THREE KINDS OF NOMINALIZATION IN ENGLISH1 By Tamina Stephenson Faculty Advisor: Pauline Jacobson Undergraduate honors thesis in linguistics, Brown University, April 2002 Introduction This project explores the noun phrase (NP) constructions exemplified in (1)-(3). (1)

(Derived nominals) John’s criticism of the book the enemy’s destruction of the city

(2)

(Gerundive nominals) John’s criticizing the book the enemy’s destroying the city

(3)

(Mixed nominals) John’s criticizing of the book the enemy’s destroying of the city

Derived nominals are morphologically unpredictable but still intuitively related to their corresponding verbs, e.g., criticism, destruction, and assignment are somehow related to criticize, destroy and assign, respectively. With all of these constructions in (1)-(3), there is an intuition that they somehow relate or correspond to the sentences in (4). (4)

(Corresponding sentences) John criticizes / criticized the book. The enemy destroys / destroyed the city. The expressions in (1) will be referred to as “derived nominals,” those in (2) as

“gerundive nominals,” and those in (3) as “mixed nominals” (“mixed” because they have the -ing verb form like gerundive nominals but use an of-phrase like derived nominals.) These have gone by some variety of names in the literature; for instance, derived nominals have also been called “de-verbal nouns,” and mixed nominals have been called “ing-of” nouns (or -ingof nouns). 1

This project comes in part out of a summer 2001 research project with advisor P. Jacobson, Argument Structure and Lexical Rules: A Cross-Theoretical Approach, funded by the Brown Undergraduate Teaching and Research Award (UTRA) program.

1

Tamina Stephenson

Honors Thesis Chapter 1

Chomsky’s “Remarks on nominalization” and the structure of nominals The seminal work on nominalization in English (Chomsky, 1970) considered, among other things, derived nominals as in (1) and gerundive nominals as in (2). Chomsky observed that derived nominals such as John’s criticism of the book seem to have the internal syntactic structure of noun phrases (NPs), whereas gerundive nominals such as John’s criticizing the book seem to have the structure of sentences. As evidence for this, he cites the fact that derived nominals can occur with adjectives and with nonpossessive determiners such as a and the, as shown in (5) below, whereas gerundive nominals cannot do so, as shown in (6). (5)

(a) John’s harsh criticism of the book (b) the criticism of the book / a criticism of the book

(6)

(a) * John’s harsh criticizing the book (b) * the criticizing the book / * a criticizing the book

Chomsky also mentions that gerundive nominals can occur with multi-verb constructions such as the perfect tense (Chomsky, 1970: fn. 7). As shown in examples such as (7)-(8), this is true provided that the main (conjugated) verb has an -ing form.2 Gerundive nominals can occur with adverbs whereas derived nominals cannot, also shown in (7)-(8). (7)

(a) * John’s harshly criticism of the book (b) * John’s having criticism of the book

[cf. John harshly criticized the book] [cf. John has criticized the book]

(8)

(a) John’s harshly criticizing the book (b) John’s having criticized the book

[cf. John harshly criticized the book] [cf. John has criticized the book]

We can also make another observation which has to do with semantics more than syntax. First notice that with typical nouns such as book or hat, a possessive NP can express many different kinds of contextually salient relationships, not just “possession.” For instance, John’s book can refer to the book that John wrote, the one he owns, the one he is reading, and so on; John’s hat can refer to the hat that John owns, the one he is

2

That is, while we get John’s having criticized the book, we don’t get *John’s willing criticize the book. The case is similar for other verbs with impoverished paradigms such as can, may, might, and auxiliary do. I assume that this is just because, for some reason, these verbs lack -ing forms. Following Chomsky, I also assume that the ill-formedness of *John’s being criticizing the book comes from a general constraint against having two -ing forms in a row. (See, e.g., Chomsky, 1970: fn. 7 and Ross, 1972.)

2

Tamina Stephenson

Honors Thesis Chapter 1

wearing, and so on. (I will discuss the possessive more thoroughly in Chapter 2.) Now notice that, at least in certain cases, the possessive NP in a derived nominal can express some relation other than that of being the agent or subject of the implied verb, whereas this is impossible with gerundive nominals and mixed nominals. This is illustrated in (9) below. (The situation will be more complicated once we consider the result/process distinction in derived nominals; however, let’s look at this simplified version for now.) (9)

Everyone in the class was required to read a piece of criticism about Google’s book and bring it in to class to report on it. Jill chose a review by Sam Oddlefoot, and Sam chose one by Jane McQuilly. (a) à John’s criticism of the book was by Mary Smith. (b) à * John’s criticizing the book was by Mary Smith. (c) à * John’s criticizing of the book was by Mary Smith.

The data from (5)-(9) is consolidated in (10)-(11) below. In addition, (12) shows how mixed nominals behave under the same tests. The patterns are summarized in Table 1. (10)

(a) John’s harsh criticism of the book (b) the criticism of the book / a criticism of the book (c) * John’s harshly criticism of the book [cf. John harshly criticized the book] (d) * John’s having criticism of the book [cf. John has criticized the book] (e) John’s criticism of the book was by Mary Smith.

(11)

(a) * John’s harsh criticizing the book (b) * the criticizing the book / * a criticizing the book (c) John’s harshly criticizing the book [cf. John harshly criticized the book] (d) John’s having criticized the book [cf. John has criticized the book] (e) * John’s criticizing the book was by Mary Smith.

(12)

(a) John’s harsh criticizing of the book (b) the criticizing of the book / ? a criticizing of the book (c) * John’s harshly criticizing of the book [cf. John harshly criticized the book] (d) * John’s having criticized of the book [cf. John has criticized the book] (e) * John’s criticizing of the book was by Mary Smith.

Table 1: Preliminary observations about nominalized expressions a / the adjectives adverbs verbal Possessive must morphology = subject / agent Typical nouns Yes/Yes Yes No No N/A Derived Yes/Yes Yes No No No Gerundive No/No No Yes Yes Yes Mixed No?/Yes Yes No No Yes

3

Tamina Stephenson

Honors Thesis Chapter 1

Notice that the behavior of mixed nominals is very similar to that of derived nominals. There are two differences: first, the mixed nominal is a bit odd with the indefinite article a; and second, it does not allow the possessive to express a relation other than that of agent or subject of the implied verb. I will return to the analysis of mixed nominals. A preliminary account in classical Transformational Grammar Before I describe Chomsky’s (1970) account of derived and gerundive nominals, I will very briefly describe the framework and assumptions for his analysis. As he makes explicit in the beginning of the paper, Chomsky assumes that the syntax of a language consists of two parts, a base and a set of transformational rules. The base includes the lexicon and a set of context-free phrase structure rules which generate syntactic deep structures of expressions. A sequence of the transformational rules may apply to these deep structures to create surface structures, by mapping syntactic structures (or “phrasemarkers”) into other syntactic structures. (Semantic interpretation is assumed to occur at deep structure, though semantics is not emphasized in Chomsky’s discussion.) This means that there are two possible ways that expressions can be generated. Some expressions are “base-generated,” meaning that they are formed only from lexical entries and the phrase structure rules, with the surface structure the same as the deep structure. Others are generated by transformations, meaning that a deep-structure created by the phrase structure rules was then mapped by transformations to a surface structure. Chomsky gives a general analysis that accounts for the contrasts in (10)-(11) above. He analyzes the gerundive nominal as a transformation acting on a sentence to yield an NP, but the derived nominal as a base-generated expression, where the derived noun is related to the corresponding verb by a semi-productive “lexical rule.” Figure 1 below gives a rough illustration of the idea behind Chomsky’s analysis. Note that lexical rules, under this conception, are relations between separate expressions that are each listed in the lexicon. The definition of a lexical rule is left somewhat vague both in “Remarks” and in the literature in general, a tradition which I will continue.

4

Figure 1

Chomsky's analysis of derived and gerundive nominals

Chomsky’s analysis of derived nominals NP Det

N'

John’s

N

PP

criticism

P

NP

of

Related by semi-productive “lexical rule”

Det

N'

the

N book

V criticize

Chomsky’s analysis of gerundive nominals S NP

VP

John

V' V criticizing

NP S

NP

NP

VP

John’s

V' V

Det

N'

the

N

criticizing

NP

Det

N'

the

N book

book

Note: These drawings are only rough schematic illustrations of the nature of Chomsky’s account and are not intended to represent his exact analysis.

5

Figure 1

Tamina Stephenson

Honors Thesis Chapter 1

A preliminary account in Categorial Grammar The syntactic framework I will use for the remainder of this project is that of Categorial Grammar. It is not my intention in this project to argue for or against any particular syntactic framework in a general sense, but just to pick one framework (the one I am most comfortable in) and explore some particular phenomena in terms of this framework. I will give a formal preliminary analysis of derived and gerundive nominals in Categorial Grammar that is fairly similar to Chomsky’s analysis given above, other than the difference in framework; but first, let me explain the basic principles of Categorial Grammar. The explanation will necessarily be a bit oversimplified, but it will hopefully make clear all the main apparatus that will be used here. First, the syntactic framework is based on a rich, recursively defined system of categories. Every expression has a category, which is determined based on the distributional properties of the expression. In general, expressions of the same category will have the same distribution, i.e., be able to appear in the same linguistic contexts; and expressions with the same distribution will usually have the same category. There are a small number of primitive, or basic categories, from which other, more complex categories can be defined. For the purposes at hand, we need only four primitive categories: S (“sentence”), NP (“noun phrase”), N (“noun,” or “common noun”), and PP (“prepositional phrase”). It should be noted that the difference between NP and N is based solely on distributional facts, and not on whether an expression consists of a single word or a phrase. So, for example, the expressions cat, black cat, and silly black cat that I saw yesterday are all ‘N’s, whereas the cat, the gray cat, and Sylvester are all ‘NP’s. Sometimes it is necessary to make finer distinctions within categories, which we can do with “features.” For example, a language with a case system might need to have categories such as N[nominative], N[accusative], etc. One such feature I will use is PP[of], which just stands for a prepositional phrase headed by of. More complex categories are defined recursively based on these primitive categories. In particular, if A and B are categories, then A/RB is a category and A/LB is a category. Non-primitive categories are always of the form A/B, which can be thought of as saying, “concatenate me with an expression of category B to give an expression of

6

Tamina Stephenson

Honors Thesis Chapter 1

category A.” Sub-scripted “L” or “R” indicates whether the “B” expression is to come to the left or right (i.e., before or after) the A/B expression.3 For example, intransitive verbs and verb phrases, e.g., smiles or eats a sandwich, are of category S/LNP, meaning that they take an NP to the left to give a sentence. Similarly, pre-nominal adjectives such as red are of category N/RN, meaning that they take an N to the right to give another N. These complex categories can themselves form part of even more complex categories. For example, a transitive verb such as like is of category (S/LNP)/RNP, meaning that it takes an NP to the right (the object) to give an expression that will take another NP to the left (the subject) to become a sentence. The combinatory rules can thus be given in the two very general phrase structure rule schemata below. (X and Y are meant to be variables over categories.) (i) (ii)

X X

à à

X/RY Y Y X/LY

This should give you an idea of the basic apparatus and notation of the framework. More important, though, is the close relationship that is posited between the syntactic and semantic combinatory processes. Crucially, the syntactic categories are meant to directly reflect the semantic types and combinatory properties of the expressions. To illustrate this, let’s start with the primitive categories; the category S is of type ; the categories NP and PP are both of semantic type ; and the category N is of semantic type . An expression of type A/B is a function from expressions of the type of B to expressions of the type of A. For example, an expression of category S/LNP is of semantic type because ‘NP’s are of type and ‘S’s are of type . An expression of category N/RN is of semantic type < , > because ‘N’s are of type . The “L” and “R” directional markings in the category do not matter for the semantics. So we can revise the phrase structure rule schemata (i)-(ii) to include semantics: (i) (ii)

X X

à à

X/RY Y Y X/LY

; ;

3

[[X]] = [[X/RY]] ( [[Y]] ) [[X]] = [[X/LY]] ( [[Y]] )

Many versions of Categorial Grammar include other methods of combination besides concatenation, but this will not be relevant to this project.

7

Tamina Stephenson

Honors Thesis Chapter 1

One consequence of this close connection between syntax and semantics is that it forces us to give a fairly formal semantic account of a phenomenon along with its syntactic account. At the very least, we have to make sure that the syntactic categories posited for expressions will yield the correct semantic types and function-argument relationships among the resulting expressions. Thus the rest of this project will pay close attention to both the semantics and the syntax of nominalized expressions. There is one more piece of apparatus which is important for the framework used here, which is the notion of a productive category-changing rule. This is a type of process that takes a certain kind of expression and maps it to another expression with changes in the phonology, syntax, and/or semantics. The input and output expressions are written as triples of . So, for example, if we treat plural nouns as being of category N[plur], then we could have the rule below: Given an expression α: α: <[α]; N; α’> Map this to an expression β: è β: <[α-/z/]; N[plur]; plural of α’> where “plural of α’ ” is the meaning of β, defined in terms of α’ Given that it is essentially impossible to give a syntactic account of anything in Categorial Grammar without discussing the semantics as well, I must now provide a few tools to talk about the meanings of nominalized expressions. I assume, again, that a noun phrase (NP) denotes an “individual” or “entity” (I use these terms interchangeably), with semantic type . A sentence (S) denotes a truth value (1 or 0), with semantic type . Now, some NPs, including nominalized expressions, seem to refer to events or propositions. I assume that these NPs are still of semantic type . The individuals they denote are, however, abstract entities corresponding either to an event or to the individual correlate of a proposition, which are roughly defined as follows: An event is a thing that happens in the world, with participants, actions, time and place, etc. An event can be an instance of a proposition if, roughly, that proposition becomes true in the context of that event. (For example, an event in which John danced is an instance of the proposition “John dances.”) There can be multiple instances of a single proposition. An event is of semantic type <ε>, which is a subtype of . An individual correlate of a proposition (due to Chierchia, 1984) is an abstract entity corresponding to a proposition, as a “thing” that exists in the world rather than 8

Tamina Stephenson

Honors Thesis Chapter 1

something which is true or false in the world. Each proposition has its own unique individual correlate. An individual correlate of a proposition is of semantic type , which is a subtype of . Zucchi (1993) argues on independent grounds that derived and mixed nominals denote events whereas gerundive nominals denote propositional individuals. This is more or less the distinction that I will make; however, the crucial difference for present purposes is simply the fact that a single proposition can correspond to many events but only one individual correlate. Now, let me finally turn to a preliminary syntactic and semantic analysis of derived and gerundive nominals within the framework of Categorial Grammar. For derived nominals, the basic idea is as follows: the expression criticism of the book is of category N (common noun), which means that it can combine freely with adjectives (category N/RN) and with determiners (NP/RN) to give an expression of category NP. This means that criticism is of category N/RPP[of]. (Note: this means that criticism, as it occurs in this construction, is not itself a common noun but in fact a relational noun, a distinction which will become important later but which we can gloss over for now.) As in Chomsky’s account, the (relational) noun criticism is related to criticize by a semiproductive lexical rule, which we will leave vaguely defined. Since criticism of the book is of category N, it can combine with the possessive (category NP/RN) to give an NP, where the possessive has its normal meaning entailing some contextually salient relation. (Again, this description of the meaning of the possessive is oversimplified and will be revised in Chapter 2.) In general, we could give a parallel analysis of mixed nominals; however, this would not account for the fact that mixed nominals are ill-formed or odd with the indefinite article a and require the possessive to denote the agent or subject relation. I will leave the mixed nominal aside for now, until we revise the analysis of both the mixed nominal and the derived process nominal, below. For gerundive nominals, we can get essentially the same effect as an S-to-NP transformation by positing a productive category-changing rule. This rule applies at the VP-level, that is, to expressions such as criticize the book (or harshly criticize the book, have harshly criticized the book, and so on), which correctly allows the full range of verbal constructions and VP phrase modifiers found in gerundive nominals. The output

9

Tamina Stephenson

Honors Thesis Chapter 1

of the rule is this same VP with -ing morphology4; this is crucially not of category N, but of category NP/LNP[‘s], which is to say that it takes a possessive NP as an argument much like a VP takes a regular NP as an argument. The semantics of the rule ensures that the NP denoted by the possessive will be the “subject” of the implied verb. One thing to notice is that the possessive is treated completely differently in the two constructions. In derived nominals, it is treated as a function that takes a noun as argument; whereas in gerundive nominals it is treated as a special argument. This means that we have two separate categories and meanings for the same possessive expression. Even when the meaning of the possessive in derived nominals is revised in Chapter 2, we will still have this problem. The role of the possessive will be generalized somewhat (though not completely) at the end of Chapter 2. This above analysis of derived and gerundive nominals in Categorial Grammar with semantics is shown below in tree form in Figures 2.a-2.b.

4

It calls for comment that the morphological change occurs on the verb itself, whereas the syntactic and semantic components of the rule apply to an entire verb phrase. I won’t be able to discuss this in depth, but such mismatches show up frequently in different phenomena and different theoretical frameworks.

