7 Components of lexical meaning
7.5 Verb meanings
Much of the recent research on lexical semantics has focused on verb meanings.
One reason for this special interest in verbs is the fact that verb meanings have a direct influence on syntactic structure, and so syntactic evidence can be used to supplement traditional semantic methods.
10The following discussion is based on Engelberg (2011: 129–130); Lyons (1977: 317ff.).
7.5 Verb meanings A classic paper by Charles Fillmore (1970) distinguishes two classes of transi-tive verbs in English: “surface contact” verbs (e.g.,hit, slap, strike, bump, stroke) vs. “change of state” verbs (e.g.,break, bend, fold, shatter, crack). Fillmore shows that the members of each class share certain syntactic and semantic properties which distinguish them from members of the other class. He further argues that the correlation between these syntactic and semantic properties supports a view of lexical semantics under which the meaning of a verb is made up of two kinds of elements: (a) systematic components of meaning that are shared by an en-tire class; and (b) idiosyncratic components that are specific to the individual root. Only the former are assumed to have syntactic effects. This basic insight has been foundational for a large body of subsequent work in the area of verbal semantics.
Fillmore begins by using syntactic criteria to distinguish the two classes, which we will refer to for convenience as thehitclass vs. thebreakclass. Subsequent research has identified additional criteria for making this distinction. One of the best-known tests is the causative-inchoative alternation.11 Break verbs generally exhibit systematic polysemy between a transitive and an intransitive sense. The intransitive sense has an inchoative (change of state) meaning while the transitive sense has a causative meaning (19). As illustrated in (20),hitverbs do not permit this alternation, and often lack intransitive senses altogether.
(19) a. John broke the window (with a rock).
b. The window broke.
(20) a. John hit the tree (with a stick).
b. * The tree hit.
Additional tests include “body-part possessor ascension” (21–22),12the cona-tive alternation (23–24),13and the middle alternation (25).14Each of these tests demonstrates a difference between the two classes in terms of the potential syn-tactic functions (subject, direct object, oblique argument, or unexpressed) of the agent and patient.
(21) a. I {hit/slapped/struck} his leg.
b. I {hit/slapped/struck} him on the leg.
11Fillmore (1970: 122–123).
12Fillmore (1970: 126).
13Guerssel et al. (1985); Levin (1993).
14Fillmore (1977); Hale & Keyser (1987); Levin (1993).
127
7 Components of lexical meaning
(22) a. I {broke/bent/shattered} his leg.
b. * I {broke/bent/shattered} him on the leg.
(23) a. Mary hit the piñata.
b. Mary hit at the piñata.
c. I slapped the mosquito.
d. I slapped at the mosquito.
(24) a. Mary broke the piñata.
b. * Mary broke at the piñata.
c. I cracked the mirror.
d. * I cracked at the mirror.
(25) a. This glass breaks easily.
b. * This fence hits easily.
These various syntactic tests (and others not mentioned here) show a high de-gree of convergence; that is, the class ofbreakverbs identified by any one test matches very closely the class ofbreakverbs identified by the other tests. This convergence strongly supports the claim that the members of each class share certain properties in common. Fillmore (1970: 125) suggests that these shared properties are semantic components: “change of state” in the case of thebreak verbs and “surface contact” in the case of thehit verbs. Crucially, he provides independent semantic evidence for this claim, specifically evidence that break verbs do buthitverbs do not entail a change of state (26).15Sentence (26a) is lin-guistically acceptable, although surprising based on our knowledge of the world, while (26b) is a contradiction. Example (27) presents similar evidence for the entailment of “surface contact” in the case of thehitverbs.
(26) a. Ihitthe window with a hammer; it didn’t faze the window, but the hammer shattered.
b. * Ibrokethe window with a hammer; it didn’t faze the window, but the hammer shattered.
(27) a. * Ihitthe window without touching it.
b. Ibrokethe window without touching it.
15Fillmore (1970: 125).
7.5 Verb meanings Without this kind of direct semantic evidence, there is a great danger of falling into circular reasoning, e.g.: break verbs permit the causative-inchoative alter-nation because they contain the component “change of state”, and we know they contain the component “change of state” because they permit the causative-inchoative alternation. As many linguists have learned to our sorrow, it is all too easy to fall into this kind of trap.
