected without its syntactic feature EVENT, N remains empty throughout the syntactic derivation. The suffix -ing is then inserted only in Phonological Form. When this happens, N-like properties of event nominals due to its presence prior to Spell Out disappear. Gerunds act like Vs:
(21) English Gerunds (GERUNDIVE NOMINALS)
(i) Gerunds lack articles and most quantifiers, and are marginal with demonstratives;
(ii) - their direct objects are not marked with of and they can have double objects;
(iii) - their AP modifiers have adverbial rather than adjectival form; and
(iv) - they are compatible with stative (and all) predicates.__
What emerges is the completely productive English gerundive nominal, italicized in (22), whose only distributional restriction is that it can appear in NP positions only where grammatical case is available.
(22) Gerundive Nominals:
a. [(*Much) Writing each other love letters on parchment ] was an imperial practice.
b. . [Your quickly and willfully penning a political tract Monday ] was dangerous.
c. . Citizens protest [ (*each/ *a) rapidly developing new roads into the hills to attract industry ].
d. [ Suddenly knowing the entire periodic table ] would be quite an experience.
e. He would disapprove of [ (*any/ *the) owing them more money ].
f. [Continually possessing too many material goods ] can worsen obsessive behavior.
(23)   '-ing'   Inserted at PF (phonology only) : GERUNDIVE NOMINALS
PF-INSERTED (MORPHOLOGICAL COMPONENT)
With gerunds/participles, the nominalising morpheme -ing is inserted "at phonological component", only, i.e. the unit goes throzugh the whole derivation as a V erb - it is interpreted as a Verb - and it taskes the function of NP only externally. This instance of Late or PF Insertion for -ing corresponds to what traditionally counts as the most regular English inflection, Le.gerundiveformation.
6
4.  THE OVERALL PICTURE FOR TRI-LEVEL INSERTION
English event nominals are productive, since a default form -ing is always available for activity verbs. Result and Event Nominals are respectively prototypical of non-productive and productive derivational morphology.
We have seen that the two types are formed by Deep and Syntactic Lexicalization, and I claim that these two types of insertion can systematically explain these traditional categories of morphology.
The three-way variation in English nominalizations is a predictable consequence of a syntax-based theory of lexical insertion at three levels in a transformational derivation.
(24)    Deep Insertion (available for LF): Non-productive idiosyncratic derivation (result nominals)
Syntax Insertion (available for LF): Productive derivation prior to "Spell Out" (event nominals) Late Insertion (available only in PF): Inflection after Spell Out (gerundive nominals)
The properties of the two main devices of traditional morphology, inflection and derivation, are both better explained in terms of a lexical theory justified by syntax than in any framework which tries to treat morphology separately. The theory of tri-level lexical insertion can recapitulated as follows:
(3)     Deep Lexicalization. Lexical heads associated with purely semantic features / satisfy lexical insertion conditions before transformations apply in a phrasal domain containing them.
(17)    Syntactic Lexicalization. Items with interpreted syntactic F (but no purely semantic /) satisfy insertion conditions as part of transformations applying in a phrasal domain containing them.
(19)    Phonological Lexicalization. Vocabulary items specified solely in terms of contextual and other uninterpretable features are inserted subsequent to any operation contributing to LF.
(26)    A:      'Deep or Phase-initial Insertion,' via the Dictionary
=> semantic + syntactic features = f + F (N,V,A,P only)........................see (4)
B:      'Insertion in Syntax' or 'Phase-final Insertion'
-> syntactic/cognitive features interpretable at LF = F (all categories)... see (11)
C: 'PF Insertion'=> syntactic features not interpretable at LF = F (all categories) such as contextual features, absence of content: STATIVE = [V, +0]; -LOCATIVE = [P, +0] etc.... see (21)
7
5.  LEXICON + PHASE BY PHASE ERIVATION
LEXICON - INTERFAi 7 "
7/7/ MEMORY
DICTIONARY
semantic features: J mldrc-siiif: b\ phnni>li)<;ical content
SYNTACTICON : syntactic/conceptual features: r lexical addressing by syntactic category and contexts
LEXICAL; \KK.WS
PF: INTERFACE WITH LANGUAGE USE/SPEECH
LF: INTERFACE WITH UNDERSTANDING/SEMANTICS
T-model of derivation
6. REFERENCES
Anderson, Stephen (1982). 'Where's Morphology?' Linguistic Inquiry, 13, 571-612.
Chomsky, Noam (1970). Remarks on Nominalizations," in Readings in English Transformational Grammar, ed. by R. Jacobs and P. Rosenbaum. Ginn, Waltham, Mass.
Emonds, Joseph (2000). Lexicon and Grammar: the English Syntacticon. Mputon de Gruyter, Berlin.
Grimshaw, Jane (1990). Argument Structure. MIT Press, Cambridge, Mass.
8