Face to Face With the Ghost in the Machine: Psycholinguistics and TESOL

There is some lack of consensus as to how far the scope of the field extends, but this special issue of TESOL Quarterly focuses on its most central concerns, namely, how language is acqu

Face to Face With the Ghost in the

Machine: Psycholinguistics and TESOL

JOHN FIELD

University of Reading

Reading, England

WHAT DOES PSYCHOLINGUISTICS TELL US?

Psycholinguistics is the study of how the mind handles language There is some lack of consensus as to how far the scope of the field

extends, but this special issue of TESOL Quarterly focuses on its most

central concerns, namely, how language is acquired, how it is stored in the mind, and how it is processed in use

Findings from psycholinguistics are relevant to TESOL in several ways Most obviously, there is a growing body of research into the psychology

of second language acquisition Researchers have investigated how learn-ers construct a new language system alongside an existing one (Schwartz

& Kroll, 2006) Established concepts from cognitive psychology have been invoked to shed light on the challenges faced by second language (L2) users (Robinson, 2001) There has been particular interest in how bilinguals coordinate their two language systems (Dijkstra & Van Heu-ven, 1998) and in how the first language (L1) suffers attrition when the second becomes the dominant one (Hansen, 2001)

Research into L1 performance also makes an important contribution

On the one hand, it provides insights into language as a general phe-nomenon All language users, whatever their L1, have to deal with the forms in which language is transmitted: They all have, for example, to assemble speech under pressure of time or to use sweeps of the eye to read printed text All human beings share a similar brain configuration, and it is reasonable to suppose that any language maps on to the opera-tions for which the brain is best fitted (Deacon, 1997, pp 115–116) On the other hand, we can also learn from research into processes specific

to the L1, which serves to identify routines that the L2 user needs to acquire It is useful to know, for example, that listeners to English use lexical stress to work out where words begin and end (Cutler, 1990); those who do not use this technique in their L1 will need to adjust to it Note that the line of argument here is not that the L2 learner must slavishly imitate the native user, but simply that years of exposure to English have enabled the native user to evolve the most efficient ways of producing and making sense of the language

At this juncture, some clarification is called for The impression may have been given that psycholinguistic enquiry is heavily normative and fails to provide for the diversity of language users and contexts of use Psycholinguists do indeed aim to identify the processes which underlie language performance in general, but they also recognize that the pro-cesses in question will vary from person to person and from situation to situation For example, cognitive models of L1 and L2 reading recognise that the way in which a text is read will depend upon, inter alia, the reader’s skill, the reader’s experience of this type of text, the reader’s familiarity with any terminology, and the reader’s purpose in reading Similarly, models of how readers construct meaning incorporate the cues provided by their knowledge of the world, the writer, and the topic; they also allow for the way a word’s range of possible senses are constrained

by the context within which it occurs

PSYCHOLINGUISTICS AND TESOL

From all that has been said so far, one would expect psycholinguistic findings to be grist for the mill of those working within TESOL, but that has not been the case There are a number of reasons why Most obvi-ously, psycholinguistics demands familiarity with ideas in both language and psychology Linguists and psychologists have different priorities and different ways of thinking As George Miller put it (1990): “Grammarians are more interested in what could be said than in what people actually say, which irritates psychologists, and psychologists insist on supplement-ing intuition with objective evidence, which irritates lsupplement-inguists” (p 321)

In addition, psycholinguistics draws its information from a range of other domains, including discourse analysis, phonetics, language pathol-ogy, computer modelling, and neuroscience This eclecticism makes the field exciting to those of us who work in it, but bewildering, to say the least, to somebody who comes to it afresh So psycholinguistics is some-times perceived as daunting to teach and study It is certainly not, but it

is quite often given a lower profile in university courses (including mas-ter of arts in TESOL programs) than it deserves, and it may end up being taught by nonspecialists.1

