Formal Approaches to Language Acquisition
University of Haifa, Israel, 7 October, 2007
It is generally agreed that the representations assumed by generative theories cannot be learned from the input. For generative linguists, this fact is a fundamental premise of arguments for the innateness of at least some aspects of these representations: since they cannot have been learned from the input, they must be available a priori. An alternative conclusion, of course, is that we need a better theory - one that does not assume representations that are unlearnable.
In this talk, I argue for a constructivist approach according to which language acquisition involves the acquisition of symbolic units (i.e. form-meaning pairings) of varying degrees of complexity and schematicity. Grammatical knowledge in this framework is captured by constructional schemas - complex units which are at least partially underspecified, i.e. contain 'slots' which can be filled with other units. To acquire such units, children must be able do two things: segment utterances and match chunks of phonology to chunks of semantic structure, and form slots by generalising over fillers attested in a particular position in the unit. Both of these abilities, I argue, are necessary for the acquisition of the lexical representations of verbs and other relational predicates: thus constructivist theories of grammatical development need not postulate any learning processes in addition to those necessary to explain lexical development. I demonstrate that children's utterances can be constructed using units which are derivable from the input, and show how these units change in the course of language development.
The study investigates lexical and morphological development in Hebrew-speaking children and adolescents from four different populations: Normal Language Acquiring (NLA) speakers from middle-high and low socio-economic status (SES), and Language-Learning Disabled (LLD) participants from both SES backgrounds.
Choice of these populations is motivated by the postulate that for children to achieve age-appropriate mastery of linguistic knowledge, they need to have access to an adequate threshold of relevant input, which may be hindered by two major factors: cognitive factors internal to the individual, and socially determined factors deriving from the external environment.
The study was designed to assess the ability to produce Hebrew nouns, verbs and adjectives in a variety of morphological structures and categories. Participants were tested orally and individually. They were presented with sentences containing a target item, and were expected to change it into a word in a different lexical class. For example, deriving an adjective from a noun as in a country that has lots of 'mountain-s' (Hebrew har-im) is ... [hararit 'mountainous'].
Results indicate that lexical and morphological knowledge increases and diversifies with age and schooling in all research populations. The following hierarchy was found among populations: NLA mid-high SES > NLA low SES ; LLD mid-high SES > LLD low SES. The resemblance between NLA L-SES and LLD H-SES groups highlights new facets of the debate regarding the impact of cognitive and environmental factors on the efficiency of the language learning system.
The talk concerns `binomial constructions' in the sense of expressions in which two nouns or nominal elements are linked without an overt predicate (e.g., in bound `construct state' compounds like bigdey^nashim, in genitive and prepositional phrases like bgadim shel nashim, bgadim le-nsashim, and with denominal adjectives as modifiers like bgadim nashiyim. The guiding motifs of the talk are (1) the relation between linguistic typology and language acquisition (2) the different expressive options that emerge as a result of processes of language change; (3) the lexicon-syntax interface in the structure and use of binomial constructions in Modern Hebrew; and (4) the long developmental route from emergence via acquisition and on to mastery of these constructions by Hebrew speaker-writers. The title of the talk is deliberately ambiguous. Its main concern is with language development in children acquiring Modern Israeli Hebrew, so that it is synchronic in focus. Yet it also touches on diachronic changes -- from the morphologically bound `construct state' (smixut xavura) constructions of classical Biblical Hebrew to more analytic, phrasal means for combining and modifying nouns in current usage.
Five types of binomial constructions in Modern Hebrew are presented at the interface between morphology (word-internal structural changes); lexicon (devices for new-word formation), semantics (sub-categorization), and syntax (compositional constructions formed by stringing of two or more nouns). These are analyzed on a continuum from fully lexicalized compounds (e.g., orex^din) via well-established but semantically transparent and syntactically `freer' expression like orex^iton, and on to fully `open-ended' combinations like orex^mesibot.