10

Figure 2.a

Categorial Grammar analysis of derived nominals

Categorial Grammar analysis of derived nominals NP John’s criticism of the book <ε> c : c = the unique contextually salient entity that is an instance of criticize’(x, b ), where x is contextually salient and c holds some contextually salient relation to j

NP/RN

N

John’s < , e > λP[the unique contextually salient member of set P that holds a contextually salient relation to j ]

criticism of the book <ε,t> The set of entities {ε : ε is an instance of criticize’(x, b ) for some x in context}

Related by morphological rule < [α α ], NP, α ’ = x > ⇒ <[α α -’s], NP/RN, λ P[the unique contextually salient member of set P that holds a contextually salient relation to x] >

N/RPP[of]

NP John j

criticism > λy[ the set of entities {ε : ε is an instance of criticize’(x,y) for some x in context} ]

Related by semi-productive derivational rule < [α α ], (S/LNP)/RNP, λ yλ λ x[α α ’(x,y)] > ⇒ < [α α -ism], N/RPP[of], λ y[the set of entities {ε : ε is an instance of the proposition α’(x,y) for some x in context}]

(S/LNP)/RNP criticize > λyλx[criticize’(x,y)]

PP[of] of the book b

PP[of]/RNP

NP

of λx[x]

the book b

NP/RN

N

the < , e > λP[the unique contextually salient member of set P]

book The set {x: book’(x)}

11

Figure 2.a

Figure 2.b

Categorial Grammar analysis of gerundive nominals

Categorial Grammar analysis of gerundive nominals NP John’s criticizing the book c : c = the individual correlate of the proposition criticize’( j , b )

NP[’s]

NP/LNP[’s]

John’s j

criticizing the book λx[the individual correlate of criticize’(x, b )]

Related by productive category-changing rule < [α α], NP, z > ⇒ < [α α-’s], NP[’s], z >

Related by productive category-changing rule < [V α], S/LNP, : Q > ⇒ < [V-ing α], NP/LNP[’s], : λx[the individual correlate of the proposition Q(x)] >

S/LNP

NP

criticize the book λx[criticize’(x, b )]

John j

(S/LNP)/RNP

NP

criticize > λyλx[criticize’(x,y)]

the book b

NP/RN

N

the < , e > λP[the unique contextually salient member of set P]

book The set {x: book’(x)}

12

Figure 2.b

Tamina Stephenson

Honors Thesis Chapter 1

The result / process distinction Since Chomsky (1970), another important distinction has been made between two different kinds of derived nominals. Two separate observations are relevant to this distinction. First, consider the derived nominal in (13): (13) John’s criticism of the book Intuitively, there seem to be at least two different readings for this. One refers to an act or event where John criticized the book; the other refers to a concrete manifestation of such an act in the form of a review, article, speech, or other verbal object. The second observation concerns the presence of arguments or complements with derived nominals. As discussed in Grimshaw (1990), a number of people have claimed that nouns cannot have obligatory arguments but only optional ones.5 For example, in (14), the object the papers is obligatory with the verb examine, but the phrase of the papers seems to be optional with the nominalized expression examination: (14)

(a) The examination took a long time. (b) (*) The instructor’s examination took a long time. (c) The instructor’s examination of the papers took a long time. (d) The instructor examined *(the papers). [Taken from Grimshaw, 1990: p. 51, (10).] Grimshaw argues convincingly, though, that the apparent optionality of the of-phrase is actually an illusion due to a widespread ambiguity among nouns. In other words, the noun examination, along with many other nouns, is ambiguous between one reading which denotes an event or process and another reading which denotes an object. In the former case, the argument is obligatory, as we can see by inserting intentional or deliberate as in (15), disambiguating examination to the action reading. (15)

(a) * The instructor’s intentional / deliberate examination took a long time. (b) The instructor’s intentional / deliberate examination of the papers took a long time. [Taken from Grimshaw, 1990: p. 51, (11).] Grimshaw cites work by several authors who have observed this type of ambiguity in the nominal system, in particular in terms of concrete or “result” nominals (as in 14.a) as opposed to “true” or “process” nominals (as in 15.b). Ultimately, she argues that the

5

Grimshaw (1990: p. 45) cites Anderson (1983-1934), Higginbotham (1983), and Dowty (1989).

13

Tamina Stephenson

Honors Thesis Chapter 1

distinction has to do with whether a noun has “complex event structure,” a notion which I will not deal with in depth in this project. Rather, I will simply use the traditional terms “result nominal” and “process nominal” to refer to the distinction between derived nominals that refer to “things” and those that refer to actions or events. I will rely largely on intuition to determine whether a particular derived nominal has a result reading or process reading, but only with the understanding that these intuitions can be tested using disambiguating modifiers as in (15). In light of these additional observations, it is necessary to refine the data and analysis presented earlier. In particular, let’s reconsider example (9a), repeated as (16). (16)

Everyone in the class was required to read a piece of criticism about Google’s book and bring it in to class to report on it. Jill chose a review by Sam Oddlefoot, and Sam chose one by Jane McQuilly. (a) à John’s criticism of the book was by Mary Smith. (b) à * John’s criticizing the book was by Mary Smith. (c) à * John’s criticizing of the book was by Mary Smith.

Intuitively, criticism of the book in (16.a) seems to have a result reading, that is, it refers to a relatively concrete thing such as an article or book, rather than to an act or event where someone criticized something. On the other hand, criticizing the book in (b) and criticizing of the book in (c) seem to clearly refer to some act or event. In fact, if we force a process reading in (a) by adding in the adverb intentional and adapting the example and context accordingly, it becomes ill-formed, as shown in (17). (17)

Everyone in the class was required to read some article where somebody intentionally criticized Google’s book, and bring the article in to class to report on it. Jill chose a review by Sam Oddlefoot, and Sam chose one by Jane McQuilly, both of whom had intentionally criticized Google. (a) à * John’s intentional criticism of the book was by Mary Smith. (b) à * John’s intentional/intentionally criticizing the book was by Mary Smith. (c) à * John’s intentional criticizing of the book was by Mary Smith.

Thus one difference between mixed nominals and derived nominals found in (10)-(12) is neutralized when the derived nominal is forced into a process reading. Recall that the other difference was that derived nominals were apparently fine with the indefinite article a, while mixed nominals were odd. Again, if we force a result reading by adding the adverb intentional (or deliberate), the two constructions are about the same.

14

Tamina Stephenson

(18) (19)

Honors Thesis Chapter 1

? an intentional criticism of the book / ? a deliberate criticism of the book ? an intentional criticizing of the / ? a deliberate criticizing of the book Before I go on, I should make another observation about derived result nominals.

It is, I believe, completely uncontroversial that these are frozen, lexicalized expressions whose relationship to a corresponding verb is not predictable. Examples of result nominals are examination (meaning a written test), assignment (meaning work assigned), criticism (meaning an instance or the body of works reviewing a piece of writing), plus many others such as consultation, decision, election, rehearsal, recital, argument, and government. Some result nominals even use the -ing ending, for example clothing, trimming, and building.6 A thorough investigation of the possible meanings for derived result nominals is beyond the scope of this project. From now on I will limit the discussion of derived nominals to process readings, only bringing up result nominals when relevant. Unless otherwise stated, I will use the term “derived nominal” to refer only to derived process nominals. Again, keep in mind that any time there is an intuitive difference between a result and process reading of a nominal, it should be possible to test this by inserting a modifier such as intentional or deliberate to force a process reading. To summarize, at first glance there are a number of differences between derived and mixed nominals on the one hand, and gerundive nominals on the other. In addition, there were a small number of differences between derived and mixed nominals. These were all shown in (10)-(12). Once we restricted the data to “process” readings of derived nominals, however, we found that these differences disappeared. If we re-cast (10)-(12) in light of these new observations, restricting ourselves to process readings of derived nominals, the data is as shown in (20)-(22) below. Then the new observations are summarized in Table 2. (20)

(a) John’s harsh criticism of the book (b) the criticism of the book (c) ? a criticism of the book [on process reading] (d) * John’s harshly criticism of the book [cf. John harshly criticized the book] (e) * John’s having criticism of the book [cf. John has criticized the book] (f) * John’s criticism of the book was by Mary Smith. [on process reading]

6

In compiling lists of nominalized expressions, I have taken freely from several sources, including the searchable online OED, Chomsky (1970), and work by Anderson, Halle, Jackendoff, G. Lakoff, Selkirk, and Wasow.

15

Tamina Stephenson

Honors Thesis Chapter 1

(21)

(a) * John’s harsh criticizing the book (b) * the criticizing the book (c) * a criticizing the book (d) John’s harshly criticizing the book [cf. John harshly criticized the book] (e) John’s having criticized the book [cf. John has criticized the book] (f) * John’s criticizing the book was by Mary Smith.

(22)

(a) John’s harsh criticizing of the book (b) the criticizing of the book (c) ? a criticizing of the book (d) * John’s harshly criticizing of the book [cf. John harshly criticized the book] (e) * John’s having criticized of the book [cf. John has criticized the book] (f) * John’s criticizing of the book was by Mary Smith.

Table 2: Revised observations about nominalized expressions a / the adjectives adverbs verbal Possessive must morphology = subject / agent Typical nouns Yes/Yes Yes No No N/A Derived result Yes/Yes Yes No No No Derived process No?/Yes Yes No No Yes Gerundive No/No No Yes Yes Yes Mixed No?/Yes Yes No No Yes Now the good news is that derived and mixed nominals seem to work the same way. Of course, this means that the analysis given earlier for derived nominals will now not work for derived (process) nominals for the same reasons that it would not work for mixed nominals before. That is, the analysis given will not account for the requirement that the possessive NP be the subject or agent of the implied verb, and it won’t account for the fact that derived (process) nominals and mixed nominals are not entirely acceptable with the indefinite article. It will require some additional apparatus to deal with these problems. Chapter 2 will give a solution to the problem of the possessive. After that, Chapters 3 and 4 will bring up additional differences between derived and mixed nominals and discuss the productivity of the three types of nominalization. Finally, Chapter 5 will explore the properties of derived and mixed nominals with respect to the distinction between mass nouns and count nouns, and will give a tentative analysis to account for the oddness of derived and mixed nominals with the indefinite article.

16

Chapter 2: Possessives and of-phrases in nominals Relational nouns and the meaning of the possessive in English The observations in the previous chapter lead us to a mystery: how is it that derived process nominals and mixed nominals act like normal nouns in so many ways, and yet seem to require a possessive determiner to express a certain relationship. I will offer a solution to this problem by calling these nominalized expressions “relational nouns.” But first, I will need to give some background on both relational nouns and the meaning of the possessive, for which I owe several important insights to Barker (1995). Possessive expressions in English, such as my, his, John’s, or the cat’s, have the same syntactic role as other expressions traditionally called determiners. In general, a determiner combines with a noun-like expression denoting a predicate or set of individuals, and returns another noun-like expression denoting an individual. Syntactically speaking, then, a determiner combines with an expression of category N (common noun) to give an expression of category NP (noun phrase). The determiner the, for example, returns as an NP something like the single contextually salient member of the set denoted by the common noun. The possessive, on the other hand, denotes something like the single contextually salient individual that bears a particular important relationship to the individual embedded in the possessive. Thus my book denotes the book that bears an important relationship to the speaker; John’s hat denotes the hat that bears an important relationship to the individual referred to by John, and the cat’s food denotes the food that bears an important relationship to the individual referred to by the cat. The natural question that arises, then, is: which sorts of “important relationships” can be expressed by the possessive in English? A few answers are listed in (23) below: (23)

Relationships expressed by the possessive in English: i. Ownership: John’s money, John’s toys (the ones that John owns) ii. Inalienable possession: John’s nose (the one that is inherently part of John) iii. Argument of a relational noun: John’s mother (the person with that relationship to John given by the meaning of mother; John fills an implied argument position of mother) iv. Conventional association: John’s chair (the one that John is sitting in) v. Other contextually salient association: John’s article about the book (the one that John brought to class)

17

Tamina Stephenson

Honors Thesis Chapter 2

Let’s look more closely at the meaning in (23.iii), “argument of a relational noun.” It is standard to treat most common nouns, such as apple or human, as one-place predicates which extensionally denote sets of individuals: each individual, that is, is either a human or not a human depending on its own properties. On the other hand, there is a class of words that are intuitively noun-like but have in their meaning some reference to a relationship between two individuals. This class includes many kinship terms such as brother, cousin, or daughter. A person is not a “brother” or “cousin” or “daughter” in an absolute sense but only with respect to another individual who they are they brother or cousin or daughter of. Following Barker (1995) and current standard practice, I will refer to these as “relational nouns,” and I will assume that the meaning of a relational noun is a two-place predicate, or a set of ordered pairs of individuals.1 So, for example, [[daughter]](a,b), or, [[daughter]](b)(a) , is true just in case individual a is the daughter of individual b, i.e., just in case ∈ [[daughter]]S where [[daughter]]S is the set characterized by the function [[daughter]] (seen as a function of ordered pairs). Or, put another way, [[daughter]] = λy[λx[daughter’(x,y)] ] This has various entailments, of course: for example, a must be female and b must be a parent of a; this is all part of the idiosyncratic lexical meaning of daughter. It should be noted that although many relational nouns denote natural kinship terms, there exist other kinds of relational nouns as well. Barker gives the examples of birthday and biography. Birthday expresses a relationship between a temporal object (a day) and a person. (Barker contrasts birthday with day, which is just a property of a temporal object). Biography, on at least one reading, expresses a relationship between a book and the person whose life the book is about. The important thing to notice about relational nouns is that when they combine with possessives, the relationship most commonly “taken” by the possessive is the same relationship expressed in the relational noun. So for example, if John denotes j, then 1

As one source for this view of relational nouns, Barker cites Lieber, 1983 (Argument Linking and Compounds in English, Linguistic Inquiry, 14.2: 251-285).

18

Tamina Stephenson

Honors Thesis Chapter 2

John’s sister refers to a person who bears the “sister” relation to j, i.e., an individual x such that [[sister]](x, j) is true. This observation holds for all of the relational nouns shown in (24), combined with a possessive under the most natural reading. (24)

(a) John’s daughter (b) John’s brother (c) John’s cousin (d) John’s mother (e) John’s child (f) John’s pet (g) John’s birthday (h) John’s biography

= individual x such that [[daughter]](x, j), that is, = λx[daughter’(x, j)] = λx[brother’(x, j)] = λx[cousin’(x, j)] = λx[mother’(x, j)] = λx[child’(x, j)] = λx[pet’(x, j)] = λx[birthday’(x, j)] = λx[biography’(x, j)]

There is one complication here which I should deal with immediately, which is this: the meanings given above for John’s daughter, John’s brother, and so on have the form of one-place predicates, that is, they should extensionally denote a set of individuals. In one sense, this is intuitively correct, since it is possible to define the sets of individuals {x: daughter’(x, j)}, {x: brother’(x, j)} and so on, that is, the set of daughters of John, the set of brothers of John, and so on. (Barker in fact assumes that expressions like those in (24) denote sets.) On the other hand, though, the expressions in (24) actually seem to have the distribution and semantic role of NPs, of category . In particular, an expression such as John’s daughter does not seem to refer to the set of daughters of John, but to one particular, salient member of that set, as evidenced by the fact that we can predicate properties of this individual as in (25): (25)

(a) John’s daughter runs. (b) John’s brother sings.

= [[runs]] ( [[John’s daughter]] ) ? = [[sings]] ( [[John’s brother]] )

To assure that the resulting expression will be of the right syntactic category and semantic type, I will follow the treatment of possessives with relational nouns from CG 113,2 which is essentially a version of Barker’s analysis. In terms of phrase structure grammar, we can use the rule in (26):

2

Course CG 113, Formal Semantics, Spring semester 2002, professor P. Jacobson.

19

Tamina Stephenson

(26)

Honors Thesis Chapter 2

Analysis of possessive NPs with relational nouns in Phrase Structure Grammar: (i) NP à NP[GEN]

RN;

[[NP]] = [[the]] ( [[RN]] ( [[NP[GEN] ) ) < , e> ( > () ) where “RN” is the category “relational noun.”

(ii) [Morphological rule]:

NP[GEN] à NP + ‘s;

[[NP[GEN]]] = [[NP]]

Under this treatment, the meaning of the possessive morpheme ‘s is just the identity function on individuals, only changing the input expression by adding possessive morphology and the syntactic feature [GEN]. (This is ignoring morphologically irregular possessive forms such as my, his, etc.; but let’s assume that our theory of morphology has an explanation for these.) Thus [[RN]] is of type > and [[NP[GEN]]] is of type , so [[RN]] ( [[NP[GEN] ) is of type . Then crucially we apply a special operator [[the]], which is of type <, e>, thus giving an expression of category . This operator [[the]] is, of course, the same as the meaning of the English word the, which takes a set and returns the single (or maximal) contextually salient member of that set. For the sake of consistency, let me recast the rule given above in Categorial Grammar. The account will be somewhat different than the one above, for a couple of reasons. First, I will treat “relational noun” as a non-primitive category. Second, since possessive NPs otherwise seem to have the distribution and role of determiners, I will want to treat the possessive NP as the function and the RN as the argument instead of the other way around. This means that the possessive NP can’t just have the same meaning as the NP, so the morphological rule for the possessive will essentially be a type-lifting rule. This is illustrated in the analysis shown in (27) below: (27)

Analysis of possessive NPs with relational nouns in Categorial Grammar: Note: w, x, y, z are variables of type ; P, Q are variables of type ; R, S, T are variables of type > (i) Relational nouns such as daughter, brother, are of category N/RPP[of] and semantic type > , which trivially gives us daughter of John, brother of John, etc. (assuming that [[of NP]] = [[NP]] ). (ii) This means that an expression such as daughter of John is of category N and semantic type . Meaning: [[S of a]] = S(a) = λy[S(a)(y)] = λy[s(y,a)] 20

Tamina Stephenson

Honors Thesis Chapter 2

(iii) The possessive morpheme ‘s is of category ( NP/R(N/RPP[of]) )/LNP and semantic type < e, <> , e> >, which means that it takes an NP to give a “possessive NP” (see below). Meaning: [[‘s]] = λx[λ λR[ [[the]] ( R(x) ) ] ] = λx[λ λR[ [[the]] ( λ y[R(x)(y)] ) ] ] = λx[λ λR[ [[the]] ( λ y[R(y,x)] ) ] ] (iv) A “possessive NP” such as John’s is of category NP/R(N/RPP[of]) and semantic type < > , e > , which means that it takes a relational noun to give an NP. Meaning: [[a’s]] = λx[λ λR[ [[the]] ( λ y[R(y,x)] ) ] ] (a) = λR[ [[the]] ( λ y[R(y,a)] ) ] (v) Thus an expression such as John’s sister is of category NP and type . Meaning: [[a’s S]]= λR[ [[the]] ( λ y[R(y,a)] ) ] (S) = [[the]] ( λ y[S(y,a)] ) This analysis is illustrated in Figure 3.a below, which shows the derivation for the sentence John’s sister sings in tree format. There is just one more loose end to tie off, which is the fact that relational nouns can occur with adjectives, as in John’s nice sister. Intuitively, the adjective nice modifies only the “sister” and not “John.” I will just assume that adjectives can be of category (N/RPP[of]) /R (N/RPP[of]), meaning that they take a relational noun to give another relational noun, with the meaning as follows: [[P2 R]]= λx[λ λy[R(y,x) and P1(y)] ] , where P1 is the normal meaning of the adjective of type . Of course, we will want to be able to derive this by a productive rule applying to adjectives. Although I won’t formalize the rule, I illustrate how the output will work in Figure 3.b, which shows the derivation of the NP John’s nice sister.3

3

One way to formalize this is to apply the “Geach Rule” to a pre-nominal adjective or in some other way make it function-compose with the category of relational nouns. It can easily be verified that this will give a meaning equivalent to the one shown above and in Figure 3.b.