Another complication is that psycholinguistic research has two very

different traditions There is an evidence-driven approach, which examines

how human beings acquire, produce, and understand language with a

1 This is particularly the case in the United Kingdom, where many departments of applied linguistics and schools of education are underinformed and understaffed in this area The situation appears to be rather better in the United States, Canada, and Australia.

view to tracing similarities of behavior between users It relies heavily on

experimental findings and on observation There is also a parallel

theory-driven approach, in which researchers adopt the assumption that the

accounts of language constructed by linguists correspond closely to what actually takes place in the mind and use them as a framework for inves-tigating the nature of language competence If anything, the gap be-tween the two has widened in recent years The theory-driven approach still adheres to the traditional notion of language as rule-governed be-havior, whereas many who espouse the evidence-driven approach are willing to contemplate the possibility that language use is driven by ex-ample rather than rule

The importance of psycholinguistic theory is increasingly being rec-ognized by specialists in second language acquisition (SLA) A number

of distinguished researchers (among them, Segalowitz, Nick Ellis, DeKey-ser, Hulstijn, Schmidt, Long, and Robinson) have drawn on concepts such as automaticity, working memory, implicit learning, and attention,

in order to provide insights into how L2s are acquired However, it also has to be acknowledged that other writers more immediately concerned

with the applications to TESOL have sometimes used the terms

psycho-linguistic and cognitive very loosely and have explained background

theory inadequately The message has to be: Caveat lector.

All of this makes the present issue of TESOL Quarterly especially timely.

Its aim is to build bridges to some of the more important ideas in psy-cholinguistics and to show their relevance to English language teaching All the articles included in this issue fall within the evidence-driven tradition They draw upon well-established principles of cognitive psy-chology and apply them to the special circumstances of the L2 learner and user There is a particular emphasis on L1 processing and how it can assist our understanding of processing in an L2, but issues connected with vocabulary, bilingualism, and acquisition are also represented The remainder of this introduction outlines some of the ideas that are touched upon by contributors The coverage is by no means compre-hensive and is not intended to offer a state-of-the-art picture of psycho-linguistics (for more detail on the various topics, see Field, 2004b) The principal concerns of psycholinguistics are taken to be:

• How individuals acquire language (whether an L1 or an L2)

• How individuals store language in their minds

• How individuals use language (how they assemble it into produc-tions and how they understand it when produced by others) These three areas of enquiry provide a structure for the discussion that follows

It might seem that psycholinguistic studies of first language acquisi-tion (FLA) would have a great deal to offer to our understanding of how

an L2 is acquired However, the point has often been made that the situation of ESL learners is very different from that of infants acquiring English as L1 ESL learners possess an already-established L1 and are often fully developed cognitively and capable of analysing input criti-cally

Acquiring Expertise

One answer has been to treat L2 proficiency not as the outcome of a constrained process that follows the course of FLA, but as a form of

expertise that has to be developed over time Widely quoted have been

Anderson’s ACT models (e.g., Anderson, 1983), which postulate that a novice in any form of expertise (driving a car, playing chess, etc.) starts

out with a form of declarative knowledge (knowledge that) which becomes transformed through practice into procedural knowledge (knowledge

how).

In the ACT models, practice brings two particular benefits, which serve to reduce the demands on a language user’s working memory Anderson himself has associated them with SLA Firstly, single steps within a larger operation become combined One way of representing this in language learning is in terms of building discrete lexical items into chunks (Wray, 2002) Secondly, and very importantly, the steps become increasingly automatic until they make minimal demands on the attention of the performer A major difference between a novice and a skilled L2 user lies in how automatic the processes are that the user

commands (DeKeyser, 2001; Segalowitz, 2003) The concepts of working

memory, automaticity, and attention are mentioned in several of the articles

in this volume (Farris, Trofimovich, Segalowitz, & Gatbonton; Field; Spelman-Miller, Lindgren, & Sullivan; Walter)

