LINGUIST List 23.853

Mon Feb 20 2012

Review: Syntax; Corpus Linguistics: Mindt (2011)

Editor for this issue: Joseph Salmons <jsalmonslinguistlist.org>



Date: 20-Feb-2012
From: Ernest Hogue <eahoguegmail.com>
Subject: Adjective Complementation
E-mail this message to a friend

Discuss this message

Announced at http://linguistlist.org/issues/22/22-2290.html

AUTHOR: Ilka MindtTITLE: Adjective ComplementationSUBTITLE: An empirical analysis of adjectives followed by 'that'-clausesSERIES TITLE: Studies in Corpus Linguistics 42PUBLISHER: John BenjaminsYEAR: 2011

E. Alan Hogue, Department of Linguistics, University of Arizona, Tucson, AZ

SUMMARYThe volume under review investigates sentences such as those in (1) and (2) (allexamples from the British National Corpus). In each case, an adjective ("sure"in (1) and "likely" in (2)) is followed by a that-clause.

(1) I'm sure that he'd never make the same mistake again.(2) It is also likely that power and water supplies would be disrupted.

The primary purpose is to compare previous descriptive treatments of adjective +that-clauses, particularly those found in Quirk et al. (1985), Biber et al.(1999), and Huddleston & Pullum (2002), to the results of Mindt's corpus analyses.

Mindt identifies a number of questions of relevance to previous treatments: theclassification of such adjectives according to matrix subjects or objects;whether that-clauses which provide an explanation, like (3), should beconsidered along with that-clauses which provide a result, (4); whether thesemantics of these adjectives has any relation to the verb phrase in thethat-clause; and, finally, whether there are possible triggers for that, as in(5), or zero-that, (6), in such clauses.

(3) Slowly she became aware that it wasn't quiet at all.(4) The answer to that is so obvious that I shan't bother to give it.

(5) But it is possible that the details took time to be formulated.(6) It's possible he's in trouble.

Methodologically, the study is corpus-driven, therefore radically inductivecompared to most other linguistic approaches (e.g., Biber 2010). A bare minimumof theory and structure is assumed beforehand, and to the extent possible anyadditional theoretical apparatus is directly motivated by analyses of the data.Practically, this means that the target of investigation is defined strictly interms of linear order of syntactic categories, and competing analyses arelargely judged by their ability to cover all cases of this pattern found in thedata. As will be seen, however, Mindt does argue that some apparent cases of theadjective + that-clause pattern do need to be considered separately.

I will begin by outlining M's major claims before giving details of theanalyses. M notes that there are several apparent types of the linear adjective+ that-clause pattern which are found in corpora but which are not consideredtogether in previous studies: the "so…that" construction(7), which Mindt dubs a resultative construction, cases with direct objects inthe matrix clause (8), and cases in which certain adjectives are preceded by theverb MAKE (9).

(7) Air is so vital that it is only possible to live for a few minutes without it.(8) I believe it possible he may be an agent for one of your creditors.(9) And she was going to make sure nothing stood in her way!

As this is a very inductive, corpus-driven study, M begins with the assumptionthat all observed instances of the linear pattern adjective + that-clause areproperly classified together and hence should be studied together. The data isthen analyzed for evidence of differences between each of these possiblesub-types, the assumption being that if these are all indeed properly distinctpatterns, we should find they behave differently in the corpus data. M concludesthat (9) is indeed properly considered a different formal pattern and shouldtherefore be excluded from the study of adjective + that-clauses. She dubs thispattern "verb MAKE + adjective 'certain', 'clear', 'sure'" and, in Chapter 4,argues convincingly that these are verb-adjective combinations and therefore nottrue examples of the adjective + that-clause pattern.

However, analyses provided in Chapters 3, 5 and 6 lead M to argue that "so…that"constructions such as (7) should be treated together with canonical adjective +that-clause constructions like (1) and (2). For M, the construction previouslystudied as adjective complementation is more precisely considered to be an"explanative" construction. This is in distinction to "so…that" constructions,which she calls "resultative". The difference between these constructions issemantic: a that-clause in a resultative indicates the result of the adjectivewhile a that-clause in an explanative provides an explanation, but otherwisethere seem to be few important formal differences between them. M concludes thatthe explanative and resultative constructions are two sub-types of the generalformal linear pattern adjective + that-clause, and therefore should be studiedtogether.

M confirms the standard classification of adjectives into experiential andevaluative using hierarchical cluster analysis. However, M argues againstextraposition analyses for cases such as those with impersonal "it".

After a few introductory chapters, Chapter 3 sets out to determine whether theadjectives which occur with that-clauses can be classified based on the subjectswith which they appear, and also to what degree previous classifications can becorroborated based on such an analysis. In particular, the author seeks to testtwo common analyses. First, she wants to show that resultatives should beconsidered together with explanatives as examples of adjective complementation.

