Auditory Sequence/Artif icial Grammar Learning 33

language experience, attention allocation, and work-ing memory resources influence lower-level auditoryorocessing at the level of the brainstem and even atihe level ofthe cochlea.

These are known as top-down influences becausea higher level of processing-higher in the sense ofbeing further from the periphery of the central ner-vous system and also in the sense of being more cog-nitive-influences a level that is lower, or earlier, inthe process. For example, phonological awarenessrnay influence backward masking performance; audi-tory brainstem responses can be affected by musicaltraining or by a persont choice of what stimulus toattend to. Thus, the causal direction of auditory pro-cessing deficits is not clear. Although perceptual defi-cits could cause problems in spoken and written lan-guage, as well as verbal working memory, attention,and other cognitive functions, problems with thesehigher functions could also influence perception.

In summary the relationship between auditoryprocessing and language impairments is difficult tocharacterize. There is a large body of evidence show-ing that auditory processing deficits are common inchildren with SLI or SRD. However, it seems unlikelythat these deficits play causal roles in spoken and writ-ten language impairments. The auditory processingdeficits are diverse in nature, and no single type ofauditory processing difficulty has been found consis-tently across studies. Regardless of the auditory pro-cessing task, group differences are not always found,and even when they are, not all individuals with SLIor SRD demonstrate de6cits. Furthermore, alterna-ttve explanations of the observed deficits are plausi-ble. Interventions that aim to improve auditoiy pro-cessing have little effect on language or reading, andaudltory processing is not consistently predictive oftanguage and academjc outcomes.

There remain, however, avenues of research thatmay reveal the nature of the relationshiDs betweena-uditory processing and language, causai or other-wtse. Our understanding of the exceedingly complexa,udrtory system, and its interactive relationship withrhe language processing system, continues to grow.^I the applied level of ameliorating SLI and SRD,most auditory training interventions investigaled sordr nave been based on a tcmporal processing deficitaccount. Other hypotheses that focus on brainstemuming or cortical oscillations may lead to differ-ent intervention rechniques that are more effective.rxplaining the entire neurodevelopmental pathway

from the perception of the speech signal to the rep-resentation of phonological, lexical, and syntacticaspects oflanguage is an ambitious goal but one thatis worth pursuing.

Carol A. MillerPennsy lvania State IIn iv e rsi ty

See Also: Developmental Dyslexia; Interyention forChild Language Impairmentq Phonological Awareness;Processing Deficits in Children With LanguageImpairmentq Specific Language Impairment (Overview);Speech Processing.

Further ReadingsBanai, K. and N. Kraus. "Neurobiology of (Central)

Auditory Processing Disorder and Language-BasedLearning Disability." In Handbook of (Central) AuditoryProcessing Disorder. Volume I, F. E. Musiek and G. D.Chermak, eds. San Diego, CA: Plural Publishing, 2007.

Banai, K. and N. Kraus. "The Dynamic Brainstem." InControversies in Cantral Auditory Procasing Disorder,A.T. Cacace and D. J. McFarland, eds. San Diego, CA:Plural Publishing, 2009.

Fey, M., et al.'huditory Processing Disorders andAuditory/Language Interventions." lcrguage, Speech,and Heaing Seruica in Schools, v.42 (20t1)-

Marler, J. A. and C. A. Champlin. "Sensory Processingof Backward-Masking Signals in Children WithLanguage-Learning Impairment as Assessed Withthe Auditory Brainstem Response." /ozrnal of Speech,La.nguq.ge, and Hmring Research, v.48 (2005) .

Miller, C. A. and D. A. Wagstaff. "Behavioral ProfilesAssociated With Auditory Processing Disorderand Specific Langua ge lmpairment!' Journal ofCommunicatiotr Disonlers, v.44 (20 | t).

Rees, N. S.'Auditory processing Factors in LanguageDisorders: The View From Procrustes' B edi Journal ofSpeech and Hearing Disorders,v.38 (1973)-

Rosen, S. 'Auditory Processing in Dyslexia and SpecificIanguage Impairmen" Journal of Phonerics, v.3 t (2003).

