L'INSTITUT DES SCIENCES COGNITIVES
ACTUALITE DU LABORATOIRE

The CNRS-URBANA WORKSHOP

6-7 December, 1999

Scientific Programme

Monday, December 6

10.00: Welcome of participants. Aims and scope of the workshop

Session 1. Development and learning
10.15: Introductory Lecture by W. Greenough: Experience and brain plasticity. Interpreting the significance of basic research for child rearing and education.
11.00: A. Christophe: Acquisition of phonology: rigidity or plasticity ?
11.30: B Ross: Concept learning and concept use for making inferences and problem solving.
12.00: M. Banich: Functional MRI studies of the Stroop task. Implications for attention and learning
12.30: General disussion
13.00: Lunch
14.00-16.00 Visit of labs

Session 2. Reading

16.00: G. McConkie: Perceptual Continuity Across Saccades: A Stimulus-based Explanation
16.30: F. Vitu: Eye guidance in reading : The role of perceptual and oculomotor factors.
17.00: T. Nazir: Early perceptual learning and reading
17.30: J. Grainger: Investigating letter position coding using masked priming
18.00: General discussion
20.00: Dinner

Tuesday, December 7.

Session 3. Linguistics

9.30: V. Deprez: Morphological plurality and the denotation of noun phrases: implications for the syntax/semantics interface.
10.00:  A. Goldberg: Learning Argument-Structure Generalizations
10.30: Break
11.00: M. Hickmann: Language and space. Developmental and crosslinguistic perspectives.
11.30: P. Dominey: A Neural network model of thematic role assignment in syntactic comprehension:  Predictions for a nonlinguistic expression of agrammatism
12.00: General discussion
12.30: Lunch
 

Session 4: Phonology

14.00: J. Cole: Speech Processing Research:  Evidence for the Development of Phonological Systems
14.30: J.L. Schwartz: How mouth, ear and eye meet language
15.00:  K. Bock: How mind meets mouth.
15.30: General discussion
16.00: Break

Session 5.
16.30: Closing discussion
17.30: End of workshop

"Experience and Brain Plasticity: Interpreting the Significance of Basic Research for Child Rearing and Education"

The last 3 decades have witnessed an enormous increase in our understanding of how the brain stores information. During development, "sculpting" of synaptic neuronal interconnections, regulated by both genetic and experiential factors, generates basic brain organization, upon which experience can further build. The fact that this process involves loss of some synapses early in life has concerned some child-advocacy leaders and educators, particularly in the United States, while evidence that experience induced growth of new connections occurs within the brain has enticed others. One accounting, proposed some years ago by Black and Greenough (e.g. Child Development, 58:539 559, 1987) and subsequently elaborated (e.g., Black, J. E., Jones, T. A., Nelson, C. A. and Greenough, W. T. Neuronal plasticity and the developing brain. Chapter 2 in N. Alessi, J. T. Coyle, S. I. Harrison and S. Eth (Editors), The Handbook of Child and Adolescent Psychiatry, Volume 6, John Wiley & Sons: New York, 1998, pp. 31-53.) proposes separable processes of "experience-expectant" and "experience dependent" "synapse plasticity," the first of which is associated with critical and sensitive periods in development and the second of which is associated with developmental and lifelong learning. The appearance of John Bruer's book on The myth of zero to three has added to the controversy regarding the degree to which the early years of postnatal brain and behavioral development depend upon the proper amount and type of "cognitive" stimulation the infant receives. We will examine implications of our knowledge of experience-expectant and experience-dependent synapse plasticity, in light of broader knowledge of how the brain functions, for views advocating intense or exclusive focus upon this period of development.We will also discuss implications of the addition of non-neural elements (e.g., supportive glial and blood vessel tissues) in response to experience, and experiential factors that govern the addition of these elements and of synapses. Finally, implications of recent findings indicating the addition of new neurons to the brain will be discussed.
 

Anne Christophe
LSCP EHESS Paris, France

"Acquisition of phonology: rigidity or plasticity?".

