LINGUIST List 14.2354

Sat Sep 6 2003

Review: Syntax: Wasow (2002)

Editor for this issue: Naomi Ogasawara <naomilinguistlist.org>


What follows is a review or discussion note contributed to our Book Discussion Forum. We expect discussions to be informal and interactive; and the author of the book discussed is cordially invited to join in. If you are interested in leading a book discussion, look for books announced on LINGUIST as "available for review." Then contact Simin Karimi at siminlinguistlist.org.

Directory

  1. Loren A. Billings, Postverbal behavior

Message 1: Postverbal behavior

Date: Fri, 05 Sep 2003 16:23:28 +0000
From: Loren A. Billings <billingsncnu.edu.tw>
Subject: Postverbal behavior

Wasow, Thomas (2002) Postverbal behavior. CSLI Publications, paperback
ISBN 1-57586-402-9, xiv+185pp, CSLI Lecture Notes 145, $25.00 (also
available clothbound ISBN 1-57586-401-0, $65.00).

Announced at http://linguistlist.org/issues/13/13-2192.html


Loren A. Billings, National Chi Nan University, Taiwan

DESCRIPTION

This monograph serves three interconnected purposes. First, it
explores the factors that influence the order of postverbal
constituents in an English clause (chapters 1 through 4). To do so,
Wasow uses statistics from elicitation experiments and corpus
studies. The second aim is to show that probabilistic factors have a
place in the study of syntax (chapter 5). This supports the book's
third purpose (chapter 6): to argue for greater use of statistical
methodology in syntax.

The book ''is based entirely on English evidence'' (p. 14), although
in a few places Wasow cites works that report how other languages
(namely: Japanese, Korean, and Polish) do not pattern in the same
way. Aside from having access to large and conveniently tagged corpora
in--and an ample supply of native speakers of--English, Wasow has a
theoretical reason to select this language. Although English has
relatively fixed order of constituents, there is a lot of variability
in how postverbal noun phrases (NPs) and other constituents are
ordered.

Three related phenomena in English studied by Wasow are heavy-NP
shift, dative alternation, and the particle-verb construction. To
begin, heavy-NP shift describes the positioning of a particularly
heavy nominal at the end of the clause. Dative alternation is
characterized by the relative optionality of ordering of a
ditransitive verb's two internal arguments (as in 'Chris gave the boy
a book' and 'Chris gave a book to the boy'). So-called particle verbs
allow their direct object to appear on either side of a
preposition-like element (as in 'Pat picked up a book' and 'Pat picked
a book up'). In each of these, the weight of a constituent can affect
its ordering relative to some other constituent.

Wasow's preface discloses a play on words in the book's title:
Chomskys (1959) review of Skinner's (1957) 'Verbal
behavior' is ''generally regarded as a landmark in the history of
linguistics--indeed of the cognitive sciences more generally.'' This
turning point was characterized by both the wholesale rejection of
behaviorism, Wasow writes, and the subsequent abandonment of
systematic procedures in how data are collected (p. xi). Wasow's book
thus looks back on the generative-syntactic enterprise, reassessing
these two issues. (Incidentally, a .pdf version of the table of
contents, preface, and chapter 1 can be downloaded from the publisher
for free: http://csli-publications.stanford.edu/pdf/1575864029.pdf;.)

Many of the studies reported in the book were conducted with
others. Among Wasow's collaborators are Jennifer Arnold, Herb Clark,
Jack Hawkins, Barbara Lohse, and Adam Yarnold. Some of these studies
are published separately as journal articles (e.g., Arnold et al. 2000
and Wasow 1997).

EVALUATION

The remainder of this review is organized as follows: I begin by
assessing individual body chapters of the book in the order that Wasow
presents them. I then briefly discuss more technical things. This
review ends with a few comments about salient issues that come up more
than once in the book.

