Language processing andlearning
3.1 Introduction
This chapter addresses the question ofhow we process and retainlinguisticinformation. Initially,the discussion delvesintothe varying modelsthat examinethe processing and production oflanguage. The chapterthen moves onto awareness oflanguage andthe roleit playsin acquisition. Cognitionin relationto memory receives some attention, butis only briefly mentioned, asthe researchto dateislacking if welook atthisfrom alinguist’s perspective. There will be further discussion below of theprocessing of languagein
connectionto memory andthe connected components. Additionally, thereis a discussion of the difficultyin cognitive processing experienced by L2learners. In particular,this
dissertationlooksto deal withtheissues L2learners experiencein an FL environment where exposuretothelanguageis ratherlimited. The chapter concludes withtheintroduction of an integrated model forlanguage processing and acquisitionin relationto memory, andthe featuresit embodies.
3.2 Three knowledge sources forlanguage comprehension and production
Accordingto Anderson and Lynch (1988) and Skehan (1998), ourlanguage processing draws onthree main knowledge sources of systemic knowledge,schematic knowledge, and contextual knowledge. Systemic knowledge, accordingto Anderson and Lynch,is comprised of syntactic, semantic, and morphological knowledge, andinthis dissertation itisinterpreted as consisting ofthe dual systems of analyticlanguage knowledge and holisticlanguage knowledge discussedin Section 2.4. Schematic knowledge, onthe other hand, refersto a person’s background knowledge, both factual and socio-cultural, and his or her procedural knowledge of howlanguageis usedin discourse, whereas contextual knowledgeisthat
26
person’s knowledge of situation (physical setting, participants, etc.) and of context (what has been and will be said).
For comprehension, Anderson and Lynch proposethatthesethree knowledge sources are drawn on,interactively,to understandthe meaning (see also Schwanenflugel,
Harnishfeger, & Stowe, 1988), which suggeststhatlinguisticinformation (i.e., systemic knowledgein Anderson and Lynch’s framework)is notthe exclusive source we depend onto extractthe message. Thatis, during oral communication, we relate whatis being saidto previous knowledgethat we have (i.e. schematic knowledge), which enables usto make very effectiveinferences aboutthe messagein question. Likewise, we referthe messagetothe array of utterancesthat arelikelyto be conveyed giventhe nature ofthe situational context or to what has been said previously (i.e., contextual knowledge), by which we narrow downthe range of probable meaningsthat we may encounter and maximizethe chance of our
deductions about meaning working. Skehan (1998)then extendsthe application of Anderson and Lynch’s modelto output production as well. Thatis,in speech production,too, we depend onthethree knowledge sourcesin orderto arrive atthelinguistic material forthe conveyance of the message. Whatis ofimportance hereisthatthe speaker frames whatisto be said bearingin mindthe comprehension capacity ofthelistener.
As will be discussedin Section 3.8,there are negativeinfluences stemming from schematic and contextual knowledge onthe growth ofthe duallanguage systemsin adult L2 learners. Unlike children, adult L2learners havetolearn a newlanguage utilizing, whether or not by choice,their existing schematic and contextual knowledge. Whilethesetwo
knowledge sources do helplearners with comprehension and productioninthe L2,their usefulness can easilyimpinge onthe need forthe development ofthe dual systems. Thisis also anissue becausetheirinterlocutors (especially native speakers) are adept at extracting theintended meaning of erroneouslearner speech withtheir schematic and contextual as well as systemic knowledge. Having derivedthe appropriate meaning, native speakers will not
27
feelthe needto providethelearners with negative feedback onthelanguage, which would instigate analysis onthelearners’ part.