Whilebreakverbs (e.g.,break, bend, fold, shatter, crack) all share the “change of state” component, they do not all mean the same thing. Each of these verbs has aspects of meaning which distinguish it from all the other members of the class, such as the specific nature of the change and selectional restrictions on the object/patient. Fillmore (1970: 131) suggests that only the shared component of meaning has syntactic consequences; the idiosyncratic aspects of meaning that distinguish onebreakverb from another do not affect the grammatical realization of arguments.
Levin (1993) builds on and extends Fillmore’s study of verb classes in English.
In her introduction she compares the break andhit verbs with two additional classes,touchverbs (touch, pat, stroke, tickle, etc.) andcut verbs (cut, hack, saw, scratch, slash, etc.). Using just three of the diagnostic tests discussed above, she shows that each of these classes has a distinctive pattern of syntactic behavior, as summarized in (28). The examples in (29–31) illustrate the behavior oftouch verbs andcutverbs.16
(28) English transitive verb classes17
touchverbs hit verbs cutverbs breakverbs body-part
possessor ascension
yes yes yes no
conative alternation
no yes yes no
middle no no yes yes
(29) body-part possessor ascension:
a. I touched Bill’s shoulder.
b. I touched Bill on the shoulder.
16Examples adapted from Levin (1993: 6–7).
17Levin (1993: 8)
129
7 Components of lexical meaning c. I cut Bill’s arm.
d. I cut Bill on the arm.
(30) conative alternation:
a. Terry touched the cat.
b. * Terry touched at the cat.
c. Margaret cut the rope.
d. Margaret cut at the rope.
(31) middle:
a. The bread cuts easily.
b. * Cats touch easily.
Levin proposes the following explanation for these observations. Body-part possessor ascension is possible only for verb classes which share the surface contact component of meaning. The conative alternation is possible only for verb classes whose meanings include both contact and motion. The middle con-struction is possible only for transitive verb classes whose meanings include a caused change of state. The four classes pattern differently with respect to these tests because each of the four has a distinctive set of meaning components, as summarized in (32).
(32) Shared components of meaning18
touchverbs contact
hitverbs motion, contact
cutverbs motion, contact, change
breakverbs change
These verb classes have been found to be grammatically relevant in other lan-guages as well. Levin (2015) cites the following examples: DeLancey (1995; 2000) on Lhasa Tibetan; Guerssel et al. (1985) on Berber, Warlpiri, and Winnebago;
Kroeger (2010) on Kimaragang Dusun; Vogel (2005) on Jarawara.
In the remainder of her book, Levin (1993) identifies 192 classes of English verbs, using 79 diagnostic patterns of diathesis alternations (changes in the way that arguments are expressed syntactically). She shows that these verb classes are supported by a very impressive body of evidence. However, she states that establishing these classes is only a means to an end; the real goal is to understand meaning components:
18Adapted from Saeed (2009: 268).
7.6 Conclusion [T]here is a sense in which the notion of verb class is an artificial construct.
Verb classes arise because a set of verbs with one or more shared meaning components show similar behavior… The important theoretical construct is the notion of meaning component, not the notion of verb class.19
Like Fillmore, Levin argues that not all meaning components are grammat-ically relevant, but only those which define class membership. The aspects of meaning that distinguish one verb from another within the same class (e.g.punch vs. slap) are idiosyncratic, and do not affect syntactic behavior. Evidence from diathesis alternations can help us determine the systematic, class-defining mean-ing components, but will not provide an analysis for the idiosyncratic aspects of the meaning of a particular verb.
As noted above, verb meanings cannot be represented as an unordered bun-dle of components, but must be structured in some way. One popular method, referred to as lexical decomposition, is illustrated in (33). This formula was proposed by Rappaport Hovav & Levin (1998: 109) as a partial representation of the systematic components of meaning for verbs likebreak. In this formula, x represents the agent andythe patient. The idiosyncratic aspects of meaning for a particular verb root would be associated with the state predicate (e.g.broken, split, etc.).
(33) [[x ACT] CAUSE [BECOME [y <state> ]]]
7.6 Conclusion
The idea that verb meanings may consist of two distinct parts, a systematic, class-defining part vs. an idiosyncratic, verb-specific part, is similar to proposals that have been made for content words in general. Fillmore (1970: 131) notes that a very similar idea is found in the general theory of word meaning proposed by Katz & Fodor (1963). These authors suggest that word meanings are made up of systematic components of meaning, which they refer to as semantic markers, plus an idiosyncratic residue which they refer to as the distinguisher.