Weight

Chapter 2 is entirely devoted to exploring, defining, and explaining
weight as a factor in determining the likelihood of heavy-NP shift,
dative alternation, and the separability of a verb from its
particle. Weight is also the one factor against which other factors
are compared in subsequent chapters. Wasow begins with the Principle
of End Weight (p. 3, quoting Quirk, et al. 1972): ''Phrases are
presented in order of increasing weight.'' He then summarizes eight
definitions of weight reported in the literature (pp. 16-17). Some are
categorical; others, graduated. (I discuss a few of these below.) The
syntactic definitions mostly refer to whether the nominal contains a
clause or prepositional phrase (PP) or whether there is a postposed
modifier.

While Wasow is generally dismissive of a phonological definition of
weight (about which I have more to say below), he contends that
multiple factors help to determine the order of constituents
(p. 23). Much of chapter 1 is devoted to pinning down which syntactic
definition of weight is crucially relevant: word count or arboreal
complexity.

Because his corpus studies fail to decide between these two
yardsticks, Wasow designed an elicitation experiment to decide the
question (pp. 32-37). I summarize that study briefly. Although
similar in length, 'what happened' and 'the situation' in a
particle-verb context (e.g., 'We talked ... over' and 'We talked over
...') are different in terms of complexity; only 'what happened'
contains a clause. Analogous contexts using dative alternation and
heavy-NP shift were also considered. For all three of these
constructions, based on subjects' responses using four-point
acceptability scales, Wasow finds that participants were more likely
to rate sentences as acceptable if NPs with internal complexity
appeared later in the clause (e.g., after both the verb and particle:
'We talked over what happened'), whereas NPs without such complexity
didn't have to be postposed to be as acceptable (e.g., 'We talked the
situation over'). That is, NP-internal complexity is a
factor--independent of the number of words in the NP.

Using a corpus study, Wasow also shows that PP-internal NPs
statistically are neither more nor less likely to affect the order in
a verb-particle construction than NPs of the same length without
internal PPs. At this point, I think Wasow criticizes Chomsky somewhat
unfairly, suggesting that ''Chomsky was wrong about particle position
being sensitive to complexity rather than length'' (p. 40), apparently
referring to the following passage:

''It is interesting to note that it is apparently not the length in
words of the object that determines the naturalness [...]. Thus 'they
brought all the leaders of the riot in' seems more natural than 'they
brought the man I saw in.' The latter, though shorter, is more
complex...'' [pp. 15-16, quoting Chomsky 1975:477]

Nothing in this quotation (not all of which I repeat here) even
implies that an internal PP increases a nominal expression's
complexity. In this sense, Chomsky appears to follow the definition
proposed by Ross 1967, which refers only to NP-internal clauses
(p. 16). Wasow is thus pinning the blame on the wrong theorist; it is
Emonds (1976:112) who allows for PPs as well as clauses in his
definition of complex NPs. Still, on the following page, in connection
with dative alternation, Wasow credits Chomsky (as well as Hawkins
1994) with having observed that structural complexity, not just
length, is relevant.

Wasow concludes that ''any weight definition must be graduated, rather
than categorical.'' He adds that ''an explanation of weight effects
must take into account the relative weights of constituents'' (p. 57).

Building upon and critiquing Hawkins (1994:57), who writes that
grammatical weight makes it easier for syntactic constituents to be
''recognized (and produced) as rapidly and efficiently as possible in
language performance'' (p. 42), Wasow argues at length that postposing
heavy constituents is primarily to the speaker's (as opposed to the
listener's) advantage (pp. 42-56; see also Wasow 1997). The details
of the corpus studies that Wasow reports are somewhat complicated, so
I only sketch the idea here. In most cases, what is easier for a
speaker to process will ease the listener's comprehension as
well. However, it is possible to find constructions where processing
load differs for these two parties. Wasow reports studies using (both
written and spoken) corpora in which the choice of orders in heavy-NP
shift and dative alternation, he contends, is influenced by planning
(rather than by decoding) an utterance. (Another corpus study
involving particle verbs failed to support either Wasow's claim or an
alternative one put forth by Stallings et al. 1998.) Wasow concludes
that ''postponing difficult constituents and keeping options open as
long as possible facilitates planning during utterance production''
(p. 56).

Newness

Discourse-new constituents also tend to be heavy. Although both
information structure and weight have been studied extensively,
according to Wasow the relationship between the two had not been
assessed prior to his studies. In chapter 3 Wasow reports
constructions that tease apart the two properties. Corpus studies and
an elicitation experiment both lead him to conclude that newness is
needed separately from weight.

