1Jean Lamarck (1744-1829)Lamarck waswrong aboutgiraffes…Theory ofinheritance ofacquiredcharacteristics…but mayberight aboutpeoplelook!a cowthrough languageand cultureComponents of LanguagePhonemes - basic unit of sound Phonological Development:learning about thesoundsystem of a languageMorphemes - smallest unit of meaningful soundSemantic Development:learningabout expressingmeaningSyntax - rules for combining wordsSyntactic Development:learning rules forcombiningwordsPragmatics - rules of conversationPragmaticDevelopment:learninghow languageis used2The case of word learning…• Children generally speak their first wordbefore they turn 1.• 18 months ~ 75 words.• 2 years ~ 250 words.• 5 years ~ 10,000 wordsThe “gavagai” problem(Quine, 1969):Gavagai!“Gavagai”?3“Gavagai”?Ears?Tail?Fur?Soft?Brown?Mouth eating leaf?Rabbit?Grass?“Rabbit essence?”Or is it- smell (n.)?- smell (v.)?- yellow?- soft? etc., etc.“Flower!”a.k.a. The “indeterminacy” problem.Different theories have differentanswers.How do young children learnto name objects?Nativist (innate knowledge of language)Chomsky, Pinker•Universal grammar - unconscious rules common to all languages.•Modularity hypothesis - the human brain contains an innate, self-contained language module.•Evidence:All children exposed to language learn language; other animalsdo not.Specific brain structures linked to specific language abilities.Children imposing grammatical structure onto simple languagesystem•Problems:Universal grammarFocus is on syntactic development; ignores communicative roleof language.4Nativists on learning words• Syntactic bootstrapping - The child comes to the task oflearning language with knowledge about syntactic kinds(e.g. noun, verb). These are used to constrain possiblemeanings.“This is a daxy one”,“Big Bird is gorping Cookie Monster”• Various Biases and constraints - The child knowssomething about how words map to meaning.–Whole-object constraint (Golinkof et al)–Taxonomic constraint (Waxman et al)–Mutual exclusivity principle (Markman)–Object (shape) and substance (material) biases(Smith et al, Soja et al)Interactionist (communicative functions of Language)Lois Bloom, Tomasello•Language is a social skill.•Children are motivated to interact and communicate.•Evidence:• Infants mostly use language to communicate with other people.• Infants’/children’s sensitivity to pragmatic cues.• Non-language-specific phoneme perception.• Infants and young children accept nonverbal sounds andgestures as labels for objects.•Problem: Nativists believe that sheer focus can not teach complexgrammar. Focuses on semantic development.Interactionists on learning words• Pragmatic cues (Tomasello, Baldwin)– Social context– Joint Attention: both parties engaged in acommunicative act, attend to the same thing (mindreading)– Intentionality: how is the word used?– Linguistic context: how does it appear in a sentence?– Syntactic bootstrapping: use the whole sentence tofigure out the meaning of the word“oops!”ConnectionistMcClelland, Elman•Language develops from strengthening neural connections inresponse to input.•Language learning results from general-purpose mechanisms.•Evidence: Infants can identify structural features of language. General-purpose connectionist models can explainoverregularization, overextensions, etc.•Problems: Most aspects of language have not been modeled yet. Built-in features and assumptions of models.5Connectionists on learning words• General purpose mechanisms - Memory, attention,associative learning.• Structured environment - General purpose mechanismslearning in a structured environment create biased learners,make word learning more