A body of experimental evidence suggests that very early in life, babies construct a phonological representation that is appropriate for their mother tongue and through which they perceive and process incoming speech. While this representation is probably plastic in that subjects can adapt to some acoustic transormations, it seems to be very rigid when it comes to perceiving and represnting phonological properties that are not relevant in the mother tongue (e.g. stress, syllabic structure, phonemes, tones, etc.). Adults from one linguistic community are functionally 'deaf' to features not used in their language (Dupoux et al. 1997), and even highly proficient bilinguals who acquired their second language by age 3 to 6 years do not generally end up with a phonological representation adapted to each of their languages (Pallier et al. 1997).

While a lot of experimental data is still missing to make this picture complete, the existing data point to an apparent paradox between plasticity on one hand and rigidity on the other. It may be useful to define better what type(s) of mechanism underlies plastic behavior, and perhaps find out what makes a transformation learnable (e.g. one-to-one remapping?) as opposed to unlearnable (e.g. create a new category?).
 

Brian Ross

"Concept Learning and Concept Use"

Concepts are important not just for classifying objects, but also because they allow us to access relevant knowledge that can be used for a variety of functions, such as predicting, understanding, explaining, and problem solving. Much work in concept learning has focused on classification learning -- how people learn to assign instances to categories. Although classification is crucial for many functions of concepts, our learning of concepts often involves much interaction with the instances and use of the concepts. Recent research shows that these uses may affect our conceptual representations, including knowledge involved in classification. I discuss the implications of these results for how we learn concepts and how we represent and use everyday categories.

Marie Banich

"Functional MRI studies of the Stroop Task: Implications for Attention and Learning"

In this talk I present a series of studies that have employed variants of the Stroop task to examine the neural mechanisms underlying attentional control.  This series of studies suggests that there is a network of brain structures involved in attentional control including prefrontal, anterior cingulate, parietal, and extrastriate regions.  Our work suggests that prefrontal regions are involved in creating an attentional "set" that tunes processing to select information that is task-relevant.  The more difficult it is to invoke such a set, the more involved are prefrontal regions.  The anterior cingulate, in contrast, appears to be more involved in selecting among competing responses.  Activity in parietal regions seems to be related to the overall difficulty of attentional selection.  Extrastriate regions become active when the information they process has, through experience, become linked to the attribute that is to be selected.  For example, if the task is to attend to color, object regions will become active if the objects being viewed are associated with a specific color (e.g., the color green with a frog).  The implications of these findings for attention and learning will be discussed.

George W. McConkie
Beckman Institute and Department of Educational Psychology University of Illinois at Urbana-Champaign

Perceptual Continuity Across Saccades: A Stimulus-based Explanation

 Each time a saccadic eye movement is made, the stimulus array is displaced on the retina.  Yet our conscious experience is one of smooth perceptual continuity, devoid of the discrete discontinuities occurring on the receptor surface.  Traditionally this has been explained by postulating a collateral signal that accompanies each saccade, which is used to remap retinal space during the eye movement, thereby spatially justifying the pre- and post-saccadic images. I will report a series of studies indicating the importance of stimulus factors in achieving this perceptual continuity and a surprising insensitivity to pre- and post-saccadic image differences.  Based on these results, a Saccade Target Theory has been proposed as an alternative, stimulus-based explanation of the continuity of perception across saccades.

Françoise Vitu
CNRS, Université René Descartes

Eye guidance in reading: The role of perceptual and oculomotor factors.

In reading, our eyes progress on the line of text by making forward saccades of variable sizes that mostly go from the center of one word to the center of the next word. On some occasions, a word is skipped, or the eyes land at the very-beginning or very-end of a word. On other occasions, the eyes may return to a previous word with a regressive saccade. The question applies of whether the on-going perceptual and linguistic processes associated with the encountered words and sentences are responsible for this variability of the ocular behavior.

In the present talk, I will present a series of recent studies that provide evidence for an influence of ongoing processes on eye guidance in reading, but that stress out at the same time the more fundamental role played by low-level visuo-motor factors associated with saccadic programming. While low-level factors would form the basis for a forward scanning of the text-to-read, higher-level perceptual and linguistic factors would intervene only occasionally, to either speed up the forward scanning, or slow it down with the execution of regressive saccades.

Tatjana Nazir
Institut des sciences cognitives - Lyon

The role of the "visual word form system" in reading: Evidence from performance of a split-brain patient.