Second, M is highly skeptical of subject extraposition analyses. Subjectextraposition is the traditional name for the notion that (10) is a marked,non-canonical variant of the unmarked, canonical structure in (11).

(10) It's true that I don't fancy John any more.(11) That I don't fancy John any more is true.

Subjects are analyzed on the basis of three criteria: matrix vs. embeddedsubject, personal pronoun vs. non-pronoun subject, and intentional vs.non-intentional subject. Once each subject is analyzed by these criteria, thedata are subjected to hierarchical cluster analysis. The results are veryrobust, indicating two semantic classes of adjectives: those expressingexperiences and those expressing evaluations. Not surprisingly, the former occurwith intentional subjects and the latter with non-intentional subjects. Whenresultatives are analyzed separately from evaluatives, the same adjectivescluster into the same groups. Thus their pattern of co-occurrence with subjectsare the same, although they are different both formally and functionally. As forextraposition, an analysis in which impersonal "it" (i.e. an extraposed subject)is separated from other pronouns produces essentially the same results. On thisbasis the author suggests that these sentences should be categorized based onlinear form rather than a supposed derivational or quasi-derivational relationship.

In Chapter 4, verb-adjective combinations are examined. Of all the observedco-occurrences of verbs and adjectives, the author singles out MAKE + 'certain','sure', and 'clear' as possible multi-word verbs. As before, the status of thesecollocations is presented as a way to interrogate another derivational analysis:here, object extraposition. Both (12) and (13) are commonly regarded as theresult of object extraposition, in which the direct object of the verb is movedto the end of the sentence, sometimes leaving behind a substitute "it" as in(13), and sometimes not as in (12).

(12) The statement makes clear that sanctions will be used only in the last resort. MAKE + "clear"(13) Officials make it clear that the actual cost is not yet known. MAKE + D.O. + "clear"

This analysis leads us to consider such sentences as essentially identical, withonly the minor difference of the presence or suppression of an overt pronoun todistinguish them. Under this analysis, therefore, it would be difficult to arguethat "makes clear" in (12) is a multi-word verb (or phrasal verb, verb-adjectivecombination, etc.), because it would be hard to explain why such a multi-wordverb can be broken up by direct objects in closely related constructions.However, if it can be shown that examples like (12) and (13) are in fact quitedifferent from each other, then such an analysis becomes feasible, while anobject extraposition analysis is placed in some doubt.

The author adopts five criteria for comparing these structures: (a) theirfrequencies, (b) the types of subjects/objects each tends to co-occur with, (c)how often adverbs are found between the verb and the adjective, (d) anydifferences in meaning, (e) complementations other than that-clauses which eachcan take.

The results suggest that MAKE + "certain" and MAKE + "sure" are multi-wordverbs, and that at least in these cases they are not closely (derivationally)related to MAKE + D.O. + "certain"/"sure". In each case, the meanings are quitedifferent, and adverbs are rarely found between the verb and adjective (whilethey occur frequently in corresponding cases with direct objects). MAKE +'clear', however, differs little from MAKE + D.O. + 'clear', consistent with thehypothesis that these are in fact closely related, and that 'make clear' is nota multi-word verb.

Chapter 5 is very similar to Chapter 3, but focuses on the relationship betweenadjectives and matrix objects. M concludes that in sentences with both a subjectand an object, it is the object type which determines the type of adjectiveused, not the subject. She again makes a similar argument against objectextraposition.

Chapter 6 returns to the question of whether explanative and resultativeconstructions should be considered separately, and to what degree they aresimilar. Here, the question is approached through analyses of the adverbs whichprecede adjectives. While Mindt admits that these constructions have quitedifferent communicative functions and convey quite different meanings, heranalysis shows that the two behave very similarly in formal terms (by which,recall, she means the linear arrangement of syntactic categories). On thisbasis, she argues that the constructions should be considered semantic variantsof an overarching adjective + that-clause category.

Chapter 7 compares cases with an overt "that" to cases with "zero-that", withthe aim of finding any contextual trigger which might increase the likelihood ofone or the other. She considers six candidates: (a) the adjective preceding, (b)the medium and genre, (c) the matrix subject, (d) the subject in thethat-clause, (e) co-reference between the subjects in each clause, and (f)intervening elements between the adjective and subject. She finds that no onecriterion has an impressive effect on the (non-)realization of "that". However,some fairly weak tendencies were found; for instance, evaluative adjectives,more formal speech, and non-intentional matrix subjects each somewhat favorovert that.

Chapter 8 examines verb phrases in the that-clause and attempts to find arelation between them and the semantics of the preceding adjectives. M firstattempts to find such a pattern in the data without any prior categorization ofadjectives. She then looks for relations between adjectives and modal verbs andsubjunctives. Finally, she looks for any relationship between VP structure usingthe distinction between experiential vs. evaluative adjectives. She concludesthat there is no relationship to be found, which contradicts a number ofprevious claims, especially those of Quirk et al. (1985), who attempted tosemantically categorize adjectives based on co-occurring VPs in adjective +that-clause constructions.