Exemplar Models

The influence of FLA can be seen more directly when we consider the

type of knowledge that is acquired by an L2 learner Recent instance-based theories (or exemplar models) propose that we acquire our L1 by

assem-bling multiple traces of the encounters we have had with speakers Thus,

a child builds up a composite representation of the category DOG by

drawing upon a set of images of most or all of the real-world animals that, over time, have had this particular label attached to them by an adult

(Hintzman, 1986) Similarly, the child learns to place an –ing form after

enjoy, a to infinitive after want and a simple stem after can because

mul-tiple exposures to these patterns are recorded in the child’s mind This hypothesis might seem implausibly wasteful, but it accords with what has been increasingly learned about the massive storage capacity of the hu-man brain (Da˛browska, 2004, p 18)

Exemplar theory has given rise to a view of SLA as

• potentially implicit and incidental, with the learner accumulating traces

without necessarily being able to express what has been acquired (Schmidt, 1994; Hulstijn, 2003)

• emergentist, with patterns of L2 knowledge being built up randomly in

a way that is determined by exposure to the target language (Larsen-Freeman, 1997; Ellis, 1998)

• example based, with the learner matching new examples of words or

syntactic structures to examples encountered earlier, rather than relying principally on abstract grammar rules (Tomasello, 2003)

• sensitive to relative frequency, since the more examples a learner has

encountered of a particular word or pattern, the more firmly estab-lished it will be in his or her mind (Bybee & Hopper, 2001) This perspective strongly informs the article in this issue by Ellis, Simp-son-Vlach, and Maynard

STORING

How language is stored receives comparatively little attention within TESOL Teachers speak of learners acquiring a set of phoneme values or

a productive vocabulary but tend not to discuss precisely what is being acquired

Variability

Among the complicating factors is the variable nature of the spoken input to which learners are exposed Words are taught in their citation forms, but in connected speech they are subject to strong reductive influences when they are not the most prominent item in an intonation group Do L2 learners then have to store a range of possible realisations

of each word as part of their oral vocabulary?

Two possible answers have just been mentioned We might assume that, instead of a single citation form of the word being stored in the mind, the language user draws upon multiple traces of it, said in differ-ent voices on differdiffer-ent occasions Another solution is to assume that many words are not only stored individually but are also embedded into recurrent chunks of language (Pawley & Syder, 1983) Within the chunk, their form remains more consistent; the information that is stored may even include a standard prosodic pattern

This is the background to an initiative by researchers at the University

of Michigan, who aim to establish a dataset of frequent formulaic chunks that are of use to those studying English for academic purposes One of the problems in constructing such a resource lies in defining what is or

is not recognized as a formulaic chunk Ellis, Simpson-Vlach, and Maynard

(this issue) investigate three factors which cause a language user to re-gard a chunk as a linguistic unit They are its length, the cumulative

frequency of the components of the chunk, and mutual information (MI),

the extent to which the components of the chunk co-occur across the corpus in question The writers report that the most important factor for L1 users proved to be MI but that for L2 users tested, it was cumulative frequency This result suggests that, even at quite an advanced level, L2 users continue to process the formulaic chunk as if it were a set of independent words

The Bilingual Lexicon

A second issue concerns the structure of the vocabulary store that a language learner draws upon Assuming a bilingual has two relatively complete systems of vocabulary, how are they distributed? Are they in separate stores but linked to a single semantic base in which fundamen-tal real-world concepts are held? Are they in separate stores, each with its own semantic base? Or are they in one single store? A number of leading researchers (e.g., Dijkstra & Van Heuven, 1998) have come to favour the last option

There are then interesting consequences for any view of how words are recognised by a bilingual reader The most widely accepted model of word recognition holds that input in the form of a group of letters

triggers a process of competition, in which a reader foregrounds a set of

possible matches for a group of letters on the page The candidates are

favoured (activated) to different degrees according to how frequent they

are and how closely they resemble the stimulus As evidence

accumu-lates, one item achieves such a high level of activation that it wins out over all the others and becomes recognised.2