The focus of the paper is an integrative re-analysis of findings from a data base that includes naturalistic samples of children's speech input and output, structured experimental elicitations, and narrative and expository texts produced in speech and writing -- ranging from the early, one-word stage via late preschool, middle childhood, and on to adolescence and adulthood. Results from these different sources yields a typologically and psycholinguistically motivated and protracted `developmental route' between emergence and consolidation in the structure and use of binomial constructions. This progression is described involves reiterated processes of integration and re-representation of earlier-emerging categories into later-acquired knowledge from initial very early lexicalized compounds to syntactic association of two associated nouns in early child grammar, via late preschool lexico-syntax by morpho-syntactic processes of new-word formation, with command of a full range of morphological changes lasting through grade-school age, and on to syntactic productivity as manifested mainly from high school age, in command of high-register, more formal syntactic constructions, typically in expository and scientific prose .
Implications of the findings for this `long developmental route' are considered in terms of the guiding themes of language typology, the lexicon-syntax interface, and the questions of what is meant by `acquisition' of a given construction or set of constructions in a given language.
A model developed for the assessment of communication and early language will be presented. The model is based on a set of assumptions of the Dynamic System Theory that highlight the importance of developing tools for evaluating fine changes overtime in behavior. The following principles become realized in the model: a) slow underlying changes in the representation of knowledge generate emergent abilities; b) observation in naturalistic contexts allow the accumulation of rich data; c) close examination of details allow for generalization of findings; d) emphasis on interaction and co-construction of knowledge. The model has five modules representing the emergence of speech as an outcome of internal re-organizations. The modules are: relation, communication, cognition, language comprehension, motor planning / sensory integration, and speech. The application of the model in real life situations will be demonstrated with examples from young children (under the age of two) with sever language impairments. Some ideas on ways of enhancing interactions between young children with profound developmental impairments (e.g., Children with Autistic Spectrum Disorders) and their parents will be discussed.
We examined the relative contribution of phonological, lexical, and contextual sources of information to word reading in the two cerebral hemispheres. With behavioral divided visual field studies we show different timelines of meaning activation in the two visual-fields. In the LVF(RH) both dominant and subordinate meanings of homographs are facilitated in both early and later SOAs. In the RVF(LH), both meanings of homophones are facilitated in the two SOAs, whereas for heterophones, facilitation for related meanings is delayed. We show that meaning activation is differentially sensitive to lexical and contextual sources of information in the visual-fields, and that phonological codes provide early sources of constraints only in the RVF(LH). We present a preliminary model for lexical disambiguation in the two cerebral hemispheres that is based on the work of Kawamoto (1993) in which orthographic, phonological and semantic codes are fully connected. The model includes two separate networks. One network incorporates Kawamoto's version, and successfully simulates the time course of lexical disambiguation in the LH. In the other network, the direct connections between orthographic and phonological units were removed. Our results suggest that in the LH meaning activation lags behind phonological processing. The results are discussed in terms of their implications for models of reading and hemispheric contribution to this process.
Syntactic movement is a central notion in syntactic theory, and an important factor in accounting for the acquisition of syntax. There are several distinct types of syntactic movement. The first classification relates to the type of element that moves, phrases undergo phrasal movement, and heads, such as verbs, undergo head movement. These movement types further divide by the position to which the constituent moves: phrases can move to an argument position (an argument movement, or A-movement), to spec-VP or spec-IP, or they can move higher up to a non-argument position, to spec-CP (termed A-bar or Wh-movement). Verbs move to I in order to collect (or check) inflection, but they can also move further to C, a movement that is obligatory in Germanic V2 languages, and is optional and stylistic in other languages, like Hebrew.
A substantial body of research has looked at the ability of children who are in the process of acquiring language to produce and understand sentences that are derived by various types of movement. With respect to A-movement, following the findings about children's difficulty in understanding passives in English (Maratsos et al., 1985), Borer and Wexler (1987, 1992) suggested that the ability to assign thematic roles to constituents that moved to another position in the sentence matures only at around the age of 4 or even 5 years. This hypothesis was termed "The maturation of A-chains". In the talk I will present data from 6 experiments with Hebrew-speaking children, as well as from the analysis of spontaneous speech, indicating that A-chains are actually acquired much earlier even before the age of two years (see also Fox and Grodzinsky (1998), Friedmann (2004, 2007); Friedmann and Lavi (2006); Guasti (2002), Snyder, Hyams, and Crisma (1995)).
