Stager & Werker 1997lpearl/courses/psych156A_2016spring/...Stager & Werker 1997 14-month-olds Here, the 14-month-olds look longer at the “lif” object when it’s labeled “neem”.

Psych156A/Ling150:AcquisitionofLanguageII

Lecture5SoundsofWords

Announcements

• BeworkingonHW1(due4/14/16)

• Beworkingonreviewquestionsforsoundsandsoundsofwords

Wordforms

Computationalproblem:Mapvariablewordsignalstomoreabstractwordforms

fwiends

friends

friends “friends”

What’sinvolvedinwordlearning

Wordlearning:mappingbetweenconcept,word,andword’svariableacousticsignal “goblin”

Timeline

fromCurtin&Zamuner2014

Learningwordformsstartsprettyearly(justbefore6months)Learningword-objectassociationscomesseveralmonthslater(reliablyat12months)

Stager&Werker1997

Learningnonsensewordsthatareminimalpairs(differbyonephoneme):‘bih’vs.‘dih’.Comparingagainstwordsthatarenot:‘lif’vs.‘neem’

“Switch”Procedure:measureslookingtime…bih…bih

Same:bih!

Switch:dih!

Habituation

Test

Stager&Werker1997

…bih…bih

Habituation

Test

14-month-olds

…dih…dih

Same:bih!

Switch:dih!

Stager&Werker1997

14-month-olds

Nolookingtimedifference=14-month-oldsdidn’tnoticethedifference!

Stager&Werker1997

Habituation

Test

8-month-olds&14-month-olds

…bih…bih

Same:bih!

Switch:dih!

Stager&Werker1997


Nodifferenceinlookingtime=14-month-oldsdidn’tnoticethedifferenceagain!

Stager&Werker1997


But8-month-oldsdid!Theyhaveadifferenceinlookingtime.Theylooklongeratthe“bih”objectwhenitislabeled“dih”-sotheymustknow“b”and“d”aredifferent.

Stager&Werker1997

…lif…lif

Same:lif!

Switch:neem!

Habituation

Test

14-month-olds

Stager&Werker1997

14-month-olds

Here,the14-month-oldslooklongeratthe“lif”objectwhenit’slabeled“neem”.Theynoticethedifference.

Stager&Werker1997

Habituation

Test

14-month-olds

Infantsunlikelytoassociatelabelwithcheckerboardpattern(thatis,totreatitlikeawordthathasareferent/meaning)

…bih…bih

Same:bih!

Switch:dih!

Stager&Werker1997

14-month-olds

Here,the14-month-oldslooklongeratthe“bih”“object”whenit’slabeled“dih”.Theynoticethedifference.

Key:Experiment2vs4

Stager&Werker1997

Keyfindings

14-month-oldscandiscriminatetheminimallycontrastingwords(Expt.4)

…buttheyfailtonoticetheminimalchangeinthesoundswhentheyarepairedwithobjects,i.e.,whentheyarewordswithassociatedmeaning(Expt.2)

14-month-oldscanperformthetask,whenthewordsaremoredistinct(Expt.3)

Therefore,14-month-oldsusemoredetailtorepresentsoundsthantheydotorepresentwords?

What’sgoingon?

Theyfailspecificallywhenthetaskrequiresword-learning.

Theydoknowthesounds…buttheyfailtousethedetailneededforminimalpairstostorewordsinmemory.

What’sgoingon?– Isthistrueforallwords?– Whendotheylearntodothis?– Whattriggerstheabilitytodothis?

Whatchildrenmaybedoing

Oneidea:Encodedetailonlyifnecessary

Ifchildrenhavesmallvocabularies,itmaynottakesomuchdetailtodistinguishonewordfromanother.(baby,cookie,mommy,daddy…)

Neighborhoodstructureidea:Whenachildknowstwowordsthatdifferonlybyasinglephoneme(like“cat”and“bat”),moreattentiontodetailisrequiredtodistinguishthem.

Prediction:Thecontentofchildren’svocabularydrivestheirabilitytonoticethedifferencebetweenwordsthatdifferminimally(ex:byasinglephoneme)


Somesupportforthisidea:

Childrenwithsmallervocabularieshavemorehighneighborhooddensitywords(Stokes2010,Stokesetal.2012a,Stokesetal.2012b).Thismayhelpchildrenkeepthewordformsseparate.

Wordsfromdenseneighborhoodsareproducedmoreaccuratelyandwithlessvariabilitythanwordswithsparseneighborhoods(Freedman&Barlow2012,Sosa&Stoel-Gammon2012).




Prediction:Thecontentofchildren’svocabularydrivestheirabilitytonoticethedifferencebetweenwordsthatdifferminimally(ex:byasinglephoneme)

Goingwiththeneighborhoodidea,lookatStager&Werker1997

“bih”and“dih”aretooclose(theydifferonlybyonephoneme),and14-month-oldkidsdon’tknowanywordscloseenoughtomotivateattentiontothe“b”/“d”differencewhenword-learning

Habituation

Test

…bih…bih

Same:bih!

Switch:dih!

Werkeretal.2002:VocabularysizemattersSame:bih!