If one subscribes to the single-store solution, then one has to accept that candidates from both a bilingual’s languages will enter the compe-tition An English–Spanish bilingual might be operating in English, but

the sight of the word animal would trigger access to the identical word in Spanish as well It has indeed been demonstrated that words like animal

with shared meanings across two languages are recognised more quickly

by bilinguals than they are by monolinguals (Van Hell & Dijkstra, 2002) The contribution by Sunderman and Schwartz (this issue) extends

earlier studies by asking what happens in the case of words such as grave,

which have (for the reader at least) one sense in Spanish but two in English As the researchers hypothesised, the existence of the additional mismatched sense weakens the advantage that is gained from a shared form

The article by Ecke (this issue) also focuses on the bilingual lexicon

It examines what occurs when a speaker cannot retrieve a word in his or

her L1 Faced with what is known in psycholinguistics as a tip-of-the-tongue

state (Brown & McNeill, 1966), speakers draw upon mental cues to the

target, including words which resemble it in form Ecke finds evidence that when a group of Spanish–English bilinguals are attempting to re-trieve a word in Spanish, they make use of retrieval cues from both their languages The number of English cues used becomes larger as ESL proficiency increases This evidence thus lends further support to the notion of a single lexical store

USING

A third area of psycholinguistics, often referred to as language

process-ing,3is concerned with how users of a language assemble utterances and how they understand the productions of others This is potentially the richest vein of all for TESOL practitioners because it can shape under-standing of what constitutes expertise in speaking, listening, reading, and writing A problem for the teaching of all four skills is that instruc-tors often lack a clear and detailed idea of the behavior that they wish to induce in their learners and thus a goal towards which their teaching might tend They fall back on conventional methods such as the com-prehension approach in reading and listening or the grading of tasks

2 A similar principle has been applied to listening.

3 Throughout, the term is used for both production and reception Some commentators limit it to reception.

and genres in speaking and writing, but these generalized approaches often fail to address the detail of what it is that constitutes a successful L2 speaker, listener, writer, or reader

It is only relatively recently that language processing theory has begun

to influence skills instruction in TESOL It is therefore noteworthy that six of the nine studies accepted for this special issue focus on skills

Speaking

A major concern in constructing models of L1 speaking has lain in an apparent incompatibility between the syntactic complexity of certain ut-terances and the speed with which they are assembled How do speakers manage, in much less than a second, to retrieve appropriate words from

a vocabulary of at least 30,000 items and to slot them into a syntactic structure? Two possible answers have already been mentioned First, the speaker is assisted by storing recurrent strings of words in the form of preassembled chunks, in which the syntax is ready-made (Wray, 2002) Second, many of the processes that support L1 speech production are highly automatic (Levelt, 1989, pp 20–22) These two features of skilled L1 speech minimise the cognitive demands that speech production im-poses on a speaker

However, with many L2 speakers, chunking and automatisation are only partially acquired—with the result that assembling an utterance requires considerably greater resources of attention The problem is that our resources of attention are strictly limited; if they are partly given over

to speaking, then there will be fewer available for other possible tasks And vice versa: If there are demanding tasks that have to be done, then the speaker’s ability to assemble an accurate and comprehensible utter-ance may be affected

This is perhaps not an issue in everyday conversation, but it becomes one in conditions where lives may be at stake In a fascinating article, Farris, Trofimovich, Segalowitz, and Gatbonton (this issue) investigate the extent to which the L2 productions of air traffic controllers are successful in conveying their intended message They investigate the communication skills of trainee controllers under conditions which in-corporate the other simultaneous demands that are likely to be made on their attention The findings make compelling reading for airline trav-ellers

Listening

Psycholinguistics distinguishes two operations in listening The first is perceptual, with the listener decoding the signal that reaches the ear