Ever since Dejerine (1892) it is established that inferior temporal structures of the left hemisphere - which today are referred to as the "visual word from system - play a crucial role in visual word processing. Damage to these cortical structures correlates with the selective loss of reading skills and brain imaging studies with healthy subjects indicate that these regions are active during processing of "readable" materials. The exact role of these regions for the reading skill, however, is not well understood. By analyzing the performance of a split-brain patient during reading of words that are presented to either of his two hemispheres the present work aims at clarifying this issue.

Jonathan Grainger
Laboratoire de Psychologie Cognitive
CNRS & University of Provence, Aix-en-Provence

Investigating letter position coding using masked priming.

        A series of lexical decision experiments used the masked prime paradigm to study letter position coding in printed word perception. Prime stimuli for 7-letter target words were formed from the first letter, the last letter, and a varying combination of internal letters and minus signs (e.g., B-LCO-Y for BALCONY). Significant priming relative to an unrelated prime condition was obtained when three adjacent internal letters of targets were present in the primes. The precise location of the triplet in the target word or the prime string did not influence priming, thus providing further evidence for relative as opposed to absolute position coding of letters. However, the precise order of letters in the triplet was critical, as was the position of the first and last letters. These results are in favor of a letter position coding scheme for long words that codes the identity and precise location of the first and last letters in a string, plus all possible internal trigrams in the string.

Viviane Deprez
Institut des sciences cognitives - Lyon

MORPHOLOGICAL PLURALITY AND  THE DENOTATION OF NOUN PHRASES:
IMPLICATIONS FOR THE SYNTAX/SEMANTICS INTERFACE

Chierchia (1998) proposes a semantic typology of bare nominals in natural languages. At the basis of this typology is a semantic parameter that governs the mapping of semantic types to syntactic categories and determines whether bare nominals are directly interpretable as arguments, as predicates or as a combination of the two on the basis of the distributional characteristics of bare nominals and the presence or absence of plural morphology in given languages.  This paper examines the semantics and syntactic properties of Haitian Creole bare nominals in view of this typology. They are shown not to fit the typology and to raise problems for the proposed semantic parameter. A syntactic proposal is put forth, which if correct, suggests that the source of crosslinguistic variation in bare nominals is more likely to reside in the morpho-syntax of nominal projections than in their semantics.

Adele Goldberg

Learning Argument-Structure Generalizations

Existing studies, both diary and experimental, establish the existence of two types of argument structure generalizations:the generalization from utterances to verb-centered categories, and the generalization from verb centered categories to more general linking patterns.  The present work provides an explanation as to why these two levels are learned, and how the two levels can be learned.   It is argued that both levels offer a degree of predictive power in interpretation, with the more abstract complement configuration providing a better predictor of overall event-type than the morphological form of the verb.  Light verbs are discussed in relation to how the particular meanings associated with linking generalizations can be learned from the verb centered categories.Parallels between the proposed account of how people learn argument structure generalizations and general categorization principles are made explicit.
 

Maya Hickmann
Laboratoire Cognition et Développement, CNRS UMR 8605
Université René Descartes, Paris V

Language and space: developmental and crosslinguistic perspectives

Spatial representation is a fundamental aspect of the everyday behavior of all living organisms. Until recently, most researchers agreed to view the development of human spatial cognition as being based on innate capacities and/or gradually constructed over time along a universal course that is language-independent. Recent crosslinguistic research, however, has begun to partially question these conclusions on the basis of findings suggesting that variable aspects of linguistic systems have an impact on the rythm and course of language acquisition, as well as on cognitive organization itself. Such findings have been reported in relation to several aspects of spatial representation (e.g., location, motion), for different developmental periods (from the emergence of language to the latest phases), in studies focusing on different levels of linguistic organization (sentence, discourse), and on the basis of different types of data (naturalistic evidence, comprehension and production experiments). This growing body of evidence has revived some versions of the ‘linguistic relativity hypothesis’, according to which the general properties of language and the particulars of a given language influence which aspects of reality are more salient or accessible to children and therefore how they select or organize spatial information during the course of development. My aim will be to suggest some of the main implications of such a hypothesis in the light of several controversies currently debated in the cognitive sciences and to propose a number of directions for future interdisciplinary research (e.g., cognitive psychology, linguistics, neurosciences) aiming at addressing these issues.

