Frequency Trajectory Gives Rise to an Age-Limited Learning Effect as a Function of Input-Output Mapping in Connectionist Networks

Frequency Trajectory Gives Rise to an Age-Limited Learning Effect as a Function of Input-Output Mapping in Connectionist Networks Martial Mermillod Patrick Bonin LAPSCO - (CNRS UMR 6024) Universite Blaise Pascal LAPSCO - (CNRS UMR 6024) Universite Blaise Pascal Tiffany Morisseau Alain Meot L2C2 - (CNRS UMR 5230) Institut des Sciences Cognitives LAPSCO - (CNRS UMR 6024) Universite Blaise Pascal Ludovic Ferrand LAPSCO - (CNRS UMR 6024) Universite Blaise Pascal Abstract According to the age of acquisition (AoA) hypothesis, words acquired early in life are processed faster and more accurately than words acquired later (see Juhasz, 2005; Johnston & Barry, 2006 for reviews). Connectionist models have begun to explore the influence of the age/order of acquisition of the items (and also their frequency of encounter) (Ellis & Lambon Ralph, 2000; Lambon Ralph & Ehsan, 2006; Zevin & Seidenberg, 2002). We explored age-limited learning effects in a connectionist model similar to that used by Lambon Ralph and Ehsan (2006) but with the use of a frequency trajectory (Zevin & Seidenberg, 2002), which refers to changes in the frequency of the words over long periods of age, since frequency trajectory is thought to better index age-limited learning effects than traditional AoA measures (Bonin, Barry, Meot, & Chalard, 2004). Our simulations show that the influence of frequency trajectory varies as a function of the mappings between input and output units in a similar type of neural network to that used by Lambon Ralph and Ehsan (2006). Introduction An important issue in psychological science is to determine whether items (words, objects, faces, etc.) which are acquired early in life are processed faster and more accurately by adults than those which are acquired later in life, namely whether there is a late influence of early acquisitions. A large number of studies have convincingly shown that words acquired early in life are processed faster and more accurately than words acquired later in life (Johnston & Barry, 2006; Juhasz, 2005 for recent reviews) using age of acquisition (AoA) norms collected from either adult ratings or from children's performance. The so-called age-of-acquisition effects have been found in a large variety of tasks (e.g., object, face and action naming, word reading, lexical decision) and in different populations (e.g., children, young and old adults, aphasics). However, despite robust AoA effects in a wide variety of lexical tasks, there is a current debate as to whether the order of acquisition of the words is per se an important factor in determining the ease of processing the words in both normal and impaired adults or whether AoA measures actually underlie other hidden factors. It is plausible that the order of acquisition of the words is a factor which is directly responsible for the ease of processing the words, and indeed this is the crucial tenet of the “AoA hypothesis”. Recent attempts to independently manipulate this factor have shown a reliable influence on the learning of artificial patterns in laboratory settings (Stewart & Ellis, 2008). However, as far as the learning of the words of a language is concerned, there are obviously, factors other than the order in which the words and/or concepts were encountered also clearly underlie the speed and accuracy of acquisition (with the result that certain words are acquired before others). These factors are truly responsible for the AoA effects found in lexical processing in adults. Among these factors are (1) the frequency of encounter of the words (e.g., during certain period of life, during one's entire life) and (2) the kind of relationships (i,e., systematic, quasi-systematic, arbitrary that exists between different types of codes (e.g., between phonological and orthographic codes, between semantic codes and phonological codes). Frequency trajectories refer to the fact that some words are more frequent during certain periods of acquisition (e.g., “dragon” during childhood) than others (e.g., “tax” during adulthood) and the words which are frequently encountered are acquired earlier than those which are encountered less frequently (Bonin, Barry, Meot, & Chalard, 2004; Hazard, De Cara, & Chanquoy, in press; Zevin & Seidenberg, 2002). But as we shall explain, the question of whether words which have been frequently encountered during a period of acquisition are easier to process later in life than words encountered less frequently also depends on the kind of relationships that exists between different types of codes (and which have to be learned). In alphabetic languages such as English or French, there are quasi-systematic relationships between sound units and orthographic units, whereas the relationships between semantic units and phonological (or orthographic) units are arbitrary. When quasi-