Auditory Sequence/Artificial Grammar LearningSpoken language development depends uponbasic auditory processing mechanisms that encode

34 Auditory Sequence/Artificial Grammar Learning

structural regularities in the input. These mechanisms,studied under the guise of sequence/sequential learn-ing and often using the artificial grammar learningparadigm, represent midlevel cognitive processes thatcould be construed as being "above" auditory sensoryand perceptual mechanisms but "below" true languageprocesses. Research has suggested that these auditorylearning mechanisms are specifically attuned to time-varying, sequential signals (like spoken language), arean important aspect of the development of language,and may be compromised in children with impover-ished experience with sound.

Sequential learning refers to general abilities thatmost higher organisms possess for learning aboutstructured patterns of information in the environ-ment that unfold over time. Sequential learningcan take place through any sensory modality and inmultiple domains, including language and commu-nication, motor and skill learning, music Perceptionand production, problem solving, and planning. Itis essential for learning both patterns of stimuli andpatterns of responses, such as those required in clas-sical and instrumental conditioning, skill learning,and highJevel problem solving and reasoning. Suchlearning is particularly important for organisms earlyin development; in fact, human infants are very sensi-tive to temporal patterns from birth, especially audi-tory patterns such as rhythm and language.

As a form of implicit learning, sequential learn-ing can occur regardless of whether the individualintends to learn the pattern and does not requireexplicit recall or recognition. For instance, sequentiallearning is preserved in amnesic patients who have noexplicit memory of the sequences. The implicit natureof (visual) sequence learning has traditionally beeninvestigated through Mary Jo Nissen and Peter Bul-lemer's serial reaction time (SRT) task. In the origi-nal task, a single visual stimulus is presented in threeor four spatially distinct locations. Each location hasa response key associated with it. As the sequence ispresented at each location, the participant presses thecorresponding key. Unbeknownst to the participant'the stimuli can be presented at the various locationsaccording to a repeating sequence. Learning is dem-' onstrated by a reduction in reaction times as the taskprogresses; when a new sequence is presented, reactiontimes increase again. Sequence learning occurs evenwhen the participant is unable to explicitly predict thenext location in the sequence and, for this reason, isthought to be a form of implicit learning.

Artificial Grammar Learning (AGt) ParadigmAnother paradigm often used to examine implicitsequence learning is Arthur Reber's artificial gram-mar learning (AGL) paradigm. The original taskinvolved presenting subjects with a string of printedletters, but the task has also been presented as asequential auditory task using nonsense syllables.ln Reber's original AGL study, adults were presentedwith "sentences" made from various combinationsof five different letters, six to eight letters in length.Subiects were asked to memorize the sentences butwere not told that they had been given either ran-domly produced sequences of letters or sequencesthat had been generated based on a complex set ofgrammatical rules. Subjects who had "grammatical sentences" learned them faster than those withrandomly generated strings of letters. In addition,after learning grammatical sentences, subjects weretold that the sentences followed a set of rules andwere asked to categorize new sentences according towhether or not they followed the rules. Despite beingunable to verbalize what the rules were, subjects cat-egorized strings of letters according to those rules atabove-chance levels. Like the SRT paradigm, learn-ing in AGL is thought to be implicit. This ability toextract rules based on Patterns in the environment isassumed to be an important component of how chil-dren learn language in a mostly incidental manner.

Infants and Young ChildrenAnother form of the AGL task that is even more akinto language learning has shown infants as young as5 months to be sensitive to statistically derived pat-terns ofauditory stimuli. In this case, participants hearsequences of nonsense syllables. The order in whichthe syllables are presented is governed by a grammarthat statistically determines the probability of whatsyllable will follow another, given the preceding #-lables. Infants, like adults, are capable of categorizingnovel grammatical and ungrammatical sequences atterlistening to grammatical sequences lor just a few min-utes. Human infants appear to be quite adept at thrstlpe of sequential-statistical learning. Impressively, thelearning takes place even after very limited exposureto the input and also occurs at very early ages, wrtnlittle or no development across the life span. For thesereasons, it is likely that these learning mechanisms areused in the service of language learning.