Peter Dominey
Institut des sciences cognitives - Lyon

A neural network model of thematic role assignment in syntactic comprehension: Predictions for a nonlinguistic expression of agrammatism

Thematic role assignment (i.e. determining who did what to whom) is a central function of the human linguistic capability.  While two sentences “John introduced Mary to Bill” and “Mary was introduce to Bill by John” have quite similar meaning, their surface structures are quite different. It appears that the ability to solve this parsing problem relies in large part on the distinction between open class words (e.g. nouns, verbs, adjectives) which have semantic content, vs. closed class words (prepositions, determiners, etc.) that have syntactic content that guides the parser.  I have developed a “dual” system model of thematic role analysis in which the application of syntactic transformations on open class items (nouns) is guided by temporal patterns of closed class items in the sentence.  The model has strong neurophysiological foundations (Dominey et al. 1998), and can be used to explain important aspects of infant’s sensitivity to linguistic structure (Dominey & Ramus 2000).  When trained, the model demonstrates a robust thematic role assignment capability, and can also be used to explain performance of Broca’s (agrammatic) aphasics.  More interestingly, the model predicts the existence of a non-linguistic expression of agrammatism, that has recently been confirmed in my laboratory (Dominey & Lelekov 2000).  The model thus provides a starting point for further developing our understanding of the initial state, the final state, and acquisition of language.

Dominey PF, Lelekov T, Ventre-Dominey J, Jeannerod M  (1998)  Dissociable Processes for Learning the surface and abstract structure sensorimotor sequences.  Journal of Cognitive Neuroscience, 10(6), 734-751.
Dominey PF, Ramus F (2000) Neural Network Processing of Natural Language : I.  Sensitivity to serial, temporal and abstract structure of alnguage in the infant.  Language and Cognitive Processes, In press.
Dominey PF, Lelekov T (2000) Non-Linguistic Transformation Processing in Agrammatic Aphasia.  Comment on Grodzinsky “The Neurophysiology of Syntax:  Role of Broca’s Area”  Behavioral and Brain Sciences, February 2000, In Press

Jennifer Cole
Department of Linguistics  and Beckman Institute University of Illinois at Urbana-Champaign

Speech Processing Research:  Evidence for the Development of Phonological Systems

 Phonology involves the study of the sound patterns of individual languages with the goal of understanding the characteristic properties of spoken language, including the nature of the cognitive representation of speech sounds, the mechanisms by which spoken language is processed and stored, and the interface between the sound system and other aspects of grammatical structure. An important question for phonology is to what extent is the phonological system shaped by constraints on the peripheral systems that process speech: the systems of auditory perception and speech articulation. While there is substantial evidence for an auditory and articulatory basis for some sound patterns in languages of the world, the mapping from the auditory and articulatory systems to the phonological system is not direct. Many complex patterns of speech perception and production are not reflected in phonological patterns, while at the same time some phonological patterns defy a reductive explanation in terms of speech processing.

My research addresses the question of explanation in phonology through studying evidence from speech processing. In this talk I report on experiments underway in my laboratory, and with colleagues at UIUC, to investigate speech perception and production for evidence of speech processing influences in phonology. In speech perception, we are examining contextual effects on consonant and vowel place identification in noise as a possible source for phonotactic constraints. In a number of distinct projects involving the acoustic analysis of speech, I am looking at the factors that govern systematic variation in speech production and relationships to factors that condition phonological sound change.  These studies examine consonant production in Spanish, as well as pitch and intonation in Japanese, Basque, Serbo-Croation and English. I am also involved in experimental research on speech production that employs psycholinguistic methods to examine evidence for grammatical constraints on phonology that influence speech production strategies. The research collectively suggests that while processing constraints can account for some patterns in phonology, other patterns suggest a role for cognitive constraints on the encoding, storage and processing of spoken language.

