neurocognitive basis of implicit learning of sequential ... cconway6/papers/ · neurocognitive...
Post on 08-Jul-2018
Embed Size (px)
Neurocognitive Basis of Implicit Learningof Sequential Structure and Its Relation
to Language ProcessingChristopher M. Conwaya and David B. Pisonia,b
aSpeech Research Laboratory, Indiana University, Bloomington, Indiana, USAbDevault Otologic Research Laboratory, Department of OtolaryngologyHead & Neck
Surgery, Indiana University School of Medicine, Indianapolis, Indiana, USA
The ability to learn and exploit environmental regularities is important for many as-pects of skill learning, of which language may be a prime example. Much of such learningproceeds in an implicit fashion, that is, it occurs unintentionally and automatically andresults in knowledge that is difficult to verbalize explicitly. An important research goalis to ascertain the underlying neurocognitive mechanisms of implicit learning abilitiesand understand its contribution to perception, language, and cognition more gener-ally. In this article, we review recent work that investigates the extent to which implicitlearning of sequential structure is mediated by stimulus-specific versus domain-generallearning mechanisms. Although much of previous implicit learning research has em-phasized its domain-general aspect, here we highlight behavioral work suggesting amodality-specific locus. Even so, our data also reveal that individual variability in im-plicit sequence learning skill correlates with performance on a task requiring sensitivityto the sequential context of spoken language, suggesting that implicit sequence learningto some extent is domain-general. Taking into consideration this behavioral work, inconjunction with recent imaging studies, we argue that implicit sequence learning andlanguage processing are both complex, dynamic processes that partially share the sameunderlying neurocognitive mechanisms, specifically those that rely on the encoding andrepresentation of phonological sequences.
Key words: implicit learning; sequence learning; language processing; modality-specificity; domain-generality; prefrontal cortex; basal ganglia
An essential characteristic of skill learning isthe necessity of encoding, representing, and/orproducing structured sequences. Because theenvironment is characterized by the regularand coherent occurrence of sounds, objects,and events, any organism that can usefully en-code and exploit such structure will have anadaptive advantage. Language and communi-cation are excellent examples of structured se-
Address for correspondence: Christopher M. Conway, Departmentof Psychology, Saint Louis University, 3511 Laclede Ave., St. Louis,MO 63103. Voice: +314-977-2299; fax: +314-977-1014. email@example.com.
quential domains to which humans are sensi-tive. That is, spoken and written language units(letters, phonemes, syllables, words, etc.) eachadhere to a semiregular, sequential structurethat can be defined in terms of statistical orprobabilistic relationships (Rubenstein, 1973).Sensitivity to such probabilistic information inthe speech stream can improve the perceptionof spoken materials in noise; the more pre-dictable a sentence is, the easier it is to per-ceive (Kalikow et al., 1977; see also Miller &Selfridge, 1950). The presence of probabilistic,structured patterns is found in almost all aspectsof our interaction with the world, whether it bein speaking, listening to music, learning a tennisswing, or perceiving complex scenes.
Ann. N.Y. Acad. Sci. 1145: 113131 (2008). C 2008 New York Academy of Sciences.doi: 10.1196/annals.1416.009 113
114 Annals of the New York Academy of Sciences
How the mind, brain, and body encode anduse structure that exists in time and space re-mains a formidable challenge for the cogni-tive and neural sciences (Port & Van Gelder,1995). This issue has begun to be eluci-dated through the study of implicit learn-ing (Cleeremans, Destrebecqz, & Boyer, 1998;Conway & Christiansen, 2006; Reber, 1993;Perruchet & Pacton, 2006). Implicit learninginvolves automatic learning mechanisms thatare used to extract regularities and patternsdistributed across a set of exemplars, typicallywithout conscious awareness of the regularitiesbeing learned. Implicit learning is believed tobe important for many aspects of skill learning,problem solving, and language processing.
