Lecture 8- Four Debates
Lecture 8- Four Debates PSY/LING 34
Popular in Language Development
Popular in Linguistics and Speech Pathology
This 46 page Class Notes was uploaded by Texana Sonnefeld on Sunday March 8, 2015. The Class Notes belongs to PSY/LING 34 at University of Arizona taught by Louann Gerken in Spring2015. Since its upload, it has received 53 views. For similar materials see Language Development in Linguistics and Speech Pathology at University of Arizona.
Reviews for Lecture 8- Four Debates
Report this Material
What is Karma?
Karma is the currency of StudySoup.
You can buy or earn more Karma at anytime and redeem it for class notes, study guides, flashcards, and more!
Date Created: 03/08/15
DISCLAIMER This note outline is made by LouAnn Gerken All text added in blue is mine Gerken L 2015 Mar 5 Four Debates University of Arizona Tucson AZ 4 ebafes 4 ebates 1How Far do children generalize 2Can associative models work can you get by without grammar 3Do children make generalizations For which they have no evidence 4Do children39s errors re ect human languages 1 How Far do children generalize Associa rionism no r very Far un ril lhey have a of of inpu r examples NoGrammar generaliza rion is based on parlicular words 10 Sfarf generalization on specifics of input Hypo rhesis Tesling and Triggering children are generalizers bu r rhe bases differ For rhe rwo lheories genera39izemuesses This deba re oF ren Focuses on verb learning 2 Can Associative models account For overgeneralization errors can t account with that without grammar Past tense error model 906d instead of went Hypothesis Testing and Triggering say no 3 Do children make generalizafions For which fhey have no evidence Associafionism and Hypofhesis no learningfrom environment g g n g s S innate they are born with knowledge to form generalizations 4 Do children s morphosyniaciic errors re ecf possible human grammars Associafionism and Hypoihesis Selecfion say if s possible ihaiL rhey won39f Triggering says they mus r This debate has an impac r on how we can in rerpre r errors eg Subjecfless senfences wanf ice creamquot Pronoun case errors her likes if 1 How Far do children generalize Verb Island Hypothesis Associationism they aren t generalized into all verbs won t assume learn verbs one at a time hit also means run Young children rst learn thematic roles associated with particular verbs They later collapse across similar verbs and create more abstract categories like subject direct object etc Eg giver GIVE givee given I gave Zelda a book caller CALL callee callen I called Rodney a Fool thrower THROW throwee thrown I threw him an egg givercallerthrower become agents and subjects verbs need to be put together you can make inferences about where a verb falls in a brand new sentence Evidence For verb islands in VThe verbisland hypolhesis slales lhal Akh rar and Tomasello 1997 children s early grammars consi51L of Look a r wha r Ernie s doing ro Ber r se5 of lexicany I r s called dacking speci c predicafe What s happening S39I39T UC39I39UT QS or verb islands Expec red answer Ernie is dacking Ber r Children a r 210 Fail ro use We new verb in a new sen rence rype A correc r answer requires a se r oF inferences Ernie is doing some rhing ro Ber r Tha r some rhing is called dacking Ernie should be We agen r and Ber r rhe rheme oF dackingquot da ve double dative Evidence against verb islands in production Conwell amp Demuth 2007 doing it based on the restrictions on their language but good at generalizing in this structure 3yearolds 08 am pilked Ernie the ballquot E double object 939V9 E 096 ill g Bert gorped the ball to Ernie E m i H O prepositional donate a I donated a statue to the museum righ I donated museum a statue wrong 02 Children trained on double 39 object Form generalized to a preposition and vice versa learned one verb in one orm Bias toward prepositional Form CI innovative l lMadelIilce vhite new form gtlack form they learned Double Obledl Prepositional Model Farm when they learned pilked they equally used both versions but did it more in the other way the properties oF English doubleobjects do not work with certain verbs eg Bert donated Ernie the ball if they learned in the prepositional way they usually used the model version added syntacticsemantic transparency because oF the presence oF a preposition Evidence against verb