21

Figure 3.a

Categorial Grammar analysis of relational nouns (part a)

Relational nouns with possessives S John’s sister sings sings’( [[the]] ( λy[sister’(y, j )] ) ) = sings’(s )

NP

S/LNP

John’s sister λR[ [[the]] ( λy[R(y, j )] ) ] (sister’) = [[the]] ( λy[sister’(y, j )] ) = some individual, let’s call it s

sings sings’

NP/R(N/RPP[of])

N/RPP[of]

John’s < > , e > λx[λR[ [[the]] ( λy[R(y,x)] ) ] ] ( j ) = λR[ [[the]] ( λy[R(y, j )] ) ]

sister > sister’

NP

( NP/R(N/RPP[of]) )/LNP

John j

’s < e, <> , e> > λx[λR[ [[the]] ( λy[R(y,x)] ) ] ]

22

Figure 3.a

Figure 3.b

Categorial Grammar analysis of relational nouns (part b)

Relational nouns with adjectives NP John’s nice sister λR[ [[the]] ( λy[R( j )(y)] ) ] ( λx[λz[sister’(z,x) and nice1’(z)] ] ) = [[the]] ( λy [ λx[λz[sister’(z,x) and nice1’(z)] ]( j )(y) ] ) = [[the]] ( λy [ sister’(y, j ) and nice1’(y) ] )

NP/R(N/RPP[of])

N/RPP[of]

John’s < > , e > λx[λR[ [[the]] ( λy[R(y,x)] ) ] ] ( j ) = λR[ [[the]] ( λy[R(y, j )] ) ] = λR[ [[the]] ( λy[R( j )(y)] ) ]

nice sister > λR[ λx[λz[R(z,x) and nice1(z)] ] ] ( sister’ ) = λx[λz[sister'(z,x) and nice1’(z)] ]

(See derivation in Figure 3.a) …

(N/RPP[of]) /R (N/RPP[of])

N/RPP[of]

nice < > , > > nice3’ = λR[ λx[λz[R(z,x) and nice1’(z)] ] ]

sister > sister’

(Related by productive category-changing rule) < …> ⇒ <…>

N/RN nice nice1’

23

Figure 3.b

Tamina Stephenson

Honors Thesis Chapter 2

Of course, this analysis does not account for the fact that possessives can also combine with normal, non-relational nouns. To do this, it is necessary to have a second meaning for the possessive which provides some relationship. This is basically what Barker (1995) does, and I refer the reader to his work for a more detailed discussion of the possessive with non-relational nouns. More importantly for the work at hand, the analysis given above predicts that when a possessive NP combines with a relational noun, the relationship that it expresses will always be the same as the one in the lexical meaning of the relational noun. Recall that when a possessive combines with a non-relational noun, there are a number of possibilities for what it will express, including ownership or other contextually salient relationships, so this is neither obvious nor trivial. Following Barker, I will argue that the prediction is correct. Before I discuss this claim in more detail, let me explain briefly what this has to do with nominalization. First, recall that with normal common nouns, the possessive can “pick out” any of a number of contextually salient relationships. Given that, the problem with derived process nominals and mixed nominals was this: these expressions act like normal nouns in many ways; yet when they combine with a possessive NP, the possessive is forced to “pick out” a particular relationship. That is, they are noun-like expressions which “force” the possessive to take on a particular meaning. But now we have evidence for another class of noun-like expressions – namely, relational nouns – which have the property of “forcing” the possessive to take on a particular meaning. We have to explain relational nouns anyway, so it would be nice if we could extend the same explanation to nominals. My strategy, therefore, will be to assimilate derived (process) nominals and mixed nominals with relational nouns. Then, if we can establish that when a possessive is combined with a relational noun, it will always express the same relationship defined by the relational noun, then this will give us a ready-made explanation for why the possessive in derived process nominals and mixed nominals must express the subject of the implied verb. We can just say that expressions such as criticism of the book or criticizing of the book are of the same category and semantic type as lexical relational nouns such as sister or cousin, and that the relation expressed in these nominals is that of

24

Tamina Stephenson

Honors Thesis Chapter 2

“subject-hood” or “agentivity.” This will also explain why derived and mixed nominals, like other nouns and relational nouns, can combine with adjectives. I should note that the situation for gerundive nominals won’t be so simple, since gerundive expressions such as criticizing the book do not behave syntactically like common nouns (or for that matter like relational nouns). Now let’s return to the claim that a possessive will always express the same relationship as a relational noun that it combines with. I will start by giving some apparently problematic examples, then present the solution proposed by Barker that will account for these, and also add in independent evidence for Barker’s solution. It is possible to construct apparent counterexamples where it seems that the possessive can express some other contextually salient relationship, as in (28)-(30).4 (28.a)

[Context: A class in sociology, psychology, etc. is discussing motherhood. Each student brings in a mother (not their own) to talk to the class.] Sue’s mother talked about her career as a writer. [Sue’s mother = the mother who Sue brought in]

(28.b)

[Context: Like (28.a) except that students bring in grandmothers.] Sue’s grandmother talked about her career as a writer.

(29)

[Context: John works at a daycare center, where each employee is assigned to take care of a particular group of children.] John’s children played with blocks today. [John’s children = the children who John takes care of]

Barker argues that these kinds of cases are actually examples of a second lexical entry that is homophonous with and semantically related to a relational noun but is nonrelational. In (28)-(29) mother and grandmother are one-place predicates that denote the set of individuals who are the mother (or grandmother) of somebody. That is, if mother1’ and grandmother1’ are the meanings of the relational nouns, then (28) and (29) use the distinct lexical entries mother2 and grandmother2, where [[mother2]] = λx[∃y[mother1’(x,y)] ] [[grandmother2]] = λx[∃y[grandmother1’(x,y)] ]

4

(28)-(29) are similar to an example given by Daphna Heller (practice talk). (30) is an example of Barker’s “daycare reading” of child (Barker, 1995).

25

Tamina Stephenson

Honors Thesis Chapter 2

There is an important piece of independent evidence for Barker’s claim: we need these second lexical entries anyway to account for the fact that normally relational nouns such as mother and grandmother can be used predicatively as in (30). (30)

(a) Jane is a mother. (b) Jane is a grandmother. (c) John is a father. (d) John is a grandfather. (e) John is a great-grandfather.

= = = = =

∃x[mother( [[Jane]], x ) ∃x[grandmother( [[Jane]], x ) ∃x[father( [[John]], x ) ∃x[grandfather( [[John]], x) ∃x[great-grandfather([[John]], x)

Of course, this account does give the correct meaning for (28), since the person who Sue brings in just has to be the mother (or grandmother) of somebody. Many “parent-like” relational nouns can be used in this way, including mother, father, parent, grandmother, grandfather, grandparent, great-grandmother, great-great-grandmother, and so on. For example, we can say that a person qualifies as being a great-grandparent if there exists some other person who is their great-grandchild. Thus the sentences in (30) are all perfectly natural, and true just in case the given conditions hold. Another piece of independent evidence for Barker’s claim is the fact that not all relational nouns can be used in this way, that is, with existential quantification over one of the arguments as in (30). For example, Barker points out that the non-relational use of child (or children) as in (29) carries an additional entailment that the individual be young, which is not present in the relational reading. That is, for example, (31.a) is compatible with the situation described in (31.b) but is a contradiction with (31.c). (31)

(a) John is an adult. (b) John is the child of Bill. (c) John is a child.

In fact, I would argue that (31.c) does not even entail that John be the child of anyone. Due to facts about the world, of course, he must be the child of someone if he is a normal human being; however, imagine a situation, e.g., in a fairy tale or myth, where a little human boy is magically created without using normal reproductive processes. If the boy created has all the characteristics of being human and is sufficiently young, it would be perfectly normal to refer to him as a child, even though he is not the child of anybody. On the other hand, let’s imagine (following a suggestion of P. Jacobson) that we lived in

26

Tamina Stephenson

Honors Thesis Chapter 2

a universe where a significant number of humans are magically created, so that they never have parents, whereas others are conceived by a set of parents and born in the normal way. Moreover, let’s imagine that a person’s status with regard to this difference has great social importance – for example, let’s say that a person is only eligible to be a full citizen if they were born in the normal way, that is, if they have parents. It would still be odd in such a universe to ask someone whether they meet this requirement by uttering, “Are you a child?” Moreover, there turn out to be many relational nouns that don’t seem to have any non-relational reading available. That is, they cannot be used predicatively and they cannot take a possessive on just any contextually salient relationship. The first fact comes out particularly well when the nouns are put into questions of the form Are you a ____?, as in (32) below. While it is perfectly natural to ask someone about their family life by uttering (a) or (b), it is very odd to ask any of (c)-(f). (Note that in all these cases, the most likely interpretation to give is an existentially quantified reading analogous to that of mother, father, etc.) (32)

(a) Are you a mother / father? (b) Are you a grandparent? (c) ?* Are you son / daughter? (d) ?* Are you a relative? (e) ?* Are you a brother / sister? (f) ?* Are you a husband / wife / spouse?

Some of these, such as son, daughter, or relative, may have an obvious pragmatic explanation: due to facts about human reproduction in this world, every person who is male is the son of somebody, and every person who is female is the daughter of somebody, and of course everyone must be a relative of somebody, living or dead. So it could be that these just aren’t interesting properties to talk about. A similar argument may be available for more distant kinship relations such as cousin, second cousin, etc.: presumably, it has no great effect on a person whether or not they are the second cousin once removed of anybody. However, this story is less convincing for words like husband. It is generally considered a quite salient property of an adult whether or not

27

Tamina Stephenson

Honors Thesis Chapter 2

they have a spouse. However, if a speaker of English were to try to find this out, they would utter something like (33.a) or maybe (33.b), but never (33.c), despite the fact that the most natural meaning to assign to (33.c) is equivalent to (33.a) and (33.b).5 (33)

(a) Are you married? (b) Do you have a wife / husband? (c) ?* Are you a husband / wife?

For that matter, it is not necessary to have taken vows for this general situation to hold; it is equally bad to ask someone, “Are you a boyfriend / girlfriend?,” even though we can (and often do) ask something like, “Are you seeing anyone?” or “Do you have a boyfriend / girlfriend?” Similarly, it is perfectly natural to inquire about whether a person has any siblings or not, as in (34.a) or (34.b), but it is very odd to ask (34.c): (34)

(a) Do you have any brothers or sisters? (b) Do you have any siblings? (c) ?* Are you a brother / sister / sibling?

With these nouns that cannot be used predicatively, a possessive NP cannot take the kind of contextually salient reading shown in the apparent counterexamples in (28)-(29). For example, going back to the example of the class visitors in (28), suppose that the requirement was that the students each bring (a) a woman who has parents, (b) a woman who is related to somebody, (c) a woman who has one or more siblings, or (d) a woman who is married. (It might help to also pretend that this class is happening in a universe where some people are magically created without normal human reproduction or birth, so that having parents or relatives become non-trivial properties.) In these cases, we cannot use Sue’s daughter / relative / sister / wife to refer to the visitor, as shown in (35). (35)

[Context: A class is discussing families. Each student brings in a __.] (a) *? Sue’s daughter talked about her career as a writer. (b) *? Sue’s relative talked about her career as a writer. (c) *? Sue’s sister talked about her career as a writer. (d) *? Sue’s wife talked about her career as a writer. [Note that (d) is bad even assuming that Sue is female and that women can’t have wives.] 5

I don’t mean to open up the issue of the semantics of questions. When I say that these questions are equivalent, I just mean that the corresponding declarative sentences are truth-conditionally equivalent.

28

Tamina Stephenson

Honors Thesis Chapter 2

It is important to distinguish here between uses of a relational noun where an argument is supplied pragmatically from the context, as opposed to uses where there is existential quantification over the argument.6 For example, let’s suppose that there are three women (call them A, B, and C) and a baby (call it D) in a room. A baby-sitter comes to take care of the baby, and needs to know which woman to talk to. If the babysitter asks, “Who is the mother?” she is not asking which one of the three women has children, but which of them stands in the mother relation to this particular baby. To avoid the semantics of questions, let’s say that woman A answers by uttering (36): (36)

A: [pointing to B] She is the mother. (= B is the mother of D)

This is a case where the noun mother has its usual relational meaning, but one argument is supplied pragmatically. This is analogous to the situation with a verb such as agree, which can pick up a complement from context as in (37): (37) Mary said that Gone with the Wind is a good book, and John agreed. (= John agreed that Gone with the Wind is a good book) On the other hand, let’s suppose that the baby-sitter is not coming to take care of that particular baby but is just trying to scout out potential clients. In that case, she might ask, “Who is a mother here?” and woman A might respond by uttering (38): (38)

A: We are all mothers. (= A, B, and C each have the property of being the mother of somebody)7

This is analogous to intransitive eat which entails that something is eaten, as in (39). (39)

John is eating (= John is eating something)

6

My examples of an argument “being supplied from context” are similar to the “bound variable readings” of relational nouns which have been discussed by people such as B. Partee and J. Mitchell. (See, e.g., Mitchell, 1986 (The Formal Semantics of Point of View, Ph.D. Dissertation, University of Massachusetts, Amherst, MA: UMass GLSA Publication) and Partee, 1991 (“Anaphoric and deictic pieces of meaning,” paper presented at the 8th Amsterdam Colloquium, Amsterdam: ITLI). 7 Incidentally, note that, despite the purpose of the baby-sitter’s question, (38) would be true even if one of the women only had children who were too old to need a baby-sitter, although if she knew why the babysitter was asking, then she might answer differently. For simplicity, let’s just assume that woman A doesn’t know the purpose of the question.

29

Tamina Stephenson

Honors Thesis Chapter 2

Incidentally, although both of these properties are idiosyncratic with respect to verbs, it does seem that it is possible in general for relational nouns to have an argument supplied pragmatically. Going back to the previous examples, then, there are situations where some of (32.c-f) can become felicitous, as in (40). (40)

[Context: a wedding or a funeral. Person A turns to person B and asks…] A: Are you a friend or a relative?

At a funeral or family reunion, you could even imagine someone turning to their neighbor and saying, “I think I’ve met you somewhere before. Aren’t you a second cousin once removed, or something like that?” where the speaker means that the addressee is a second cousin once removed of the speaker, or perhaps (at a funeral) of the deceased. The point of all this is that if we eliminate cases where arguments are supplied pragmatically, then it is fairly idiosyncratic which relational nouns have non-relational counterparts. At least it is sufficiently idiosyncratic that it is reasonable to claim that derived and mixed nominals simply do not have that sort of second reading. To summarize, then, I have hopefully convinced you of two things: (1) if we treat relational nouns as two-place predicates, then it is simple to have the possessive directly “hook into” this relation; and (2) when the possessive seems to be able to give some other contextually salient relation with a relational noun, this is really a case of an idiosyncratic second lexical entry which is in fact a non-relational, normal noun. The upshot is that if we can just treat derived and mixed nominals as relational nouns, it should be easy to show why the possessive “subject” must be the subject of the implied verb. It only remains to formalize how these can be treated as relational nouns, which I do below. Analysis of derived and mixed nominals as relational nouns Let’s first assume that derived nominals such as John’s criticism of the book and mixed nominals such as John’s criticizing of the book are of semantic type <ε>, and gerundive nominals such as John’s criticizing the book are of semantic type . Thus all of the nominalized expressions denote individuals. (Remember that <ε> stands for “events” and for the individual correlate of a proposition, both of which are subtypes of individuals.) This correctly predicts that we can put these expressions into positions typically filled by NPs, such as the subject position of a sentence: 30

Tamina Stephenson

(41)

Honors Thesis Chapter 2

(a) John’s criticism of the book was surprising. (b) John’s criticizing of the book was surprising. (c) John’s criticizing the book was surprising.

Each of these sentences seems to be predicating the property [[surprising]] of some kind of event-like or propositional entity, namely that relating to the fact that John criticized the book. Exactly what is the nature of this kind of “individual” is an interesting question which I will not look at in depth. As mentioned in Chapter 1, Zucchi (1993) argues that expressions like (41.a-b) denote events while those like (41.c) denote propositions, a distinction that is motivated by several differences in the distribution and possible meanings of the expressions. Although I have adopted those labels, the analysis outlined below will remain largely neutral on this question of what exactly is the nature of eventlike and propositional individuals. Now, let’s see how we can get a meaning for a nominalized expression from the meaning of a verb. To simplify the exposition, I would like to consider intransitive verbs first. To do this, I will need to make a few observations and assumptions. Consider an intransitive verb such as growl. Just as with transitive verbs, we can nominalize this to give an expression such as the lion’s growling; however, in a sense we cannot tell whether this is a mixed nominal or a gerundive nominal, since the only difference between these two constructions was whether or not they contained an of-phrase with their direct object, and growl has no direct object. In fact it seems that it can be either one, since it can occur either with an adjective (like a mixed nominal) or an adverb (like a gerundive nominal): (42)

(a) The lion’s loud growling surprised us. (b) The lion’s loudly growling surprised us.

Note that the insertion of an adjective or adverb essentially disambiguates this expression into being a mixed nominal or a gerundive nominal. In the discussion below, I will assume that expressions such as the lion’s growling are mixed nominals. Keep in mind that you can always mentally insert an adjective such as loud to ensure that it is a mixed nominal instead of a gerundive nominal. Consider the verb growl in comparison to the mixed nominal the lion’s growling. The word growl is an intransitive verb, which is to say that it is of semantic type ,

31

Tamina Stephenson

Honors Thesis Chapter 2

that is, a one-place predicate. Let’s suppose that the grammar contains a type-shifting rule which takes a one-place predicate such as growl and gives back a two-place relation between growling-events and the individual responsible for them, i.e., the individual who growls. Such a rule is given in (43). (43)

Type-shifting rule applying to intransitive verbs to give mixed nominals8 α: < [α] ; S/LNP ; : α’> ⇒ β: < [α-ing] ; N/RPP[of] ; β’: > > β’ = λx[λε[ε is an instance of the proposition α’(x) ] ]

Thus the nominalized form growling is of the same type as a relational noun such as sister, where growling’(x,y), or growling’(y)(x), is true just in case x is an event-like individual that is an instance of the proposition that y growls. Thus growling is the right category to combine with the possessive, as shown in Figure 4.a. Note that growling could, of course, combine with an of-phrase such as of the lion to get a common noun growling of the lion, which can then combine with a determiner such as the to get an NP. This is shown in Figure 4.b.

8

There is one problem with the formulation of this rule, which is that we need to restrict it to apply to lexical verbs and not to any verb phrase. Otherwise this rule will incorrectly predict that expressions such as *John’s harsh criticizing the book will be well formed. How to ensure that the input will be a “word” instead of a phrase is beyond the scope of this project.