EVALUATIONM's study offers an interesting example of the pros and cons of studyinglanguage with a highly inductive approach. Such research can and does have asalutary effect, as when M provides evidence which calls into question therelevance of subject/object extraposition analyses in cases with impersonal"it", an analysis that is so old and familiar that it is often simply assumed.On the other hand, the evidence provided against movement analyses have acircumstantial flavor, because she does not provide any principled basis uponwhich to evaluate frequency evidence, and probably there is no such basis. Forinstance, M shows that the presumed ''canonical'' or non-extraposed orders in somecases appear very infrequently, if at all, in the corpus. The assumption here isthat non-extraposed sentences should be more frequent than sentences resultingfrom movement, or at least that they should not be rare. On this assumption thefrequency data look quite convincing: the extraposition analysis in these casesmakes the wrong prediction. This may have worried more deductively inclinedlinguists in the days of Chomskyan deep and surface structure, and in particularin the short-lived framework of generative semantics, but the more common moderntheories of grammar do not make use of deep and surface structures, and there isno reason, in principle, why sentences with extraposition should be less commonthan sentences without it. Depending on the framework, movement analyses ingeneral have become more or less metaphorical. Even in principles and parameterstheory, in which some element may be said to have moved, this no longer impliesthat an extraposed sentence has undergone a derivation from an otherwisewell-formed base sentence. And even if extraposition sentences are derived frommore basic non-extraposed sentences, it's still questionable whether we shouldexpect to find more derivationally simple sentences in corpora. What peopleactually speak and write is influenced by many extra-grammatical factors, notleast sociocultural ones. This is not to suggest, as some do, that corpusstudies are useless. But when corpus data is used to infer the nature of grammaritself, questions of this kind inevitably arise. I would have liked to see morediscussion of this.

Something underlying this objection for many linguists, I suspect, would be thatextraposition has turned out to be a highly useful theoretical apparatus, onewhich unites a good deal of otherwise disparate phenomena cross-linguistically.This is perhaps where the difference between inductive and deductive approachesis most stark, and perhaps unbridgeable. Those working the Chomskyan vein tendto judge their theories based on theory-internal factors such as elegance andexplanatory power, while they tend to downplay data from actual language use.Mindt tends to judge theories based on their ability to cover all (or thegreatest part) of her corpus data, employing a minimum of abstraction. Theformer approach can cover a great deal of data (as it is defined), but runs therisk of providing analyses which are based on faulty and sometimes untestableassumptions. Mindt's approach minimizes the risk of untestable assumptions(beyond what is absolutely necessary -- for instance that there are verbs andnouns), but suffers from a necessarily very narrow scope which may be in dangerof missing important generalizations.

However, although more theoretically-oriented linguists may dismiss thesearguments, based as they are on frequencies, co-occurrence relations, and thelike, more empirical evidence is always better than less, and even if M does notclose the book on extraposition in these constructions, her analysis gives goodreason to revisit this and other assumptions. This skepticism towardstheoretical assumptions demanded by the corpus-driven approach can lead touseful reevaluations of what may seem obvious truths.

While this book should be of interest to anyone studying adjectivecomplementation in English, it would probably be of little value to thosestudying other constructions or languages. With virtually no theoreticalapparatus it is hard to see how the insights revealed here can be related toother linguistic phenomena. This, however, is not meant as criticism of thisstudy in particular, which sets out to do precisely what it does, and does itvery thoroughly and methodically.

REFERENCESBiber, Douglas. 2010. Corpus-based and corpus-driven analyses of languagevariation and use. In The Oxford Handbook of Linguistic Analysis, Bernd Heine &Heiko Narrog (eds), 159-191. Oxford: OUP.

Biber, Douglas, Johansson, Stig, Leech, Geoffrey, Conrad, Susan & Finegan,Edward. 1999. Longman Grammar of Spoken and Written English. London: Longman.

Huddleston, Rodney & Pullum, Geoffrey K. 2002. The Cambridge Grammar of theEnglish Language. Cambridge: CUP.

Quirk, Randolph, Greenbaum, Sidney, Leech, Geoffrey, Svartik, Jan. 1985. AComprehensive Grammar of the English Language. London: Longman.

ABOUT THE REVIEWERAlan Hogue is a graduate student in linguistics at the University ofArizona, with primary research interests in syntax, sentence processing andcorpus linguistics. He is keenly interested in the philosophy andmethodology of linguistic theories, as well as evolutionary theory, formallanguage theory, and computability theory. He currently uses a combinationof corpus-based and experimental methods to study the effect of structuralpriming on elided structures.

Page Updated: 20-Feb-2012