The origin of abstract concepts: Comparing children with Down Syndrome and Williams Syndrome. Michele Burigo, Sara Martelli, Renato Borgatti, Alessandro Tavano & Maria Luisa Lorusso Scientific Institute “E. Medea”, Bosisio Parini, Italy

Introduction The difference between concrete and abstract concepts has been discussed for more than 30 years, and the scientific community has still not reached an agreement on the theory that explains this divergence. The most accredited approaches are: the Dual-Coding Theory (Paivio, 1986) claims that the fundamental difference between abstract and concrete concepts is that concrete concepts can be processed by both a language-like code and imagery information, whereas abstract concepts can be processed only by the language-like code, the Context Availability Theory (Schwanenflugel & Shoben, 1983) argues that it is easier to think of a context for concrete objects than for abstract ones. The issue here is not whether a context can come to mind at all, but how long it takes to retrieve or construct it based on information in memory, the Contextual Constraint Theory (Wiemer-Hastings, Krug and Xu, 2001) proposes that an abstract concept can occur in many or few kinds of context, depending on the situation where it has been used. Thus an entity that is not strongly constrained is more abstract than an entity that is contingent on a fairly extensive set of constraints. All these theories assume that the distinction between concrete and abstract concepts is based on differences emerging at a representational level. However, the difference between concrete and abstract concepts may emerge during the concept learning stage. Two working hypotheses are here contrasted: an Embodiment Hypothesis and an Abstraction from Language Hypothesis. The former, according to cognitive linguistics, sees abstract knowledge as a “conceptual metaphor” (i.e., the use of a concrete conceptual domain of knowledge to describe an abstract conceptual domain) (Lakoff, 1980) and assume that learning abstract concepts is grounded in concrete knowledge, which in turn is grounded in our bodily experience of the world (Embodiment Theory). The alternative theory claims that abstract language is learned through the process of extracting statistical regularities from language (Landauer & Dumais, 1997; Burgess & Lund, 1997). In order to contrast these alternative theories we focused on two populations whose specific patterns of impairments offer a direct test of the two views: children with Williams syndrome and children with Down syndrome. Clinical assessment and predictions Williams syndrome (WS) is a rare genetically based neuro-developmental disorder caused by the deletion of one copy of a small set of genes on chromosome 7 (7q11.23). Children with WS are characterised by major visual-spatial impairments accompanied by closer-to-normal linguistic skills (but see Bellugi & George, 2001, for a more detailed review of the neuropsychological data). Down syndrome (DS) is a chromosome abnormality caused by trisomy of chromosome 21 and children affected by this syndrome are mainly characterised by severe linguistic impairments with relatively spared visual-spatial skills. Therefore, according to the Embodiment hypothesis we should find a more extensive use of abstract concepts within the DS population whose embodiment process would be facilitated by better visualspatial functions. On the other hand, more extensive use of abstract concepts within the WS population should indicate that abstract concepts are learned via Abstraction from Language mechanisms.