This proposal is controversial, but there do seem to be some good reasons to distinguish systematic vs. idiosyncratic aspects of meaning. As we have seen, Fillmore and Levin demonstrate that certain rules of syntax are sensitive to some components of meaning but not others, and that the grammatically relevant com-ponents are shared by whole classes of verbs. Additional motivation for making
19Levin (1993: 9–10).
131
7 Components of lexical meaning
this distinction comes from the existence of systematic polysemy. It seems logical to expect that rules of systematic polysemy must be stated in terms of systematic aspects of meaning.
However, there is no general consensus as to what the systematic aspects of meaning are, or how they should be represented.20 Some scholars even deny that components of meaning exist, arguing that word meanings are atoms, in the sense defined in §7.4.21 Under this “atomic” view of word meanings, lexical entailments might be expressed in the form of meaning postulates like the following:
(34) ∀x[STALLION(x) → MALE(x)]
∀x[BACHELOR(x) → ¬MARRIED(x)]
Many scholars do believe that word meanings are built up in some way from smaller elements of meaning. However, a great deal of work remains to be done in determining what those smaller elements are, and how they are combined.
Further reading
Engelberg (2011) provides a good overview of the various approaches to and controversies about lexical decomposition and componential analysis.
Lyons (1977: 317ff.) discusses some of the problems with the binary feature approach to componential analysis. The first chapter of Levin (1993) gives a very good introduction to the Fillmore-type analysis of verb classes and what they can tell us about verb meanings, and Levin (2015) presents an updated cross-linguistic survey of the topic.
20For one influential proposal, see Pustejovsky (1995).
21E.g. Fodor (1975) and subsequent work.
7.6 Conclusion
Discussion exercises
A. Componential analysis of meaning. Construct a table of semantic components, represented as binary features, for each of the following sets of words:
1. bachelor, spinster, widow, widower, husband, wife, boy, girl 2. walk, run, march, limp, stroll
3. cup, glass, mug, tumbler, chalice, goblet, stein
B. Locative-alternation (“spray-load”) verbs.a Based on the following examples, fill in the table below to show which verbs allow the goal or location argument to be expressed as direct object and which verbs allow the displaced theme argument to be expressed as direct object. Try to for-mulate an analysis in terms of meaning components to account for the patterns you find in the data.
(1) a. Jack sprayed paint on the wall.
b. Jack sprayed the wall with paint.
(2) a. Bill loaded the cart with apples.
b. Bill loaded the apples onto the cart.
(3) a. William filled his mug with guava juice.
b. * William filled guava juice into his mug.
(4) a. *William poured his mug with guava juice.
b. William poured guava juice into his mug.
(5) a. Ailbhe pushed the bicycle into the shed.
b. #Ailbhe pushed the shed with the bicycle. [different meaning]
(6) a. Harvey pulled me onto the stage.
b. #Harvey pulled the stage with me. [different meaning]
(7) a. Libby coated the chicken with oil.
b. ?*Libby coated the oil onto the chicken.
133
7 Components of lexical meaning
(8) a. Mike covered the ceiling with paint.
b. * Mike covered the paint onto the ceiling.
Verb Theme = object Location = object
fill no yes
load spray cover coat pour push pull
aAdapted from Saeed (2009), ch. 9.
Homework exercises
Causative/ inchoative alternation.a Rappaport Hovav & Levin (1998:
102–105) propose a semantic explanation for why some change of state verbs participate in the causative/inchoative alternation (John broke the window vs.the window broke), while others do not. They suggest that verbs which name events that must involve an animate, intentional and volitional agent never appear in the intransitive form. This hypothesis predicts that only (but not necessarily all) verbs which allow an inanimate force as subject should participate in the alternation, as illustrated in (a–b).
Your tasks: (i) construct examples like those in (a–b) to test this prediction
7.6 Conclusion
for the following verbs, and explain what your examples show us about the hypothesis:melt, write, shrink, destroy; (ii) Use Levin & Rappaport Ho-vav’s hypothesis to explain the contrasts in sentences (c–d).
a. A terrorist/*tornado assassinated the governor.
*The governor assassinated.
b. The storm broke all the windows in my office.
All the windows in my office broke.
c. The sky/*table cleared.
d. Paul’s window/*contract/*promise broke.
aAdapted from Saeed (2009: 298), ex. 9.3.
135