LINGUIST List 4.1102

Sun 26 Dec 1993

Review: Beth Levin, English Verb Classes and Alternations

Editor for this issue: <>

Directory

1. T. Daniel Seely, Beth Levin, English Verb Classes and Alternations
2. Paul Kershaw, Beth Levin, English Verb Classes and Alternations

"English Verb Classes and Alternations: A Preliminary
Investigation." By Beth Levin. Chicago and London: The
University of Chicago Press, 1993. Pp. xviii, 348.

 Reviewed by T. Daniel Seely, Eastern Michigan
 University.

 Beth Levin's "English Verb Classes and Alternations: A
Preliminary Investigation" (hence- forth EVC) is an excellent
reference book. It presents syntactic and semantic information
which is valuable and easy to use. The book is rich in well-
organized data (there are thousands of entries in the verb index
and the bulk of the book is made up of dozens of diathesis
alternations and verb classes), it is thoroughly documented (there
are some 800 references), and it has important theoretical
implications (nicely traced in the Introduction). It is, in short, an
impressive accomplishment and it has become an indispensable part
of my linguistics library.

 EVC is a "set of resource materials" which can be used in a great
many ways (more on that later). But it is not just a disconnected list
of verb alternations and semantic subgroupings. Rather, it
represents a well-conceived enterprise whose overall coherence can
be found in the "hypothesis of semantic determination," viz

(1) The meaning of a verb determines its syntactic
 behavior.

 [Important corollaries of (1) are:

 a. If the members of a set of verbs S share some
 meaning component M, then the members of S can be
 expected to exhibit the same syntactic
 behavior(s), and

 b. If the members of a set of verbs S exhibit the
 same syntactic behavior(s), then the members of S
 can be expected to share some meaning
 component(s)]

This hypothesis is the book's guide, an extensive elaboration of its
requirements and implications is the book's substance. The
Introduction to EVC, for example, presents, illustrates, and defends
(1), Part I systematizes many of the syntactic behaviors (verb
alternations) relevant to it, and Part II contains some of the "...
semantically coherent verb classes" that result from using it as a
probe for "linguistically relevant aspects of verb meaning" (an
exploitation of(1)b). That is the frame, let me now add some of the
picture.

 As is made clear in the Introduction, (1) is controversial but worth
pursuing; controversial in light of apparent counterexamples, but
worth pursuing because

(2)a. some of the counterexamples turn out on closer
 inspection to actually support it,

 b. it has important theoretical implications
 regarding the nature and content of the lexicon,
 and

 c. it gives rise to a powerful research tool.

The very readable Introduction covers each of these points; I
summarize below.

 The counter-example that Levin considers, from Rosen (1984), is
this: The Italian verbs "russare" (snore) and "arrossire" (blush) share
the meaning component "bodily process" while exhibiting disparate
syntactic behavior--the former is unergative, the latter unaccusative.
(1) must therefore be wrong. As for the defense, Levin argues that
the case is damaging only if there is reason for assuming that "...the
semantic notion "bodily process" plays a part in determining a verb's
status with respect to the Unaccusative Hypothesis." As it turns out,
there are other semantic components of these verbs (involving
activity vs change of state) which predict divergent behavior and
which arguably do relate to the Unaccusative Hypothesis. Further
details need not concern us here (indeed, the discussion in the
Introduction on this point is essentially review of other work, by
Levin et al and others). The more general point is that (1) requires
determining the appropriate meaning components, something which
while necessary is not always easy. [Moreover, maintaining (1) in
the face of apparent counter-examples encourages us to look for, and
perhaps discover, meaning components which might not have
emerged otherwise.]

 Turning now to (2)b, the theoretical implications of the hypothesis
of semantic determination are important indeed. A traditional view
of the lexicon is wonderfully captured by Di Sciullo and Williams
(1987). "The lexicon," they write, "is like a prison--it contains only the
lawless, and the only thing that its inmates have in common is
lawlessness."

[NOTE that Levin herself gives the earlier and somewhat less vivid
statement of Bloomfield (1933): "The lexicon is really an appendix of
the grammar, a list of basic irregularities." I use the Di Sciullo-
Williams' quote (with full realization that there are great differences
between Bloomfield and Di Sciullo-Williams) because I like it a great
deal but also to set acontext for a mild criticism. Although EVC has
a tremendous bibliography, there are a number of conspicuous gaps.
There is no reference to Di Sciullo and Williams, for instance. Nor is
there reference to work on s- and c-selection by Grimshaw (1979)
and Pesetsky (1982), work which is clearly and directly relevant to
(1).]