32

Figure 4.a

Derivation of the lion's growling as a mixed nominal

An intransitive verb in a mixed nominal NP the lion’s growling <ε> λR[ [[the]] ( λy[R(l )(y)] ) ] ( λx[λε[ ε is an instance of growl’(x) ] ] ) = [[the]] ( λy[ λx[λε[ ε is an instance of growl’(x) ] ] (l )(y)] ) = [[the]] ( λy[ y (of subtype <ε>) is an instance of growl’(l ) ] ) = the unique contextually salient instance of the proposition growl’( l )

NP/R(N/RPP[of])

N/RPP[of]

the lion’s < > , e > λx[λR[ [[the]] ( λy[R(y,x)] ) ] ] ( l ) = λR[ [[the]] ( λy[R(y, l )] ) ] = λR[ [[the]] ( λy[R(l )(y)] ) ]

growling > λx[λε[ ε is an instance of growl’(x) ] ]

NP

( NP/R(N/RPP[of]) )/LNP

the lion l

’s < e, <> , e> > λx[λR[ [[the]] ( λy[R(y,x)] ) ] ]

Related by productive category-changing rule < [α α] ; S/LNP ; : α’> ⇒ < [α α-ing] ; N/RPP[of] ; > : λx[λε λε[ ε is an instance of the proposition α’(x) ] ] >

S/LNP growl growl’

33

Figure 4.a

Derivation of the growling of the lion as a mixed nominal

Figure 4.b

An intransitive verb in a mixed nominal with an of -phrase NP the growling of the lion <ε> [[the]] ( λε[ ε is an instance of growl’(l ) ] ) = the unique contextually salient instance of growl’( l )

NP/RN

N

the < , e> [[the]]

growling of the lion <ε,t> λε[ε is an instance of growl’(l ) ]

N/RPP[of]

PP[of]

growling > λx[λε[ε is an instance of growl’(x) ] ]

of the lion l

Related by productive category-changing rule < [α α] ; S/LNP ; : α’> ⇒ < [α α-ing] ; N/RPP[of] ; > : λx[λε λε[εε is an instance of the proposition α’(x) ] ] >

S/LNP growl growl’

34

Figure 4.b

Tamina Stephenson

Honors Thesis Chapter 2

Now, let’s extend this analysis to transitive verbs. Consider the word criticize. This is an expression of semantic type >, that is, a two-place predicate. We can extend the rule in (43) to apply to transitive verbs, as in (44) below. This rule takes a two-place relation between “criticizers” and “criticizees,” and returns a three-place relation between criticizing-events, criticizers, and criticizees. (44)

Type-shifting rule applying to transitive verbs to give mixed nominals α: < [α] ; (S/LNP)/RNP ; > : α’> ⇒ β: < [α-ing] ; (N/RPP[of])/RPP[of] ; β’: >> : β’ = λw[λx[λε[ε is an instance of the proposition α’(w)(x) ] ] ] > = λw[λx[λε[ε is an instance of the proposition α’(x, w) ] ] ] >

The output of this rule thus denotes a three-place predicate, criticizing’, which is true of a triple (z,x,w) just in case z an event that is an instance of the proposition that x criticizes w. That is, the output of (44) is essentially a three-place relational noun. Under the formulation we have given for the possessive, an expression such as John’s or the lion’s can only combine with an expression denoting a two-place relation, not a three-place relation. Thus as it stands, the possessive can’t combine directly with criticizing under this analysis. On the other hand, an of-phrase can come in as an argument for either a two-place or a three-place nominal. So criticizing must instead combine first with an ofphrase such as of the book, at which point the expression criticizing of the book will be of the right type to combine with the possessive. In fact this correctly predicts that a nominal derived from a transitive verb cannot take both of its arguments in the form of the possessive, as in (45). (45)

(a) * John’s (the) book’s criticizing (b) * the lion’s deer’s hunting

Unfortunately, this does not explain why it is also apparently impossible for the same nominal to take both arguments in the form of of-phrases, as in (46). (46)

(a) * the criticizing of the book of John (b) * the hunting of the deer of the lion

I don’t have a good explanation for this fact. It could possibly be related to an observation which has often been made that there don’t seem to be any three-place verbs

35

Tamina Stephenson

Honors Thesis Chapter 2

that can form nominal expressions along these same lines.9 For example, it is impossible to form the NP *John’s giving of Mary of the book corresponding to the sentence John gave Mary the book. If there were some reason why it was just impossible to apply two of-phrases to the same three-place predicate, then this would account for both of these facts. Again, though, it is not clear what this reason would be, so I’ll leave this aside as an open problem. There are also some apparent counterexamples to the claim that possessives cannot combine directly with nominals related to transitive verbs, shown in (47). These types of examples are discussed in terms of passivization in Chomsky (1970), and are sometimes referred to as “passive nominals” because the possessive NP seems to have the same role as the subject in a passive construction. (47)

(a) the city’s destruction (b) John’s murder (c) Kennedy’s assassination

(= the city was destroyed) (= John was murdered) (= Kennedy was assassinated)

I claim that these are idiosyncratic homonyms of the standard derived nominal. There is simply a second lexical item for destruction (and for murder and for assassination) that denotes a 2-place relation between an event and the individual destroyed (or murdered or assassinated) in that event. This is similar to the second lexical entry for mother which is a non-relational noun. One piece of evidence for the claim that these are idiosyncratic is the fact that they cannot be replaced with a -ing form, as shown in (48). (48)

(a) ?* The city’s destroying (* by the enemy) (b) ?* John’s murdering (c) ?* Kennedy’s assassinating

These can only describe the situations where the city, John, and Kennedy are the agents of the action, and are grammatical only to the extent that (49.a)-(49.c) are. (49)

(a) ?* The city destroys (b) ?* John murders (c) ?* Kennedy assassinates

9

See, for example, Grimshaw, 1990 (pp. 70-80) and M. Rappaport, 1983 (“On the nature of derived nominals,” in L. Levin, M. Rappaport, and A. Zaenen, eds., Papers in Lexical-Functional Grammar: pp. 113-142, Indiana Linguistics Club).

36

Tamina Stephenson

Honors Thesis Chapter 2

Returning to the rule in (44), so far this only accounts for mixed nominals, not derived or gerundive nominals. It is simple to extend this to derived nominals, however, if we assume that criticism is derived from criticize by a rule like (44) except for the phonology. (In fact I will argue in Chapters 3 and 4 that derived nominals are not generated by any productive rule, but this will only have a minor effect on the analysis.) Now we can give a derivation for the derived nominal John’s criticism of the book and the mixed nominal John’s criticizing of the book, as shown in Figure 5.a and Figure 5.b.

37

Derivation of John's criticism of the book

Figure 5.a

Derivation of a derived nominal NP John’s criticism of the book λR[ [[the]] ( λy[R(j )(y)] ) ] ( λx[λε[ε is an instance of criticize’(b )(x) ] ] ) = [[the]] ( λy[ λx[λε[ε is an instance of criticize’(b )(x) ] ] (j )(y)] ) = [[the]] ( λy[y (of subtype <ε>) is an instance of criticize’(b )(j ) ] ) = the unique contextually salient instance of criticize’(b )(j )

NP/R(N/RPP[of])

N/RPP[of]

John’s < > , e > λx[λR[ [[the]] ( λy[R(y,x)] ) ] ] ( j ) = λR[ [[the]] ( λy[R(y, j )] ) ] = λR[ [[the]] ( λy[R(j )(y)] ) ]

criticism of the book > λw[λx[λε[ε is an instance of criticize’(w)(x) ] ] ] (b ) = λx[λε[ε is an instance of criticize’(b )(x)] ]

NP

( NP/R(N/RPP[of]) )/LNP

John j

’s < e, <> , e> > λx[λR[ [[the]] ( λy[R(y,x)] ) ] ]

(N/RPP[of])/RPP[of]

PP[of]

criticism > > λw[λx[λε[ε is an instance of criticize’(w)(x) ] ] ]

of the book b

Related by semi-productive derivational rule < [α α], (S/LNP)/RNP, > : α’ > ⇒ < [α α-ism], (N/RPP[of])/RPP[of], >> : λw[λ λx[λε λε[εε is an instance of the proposition α’(x,w) ] ] ] = λw[λ λx[λε λε[εε is an instance of the proposition α’(w)(x) ] ] ] >

(S/LNP)/RNP criticize > criticize’

38

Figure 5.a

Derivation of John's criticizing of the book

Figure 5.b

Derivation of a mixed nominal NP John’s criticizing of the book λR[ [[the]] ( λy[R(j )(y)] ) ] ( λx[λε[ε is an instance of criticize’(b )(x) ] ] ) = [[the]] ( λy[ λx[λε[ε is an instance of criticize’(b )(x)] ] (j )(y)] ) = [[the]] ( λy[y(of sub-type <ε>) is an instance of criticize’(b )(j )] ) = the unique contextually salient instance of criticize’( b )( j ) [= criticize’( j , b )]

NP/R(N/RPP[of])

N/RPP[of]

John’s < > , e > λx[λR[ [[the]] ( λy[R(y,x)] ) ] ] ( j ) = λR[ [[the]] ( λy[R(y, j )] ) ] = λR[ [[the]] ( λy[R(j )(y)] ) ]

criticizing of the book > λw[λx[λε[ε is an instance of criticize’(w)(x) ] ] ] (b ) = λx[λε[ε is an instance of criticize’(b )(x) ] ]

NP

( NP/R(N/RPP[of]) )/LNP

John j

’s < e, <> , e> > λx[λR[ [[the]] ( λy[R(y,x)] ) ] ]

(N/RPP[of])/RPP[of]

PP[of]

criticizing > > λw[λx[λε[ε is an instance of criticize’(w)(x) ] ] ]

of the book b

Related by productive category-changing rule < [α α], (S/LNP)/RNP, > : α’ > ⇒ < [α α-ing], (N/RPP[of])/RPP[of], >> : λw[λ λx[λε λε[εε is an instance of the proposition α’(x,w) ] ] ] = λw[λ λx[λε λε[εε is an instance of the proposition α’(w)(x) ] ] ] >

(S/LNP)/RNP criticize > criticize’

39

Figure 5.b

Tamina Stephenson

Honors Thesis Chapter 2

Derived result nominals Derived result nominals can keep the analysis we gave in Chapter 1. (Recall that derived result nominals are idiosyncratic forms such as criticism on its reading meaning “a critical paper” or “a critique.”) They are idiosyncratic anyway, both semantically and morphologically, so it is not a problem to just list them in the lexicon. Moreover, recall that the meanings for derived nominals involved existential quantification over one of the arguments; in some cases we might also want some extra entailments (such as possibly being written down, in the case of criticism). This means that derived result nominals are analogous to the second, non-relational lexical entries for relational nouns such as mother; when they have extra entailments, they are analogous to the non-relational child. Revised analysis of gerundive nominals The type of analysis given above for derived and mixed nominals will not work for gerundive nominals, because of the observations made about them in Chapter 1, in particular the fact that they cannot combine with adjectives or most determiners. Therefore we could just let gerundive nominals keep the sort of analysis they were given in Chapter 1. However, the possessive does seem to have a very similar role in all three constructions, so it would be nice to try to generalize it. The following analysis is in a sense more complex than the original analysis of gerundive nominals, but the role of the possessive NP is much more similar to what it is with derived and mixed nominals. First, recall that in the analysis given in Chapter 1 (Figure 2.c), an expression such as criticizing the book was assigned the syntactic category NP/LNP[-s] and semantic type and its meaning was that function which takes an individual x and yields the proposition-like individual corresponding to the proposition x criticizes the book. This simply took as an argument a possessive NP, that is, an NP with the feature “[-s].” Under this analysis, we have two different versions of the possessive, one of which is a function applying to relational nouns and one of which is basically the same as an NP (in addition to the possessive which applies to non-relational nouns, which I continue to ignore). However, we could instead do this: Let’s forget about the special feature [-s], and say that an expression such as criticizing the book is of category NP/LNP, with the same meaning

40

Tamina Stephenson

Honors Thesis Chapter 2

as before. This means that it can combine with an NP such as John to get John criticizing the book – which, incidentally, is another acceptable form of nominalization. Now, we can posit a second lexical entry for the possessive which is slightly different from the one previously given but enticingly similar to it, as shown in (50) below. Note that R is a variable of type > and U is a variable of type . (50)

Previously: ‘s-1: category ( NP/R(N/RPP[of]) )/LNP [[‘s-1]] = λx[λR>[ [[the]] ( λy[R(y,x)] ) ] ] = λx[λR>[ [[the]] ( λy[R(x)(y)] ) ] ] = λx[λR>[ [[the]] ( R(x) ) ] ] So if [[John]] = j, then [[John’s-1]] = λR[ [[the]] ( λy[R(j)(y)] ) ] 2nd reading:

‘s-2: category ( NP/R(NP/LNP) )/LNP [[‘s-2]] = λx[λU[U(x)] ] So [[John’s-2]] = λU[U(j)]

In the first case John’s selects for an argument of type >, applies this to [[John]], and finally applies the operator [[the]] to the result. In the second case, John’s selects for an item of type and applies this function to [[John]]. (The [[the]] operator cannot apply in this case because there is no set from which to select an individual.) To generalize the meaning of the possessive to cover both kinds of cases, we could say something like the following: (1) The possessive morpheme ‘s works by first combining with an NP as argument, which denotes some individual, let’s call it a. This expression then combines with some expression of type , with meaning ëx[…]. Let’s call this expression B. Now, to get the meaning of the entire expression, the semantics tells us to first apply the meaning of B to the argument a. In other words, a possessive NP is basically an argument of type that has been type-lifted over a function of type . (2) Then, if the result is an individual, that individual is the denotation of the entire expression; if the result is a set of individuals, then apply the operator [[the]] to select the single maximal contextually salient member of that set. This isn’t really a very good generalization, since we have to resort to the “if/otherwise” type of wording in part (2). On the other hand, part (1) is reasonably general, and corresponds to the fairly well accepted process of type-lifting, so hopefully this is on the right track. This revised analysis of gerundive nominals is illustrated below in Figure 6.

41

Figure 6

Revised analysis of gerundive nominals

Categorial Grammar analysis of gerundive nominals NP John’s criticizing the book λU[U( j )] ( λx[the individual correlate of criticize’(x, b )] ) = λx[the individual correlate of criticize’(x, b )] ( j ) = the individual correlate of criticize’( j , b )

NP/R(NP/LNP)

NP/LNP

John’s < , e > λx[λU[U(x)] ] ( j ) = λU[U( j )]

criticizing the book λx[the individual correlate of criticize’(x, b ) ]

NP

( NP/R(NP/LNP) )/LNP

John j

’s < e, < , e > > λx[λU[U(x)] ]

Related by productive category-changing rule < [V α], S/LNP, : P > ⇒ < [V-ing α], NP/LNP, : λx[the individual correlate of the proposition P(x)] >

S/LNP criticize the book λx[criticize’(x, b )]

(S/LNP)/RNP

NP

criticize > λyλx[criticize’(x,y)]

the book b

42

Figure 6

Tamina Stephenson

Honors Thesis Chapter 2

To sum up, we can now account for the role of the possessive and the of-phrase in derived nominals like (1) and mixed nominals like (3), using the apparatus of relational nouns and a meaning for the possessive which has a way to “hook into” the relation expressed by a relational noun. Since these nominals are derived from transitive verbs, which express a relationship between the subject and object of an action, the most natural relation to be expressed in the nominals is that of being the subject and object of an action, which is exactly what the possessive and of-phrases seem to express. This account could not be extended directly to cover gerundive nominals because of the differences in distribution of the constructions, but it could be partly generalized to make the analysis of gerundive nominals more similar to that of derived and mixed nominals.

43

Chapter 3: The distribution of modifiers with nominals Adjectives and adverbs with nominalized expressions In this chapter I will explore the question of what kinds of modifiers can appear with derived, mixed, and gerundive nominals. In Chapter 1, we saw that in general, derived and mixed nominals can appear with adjectives but not adverbs, whereas gerundive nominals can appear with adverbs but not adjectives. This was one way in which derived and mixed nominals behaved like nouns while gerundive nominals behaved like verb phrases or sentences. The analysis given in chapter 2, treating derived and mixed nominals as relational nouns (type >), but mixed nominals as functions of type , was consistent with this generalization. However, it turns out that this empirical generalization is somewhat oversimplified. It is true that adjectives can appear with derived and mixed nominals but not with gerundive nominals, but the situation with adverbs is a bit more complicated. In particular, I claim that there is a class of adverbs which can appear with mixed nominals, at least in certain contexts, though they cannot ever appear with derived nominals. Before I present this claim, it is important to distinguish between two major classes of modifiers that could be classified as adverbial since they can modify verbs or verb phrases. The first is the type of adverb, such as slowly or harshly, which can appear typically either at the end of a verb phrase or immediately before the verb, as in (51)-(52). (51)

(a) John criticized the book harshly. (b) John harshly criticized the book.

(52)

(a) The enemy destroyed the city slowly. (b) The enemy slowly destroyed the city.

Of course, these cannot combine with nouns. (53)

(a) * That rebuke harshly was unfair.

[cf., That harsh rebuke was unfair.]

The other kind of modifier I will consider includes various expressions which typically go at the end of a verb phrase and often denote temporal or spatial aspects of events. These include prepositional phrases such as on Tuesday or to Providence, as well as single words such as yesterday. These cannot appear immediately before the verb, as illustrated in (54)-(55) below.

45

Tamina Stephenson

(54)

(a) Mary sang a song yesterday. (b) * Mary yesterday sang a song.

(55)

(a) Mary drove to Providence. (b) * Mary to Providence drove.

Honors Thesis Chapter 3

It turns out that the second kind of “adverbial” modifier has a much wider distribution than the first. Chomsky pointed out that many nouns seem to take prepositional phrases as complements, citing examples such as the weather in England, the message from Bill to Tom about the meeting, the prospects for peace, and many others (Chomsky, 1970: p. 196). In fact, single words such as yesterday can also do so provided that the meaning of the noun has an appropriate temporal component (i.e., denotes what is intuitively some kind of event). For example, the party yesterday can be equivalent to the party that took place yesterday. Consider (56.a), for instance. (56)

(a) Mary told Bill all about the party yesterday. (b) * Mary told Bill all about the yesterday party.

Sentence (56.a) has a reading where the discussion of the party could have taken place today, provided that the party took place yesterday. Similar examples can be constructed using other nouns such as sunset, lecture, class, weather, or even event. Note, though, that they can appear only before the noun, not after, as in (56.b) above. Since this second kind of modifier can appear with normal nouns such as party, it is not surprising that they can also occur with derived and mixed nominals, as in (57.a-b). Again, of course, they can only appear after the nominalized expression. (57)

(a) The enemy’s destruction of the city yesterday shocked the inhabitants. (b) The enemy’s destroying of the city yesterday shocked the inhabitants. (c) * The enemy’s yesterday destruction of the city shocked the inhabitants. (d) * The enemy’s yesterday destroying of the city shocked the inhabitants.

Thus when I talk about whether derived and mixed nominals can appear with adverbs, I am only talking about the first class of adverbs – those like harshly or slowly, which can appear either before a verb or at the end of the verb phrase. Keeping this in mind, let’s see what happens when we try to combine derived and mixed nominals with adjectives and with adverbs, as in (58)-(59).

46

Tamina Stephenson

Honors Thesis Chapter 3

(58)

(a) The enemy’s slow destruction of the city took six months. (b) Her harsh criticism of the book surprised the author. (c) * The enemy’s slowly destruction of the city took six months. (d) * Her harshly criticism of the book surprised the author.

(59)

(a) The enemy’s slow destroying of the city took six months. (b) Her harsh criticizing of the book surprised the author. (c) * The enemy’s slowly destroying of the city took six months. (d) * Her harshly criticizing of the book surprised the author.

According to the data in (58)-(59), both derived and mixed nominals can occur with adjectives, but neither with adverbs. This is just what we saw in Chapter 1. But now consider (60.a). (60)

(a) The burning or otherwise destroying of records is illegal. (b) * The burning or otherwise destruction of records is illegal.

