On the processing and representation of blends in long term memory

In this work we analyse the mental representation of blends (portemanteau words), particularly focusing on French. Blends are assimilable to compounds in that they combine two lexemes at least. However, in blends, the bases often appear in a reduced form. For instance, the word motel is made by combining the lexeme motor and the lexeme hotel. Blends formation depends on phonological and semantic constraints, and can generally be divided into several types (Fradin, 2000 ; Grésillon, 1984, Lehrer, 1996). Although some phonological patterns seem to be more natural than others, the base lexemes are in general clipped, thus violating a universal principle such as the integrity of the base. Blends can be then defined as being a " less regular way to coin new lexical units than compounds " (Fradin, 2000). Moreover, as suggested by some authors (Fradin et al. 2009 ; Grésillon, 1984; Piñeros, 2004), in most cases blends result from the linear superposition of a phonological sequence shared by both their base lexemes, rather than from their deletion. The process of blending, then, can be assimilated to haplology (cf. Fradin et al., 2009). Semantically, blends are generally interpreted on the basis of the meaning of their base-morphemes. If blends result from a superposition governed by a constraint of linearity of the constituents, they should be stored in memory as whole forms, closely connected to their base-lexemes representations. The aim of the present paper is to investigate this issue. Through a series of psycholinguistic experiments, we manipulated blends and constituent lexemes, in order to examine, on the one hand, how blends are cognitively processed by readers during visual recognition, and on the other hand, in which format blends are coded in long term memory. Finally, we will present a cognitive architecture of lexical access that integrates lexemes and blends