The literature on information structure is fraught with distinctions
and overlapping definitions (e.g., topic vs. comment, presupposition
vs. focus, theme vs. rheme, and given vs. new). Wasow masterfully
boils down the vast literature--functionalist and generative alike--to
just a few paragraphs. Following Gundel (1988), Wasow sees two
distinct claims: ''State what is given before what is new in relation
to it'' and ''Provide the most important information first''
(pp. 62-63).

Several statistical studies, also reported in Arnold et al. (2000),
show that newness to the discourse increases an NP's likelihood to
appear later in the postverbal domain. First, a corpus study--with
quite a bit of coding and culling by hand--was conducted. For both
heavy-NP shift and dative alternation, information structure was found
to be highly significant (independently of weight). In addition, an
intricate and rather ingenious experiment--with subjects telling other
participants to give objects of various sorts to toy animals--was also
used to determine the interaction of newness and weight. Using a
logistic regression, Wasow shows that weight (p[robability of error] <
.005) and discourse status (p < .001) are highly significant, as is
the interaction of these two factors with participant identity (p <
.001 for both).

Other factors

Rounding out the empirical part of the book, chapter 4 considers three
additional factors which--in addition to weight and information
structure-- appear to influence postverbal ordering. These are
semantic connectedness, lexical bias, and avoiding ambiguity.

Previously unpublished studies (done by Lohse, Hawkins, & Wasow) show
that particles that are dependent on the verb ''occur adjacent to the
verb more frequently than independent particles'' (p. 86). A particle
is dependent on the verb if its lexical meaning does not contribute to
sentential meaning. (For example, 'I ate the bagel up' does not
entail that the bagel went up.) Semantic connectedness, Wasow
reports, is not a very strong effect but still significant (if weight,
itself a stronger effect, is controlled for).

Next, Wasow reports that the choice of lexical verb itself appears to
be a factor in how likely the following constituents are to appear in
one order or the other. For example, among ditransitive verbs, a
corpus study shows that _tell_ invariably uses the V-NP-NP order,
while _sell_ almost always uses the V-NP-PP order. Other verbs fall
somewhere in between (namely: tell > give > show > hand > fax > bring
> send > sell ). Wasow allows for the possibility that this factor
may derive from something else; at this point he admits that he cannot
isolate what such a factor is.

The bulk of chapter 4 is then devoted to whether avoiding ambiguity is
a factor influencing postverbal ordering (independent of other factors
mentioned so far). Despite the vast literature showing that ambiguous
sentences are harder to process, Wasow ''found no compelling evidence
that speakers use ordering to avoid attachment ambiguities''
(pp. 88-89), adding that his results are not entirely unequivocal.

In some corpus studies of written texts there is apparently an
ambiguity effect. However, Wasow adds (pp. 95-96), ''when an
interaction factor is added to the analysis, length and the
interaction of length with ambiguity are both predictors of shifting,
but ambiguity is not. These results suggest that ambiguity avoidance
may exert some influence on constituent ordering, but the influence is
quite limited.'' Wasow conjectures that writers--and especially
editors, who serve as both readers and writers--are more likely than
speakers to anticipate ambiguity in a text (p. 94). For this reason,
he searched the (spoken) Switchboard corpus; here he found not even
one example that could be explained using avoidance of ambiguity. Nor
were two experiments he ran able to detect an ambiguity factor in
spoken situations. Wasow correctly (in my view) conjectures that
prosodic cues often disambiguate speech; 'we send the faculty lists'
can be read as mono- or ditransitive: 'we send [the faculty lists]' or
'we send [the faculty] [lists]' (respectively). These two parses are
not homophonous, however. The bracketed part of the latter parse
contains a compound noun with more prosodic prominence on the first
lexical word: [the FA.cul.ty lists].