Switch:dih!

TestStager-Werkertask

20-month-oldsnotice

Werkeretal.2002:Vocabularysizematters


Same:bih!

Switch:dih!

14month-oldsdon’t



Same:bih!

Switch:dih!

17-month-oldsdo



Same:bih!

Switch:dih!

Zoominonthe17-month-olds



Thosewithasmallvocabularylooklike14-month-olds-theycan’ttellthedifferenceforanovelwordtheyhaven’theardmuch.



Thosewithalargevocabularylooklike20-month-olds-theycantellthedifferenceforanovelword,eventhoughtheyhaven’thearditmuch.



Implication:PerformanceonStager-Werkertaskwithnovelwordsdoesdepend(somehow!)onhowmanywordsthechildknows.


Werkeretal.2002:PerformanceonStager-Werkertaskwithnovelwordsdependsonhowmanywordsthechildknows.

Morevocabulary= morenecessarydistinctions?

Implication:Thecontentofchildren’svocabularydrivestheirabilitytonoticethedifferencebetweenwordsthatdifferminimally(ex:byasinglephoneme)

Prediction:Thisshouldapplytofamiliarwordstoo.Specifically,childrenwithsmallvocabulariesshouldhavetroublenoticingphonemicdifferencesinfamiliarwords.

Swingley&Aslin2002:Familiarwordtests

ButEnglish14-month-oldsnoticedthedifferencebetweencorrectpronunciationsandmispronunciationswhenthewordswerefamiliar!

Maybethese14-month-oldsjusthappentohavelargevocabularies?

Swingley2005:Familiarwordsforyoungerchildren

(Dutch)11-month-oldsnoticedthedifferencebetweencorrectpronunciationsandmispronunciationswhenthewordswerefamiliar(HeadturnProcedure:testsabilitytohearsounddifferences)

Swingley2005:Familiarwordsforyoungerchildren

Butthisisbeforethey’velikelylearnedmanywords…soitprobablyisn’tjustthenumberofwordstheyknow(andwhichwordstheyknow)thatdrivesthedetailedrepresentationsofthesoundsinthewords.

Point:Vocabularycan’tbetheonlythingdeterminingchildren’sabilitytodistinguishthesoundsofwords.Sowhat’stheproblemwiththe14-month-oldsintheStager-Werkertask?

(Dutch)11-month-oldsnoticedthedifferencebetweencorrectpronunciationsandmispronunciationswhenthewordswerefamiliar(HeadturnProcedure:testsabilitytohearsounddifferences)

Wasthetasktoohardfor14-month-olds?

Maybetheproblemwiththe14-month-oldinfantswasthattheswitchtaskwastoohard-theyhavetobeveryconfidentthattheclosemispronunciationofthenewword(dihfornovelwordbih)isnotactuallycloseenough

Yoshida,Fennell,Swingley,&Werker(2009)

Whatwouldhappenifwehabituated14-month-oldchildrentheusualwayfortheSwitchprocedure,butthentestedthemadifferentwaythatdidn’trequirethemtobeasconfidentaboutthecorrectpronunciationofaword’sform?

TheVisualChoiceTask“Preferentiallooking”

Atwo-alternativeforcedchoicelookingtaskthatcomparesvisualfixationstotargetanddistractorobjects

Golinkoff,Hirsh-Pasek,Cauley&Gordon1987

“Where’sthedog?”

Familiarobjectbettermatchforfamiliarword

TheVisualChoiceTask“Preferentiallooking”

Atwo-alternativeforcedchoicelookingtaskthatcomparesvisualfixationstotargetanddistractorobjects

Golinkoff,Hirsh-Pasek,Cauley&Gordon1987

“Where’sthetog?”

Novelobjectisabettermatchfornovelwordformandimportantlyfamiliarobjectisapoormatch-infantknowsfamiliarword.

Yoshida,Fennell,Swingley,&Werker2009

“bin” “din”Novellabels

Test:14-month-olds“Where’sthebin?”

14-month-oldinfantslooksignificantlymoreatthecorrectnovelobject-theydohavedetailforwords!

TheproblemwiththeStager-WerkerTask

Maybetheproblemwiththe14-month-oldsintheStager-Werkertaskwasthattheyencodedthephoneticformswithlowconfidence.So,whentestedontheoriginalswitchtask,theydidn’thaveenoughconfidenceintheirrepresentationofthenovelformtorealizeitwasthewronglabelforthenovelobject.

Yoshidaetal.2009:“Callingadinobjectbythewordbinisnotgoodpronunciationtothe14-month-old,butneitherisitcategoricallyincorrect.”

Anothermethodologicalcheck

Fennell&Waxman2010:14-month-oldscanpassthisswitchtaskifthecommunicativepurposeofthenovelwordlabelismademoresalient.

bih…bih…bih…bih…bih…Habituation …Ilikethebih…lookatthebih…

Stager&Werker1997 Fennell&Waxman2010

Issue:Isbihalabellike“toy”?Anexclamationlike“wow”?Somethingelse?

Non-issue:bihisdefinitelyalabelfortheobject.

