Suivez aussi la conférence en visio ICI
Abstract by Denis Delfitto and Maria Vender (University of Verona)
- Human languages are externalized as linear sequences of hierarchically-encoded atomic units. The nature of the interplay between the cognitive development of these hierarchical representations and their linearization on the string is still poorly understood. In this presentation, we will explore the relationship between precedence and containment by capitalizing on the results of a new experimental paradigm that has already provided interesting insights (Vender et al., 2019, 2020).
More specifically, we report the results of five Artificial Grammar Learning studies developed adopting a Serial Reaction Time task, a modified Simon Task in which the sequence of the stimuli is determined by the rules of the Fibonacci grammar (Fib) or of its modifications Skip and Bif (all non-canonical grammars categorized as asymmetric Lindenmayer-systems).
We will argue that given their peculiar properties (especially self-similarity), these grammars lend themselves to experimentally address the interplay between statistically-based and hierarchically-based computations. For instance, though all three grammars share the same transitional regularities, learnable in purely statistical terms, they crucially differ in their structure : as a matter of fact, only Fib is characterized by the presence of so-called k-points, which formally provide a potential bridge to hierarchical reconstruction.
In our studies, we examined both children’s and adults’ implicit learning skills, assessing learning of the statistical regularities in Fib, Skip and Bif, while also exploring the presence of hierarchical learning, in terms of the ability to predict k-points.
Results of all studies unanimously provide evidence not only for the presence of statistically-based sequential learning, but also for hierarchical learning in Fib. Based on these results, we argue that the relations of precedence and containment are not antagonistic ways of processing a temporally ordered sequence of symbols, but are strictly interdependent implementations of an abstract mathematical relation of linear ordering within a bidimensional computational space.
We propose that the construction of this multidimensional space, hence of hierarchical structures in addition to purely sequential ones, is primarily determined by labeling requirements, with the labeling algorithm emerging as the solution to the problem of mapping precedence into containment.
References
Vender M, Krivochen DG, Phillips B, Saddy D, Delfitto D. Implicit Learning, Bilingualism, and Dyslexia : Insights From a Study Assessing AGL With a Modified Simon Task. Front Psychol. 2019 Jul 26 ;10:1647.
Vender M, Krivochen DG, Compostella A, Phillips B, Delfitto D, Saddy D. Disentangling sequential from hierarchical learning in Artificial Grammar Learning : Evidence from a modified Simon Task. Koizumi M, editor. PLoS ONE. 2020 May 14 ;15(5):e0232687.