islands in comprehension Naigles 1990 IPLP 25montholds The duck is gorping the bunnyquot Verb learning suggests that they have some idea about subjectagents and direct objectthemes even when a verb is heard for the first time they have some sort of guess as to what it is YES they are good generalizers with verbs How Far do children generalize Summary Although there is some con icting evidence it appears that children as young as 2 years m generalize across verbs However their willingness to do so may depend on the amount and potentially type oF evidence that they have been given and also the particulars oF the task less willing to generalize in the first type of example but in the last two they are 2 Can associative models work models that don t assume they learn a grammar One oF the primary examples oF children s ability to generalize beyond the input they get are past tense errors like goed Learning appears to come in three stages pretty Pity REGRESSIONS when learning phonology of the language pretty There are two main accounts oF how these errors arise Past tense errors Grammatical view On the grammar view past tense errors re ect the Formation and overapplication oF a rule Early performance re ects prerule stage The past tense rule needs to account For the Following Forms snow gt snowed d hug gt hugged d kiss gt kissed t bat gt batted ed nod gt nodded ad What s the rule if it ends in voiced consonant you add d ends in voiceless consonant add t ends in t or C then ed Overgeneralizafion errors Examples From Amahl Smifh 1973 I020428 brokened 39braukand I020513 bringed brigd I corned kAmd I020626 sayed seid 020721 drinked drigkt 39020927 breaked breiktl clothses kbasaz cutten 39kAtanj cuttened 39kAtand leaved lixvd tooken Wukanj 021013 biten 39baitan1 drinked drigkt leaved lixvd says taked teikt Pasl lense errors Grammalical view Pas r rense simpli ed presen r rense ending in voiced phoneme excep r d plus d presen r rense ending in voiceless phoneme excep r r plus r presen r rense ending in r or d plus ad Children apply rhis rule ro all Forms even rhose rha r are excep rions in adul r English eg goed ins read oF wen r Associa rive models canno r easily accoun r For rhis Pasi Tense Errors Associaiionis r View The pas r rense rule is represen red in rhe mind on rhe Hypo rhesis Selec rion and Triggering views no r on rhe Associa rionis r view Wi rhin rhe Associa rionis r view implemented in computer model There are no phonological rules Children s rore connec rions be rween a presen r and pas r rense version of each word Given rhis how can children possibly make over generaliza rion errors Learning view Stage 1 Many irregulars learned Stag IPA transcription Word IPA transcription of past tense go go want take teik tUk have haev haed eat it eit give gIv geiv kiss kIs kIst hug hAg hAgd Learning view Stage 2 More regulars learned IPA transcription Stage 2 Word IPA transcription of past tense snow sno snod bake beik beikt like aik aikt live Iv IIvd save seiv seivd want want wantad scratch skraet skraetJt Sfage 2 Learning Presenf Pasi pairs Present tense Present Tense in Triphones Past Tense in Triphones kiss kI km 15 k1 kIs Ist st hug hA hAc Ag hA hAg Agd gd snow sn sno no sn sno nod od bake bei beik eik bei beik eik kt like lai Iaik aik Iai laik aik kt live II IIV Iv II IIv Ivd vd save sei seiv eiv sei seiv eivd vd want wa wan ant nt wa wan ant nta tad ad scratch sk krae raetj aetj sk krae raetj Itjt tjt the model found the shortcut of errors children make k gt kt g gt gd 0 gt od U gt Ut generalizafion over many examples Stage 3 Irregulars relearned go k kt broke teach g gd dug di9gt 0 39 0d went break U 39 U t taught Generalization ignored For irregulars Then they learn how to do it the right way then A Crificism of fhe No Grammar Approach The model makes weird39 errors rha r probably no would make Predicts thattheywill make odd pasttense errors Presenf fense Pasf fense squaf squakf mail membled four foureder mafe maded Pasi iense errors Summary The pas r rense rule is represen red in rhe mind on rhe Hypo rhesis Selec rion and Triggering views no r on rhe Associa rionis r view Is fhe Associa rionisf view a work in progress more work is needed fo gef rid oF weird errors or Fundamenfally awed fhe weird errors are a sign of a deep problem rha r requires a grammar 3 Do children make generalizafions fhaf aren39f