The second is conceptual, with the listener building meaning by contex-tualising the words that have been decoded In accounts of L2 listening, writers sometimes imply that the two operations are alternatives rather than mutually dependent (see Field, 2004a, for a discussion) A received view has developed that decoding is the lesser partner because any prob-lems in this area can usually be redressed by the use of what is loosely

termed context This line of argument ignores the fact that a small error

of word recognition can have an impact on the understanding of the whole utterance and indeed of the whole discourse Recently, opinion has shifted and there has been a renewed interest in how input contrib-utes to the meaning that is extracted (Field, 2008)

If we are to learn more about the way in which an L2 listener combines cues from input with cues from context, then it is imperative to know how much of the input is likely to be successfully decoded and if there are biases in the way the listener distributes his or her attention Field (this issue) cites evidence from psycholinguistics that content and func-tion words are processed differently In English, fast-track decoding for function words is assisted by their association with weak quality syllables (Grosjean & Gee, 1987) This characteristic raises the question of how L2 listeners handle function words They might find them easy to identify because of their high frequency or hard to identify because of their low prominence The answer provides an indication of the type of intake that

an L2 learner of English derives from an utterance: to what extent it contains elements of syntax and to what extent it is lexically based

Writing

TESOL practitioners are generally familiar with what is termed a

pro-cess approach to the teaching of L2 writing, in which learners are

en-couraged to draft, review, and revise their texts, often working in pairs or groups (White & Arndt, 1991) They may not be aware that this approach largely grew out of early psycholinguistic models of L1 writing which identified the phases through which a writer proceeds Most models (e.g., Hayes & Flower, 1980; Kellogg, 1994) mention four main phases: planning, translation (from ideas to language), execution, and editing.4 The important point about the phases is that they do not form a neat sequence: They are recursive, with the writer free to go back at any moment and change what was planned or what was written Insights into these writing decisions are obtained by two methods: (a) by asking writ-ers to produce a verbal report of what is in their minds as they write, and

4 These terms are not universally employed.

(b) by maintaining a record of a writer’s keystrokes and thus of all the revisions made Spelman Miller, Lingren, and Sullivan (this issue) are leading exponents of keystroke logging as a means of investigating L2 writing processes In their article they provide a detailed rationale for the method and give concrete examples of the informative data it provides Their particular interest here is to track longitudinally the developments

in L2 writing skills across 3 years of study

Reading

A point shared by two of the three articles on reading in this issue, and one that may appear surprising, is the part played by phonology Firstly,

it is now generally accepted that readers of alphabetic scripts use two

routes for identifying words: a lexical (or whole word) one and a sublexical

one based on grapheme–phoneme relationships (Coltheart, 1978) This dual approach is even employed by users of a relatively opaque spelling system such as the English one However, the ability to recognise discrete phonemes (and thus to make the letter-sound connections that an al-phabetic system requires) may not be innate Widely quoted studies in L1 psycholinguistics (e.g., Morais, Cary, Alegria, & Bertelson, 1979) have suggested that illiterate individuals are incapable of recognising and manipulating phonemes (a typical task is to ask them what word remains

after deleting the first sound from gold) Commentators have concluded that phonological awareness (the knowledge that speech in any language

draws upon a finite system of sound values)5 may be a byproduct of learning to read using an alphabetical writing system (see Goswami & Bryant, 1990 for a critical review of this theory)

If one accepts that this faculty is acquired in the way described, then matching regular written forms to their spoken equivalents must be more difficult for those whose L1 does not have an alphabetical system McDowell and Lorch (this issue) compare readers from mainland China

who were introduced to alphabetic principles in the form of pinyin (an

early reading script) with readers from Hong Kong who mastered logo-graphic Chinese characters without pinyin and thus have had no expo-sure to phoneme-level analysis in L1 McDowell and Lorch expand on earlier studies by adding a third condition: They include a group of learners with formal phoneme training in the form of exposure to the International Phonetic Alphabet

There is a second way in which phonology is implicated in reading

5Note that this is entirely distinct from the concept of phonological working memory, discussed

by Walter (this issue).