The issue of prosodic cues aside, however, I think that Wasow is
missing an opportunity to support his claim elsewhere (chapter 2 and
Wasow 1997) that End Weight serves the speaker's planning, not the
listener's recognition. Although I am not that familiar with the
psycholinguistic literature on ambiguity and processing, it seems to
me that ambiguity is mainly relevant to recognition (i.e., listening
and reading), not production--especially not natural, spoken
production; see, e.g., Ferreira & Dell (2000) or Snedeker & Trueswell
(2003). If the listener's and the speaker's tasks are truly distinct,
as Wasow argues quite plausibly, then avoiding ambiguity shouldn't
necessarily be relevant to a speaker's algorithm for determining the
order of postverbal constituents.

Wasow is careful to hedge himself when discussing increasingly less
robust factors in the determination of postverbal ordering. Weight is
certainly a factor (chapter 2), as is newness--independent of weight
(chapter 3). The additional factors assessed in chapter 4--semantic
connectedness, lexical bias, and avoiding ambiguity--are only
plausible candidates. Whether these turn out to be separate factors or
not is beside the point. Throughout the book, Wasow maintains that
just one or two factors alone fail to account for all
postverbal-ordering data fully.

Theoretical and methodological issues

Chapters 4 and 5 shift the discussion back to syntactic theory. These
two chapters also ask whether the kind of findings of the previous
chapters have a place in such a theory.

Wasow begins chapter 4 (called ''Theoretical and metatheoretical
issues'') by summarizing in detail one influential work on
double-object constructions in English, Larson (1988), and how it
fails to account for certain more probabilistic aspects associated
with weight. Among other things not adequately handled by
generativists (although in this case Wasow does not criticize Larson
1988 as such), sentences like *'Mike talked to about politics
yesterday my friends' (Ross 1967:125) are not unacceptable in every
instance. The following is an attested counterexample: '[...] a
landowner's right to rely on, in good faith, a city's representation'
(p. 128, quoting the Orlando Sentinel). While such counterexamples
are rare, Wasow writes (p. 129), ''they are acceptable and do
occur. Consequently, any analysis that rules them out is
inadequate. The challenge is to come up with a theory that permits
such examples while also addressing their infrequency.'' (I might add
that Wasow's four counterexamples, all taken from newspapers, involve
an adjunct separating the preposition from its complement.) Wasow
later writes that ''generative grammarians need to provide a
justification for the systematic exclusion of frequency information
from their models of linguistic competence.'' At the same time,
''there is clearly a heavy burden of proof [on the part of
anti-formalists] to show how a purely probabilistic system of
constraints can capture the many discoveries that have come out of the
generative tradition'' (p. 133).

Primarily in chapters 4 and 5, Wasow cites no less than 20 works by
Chomsky in connection with the related issues of probability and
statistical methodology, almost always citing exact page numbers and
often quoting lengthy excerpts and interacting with them. (However, I
find it interesting that Wasow doesn't actually interact with Chomsky
1959, despite that review being so influential in the field and
instrumental in the choice of the book's title.) Among the many issues
Wasow raises are the distinction between I(nternal)- and
E(xternal)-language and the nearly parallel issue of competence
vs. performance. Keeping (categorical) grammar separate from
probabilistic factors, Wasow claims, is a conceptual issue that
appears to inhibit descriptive adequacy (p. 140); ''Chomsky's
extensive writings on I- language vs. E-language seem to presuppose
that linguists must choose one or the other as the definition of
language'' and then adds forcefully (p. 146): ''I reject that
presupposition.''

What makes one theory better than another? Chapter 5 (which Wasow
entitles ''Methodological issues'') boils the answer down to
simplicity, generality, and coverage. That is, generative
syntacticians use these three criteria as guidelines. Although Wasow
is clearly sympathetic to the idea that a child learns a language in a
very short time span, he writes that appeals to rapid learnability
''simply have no basis without some formal foundation'' (p. 152);
these things must be confirmed empirically, he implies. Regarding
generality, Wasow attempts to point out an inconsistency in Chomsky's
writings about language as a biological system on the one hand and
redundancy on the other. If language is an organ that has adapted into
its present form, then it is bound to have parts that perform
identical functions. Wasow's discussion of coverage is more
autobiographical. Writing about his own work, Wasow (1972), he
stresses that some crossover phenomena are absolutely unacceptable
while others are merely degraded. (Since that work, generativists have
made a clear distinction between strong and weak crossover primarily
because strong-crossover violations were thought to derive from a
different principle.) Wasow's point is that hardly anyone has
mentioned that weak-crossover violations are not absolutely
unacceptable; generative syntacticians just aren't interested in
graduated judgments.

