Morphological Processing 


The organization of words in the mental lexicon has been one of my main themes of research, dating back my Ph.D. dissertation. My research in this field has focused around the concept of morphological paradigms. These are sets of words that share the same morphological constitutent (e.g., roots, stems, etc.). In a number of studies we found that the size of these paradigms (the famiy size effect) is a consistent predictor of lexical decision reaction times. This observed across languages with very diffferent morphological structures, including Dutch (Moscoso del Prado, Deutsch, et al., 2005), Hebrew (Moscoso del Prado, Deutsch, et al., 2005), and Finnish (Moscoso del Prado, Bertram, et al., 2004). Furthermore, non-native bilinguals are affected by these paradigms in both languages they speak. Surprisingly, bilinguals are sensitive to the morphological properties of words in their second language, even when they are tested in settings and taks that do not involve their second language. Paradigms in the second language interfere with the subjects' first language processing, reflecting the large degree of automaticity of morphological processes (Dijkstra, Moscoso del Prado, et al., 2005). In further research, in collaboration with several colleagues, we have developed a parsimonious information-theoretical account of morphological paradigms (Moscoso del Prado, Kostić, & Baayen, 2004), that can also characterize the regularities between related paradigms (Milin, Filipović-Ɖurđević, & Moscoso del Prado, 2009).

Morphological relations are formed by a combination of formal (orthographic and phonological) and semantic factors. In several studies, we have documented the sensitivity of morphological effects to semantic distinctions (Moscoso del Prado, Bertram, et al., 2004, Moscoso del Prado, Kostić, & Baayen, 2004, Moscoso del Prado, Deutsch, et al., 2005, Baayen & Moscoso del Prado, 2005). Importantly, these semantic relations are part and parcel of the morphological relation. Some current models of morphological processing have claimed that it proceeds in two discrete stages, first accessing the formal properties (i.e., "affix stripping"), and only then dealing with the semantic details. However, joint consideration of the studies published in the literature reveal that semantic effects happen indeed very early (Feldman, O'Connor, & Moscoso del Prado, 2009), which is in line with reports of very early semantic processing I observed in ERP studies (Moscoso del Prado, Hauk, & Pulvermüller, 2006). This has led to an interesting discussion on what type of model of morphological processing does the literature actually support. In a sequence of studies, in collaboration with Laurie B. Feldman, we have observed early semantic effects in morphological processing across multiple languages, and a more detailed meta-analysis of the literature reveals that these semantic component is indeed unavoidable.

The automatic semantic relations between morphologically related words do not seem to fit well into affix-stripping models of morphology. Rather, the system seems more like a messy system in which direct "links" exists between morphologically related words, in line the the current "Word and Paradigm" theories of morphology that are being developed in linguistics. From a modelling perspective, this fits well into disributed connectionist models of morphological processing. I have shown that distributed connectionist models naturally give rise to the paradigmatic effects reported above (see my dissertation for a detailed model, and Moscoso del Prado, Ernestus, & Baayen, 2004, for a more specific model). Currently, in collaboration with Farrell Ackermann, Jim Blevins, and Rob Malouf, we are trying to extend the information theoretical model to provide a more general account of morphological relations.