 Continuing our main discussion, lexical information should be
minimized. If (1) is true, lexical information can be minimized in
rather dramatic fashion. As Levin explains (and I quote here
somewhat extensively to give the reader a feel for Levin's very
accessible style):

 If the syntactic properties of a verb indeed follow
 in large part from its meaning, then it should be
 possible to identify general principles that derive
 the behavior of a verb from its meaning. Given such
 principles, the meaning of a verb will clearly have a
 place in its lexical entry, but it is possible that
 the entry will need to contain little more. And
 since a word's meaning is necessarily idiosyncratic,
 the inclusion of a word's meaning in its lexical
 entry conforms to Bloomfield's characterization of
 the lexicon as a locus of idiosyncrasy. (p. 11)

And a bit later Levin summarizes:

 Taking this approach seriously requires a
 re-evaluation of previous assumptions concerning the
 contents of lexical entries, since it suggests that
 they may contain less information than has sometimes
 been proposed. Specifically, if there are indeed
 correlations between verb meaning and verb behavior,
 some properties that might have been included in
 lexical entries because they were thought to be
 idiosyncratic could turn out on further examination
 to be predictable from verb meaning and could be
 eliminated from a verb's lexical entry. (p. 12)

Much recent work, as Levin notes, considers these conceptually very
satisfying ideas. Levin's work represents a significant step in
creating the empirical basis for realizing them.

 (2)c, the final point, is that the hypothesis of semantic
determination gives rise to a powerful research tool. What this
amounts to is an exploitation of (1)b. If the members of some set of
verbs behave alike with respect to diathesis alternations (i.e. if they
exhibit the same syntactic behavior), then there should be some
meaning component which they have in common. And "the
availability of this technique for investigating word meaning is
important since it can be quite difficult to pin down the meanings of
words using introspection alone." (p 15)

 The members of the set of verbs {CUT, hack, saw,...}, for example,
participate in the same alternations. Using "cut" as the
representative case, they are found in the middle construction (3),
the conative alternation (4), and the body-part possessor ascension
alternation (5):

(3)a. Kimi cut the bread
 b. Bread cuts quite easily.

(4)a. Kimi cut the bread.
 b. Kimi cut at the bread.

(5)a. Kimi cut Bill's arm.
 b. Kimi cut Bill on the arm.

This contrasts with the sets {BREAK, crack, rip, shatter, snap, ...},
{TOUCH, pat, stroke, tickle, ...}, and {HIT, bask, kick, tap, ...}.
Using the uppercase verb as the representative, the following pattern
of behavior emerges (I use Levin's convenient table in the interest of
space, the reader can easily plug in the relevant examples):

 TOUCH HIT CUT BREAK
Conative: No Yes Yes No
Body-Part Ascension: Yes Yes Yes No
Middle: No No Yes Yes

Levin then shows that these sets have common meaning
components. Simplifying, the BREAK group doesn't require contact
nor motion but does involve change of state. The CUT group
requires contact and motion; Hit requires contact and motion but is
not change of state, and TOUCH needs contact, no motion, and is
not change of state. Overall, then, there is a clear relation between
the syntactic behavior and key elements of meaning. And it is also
clear that without the syntactic pattern as a guide, we might not
have grouped the words together to look for the shared meaning.
Consequently, Levin's research tool bears fruit.

Levin's book is rich in detail and it points toward extremely fruitful
further research. The issues above are important as is the book's
more general aim of "[paving] the way toward the development of a
theory of lexical knowledge." Progress can be made only if the
foundations are put into place. Levin's book is a solid foundation
indeed.

 As a final note, let me point out that although my comments have
focussed on (some of) Levin's theoretical underpinnings, I have
found many practical uses for the book. It has helped in making up
exercise sets for syntax, semantics, and morphology classes, for
example, it made checking the verbs of example sentences in a
psycholinguistic study much easier, and it has been invaluable (my
student's tell me) for creating exercises of various sorts in TESOL. It
is, after all, a reference work and like all good references it is limited
only by the imagination of its user. At one point Levin states "... I
hope that [this book] will be a valuable resource for linguists and
researchers inrelated fields." A hope most certainly realized!

REFERENCES

Bloomfield, L. (1933) "Language," Holt, New York.
Di Sciullo, A. M., and E. Williams (1987) "On the
 Definition of Word," Linguistic Inquiry Monograph 14,
 MIT Press, Cambridge.
Grimshaw, J. (1979) "Complement Selection and the
 Lexicon," Linguistic Inquiry 10, 279-326.
Pesetsky, D. (1982) "Paths and Categories," MIT
 dissertation.
Rosen, C. (1984) "The Interface between Semantic Roles
 and Initial Grammatical Relations," in D. M.
 Perlmutter and C. Rosen, (eds) (1984) "Studies in
 Relational Grammar 2," University of Chicago Press,
 Chicago, Il.

Comments on: Levin, Beth 1993 English Verb Classes and Alternations: A
preliminary investigation. The University of Chicago Press.-- Paul Kershaw,
Michigan State University, KershawPStudent.MSU.Edu

Description of the content: This book may be divided into three parts, to wit,
the introduction and parts one and two. The introduction, pp. 1-21, lays down
the theoretical foundation of the book. The book attempts to "delimit... and
systematiz[e] the facets of verb behavior" on the assumption that "the behavior
of a verb, particularly with respect to the expression and interpretation of
its arguments, is to a large extent determined by its meaning" (both quotes, p.1).