Method In order to test which of the two groups shows a more extensive use of abstract concepts, children’s narratives of the picture story “Frog, where are you?” (Berman & Slobin, 1994) were collected in both clinical groups. All the narratives were coded in CHILDES format. A first lexical analysis excluded closed-class and auxiliaries from the story, while the verbs were all coded in the infinite form. Afterwards, the full list of words was rated by 18 adult native speakers for concreteness, on a 1 (totally abstract) to 7 (totally concrete) scale. Two classes of concrete and abstract words were created using a dichotomised code where the words rated 4 and below were treated as abstract while the words rated above 4 were considered concrete. Subsequently, an index of concreteness of language was calculated expressing the percentage of concrete words over the total number of words (e.g., a subject that produced 3 concrete words over a total of 10 words should have a percentage of 30%). Results The two groups of children were matched on different variables. The first analysis compared the DS group and the WS groups matched on Mental Age. A pairwise comparison showed that WS children were more abstract (M = 63.92) than DS children (M = 73.94) (T(18) = -3.67, p = .0018) corroborating the Abstraction from Language Hypothesis. We also run a number of correlations in order to investigate whether any linguistic variable could be predictive of the index of concreteness. Linguistic skills were measured by 4 tests: Naming test (De Agostini et al, 1998; Italian adaptation Fabbro, 1999), Sentence Repetition (Vender et al, 1981) Token Test (Ferrari et al, 1981) and British Picture Vocabulary Scale (BPVS) (De Agostini et al, 1998; Italian adaptation Fabbro, 1999). The tests were further grouped in order to selectively compare Semantic components (Naming + BPVS) vs. Syntactic components (Token Test+ Sentence Repetition). In a different grouping we also compared Comprehension (BPVS + Token Test) and Production (Naming + Sentence Repetition) components. Separate correlations for DS, WS and Controls did not reveal any correlation with concreteness. If including all children (DS, WS, C = 60ss) concreteness negatively correlates with Comprehension (r = -.315; p = .008) and Syntax (r = .248; p = .041). Discussion Two hypotheses are here contrasted: The Embodiment Hypothesis which assumes that learning abstract concepts is grounded in learning concrete information, which in turn is grounded in our bodily experience of the world, and the Abstraction from Language Hypothesis that claims that abstract language is learned through the process of extracting statistical regularities from the language. For this purpose children with Williams syndrome and children with Down syndrome were compared on narrative production. Data analysis revealed that WS children present, overall, a more extensive use of abstract concepts than DS children matched on mental age. This outcome supports the Abstraction from Language hypothesis. In fact, an advantage for children with Ws in the use of abstract language is only possible if the visual spatial system is not, or only marginally, involved in the acquisition of abstract concepts. On the other hand, given that abstract concepts seem to be learned through the process of extracting statistical regularities from the language, DS children present a more concrete language because their linguistic difficulties hamper this process. The correlation found in the sample of subjects further suggest that the linguistic skills more directly involved in the acquisition of abstract concepts are to be grounded in the Syntactic rather than in the Lexical-Semantic components and in Comprehension rather than in Production. This is also compatible with the Abstraction from Language Hypothesis, where the crucial process is the extraction of regularities in input language and process that can also be seen as the basis of syntactic competence.

References Bellugi, U. and George, M. (2001). Journey from cognition to brain to gene. Perspective from Williams Syndrome. Cambridge, Massachusetts; The MIT Press. Berman, R. A., and Slobin, D. I. (1994). Relating events in narrative: A cross-linguistic developmental study. Hillsdale, NJ: Lawrence Erlbaum Associates. Burgess, C., and Lund, K. (1997). Modelling parsing constraints with high-dimensional context space. Language and Cognitive Processes, 12, 177-210. De Agostini, M., Metz-Lutz, M.N., Van Hout, A., Chavance, M., Deloche, G., Pavao-Martins, I., and Dellatolas, G. (1998). Batterie d'évaluation du langage oral de lenfant aphasique: Standardisation française (4-12 ans). Oral language evaluation battery of aphasic children: a French standardisation. Revue de Neuropsychologie, 8, 319-367. Ferrari, E., De Renzi, E., Faglioni, P., & Barbieri, E. (1981). Standardizzazione di una batteria per la valutazione dei disturbi del linguaggio nell’età scolare. Neuropsichiatria Infantile, 235, 148-158. Landauer, T. K. and Dumais, S. T. (1997). A solution to Plato's problem: the Latent Semantic Analysis theory of acquisition, induction and representation of knowledge. Psychological Review, 104(2) , 211-240. Lakoff, G. and Johnson, M. (1980). Metaphors We Live By. Chicago: University of Chicago Press. Paivio, A. (1986). Mental representations. A dual coding approach, Clarendon Press, Oxford. Schwanenflugen, P. and Shoben, E. J. (1983). Different context effects in the comprehension of abstract and concrete verbal material. Journal of Experimental Psychology: Learning, Memory and Cognition, 9, 82-102. Vender, C., Borgia, R., Cumer Bruno, S., Freo, P., & Zardini, G. (1981). Un test di ripetizione di frasi. Analisi delle performances di bambini normali. Neuropsichiatria Infantile, 243, 819-831. Wiemer-Hastings, K., Krug, J. and Xu, X. (2002). Imagery, context availability, contextual constraint, and abstractness. Proceedings of the 23rd Annual Conference of the Cognitive Science Society, 1134-1139.

This work has been funded by the EU Community as part of the NEST Project “ABSTRACT”.

90 Williams Down

% Concrete/Tot

80

70

60

50 Mental age