Haifa Workshop on Psycholinguistics
University of Haifa, Israel, Monday, January 4th, 2010
CLAN tutorial, Tuesday, January 5th, 2010
A considerable body of research showed that children's early multiword utterances are constructed using rote-learned phrases or lexically based patterns with slots. In a series of works, Lieven and colleagues propose two operations, superimposition and juxtaposition, to account for the acquisition of increasingly more abstract constructions. Specifically, Dabrowska and Lieven (2005) show how this model can account for the acquisition of wh-questions in English.
We identify some drawbacks in the proposed model, the primary one being that it is vastly over-generating. There is no principled way to determine which operations should be applied, and how many times. We conduct a computational evaluation of the predictions of the model and show that its ability to generate many observed utterances is unsurprising, since it is similarly able to generate obviously ungrammatical ones. We conclude that more robust evaluation methodology is needed for assessing the performance of language learning algorithms.
Joint work with Sheli Kol and Bracha Nir
Studies examining the role of repeated utterances in child language acquisition have focused to date mainly on the way by which children learn to appropriately re-use the language that they hear and on the different pragmatic functions that such repeated utterances fulfill. However, recent usage-based research has suggested that repetition is not only functionally well motivated but also statistically prominent in child speech (Bannard & Lieven, 2009). The present study examines several sets of longitudinal speech samples in both English and Hebrew. Analysis shows that both in the first and the second year of life, around half of all child utterances are exact repetitions of previously heard utterances -- either of the child herself or of her caretaker; that the bulk of utterances that are repeated can be found in the immediate surrounding context of the same recording session; and that, with age, re-used utterances can be found in previous sessions as well. The theoretical implications of these quantitative results will be discussed.
Mental or cognitive verbs such as think, understand or believe refer to actions and states that are carried out within the mind. As such, they play an important role in the acquisition of the verb lexicon, on the one hand, and the development of Theory of Mind, on the other. This is the first systematic work to investigate the acquisition of Hebrew mental verbs from a psycholinguistic perspective. Our first goal was to reach a working definition for mental verbs, since the literature differs on their nature and functions. This was done in the first part of this study, which is an analysis of Hebrew mental verbs drawn from a current dictionary. This analysis resulted in a corpus of Hebrew mental verb types classified by semantic-pragmatic, morphological and syntactic properties, with a prominent place for binyan functions. The second part of this study examined the occurrence of mental verb tokens and types (as classified above) in the peer talk of Hebrew-speaking children in six consecutive age groups across early childhood. This analysis shows not only an increase in the number and diversity of Hebrew mental verbs between the ages of 2-8, but also clear developmental changes in the distribution of semantic-pragmatic, morphological and syntactic mental verb classes and in their discourse contexts. A dramatic quantitative and qualitative increase occurred at age 5-6, with the advent of `the language of literacy'. Finally, the developmental analysis revealed that in addition to primary mental verbs, all verbs take on secondary mental facets as part of later, school-age lexical development.
Joint work with Topaz Egoz
Objectives: The study examined whether children's phoneme recognition varied as a function of the phoneme's identity: Modern Standard Arabic (hereafter, MSA) phonemes versus Spoken Arabic Vernacular (hereafter, SAV) phonemes. Earlier research has addressed this question using phoneme segmentation tasks, the most widely used tests of phonological awareness. Yet, these tasks require phonological production (Saiegh-Haddad, 2003, 2004, 2007). Thus, even though these studies tested only those children who could accurately articulate the target phonemes, in particular the critical MSA phonemes, the possibility that phonological production at the output phonological stage played a role in the observed difficulty with MSA phonemes was difficult to rule out. In the light of that, the current study aimed to test the effect of phoneme identity (MSA versus SAV) on a phoneme recognition task that does not require phonological production.
A second objective of the study was to test two hypotheses regarding the specific source of children's difficulty in accessing MSA phonemes. The first is the phonological encoding difficulty hypothesis, according to which the difficulty lies in the encoding of the full segmental phonological representations of Standard words in long term memory. If this hypothesis is valid, children are expected to find Standard phonemes harder to recognize when they are presented with phonetically neighboring Spoken phonemes. The second hypothesis is the phonological processing interference hypothesis, according to which the difficulty lies in the reactivation of a fully specified phonological representation and occurs as a result of interference from the Spoken representation of the word. If this hypothesis is correct, phonetically neighboring phonemes should not undermine the recognition of Standard phonemes, especially as no phonological production is required. Instead, cognate phonemes---the specific phonemes that usually correspond to the Standard phonemes in the specific Spoken vernacular---should be more powerful distractors.
Method: The study tested phoneme recognition among 60 Arabic native speaking kindergarten children using a picture selection task that carefully manipulated the type of phoneme targeted (MSA versus SAV) as well as the type of distractor (cognate phoneme versus phonetic neighbor). Findings: The results showed that children's recognition of Standard phonemes was poorer than that of Spoken phonemes. This finding is consistent with earlier research (Saiegh-Haddad, 2003, 2004, 2007) and was interpreted as indicating a deficiency in the phonological representations of Standard words (Elbro, 1996).The results also revealed that children's difficulty in accessing Standard Arabic phonemes was due to a difficulty in the phonological encoding of Standard words rather than due to a processing difficulty (Saiegh-Haddad, Levin, Hende, and Ziv, 2009).
