Evidence for Item Based Development E. Bates and J. C. Goodman, On the Emergence of Grammar From the Lexicon M. Tomasello, The item-based nature of children's early syntactic development 1 Evidence for Item Based Development Introduction What is item-based development? Studies in Lexically Based Grammar Studies in Item Based Development Conclusion 2 Introduction What is Item Based Development? Hypothesis: children’s early utterances are organized around particular words and phrases. Speech is not abstract So why is early speech perceived as grammatical? Children imitate and reproduce adult utterances, appearing to posses a knowledge of grammar 3 Introduction Item based nature is most evident in the use of verbs Children tend to retain sampled sentence structure for each particular verb, hence: A child might use “cut” in the schema Cut___ alone The same child will use in more complex schemas for “draw” like Draw___, Draw___on____,Draw___for___ When children learn the determiners a and the, they use each with a different set of nouns, usually mutually exclusive The cat a dog The house a plate The … a… 4 Introduction Studies in Italian regarding verb use show that: 47% of all verbs used, were used in one form only 40% were used in two or three forms The remaining 13% were highly irregular forms (frequently used by adults), which could not be learned from generalization Study group were 3 children, aged 18 months to 3 years The verbs in question had 6 possible forms (first person singular, second person singular etc…) 5 Introduction Transitive and Intransitive Transitive (Subject-Verb-Object) Intransitive (Subject-Verb) Can children use verbs they’ve heard in an intransitive context in a transitive way? Experiment Children were introduced a novel verb with a picture. For example “The sock is tamming” with a matching cartoon 6 Introduction Later, they were encouraged, with another cartoon, to reply to the question “What is doggie doing?” We could expect the child to say something like “The doggie is tamming the car” This would be creative, as the child has taken a novel verb, and taken it into a new, transitive, context Very few children produced the transitive reply As a control for these results, other children were exposed to the transitive form, and they had no trouble reproducing it We’ll return to these studies later 7 Introduction Why should we care? Chomsky’s nativist approach claims Language acquisition takes place quickly and effortlessly because children have full linguistic competence at birth Language acquisition relies only indirectly on the language they are exposed to Children are creative in early stages, because of innate grammar proficiency Item based development disagrees with this, and questions the presence of adult grammar in children’s language 8 Definitions Grammar Grammar is the discovery, enunciation, and study of rules governing the use of language. The set of rules governing a particular language is also called the grammar of the language. Or as previously described: A set of sentences with a finite structural description. Lexicon What words are, how the vocabulary in a language is structured, how people use and store words, how they learn words, the history and evolution of words, types of relationships between words as well as how words are created. Lexicon is a word of Greek origin (λεξικόν) meaning vocabulary 9 First Debate Grammar from the Lexicon What does this mean? Grammar has vocabulary qualities Grammar and vocabulary are learned the same Same mental mechanisms used for both Chomsky Grammar cannot be learned! (in finite time) It is different from vocabulary 10 Nature of Debate Epistemology Empiricism Vs. Nativism Plato Vs. Aristotle Do we have a special grammar organ, or are we just really smart? Really smart = innate abilities not specific to language Soft empiricist claim 11 Emergentism Emergentism Solutions to a problem are unpredictable We will explore the Emergentist approach Emphasizing the union between grammar and lexicon 12 The Giraffe Is the giraffe’s neck a “leaf eating organ”? Is the giraffe itself a leaf eating organ? 13 Humans and Giraffes Hypotheses: Human grammar has evolved like the giraffe’s neck Human beings have walked into a problem space that other animals cannot perceive Appearance of language applied pressure on neural mechanisms in the brain So… Human beings have symbols for everything When these symbols appeared together, grammar emerged 14 Evidence Two types of evidence grammar and the lexicon same mental systems (neural mechanism) Strong relation between grammar/lexical development Overlap in symptoms of brain damage The same mental systems for grammar and the lexicon have other roles Same mental systems do other things 15 Evidence for Item Based Development Introduction Studies in Lexically Based Grammar Normal Children Atypical Populations Lexicon and Grammar in the Adult Brain Studies in Item Based Development Conclusion 16 Studies Research in normal children Relation between lexical development and grammar complexity Target group: normal children, 8 to 30 months of age Early language development in atypical populations Comparison with normal children Early/late talkers, focal brain injury, Williams and Down Syndrome, SLI 17 Studies Grammar and lexicon in the adult brain Does modularization occur in a later stage? We will examine neurological patients 18 Development in Normal Children General Maturation of speech development (English) Phonology (reduplicative babbling) - 6 to 8 months Meaningful speech – 10 to 12 months Additional 4 to 8 months in one word stage Burst in vocabulary growth (combinations) - 16 to 20 months Second burst, morphological – 24 to 30 months