In this book, Stroik and Putnam take on Turing's challenge. They argue that the narrow syntax – the lexicon, the Numeration, and the computational system – must reside, for reasons of conceptual necessity, within the performance systems.
AUTHOR: Wulff, Stefanie TITLE: Rethinking Idiomaticity SUBTITLE: A Usage-based Approach SERIES TITLE: Corpus and Discourse PUBLISHER: Continuum YEAR: 2008
Benet Vincent, PhD Student, Department of English, University of Birmingham, UK
This monograph, developed from Wulff's PhD thesis, investigates idiomaticity in an attempt to discover what information users of English draw on to decide that a particular phrase is more or less idiomatic or idiom-like. Taking 39 examples of the Verb Noun Phrase (V NP)-construction, Wulff combines judgment data and corpus investigation, allied with statistical procedures, to study the variables seen to influence our judgments of idiomaticity, namely compositionality and flexibility. Innovative methodological approaches for investigating these variables are introduced and described in detail and conclusions are drawn regarding the results of the enquiry and the effectiveness of these methods judged according to cognitive linguistic criteria. Wulff also suggests further areas of research based on her findings using the methodology introduced.
The book starts by setting out Wulff's main question -- How it is that speakers judge a phrase such as ''take the plunge'' to be more idiomatic than one like ''write a letter''. Idiomaticity is seen as a psychological characteristic that a particular phrase or construction has, to a greater or lesser extent, which is therefore open to judgment by speakers of a language. While Wulff stresses that idiomaticity, as a psychological construct, cannot be directly observed in a corpus, she points out that it may be possible to pin down this complex concept through corpus-based research. This is because, according to the usage-based grammar perspective adopted in this study, language users base their linguistic behaviour on distributional information derived from their language environment. Therefore, it is possible to use the distributions of constructions found in a balanced corpus, such as the British National Corpus (BNC), to postulate which factors are most salient when idiomaticity judgments are made by speakers of British English. The usage-based approach also allows Wulff to conceive of two continuums: the unobservable 'idiomaticity continuum', on which speakers place each exemplar of a construction according to their perception of its distributional characteristics; and the observable 'idiomaticity variation continuum', on which each construction lies on by virtue of its actual behaviour.
Chapter 1 -- 'Theoretical Issues' -- establishes the main theoretical underpinning of the work by reviewing previous research indicating that a comprehensive view of idiomaticity needs to take into account factors beyond mere non-compositionality, to include various types of flexibility as well. Wulff shows how discourse analysis, phraseology and psycholinguistics have all contributed to an understanding of idiomaticity as a scalar, multifactorial concept. She also points out the importance of this view of idiomaticity in cognitive linguistics, and more specifically, construction grammar (Goldberg, 2006). In common with other constructions, the storage and recognition of idioms depends on interplay between compositionality, conventionality, schematicity and frequency. However, their low degree of schematicity, that is, the fact that they do not show a great deal of lexical variation, in contrast to many other constructions, makes idioms good candidates for corpus study, since they are easier to identify. This explains Wulff's decision to focus on 'fully lexically specified V NP phrases like ''make a point'' or ''take the plunge''' (p. 18).
Chapter 2 is entitled 'Methodological Issues'. The first methodological issue referred to is Wulff's advocacy of the methods of quantitative corpus linguistics over 'traditional corpus linguistic methodology' (p. 21). The main argument presented is that the methods of quantitative corpus linguistics offer the best ways of investigating hypotheses, particularly those deriving from cognitive linguistics, because of the close match between the methods and the theories. This does not, however, prevent Wulff from following Gries (2008) in pointing to similarities between the products of traditional corpus linguistics, such as such as Pattern Grammar (Hunston & Francis, 1999), and those of cognitive linguistics, such as Goldbergian construction grammar.
The next methodological issue relates to the data sample, a selection of 39 V NP-constructions, 33 of which are the most frequently occurring V NP-constructions listed in the Collins Cobuild Dictionary of Idioms (2002), and 6 further constructions selected at random from the most frequently occurring V NP-constructions in the BNC.
The chapter ends with a description of the idiomaticity judgment experiment which Wulff carried out, in which asked the informants -- first year English linguistics undergraduates at the University of Sheffield -- to judge the constructions in terms of relative idiomaticity by looking at the constructions in sentences designed to reflect each construction's particular behaviour. Wulff demonstrates how the results of this experiment rank the constructions as might be expected from the literature: the constructions with the lowest rankings, such as 'write X letter' and 'tell X story', are those not included in the Dictionary of Idioms, and higher rankings tend to be given to less compositional or literal and more metaphorical and opaque constructions, with groupings approximately following those identified elsewhere (e.g. Cacciari and Glucksberg, 1991). She also points out that the gradual increase in judgment values, with none of the constructions being given extreme idiomaticity scores, is consistent with the lack of truly 'non-decomposable constructions like ''chew the fat''' (p.30, p.32), which are too rare to occur frequently in the BNC and so are not included in the study. Finally, and significantly, the rankings given across informants were found to be very consistent.