Discussion: The talk will discuss the implications of the findings for language and literacy development in diglossic Arabic (Perfetti, 2007).
Elbro, C. (1996). Early linguistic abilities and reading development: A review and a hypothesis. Reading and Writing: An Interdisciplinary Journal 8, 453-485.
Perfetti, C. (2007). Reading ability: Lexical quality to comprehension. Scientific Studies of Reading 11, 1-27.
Saiegh-Haddad, E. (2003). Linguistic distance and initial reading acquisition: The case of Arabic diglossia. Applied Psycholinguistics 24, 431-451.
Saiegh-Haddad, E. (2004). The impact of phonemic and lexical distance on the phonological analysis of words and pseudowords in a diglossic context. Applied Psycholinguistics 25, 495-512.
Saiegh-Haddad, E. (2007). Linguistic constraints on children's ability to isolate phonemes in Arabic. Applied Psycholinguistics 28, 605-625.
Saiegh-Haddad, E. , Levin, I., Hende, N. and Ziv, M. (2009). The linguistic affiliation constraint and phoneme recognition in diglossic Arabic. Journal of Child Language. In Press.
The talk will focus on two main issues: the comprehension and production of sentences derived by Wh-movement in syntactic SLI, and the syntactic abilities of other subtypes of SLI.
To explore the nature of the syntactic deficit in syntactic SLI, results from comprehension, repetition, reading, and production tests will be presented. The tests were aimed at assessing various types of Wh movement in 20 school-aged Hebrew-speaking children with syntactic SLI (some of them reported in Friedmann and Novogrodsky, 2004, 2007, 2008, in press; Friedmann, Gvion, and Novogrodsky, 2006; Novogrodsky and Friedmann, 2006). Tasks of sentence-picture matching, comprehension questions and paraphrasing of written and heard sentences suggested that children with syntactic SLI have a deficit in the comprehension of sentences that are derived by Wh movement: object relatives, topicalization structures, and object Wh questions. The results also indicated that not all sentences derived by movement are impaired. Whereas sentences in which the moved element crosses another argument of the verb pose severe comprehension problems, sentences that are derived by Wh-movement but maintain the canonical order, such as subject relatives and subject questions are comprehended well. Moreover, when the sentence contains only one argument, or when the other argument in the sentence is of a different type than the moved argument, comprehension is fine, even in sentences derived by Wh-movement. The deficit seems to relate to the assignment of thematic roles to the moved constituent, when this assignment needs to cross another argument of the same type. The SLI deficit in the assignment of thematic roles in manifested in production as well --- children with syntactic SLI tend to produce subject relatives instead of object relatives, and to produce object relatives with only one overt NP, thus avoiding the need for crossing dependencies. The data from sentence repetition and sentence production indicate that unlike thematic role assignment, syntactic structure building, and in particular the projection of CP, are unimpaired in school-age children with syntactic SLI.
The second focus of the talk will be on the modularity within the language system as seen from selective developmental impairments. For this aim, the lexical and phonological abilities of children with syntactic SLI will be explored, and then the syntactic ability of children with lexical, phonological, or pragmatic deficits (LeSLI, PhoSLI, and PraSLI) will be presented. The assessment of lexical retrieval and phonological abilities in the children with SySLI indicated that whereas their comprehension and production of sentences derived by Wh-movement were severely compromised, the lexical and phonological abilities of many of them were unimpaired. Interestingly, for some children with SySLI lexical-syntactic information such as predicate argument structure was unimpaired as well.
The assessment of sentences derived by Wh-movement in 7 children with a lexical retrieval deficit (LeSLI), 7 children with phonological impairment (PhoSLI), and 6 children with a pragmatic deficit (PraSLI) indicated that children with a selective deficit to lexical retrieval, phonological ability, or pragmatic ability can still have normal comprehension and production of sentences derived by Wh-movement. The implications of these findings will be discussed with reference to modularity of language, theories that ascribe the syntactic deficit to phonological or lexical deficits, as well as to the diagnosis and treatment of children with SLI.
Most models of visual word recognition assume that lexical organization mimics the alphabetic principle so that the processing system is tuned to the s linear orthographic structure. Hence, current models of reading share the principle that orthographic similarity is the main constraint that governs lexical architecture and lexical access in alphabetic orthographies. In this context, research in Hebrew, a Semitic language, provides a unique perspective. This is because, on the one hand, Hebrew has an alphabetic orthography where sequential and contiguous letter strings represent phonemes, and orthographic processing in that language should, therefore, be similar to that of Indo-European languages. However, on the other hand, Hebrew has a Semitic morphology where all verbs and the great majority of nouns and adjectives are morphologically complex and are composed of root derivations. In the last decade, our extensive investigation has consistently suggested that Hebrew words are automatically decomposed into their morphemic constituents and that words in the Hebrew mental lexicon are organized by clusters of root morphemes rather than by simple letter-sequences. By this view, the aligning of words one next to the other in the mental lexicon, and the consequent processes of lexical access in two languages that employ the alphabetic principle in writing, but have very different morphological structures may differ qualitatively. We therefore adopt an "ecological" theory of visual word recognition, and suggest that any model of reading should consider first the morphological characteristic of the language.
