Review of: Bayley, R. & Preston, D.R. (1996), Second Language Acquisition and Linguistic Variation, John Benjamins
|In EUROSLA Clarion|
You’ve played Zork. You’ve despaired in the Dungeons of Doom. You’ve marched with the Mutant Mules of Alpha Centauri. You’ve solved the secrets of the Stone World, the Water World, and the Tree World. Are there any computer frontiers left to conquer?
Here comes the blockbuster of them all — VARBRUL! Bigger than an exploding ANOVA, faster than a multiple regression, > 2 outrunning c2. CHECKTOK your tokens, CROSSTAB your cells, TVARB your results, TSORT, MAKECELL and finally MVARB. Feed in the figures and the solutions pop out. Want to know if L2 transfer declines as L2 development progresses? Feed in some figures from Portuguese-speaking learners and the answer is plain — yes! Want to know if phonological and grammatical processes are distinct in L2 learners’ knowledge of the English inflectional "-ed"? Feed in some data on Chinese learners of English and the answer’s there —yes! Want to see if Alberto really was a wimp? Feed in the data, turn on the logistic regression method, and out comes the answer—his negation improved over time! And it comes with a 54 page set of handy hints for the user. Transform your research! Let the computer take the load!
Or at least this is the impression left by much of this book—a paean of praise for the VARBRUL program, a statistical package devised to handle the messy facts of sociolinguistic variation, used by most of the contributors here to handle their quantitative data. Everyone gets attached to one program or another. My students mutter SPSS as if that in itself magically showed their research was valid; my own current data goes into ERTSCODE. Statistical packages it seems to me are as good as the statistical approaches they incorporate, whether ANOVAs, regressions or whatever. It is the underlying understanding of what a statistic is and when it is appropriate to use it that matters, not whether it is in a package. This reviewwould not have existed in this form but for the spelling checker, but that doesn’t mean the responsibility for the spelling isn’t ultimately mine. Perhaps someone should put together a collection on how to use Word for Windows for second language acquisition research—the usefulness of Find as a way of hunting for examples for instance. Much of this volume consists of analyses based on VARBRUL; one chapter is indeed a step-by-step handbook for using VARBRUL. For the convert this may be of some use if they can take in such formulas as:
p= po x ... x pn [po ... pn] + [(1-po) x ... x (1-pn)]
repeated twice in this book. But one suspects that written instructions for a computer program are only of use when used alongside the program to solve some dummy problem.
The main gap for the non-convert is an outline of the reasons for preferring VARBRULover other packages. If its core is indeed regression analysis, then why not use any package that includes this? Indeed Berdan’s contribution uses ‘logistic regression’ from SPSS which is ‘the more generalised procedure implemented in the VARBRUL and GOLDVARB analyses...’ Though users of Macintosh version of VARBRUL tell me of its user-friendliness, this is far from the impression left on the reader. It is not clear why researchers who are engaged in a range of research activities should not instead get to know a basic stats package that happens to include regression analysis. If VARBRUL is indeed the vibrant heart of variation research, its advantages over other statistical approaches need to be made clear. Philip Scholfield’s recent book Quantifying Language manages without VARBRUL, or indeed regression analysis. Perhaps sociolinguistics puts unique demands upon a statistics package but outsiders need to have this explained to them in greater detail.
So what does the term ‘variation’ mean in this book? We are firmly told by Preston that it is not merely ‘socially sensitive pragmatics’ and he takes various SLA researchers to task for their misinterpretation of the concept. The book is within the Labovian tradition of looking at differences in form relate to social and linguistic factors. Types of variation dealt with here are:
- variation in VOT forEnglish /p t ¶ q / by Italian learners (Flege et al), showing some effects of age of arrival in Canada but with ‘70% of variance unaccounted for’.
- variation inEnglish consonant clusters by Portuguese-speaking learners, showing interaction of universal hierarchies with L1 transfer (Major).
- variation inEnglish /t/ /d/ past tense allomorphs by Chinese learners scored according to grammatical category, syllable structure and assimilation (Bayley).
- variation inEnglish past tense marking by Spanish learners, showing multiple constraints by phonological environment and discourse (Adamson et al).
- variation in English article use by Czech or Slovak learners, showing the importance of form-function relations (Young).
- variation in ne deletion in French by Irish learners, showing dramatic increase after a period in France (Regan).
- variation ofEnglish negation in the speech of one Spanish learner, showing gradual approximation to native use over time (Berdan).
- variation inEnglish /r/ in Puerto Rican learners of English, based upon their perceptual systems (Labov).
In the polemical first chapter, which ‘sets the tone for the book’ (p.xiv), Preston accuses many researchers of taking the easy way out by investigating social sensitivity and avoiding ‘the hard stuff of phonology, morphology, syntax and semantics’. Of the eight articles outlined above, 4 are aboutEnglish phonology, 3 about English grammatical morphemes, one about French grammatical morphemes. In other words the bulk concerns certain familiar aspects of phonology and grammatical morphemes in English. To outsiders it has seemed that sociolinguists have often become obsessed with certain phonological issues and grammatical inflections rather than ‘the hard stuff’’ of syntax, the lexicon, discourse and semantics; the balance here does not reassure one. Nor is the wealth of variation in language served by treating only English, with the exception of Regan, with all contributors emanating from universities in the USA apart from one in Ireland and one in Canada.
It is indeed time thatsecond language acquisition research approached language variation seriously across a range of linguistic topics and a range of languages. This book touches on a small fragment of what could be possible, limited by a partisan approach wedded to a particular technique. It would have been useful to have had a book that gave a clear introduction to VARBRUL and a rationale for its use; indeed some teachers of sociolinguistics have already produced teaching treatments of VARBRUL for their students and at least one is rumoured to be heading for publication. It would also have been useful to have a volume that covered a spectrum of variational studies and methodologies used for second language acquisition research. Most of the contributors, however, are following well-known tracks, mostly already known from their other publications; the papers could well have been published in journals rather than in a collection. The value-added features are the fiery introductory chapter by Preston, the technical user’s guide to VARBRUL, and the contribution by William Labov, which does not have the kind of power normally associated with his writing. Potential purchasers will have to decide whether it is worth paying the kind of hard-back prices we are now used to from Dutch linguistics publishers to obtain these benefits.
Scholfield, P. (1996), Quantifying Language, Multingual Matters, Clevedon