Test dih! Lookatthedih…

(Thisisfineifitmeans“wow”!) (Thisisdefinitelystrange,giventhehabituation.)



Kitty! Car! Shoe! Bih!

Fennell&Waxman2010

Non-issueagain:bihisdefinitelyalabelfortheobject.

Lookatthedih…

(Thisisdefinitelystrange,giventhehabituation.)

Thecommunicativeintentofthenovelwordcanalsobemadeclearbytrainingitemsthatshowfamiliarobjectsandlabels.



Whenthere’sclearintentforthenovelwordtobealabel,14-month-oldscanpasstheSwitchtaskjustfine.

Kitty!Lookatthebih!

Wow!



Whenit’snotclearthenovelwordisintendedasalabel(infact,itseemstobemoreofanexclamationlike“wow”),14-month-oldslookjustliketheydidintheStager&Werker(1997)experiment.

Kitty!Lookatthebih!

Wow!

Whydoeshavingafamiliarwordhelp?

Idea:Childrenbuildupmoreconfidenceinthewordformthemoretimestheyhearit.

{p/b/d/g}{a/o/u}{l/r}=“pall”,“dor”

… “gull”,“ball”

(p/b}{a}{l/r}=“pall”,“ball”,

… “bar”,“par”

{b}{a}{l}=“ball”

Whydoeshavingafamiliarwordhelp?

Idea:Childrenbuildupmoreconfidenceinthewordformthemoretimestheyhearit.

Also,notallpositionsinthewordarecreatedequalwithrespecttohowwellinfantsrememberthem.Forwordswithmorethanonesyllable,seven-month-olds(Benavides-Varela&Mehler2014)andnewborns(Ferryetal.2015)rememberthefirstandlastsyllablesbest.

http://www.sciencedaily.com/releases/2014/09/140908083348.htmhttp://www.sciencedaily.com/releases/2015/07/150721081725.htm

elephant{ɛ/a/ɔ}{l/r/d/t}{ɛ/a/ɔ}{f/v/p/b}{æ/a/ə}{n/m}{t/p/k}

{ɛ/a}{l/r/d/t}{ɛ/a/ɔ}{f/v}{æ/ə}{n/m}{t/k}

{ɛ}{l/r/d}{ɛ/ɔ}{f}{ə}{n}{t}

…

…

Buildingupanabstractrepresentationfrommanysamples

Thiessen&Yee2010:Earlywordformrepresentationsretaincontextualandperceptualfeaturesassociatedwithchildren’spriorexperiencewithwords.

15-month-oldslearnednovelnamesforobjectsthatbeganwitheither[t]or[d].

dawbowvs.tawgoo

Thiswasmeanttodrawattentiontothedifferencebetweenthesephonemes.



15-month-oldslearnedthenameofanovelobject,called

yad



Similartothe14-month-oldsinStager&Werker1997,whenthisnamewasswitchedtoyat,theydidn’tnotice.

yat



15-month-oldslearnednovelnamesforobjectsthatbeganwitheither[t]or[d].

dawbowvs.tawgoo

Thiswasmeanttodrawattentiontothedifferencebetweenthesephonemes.



Iftheywereabletorepresentthe[d]vs.[t]distinctionabstractly,dawbowandtawbooshouldhelpremindthemthat[d]and[t]aredistinct.So,iftheothernovelobject’snameisswitchedfromyadtoyat,theyshouldnotice.

Buttheydidn’t…probablybecausethisisadifferentacousticcontext(word-initialvs.word-final).



Thissuggeststhey’rerepresentingalotofextracontextualandperceptualdetailaboutthe[d]and[t]examplestheyheard,whichcausesthemnottorecognizethosesounds(andtheimportantdifferencesbetweenthem)whenthey’reusedinthethirdnovelword.



Check:Whenthey’rehabituatedtonovelwordsthatusethesameacousticcontextasthetestword…

boeyadvs.gooyat



Nowtheydobetterattellingthatthiscontrastisrelevantinthesamecontext.

Same Switchyad vs. yat



Thissuggeststhattheyaredetectingthedifferencebetween[d]and[t],butnotattheabstractlevelthatwouldallowthemtorecognizethatdifferenceindifferentacousticcontexts.

Theyhaven’tyetabstractedtothephonemicleveladultsuse.

Recap:Sounds,words,anddetail

Word-learningisveryhardforyoungerchildren,sodetailseemstobeinitiallymissedwhentheyfirstlearnwords.

Whenchildrenaretestedwithavisualchoicetask,theyshowmoreknowledgeofdetailedwordformsthanwhentheyaretestedwithaSwitchproceduretask.

Manyexposuresareneededtolearndetailedwordformsattheearlieststagesofword-learning,sothatthewordformsarerepresentedattheappropriateabstractlevel.

Questions?

YoushouldbeabletodoallthequestionsonHW1andallthereviewquestionsforsounds&soundsofwords.

Stager & Werker 1997lpearl/courses/psych156A_2016spring/...Stager & Werker 1997 14-month-olds Here, the 14-month-olds look longer at the “lif” object when it’s labeled “neem”.

Documents