Theories Of Speech Perception

Theories of Speech Perception Motor Theory (Liberman)– Close link between perceptionand production of speech Use motor information tocompensate for lack ofinvariants in speech signal Determine which articulatorygesture was made, inferphoneme– Human speech perception isan innate, species-specificskill Because only humans canproduce speech, onlyhumans can perceive it as asequence of phonemes Speech is special Auditory Theory– Derives from generalproperties of the auditorysystem– Speech perception is notspecies-specific

Wilson & friends, 2004 Perception/pa//gi/BellBurst of white noise Production/pa//gi/Tap alternate thumbs

Wilson et al., 2004 Black areas are premotorand primary motor cortexactivated when subjectsproduced the syllables White arrows indicatecentral sulcus Orange represents areasactivated by listening tospeech Extensive activation insuperior temporal gyrus Activation in motor areasinvolved in speechproduction (!)

Wilson and colleagues, 2004

Is categorical perception innate?

Manipulate VOT, Monitor Sucking

4-month-old infants: Eimas et al. (1971)20 ms20 ms(Different Sides) (Same Side)0 ms(Control)

Is categorical perception speciesspecific? Chinchillas exhibit categorical perceptionas well

Chinchilla experiment(Kuhl & Miller experiment)“pa pa pa pa ”“ba ba ba ba ”

Train on end-point “ba” (good), “pa” (bad) Test on intermediate stimuli Results:– Chinchillas switched over from staying torunning at about the same location as theEnglish b/p phoneme boundary

VOT “identification” by chinchillas(Kuhl & Miller, 1981)

Categorical perception, Take 2 Natural discontinuities in many sensorysystems; many of these are commonacross mammalian species Some stimulus differences are hard;others are easy Language takes advantage of “naturalboundaries”

Categorical Perception &Auditory Theory Categorical perception may arise fromrapid decay of auditory memory– not unique to speech People have some ability to discriminatesounds within a phoneme– judgments may reflect decision process ratherthan perception

Motor Theory versusAuditory Theory Close link between speech perception andspeech production systems– Motor Right! Some properties of speech perception (e.g.categorical perception) general auditoryproperties– Auditory Right! Speech perception probably not innate speciesspecific– Motor Wrong

Comprehension Recognize Word– Phonological Info– Visual Info Retrieve Information– Syntactic Info– Semantic/Pragmatic Info Integrate Syntactic & Semantic/Pragmatic Info Store Gist Representation

Word Recognition Serial– Comprehensioninvolves analysis atseveral different levelsin turn Interactive– Various sourcesinteract and combineto produce efficientanalysisSerialInteractive

Bottom-up Processes Acoustic InfoPhonetic InfoPhonemic InfoWords & Sentences

Top-Down Processes To what extent does knowledge of what speakeris saying impact processes necessary forunderstanding speech?

Phonemic Restoration Effect Legislature Sentences

McGurk Effect

McGurk EffectLips say “ba” /ba/ bilabial /ga/ velar /da/ dentalSound signal “ga”Subjects hear “da”

What’s the relevance? What does this stuff have to do withinteractive vs. serial models? Context Effects– Interactive Models use all sources ofinformation for rapid word ID– Serial Models inefficient & slow

Marslen-Wilson’s Cohort Model Mentalrepresentations ofwords activated (inparallel) on the basisof bottom-up input(sounds) Can be de-activatedby subsequent input– bottom-up(phonological)– top-down (contextual)

Uniqueness and Recognition When we hear the beginning of a word this activates ALLwords beginning with the same sound: the “word initialcohort”. Subsequent sounds eliminate candidates fromthe cohort until only one remains (failure to fit with contextcan also eliminate candidates)t- tea, tree, trick, tread, tressle,trespass, top, tick, etc.tr- tree, trick, tread, tressle, trespass,etc.tre- tread, tressle, trespass, etc.tres- tressle, trespass, etc.tresp - trespass.

Uniqueness and Recognition The uniqueness point is the point at which aword becomes uniquely identifiable from itsinitial sound sequenceE.g. “dial”dayl UP“crocodile”krokod aylUP For non-words there is a deviation point: a pointat which the cohort is reduced to zeroE.g. “zn owble” would be rejected with a faster RT than “thousaj ining”DPDP

Uniqueness and Recognition The recognition point is the point at which,empirically, a word is actually identified Empirical studies show that recognition pointcorrelates with (and is closely tied to) theuniqueness point.– phoneme monitoring latencies correlate with a prioricohort analysis (and one way to recognise word initialphonemes is to recognise the word and to know itbegins with e.g. /p/)

Cohort Model(Marslen-Wilson & Tyler) Words consistent withinput become active– Cohort – set of wordsconsistent with first syllable Words in the cohorteliminated when theybecome inconsistent withinput Words eliminated due tocontextual incongruity Processing ends whenthere is one word left inthe cohort/ka/cat captain catchcapitalism/kap/captain capitalismCommunism is slightlydifferent from /kap/capitalism

Marslen-Wilson & Tyler NormalThe church was broken into last night.Some thieves stole most of the lead off the roof. SyntacticThe power was located in green water.No buns puzzle some in the lead off the text. RandomIn was great power water the located.Some the no puzzle buns in lead text the off.

Marslen-Wilson & 500Monitoring Time

Activation in the Revised antcaptaintimecaptivecaptin

TRACE Like the interactive-activationmodel of printed wordrecognition, TRACE has threesets of interconnecteddetectors– Feature detectors– Phoneme detectors– Word detectors These detectors span differentstretches of the input (featuredetector span small parts,word detectors span largerparts)The input is divided into “timeslices” which are processedsequentially.

Phoneme boundary12PdetectorPPBdetector.3456PPPP.789PP.BBB

If there are feature detectors, can we tire one of them out?

Selective adaptation1. Do phoneme identification test(e.g., “ba-pa” continuum)2. Play a stimulus from one of the endpoints many times (e.g., 100 times)3. Repeat phoneme identification test

Selective adaptationPre-adpatationphonemeboundary% “ba” -100 -90 -80 -70 -60 -50 -40 -30 -20 -10 0 10 20 30 40 50 60“pa”“ba”REPEAT -100 “ba”100 times for one minuteVoice Onset Time continuum

The power was located in green water. No buns puzzle some in the lead off the text. Random . model of printed word recognition, TRACE has three sets of interconnected detectors – Feature detectors – Phoneme detectors – Word detectors These detectors span different