One important research goal is to estab-lish whether implicit learning is subserved bya single, domain-general mechanism that ap-plies across a wide range of tasks, input, anddomains, or instead consists of multiple task-or stimulus-specific subsystems. This is espe-cially important if we are to usefully applyknowledge gained from laboratory studies ofimplicit learning to more real-world examplesof language and skill learning. Two bodies ofevidence have favored the former conclusion.First, many studies have demonstrated implicitlearning across a wide range of stimulus do-mains and tasks, including but not limited tospeech-like stimuli (Gomez & Gerken, 1999;Saffran, Aslin, & Newport, 1996), tone se-quences (Saffran, Johnson, Aslin, & Newport,1999), visual scenes and geometric shapes(Fiser & Aslin, 2001; Pothos & Bailey, 2000),colored light displays (Karpicke & Pisoni,2004), and visuomotor sequences (Cleeremans& McClelland, 1991). Second, other studiesusing the artificial grammar learning (AGL)paradigm (Reber, 1967) have shown that par-ticipants can transfer their knowledge gainedfrom one stimulus domain (e.g., visual symbols)to a different domain (e.g., nonsense syllables)if the underlying rule structure is the same(Altmann, Dienes, & Goode, 1995; Brooks &Vokey, 1991; Gomez & Gerken, 1999; Reber,1969). Thus, implicit learning is argued to re-
sult in knowledge representations that are ab-stract or amodal in nature, independent of thephysical qualities of the stimulus (Reber, 1993).
Despite this apparent evidence for a single,domain-general, and amodal implicit learningskill, there is reason to believe that implicitlearning may be at least partly mediated by anumber of separate specialized neurocognitivemechanisms. First, many views of the mind en-compass to a greater or lesser extent the notionof functional specialization (Barrett & Kurzban,2006; Fodor, 1983). As an example, workingmemory (Baddeley, 1986) consists in part ofmultiple, modality-specific processing compo-nents. Second, some of the transfer of knowl-edge data discussed above may suffer frommethodological concerns (Redington & Chater,1996); even if one disregards such concerns, itis not immediately clear that the results nec-essarily support the notion of amodal knowl-edge gained through implicit learning. Third,the fact that implicit learning has been demon-strated across numerous input types and tasksdoes not necessarily imply a single, domain-general system. It is just as possible logicallythat there may exist multiple implicit learningsubsystems that all have similar computationalprinciples, but that only some are engagedfor specific task and input demands (Conway& Christiansen, 2005; Goschke, 1998; Seger,1998). Finally, consistent with a multiple subsys-tems perspective, correlational analyses suggestthat implicit learning is relatively task-specific(Feldman, Kerr, & Streissguth, 1995; Gebauer& Mackintosh, 2007).
In the first section below, we review recentbehavioral work that examines the issues ofdomain-generality and modality-specificity inimplicit learning. Our investigations explorethe effect of sensory modality on implicit learn-ing of sequential structures and the contribu-tion of such abilities to language processing.Following the presentation of these studies, wereview recent neural evidence that further illu-minates the underlying neurocognitive basis ofimplicit sequence learning. Finally, we integratethe behavioral and imaging data and offer an
Conway & Pisoni: Implicit Learning and Language 115
account of the relation between implicit learn-ing and language processing.
Cognitive Basis of Implicit Learning
Recent Behavioral Studies
The following three studies all use the ar-tificial grammar learning (AGL) methodology(Reber, 1967). In a standard AGL task, a finite-state grammar is used to generate stimuli thatconform to particular rules that determine theorder in which each element of a sequence canoccur (Fig. 1). Participants are exposed to therule-governed stimuli under incidental and un-supervised learning conditions. Following ex-posure, participants knowledge of the complexsequential structure is tested by giving them atest in which they must decide whether a set ofnovel stimuli follow the same rules or not. Gen-erally, participants display adequate knowledgeof the sequential structure despite having verylittle explicit awareness of what the underlyingrules are; in fact, most participants reportthey were guessing during the test task. Thesimilarities between implicit learning of arti-ficial grammars and language acquisition arenotable: both appear to involve the automaticextraction of sequential structure from a com-plex input domain that results in knowledgethat is difficult to verbalize explicitly (Cleere-mans et al., 1998; Reber, 1967).
In the following sets of studies, we examinethree important issues: the nature of modalityconstraints affecting implicit learning, the ques-tion of whether learning is mediated by multi-ple, independent learning mechanisms, and theextent to which implicit learning is a fundamen-tal component of language-processing abilities.
To rigorously examine the effect of modalityin implicit learning, we had three groups of par-ticipants engage in an AGL task, each assignedto a different sense modality: audition, vision,or touch (Conway & Christiansen, 2005). In
Figure 1. An example of an artificial grammarused to generate sequences in implicit learnin