in fhe inpuf lSmom holds Infermodal Preference Looking Procedure Look A yellow boHle Do children make generalizafions fhaf aren39f in fhe inpuf CONTROL CONDITION Now look When do you see now Do children make generolizofions fhof oren f in fhe inpuf NOVEL CONDITION Now look Do you see anofher one meaning another yellow bottle Infan rs rhink rhal one refers lo lhe enlire NP 350 300 250 200 leamiliar Ye 0W I novel blue 150 100 050 Mean Looking Tlme seconds 000 1 Control Ana pnoric look more at the yellow bottle Fig 1 Mean looking lime in seconds to the No text images in each condition From Lidz ef al 2003 Wha r inFan rs know abou r syn rax bu r couldn l have learned Experimental evidence For syn rac ric s rruc rure of 18 mon rhsquot But how could they learn that Look a yellow bottle Can you nd another one could be taken as another bottle or another yellow bottle Because a yellow bottle is also a bottle the learner doesn39t get any direct evidence that one refers speci cally to yellow bottle Don t have to have it as builtin knowledge ggfggvlf r gggom However if you only encounter situations in which one refers to Ejective noun the whole NP and never to just the noun you have good evidence For the hypothesis that one refers to the whole NP Summary of learning errors From no inpuf Children show a remarkable abilily lo inlerpre r language lhe way lha r adulls do From an early age They migh r achieve rhis abili ry wilh inna re knowledge oF language Triggering They migh r ins read use indirec r evidence lack oF anoiher onequot referring lo jus r rhe noun Hypo rhesis Selec rion 4 Do children s errors re ecf human language Subjecfless senfences and Pronoun case errors Children s subjeciless senlences Examples Wan r ice cream Push himquot Triggering 1 Languages ei rher require over r subjec rs eg English or no r eg Spanish 2 Children are born wilh lheir prodrop parame rer se r lo allow subjeclless sen rences Children s subjectless sentences How do children change the setting Failure to hear any sentences without subjects no Eat your peas They heara lot of sentences without subjects Hearing sentences with subjects in which the subject does not refer to anything It39s raining Subjectzthe sky There are three cats on the roof empty subject Key observation Children begin producing subjects at the some time as producing such semantically empty subjects What is the trigger If triggering is right then this will support that argument Subjecfless senfences Inpui based accouni Bu r seniences like If s rainingquot have been in ihe child39s inpuf From bir rh Why do rhey suddenly noiice ihem Fur rhermore children omi r many par rs of words and seniences including sub iecis These omissions migh r re ec r dif cul ries in produc rion no r an incorrec r grammar Three s rudies Subjecfless senlences Inpulbased accounl Valian 1995 Englishspeaking should have IF Englishspeaking children ini rially have rhe same prodrop Z Pa ig r grrg sentences selling as children learning a prodrop language like I ralian bo rh groups should produce pronoun subjec rs equally inFrequen rly A s rudy of lhe spon raneous Speech oF children learning English vs I ralian revealed rha r EnglishSpeaking children produce pronoun subjec rs 5080 of rhe rime I ralian children produce pronoun subjec rs less rhan 5 of lhe rime This is evidence rha r EnglishSpeaking children do no r have rhe same syn rax as I ralian children Subjeclless senlences Inpul based accounl Bu r who r causes Englishspeakers omissions Bloom 1993 Examined rhe produc rions oF Adam Eve 84 Sarah in rhe Brown corpus oF CHILDES Found rho r children were less liker ro produce subjec rs For longer u r remnces rhon shor rer u r remnces Wanf some more juice sentenceswithoutsubjectslonger 0 ll I Sing sentences WIth subject 2 shorter This nding sugges rs rho r children omi r subjec rs due ro cons rmin rs on produc rion Subjecfless senlences Inpul based accounl BumL why do children omhL subjecls in parlicular Gerken 1991 SubjecmL pronouns are unslressed syllables in uHerance inilial posilion hardtime omitting syllables We know lhalL children also omhL inilial unslressed syllables From words like giraffe Could subjecmL omissions be lhe resuhL oF dif cully producing unslressed inilial syllables IF so lhey should also omil inilial deferminers like lhe more lhan deferminers laler in lhe senlence Subjeclless