Unifying Awareness and On-Line Studies of Speech: A tentative Framework ' 9

Jacques Mehler Laboratoirc de Science Cognitiv e et Psycho linguistique . CN RS-EHESS Paris France Generally, studies of speech recognitio n are related to theories of performance while studies of awareness are thought to bear upon language competence. In our concept ion, both area s or resea rch contribute to our unders tandlng of processing and of the represeetanons that the subjeds use when listening to speech. We present a unitary framework within which it becomes pouibk: to incorporate the results from on-line speech rccognitlon studies and from studies of the awareness that the language user has of speech sc:gmen~. In puticular, we argue that it is necessary to include a descriptlon of the manner in which acoustic-phonet ic information is transduced , and represented in order for us to understand how subjects come to decide to respond or not in a psycholinguistic expe riment. Particular attention is given to the data from on-line chunk detection experiments and to the potential role of orthographic representation. As we all know, learning to speak and learning to read are very different processes. While one acquires speech by mere exposure, one learns to read the hard way. namely. at school. Not only is schooling required , but many years of instruction are needed to turn illiterates into skilled readers. Nonetheless, we have always been fascinated by the observation that the very young child seems to be unaware that one can view speech as a sequence of phonemes. Research initiated by Liberman and her colleagues (Liberman, Shankweiler, Fischer, & Carter. 1974). but greatly expanded by a team of Belgian psycholingu ists led by P. Bertelson (see the collection of papers edited by Bertelson, 1986), suggests that phonemic awareness arises while children are learning to read thanks to the conversion of graphemes into phonemes (Morais, Cary, Alegria, & Bertelson, 1979). This finding has received more attention from educational psychologists than from psycholinguists. This is surpri sing since, regardless of whether they study speech perception or reading. psycholinguists relate their findings to the units used to process speech, see Savin and Bever (1970), Spoehr (1981), Taft and Forster (1976), 60 E. Dupoux and J. Mehfu and Mehler (1981), among many others. We should acknowledge, from the beginning of this chapter, that our intuitions lead us to expect that metalingu istic competence as it is explored in awareness studies reflects important properties of the speech perception apparatus. Tacitly, psycholinguists have often behaved as if the study of metalinguistic competence and that of on-line speech perception are quite unrelated. Metalinguistic studies supposed ly reflect the knowledge of the language while on-line studies, at best , reflect the subjects ' encapsulated processes (see Pylyshyn, 1984). Tho ugh this attitude, to the best of our knowledge, has never been explicitly defended, observing how the fields have grown shows that the above assertion describes an underlying reality. Our belief is that this behaviour is profoundly misguided. It is not that we are willing to give up the distinction between competence and performance or between knowledge and encapsulated processes. Rather, we believe that the methods used in studies of awareness and on-line speech perception rely on representations brought about by the encapsulated processing system. In order to play metalinguistic games the subject has to use some representation or another. Whichever representation he/she might use it must be one that is computed by the processing system. Likewise, in chronometric studies in order to perform speeded tasks, and even to memorize targets that have to be detected, subjects have to rely on some representations. These representations might be phonetic, phonemic or even orthographic, and thus reflect the knowledge of language that the subjects have attained, just as in the studies of awareness. The purpose we pursue here is to try and present a preliminary framework to illustrate the reciprocal advantage one might draw by formulating a unitary model within which both awareness and chronometric studies can be couched. The model, which will be detailed below, suggests that subjects behave on the basis of an encapsulated information processing system, a representational system of the surrounding information, and a non-modular decision mechanism. An important consequence of adopting such a unitary model is that psycholinguists will have to include considerations about literacy in the interpretation of on-line studies. Indeed, literacy adds representations which make it possible for subjects to perform in tasks like phoneme-monitoring experiment which they would otherwise be unable to do. Given this observation, it is surprising that psycholinguists have neglected the potential effects of cultural representations. A few exceptions can be mentioned. For instance, the ability to detect rhymes in spoken words is greatly affected by orthographic similarity (Seidenberg & Tanenhaus, 1979; see also Tanenhaus, Flanigan, & Seidenberg, 1980). Likewise, Cutler, Norr is, and van Oayen (1990) speculate that the behaviour of subjects in a vowel detection task may be influenced by orthography (see also Taft & Hambly, 1985). Psycholinguists acknowledge that literacy adds new represen tations to the existing ones. In our framewo rk, we have to say that these new representations can have a potentially large impact on subjects' performance in