Mastering of morphological and syntactic structures – 3 to 3.5 years Appears like maturation of three mental modules Phonological Lexical Grammatical 19 Grammar, Comprehension Production Zones of acceleration for each domain are separated by many weeks Lets try and find a connection 20 Vocabulary and MLU Correlation between vocabulary and MLU Best indication for 28 month MLU is 20 month vocabulary Correlation is not cause 21 Cross Sectional Grammar Complexity Individual differences around the grammar on vocabulary function are rather small (small s.d.) 22 Grammar and Expressive Vocabulary Tight correlation between grammar and vocabulary Clear dissociation between words comprehended and words produced 23 Grammar and Expressive Vocabulary Fan shaped pattern Implies that word comprehension is a prerequisite for expressive grammar not sufficient Comprehension and production can dissociate Grammar We expect vocabulary to put a ceiling on grammar complexity, until a threshold is reached grammar will then decouple with vocabulary Instead, grammar and vocabulary remain tightly coupled 24 Important Points Study follows children through critical stage in development Is the correlation we found a correlation of grammar with itself? Vocabulary includes many prepositions, articles and other grammatical words Removal of such words yields close results Similar study conducted in Italian Similar results 25 Longitudinal Study We can see that the link between grammar and lexical development extends to longitudinal studies as well 26 Explaining the Link Perceptual bootstrapping Logical bootstrapping Grammatical function words are short, low in stress and difficult to perceive Children cannot understand relational terms, until they understand what they relate to. So, grammar depends on the lexicon Syntactic bootstrapping Children exploit sentential information to extract the meaning of a novel word. Grammar words are thus obtained 27 Explaining the Link Nonlinear dynamics of learning in a neural network Experiments in neural network learning (even past tense learning) has resulted in non-linear curves Lexically based grammar The relation observed would be exactly what we would expect, if grammar is part of the lexicon 28 Development in Atypical Populations We would like to find a pediatric population which displays a dissociation between grammar and the lexicon We shall examine Late and early talkers Early focal lesions Williams Syndrome and Down Syndrome Specific Language Impairment (SLI) 29 Late and Early Talkers Late talkers Early talkers Children of age 18 to 24 months who are in the bottom 10th percentile for expressive vocabulary Children of age 12 to 24 months who are in the top 10th percentile for expressive vocabulary Do grammar and vocabulary dissociate in these two groups? 30 Late and Early Talkers Grammar-on-vocabulary function for two children Age 16 to 30 months Age is a poor predictor of vocabulary and grammar 31 Two Case Studies 32 Two Case Studies MW SW 17 months old, expressive vocabulary 596 words, MLU 2.13 21 months old, expressive vocabulary 627 words, MLU 1.12 Deductions It appears that SW is lagging in grammatical development (just begun combining words). Possible dissociation Despite huge vocabulary, her grammatical level is average for her age However, SW displays advanced morphology (knows falling and fell) Dissociation could be explained by short auditory memory We will use this argument again 33 Early Focal Lesions Assuming different neural mechanisms for grammar and vocabulary We expect to find dissociation between grammar and lexicon, in relation to congenital(Present at birth) brain injuries We also expect conformity with classic adult aphasia studies (discussed later) Delay in grammar development for left frontal damage (Broca’s area) Delay in lexical development for posterior left damage (Wernicke’s area) 34 Early Focal Lesions No evidence in favor of predictions With older children Plastic reorganization of brain for early focal lesions Studies conducted during first stages of language acquisition might prove insightful Lesion site has impact on lingual development Target group: 10 months to 12 years 35 Classic Language Areas Adults suffering damage to Broca’s area usually display inability to produce or comprehend grammatically complex sentences Damage to Wernicke’s area usually manifests in impairment of comprehension, and natural sounding speech without meaning 36 Early Focal Lesions Absence of left right differences Wernicke’s area Absence of global differences Small but reliable disadvantage in word comprehension for right focal damage Left temporal cortex Delayed in expressive language (10 to 60 months) Reliable disadvantage Broca’s area No effects recorded for Broca’s Area Front damage is symmetric for right and left (during 19 to 31 months period) Conclusion: temporal lobe of left hemisphere is critical, but the frontal lobes become involved in later stages 37 Early Focal Lesions Disappearance of left temporal effect Children with (any) early focal lesion rank below average at the age of 5 to 7 years Left lesion disadvantage disappears at this age, indicating that some plastic reorganization has taken place (the discussed Wernicke’s area) 38 Conclusions For Early Focal Lesions Study group ranks well within 10th to 90th percentile of normal population in grammar-on-vocabulary function In a normal group of 19 children, we would also expect 14 children in the outskirts 39 Williams Vs. Down Williams (WMS) and Down (DNS) syndromes Down Both constitute a form of genetically based mental retardation Mean IQ’s between 40 and 60 Contrast in grammar development Language abilities below mental age Severe function word omissions and structural simplifications Williams Below mental age Language abilities surprisingly good compared with other mental abilities 40 Williams Vs. Down When do groups separate? Both groups are late talkers, seriously delayed in word comprehension and production during the infant scale (8 to 16 months in normal children) Though still 2 years delayed in vocabulary, during toddler scale (16 to 30 months in normal children), WMS children display good grammar capabilities (Within 10th to 90th percentiles) DNS children remain at a disadvantage First evidence of dissociation WMS usually score low on visual short term memory DNS usually score low on auditory short term memory A result of perceptual impairment? 