senlences Inpul based accounl 2yearolds were asked ro imifafe senfences like The bear kissed Janequot Jane kissed fhe bear He kissed Janequot Jane kissed him The quesfion was whefher children would omif fhe and he From rhe beginning of sen rences more Frequen rly rhan From rhe middle or end of sen rences Produced 100 75 50 25 0 Data From Gerken 1991 I Stressed I Unstressed Pronoun 3 Subject I Unstressed Determiner Object YES they do omit unstressed syllables at the beginning of the sentence Subjeclless senlences Summary Englishspeaking children s omissions oF subjec rs may be might due ro an incorrec r selling of an inna re parame rer However several s rudies sugges r rha r Children a r remp r ro produce subjec rs consis ren r wi rh rheir inpu r more about the limitations They may omi r SUijC39I39S due 39I39O COHS39I39T ClllTI39S on rhe on the language and not the grammaritse leng rh and s rress pa r rerns oF sen rences rhey produce So rhe problem may in Fac r rela re more ro limi rs on wha r rhey can say and no r rhe grammar Skipped until summary at end Morphological Case Errors You already know lhaf morphological case is a sysfem oF marking which noun in a senfence is rhe agen r rheme elc Many languages like Russian and Polish mark all nouns For case English makes very little use of case marking and only marks the case of pronouns Morphological Case Errors Table 87 English Pronoun Case 1st 3rd 1 st 3rd person 2nd person person person Case singular person singular plural plural Nominative I you heshe we they Objective me you himher us them Genitive my your hisher our their Morphological Case Errors Children learning English make errors in pronoun case Him can l seequot My raked if offquot Our go r one eachquot I pull she ping pong ball I ride he I wan r rheyquot The mos r Frequen r error involves subs ri ru ring objec rive or geni rive case eg memy herher For nomina rive I she Morphological case errors Triggering view Shii rze amp Wexler s hypo rhesis Verb rense is wha r gives case ro pronouns He walks home everyday case in normal clause Jane wan red him ro walk home case in in ni rive His walking home upse r Be rh geni rive case in gerund Children have no r mas rered rense so rhey canno r use if ro give case ro pronouns Him walk home Objec rive case is rhe deFaul r case Wha r me worry Therefore children produce rhe objec rive case ins read of lhe nomina rive case Morphological case errors Learning view Rispoli argues fhal lense and case develop independenlly iF you only look al small developmem al windows area oF disagreemenl Children39s errors are due lo lhe combined slrenglh of objeclive and genilive pronoun Forms which are more similar lo each olher lhan eilher is lo lhe nominalive case The objeclive and genilive Forms overwhelm lhe nominalive Forms especially For cerlain pronouns especially in senlence nal posilion Rispoli s accounl predicls more errors For Feminine lhan masculine which is lrue Morphological Case Errors Learning View Table 87 English Pronoun Case Genitive 1st 3rd 1st 3rd person 2nd person person person Case singular person singular plural plural Nominative I you heshe we they Objective Morphological case errors Summary The Triggering vs AssociaiioniSIL accounfs vary on whefher case and fense are relafed in ch dren The Triggering accoumL has dif culfy dealing wifh children39s producing genifive case The AssociaiioniSIL accoumL makes predicfions abounL relafive error rafes across Forms some of which have been fesfed and shown fo be frue 4 Debates Summary Is there a grammar Children are amazing generalizers contrary to verbisland hypothesis Grammarless approaches to generalization may be Fatally Flawed supporting Hypothesis Selection or Triggering against Associationism Is there innate linguistic knowledge Anaphoric one prodrop and pronoun case data are explainable in a number of different ways The approaches WITH a grammar seem to do better so Hypothesis selection and triggering There IS a grammar a set of generalizations and rules and not what they take away with innate knowledge HYPOTHESIS SELECTION SEEMS TO BE SUPPORTED THE MOST
Are you sure you want to buy this material for
You're already Subscribed!
Looks like you've already subscribed to StudySoup, you won't need to purchase another subscription to get this material. To access this material simply click 'View Full Document'