Memory effects to be considered in the design of measurements of the influence of context on speech perception.

One of the experimental parameters needed for the design of testing procedures for measuring context utilization in speech perception is the magnitude of memory effects. Such memory effects were studied using data obtained from normal-hearing subjects in experiments during which test words were presented under noise either in isolation or in the context of sentences. Each test word was presented twice; during each presentation the subject heard the test word several (five or six) different times, under progressively higher signal-to-noise (S/N) ratios. When the second presentation of a test word occurred during the same 1 h session as the first presentation, the observed memory effect was very large, being equivalent to a shift in S/N ratio of the order of 5 dB. In experiments during which the second presentation of a test word occurred after an interval of several days following the first presentation, the observed memory effect was equivalent to a shift in S/N ratio of the order of 1 dB.