Wasow asks why more generativists don't confirm their theories
empirically. He comes up with two answers. First, many analyses are
formulated too imprecisely to be testable against corpora. I find this
claim a little far-fetched. In his own studies, Wasow had to make many
judgment calls in order to allow statistical measurement. Some
judgments were reasonable; others, quite a matter of
preference. Regardless of whether it makes sense to me or anyone else,
a judgment call is still a weakness in the process of empirical
verification. Wasow's other answer is that theorists suspect that
their models won't stand up to empirical scrutiny. I would add a third
reason: laziness. Many linguists are comfortable with philosophizing
about language or (at most) being introspective about it. Enough
linguists agree that ideas count for more than the facts, making for
the current situation.

Considering that Wasow is apparently trying to win over generativists
in this regard, it might help if some additional readings were
presented clearly. Some of these are scattered throughout his
bibliography: Abney (1997), Manning & Sch�tze (1999), and
Sch�tze (1996). However, at least one important work is not
cited at all in the book: Cowart (1997). And Sorace & Keller (to
appear) is worth mentioning as well. All these are listed in my
references below. More technical matters

As in every review, there are details that need to be
addressed. Generally, the book is handsomely put together. Here are
some additional details:

The front matter contains Library of Congress
cataloging-in-publication data, the acid-free standard of the book's
paper, and a relatively complete table of contents. At the back of the
monograph, following the American Psychological Association (APA)
style used by psycholinguists in general, Wasow's references list
authors' and editors' surnames but only initials of their given
names. (In a sense, this choice between APA and the styles used in the
general-linguistics journals is another indicator of how divided the
theorists and psycholinguists are from each other.) Just a few
citations in the book do not appear in the references: Erteschik-Shir
1979 (p. 8), Giv�n 1983 (p. 68), Manning and Sch�tze
1999 (p. 132, fn. 7). In addition, Hawkins (2001) has now
appeared. Citations for all of these are listed my references below. I
also found the name and subject indexes to be mostly (but not
entirely) complete. The only truly confusing typo that I found
involved the labels under pairs of bars in a graph (p. 56); based on
the accompanying text's numbers, the labels ''Brown Corpus'' and
''Switchboard Corpus'' should just be swapped. The remaining typos--I
counted about a dozen--were not that distracting; many refer to the
wrong example/footnote/volume/page number or year.

Regarding the statistics, I did not detect any serious flaws in the
design and methodology. However, the level of detail in how the
elicitation and corpus studies reported in chapters 2 through 4 should
not qualify the book as the primary report of a piece of
social-science research. I say this not so much as a criticism of the
book but to prepare some readers familiar with such research who may
expect more detail. Some of the studies reported in the book have
appeared separately as peer-reviewed articles in journals. Despite
Wasow's longstanding interaction with psycholinguists, this book is
nonetheless aimed at mainstream syntacticians. As such, the way some
of the statistics is reported might raise a social scientist's
eyebrow. In one case, however, I also found some inconsistencies in
how the findings are reported. In addition to other justification of
internal complexity in determining grammatical weight, Wasow ran
(apparently two-way) analysis-of- variance (ANOVA) tests. I repeat his
prose and the following quantitative findings (reformatted here with
line numbers for expository purposes):

''Analyses of variance also reveal that complexity significantly
affects preferences in ordering. For heavy NP shift and the dative
alternation, the results are significant by both subjects and
items. For the verb-particle construction, they are significant only
by subjects. [...]

1. [Heavy-NP shift]. F1(1,87)=13.5, p<.001; F2(1,3)=10.6, p<.05.