In fact, infants and young.hfld..n h"u. b"tnshown to be capable ofa whole host ofskills requiring

Auditory Sequence/Artificial Grammar Learning 35

sorne level of understanding of sequential patterns'ii"...tng" from categorizing novel versus familiar,*u"naat p."t"nred in a single--sensory modalily, tomut.tting sequences across different modalities, to,"qu"naing their own motor actions in naturalis-ii.'r"ttingi. However' relatively,little is known aboutrhe deveiopment of sequcntial learning skills' Forsequence learning with visual stimuli, findings aresomewhat ambiguous. The ability to discriminatesimple, familiar, statistically determined sequencesfrom unfamiliar sequences aPPears to be present by2 months and to remain stable with no developmentacross infanry However, not all methods of testinghave found clear evidence ofvisual sequence learningin all infants. It may be that the complexity of the taskor mode of response is relevant. In contrast, studiesof older children and adults have found age-relatedimprovement in visual sequence learning. Althoughboth adults and children ages 7 to 11 years learn visu-ally presented sequences, adults have been shown todo so with greater speed and magnitude.

Development of Auditory Sequence LearningFindings on the development of auditory sequencelearning are less ambiguous. Infants appear to be sen-sitive to auditory sequence information in the formof rhythm, tones, and nonsense syllables by 5 monthsor earlier. By 8 months, they can learn complex tran-sitional probabilities such as are present in the AGLtask, and this ability remains stable across the lifespan with little or no development. The discrepancyin the development of visual and auditory forms ofsequential learning (i.e., that visual sequence learn-mg appears to remain static across infancy and thenrmProves from childhood to adulthood, while audi-tory sequence learning as measured by the AGL taskdoes not appear to develop at all) may be explainedby two sound-related ohenomena.

First, some research has suggested that there is anauditory dominance in memory which may includeimplicit memory in early childhood that shifu to avrsual dominance in later childhood and adulthood.For example, when visual and auditory stimuli arepaired and participants are subsequently tested for rec-ognltton of original pairs with correct pairings, repair-utgs of original stimuli, or pairings that include eithernew visual or new auditory stimuli, children aroundne age of4 years recognize when the auditory stimulusts new more reliably tJran do adults. On the other hand,adults and older children perform better on the visual

stimuli. Thus, it is possible that this auditorybias mightexplain the different developmental patterns observedfor auditory versus visual sequential learning.

The second sound-related phenomenon is thatthere appear to be modality constraints for the learn-ing of sequential and spatial patterns: an overall audi-tory superiority for temporal tasks, such as sequencelearning, but a visual superiority for spatial tasks.When adults are exposed to stimuli that are audi-tory (sequences of tones ofvarying frequency), visual(black squares presented sequentially at differentlocations), or tactile (sequences of vibrotactile pulsesdelivered to different fingers) that are generated by anartificial grammar, they show twice as much learn-ing for auditory sequences than for visual or tactil€sequences. In addition, with faster rates of sequencepresentation, visual learning declines, whereas audi-tory learning is not affected. It is likely that the auditory superiority effect seen in sequential learning isrelated to the fact that sound itself is inherently a tem-porally arrayed, sequential signal in which elemcntsare defined specifically by the timing or serial order inwhich they are presented.

Despite these modality constraints, sequentiallearning also appears to be a domain-general cog-nitive mechanism that is used across a number ofdomains. The link between sequential learning andlanguage development specifically is supported byevidence that infants' visual sequence learning abilityis correlated with both their receptive language abil-ity and their ability to communicate using gesture.These findings are in agreement with those showingthat visual sequence learning ability is correlated withlanguage processing ability in adults and in childrenwith hearing impairment. Language learning is gener-ally viewed to be an unconscious process, with peopleoften using and understanding language without anexplicit understanding of the grammatical rules thatgovern it. Thus, it is likely that language is acquiredthrough implicit learning mechanisms, such assequence learning.

