Mechanisms of aural encoding: IV. Hear-see, say-write interactions for vowels

Six vowel sounds were presented in rondom orders for recall from short-term memory. Four groups of 25 Ss each in a2by 2 design either heard or saw the words, and either had to say them or write them. In general, accuracy was greater when the sounds were seen rather than heard, especially when they were verbatty reported. The major concern was the degree to which the major distinctive feature systems could predict the intrusion errors in each of the four conditions. The best prediction occurred in the auditory-input/verbal-output condition, as would be expected from the models on which distinctive feature specifications are based. However, as in a previous study by these authors, the most accurate predictions were not always made by the same combination of distinctive features. It is concluded that distinctive features are identifled in the initial encoding bef ore any retrieval processes are activated and that further errors made at the time of retrieval are aho related to distinctive features. Most importantly, however, both at initial registration and at retrieval, different features are implicated, depending upon the sound presented for recall and the input-output modalities involved.