Part 1 consists of a series of diathesis alternations which distinguish various
verb classes, organized into major subtypes of alternations. For instance, the
alternations in section consist of transitivity alternations, "involving a
change in the verb's transitivity" (p. 25), exemplified by alternations between
NP V NP and NP V frames (i.e., +[__ (NP)] subcat) and between NP V NP and
NP V PP frames. This set includes alternations such as in (1)
(= Levin's (12)), (2) (= Levin's (38)), and (3) (= Levin's (113)):
(1) a. Jane broke the cup.
 b. The cup broke.
(2) a. Mike ate the cake.
 b. Mike ate.
(3) a. Jill met with Sarah.
 b. Jill met Sarah.

Part two consists of a series of verb classes, in each of which the members
have some semantic commonality, and together observe certain alternations and
properties. The idea is to minimize the cognitive load required by the
lexicon. Rather than knowing, for instance, that "flinch" does not allow an
object NP, and therefore does not participate in causative alternations, nor
allow a cognate or a reaction object, a speaker of English need only know that
"flinch" is a flinch verb, as are "cower", "cringe", "recoil", "shrink", and
"wince", and possibly "balk", and that the properties of flinch verbs in
general is as given above (for flinch verbs, p. 223).

Comments:

First of all, as a reference book, this book is well laid out. Most sections
consist of a list of references, a group of examples, and comments on the
section. Part one seems fairly well organized, although I wondered why
conative alternations (1.3, pp. 41-42, as in (84) Paula hit (at) the fence.)
deserved a section separate from preposition drop alternations (1.4, pp. 43-44,
as in (102) Martha climbed (up) the mountain. and (113) Jill met (with) Sarah).
Also, some sections lack adequate comments, and it often seemed more
appropriate the give the comments before, not after, the examples -- it is
often distracting to read the examples before finding out what they're examples
of.

There were a few other unnecessary distractions. One was the definite
preference for feminine proper names, which struck me as just as annoying and
inappropriate as the traditional preference for masculine proper names
(especially as agents). Also, for instance, Levin writes of one alternation,
2.13.3 Possessor and Attribute Alternation, "This alternation should probbly
not be recognized as a separate alternation, because it arises simply as a
consequence of the fact that admire-type psych-verbs show both possessor and
attribute objects" (p. 76). Why list an alternation and then qualify it by
saying it shouldn't be there? Why not eliminate it altogether? As with the
conative/prep-drop distinction, there may be a motivation, but Levin doesn't
express one.

Finally, as specific criticisms go, Levin perpetuates the implicit binary or
tertiary acceptibility judgements even when they don't seem necessary. For
instance, 1.2.2 Understood Body-Part Object Alternation (pp. 34-35) involve the
omission of conventionally understood objects ((40) The departing passenger
waved (his hand) at the crowd. vs. (41) Jennifer craned *0/her neck.) This
should predict that the acceptability of dropped objects should become more
acceptable if the verb is contextualized. For instance, while (44b) Celia
braided. seems odd out of context, it can easily be made (more) acceptable in
context (e.g., Janine permed her hair, and Stephanie combed it out straight,
but Celia always braided.). This seems like the sort of prediction Levin would
want to make, but doesn't.

What bothered me most, though, was the brevity of the theoretical section -- 21
pages. After all, Levin is accounting for one aspect of verbal behavior that
is typically reserved for syntax (i.e., the subcategorization of verbs for
arguments and theta-roles) by using semantic grouping. It would seem like this
book, then, would be an excellent forum for discussing the role that semantics
has in syntax, if any. It's not certain at what level these alternations take
place: are they pre-syntactic (that is, is one of several possible argument
structures for a verb determined at a lexical level, with this information
passed on to the syntax)? Or is the argument structure determined during the
syntactic derivation, with an interplay between syntax and semantics?
Specifically, I am interested in how the material laid out here meshes with a
theory such as in Grimshaw's 1990 Argument Structure (MIT Press), which Levin
cites examples from but does not (as far as I can tell) discuss in the brief
theoretical section. In Grimshaw, specific roles are for the most part
irrelevant (only the number and configuration of roles is important); for
Levin, specific roles seem very important.

To be sure, Levin is discussing the lexicon here, and not syntactic
derivations. But that is not as straightforward a defence as it sounds, since
she doesn't make clear exactly what the difference is. The idea of classing
verbs according to the syntactic behavior/argument structure is hardly new; the
idea of defining these classes by semantics, as Levin does, is tantalizing.
But the "preliminary" in the title definintely needs stressing: without an
expanded theoretical background justifying and delineating what it is she's
doing, its precise purpose is not clear. Persuant to this, it is this
theoretical background which I think to be the most relevant for discussion
here on the Linguist List.