The numerous studies that focused on the acquisition of Wh movement found that structures that involve Wh movement do not occur from the onset of sentences production, and that relative clauses, for example, appear around age 2;6 in production (Berman, 1997; Crain, McKee, and Emiliani, 1990; de Villiers, de Villiers, and Hoban, 1994; Labelle, 1990, 1996; McKee, McDaniel, and Snedeker, 1998; Varlokosta and Armon-Lotem, 1998). In comprehension relative clauses are mastered even later, at around the age of five or six (de Villiers et al., 1994; Friedmann and Novogrodsky, 2004, 2007; McKee et al., 1998; Roth, 1984; Sheldon, 1974; Tavakolian, 1981). These findings have been accounted for by reference to children's inability to handle the movement that occurs in this structure. Wexler (1992; and also Guasti and Shlonsky, 1995) suggested that linking operators, operators that must be co-indexed with an antecedent and transfer referential features, mature late. Namely, what matures later is the ability to co-index an operator that moves to a non-argument position with an element in the matrix clause. This is the case in object relatives: an empty operator moves from object position within the embedded clause to spec-CP of the embedded clause, and is co-indexed with an NP in the matrix clause, and therefore it can account also for the late acquisition of object relatives. In the talk I will present experiments that relate to the acquisition of comprehension of object- and subject relatives in Hebrew in sentence-picture matching task and in act-out task in Hebrew-speaking children and the results of several experiments that assessed the production of relative clauses in Hebrew. I will also present data on the acquisition of Wh questions of various types and of topicalized structures.
Studies that explored the acquisition of verb movement (thorough I) to C (V-C for short) showed that structures that involve this movement also do not occur early on. V-C movement was found to be acquired later than V-I (Deprez and Pierce, 1993; Soares, 2003), and in Hebrew after age 4 (Friedmann and Novogrodsky, 2003; Zuckerman, 2001). I will present data on the acquisition of this optional movement in Hebrew via an imitation study.
Finally, not much is known about the order of acquisition of the three types of movement within a child, and whether the order is uniform across children, or whether each movement type kicks in at a different stage for different children. I will present a study that tried to establish the order of acquisition of Wh movement, A-movement and V-C movement within the same child, using an imitation task that included 60 Hebrew-speaking children, each repeating 80 sentences derived by different types of movement. The results indicated quite clearly that movement is an important tool in accounting for syntactic acquisition. Sentences that share the same movement types are acquired together, and although different children acquire movement at different ages, the order of acquisition of A, Wh- and V-C movement is fixed across children. First they acquire A-movement, then Wh-movement, and only then V-C movement.
The discourse about the nature and origins of human knowledge has been going on for about 2,000 years of intellectual history. In the past few decades, linguists and psycho-linguists have joined the discussion, but children's voice is yet to be heard. This talk presents a conceptual framework for the elucidation of children's and adolescents' theories about linguistic knowledge, its acquisition and development. This framework brings together for the first time two central disciplines: Na´ve theories and Linguistic and psycho-linguistic theories and approaches. The connecting link between these two disciplines is linguistic literacy that connects between the two disciplines in regards to the relationship between oral and written language as well as to later language development . The connection to linguistic theories touches upon the nature of linguistic language and its explicit description, whereas the connection to na´ve theories touches upon the influence of literacy on thought and its representational nature. In spite of the fact that these three areas are usually discussed separately, the approach taken here connects between them to clarify the basis for the investigation of naive speakers' theories about language and its acquisition.
All the psycho-linguistic theories and approaches attempt to solve the problem of language acquisition. In spite of the great differences among them and the relative weight that they assign to various issues, they all deal with the nature of linguistic knowledge, with the relationship between language and thought, and with the nature-nurture debate. And yet, theories are not the domain of experts alone. Within cognitive approaches to the development of thought and learning, a conception has grown according to which children organize different areas of knowledge in mental structures similar to scientific theories (Gopnik and Meltzoff; 1999). According to this approach, children hold beliefs about phenomena in the world, that -- were they to be held by scientists -- we would call them scientific theories.
