Two experimental tasks, a speech segmentation and a short-term memory task, were presented to children who began to learn to read following either the "phonic" or the "wholeword" method. The segmentation task required the child to reverse two segments (either two phones or two syllables) in an utterance. The phonic group performed significantly better than the whole-wordgroup in the "phonic reversal" task, but no difference appeared in the "syllable reversal" task. This indicated III that most children by the age of 6 years are ready to discover that speech consists of a sequence of phones and (2) that the moment at which they do it is influenced by the way they are taught to read. In the memory task, the children recalled series of visually presented items whose names either rhymed or did not. The difference in performance for the rhyming and nonrhyming series was significant in both groups. It was no greater for the phonic than for the whole-word group and was uncorrelated with the "phonic reversal" task. These results are discussed in connection with the distinction between ways of lexical access and ways of representing verbal information in short-term memory.
451Although children are competent speakers and hearers of language when they begin to learn to read, that competence does not guarantee their success at reading. In order to read, the child must master a new code that gives access to linguistic knowledge from vision rather than from sound. The new code apparently presents considerable difficulties to a substantial number of children, and some fail to cope with it at all. Since the alphabet represents speech at the level of the phoneme, to read in an alphabetic system requires the ability to explicitly analyze speech into phones, at least during the initial stages of the Iearning-to-read process, as well as for reading new words.' Several authors (see, for instance, Liberman, 1971Liberman, , 1973 Liberman, Shankweiler, Fischer, & Carter, 1974;Rozin, Poritsky, & Sotsky, 1971) have claimed, quite convincingly, that one of the most important causes of difficulty in learning to read may be the inability to segment speech into phones. Indeed, some underlying capacity is necessary for the ability to develop, and no training, per se, can be a sufficient condition. But it seems, on the basis of illiterate adults' performance in segmentation tasks (Morais, Cary, Alegria, & Bertelson, 1979), that the ability does not appear spontaneously, so it can be hypothesized that it generally emerges in the learning-to-read situation itself. The most important question is thus how the child becomes aware of the phonetic structure of speech during the learning-to-read process. The development of The present research was partially subsidized by grants to the first author from the Belgian Fond de la Recherche Fondamentale Collective (Contract 2.4535.79) and from the Fond de la Recherche Scientiflque Medicale (Contract 3.4553.79). We wish to thank Virginia Mann for a number of interesting comments and suggestions and Mireill...