In (60.a), a mixed nominal appears directly preceded by the adverb otherwise, and yet the resulting expression is well formed. On the other hand, if we replace destroying with the corresponding derived nominal destruction, as in (60.b), it becomes ill-formed. Again, this is not to be confused with the second type of modifier discussed above, which can appear after verb phrases or nouns. If we take out otherwise and put in a roughly equivalent modifier of this second kind, such as in some other way, both the derived and the mixed nominal become acceptable, as illustrated in (61). (61)

(a) The burning of records or the destroying of them in some other way is illegal. (b) The burning of records or the destruction of them in some other way is illegal.

Note also that the well-formedness of (60.a) does not depend on the fact that both nominalized expressions in the coordinate structure (burning … or destroying…) are mixed nominals. Neither does the ill-formedness of (60.b) depend on the fact that one is a derived form and one is a mixed form. It is fine in this kind of sentence for one of the nominalized expressions to be a derived form and the other a mixed form, provided that the one preceded by otherwise is a mixed form. This is illustrated in (62). (62)

(a) The destruction or otherwise damaging of records is illegal. (b) * The destruction or otherwise removal of records is illegal. [cf. the destruction or otherwise removing of records]

47

Tamina Stephenson

Honors Thesis Chapter 3

At this point, you might be wondering if this is just something special about the adverb otherwise. Maybe it isn’t truly an adverb, or maybe it is being used in some special meta-linguistic way, where “destruction or otherwise damaging of records” means something like “destruction, or we could also include all kinds of damaging of records.” One or both of these objections may or may not be true, and would seem very likely if otherwise were the only adverb with the property of appearing with mixed nominals (although this would still beg the question of why it cannot appear with derived nominals). However, it turns out that there is a whole other class of adverbs that show the same pattern as otherwise, including practically, pretty much, almost, and completely. Using a simple text-based web search engine, I was able to find attested examples of many of these, which are sampled in (63) below.1 In (64), I have constructed some similar examples with different adverbs. In the cases where there exists a corresponding derived nominal, I have added in a constructed example showing that the same adverb is unacceptable with the derived form. (63)

Attested examples of adverbs with mixed nominals: (underlines added) [from a junior-senior high school choral department handbook] (a) There shall be no defacing or otherwise damaging of any Choral Department property. [ http://www.msdwc.k12.in.us/msdclass/nevm/handbook.htm ] [from an Irish Parliamentary debate about agricultural law codes] (b) In an action for damages for the shooting or otherwise destroying of a dog it shall be a good defence if the defendant proves [that] …. [ http://www.oireachtas-debates.gov.ie/S.0052.196006080004.html ] vs. *… the shooting or otherwise destruction of a dog …. [from an amateur travel report ] (c) [T]hey have a wonderful museum and show a great movie depicting the building, burying, and finally almost destroying of the statues. [ http://china.bgmm.com/Xian.htm ] vs. *… and finally almost destruction of the statues. [from the description of an online race-car simulation game] (d) [R]olling and completely destroying of the car is possible. [ http://www.f1gamers.com/f1/games/info.php?game=gp3&info=features ] vs. * Rolling and completely destruction of the car is possible.

1

Google Advanced Search form, http://www.google.com/advanced_search?hl=en. 4/11/2002. My original observation about otherwise came from a similar online search from the fall of 2001.

48

Tamina Stephenson

Honors Thesis Chapter 3

(64) More examples of adverbs with mixed nominals (constructed): (a) Vandalism is so prevalent in that town because the people don’t care to prevent it. The damaging and even practically destroying of property is considered a fact of life. vs. (a’) * Vandalism is so prevalent in that town because the people don’t care to prevent it. The damaging and even practically destruction of property is considered a fact of life. (b) Vandalism is so prevalent in that town because the people don’t care to prevent it. The damaging and even pretty much destroying of property is considered a fact of life. vs. (b’) * Vandalism is so prevalent in that town because the people don’t care to prevent it. The damaging and even pretty much destruction of property is considered a fact of life. This means that the generalizations we made previously about the distribution of modifiers with nominals were not quite right. It is true that adjectives can appear with derived and mixed nominals but not with gerundive nominals; and it is true that adverbs can appear with gerundive nominals but not derived nominals. But as we have seen, some adverbs can appear with mixed nominals in at least some contexts. Possible explanations for adverbs with mixed nominals Intuitively, it seems that the adverbs used in (60)-(64) fall into a natural semantic class. All of them except otherwise seem to express something like the extent to which an action is completed. I have no good explanation for why anything about the meaning of these particular adverbs would make them more felicitous with mixed nominals than other adverbs. (They do have some syntactic similarities, which I will return to below.) One might notice, though, that most of these adverbs do not have corresponding morphologically related adjectives. (Note that the meaning of practically used above is not the one that would be derived from practical.) Thus, one might propose that there is some sort of blocking effect or constraint such that an adjective would be preferred when available. This cannot be the right way to account for it, however, since the adverb completely does have a corresponding adjective, complete, with essentially the same meaning. The same is true of the adverb virtually, as shown in (65). Of course, the adjective is also acceptable with the mixed nominal:

49

Tamina Stephenson

(65)

Honors Thesis Chapter 3

(a) * Vandalism is so prevalent in that town because the people don’t care to prevent it. The damaging and even virtually destruction of property is considered a fact of life. (b) … The damaging and even virtual destruction of property is considered a fact of life. (c) Vandalism is so prevalent in that town because the people don’t care to prevent it. The damaging and even virtually destroying of property is considered a fact of life. (d) … The damaging and even virtual destroying of property is considered a fact of life.

Another possibility might be to say that there is something special about coordination that causes this phenomenon. It is certainly true that these facts are brought out more clearly in coordinated structures, but this won’t explain the situation. First of all, under any accepted account of coordination, it is assumed that a conjunction like or simply combines with two well-formed expressions of the same syntactic category and semantic type to make a larger expression of the same type. This boils down to the following phrase structure rule or some variant of it: (66)

X3 à X1

Conj

X2 ;

Conj à and, or, …

Unless we can come up with independent evidence that a conjunction somehow allows one of its conjuncts to not actually be a well-formed expression of the right category, then it’s not clear in what sense the facts given in (60)-(64) could come from a property of coordination. More importantly, even if we could think of such a reason, it would not explain why this effect occurs only with mixed nominals and not with derived nominals or, for that matter, with regular nouns. So let’s try another tack. If the adverbs otherwise, completely, etc. have no way to combine directly with a noun-like expression such as criticizing or criticizing of the book after it is formed, then they must combine with the expression at some point before the noun is formed. In other words, there are two obvious possibilities: (1) that they combine with verb phrases such as criticize the book, i.e, expressions of type ; or (2) that they combine with single verbs such as criticize, i.e., expressions of type > or . It’s easy to see that (1) will not work. Let’s try to derive the NP from (60), the burning or otherwise destroying of records. The word destroy is of type >, so it

50

Tamina Stephenson

Honors Thesis Chapter 3

cannot combine with the adverb otherwise until it takes an object. It takes the object records to become destroy records, of type , which can then combine with the adverb otherwise to get otherwise destroy records, also of type . Let’s assume for a moment that this can undergo the process that generates mixed nominals and become otherwise destroying (of) records, an expression of type >. On the other hand, on the left side of the conjunction, the verb burn is of type >, so it undergoes the process generating mixed nominals to become burning, of type >>. Now we cannot combine burning with otherwise destroying (of) records using or, because they are not the same type. On the other hand, let’s derive this NP again assuming that the adverb otherwise can combine directly with destroy. The verb destroy is of type , so it combines with otherwise to give otherwise destroy, also of type >. This then undergoes the process generating mixed nominals to become otherwise destroying, of type >>. On the other hand, burn is also of type >, so it also goes through the process for mixed nominalization to become burning, of type >>. (These both have the same syntactic category as well, of course, which is that of transitive verbs.) Now burning and otherwise destroying are of the same type, so they can conjoin with or to become burning or otherwise destroying, also of type >>. This, finally, can take the object of records to become a relational noun, of type >. Therefore, out of the various obvious possibilities for combining otherwise (or almost, practically, etc.) with a mixed nominal, the most plausible way is to have the adverb combine directly with the verb. There is in fact some important syntactic evidence, which I have ignored up till now, that suggests that these particular adverbs combine with verbal expressions differently than other adverbs. This has to do with their distribution just in normal sentences: Note that typical adverbs, such as harshly or happily, can appear either at the beginning or the end of verb phrases, as illustrated in (67)-(68) below. (67)

(a) John harshly criticized the book. (b) John criticized the book harshly.

(68)

(a) John happily read the book. (b) John read the book happily. 51

Tamina Stephenson

Honors Thesis Chapter 3

As I brought up in the beginning of the chapter, there are some adverbial expressions that only occur at the end of the verb phrase, such as yesterday; let’s disregard these for now. On the other hand, this new class of adverbs (otherwise, practically, almost etc.), with the exception of completely, generally appear only at the beginning of the verb phrase, before the verb; unless accompanied by certain marked patterns of intonation, they are odd at the end of the verb phrase.2 Examples of this are given in (69)-(73) below. (69)

(a) If you shout or otherwise make noise, your neighbors will get mad. (b) ?* If you shout or make noise otherwise, your neighbors will get mad.

(70)

(a) The enemy practically destroyed the city. (b) ?* The enemy destroyed the city practically.3

(71)

(a) The patient almost died from the disease. (b) ?* The patient died almost from the disease. (b’) ?* The patient died from the disease almost.

(72)

(a) The loud concert virtually deafened us. (b) ?* The loud concert deafened us virtually.

(73)

(a) The enemy pretty much destroyed the city. (b) ?* The enemy destroyed the city pretty much.

There are various ways one might formulate the difference between these two classes of adverbs – that is, those that can appear either at the beginning or the end of the verb phrase, and those that can only appear at the beginning. One possibility is to say that otherwise, practically, almost and so on combine at the verb-level, whereas the others combine at the VP-level. I don’t have any independent evidence for this kind of analysis over one in which all adverbs combine at the VP-level; but this would be consistent with the observation that only this second class of adverbs can appear in mixed nominals. (Note that the behavior of completely, which was exceptional in being able to appear in mixed nominals and also at the end of a verb phrase, could be accounted for if we let

2

I happened upon this observation while I was working on the final revisions of this project, and I regret that I did not have time to determine who may have made it before. I apologize for omitting anyone who ought to have been given credit here. 3 Here, take seriously my note about “marked intonation.” It's true that you can get practically at the end of a sentence like this, but it strikes me as very odd unless you put heavy stress on destroy or destroy the city. This may be a sort of meta-linguistic use of the adverb practically.

52

Tamina Stephenson

Honors Thesis Chapter 3

completely be ambiguous between the two types of adverb.) I will have to leave it to future work to determine the best way to deal with these different classes of adverbs. In any case, as discussed above, it seems most likely that otherwise, almost, and so on combine in at the verb-level when they appear in mixed nominals, so let me show a derivation of this. It will be illustrated using the adverb otherwise, so I should first say a few words about the meaning of otherwise. Consider a verb phrase such as burn or otherwise destroy records. Intuitively, part of the semantic contribution of otherwise includes the understanding that there is an expansion of the type of activities include in the discussion. That is, otherwise destroy records is an expansion of the type of activities included under burn [records]. More precisely, it is understood that the set of individuals who burn records is a proper subset of the set of individuals who destroy records; if you prefer, the set of events in which someone burns records is a proper subset of the set of events in which someone destroys records. The meaning of burn or otherwise destroy records thus has the same extension as just destroy records. But another part of the semantic contribution of otherwise is a sense in which this extension is being divided into two categories, in this case, the category of record-destroying involving burning, and not involving burning. So I would say that what otherwise does is the following: It combines with a verb such as destroy, whose meaning can ultimately be given as a set (a set of events, a set of destroyer/destroy-ee pairs, etc.), and it gives back that same set minus a contextually salient subset of it. Let me give this slightly more formally. Assume that f is a variable of type >, where “…” is possibly null – for the purposes of the examples here, we can assume that f is of type >. The notation fS refers to the set which f characterizes. Now we can give the meaning of otherwise as in (74): (74)

[[otherwise]] = λf[g: gS = fS – hS] with f, g, and h of the same semantic type, hS ⊂ fS, and h salient in context

There may be other possible ways to formulate the meaning of otherwise, but this will suffice. Now, Figure 7 below shows the derivation of the mixed nominal burning or otherwise destroying of records. For simplicity, I will assume that or is of a special category, CONJ, which takes two expressions of the same category, one on each side, to give another expression of the same category. In other words, I just assume that or is

53

Tamina Stephenson

Honors Thesis Chapter 3

introduced by a phrase structure rule such as (66). It should be fairly trivial to translate this into your favorite theory of syntax. The final meaning requires us to choose an h of type > from context that fulfills the right requirements. Since burn is present in the expression, [[burn]] can fill this role.

54

Figure 7

Derivation of burning or otherwise destroying of records

Derivation of a mixed nominal with an adverb N/RPP[of] burning or otherwise destroying of records > λw[λx[λε[ε is an instance of burn’(w)(x) or g(w)(x) ] ] ] (r ) = λx[λε[ε is an instance of burn’(r )(x) or g(r )(x) ] ]; gS = destroy’S – hS; hS ⊂ destroy’S salient in context

(N/RPP[of])/RPP[of]

PP[of]

burning or otherwise destroying > > λw[λx[λε[ε is an instance of burn’(w)(x) or ε is an instance of g(w)(x)] ] ]; gS = destroy’S – hS; hS ⊂ destroy’S salient in context

of records r

(N/RPP[of])/RPP[of] burning > > λw[λx[λε[ε is an instance of burn’(w)(x)] ] ]

[CONJ] or or’

Related by productive rule < [α α], (S/LNP)/RNP, > : α’ > ⇒ < [α α-ing], (N/RPP[of])/RPP[of], >> : λw[λ λx[λε λε[εε is an instance of α’(x,w) ] ] ]

(N/RPP[of])/RPP[of] otherwise destroying > > λw[λx[λε[ε is an instance of otherwisedestroy’(w)(x)] ] ] = λw[λx[λε[ε is an instance of g(w)(x)] ] ]; gS = destroy’S – hS; hS ⊂ destroy’S salient in context Related by productive rule < [α α], (S/LNP)/RNP, > : α’ > ⇒ < [α α-ing], (N/RPP[of])/RPP[of], >> : λw[λ λx[λε λε[εε is an instance of α’(x,w) ] ] ]

(S/LNP)/RNP

(S/LNP)/RNP

burn > burn’

otherwise destroy > λf[g: gS = fS – hS](destroy’) = g: gS = destroy’S – hS; hS ⊂ destroy’S salient in context

( (S/LNP)/RNP ) /R (S/LNP)/RNP )

(S/LNP)/RNP

otherwise < > , > > λf[g: gS = fS – hS]; hS ⊂ fS; h salient in context

destroy > destroy’

55

Figure 7

Tamina Stephenson

Honors Thesis Chapter 3

If something like the analysis in Figure 7 is correct, then the process creating mixed nominals out of transitive verbs must be able to apply to “complex verbs” such as otherwise destroy, whereas the process creating derived nominals cannot do so. This means that there is an important sense in which mixed nominalization is more productive than derived nominalization. This shouldn’t come as a great surprise, of course. After all, derived nominal forms are morphologically unpredictable and do not exist for all verbs. This is what led Chomsky and others to account for them using a semi-productive “lexical rule” rather than a transformation or a productive category-changing rule. Under this type of analysis, a derived nominal is a noun-like expression from the very beginning, so there is never a place for an adverb to come in. At the other extreme, the process creating gerundive nominals is completely productive, as well as morphologically regular, and seems to be able to apply to virtually any verb phrase. The process creating mixed nominals is apparently somewhere in the middle between the two extremes. The morphology is completely predictable. Also, the process can apply to virtually any transitive or intransitive verb in the lexicon, and also a certain number of “complex verbs” consisting of an adverb plus a verb. On the other hand, it cannot apply to all “complex verbs”; for example, we saw in Chapter 1 that it cannot apply to those containing auxiliary verbs such as have destroyed (which would give *having destroyed of the city). Ultimately we would want to give a precise characterization of just when it can apply to this kind of complex verb, and why. This still eludes me at this point, but given the syntactic and semantic similarity of the adverbs from (60)-(64) it seems promising to appeal to some property of a class of adverbs. In summary, then, I have shown in this chapter that there actually are cases where a mixed nominal can be modified by a pre-verbal adverb. More precisely, a mixed nominal can sometimes be formed from a “complex verb” that already includes an adverb. Derived nominals, on the other hand, absolutely cannot do so. This gives evidence that mixed nominalization is more productive than derived nominalization in the sense that it can apply more generally to input expressions of a certain type. On the other hand, its productivity is much more limited than that of gerundive nominals, and seems to possibly be constrained by a property of a certain class of adverbs.

56

Chapter 4: The productivity of the three kinds of nominalization Goals of the chapter: Previous chapters have been devoted to exploring and accounting for the output of the processes that yield nominals: making sure that they occur with the right kinds of modifiers, giving them the right semantic type and argument structure, and so on. At this point I will take a closer look at the nominalization processes themselves, in particular at how productive they are. Now, we can easily see that derived nominalization is not productive in a morpho-phonological sense, and mixed and gerundive nominalization clearly are productive. A more interesting question is that of semantic productivity, that is, the extent to which the meaning of a nominalized expression (derived, mixed, or gerundive) can be predicted by the meaning of its corresponding verb. I will ultimately argue that derived nominalization, even in the case of a true “process” nominal, is not a productive process in this respect, whereas both mixed nominalization and gerundive nominalization are productive over their respective domains. I will use two main types of evidence for the claim that derived nominals are not productive. The first is strictly semantic: I will show that certain meanings of verbs are available to mixed nominals but not to derived nominals, meaning that derived nominals are semantically “frozen.” The second type of evidence comes from the distribution of derived and mixed nominals. Using attested examples, I will argue that corresponding derived and mixed nominals (e.g., destruction of and destroying of) are often both available in many of the same contexts, with roughly the same meaning. This indicates that derived nominals do not have a “blocking effect” on mixed nominals, suggesting that, contra Harley & Noyer (1997), the two are not just “regular” and “irregular” forms of the same process, but are separate processes which happen to act similarly. Notions of “regular” and “productive” For the sake of precision, let me first define a few terms. Regular, as I will use it, applies to a particular form or expression, especially something that could be the output of a rule. For example, the verb form drives is regular, whereas the form driven is irregular. In general, it only makes sense to talk about an “irregular” form when there is some other form that would be “regular” and which would have the same meaning, but whose existence is blocked by the existence of the irregular form. For example, it makes

57

Tamina Stephenson

Honors Thesis Chapter 4

sense to call driven an irregular participle form of drive because it means the same thing that the participle *drived would mean, but we must use driven instead of *drived. Productive, as I will use it, applies to a rule or process itself. For example, the morphological process that adds -s to a verb stem to give a third-person singular form (e.g., drives) is productive, whereas the one that changes the end of a verb to -ought to give a past tense (e.g., thought, sought) is not productive. Finally, let’s call a process semi-productive if it applies to expressions in such a way that the output is relatively predictable, but is fairly idiosyncratic in terms of which expressions it can apply to. For example, the relation between verb patterns such as swim-swam-swum, begin-beganbegun, and so on, is semi-productive in this sense. (Obviously, where the line is between “semi-productive” and “unproductive” is somewhat vague.) In general, then, productive processes yield regular forms, while unproductive processes yield irregular forms. Semiproductive processes yield forms that may in a sense be regular with respect to that process, but are irregular forms strictly speaking. The examples I have given in these definitions all have to do with morphophonology, but these terms can apply equally to other aspects of an expression such as its meaning. Moreover, there are cases where a form may be regular in one sense but not in another. For example, the irregular past participles driven and gone are semantically regular in that they mean exactly the same things that *drived and *goed mean (or would mean if they were well formed). In particular, even if we have to list something like gone separately in the lexicon, its meaning can be represented as something like Pastpart(go’), directly accessing the lexical meaning for go to give the meaning of gone. So the words driven and went are morphologically irregular but semantically regular.1 On the other hand, consider a word like building, on the reading meaning ‘structure.’ This is morphologically identical to the gerundive form of build, but its meaning is related to the meaning of build in an idiosyncratic way, just as the result reading of criticism was related idiosyncratically to the meaning of criticize. (In fact it makes the most sense to

1

This may seem a bit circular, since we consider driven and gone to be “forms of” drive and go mainly because they mean the right thing. However, synonymy is not a sufficient condition for one expression to be considered a form of another. In particular, as I mentioned above, a crucial property of driven and gone is that we use them instead of *drived and *goed, that is, the regular forms are blocked by the existence of the irregular forms.