The talk analyzes responses to single and multiple association tests, part of a larger battery of tests designed to investigate the role of semantic compared with structural factors in speakers' responses to derived nouns in Hebrew in relation to three types of variables: Structural -- root transparency (full or defective consonantal roots), Usage -- familiarity/frequency and concreteness (derived from independently administered questionnaires), and Development of age-schooling level (grade-school, high-school, and adult participants). Semantic-pragmatic responses accounted for around two-thirds of the responses across the population, mitigated by strong effects for all variables, so that, for example, words of low frequency-familiarity elicited relatively more morphologically structure-based associations across the population, and most markedly among younger participants. Implications are discussed in relation to the nature of word-associations as shedding light on the mental lexicon in general and on the processing of derived nouns in Hebrew in particular.
A key assumption underlying effective communication is that speakers of a language tend to use the same words to express certain meanings. Given that linguistic forms are arbitrary and semantically opaque, this assumption of conventionality reduces the effort that speakers must expend in deciding how to refer to things, and also that of addressees in determining what a speaker means. A crucial clause of this assumption, nonetheless, is that speakers and addressees need to differentiate between forms that are conventional from those that are not, and between speakers who are conventional from those who are not. In the present talk, I will present evidence on the emergence in children of this sophisticated assumption of conventionality. The first studies will demonstrate that preschoolers assume that while speakers are expected to know the common names of objects, they are not expected to know the proper names or arbitrary facts associated with novel objects. The second set of studies reveals that children's responses to speakers' request for the referent of a novel name vary according to how knowledgeable of conventional names the speakers are. I will discuss possible mechanisms by which children arrive at this nuanced assumption of conventionality.
The distinction between competence and performance goes back to Chomsky 1965, where he defines competence as an idealized capacity, and performance as the production of actual utterances. Competence, the speaker's knowledge of his/her language, is regarded as stable and unaffected "by such grammatically irrelevant conditions as memory limitations, distractions, shifts of attention and interest, and errors (random or characteristic)" (Chomsky 1965, p.3). Performance refers to concrete realizations of this idealized system. Under this view, competence is primary to performance, and in a way enables it. Without the knowledge of a system, how can there be individual instantiations of that system?
The competence-performance issue will be examined by studying compound formation in ABSL. Comparison across 18 signers (in three different groups) and 51 compound types in ABSL reveals an impressive degree of variation in the language community (Meir et al in press). On average, less than 50% of the signers in each group produced the same form for a specific compound. Only two compounds had one unified form in one of the groups. The number of variants for all other compounds ranged from two to eight. However, we find that some regularities emerge in specific types of compounds: compounds containing a size-and-shape specifier and compounds denoting place names. We also find a relationship between conventionalization and grammaticalization of compounds: as particular forms become conventionalized in the community, both morphological and phonological structures begin to emerge.
These findings are interpreted here as suggesting that language structure starts locally, in very specific domains, driven by properties of specific lexical items. It may be that once a construction is conventionalized, it can serve as a tool or frame for creating new lexical items in a faster way, by adding lexical items into the slots. ABSL compounds, then, may show us how instances of production (performance) may lead to the creation of structure (competence).
Al-Fityani, Kinda (2007). Arab sign languages: A lexical comparison. Technical Report. Center for Research in Language Newsletter, 19, (1).
Chomsky, Noam (1965). Aspects of the Theory of Syntax. Cambridge: MIT Press.
Meir, Irit, Aronoff, Mark, Sandler, Wendy & Padden, Carol (in Sign languages and compounding. In S.Scalise & I.Vogel (Eds.): Compounding. John Benjamins.
Sandler, Wendy, Meir, Irit, Padden, Carol, & Mark Aronoff (2005). The emergence of grammar: Systematic structure in a new language. PNAS, 102(7), 2661-2665.
Sandler, Wendy, Aronoff, Mark, Meir, Irit and Carol Padden (2009). "The Gradual Emergence of Phonological Form in a New Language", manuscript under review.
Construction Grammar has proposed that language is a vast collection of constructions varying mostly in size, range, and specificity from small morphemes to large phrases. I will argue that this view fails to recognize the fundamental computational differences between the ways in which lexical processing in the ventral/posterior stream differs from combinatorial processing in the dorsal/anterior stream. Crucially, the interplay between these two streams involves a carefully orchestrated timed interaction that must be carefully constructed and proceduralized during language learning. Language disorders can be viewed as distributions of the smoothness of interactions between these areas and final motor output.
email@example.com, modified Saturday January 02, 2010.