Chapter 3 -- 'Compositionality' -- moves on to a discussion of the concept of compositionality and how a corpus might be used to measure it. The definition of compositionality given (i.e. that 'compositionality is a function of the semantic similarity of the constituent words and the phrasal expression' (Wulff 2009: 135)) underpins the method, that is, comparing the numbers of collocates a construction has in common with its component words; an important innovation being that both constituents of the V NP-phrase are taken into account. Wulff presents two ways of measuring the compositionality of V NP-constructions: 'weight' compares the number of collocations each element has amongst the total collocations of the construction; 'share' determines the number of construction collocates as a proportion of collocates of the constituent verb and noun. This second measurement is found to distinguish more effectively between intuitively compositional and non-compositional items, with some provisos.
Chapter 4 -- 'Flexibility Measures' -- moves on to consider how flexibility can be quantified and the relationship between such measures and idiomaticity. Idioms are traditionally considered to be relatively fixed in their behaviour, not undergoing certain syntactic, morphological and lexical changes, and Wulff presents a systematic attempt to measure this fixedness. The chapter focuses on three main types of flexibility: tree-syntactic flexibility, or the tendency of a construction to occur in different types of syntactic patterns; lexico-syntactic flexibility, the extent to which a phrase allows external or internal modification; and morphological flexibility, relating to inflections as well as variations in the determiner slot of a construction. Each of these types of flexibility is further divided into an exhaustive list of 'flexibility parameters' and 'parameter levels': to give an example, 'Tense' is a flexibility parameter and it has the parameter levels 'Past/Present/Future/Nonfinite'.
Two different measures are introduced, both based on the flexibility seen in a representative sample of V NP-constructions derived from the International Corpus of English. One of these, based on Barkema (1994), measures flexibility as a function of difference from the baseline measure seen in the sample, while the other aims to give a relative flexibility figure, or 'entropy', for each parameter compared to total flexibility, that is, all possibilities happening equally often. This way of calculating entropy arguably does not give a reliable measure for lexico-syntactic flexibility, which should be shown as higher when a feature, such as an adjectival modifier, is often present, rather than when it is as likely to be present as absent. Therefore, Wulff adapts the entropy measure to indicate that a construction frequently occurs with extra lexical material and renames it 'directional entropy'. Wulff concedes there is a small problem here in that in some cases, such as 'break X ground', V NP constructions are modified almost invariably -- in this case by 'new' -- which does not indicate flexibility, but rather fixedness, and further suggests that 'break X ground' is not really a V NP construction, but more likely a V ADJ NP construction.
Detailed results are reported for these measures with the aid of graphs; correlations between the various variation parameters and parameter levels and measures of compositionality are also provided. This means that different perspectives are brought to bear on the relative flexibility of each V-NP construction in the study.
Chapter 5 -- 'The Idiomatic Variation Continuum' -- presents a Principal Component Analysis (PCA) of the data, which indicates groups of variables that are related to each other, taking into account the flexibility parameters as well as corpus frequency and compositionality scores. This analysis reduces all 20 variables to a total of 8 components listed in order of importance: component 1 consists of tree-syntactic flexibility and Voice, with contributions from Det (determiners), Person, and AttrNP (presence of a pre-modifying noun); component 2 consists of NumV (whether the verb is singular, plural or non-finite) and Mood; component 3 contains Addition and NoAdv (presence of adverbials); component 4 includes compositionality and corpus frequency; component 5 consists of the parameter relating to Gerund formation; component 6 is that relating to prepositional phrases, with contributions from Tense and NumNP (the number of the noun phrase in the V-NP construction); component 7 consists of AttrAdj (the presence/absence of attributive adjectives) with a contribution from Aspect; component 8 contains the flexibility parameter Negative. The first 4 of these components are found to account for more than half the variance in the data. Two of these parameters, verb number and verbal mood, have escaped attention in the past and hence represent new findings. Another interesting, though tentative finding, regards the lower than expected importance of compositionality.
In Chapter 6 -- 'The Idiomaticity Continuum' -- Wulff moves on to discuss the correspondence between the corpus-derived findings and the idiomaticity judgments. The results of a Multiple Regression Analysis (MRA) show a high overlap between those parameters that are found to be important and those which the components analysis in Chapter 5 revealed as relevant. This strongly suggests that there is a correlation between idiomaticity as a psychological construct and a construction's location on the idiomaticity variation continuum.