2. Verb-Particle. F1(1,87)=53, p<.001[;] F2 p>.1.

3. Dative alternation (goal). F1(1,86)=14, p<.001[;] F2(1,1)=625, p<.05.

4. Dative alternation (theme). F1(1,86)=17, p<.001[;] F2 p>.1.'' [p. 38]

I will spell out only the technical notations relevant to my
critique. The two factors are the subjects (F1) and the data items
being compared (F2). I see some problems with this excerpt. The first
is a matter of convention: Wasow twice lists statistically
insignificant findings: ''F2 p>.1'' (in both instances). I can
understand reporting figures that are close to being significant, but
here the probability of error is over twice that of .05 (= statistical
significance). Again, this is not an error as I see it but rather just
a deviation from notational convention; perhaps Wasow lists these
figures just for ease of comparison. The second problem is more
substantial, however: Wasow writes, ''For [...] dative alternation,
the results are significant by both subjects and items.'' However,
this is not consistent with the findings on line 4. The results were
significant for both subjects and items only for line 3, in which the
goal (or indirect object) was manipulated for complexity. One final
qualm I have with this excerpt is the number of subjects. Earlier
(p. 34), Wasow mentions that there were 88 completed questionnaires,
one of which was discarded; ''one (randomly selected) participant's
responses to each of the other questionnaires were also thrown out.''
It appears that this balancing was performed for the follow-on ANOVA
tests discussed only within the two dative-alternation comparisons on
lines 3 and 4--i.e., ''(1,86)''--and not for lines 1 and 2, with
''(1,87)''; the number after the comma in these formulas is the number
of subjects (88 or 87) or items (four) minus one.

This passage aside, however, I found the level of detail sufficient to
guide a grammarian through the statistics without too much discussion
of methodological matters. In addition, jargon (e.g., the name of
statistical tests) was generally spelled out clearly for the
uninitiated.

Recurrent issues

In the remainder of this review, I touch on some issues that come up
more than once in Wasow's book. These topics are the rift between
experimental and theoretical linguistics, extrasyntactic issues
overlooked in the monograph, and ''Why only after the verb?''

In a number of places, especially the somewhat autobiographical
preface, Wasow talks about his transformation from grammarian to
psycholinguist. (See also Wasow's early empiricist leanings in
connection with his work on weak crossover, discussed above.) The
following is also a telling vignette: ''My change in methodology was
quite naturally accompanied by a change in the questions I tried to
answer. In particular, instead of seeking to formulate and justify
formal analyses (rules, lexical entries, parameters, or what have you)
within some generative theory, I asked, more straight- forwardly, what
led speakers (and writers) to select one constituent ordering rather
than the other.

''I quickly discovered that most generative linguists had little
interest in the new question I was asking. Psycholinguists--especially
those trained as psychologists--thought it was worth pursuing [...]. I
observed that psychologists spend most of their time and effort on
insuring the quality of their data--that is, on experimental design
and analysis. That emphasis is strikingly different from generative
linguists', and I have on occasion found myself wondering why
psychologists give so little attention to developing high-level
theories. It seems to me that both fields would benefit from acting a
little more like the other.'' [pp. xiii-xiv]

Just as Wasow argues that probabilistic and categorical effects should
not be isolated, I'm surprised that he continues to isolate the fields
of psycholinguistics and theoretical linguistics. Perhaps this is just
sad commentary on his part. If empirical confirmation of theoretical
claims is necessary, then perhaps the wall between the two fields
needs to come down.