3.3 Noticing, attention, and awareness
Accordingtothe Noticing Hypothesis proposed by Richard Schmidt (1990,1994,1995; Schmidt & Frota, 1986),thereis nolearning without noticing, andattentionis necessary for noticingto occur. In orderto fully understand Schmidt’s concept, however,two key words need elucidation: namely, noticing and attention. First, as for attention, Tomlin and Villa (1994) seeit as consisting ofthree separate butinterrelated networks ofalertness/readiness (i.e., attentional resourcesthat are not depleted), orientation (i.e., allocation of attentional resources), and detection (i.e., recognition ofinput). Tomlin and Villa proposethat (1)the cognitive process of detectionis a necessary condition for anytype oflearningtotake place, (2) alertness/readiness may promote orientation and detection, (3) orientation may facilitate orinhibit detection, and (4) awareness (inthe sense of memory or understanding of
something) also may enhance detection, becauseit may enhancethe operation of alertness and orientation.
If we deconstructthe notion of attentioninthis way,then,the concept of noticingin L2 learning can be understood asinvolving some deeper cognitiveprocessthan mere detection. For example, when someone has noticed something, he or she may have (1) recognized a detected form as non-existentin his or her current L2 knowledge system (a phenomenon Doughty and Williams (1998) call noticing aform); (2) formed a hypothesis about a detected form; (3) realizedthat a particular part of what he or she said or wroteinthe L2 was different from how a native speaker ofthe L2 would expressit (referredto as noticing a gap by
Schmidt and Frota (1986) and Swain (1995)); (4) rejected aninterim rulein his or her L2 representational system (a further process of noticing a gap); (5) modified a hypothesisthat was contradicted (a yet further process of noticing a gap); or (6)identifiedthat a hypothesis
28
was confirmed (a rule-strengthening or exemplar-generation effect suggested by Skehan (1998)). Another attemptto characterize noticing has been made by Robinson (2003).
Accordingto Robinson, noticingis defined as detection plus awarenessthrough either oftwo types ofrehearsal in working memory: maintenance rehearsal (data-driven,instance-based processing) and elaborative rehearsal (conceptually driven, schema-based processing).
3.4 Working memory
In orderto properly grasp Robinson’s definition,then,the notion of working memoryis in need of clarification. Models of working memory have been developed by cognitive psychologists sincethe beginning ofthe 1960s. Accordingto Baddeley’s (2000,2007; Baddeley, et al., 1974) multicomponent model, working memory consists of four
components:the phonologicalloop,the visuospatial sketchpad,the episodic buffer, andthe central executive. Baddeley’s model primarily postulatesthat differenttypes ofinformation are stored and manipulatedin different working-memory workspaces (namely,the firstthree components), which are orchestrated andlinkedtolong-term memory bythelast component, the central executive.
Cowan’s (1988,1995,2005) embedded-processes model, onthe other hand, emphasizes the potentiallyinfinite power oflong-term memorythat expeditesthe operation of working memory. Accordingto Cowan’s model,the performance of working memory can be highly restricted because attentional focusis only ableto handle a small number of chunks at atime, threeto five chunksin normal adults (Cowan, 2001).10 What allows working memoryto operate (potentially far) beyondits attentionallimitationsis a summoned subset oflong-term memory, whichisinthe state of high activationin working memory. Itis postulatedthat
10 Theissue ofthe attentional capac ityin working memory was first discussed by Miller (1956). Usingthe famous phrase magical number seven, Miller proposedthatthe maximum number of pieces ofinformationthat a young adult can sustain at atimein working memoryis seven plus or minustwo. The number proposed by Cowanis,then, about half of Miller’s original proposal.
29
thereis nolimitto activation of representationsinlong-term memory. Thus, at any given moment,the more activation fromlong-term memory,the morethat can be handled by working memory.