Jean-Luc Schwartz
Institut de la Communication Parlée, CNRS-INPG-Univ. Stendhal, Grenoble
ICP, INPG, 46 Av. Félix-Viallet, 38031 Grenoble Cedex 1
 tel (33) 4 76 57 47 12        fax (33) 4 76 57 47 10

How mouth, ear and eye meet language ? A Theory of Speech-Perception-for-Action-Control

Let us start from a radical substance-based position, in which speech stems from a volitional vocalisation system involving the cingular cortex and the supplementary motor area, and providing a basic phonological principle, a "frame" consisting in syllabic oscillations; and from a vocal self-monitoring system, probably recruiting the Broca-Wernicke circuit, and allowing a "content" to emerge and provide phonological parameters, specifying segments. In this view, the present contribution is rather focused on vocal self-monitoring. After the seminal work by Sperry on corollary discharge enabling to predict the perceptual consequences of one's action, Frith's experiments extended this concept to the on-line following of one's own voice. The extrapolation towards the possible ability to follow on-line the vocal gestures of somebody else was suggested by Liberman’s Motor Theory, and received some sort of echo in the so-called "mirror neurons" recently discovered by Rizzolati et al. in the monkey's brain, and displaying similar kinds of responses to the monkey's action and action perception.

A theory of speech perception should, in our view, tell us how a listener might follow the vocalisations of his speaking partner, in order perhaps to understand them, but at least certainly to imitate and learn: in other words, how perception enables a listener to specify the control of his future actions as a speaker. Conversely, this theory should also tell us how the perceptual representations of speech gestures transform, deform, shape the speaker's gestures in the listener's mind, and hence provide templates that in return contribute to specify the control of the speaking partner's own actions. And lastly, this theory should be able to show how the choice of speech units inside the phonological system may be constrained and patterned by the inherent limitations and intrinsic properties of the speech perception system – and its indissociable companion, the speech production system.

In summary, we defend the view that speech perception is the set of perceptual (auditory, visual, if not tactile) processes allowing, at the segmental level, to recover and specify the timing and targets of speech gestures, supplying a set of representations for the control of one's own actions and the specification of somebody else’s actions, and able to intervene in the structuring of phonological systems. Therefore, our research programme is concerned with the determination of perceptual processing and representations allowing action control and related, through perception-action interactions, to certain aspects of principles and parameter settings in phonology.

Our approach, centred on the co-structuring of the perception and action systems in relation with phonology, is clearly different from both « auditory » theories in which the sensory-interpretative chain is considered independently of the patterning of sounds by speech gestures, in the search of some "direct link" between sounds and phonemes; and from « motor » theories in which perception is nothing but a mirror of action, in the claim of a "direct link" between sounds and gestures. It is rather focused on multimodal percepts regularised by motor constraints; or speech gestures shaped by multimodal processing.

References

Abry, C., Boë, L.J., Laboissière, R., & Schwartz, J.L. (1998). A new puzzle for the evolution of speech? Behavioral and Brain Sciences, 21, 512-513.

Robert-Ribes, J., Schwartz, J.L., Lallouache, T., & Escudier, P. (1998). Complementarity and synergy in bimodal speech: auditory, visual and audiovisual identification of French oral vowels in noise. Journal. Acoust. Soc. Am., 103, 3677-3689.

Schwartz, J.L., Abry, C., Boë, L.J., & Cathiard, M.A. (in press). Phonology in a Theory of Perception-for-Action-Control. In J. Durand, B. Laks (eds.) Phonology : from Phonetics to Cognition. Oxford University Press.

Schwartz, J.L., Arrouas, Y., Beautemps, D., & Escudier, P. (1992). Auditory analysis of speech gestures. In M.E.H. Schouten (ed.) The Auditory Processing of Speech – From Sounds to Words (pp. 239-252). Speech Research, 10, Berlin : Mouton de Gruyter.

Schwartz, J.L., Boë, L.J., Vallée, N., & Abry , C. (1997a). Major trends in vowel system inventories. Journal of Phonetics, 25, 233-254.

Schwartz, J.L., Boë, L.J., Vallée, N., & Abry , C. (1997b). The dispersion-focalization theory of vowel systems. Journal of Phonetics, 25, 255-286.

Schwartz, J.L., Robert-Ribes, J., & Escudier, P. (1998). Ten years after Summerfield ... a taxonomy of models for audiovisual fusion in speech perception. In R. Campbell, B. Dodd & D. Burnham (eds.) Hearing by eye, II. Perspectives and directions in research on audiovisual aspects of language processing (pp. 85-108). Hove (UK) : Psychology Press.

Kay Bock
"How Mind Meets Mouth"

People reliably create precise changes in other people's minds.  They do this not with ESP, but with mere talk.  What is the nature of the human ability to convey specific ideas in specific circumstances to specific listeners? I will survey several lines of research on language production that explore the cognitive capacities and limitations of turning thoughts into speech.