58

Tamina Stephenson

Honors Thesis Chapter 4

think of this building as a derived result nominal that happens to end in -ing.) Thus one might say that in a sense building is morphologically regular but semantically irregular. The three kinds of nominalization: Now let’s turn to the question of how productive the three kinds of nominalization are. Switching around my customary order, let’s consider gerundive nominals first. Whether or not the analysis I have given for gerundive nominals is ultimately the correct one, it is clear that gerundive nominalization is a completely productive process, both morphologically and semantically. This was one of the main points of Chomsky (1970). Essentially any verb phrase at all can undergo this kind of nominalization, and the resulting morpho-phonological form of the output is completely predictable from that of the input: the ending -ing is simply added to the main verb. Moreover, we can observe that the meaning of the resulting expression is completely predictable from the meaning of the verb phrase (and the subject). That is to say, the process creating gerundive nominalization is both morphologically and semantically productive. Mixed nominalization is also a productive process. It is more restricted as to the types of verbal expressions it can apply to, but it does not idiosyncratically apply to some verbs and not others. Even when there exists an idiosyncratic -ing form of a transitive or intransitive verb, we still always find another, homophonous form that is a regular mixed nominal related to the verb in the normal way. (For example, building is a frozen form in the Capitol Building, but regular in John’s building of the tower.) In particular, if we analyze mixed nominalization as a rule applying directly at the level of a transitive or intransitive verb from the lexicon, i.e., single words, then this rule applies completely productively over that domain. (As we saw in Chapter 4, it can even apply in some cases to multi-word “verbs” consisting of an adverb plus a verb, although this is somewhat restricted.) As with gerundive nominals, the morpho-phonological change is always the same: it adds the -ing ending to the verb and manipulates its argument structure such that the subject can be a possessive NP and the object an of-phrase. Moreover, the meaning of the resulting expression is always predictable from the meaning of the input verb (and its arguments). Thus mixed nominalization, too, is both morphologically and semantically productive.

59

Tamina Stephenson

Honors Thesis Chapter 4

Finally, let’s consider derived nominals. This is a largely unproductive process, morphologically speaking. There are a number of semi-productive morphological processes that give rise to derived nominals – e.g., the endings -ism, -tion, -ment, and others, sometimes with fairly predictable accompanying vowel changes in the verb stem. Certainly not all of these processes can apply to all verbs, and not all verbs seem to have corresponding derived nominals at all. This is true for both process nominals and result nominals. Now, the question is, are derived nominals semantically productive? That is, given any pair of a root verb and its corresponding derived nominal, can we predict the meaning of the derived nominal from that of the root verb? Of course, in the case of result nominals, clearly we can’t. For example, an assignment is a thing that is assigned, whereas a criticism is not a thing that is criticized, but rather a thing in which something else is criticized. But what about process nominals? Under the analysis developed in Chapter 2, derived process nominals are predicted to be completely semantically regular – in fact, they were predicted to have the same meaning of the productive mixed nominals. For example, the meaning of the derived process nominal John’s criticism of the book was predicted to be exactly the same as the meaning of the mixed nominal John’s criticizing of the book. These are repeated in (75) below. (75)

(a) [[John’s criticism of the book]] [from Figure 5.a] = [[the]] ( λy[y (of subtype <ε>) is an instance of criticize’(b)(j) ] ) = the unique contextually salient instance of criticize’(b)(j) (b) [[John’s criticizing of the book]] [from Figure 5.b] = [[the]] ( λy[y (of sub-type <ε>) is an instance of criticize’(b)(j)] ) = the unique contextually salient instance of criticize’(b)(j)

In particular, both meanings directly refer to the lexical meaning of criticize. That is, if the analysis given in Chapter 2 is correct, then derived and mixed nominalization are both equally productive semantically even though only one is fully productive morphologically. At first glance this seems to be right: John’s (intentional) criticism of the book and John’s (intentional) criticizing of the book do mean about the same thing and seem to relate to the verb criticize in roughly the same way. Therefore it would seem that derived nominalization is morphologically unproductive but semantically productive.

60

Tamina Stephenson

Honors Thesis Chapter 4

This view of derived nominals agrees with that given in Harley & Noyer (1997), who argue that what we are calling derived process nominals and mixed nominals are actually a single morphological process. Derived nominals, they claim, are simply irregular forms; the default or “elsewhere” form is the verb plus the -ing ending. This predicts two things: first, that a derived nominal such as destruction should be just like an irregular verb form such as gone: it should mean the same thing that the regular form destroying would mean, just with an idiosyncratic form. And second, it predicts that there should be a blocking effect that will prevent a mixed nominal (-ing form) from being used when there is a derived (idiosyncratic) form available for the same verb, just as the existence of the form gone prevents us from using *goed as a past participle of go. I will argue that neither of these predictions are correct, however, and that consequently the analysis of derived nominals developed in Chapter 2 must be revised slightly. First I will consider the semantic prediction. It turns out that there are examples of derived nominals that can be disambiguated to a process reading and mean roughly the same as the corresponding mixed nominal in most contexts, and yet which do not seem to have the same direct access to the meaning of the corresponding verb. This comes out when we consider verbs which have a special kind of extended or metaphorical reading. For example, there is a somewhat conventionalized metaphorical use of destroy that means ‘to criticize [e.g., an academic paper] particularly harshly.’ Similarly, there is a conventionalized use of inhale to mean ‘eat voraciously or very quickly.’ In these cases, the special reading is available in the mixed nominal (e.g., destroying of, inhaling of) but not, in general, in the derived nominal (e.g., destruction of, inhalation of). This is shown for destroy in (76). (76)

(a) Sam thoroughly destroyed my paper in his review. [destroyed = criticized it harshly) (b) Sam’s thorough destroying of my paper made me think that he has no respect for any of my work. (c) ?? Sam’s thorough destruction of my paper made me think that he has no respect for any of my work.

The special extended reading of destroy used in (76.a) is available for the mixed nominal in (b), but not for the derived nominal in (c). The derived nominal in (c) only allows the reading where Sam literally destroyed the paper, by tearing it up, burning it, deleting the

61

Tamina Stephenson

Honors Thesis Chapter 4

file, or some other means. This pattern also shows up on another special reading of destroy, meaning something like ‘to have a decisive victory over.’ This was attested in a college newspaper headline, “UCSB Men’s Basketball ends losing streak with destroying of Fullerton,”2 which I recast in example (77) below: (77)

(a) UCSB destroyed Fullerton, beating them by 25 points and ending a losing streak. [destroyed = had a decisive victory over] (b) UCSB’s destroying of Fullerton ended a long losing streak. (c) ?? UCSB’s destruction of Fullerton ended a long losing streak.

Again, the special meaning of destroy used in (77.a) is available for the mixed nominal in (b) but not for the derived nominal in (c); sentence (c) only allows the reading where the UCSB team somehow literally destroyed the Fullerton team. Another example is the verb inhale, which can be used to mean ‘eat quickly,’ as in (78) below. (78)

(a) The kids inhaled their food, they were so hungry. [inhale = eat quickly] (b) The kids were really hungry after such a long day. There was a lot of inhaling of food going on at dinner that night. (c) ?? The kids were really hungry after such a long day. There was a lot of inhalation of food going on at dinner that night.

Again, (78.a) and (78.b) would be appropriate in a case where the kids ate their food very quickly but did not choke on it; whereas (78.c) describes the alarming situation where the kids were actually breathing food particles into their lungs. Crucially, the derived nominals destruction and inhalation do have robust process readings which are roughly equivalent to their corresponding mixed nominals and seem to correspond to the meaning of destroy and inhale in a predictable way. We can tell because we can disambiguate the derived forms into process readings, and after we have done so X’s destruction of Y is roughly the same as X’s destroying of Y, and similarly for X’s inhalation of Y and X’s inhaling of Y. Yet when the verb is used in an extended or novel way, the two kinds of nominals are no longer equivalent.

2

UCSB Daily Nexus, January 29, 2001. Online at http://www.ucsbdailynexus.com/sports/2001/269.html.

62

Tamina Stephenson

Honors Thesis Chapter 4

I claim that this pattern is robust and will show up with verbs in general. Unfortunately, at this time I cannot give any more examples to support this claim other than the two readings of destroy and one of inhale above. It is particularly difficult to find examples of this phenomenon, even when we can think of these special “extended” meanings of verbs, for two main reasons. First, many verbs do not have derived nominal forms at all, or their derived forms have only result readings. In this case, the meaning of the derived form would not be expected to directly access the meaning of the verb anyway. Second, there are some cases where the derived form “accidentally” does allow the special extended reading. For example, I believe there is a special reading of criticize that means ‘to review, to do literary criticism on,’3 as in, The literature scholar criticized the book (where the scholar didn’t necessarily say anything bad about it). To the extent that we can use this as a verb in this way, we can also get this meaning with the mixed nominal criticizing, as in the literature scholar’s criticizing of the book. However, we can also get this reading in the literature scholar’s criticism of the book – in fact, this is one of the more salient readings of criticism. This does not constitute a counterexample to my claim, of course, because there is no reason why a derived nominal might not happen to have a second, frozen meaning that corresponds to an extended meaning of the verb. The important point is that there are verbs, such as destroy and inhale, which have special extended meanings that are not available with the derived nominal. In theory, the few examples I gave above would suffice to prove this point. Crucially, though, my claim makes one more prediction: that there will never be a case where some special, extended meaning of a verb is available in a derived process nominal but not available in a mixed nominal. This would be a real counterexample to the claim. As far as I can tell, though, this prediction is correct, and I will have to leave it to future research to find additional examples of this phenomenon and verify that it has no true counterexamples. We already knew that derived nominalization was morphologically unproductive. Now we have evidence that it is semantically unproductive as well. That is, the meaning of destruction, for example, shouldn’t be represented in terms of instances of destroy’, with direct access to the lexical meaning of destroy. Instead, it should be instances of

3

Not being a literary critic, I don’t have particularly good intuitions about this reading of criticize; therefore I apologize if the facts I cite are not quite correct.

63

Tamina Stephenson

Honors Thesis Chapter 4

something, where that “something” is a predicate that is very similar to the lexical meaning of the verb destroy but may be idiosyncratically different from it. For example, it may be less flexible. Similarly, under the analysis given in Chapter 2, the meaning of inhalation would have been represented using direct access to the lexical meaning of inhale; but it does not seem to have this direct access. Instead, inhalation is an instance of something, where that “something” is a predicate similar to the meaning of the verb inhale. Revised meanings of these derived nominals are given in (79)-(80) below, along with that of their corresponding mixed nominals for comparison. (79)

(a) [[John’s criticism of the book]] [cf. (75.a)] = [[the]] ( λy[y (of subtype <ε>) is an instance of P-criticize(b)(j) ] ) = the unique contextually salient instance of P-criticize(b)(j) where P-criticize is a 2-place predicate that is similar to criticize’ (b) [[John’s criticizing of the book]] [from Figure 5.b] = [[the]] ( λy[y (of sub-type <ε>) is an instance of criticize’(b)(j)] ) = the unique contextually salient instance of criticize’(b)(j)

(80)

[Let [[the kids]] = k, [[the food]] = f] (a) [[the kids’ inhalation of food]] = [[the]] ( λy[y (of subtype <ε>) is an instance of P-inhale(f)(k) ] ) = the unique contextually salient instance of P-inhale(f)(k) where P-inhale is a 2-place predicate that is similar to inhale’ (b) [[the kids’ inhaling of food]] [from Figure 5.b] = [[the]] ( λy[y (of subtype <ε>) is an instance of inhale’(f)(k) ] ) = the unique contextually salient instance of inhale’(f)(k) This finding goes against Harley & Noyer’s (1997) claim that derived process

nominals and mixed nominals are in fact a single process. On the contrary, derived nominals are frozen and idiosyncratic, even in seemingly regular process readings. Their meanings are almost, but not quite derivable from the meanings of the verbs. In some cases, of course, the meaning of a derived nominal may happen to be exactly what is expected, that is, exactly the same as the mixed nominal, but in other cases it may differ from it subtly or strikingly. This is very different from the situation with irregular verbs. For example, consider went, which is the past tense of go. A suppletive form, this is completely idiosyncratic as far as morphology, but it always means exactly the same thing as the present tense go except for the temporal component. So, for example, any 64

Tamina Stephenson

Honors Thesis Chapter 4

expression with go in the present tense, e.g., go crazy, will automatically have a past tense form with went, e.g., went crazy. Even when go to the bathroom (or sometimes just go) is used on the particularly idiosyncratic reading meaning ‘to excrete waste’, the past tense is still went. For example, in my dialect at least, one can say things like, “My puppy went to the bathroom on the rug,” or “The cat meowed and meowed at them to let him out, but they didn’t pay attention and he finally just went on the rug.” This is not what happens with derived nominals such as destruction and inhalation. They usually have a meaning corresponding to the verbs destroy and inhale, but not always, particularly when these are used on certain “special” readings. At this point, I should address one potential point of confusion. I am arguing in this chapter that derived nominals are unproductive in an essential way, whereas mixed nominals are productive. However, all the facts around the role of the possessive, for example, the fact that the possessive NP has to be the “subject” or “agent” of the embedded predicate, still hold. The productivity in question relates to exactly what 1- or 2-place predicate is embedded in the meaning of the nominal, and whether or not it refers directly to a verb in the lexicon; but the basic structure of the meaning of derived and mixed nominals is still the same. Both derived and mixed nominals still denote instances of propositions, and in particular they denote relations between participants and events that are instances of propositions. In other words, despite the difference in productivity between derived and mixed nominals, they are still both relational nouns and combine with possessive NPs and of-phrases the same way relational nouns like brother do. Also, one might object that derived process nominals ought to have some direct access to the meaning of the corresponding verb, since even though the embedded predicate doesn’t always mean exactly the same thing as the verb, it does always have the same argument structure. For example, what I’m calling P-criticize has the same argument structure as criticize’, P-inhale has the same argument structure as inhale’, and so on. But actually, this is not quite the case either. There do exist derived nominals whose argument structure differs from that of the corresponding verb, namely, the so-called “passive nominals” (e.g., the city’s destruction or Kennedy’s assassination) which were discussed in Chomsky (1970) and in Chapter 2. These have embedded predicates which are clearly “related” to destroy, assassinate, etc., but have a change in argument structure.

65

Tamina Stephenson

Honors Thesis Chapter 4

Now, let me return to my argument about the productivity of derived and mixed nominals, in particular my argument against the view of Harley & Noyer. As I mentioned before, Harley & Noyer’s account also made a second empirical prediction: that there should be a blocking effect that prevents speakers from using a mixed nominal for a verb when there is a derived process nominal available for that verb. Under their view that derived and mixed nominalization are a single process, we would expect to see such a blocking effect, just as we do with irregular verb forms; under the account I have argued for above, however, derived and mixed nominalization are two separate processes (in fact, one is not a “process” at all), and so we would not expect to see such a blocking effect. Using data from some very informal corpus searches, I will argue that the evidence goes against a blocking effect, and hence on the side of an account of derived and mixed nominalization being two separate processes. Before I go on with the evidence, I should provide a disclaimer and an explanation. The disclaimer is that, while I will be using attested examples to help prove my point, I do not intend this to be a rigorous corpus study in either scope or methodology. My “corpus” was created simply by entering phrases such as “destroying of” or “destruction of” or “’s destruction of” into a regular Internet search engine. This means that I had to choose which particular derived and mixed nominal forms to look at, thus decreasing my total pool of data and increasing its “hand-picked” nature. This process also generates a fair amount of “garbage,” of course, in the form of both irrelevant and ill-formed examples. Therefore, I do not attempt to give any statistical analyses of the data, nor do I assume that my body of data is an accurate sample of written corpora. Rather what I will be doing is pulling out specific examples to illustrate my point, using only those that I consider well-formed and reserving the right to edit or adapt them slightly to aid exposition (provided that they are still well formed). In particular, I will try to show that there are many contexts where either a derived (process) nominal or a mixed nominal may be used with roughly the same meaning. First, consider the following attested examples of the mixed nominal destroying:

66

Tamina Stephenson

(81)

Honors Thesis Chapter 4

Attested4 examples of mixed nominals with destroying: (underlines added) [from computer equipment manual] (a) Negligence, but also untrained handling, may lead to the destroying of equipment or data which can severely disrupt further operation of the IT system. [ http://www.bsi.bund.de/gshb/english/t/t3002.htm ] [from the user guidelines for a university computer network] (b) The reading, copying, altering or destroying of files or software stored in a user directory to which you have not been given authorized access, even if the files or software are not explicitly protected, is prohibited. [ http://www.uwrf.edu/its/policies/general/phtml ] [from a news story] (c) The destroying of ammunition from the Margetici dump, near Sokolac, was continued today, a few kms away from the site. [ http://www.hri.org/news/agencies/srna/96-08-19.srna.html ] [from a guide to insect control by an environmental agency] (d) Thorough annual pruning and destroying of galls helps reduce populations. [ http://www.mb.ec.gc.ca/pollution/pesticides/ec00s08.en.html ] [from a news/analysis story] (e) Although it is not a direct consequence of Taliban’s religious zeal in hacking and destroying of Buddha icons and statues in their country, it has more or less coincided with their latest shocking “achievement”. [ http://www.defencejournal.com/globe/2001/apr/sino-russian.htm ]

The examples in (81) come from various domains, two having to do with electronic data, one with disarmament, one with pest control, and one with religious statues. It turns out that a derived nominal is also felicitous in contexts similar to all of these, as indicated by the additional attested examples in (82). (82)

Attested examples of derived nominals with destruction: (underlines added) [from computer equipment manual] (a) A regulated procedure for the deletion or destruction of data media will prevent misuse of stored data. [ http://www.bsi.bund.de/gshb/english/s/s2167.htm ]

4

All attested examples in this project were found using the advanced search form at http://www.google.com/advanced_search?hl=en, during searches performed between 10/2001 and 4/2002.