Thelast model of working memory reviewedinthis sectionisthe one proposed by Ericsson and Kintsch (1995). Exploring furtherthelinkage between working memory and long-term memory, Ericsson and Kintsch arguethat humans use skilled memory,termed long-term working memory,in most everydaytasks such asreading. When we comprehend the complex relations betweenthoughts expressedin along novel or a scientifictext, for instance, we must be ableto manage alot morethanthree or four or even seven chunksin working memory. Accordingto Ericsson and Kintsch, we are capable of accomplishing such a cognitively demandingtask because we can store most oftheinformation neededto
successfully carry outthetaskinlong-term memory by means oflinkingthatinformationto retrieval structures, which can be understood as readily retrievable constructionsin
long-term memory. By relying on retrieval structures, we only needto hold a minimal number of conceptsin working memory serving as cuesthat are sufficientto retrieve everything connectedtothem bythe retrieval structures. Ericsson and Kintsch refertothe quickly accessibleinformation by way of retrieval structuresthattechnically nullifythelimit of attentional focus aslong-term working memory.11
The study of working memory has produced different models, butthese have been developed by specialistsin cognitive psychology, not by language acquisition researchers. Thus, any attemptto make a connection betweenthose models of working memory andthe accounts of SLA proposed bylanguage acquisition scholars has to be made carefully. Inthis dissertation,those assumptionsinthe model proposed by Ericsson and Kintsch aretentatively
11 See also Guida, Gobet, Tard ieu and Nicolas (2012), Guida and Tardieu (2005), and Guida, Tardieu, and Nicolas (2009) fortheir proposal of the “personalisation method” as a wayto operationalizethe long-term working memory.
30
adopted.12 The postulatedlong-term working memory by way of retrieval structures,then, can beinterpreted as a potentially unlimited amount of operationalinformationin working memory (eitherinput from outside or self-generated output)thatis attention-free and networked withthethree sources of knowledgeinlong-term memory (see Section 3.2).
3.5 Transfer-appropriate processing
This dissertation hasthus far given an overview ofthe processes by which formulaic sequences are acquired, and examined general cognitive accounts oflearning with respectto working memory. In relationtothe focus ofthis dissertation, namelytheteaching of
formulaic sequences, one other cognitive account oflearning needsto beintroduced: transfer-appropriate processing. Segalowitz and Lightbown (1999) arguethat memory performanceislargely regulated bythe relationship between howinformationisinitially encoded and howitislater retrieved. Theideaisthat when a person acquires new
information,thatinformationis encodedin a prompt-dependent or context-sensitive fashion; therefore,that person’s successful retrieval ofinformation previouslylearnedis facilitated or lessened accordingtothe extentto whichthe cognitive operations exerted atthetime of recollection corresponds withthose previously engagedin atthetime of acquisition or learning. Segalowitz and Lightbown explainthatthis effect occurs becausetheinternal cognitive state oftheindividual affords him or her cluesto assist with recollection;ifthe retrieval cuestriggered atthetime of recollection matchthe cues encoded duringlearning of theinformationin question,then retrieval will be readilyinvoked (for empirical evidence see Blaxton, 1989; Roediger & Guynn 1996). As recently summarized by Segalowitz (2010):
12 For more extensive systema tic reviews and comparison oftheories of working memory, seethe volumes edited by Miyake and Shah (1999) and Conway, Jarrold, Kane, Miyake, and Towse (2008).
31
theease of retrieval (and hencethe fluency of action dependent onthat retrieval) will dependinlarge measure onthe degreeto which brain region activation patterns atthe time of retrieval overlapthe patternsthat were active atthetime of study. (p. 65) The reason for introducing heretheidea oftransfer-appropriate processing isthatitis closely relatedtothe framework ofthethree knowledge sources reviewedin Section 3.2. Thatis, successfullearning (andteaching) of alinguisticitem, whether formulaic or grammatical,is expectedtotake place whenits encoding coincides with simultaneous encodingin memory or activation of theschematic and/or contextual knowledgethat will subsequently be called up insituations where thelanguage itemis beingused. Thisthinking will be revisited when a synthesis of allthe keyideas described so faris presentedin Section 3.8.
