Recognising morphologically complex forms
Our discussion of word recognition processes has so far largely avoided the issue of how we recognise, process and store words that are made up of more than one meaningful element, i.e. morphologically complex words like cats, children or predisposition. A determining factor in this con text is the relative productivity of morphological processes, that is, the predictability of a certain way of constructing a morphologically complex form. As discussed in Chapter 4, productivity is sometimes also linked to the distinction between inflectional morphology adding affixes to mark grammatical information such as plural [cat – cats], past tense – [jump-jumped], or subject-verb agreement [he swims vs I swim] and derivational morphology adding affixes to make a different kind of word, such as making a noun disposition from a verb dispose (Bauer, 2001; Frauenfelder & Schreuder, 1992).
Inflections tend to be more productive than derivations, though there are differences within each of these sets. For instance, examples of making a plural in English by adding s, as in cats, are plentiful, but making a plural by adding -ren and changing the vowel in the word (as in child-children) is very unusual. The more frequent or regular patterns are the ones that children identify for themselves as they learn their native languages, and which for a while they apply inappropriately in their own speech. So for instance children learning English as their first language will produce forms like foots (or even feets) instead of feet. See also the discussion in Chapter 4 of the wug test.
The corresponding rule-governed process in word recognition would be to look for a form like as under the entry for cat. Just as it is claimed that the language production system assembles as from the base word a and the highly productive affix-s, following a rule like to make a plural add -s to the stem’, so the word recognition system would disassemble the input as through a process of morphological decomposition. This is reason ably straightforward in the kind of left-to-right’ processing system that we have been considering for spoken language processing, since the base word a would be contacted on the basis of the first three phonemes of as in any case. Indeed, surveys of the morphological characteristics of the world’s languages suggest that accessing an inflected form via its stem may be a common strategy, since inflectional affixation is predominantly done through suffixes, i.e. adding elements to the ends of words rather than to their beginnings or to some point within the word.
Processing inflections
There is considerable evidence from both spoken and visual word recognition studies concerning how we process inflected forms. We have seen above how the frequency with which a word has been encountered can influence recognition speed and accuracy. It has also been shown that the important frequency statistic is not that of a base form (e.g. cat) on its own, but the combination of the base form and its inflected forms so a and as together. This combined frequency can be a better predictor of recognition time. This suggests that each time we encounter an inflected form we also encounter the base word, adding to its frequency count. It has also been shown that inflected forms are hard to access – out of context, /daz/ and /pækt/ tend to be reported as ae and a rather than as and a e , even though the latter are of higher frequency. Note though that this may be a consequence of presenting words in isolation – citation forms tend to be uninflected forms.
There is clearly a difference between regular and irregular inflection. As we have seen, children tend to regularise irregular forms, suggesting that these do not have a strong hold on their processing system, even for frequently encountered irregular forms like went. In addition, evidence from repetition priming in adults supports this processing distinction as well as providing further evidence for the close links between inflected forms and base words. Equally strong priming is found for identical prime--target pairs so the word o preceded several items earlier by the same word pour, compared to a condition without this earlier instance of o and for a condition where the form o is preceded not by the same word but by the inflected word poured. Compare this with the finding that although there is some priming of swim by its irregular past form swam, it is not as strong as primming of swim by swim itself.
Differences in the processing of regular and irregular forms are reflected in ERP studies. These have indicated that different brain areas are activated, with the regular forms showing activation in areas known to be involved in rule processing and the irregular forms leading to activation of areas that are associated with lexical storage (Newman, llman, Pancheva, Waligura & Neville, 2007).