As mentioned earlier, the temporal and sequential nature of sound and the superiority of auditionfor processing sequentially presented informationmay make sound and hearing the primary means forunderstanding temporal and sequential events. Thus,hearing may provide critical exposure to sequen-tially presented information, which may help scaf-fold the development of sequential learning mecha-nisms. As such, early sound deprivation may result

36 Autism and Language Development

in domain-general disturbances to sequence learn-ing across all sensory modalities. Because so muchdaily information is presented sequentially and somany behaviors depend upon one's ability to inter-pret and produce sequential behaviors, impairmentsto sequence learning abilities could have far-reachingconsequences, particularly for language learning.Research, in fact, corroborates these extrapolations.

Not only do deaf adults show disturbances to non-auditory functions related to time and serial order, butchildren born deafwho later gain some level of hear-ing through cochlear implants show greatly depressedability to complete both motor learning tasks andvisual sequential learning tasks compared to normal-hearing peers. In addition, deaf chiidren who gainhearing through cochlear implants vary greatly intheir ability to develop spoken language after implan-tation. While some catch up to their normal-hearingpeers, others continue to display languageJearningdeficits that cannot easily be explained by early lackoflanguage exposure alone. It is possible that individ-ual differences in natural sequence learning abilitiesand the extent to which they were disturbed by earlysound deprivation could lead to profound differencesin the ability to learn spoken language after implan-tation. This notion is supported by research showingthat implicit sequence learning abilities are correlatedwith standardized measures oflanguage outcomes forchildren with cochlear implants.

Finall* there is also evidence that variations insequential learning skill may contribute to the defi-cits observed in certain language and communicationdisorders such as specific language impairment (SLI),dyslexia, and autism. For example, children with SLIlearned the pattern of a visual SRT task slower thantypically developing (TD) peers. Once the patternwas learned, children with SLI who had grammaticaldeficits continued to have slower reaction times thanTD peers, while children with SLI who had vocabu-lary deficits did not, suggesting a specific link betweensequence learning and grammar skills.

In addition, children with SLI need twice as muchexposure to learn transitional probabilities in an audi-tory AGL task as TD peers, and they show even morediffrculty learning sequences of tones. Research onSRT tasks with dyslexic individuals has been mixed,but they seem to show impaired sequence learn-ing and increased spatial context learning comparedto TD peers. In addition, dyslexic students' readingabilities are positively correlated with their sequence

learning abilities and negatively correlated with theirspatial context learning. Thus, it is not a general cog-nitive or even implicit learning deficit but a specificdeGcit in implicit sequence learning that appears to berelated to poor reading in individuals with dyslexia.

Joanne A. DeocampoChristopher M. ConwayG eorgi a St ate Univers ity

See Also: Distributional Knowledge and LanguageLearning; Interrelationship of Language and CognitiveDevelopment (Overview); Language Development inChildren With Cochlear Implants; Prediction in LanguageLearning, Role of; Statistical Learning.

Further ReadingsConway, C. M. and M. H. Christiansen. "Modality-

Constrained Statistical Learning of Tactile, Visual, andAuditory Sequences )' Journal of Experimenul Psychologr:Learning Memory, and Cognition,v.3l (2005) .

Conway, C. M., D. B. Pisoni, and W G. Kronenberger."The Importance of Sound for Cognitive SequencingAbilitiesl' Current Directions in Psychological Science,v.l8/5 (2009).

Nissen, Mary Jo and Peter Bullemer. "AttentionalRequirements of Learning: Evidence From PerformanceMeasures." Cognitive Psychology, v.l9 (1987).

Reber, Arthur S. "Implicit Learning ofArtificialGrammarsl' Journal of Verbal Learning and VerbalBehavior, v.6 (1967)-

Robinson, Christopher W. and Vladimir M. Sloutsky.'Auditory Dominance and Its Change in the Course ofDevef opment." C/rild Development, v.7 5 I 5 (2004) -

Autism and LanguageDevelopmentCommunication impairments have been recognizedas one of the core features of Autism Spectrum Dis-orders (ASD) along with social deficits and restrictedand repetitive behaviors. Approximately 20 percent ofthe ASD population does not acquire any functionalexpressive language. Communication deficits in ASDinclude a variety of impairments such as use ofstereo-typed speech or delayed echolalia (e.g., repeating linesfrom a favorite cartoon) and difficulties initiating and