3.6 Inherent difficulties of alinguistic feature
This section can be seen as supplementingthe descriptions ofthe characteristics of formulaiclanguage offeredin Chapter 2, but herethe focusis onidentifyingthose features of FSsthat makethem difficulttolearn, some of which especially pertainto adult L2learners. In so doing,it appliesthe framework usedin Matsuzaki (2011), which focuses onthe article system andthe difficultiesitis notorious for presentingto Japanese EFLlearners. The inherent difficulties of alinguisticitem, whether formulaic or grammatical, can be measured from a number of perspectives, each of whichis discussed here. What followsinthis section are,therefore,the aspects of anylinguistic featurethatinfluencethe relative ease or difficulty for cognitive operations reviewedin Sections 3.3 and 3.4.
3.6.1 Frequency
One category for assessinginherent FS difficulty hasto do withthe frequency ofinput thatthelearner receives. Anitemthat occurs frequentlyininput, whetherlexical or
32
grammaticalin nature, has a better chance of beinglearnedthan onethatisinfrequent. For instance,thereislittle doubtthatthe sequencenot often islearned beforeitsless frequent equivalent oncein a blue moon. A simple questionthenis whetherto choosetoteachless frequentitems asthey haveless opportunities of being noticed bythelearner, or whetherto prioritize more frequent ones asinfrequent ones arelikelyto beless useful.
3.6.2 Perceptual saliency
Another aspectthat contributestolearning difficultyis perceptual saliencyininput. For example,in oral communication,the definite thein most casesis unstressed andthus
imperceptible, which makesthe perceptual differencein meaning betweenDo you have time? vs. Do you have thetime? nearlyimpenetrable for L2learners. Giventhe pervasiveness of such articlesinthe case of English,then, relying solely on naturalinput seems fundamentally insufficient forlearnersto notice and eventually acquire formulaic sequences containing such perceptually non-salient features.
3.6.3 Communicativeload
Another factorthat canincreaselearning difficulty for anitem, or part of a multiword string,is whenit carrieslittle communicativeload. For example, as awkward asthe utterance If I am you, I will… might soundtothe ears of native English speakers, a non-native speaker would be able to deliver her or hisintended message withthe sentencein spite ofthe
erroneous word usage. Sinceitisimpossible for someoneto be someone else,the communicativeload ofthe use of pasttensein a counterfactual sentenceislow. Thereis nothing difficult forthe hearerto understandinIf I am you, I will… Even withoutthe erroneous modal will there would be no diminishment ofintelligibility. Compoundingthis difficulty forlearnersisthetendencyin normal communication for native speakersto not botherto correctthe error (except whenthey areindeed unsure ofthe message), asto do so
33
would cause unnecessaryinterruptions andimpedethe flow of communication. Thetendency for suchleniency onthe part of native speakersislikelyto be stronger in a foreignlanguage context wherethey are accustomedto speechthatis not nativelike.
3.6.4 Form-meaning-function complexity
An additional aspect of FSsthat further explainstheir difficultyisthe notorious
complexity ofthe relationships betweentheir form, surface meaning, and functions. A single exampleis offered here, but one whichillustratesthe point well. There are a great many examplesthat clearly showthe functional complexity of formulaic sequences andthe difficultiesthey cause L2learners. The one offered here shows howthere can be multiple layers of complexity even within a single FS.
Japaneselearners overuse will, for example,in some situations and underuseitin others. So, where a native speaker might ask Are you going outtonight?, a Japaneselearner will typically usewill. Conversely, Japaneselearners will nottypically usethe shortened form of will to make an offer, asinI’ll doit.Norindeed do advanced Japaneselearnerstendto use will asit hasjust been usedtwice here with typicallyand once more belowinthis paragraph withthe same function but without typically.In additiontothe form-function complexity of will, thestatement I will be going out contains a potential form-function difficulty forlearners intheout/outside distinction. Japaneselearners often have difficulty graspingthe semantic distinction between out and outside. Sinceout is usedin so many waysthat makethe usage ofthe word very uncleartothem,the wordis not easyto use, andthusthey will often saygo outside.
3.6.5 Grammatical reliability
Addingtothe form-function complexity are a huge number ofidiomatic statementsthat do not follow syntactic conventions. For example,inthe case of Japaneselearners of English,