Moving next to extrasyntactic factors, I found some of the discussion
of phonology and semantics--and the lack of attention to morphology
altogether--a bit worrysome. I address these three subdisciplines in
reverse order. To begin, Wasow generally ignores the morpholexical
effects of dative alternation. To be fair, one can't discuss
everything in in a single monograph. Still, some references might have
been useful. Turning next to semantics, I found the discussion scant
and erratic. For example, Wasow briefly considers spray/fill verbs as
in the following alternation: 'Pat sprayed the wall with red paint'
and 'Pat sprayed red paint on the wall' (p. 9). However, because of
''meaning differences'' between the variants, this construction isn't
investigated further. Agreed, there are meaning distinctions, but
aren't equally elusive semantic differences to be found in dative
alternation (i.e., between 'Pat bought the boy a book' and 'Pat bought
a book for the boy')? Turning finally to phonology, there was quite a
bit of discussion of prosody, although what Wasow does mention
disquiets me somewhat. In this respect, Wasow is a bit hasty in
dismissing the one definition of weight based on prosody: ''the
dislocated NP [in heavy-NP shift] is licensed if it contains at least
two phonological phrases'' (p. 16, quoting Zec & Inkelas 1990:377). As
a counterexample, Wasow offers the following (witnessed at a recent
meeting in Chicago of graduate-school deans): 'Can you clarify for me
this business of bringing into the degree requirement teaching?' He
adds that 'teaching' here, the lone word in the shifted nominal, ''was
destressed'' (p. 21). I'm not sure what Wasow means. Surely, if rising
intonation (characteristic of yes/no questions in General American
English) had been pronounced, it would be very difficult for this word
to be pronounced without some degree of prosodic prominence. Because
the other, syntactic definitions of weight also have counterexamples,
it would have been more fair for Wasow to have subjected this test, by
Zec & Inkelas (two graduates of Wasow's department, no less), to the
same level of empirical scrutiny. These quibbles aside, this book is
aimed at syntacticians and--ably--discusses primarily syntax. Yet
another niggling aspect of Wasow's book left me wanting. What is it
about the verb phrase that allows noncategorical factors to creep into
the ordering of constituents? Maybe the predicate must be uttered
before planning of subsequent elements' arrangement (p. 56). My own
work is on Tagalog (Austronesian), where the verb is generally initial
and all other elements are arranged by weight and discourse
factors. (This is the reason I became interested in Wasow's book.)
Kroeger (1993) argues for a flat, exocentric structure for VP (which
includes the subject and other NPs) as opposed to normal projection
for other categories in the clause. Although Kroeger's analysis seems
a bit ad hoc, at least it attempts to codify constituent-order freedom
formally. Wasow neglects this issue--aside from pointing out a few
constructions in chapter 1 (pp. 6-7, 9-11) in which some
otherwise-preverbal heavy elements can appear at the end of the
clause).

Closing comments

Wasow is careful to say that he is no Skinnerian (pp. xi,
esp. xiv). In a way, however, this book signifies a swing of the
pendulum toward empirical, scientific linguistics. In this connection,
Wasow's book has also shown me scientific linguistics. In this
connection, Wasow's book has also shown me how much of a philosopher's
role Chomsky has taken in the past half century, preparing the way for
the advent of true science in the study of language. In a sense, it
is futile to expect scientific rigor from Chomsky (no more than
Solzhenitsyn can be expected to show political savvy). Still, I cannot
accuse this work of being mere methodological moaning--as
(meta)bemoaned by Pullum 1983--because Wasow puts forth several
concrete studies as a way to strengthen _existing_ methodology. As a
generativist with longstanding and solid credentials, Wasow is in a
very good position to present this kind of proposal. Wasow and his
colleagues at Stanford and other universities in California have given
the Chomskyan movement a run for their money, mainly in their
theories' applicability to teaching computers to use language. Now we
see a chance for computational ideas like corpus searching (and other
statistical methods) to teach syntacticians to be scientists as well.

REFERENCES

Abney, Steven (1997) ''Statistical methods and linguistics.'' Judith
L. Klavans & Philip Resnik, eds. The balancing act: combining
symbolic and statistical approaches to language. Cambridge, MA: MIT
Press, 1-26.

Arnold, Jennifer, Thomas Wasow, Anthony Losongco, & Ryan Ginstrom
(2000) ''Heaviness vs. Newness: The effects of complexity and
information structure on constituent ordering.'' Language 76, 28-55.

Chomsky, Noam (1959) ''A review of Skinner's Verbal Behavior.''
Language 35, 26-58. (Prefaced and reprinted in Leon A. Jakobovits &
Murray S. Miron, eds. Readings in the psychology of
language. Prentice-Hall, 1967, 142-143.)

Chomsky, Noam (1975) The logical structure of linguistic
theory. Chicago: University of Chicago Press.

Cowart, Wayne (1997) Experimental syntax: applying objective methods
to sentence judgments. Thousand Oaks, CA: Sage Publications.

Emonds, Joseph (1976) A transformational approach to English
syntax. New York: Academic Press.