The final chapter -- 'Towards a New Model of Idiomaticity' -- concludes by reviewing the contributions made by the study and suggesting further applications of the methodology, for example, to assess the degree of semantic bleaching of verbs, that is, whether they are losing lexical meaning and becoming more like function words. The results are also seen as providing further support for a performance-based approach to language and one in which grammatical categories emerge from the data. This leads into a discussion of how the findings relating to idiomatic variation might be integrated into a model of the constructicon, the hypothesized storage and retrieval facility for constructions in the mind. Wulff proposes a new axis of idiomaticity which cuts across the constructicon at the level of lexically specific complex constructions and ranges between low and high idiomaticity. This axis allows for the inclusion of information regarding the variables that may be important for judging the idiomaticity of a particular construction as part of the construction and is thus an important addition to the model.
'Rethinking Idiomaticity' is a rigorous investigation of the relationship between idiomaticity and performance data which speakers may draw on in making idiomaticity judgments. In focusing on a relatively small data set, which is then extensively analysed, Wulff is able to obtain some interesting results. A number of innovative methods for measuring variables such as compositionality, are presented, which will no doubt inform future research. This work also makes some persuasive arguments for factors that influence idiomaticity judgments which might be tested on different data sets.
Naturally, for the curious and critical reader, some questions remain. Perhaps the most important of these relate to the definition of 'construction', and its relationship to 'idiom', as applied to individual V NP-constructions, and what exactly the idiomaticity informants were being asked to judge. The fact that several of the constructions included in this study, such as 'hold X breath', 'play X game' and 'draw X line' are polysemous is not really discussed, so it is not clear whether Wulff distinguished between them or not. However, corpus linguistic research such as Sinclair (1991), Hunston & Francis (1999) and Hoey (2005) has consistently found that a change in meaning accompanies change in grammatical behaviour, suggesting that a lack of distinction between different meanings may well affect results. Moreover, to take the example of the construction 'hold X breath', the example sentence provided to the informants, 'She held her breath', appears indeterminate between metaphorical and literal interpretations. It seems likely that one's interpretation of this sentence will influence one's idiomaticity judgment, having some effect on the results, but this possibility is not mentioned. Thus, basing findings, as seems to have happened, on idiom judgments derived from sentences which are either literal or metaphorical, or ultimately ambiguous, without further asking informants how they made their choices, seems to raise some questions about such findings.
A note should also be made regarding editorial control over this book. It is disappointing to find numerous deficiencies, both in terms of editing and proofreading, especially bearing in mind that books of this kind are invariably expensive. On several occasions, for example, numbers stated in the text do not match those presented in figures or tables, and a number of typographical mistakes remain.
In terms of audience, this book, with its high emphasis on statistical approaches and its basis in cognitive linguistics, will be of interest to practitioners of quantitative corpus linguistics, particularly readers of the Journal of Corpus Linguistics and Linguistic Theory. However, despite Wulff's generally clear descriptions and rationales, the large number of descriptions of statistical procedures may be off-putting for readers outside this area. Readers not used to this field may also baulk at Wulff's almost exclusive focus on abstractions at the expense of examples drawn from the corpus.
Barkema, Henk. 1994. Determining the syntactic flexibility of idioms. In Udo Fries, Gunnel Tottie & Peter Schneider (eds), Creating and using English language corpora, 39-52. Amsterdam: Rodopi.
Cacciari, Cristina & Sam Glucksberg. 1991. Understanding idiomatic expressions: the contributions of word meanings. In Greg B. Simpson (ed), Understanding word and sentence, 217-40. The Hague: North Holland.
Goldberg, Adele. 2006. Constructions at work. Oxford: Oxford University Press.
Gries, Stefan Th. 2008. Phraseology and linguistic theory: a brief survey. In Sylviane Granger & Fanny Meunier (eds), Phraseology: An interdisciplinary perspective, 3-25. Amsterdam/Philadelphia: John Benjamins.
Hunston, Susan & Gill Francis. 1999. Pattern grammar. Amsterdam/Philadelphia: John Benjamins.
Sinclair, John. 1991. Corpus, concordance, collocation. Oxford: Oxford University Press.
Wulff, Stefanie. 2009. Converging evidence from corpus and experimental data to capture idiomaticity. Corpus Linguistics and Linguistic Theory 5(1). 131-159.
ABOUT THE REVIEWER
ABOUT THE REVIEWER:
Benet Vincent is PhD student in the Department of English at the University
of Birmingham. He is investigating the association between modal meaning
and the V wh pattern using the BNC. He formerly worked for 15 years in the
field of English language teaching, latterly at Sabanci University, in
Istanbul. His research interests include phraseology, corpus linguistics,
discourse analysis, writing in a foreign language and language learning.