67

Tamina Stephenson

Honors Thesis Chapter 4

[from the user guidelines for a university computer network] (b) Appropriate security measures shall be taken against unauthorised access to, or alteration, disclosure or destruction of, personal data and against accidental loss or destruction of personal data. [ http://www.lboro.ac.uk/admin/central_admin/policy/data_protection.html ] [from an international agreement on chemical weapons] (c) In accordance with provisions of this Agreement, the Parties undertake … to co-operate regarding methods and technologies for the safe and efficient destruction of chemical weapons…. [ http://projects.sipri.se/cbw/docs/cw-bda.html ] [name of an environmental agency] (d) Coalition to Prevent the Destruction of Canada Geese. [ http://www.icu.com/geese/doc3.html ] [from a news/analysis story] (e) In a growing intellectual challenge to Islamic practices abroad, leading Southern California Muslim scholars Thursday denounced the ruling Taliban's destruction of Buddhist statues in Afghanistan as contrary to their faith's laws and traditions. [http://www.institute-for-afghan-studies.org/History/NATIONAL%20TREASURES/statues_destroyed/islamic_reaction.htm] The use of destruction in each of (a)-(e) in (82) is very similar to the use of destroying in the corresponding example in (81), both in meaning and context of use. Now consider the following examples of nominals using assigning and assignment, in (83) and (84). (83)

Attested examples of mixed nominals with assigning: (underlines added) [from guidelines to a university’s student appeal procedures] (a) The Faculty Handbook states "The assigning of grades is the responsibility of the individual instructor in every case. [ http://www.vt.edu/vt99/academics/gcat/gcaAppeals.html ] [from a university library web page] (b) At the moment, however, some of the significant value added to a resource by librarians, creators, and users, is in the assigning of evaluative metadata. [ http://www.educationau.edu.au/edna_cataloguing/assigning.html ]

Again, we find the derived nominal assignment can also show up in these contexts:

68

Tamina Stephenson

(84)

Honors Thesis Chapter 4

Attested examples of derived nominals with assignment: (underlines added) [from an organization for academic integrity] (a) … a professor’s assignment of a grade is final and within a professor’s right to academic free expression. [ http://www.academia.org/news/thirdcircuitdeals.html ] [from a university library web page] (b) Each processing center has its own policy on the assignment of call numbers to non-book materials. [ http://www.uky.edu/Libraries/bp1.html ]

Finally, a similar pattern of examples of nominalized expressions with acquitting and acquittal is shown in (85) and (86). Once again, we find examples of the mixed nominal acquitting in contexts where the derived form acquittal is also felicitous. (85)

Attested examples of mixed nominals with acquitting: (underlines added) [from a book review] (a) … Bruun and Crosby have also created a coherent, interesting, and unique way to follow the history of America from the pre-Columbus New World to the acquitting of William Jefferson Clinton. [ http://www.independentrepublican.com/bookreviews.html - past reviews ] [from a political or news/analysis article] (b) The cause of this outrage was the acquitting of Hank Borzworski from his charge of quadruple infanticide. [ http://members.tripod.com/~dystopian_society/page51.html ]

(86)

Attested examples of derived nominals with acquittal: (underlines added) [from a political or news analysis article] (a) Within a day of the Senate’s acquittal of Bill Clinton, ABC and CBS identified the Republicans as the party hurt the most. [ http://www.mediaresearch.org/news/mediawatch/1999/mw19990222p1.html ] [from a political or news/analysis article] (b) He said this in response to an all-white jury’s acquittal of a white assailant who attacked an innocent black pedestrian. [ http://www.geocities.com/WestHollywood/Castro/3242/writings/secular.html ]

The examples in (81)-(86) indicate that there is no strong blocking effect against mixed nominals in favor of available derived nominals. In all fairness, I should mention that when I did these web searches, there would generally be considerably fewer results for a

69

Tamina Stephenson

Honors Thesis Chapter 4

mixed nominal than for the corresponding derived nominal. Again, this is simply a rough observation and has not been substantiated by any statistical tests, but even if it had been, it would not be enough to prove a blocking effect, in my mind. It could show that there is some preference for derived forms for some reason, as there very well might be, but it would not show that the mixed nominals are blocked. In fact, any search for a mixed nominal that got any results at all (that is, any well-formed, relevant results) proves just the opposite – that mixed nominals are not blocked. This is entirely different from a true blocking effect such as the one that prohibits *goed as the past tense of go; in this case, even if a corpus search found attested examples of goed as the past tense of go, speakers of English would reject them as ill-formed. In summary, then, I have argued in this chapter that an account of derived process nominals as simply irregular forms of mixed nominals fails in two respects. First, it fails to account for the fact that certain “special” readings of verbs are available for mixed nominals but not for derived nominals, even in true “process” readings. Second, it incorrectly predicts that there should be a blocking effect preventing mixed nominals from occurring when there is a derived form available. My proposal is to treat derived and mixed nominals as two separate, though similar, phenomena. Derived nominals are always frozen, idiosyncratic forms, even on seemingly productive process readings; there is a predicate embedded in the meaning of a derived nominal, but this predicate is not necessarily exactly the same as the meaning of the corresponding verb. Mixed nominals, on the other hand, are completely productive, with a meaning that is predictable directly from the verb. In other words, derived and mixed nominals have meanings of the same general form, (functions to instances of propositions with some predicate), but they do not represent the same process. Put another way, derived nominals are expressions just like mixed nominals except that the “verb” that they correspond to is not any actual verb in English, but just some predicate defined within the meaning of the derived nominal. This analysis accounts for the fact that some special readings of a verb will not be available for a derived nominal, since the predicate embedded in the meaning of the derived nominal can simply be slightly different from the meaning of the verb. This also allows for the occurrence of corresponding derived and mixed nominals in similar contexts: since they do not represent a single morphological process, we do not expect a

70

Tamina Stephenson

Honors Thesis Chapter 4

blocking effect. These revisions do not need to change the basic properties of derived and mixed nominals; they can still be analyzed as relational nouns denoting a relation between events and event participants. Both derived and mixed nominals make reference to some verb-type predicate in their meanings; the difference is that mixed nominals refer directly to a verb in the lexicon, whereas derived nominals have a predicate defined right in the meaning of the nominal and do not refer to actual verbs. Finally, these factors have no bearing on gerundive nominals, so we can keep the account of gerundive nominals that was developed in Chapters 1-2.

71

Chapter 5: Mass nouns, pluralities, and the meaning of nominals The distribution of nominals with mass and count determiners In Chapter 2, derived and mixed nominals were analyzed as relational nouns. Among other things, this analysis accounted for the fact that these expressions can combine with adjectives and the definite article the, just like ordinary common nouns and relational nouns. However, recall from Chapter 1 that derived process nominals and mixed nominals were odd with the indefinite article a, even though they were fine with the. Gerundive nominals are ill-formed with both a and the. In contrast, a normal relational noun such as sister [of John] is fine with either one. (87)

(a) the criticism of the book (by John) (b) ? a criticism of the book (by John) [on process reading]

[repeated from (20.b)] [repeated from (20.c)]

(c) the criticizing of the book (by John) (d) ? a criticizing of the book (by John)

[repeated from (22.b)] [repeated from (22.c)]

(e) * the criticizing the book (by John) (f) * a criticizing the book (by John)

[repeated from (21.b)] [repeated from (21.c)]

cf.

(g) the sister of John / a sister of John

It also seems to be impossible to pluralize derived, mixed, and gerundive nominals. (88)

(a) * John’s three (intentional) criticisms of the book [on process reading] (b) * John’s three criticizings of the book (c) * John’s three criticizings the book / * John’s three criticizing the books cf.

(d) three sisters of John (e) John’s three sisters

In this chapter, I will explore one possible way to treat nominalized expressions that will help to account for these facts. The sketch I will give is meant as a tentative treatment, and it will require significant further work to determine exactly how this can be fleshed out and formalized. There may be other ways to account for this, and there are most likely some facts which I have overlooked or oversimplified; however, this is hopefully a promising start. At first glance, the patterns illustrated in (87)-(88) seem to indicate a way in which nominalized expressions are qualitatively different from normal nouns. For

73

Tamina Stephenson

Honors Thesis Chapter 5

gerundive nominals, this may very well be the case; but for derived and mixed nominals, there actually are nouns in English with exactly the same distribution – a substantial class of them. These are, of course, those nouns such as furniture, water, rice, and many others, which are traditionally labeled “mass nouns,” as opposed to “count nouns” such as chair, lake, and noodle. Mass nouns pattern like the derived and mixed nominals in (87)-(88) above, whereas count nouns pattern with the noun sister of John, as in (89). (89)

(a) the furniture (b) * a furniture cf. (c) the chair / a chair (d) * three furnitures cf. (e) three chairs

Thus a plausible initial conjecture is that derived and mixed nominals are simply mass nouns. To test this, let’s look at a more complete set of data on the distribution of mass and count nouns with these three kinds of nominals. This is compiled below in the form of tables. First, for reference, Table 3 shows the distribution of normal mass and count nouns with a variety of determiners and quantifiers. Both kinds of nouns can appear with the definite article the, the demonstrative that, and the possessive, but with the other constructions shown mass and count nouns are more or less in complementary distribution. Count nouns occur in the singular with a, one, each, and every, and in the plural with some, those, three, many, and all. Mass nouns occur in the singular with some, much, and all, and never occur in the plural. Also, count nouns can occur as a bare argument only in the plural, whereas mass nouns can do so only in the singular, and only singular count nouns can appear with the in a generic or natural kind reading. Table 3

Distribution of typical mass and count nouns Count noun – chair Mass noun – furniture the + sing. the chair the furniture that + sing. that chair that furniture possessive – NP-’s + sing. John’s chair John’s furniture indefinite – a + sing. one + sing. each + sing. every + sing.

a chair one chair each chair every chair

* a furniture * one furniture * each furniture * every furniture

74

Tamina Stephenson

some + sing.

Honors Thesis Chapter 5

some furniture

(so / how) much + sing. all + sing.

* some chair [relevant reading1] * (so / how) much chair * all chair

some + plur those + plur. three + plur. many + plur. all + plur

some chairs those chairs three chairs many chairs all chairs

* some furnitures * those furnitures * three furnitures * many furnitures * all furnitures

bare argument – sing. bare argument – plur.

* chair chairs

furniture * furnitures

Generic / Natural kind the

the dog is a mammal * the dogs are mammals

* The mold is a fungus.

(so / how) much furniture all furniture

Tables 4 and 5 include corresponding data for derived and mixed nominals, arranged like that in Table 3. I should note at the outset that there are a number of cases in Tables 4-5 for which it was difficult to get clear, consistent acceptability judgments from myself or other informants. Although some of these might be slightly more acceptable than others, I have marked them all with *? to reflect the fact that they are generally better than those marked with “*” and worse than those marked as well formed. Table 4 the + sing. that + sing.

NP-’s + sing. indefinite – a + sing. one + sing.

Distribution of a derived nominal The destruction of $1000 worth of property (by vandals) was a criminal act. [to a neighbor]: I’m not mad about the time when Johnny threw a baseball through my window – that destruction of my property was an accident. What I’m mad about is the time when Susie purposely knocked over my fence. The vandal’s destruction of $1000 worth of property was a criminal act. ? Three crimes were reported last Friday. A theft occurred at 4 pm, an assault at 5 pm, and a destruction of property (by vandals) at 6 pm. ? Destruction of property by vandals was reported three times

1

I believe that some is ambiguous between what I would call the “logic class” reading (which gives expressions such as some man) and a more everyday reading which must refer to a quantity of something. I as that you ignore the “logic class” reading for the purposes of the data at hand.

75

Tamina Stephenson

on Thursday. One destruction of property occurred in a dorm, one in a classroom building, and one off-campus. ? Last year there were several well-publicized cases of destruction of property by vandals. Each destruction of property led to a public outcry about the condition of the city. ? Last year there were several well-publicized cases of destruction of property by vandals. Every destruction of property led to a public outcry about the condition of the city. Some destruction of property occurred on Friday.

each + sing. every + sing. some + sing. (so / how) much + sing. all + sing. some + plur those + plur.

three + plur. many + plur. all + plur bare argument – sing. bare argument – plur. Generic / kind the

Table 5 the + sing. that + sing.

NP-’s + sing. indefinite – a + sing.

Honors Thesis Chapter 5

How much destruction of property occurred during the robbery? All destruction of property will be prosecuted. * Several crimes were reported last Friday. There were some robberies, some assaults, and some destructions of property. * [to a neighbor]: I’m not mad about the time when Johnny threw a baseball through my window or the time when Billy ran into my mailbox – those destructions of my property were an accident. What I’m mad about is the time when Susie purposely knocked over my fence. * Several crimes were reported last Friday. There were two robberies, two assaults, and three destructions of property. * Several crimes were reported last Friday. There were a few robberies, some assaults, and many destructions of property. * All destructions of property will be prosecuted. Destruction of property is illegal. * Destructions of property are illegal. The destruction of property is illegal. * The destructions of property are illegal.

Distribution of a mixed nominal The destroying of $1000 worth of property (by vandals) was a criminal act. ? [to a neighbor]: I’m not mad about the time when Johnny threw a baseball through my window – that destroying of my property was an accident. What I’m mad about is the time when Susie purposely knocked over my fence. The vandal’s destroying of $1000 worth of property was a criminal act. ? Three crimes were reported last Friday. A theft occurred at 4 pm, an assault at 5 pm, and a destroying of property (by vandals) at 6 pm. 76

Tamina Stephenson

one + sing. each + sing. every + sing. some + sing. (so / how) much + sing. all + sing. some + plur those + plur.

three + plur. many + plur. all + plur bare argument – sing. bare argument – plur. Generic / kind the

Honors Thesis Chapter 5

? Destroying of property by vandals was reported three times on Thursday. One destroying of property occurred in a dorm, one in a classroom building, and one off-campus. ? Last year there were several well-publicized cases of destroying of property by vandals. Each destroying of property led to a public outcry about the condition of the city. ? Last year there were several well-publicized cases of destroying of property by vandals. Every destroying of property led to a public outcry about the condition of the city. Some destroying of property occurred on Friday. How much destroying of property occurred during the robbery? All destroying of property will be prosecuted. * Several crimes were reported last Friday. There were some robberies, some assaults, and some destroyings of property. * [to a neighbor]: I’m not mad about the time when Johnny threw a baseball through my window or the time when Billy ran into my mailbox – those destroyings of my property were an accident. What I’m mad about is the time when Susie purposely knocked over my fence. * Several crimes were reported last Friday. There were two robberies, two assaults, and three destroyings of property. * Several crimes were reported last Friday. There were a few robberies, some assaults, and many destroyings of property. All destroyings of property will be prosecuted. Destroying of property is illegal. * Destroyings of property are illegal. The destroying of property is illegal. * The destroyings of property are illegal.

It turns out that derived and mixed nominals have a distribution that is very similar to that of run-of-the-mill mass nouns, but not quite identical to it. In particular, like mass nouns, they can appear in the singular with determiners such as some, much, and all, and as singular bare arguments, and they can never appear in the plural. But unlike mass nouns, in certain contexts they are not entirely ill-formed with singular count determiners such as a, one, each, and every. I will return to a possible explanation of this data below. In the meantime, Table 6 shows the very different situation which holds for gerundive nominals.

77

Tamina Stephenson

Table 6 the + sing. that + sing.

NP-’s + sing.

Honors Thesis Chapter 5

Distribution of a gerundive nominal * The destroying $1000 worth of property (by vandals) was a criminal act. * [to a neighbor]: I’m not mad about the time when Johnny threw a baseball through my window – that destroying my property was an accident. What I’m mad about is the time when Susie purposely knocked over my fence. The vandal’s destroying $1000 worth of property was a criminal act.

indefinite – a + sing.

some + sing.

* Three crimes were reported last Friday. A theft occurred at 4 pm, an assault at 5 pm, and a destroying property (by vandals) at 6 pm. * A vandal’s destroying property was reported three times on Thursday. One destroying property occurred in a dorm, one in a classroom building, and one off-campus. * Last year there were several well-publicized cases of a vandal destroying property. Each destroying property led to a public outcry about the condition of the city. * Last year there were several well-publicized cases of a vandal destroying property. Every destroying property led to a public outcry about the condition of the city. * Some destroying property occurred on Friday.

(so / how) much + sing. all + sing.

* How much destroying property occurred during the robbery? * All destroying property will be prosecuted.

some + plur

* Several crimes were reported last Friday. There were some robberies, some assaults, and some destroyings property / ‘destroying property’s * [to a neighbor]: I’m not mad about the time when Johnny threw a baseball through my window or the time when Billy ran into my mailbox – those destroyings my property / ‘destroying my property’s were an accident. I’m mad about the time when Susie purposely knocked over my fence. * Several crimes were reported last Friday. There were two robberies, two assaults, and three destroyings property / …. * Several crimes were reported last Friday. There were a few robberies, some assaults, and many destroyings property / …. * All destroyings property / All ‘destroying property’s will be prosecuted.

one + sing. each + sing. every + sing.

those + plur.

three + plur. many + plur. all + plur

bare argument – sing. bare argument – plur. Generic / kind the

Destroying property is illegal. * Destroyings property / ‘destroying property’s are illegal. * The destroying property is illegal. * The destroyings property / ‘destroying property’s are illegal. 78

Tamina Stephenson

Honors Thesis Chapter 5

Whereas derived and mixed nominals behave something like a certain class of nouns, gerundive nominals have a much more limited distribution, in fact hardly occurring with any determiners or quantifiers at all: they can appear with a possessive NP, and as a singular bare argument, and that’s it. This is consistent with the observation made in Chapter 1 (as well as by Chomsky) that gerundive nominals don’t seem to have the internal structure of noun phrases. To move towards an account of the distribution of the three kinds of nominals, I will first have to give some background on an account of the mass/count distinction. Atoms, pluralities, and the denotation of mass and count nouns Let me make a few assumptions about mass and count nouns. First and foremost, I assume that the distinction in question is primarily a semantic one. It would be easy enough, of course, to posit some ad hoc syntactic feature, say [± COUNT], that could be attached to nouns; determiners could be specified as to which kinds of nouns they can combine with, and thus the distribution of count and mass nouns could be accounted for. However, a semantic account of the difference between the two noun classes could potentially have much more explanatory value, so that is what I will pursue. In particular, I will make a number of assumptions about the meaning of mass and count nouns. These are based closely on Chierchia (1998), although the account I follow will differ slightly from his. For a more complete discussion of the semantics of plurals and mass nouns, see, e.g., Chierchia (1998), Lasersohn (1995), and references cited therein. The assumptions I make are the following: First, there exists in the domain of discourse a set of entities that are represented as being individuated, separate from one another. These are “atomic individuals,” or “atoms,” and can be illustrated as in (90). (90)

Set of atoms: {

e1,

e2,

e3,

e4,

e5,

e6,

e7,

…

}

Second, it is possible to form more complex individuals by combining these atoms together into larger individuals. You can think of this either as set-formation or as some kind of Boolean “join” operation, whichever you are more comfortable with. These are “plural individuals,” or “pluralities,” and can be illustrated as something like (91) below.