41 Williams Vs. Down 42 Specific Language Impairment Definition: A delay in expressive language abilities that is 1 standard deviation below average The term specific may be misleading Low attention span also diagnosed Studies show that grammatical morphology is highly effected This dissociation can also be explained by a difficulty of processing rapid auditory data 43 Grammar and Lexicon in the Adult Brain Conclusions at this point Adult neural mechanisms We’ve seen an interdependence between grammar and lexicon compatible with unified grammar/lexical approach Does modularization occur in later stages This is not incompatible with findings so far We will present two kinds of evidence Neural imaging of lexical and grammatical processing Dissociation between lexicon and grammar in patients with focal brain injury (or lack of) 44 Grammar and Lexicon in the Adult Brain Some points to keep in mind All knowledge is in the brain Short of finding neural activity at birth, can’t know source of knowledge (innate/acquired) Differences in experience must be accompanied in differences in neural activity Different responses to two classes of stimuli would require these classes be associated with different patterns in the brain Different brain activity accounted for Nouns vs. Verbs Animal words Vs. Tool Words High/low frequency words Classifying by brain activity would result in two many “systems” Difficulty in classifying by neural activity 45 Grammar and Lexicon in the Adult Brain Localization and domain specificity are not the same If an area is used for language processing, it does not imply dedication Difficult to prove negative things (like dedication) Broca’s area known to mediate some motor tasks as well as language 46 Definitions Aphasia Aphasia is a loss or impairment of the ability to produce or comprehend language, due to brain damage. It is usually a result of damage to the language centres of the brain (like Broca’s area). 47 Different Arguments Neural Imaging So far no convincing study conducted Evidence of dissociation between grammar and lexicon exist, but vary from study to study Adult aphasia presents a more interesting challenge Damage to Broca’s area known to create grammatical difficulties Is Broca’s area central for grammar processing? Damage to Broca’s area results in processing impairments that transcend language 48 All Aphasic Patients have Lexical Difficulties Anomia Deficit in word retrieval All Aphasic patients have some sort of Anomia Hence, a grammar deficit is always accompanied by a lexical deficit 49 Expressive Agrammatism Studies in English show that patients of Broca’s aphasia suffer from agrammatism, while Wernicke’s aphasia patients do not Agrammatism for Wernicke’s aphasia only detected in highly inflected languages (like German and Czech) English is poor in inflections So, the above hypotheses is a result of studies conducted in English! The following table summarizes agrammatism in different populations 50 Expressive Agrammatism 51 Similar Symptoms Review of the table reveals that Patients with agrammatical symptoms, have similar symptoms relating to the lexicon Patients with omission pattern in grammar, have word retrieval failures (common in Broca’s aphasia) Patients who display word substitution in grammar (in instead of at) also display substitution in vocabulary (paraphasia). Common to Wernicke’s and WMS Etc Results suggest that grammatical and lexical deficits have common cause 52 Receptive Agrammatism Receptive agrammatism Is characterized by a difficulty of processing inflections and closed-class words. More difficulties with noncanonical word order types For example The rabbit is being thrown by the bear is more difficult than the bear is throwing the rabbit present in normals as well, under the influence of noise, or other interference Closed-class words – part of the vocabulary of a language that isn’t likely to change (such as pronouns) 53 Receptive Agrammatism Not unique to aphasia populations, present in normals under adverse conditions 54 Evidence for Item Based Development Introduction Studies in Lexically Based Grammar Studies in Item Based Development Investigating children’s verb usage Building a Usage Based Model Conclusion 55 Recalling Previous Experiment Transitive and Intransitive Transitive (Subject-Verb-Object) Intransitive (Subject-Verb) Can children use verbs they’ve heard in an intransitive context in a transitive way? Experiment Children were introduced a novel verb with a picture. For example “The sock is tamming” with a matching cartoon Later, they were encouraged, with another cartoon, to reply to the question “What is doggie doing?” 