The starting point is thus, that language is not merely a medium of communication, but that it is an area of knowledge and, like in other areas, it can be assumed that children hold theories about language too. To clarify this claim, this talk will present three research traditions that come together to create one basis.
Linguistic approaches to language acquisition presuppose a set developmental course that leads to mature performance. With respect to disordered populations normalcy is checked against achievements, assuming that those reflect a similar-to-the-normal path to acquisition. This assumption reflects the impact of linguistic theories of acquisition as well as the idea of brain plasticity early in development. Karmiloff-Smith and colleagues (2001; 2005) have challenged these assumptions, arguing that in congenital disorders that affect the brain, the developmental course necessarily takes a non-normal route.
Our project was designed to provide data that will bear upon this debate. Free speech from 90 typically-developing children (TD) served to construct developmental curves that trace the acquisition of morphological, syntactic and lexical features, starting with MLU 1.2 and ending when the children reached an MLU of 4.6-6. Five features served as 'core': percent of Hebrew nominal sentences, percent of complex sentences, errors of agreement, tensed marked verbs and dropped subjects (Pro-drop). These were chosen because they featured in theoretical models of normal acquisition. Eight 'peripheral' grammatical features along with vocabulary acquisition were monitored as well.
We are currently engaged in a longitudinal follow-up of children with Williams syndrome or Down syndrome, beginning with two-word combinations. Assuming that extent of delay will differ among children, the following questions are raised:
In the current presentation longitudinal data from four children with Williams syndrome will be presented. Individual developmental curves will be compared and contrasted and tentative answers to the above questions will be suggested.
One of the first inflectional systems to emerge in Hebrew-speaking children as in other languages is that of noun plurals, however full command of the noun plurals system and its items is not achieved before the end of gradeschool and even beyond. The early emergence of noun plurals in toddlers derives from its status as a semantically transparent, predictable and regular inflectional system, with general and obligatory applicability -- hence high type and token frequency. The delay in its mastery derives from the morpho-phonological complexity of the Hebrew plural systems in the two genders, with regular, irregular and partially predictable suffixes and stem changes.
The Hebrew noun plural system is of psycholinguistic importance in single and dual considerations of early inflectional acquisition, but also as the first system that enables children to gain understanding of the complexity of Hebrew linear morphology (Ravid, 2006). The talk will present a cluster of studies using innovative experimental and semi-naturalistic elicitation methods, which will shed new light on the acquisition of noun plurals in Hebrew
The present study examines the prosodic characteristics of the Hebrew vocabulary addressed to children between 0;9 and 3 years of age. A total of 228,946 tokens and 8,075 types consisting of longitudinal speech samples were examined. Words were analyzed according to their prosodic patterns -- number of syllables and stress patterns -- as well as by their lexical category, distinguishing not only between Open Class and Closed Class but, also between a third innovative class referred to as a Between Class (Berman, 2001).
Results indicate that Hebrew CDS consists mainly of mono- and bi-syllabic words (with significant differences emerging between lexical categories), and that the most common stress pattern is word-final (with parallel distributions found for all lexical categories). Additional analyses showed that bi-syllabic verbs take word-final stress, but that bi-syllabic nouns are almost equally distributed in terms of trochaic and iambic stress patterns. Finally, a developmental analysis indicates a significant increase in the frequency of iambic words in Hebrew CDS as a function of age.
We discuss the implications of these findings with respect to the use of prosody for word segmentation and assignment of lexical class in infancy. For example, these findings suggest that relying on stressed syllables as a cue for word beginning may not be an efficient strategy for young Hebrew learning infants as it is for young learners of trochaic languages, since the majority of words in the input begin with a weak syllable. Additionally, our findings indicate that word length may serve as a perceptual cue assisting Hebrew learning infants in distinguishing between lexical classes and thus as a means for tapping into lexicon and syntax.
Joint work with Kenji Sagae, Eric Davis, Alon Lavie and Brian MacWhinney
email@example.com, modified Sunday November 24, 2013.