79

Tamina Stephenson

(91)

Honors Thesis Chapter 5

Set of pluralities: (a) { <+e1+e2+>, <+e2+e4+e6+>, <+e1+e2+e3+e4+e5+f+>, <+e3+e6+>, … } or (b) { {e1, e2}, {e2, e4, e6}, {e1, e2, e3, e4, e5, e6}, {e3, e6}, … }

I personally prefer some kind of notation like that in (91.a), which distinguishes plural individuals from sets, since plural and atomic individuals both ought to have the same semantic type, namely type . On the other hand, pluralities do have some of the same properties as sets; for example, it makes sense to invoke the notion of cardinality to account for the role of plural determiners such as three. Also, these plural individuals should be thought of as combinations of atoms, strictly speaking, meaning that, like sets, they are not dependent on order and there is no sense in which one atom can occur more than once in a particular plural individual. Chierchia uses set notation like that in (91.b), however I’ll stick to that shown in (91.a). The denotation of a singular count noun such as chair is, then, a set of atoms, where each atom is a specific, fully individuated chair. For example, let’s suppose that in a given world, there are exactly four chairs, which we will denote by c1, c2, c3, and c4. Then the denotation of chair is as given in (92). (92)

[[chair]]S

=

{

c1,

c2,

c3,

c4

}

Now, there is a process of pluralization which takes a singular count noun like chair and returns the same noun with plural morphology. The semantics works such that the denotation of the plural noun is the set of all pluralities which can be made by combining two or more of the atoms in the denotation of chair. So the denotation of chairs is as shown in (93). If we think of plural individuals in terms of sets, then the denotation of a plural noun is that subset of the power set of the denotation of the singular noun in which all members are of cardinality greater than or equal to 2. (93)

[[chairs]]S

=

   

<+c1+c2+>, <+c1+c3+>, <+c1+c4+>, <+c2+c3+>, <+c2+c4+>, <+c3+c4+>, <+c1+c2+c3+>, <+c1+c2+c4+>, <+c1+c3+c4+>, <+c2+c3+c4+>, <+c1+c2+c3+c4+>

   

Notice that the denotation of a plural noun will thus always have the important property of forming a lattice; that is, it will have overlapping elements and elements of different

80

Tamina Stephenson

Honors Thesis Chapter 5

sizes, with every smaller element being contained within some larger element, and in addition it will have one maximal element of which every other element is a part. On the other hand, the denotation of a mass noun such as water or furniture does not contain atoms. Rather, it consists of a lattice much like that for a plural count noun, except that there is no “bottom” or basic unit to the lattice, or, put another way, the “atoms” are infinitely small. You can think of it as the power set of the (infinite) set containing all the (infinitely small) “atoms” of water, furniture, etc. Such a denotation can be partially illustrated as something like (94). (94)

[[water]]S

=

 … <+w1+…+wa+>, <+wa+…+wb+>,   <+wb+…+wN+>, <+w1+…+wb+>, <+wa+…+wN+>,   <+w1+…+wa+wb+…+wN+>, … <+w1+…+wN+> 

[where a, b, N are positive integers with a < b < c < N, and N can be arbitrarily large] The idea is that any of the individuals listed in the denotation of water can be infinitely subdivided into other individuals, and thus every individual in the denotation will be a plurality. This means, of course, that there is no sensible definition of cardinality for the pluralities denoted by a mass noun, though the cardinality is arguably always greater than 1. This intuitively makes sense for a noun like water, since (barring knowledge of chemistry) we think of water as being a continuous, fluid substance with no smallest unit. It does seem a bit strange to apply this to a noun like furniture, since there is arguably a smallest unit of furniture, namely a single piece (since, for example, a single table would count as furniture but one of its legs would probably not). The point, though, is that the meaning of a noun like furniture is represented in the lexicon as if it could be infinitely subdivided in this way, even though in reality there is a limit on this. This is really just the same situation we have with water, which we continue to think of as being continuous and infinitely subdividable even when we have learned that there is in fact a minimal unit of water, namely a molecule. The “molecules” of furniture are simply big and easy to spot. There are other mass nouns, such as pasta or sand, whose minimal units are visible but relatively small. Of course, we would expect to find that other languages with a mass-count distinction may make different choices about which nouns to represent as atoms and which as pluralities, where nouns corresponding to furniture would be more 81

Tamina Stephenson

Honors Thesis Chapter 5

likely to be represented as atoms than nouns corresponding to sand or especially water. Although I won’t give a cross-linguistic survey here, this does seem to be the case. (See Chierchia, 1998 for a discussion of mass and count nouns across languages.) Given this account of the difference between mass and count nouns, we can now give rough formulations of the meanings of the determiners and quantifiers from Tables 3-6 to explain their distribution. For the moment, let’s only consider typical mass and count nouns such as chair, furniture, dog, and mold, as are used in Table 3. I’ll return below to the place of nominalized expressions in this pattern. pluralization: Given the definition for pluralization illustrated in (93), it will only apply to sets of atoms. In other words, as Chierchia puts it, mass nouns cannot be pluralized because they are already plural. Basically, we can’t pluralize a mass noun for the same reason that we can’t re-pluralize a noun that is already plural. Of course, the natural question to ask is why we can’t pluralize a plural noun. Think of it this way: Let’s say we were to apply the pluralization process to a count noun such as chairs that was already plural. What we would get would be the set containing all combinations of the plural individuals contained in the set denoted by chairs with cardinality at least 2. Now, all of the individuals in the chairs-set already have a cardinality of at least 2, so all of these individuals will be in the new set. Also, we can take any combination of these individuals – but all combinations are already included in the original set, keeping in mind that one atom can’t occur more than once in a plural individual. Therefore at the end we would get out the same set we put in. The same thing would happen if we tried to apply pluralization to a mass noun such as water (ignoring for the moment the requirement on the cardinality of the resulting pluralities). Therefore all we need is some general principle to the effect that we can’t apply this sort of vacuous process, and this will explain why mass nouns never occur in the plural. the: In Chapter 2, I assumed that [[the]] applied to a set to give the unique contextually salient member of that set. To extend this to pluralities, we can re-define [[the]] as giving the maximal contextually salient member of the set (see, e.g., Chierchia, 1998: p. 346). This will correctly predict that the furniture in the second sentence of (95.a) below refers to all of the furniture in the store, not just some of it or one particularly salient piece of it, and similarly for the books in (95.b).

82

Tamina Stephenson

(95)

Honors Thesis Chapter 5

(a) Mary’s store sells furniture and bedding. The furniture is in the front of the store and the bedding is in the back. (b) Mary’s store sells books and stationery. The books are in the front of the store and the stationery is in the back.

Note that there is no sense in which one atom can be more “maximal” than another, so when [[the]] applies to a set of atoms, it will only be felicitous if the set is a singleton. On the other hand, when it applies to a set of pluralities (as with either a mass noun or a plural count noun), it will correctly give the maximal salient member of the set. This correctly predicts that the can occur with singular count nouns, plural count nouns, and singular mass nouns. that: The meaning of the indexical that can be pretty much the same as the except that the domain of contextual salience is very tightly constrained to include just those individuals that have been specifically indicated, e.g. by pointing at them. It also has an additional requirement that the individual not consist of two or more atoms. This predicts that mass nouns and singular count nouns will occur with that, but plural nouns will not. possessive NPs: The way the meaning of the possessive was defined in Chapter 2, it actually contained the operator [[the]]. Thus, as long as there is an appropriate relationship for the possessive to express, it makes sense that it can apply to singular count nouns, plural count nouns, and singular mass nouns. Note that the way we redefined [[the]] above makes the correct prediction about the meaning of the possessive with plurals and mass nouns, since John’s sisters or John’s property normally refers to all of his sisters or all of his property. indefinite a and numeral one: There are two plausible ways to characterize the meaning of the indefinite article. One is that it quantifies over the set denoted by its argument noun, requiring that there exist some atomic individual in that set that has a given property. (Thus the sentence A cat is running means that there exists an atomic individual in the cat-set that is running.) This makes sense when applied to a singular count noun, so singular count nouns can occur with a. However, since the sets denoted by plural and mass nouns never contain any atomic individuals, they cannot occur with a. The other plausible way to characterize the meaning of a is to say that it just picks out some atomic individual from the denotation of its argument noun, in which case the same

83

Tamina Stephenson

Honors Thesis Chapter 5

reasoning works. Either way, we can assume that the numeral one will do basically the same thing as the indefinite a. each and every: These can both be analyzed as quantifying over the atomic individuals in the denotation of the argument noun, requiring that all of them have a given property. (Thus the sentence Every cat chases its [own] tail means that for every atomic individual in the cat-set, that individual has the property of chasing its [own] tail.) Again, since the denotation of a singular count noun always contains atoms, but the denotation of a plural or mass noun never does, singular count nouns will occur with these quantifiers but plural and mass nouns will not. some: On the relevant reading, some existentially quantifies over the denotation of the argument noun, requiring that there exist some plural individual in the set with a given property. Or, alternatively, it picks out some plural individual. This is just like a in reverse, so some will occur with plural and mass nouns but not singular count nouns. much: This will do something like quantify over the individuals in the domain of the argument noun, requiring that there exist some (plural) individual of sufficient size with some given property, with the added requirement that the individual must be infinitely subdividable. This will predict that much only occurs with mass nouns. all: This works like every, except that it quantifies over the plural individuals in the denotation of the argument noun, requiring that every plural individual have a given property. This predicts that all will occur with plural and mass nouns, but not with singular count nouns. those: This works just like that, except that it requires that the individual picked out must consist of two or more atoms. Thus those only occurs with plural count nouns. three: This works like a / one, except that it looks for the existence of a plurality consisting of three or more atoms. Thus three only occurs with plural count nouns. many: This will quantify over the individuals in the domain of the argument noun, requiring that there exist some plural individual of sufficient cardinality with some given property, with the added requirement that the individual must not be infinitely subdividable. This predicts that many will only occur with plural count nouns. bare arguments: Let’s assume that there is a general process in English by which any common noun denoting more than one individual can be turned into a noun phrase

84

Tamina Stephenson

Honors Thesis Chapter 5

that denotes the maximal individual in the original denotation. This will correctly predict that plural and mass nouns can occur as bare arguments, but singular count nouns cannot. generic / natural kind the: Finally, let’s assume that there is a general process by which certain nouns whose denotation includes two or more atoms can be turned into another noun containing the single individual representing a “kind” corresponding to all the atoms in the denotation of the original noun.2 Once this occurs, the denotation of this “kind” will be a singleton set, so the determiner the can pick out that individual. This predicts that only singular count nouns will occur with the generic the. The denotation of derived and mixed nominals You may have noticed that I have defined some of the expressions above rather carefully. For example, the meaning of the indefinite a makes reference to some atomic individual in the denotation of the noun, but does not require that the noun denote only atoms. Similarly, all limits its quantification to plural individuals in the denotation of the argument noun, but does not require that all individuals in the denotation be plural. So far, just looking at typical mass and count nouns, these technicalities haven’t mattered since every noun consists either entirely of individuals or entirely of pluralities (and in those consisting of pluralities, either all the pluralities are made up of atoms or none of them are). They become important, however, in the account I am about to propose for the denotation of derived and mixed nominals. (I hope you will forgive my cheating a bit by sneaking them in above.) It is a fairly simple proposal. We have seen above that derived and mixed nominals are similar to mass nouns, so let’s suppose they have in their denotation a lattice just like that of a mass noun like water, which is infinitely subdividable (or, put differently, is built up from infinitely small “atoms”). But let’s suppose that the denotation of a derived or mixed nominal contains, in addition, a set of atoms, that is, individuated units. Intuitively, these atoms can be thought of as “events” to the extent that events can be individuated from one another; whereas the mass-type individuals in the lattice can be thought of as “processes” or “actions” which are more or less continuous through blocks of time and therefore have no smallest unit. In other words,

85

Tamina Stephenson

Honors Thesis Chapter 5

the denotation of a derived or mixed nominal is the union of a mass-noun denotation with a singular count-noun denotation.3 That is, there are individuals that are quantities, which form a lattice, and then there are also other individuals that are atoms. The denotation of a derived or mixed nominal can be illustrated as in (96) below. (96)

Denotation of a derived or mixed nominal = Singular count noun ∪ Mass noun

    

a1, a2, a3, a4, a5, a6, a7, … <+m1+…+ma+>, <+ma+…+mb+>, <+mb+…+mN+>, <+m1+…+mb+>, <+ma+…+mN+>, <+m1+…+ma+mb+…+mN+>, … <+m1+…+mN+>

    

[where a, b, N are positive integers with a < b < c < N, and N can be arbitrarily large] Crucially, under this view derived and mixed nominals are not ambiguous between a reading with a count-type denotation and a reading with a mass-like denotation. (If this were the case, then we would expect them to appear with all kinds of determiners, both mass and count, as well as be able to undergo pluralization, which is clearly not the case.) Rather, the denotation is a set that includes both types of entities as members. Also, the way I am formulating it, the atomic individuals in the denotation do not form a lattice but are only listed as singular individuals. Since this type of denotation includes, among other things, pluralities that are infinitely subdividable, it is correctly predicted that the mass determiners some, much, and all can occur with derived and mixed nominals, and that they will be able to appear as bare arguments. Also, if we formulate pluralization as a process that can only apply to a whole set, then a derived or mixed nominal cannot be pluralized because part of its denotation is in a full lattice structure, which is to say that in a sense it is “already plural.” And finally, to the extent that the denotation includes individuated atoms, they can occur with certain singular count determiners and with the generic or natural kind the.

2

See Chierchia (1998) for a more detailed discussion of kinds in language. For discussion of what sorts of nouns can be represented as kinds, Chierchia cites Carlson (1997) and Krifka et al. (1995). 3 I appreciate P. Jacobson’s particularly helpful suggestion that I treat this as the union of a mass-type denotation with a count-like denotation. Before I had tried to set it up so that a lattice was built up using both atoms and mass-like units as the basic level of the lattice.

86

Tamina Stephenson

Honors Thesis Chapter 5

Again, this account is tentative and speculative. For one thing, it depends on a number of seemingly arbitrary stipulations about the meaning of various determiners. In particular, it depends on an account where some determiners require only part of the denotation of the argument noun to be of a certain type, whereas other determiners (as well as the process of pluralization) have their requirements over the entire denotation of the noun. It would require further research to see whether there is any independent evidence for this. However, it seems plausible at this point that we could give a semantic account of the distribution of derived and process nominals with various determiners if we could nail down exactly what kind of individuals are included in the denotation of the nominals. If nothing else, it certainly seems that in order to account for this distribution, we need to carefully consider the difference between mass and count nouns in English. In some way or other, derived and mixed nominals are almost, but not quite, mass nouns. The denotation of gerundive nominals I have said little about how gerundive nominals fit into this picture, a question which I will address briefly at this point. Recall from the last part of Chapter 2 that an expression such as criticizing the book was assigned the semantic type . This means that there is no sense in which it denotes a set of individuals, plural or otherwise. Therefore it is not the right type of expression to be pluralized or to combine with determiners other than a possessive. From a purely syntactic standpoint, while a gerundive nominal is being built up there can never be any item of category N, so there is no way for normal determiners to combine with it. The account given in Figure 6 thus correctly predicts that gerundive nominals will have a limited distribution compared to derived and mixed nominals. Summary and Conclusions: Let me conclude with an overview of the purposes and accomplishments of this project. I have looked at a number of aspects of derived, mixed, and gerundive nominalization. In Chapter 1, I introduced the subject and outlined the relevant empirical facts to be accounted for. In order to find the correct facts, it was necessary to establish the distinction between derived result nominals and derived process nominals. In Chapter 2, I developed an account for derived process nominals and mixed nominals that

87

Tamina Stephenson

Honors Thesis Chapter 5

treats these expressions as relational nouns, which accounted for a number of facts from Chapter 1, in particular the role of the possessive “subject” of the nominals. In Chapter 3, I observed that – contrary to the generalization made in earlier chapters – there are certain situations where adverbs can appear with mixed nominals, though they never can with derived nominals. Although I was unable to give a full explanation for when this can occur, it did indicate that the process generating mixed nominals is somewhat more productive than previously assumed, and certainly more productive than the process generating derived nominals, if the latter can be called a process at all. In Chapter 4, I pointed out two more differences between derived and mixed nominals which indicated that derived and mixed nominals do not represent exactly the same process. These findings required a slight revision in the semantics of derived nominals but were otherwise consistent with the analyses developed in earlier chapters. Finally, in the current chapter, I explored the behavior of derived and mixed nominals compared to regular count and mass nouns, arguing for a tentative analysis where the denotation of derived and mixed nominals comprises both count-type entities and mass-type entities.

88

Tamina Stephenson

Honors Thesis Bibliography

References Barker, Chris, 1995. Possessive Descriptions. Stanford, California: CSLI Publications. Chierchia, Gennaro, 1984. Topics in the Syntax and Semantics of Infinitives and Gerunds. Ph.D. dissertation, University of Massachusetts at Amherst. Published in 1988 by Garland, New York. Chierchia, Gennaro, 1998. Reference to Kinds across Languages. Natural Language Semantics, 6.4: 339-405. Chomsky, Noam, 1970. Remarks on Nominalization. In Roderick A. Jacobs and Peter S. Rosenbaum (eds.), Readings in English Transformational Grammar: 184-221. Waltham, Massachusetts: Ginn. Grimshaw, Jane, 1990. Argument Structure. Cambridge, Massachusetts: MIT Press. Harley, Heidi, and Rolf Noyer, 1997. Mixed nominalizations, short verb movement and object shift in English. In Proceedings of NELS 28, GLSA, University of Massachusetts, Amherst. Heller, Daphna, presentation. Possession as a lexical relation: Evidence from the Hebrew Construct State. Practice talk, Rutgers University. Handout taken from presentation at SURGE (Rutgers semantics research group), March 26, 2002. Lasersohn, Peter, 1995. Plurality, Conjunction, and Events. Boston: Kluwer Academic Publishers. Ross, John R., 1972. Double -ing. Linguistic Inquiry, 3: 61-86. Sag, Ivan A., and Thomas Wasow, 1999. Syntactic Theory: A Formal Introduction. Stanford, California: CSLI Publications. Zucchi, Alessandro, 1993. The Language of Propositions and Events. Boston: Kluwer Academic Publishers.

89