56 Recalling Previous Experiment The above experiment was conducted with children aged 2-3 years old Other studies have shown that children of age 3-4, have no difficulty assimilating a novel verb and using it creatively 57 Similar Experiments in English Novel verbs were presented in different sentence frames Presentational construction (This is called groping) Imperative construction (Tam, Anna!) Passive construction (Ernie is getting meeked by the dog) The children were encouraged to produce transitive sentences Children under the age of 3 were very poor in creative constructions 58 Inducing non-Grammatical English Presenting 3 novel verbs Age groups: 2;8, 3;6, 4;4 Verbs introduced One in normal SVO (transitive) form Ernie meeking the car One in SOV form Ernie the car tamming One in VSO form Gropping Ernie the car Almost everyone produced SVO forms with the verb they heard in that form When encouraged to use the incorrect forms The older children corrected the verb to normal transitive form The younger children generally produced the illegal forms in which the verb was introduced These results are inconsistent with an innate proficiency in grammar 59 Additional Attempts Key to Graph 60 Additional Attempts As we can see, creativity improves with age 61 Two Supporting Facts Perhaps young children are reluctant to use novel words in novel ways? Studies show that children freely use novel nouns in novel sentence frames Perhaps children have production difficulties? Children participating in the studies proved no better in comprehension, than they did in production 62 Introduction As we’ve seen, the above results contradict Chomsky’s nativist approach Chomsky’s nativist approach claims that Language acquisition takes place quickly and effortlessly because children have full linguistic competence at birth Language acquisition relies only indirectly on the language they are exposed to. Children are creative in early stages, because of innate grammar proficiency 63 Possible Nativist Retorts Performance Limitations Children have performance limitations that inhibit the expression of innate knowledge However: children display no limitations when learning new nouns, or reproducing familiar sentence frames with novel verbs Genes for adult like grammar turn on later Perhaps early speech is item-based, but appropriate “brain circuits” turn on at later stages Problem: All the above experiments were conducted on the English transitive form, which children could produce with other verbs 64 Usage Based Model Nativist model is lacking Usage Based Model At odds with empirical data A new model is needed This approach tends to characterize a child’s language in terms of cognitive and communicative processes involved Children begin categorizing concrete nouns quite early Only later do children analyze the syntactic structure of their item-based constructions Adult end point Instead of abstract grammar, we have an inventory of symbolic This resembles the lexically based grammar we spoke of earlier resources 65 Usage Based Model Processes involved Imitative learning Children reproduce adult utterances, but not only reproduce, but for the same communicative purpose (they recognize it has meaning) Imitative learning for all constructions Children find abstract categories and schemas Children find patterns Done in concrete nouns quite early (Daddy’s car into Daddy’s_____) Children see both structural and functional similarities in sentences like dad kisses mommy, I hit Jeffery Hypothesis: a critical mass of verbs is necessary 66 Usage Based Model Processes (continued…) Children combine structures and schemas For example, the child combines See___(Mommy/Ball) with Daddy’s___(Car) into See daddy’s car Child must realize, that Daddy’s car is somehow equivalent to Mommy or Ball 67 Evidence for Item Based Development Introduction Studies in Lexically Based Grammar Studies in Item Based Development Conclusion Grammaticalization Lexically based grammar 68 Conclusion So we reject the Nativist Account Where does language come from? Hypothesis Grammatical structures do not come from the human genome Children do not invent grammar A reasonable theory would be that once the homo sapiens learned symbolic communication, a string of successive symbols began to take form. This is called Grammaticalization Grammaticalization processes are well attested to in literature of the recent past 69 Grammaticalization What is it The inventory of symbolic conventions is universal (the existence of a past, all humans have hands etc.) Peculiarities of each language are governed by what that community thinks it’s important to talk about The structures and conventions of a language evolve, adapt and change. This is called Grammaticalization 70 Examples From English The future tense with the word will (used for volition). “I will it to happen” turns to “It will happen” Go was used to indicate movement, so “I am going to the store” turned into “I am going to sleep” The past perfect tense, with have, is most likely to have derived from sentences like “I have a broken finger”, turning into “I have broken a finger” Phrases like “On the top of” or “In the side of” turn into “On top of” or “Inside of”, eventually reducing to “atop” or “inside” 71 Conclusion So how do we account for abstraction? Chomsky noted that abstraction must be contributed from the individual child’s mind (The sentences themselves are not abstract) It is difficult to imagine children applying abstract properties to the language through some innate capability In accord with recent data, it is possible to imagine children using their cognitive and vocal auditory processing skills on the historical product of Grammaticalization 72 The Origin of Language And so we can hypothesize Human language originated from our adaptation to symbolic communication The grammatical structures of modern languages are due to the process of historic Grammaticalization and the analysis of that product using Imitation Schema formation Structural combining Done by separate individuals 73