Erteschik-Shir, Nomi (1979) ''Discourse constraints on dative
movement.'' Talmy Giv�n, ed. Discourse and syntax. (= Syntax
and Semantics, 12.) New York: Academic Press, 441-467.

Ferreira, Victor S., & Gary S. Dell (2000) ''The effect of ambiguity
and lexical availability on syntactic and lexical production.''
Cognitive Psychology 40, 296-340.

Giv�n, Talmy. (1983) ''Topic continuity in discourse: an
introduction.'' Talmy Giv�n, ed. Topic continuity in
discourse: a quantitative cross- language study. Amsterdam: John
Benjamins, 1-41.

Gundel, Jeanette K. (1988) ''Universals of topic-comment structure.''
Michael Hammond et al, eds. Studies in syntactic typology. Amsterdam:
John Benjamins, 209-239.

Hawkins, John A. (2001) ''Why are categories adjacent?'' Journal of
Linguistics 37, 1-34.

Kroeger, Paul (1993) Phrase structure and grammatical relations in
Tagalog. Stanford, CA: CSLI Publications.

Larson, Richard (1988) ''On the double object construction.''
Linguistic Inquiry 19, 335-391.

Manning, Christopher D., & Hinrich Sch�tze (1999) Foundations
of statistical natural language processing. Cambridge, MA: MIT Press.

Pullum, Geoffrey K. (1983) ''The revenge of the methodological
moaners.'' Natural language and Linguistic Theory 4,
583-588. (Reprinted in 1991: Geoffrey K. Pullum. The great Eskimo
vocabulary hoax and other irreverent essays on the study of
language. Chicago: University of Chicago Press.)

Quirk, Randolph, Sidney Greenbaum, Geoffrey Leech, & Jan Svartvik
(1972) A grammar of contemporary English. London: Longman.

Ross, John Robert (1967) Constraints on variables in syntax. Doctoral
dissertation, Massachusetts Institute of Technology. Sch�tze,
Carson T. (1996) The empirical base of linguistics. Chicago:
University of Chicago Press.

Skinner, B.F. (1957) Verbal behavior. New
York. Appleton-Century-Crofts. (Reprinted in 1992 and 2002,
B.F. Skinner Foundation.)

Snedeker, Jesse & John Trueswell (2003) ''Using prosody to avoid
ambiguity: effects of speaker awareness and referential context.''
Journal of Memory and Language 48, 103-130.

Sorace, Antonella, & Frank Keller (to appear) ''Gradience in
linguistic data.'' Lingua.

Stallings, Lynne M., Maryellen C. MacDonald, & Padraig G. O'Seaghdha
(1998) ''Phrasal ordering constraints in sentence production: phrase
length and verb disposition in heavy-NP shift.'' Journal of Memory and
Language 39, 392-417.

Wasow, Thomas (1972) Anaphoric relations in English. Doctoral
dissertation, Massachusetts Institute of Technology.

Wasow, Thomas (1997) ''End-weight from the speaker's perspective.''
Journal of Psycholinguistic Research 26, 347-361.

Zec, Draga, & Sharon Inkelas (1990) ''Prosodically constrained
syntax.'' Sharon Inkelas & Draga Zec, eds. The phonology-syntax
connection. Stanford, CA: CSLI Publications, 365-378.

ACKNOWLEDGEMENTS

As a linguist who does not work very much with statistics or at all on
processing, I have relied on the following kind colleagues: Hsi-chin
Janet Chu (National Taiwan Normal Univerisity), T. Florian Jaeger
(Stanford University), and Irina A. Sekerina (City University of New
York); although their comments have been extremely helpful, only I am
to blame for any lapses in interpreting their sage advice or for any
other errors that remain in this review. I also contacted the book's
author, who was kind enough to confirm the missing references. 

ABOUT THE REVIEWER

Loren A. Billings (Ph.D., Princeton, 1995) is associate professor of
linguistics at the Department of Foreign Languages and Literature,
National Chi Nan University, Taiwan. Concentrating on the Slavic
(Indo-European) and Central Philippine (Austronesian) languages, he
specializes in the interfaces of syntax, morphology, and
phonology. His publications also look at the interactions of animacy
and referentiality on constituent order.
Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue