Sound coding in verbal learning with and without restrictions on decoding

.80 .70 V) ~ .60 .60 oU V) S.50 .50 lOW/..{ c.:: .40 Fig. I.lnteraction of mean and variability groups as a function of run position. ",G"' v A 1ow~ ~o ol,-_~_~_~_ ol~~_~__EARL Y MtDDLE LA TE MEAN GROUPS DISCUSSION As was observed before, early run scores for all groups are significantly lower than either intermediate or late run scores. This is to be expected in that the optimal strategy for each S is still weighted heavily for the prediction of the previous event, particularly so due to the random nature of the length of runs. Indeed, this is an essential feature ofthe k-span model ofBurke & Estes (1957), where S bases his response on any one event as a result of the last k events. The early run scores for the shorter run length groups were significantly lower than those of the groups exposed to longer run lengths, while no significant differences were noted in the late TUn scores between the groups. I t would seem, then, that there is a tendency on the part of the short run groups to perseverate the response favorable to the last tone run heard. As was suggested by Gambino & Myers (1966), designs employing feedback techniques increase the number of reinforcements for the prediction of the alternate event for the shoTt TUn groups. In the present context, feedback to S was eliminated with the net effect that differences in recency disappeared between groups with different mean run lengths. This seems to support the conclusion that the differential amount of reinforcement between groups is responsible for differences in recency formerly attributed to mean run length. However, this conc1usion may be more relevant to detection psychophysics than to probability learning experiments, due to the notable lack of explicit reinforcement. Perhaps the most salient feature in the resul ts of the nonfeedback design is the increased negative recency effect with the high variability groups. It would appcar that an S operating under uncertainty duc to both thc stimulus Psychon.Sci.,1969,VoI.17(2) EARL Y MiDDLE LA TE YARIANCE GROUPS and TUn stability seems to choose a strategy involving the selection ofthe alternate event to an increasing degree shortly after the mean of the TUn length distribution. As an extension of the work ofVerplanck, Collier, & Cotton (1952) with sequential dependencies, the concept of negative recency in probabilistic learningmodels may be direct1y associated with the declining critical ratios reported by these authors. They found statistically significant nonindependence when as many as 10 responses intervened between correlated responses. This appears to be an accurate observation for relatively stable run sequences. However, as the variability ofthe run lengths increases, there is a decline in the sequential dependencies. In a later study by Collier & Verplanck (1958), the degree of dependency in run sequences was found to be directly associated with the relative strength of the stimuius. In the tradition al two-choice probability learning situation, S is presented with what might be considered to be a relatively strong stimulus in the context of a detection experiment. It would seem, then, that the sequential dependencies reported in the psychophysicalliterature are extensions of effects found in the context of probability learning experiments. REFERENCES BURKE, C. J., & ESTES, W. K. A component model for stimulus variables in discrimination learning. Psychometrika, 1957,22,133-145. COLLIER, G., & VERPLANCK, W. S. Nonindependence of successive responses at the visual threshold as a function of interpolated stimuli. Journal of Experimental Psychology, 1958,55,429-437. GAMBINO, B., & MYERS, J. L. Effect of mean and variability of event run length on two-choice learning. Journal of Experimental Psychology, 1966,73,56-67. . JARVIK; M. E. Probability leiuning and anegative recency effect in the serial anticipation of alternate events. Journal of Experimental Psychology, 1951,41. 291-297. JONES, M. R., & MYERS, J. L. A comparison of two methods of event randomization in probability learning. Journal of Experimental Psychology, 1966, 72, 909-911. KINCHLA, R. A. A leaming factor in visual discrimination. In R. C. Atkinson (Ed.),Studies in mathematical psychology. Stanford, Calif.: Stanford University Press, 1964. Pp. 233-249. VERPLANCK, W. S., COLLIER, G. H., & COTTON, J. W. Nonindependence ofsuccessive responses in measurements of the visual threshold. Journal of Experimen tal Psychology, 1952,44,273-282. NOTE 1. Now at Psychology Department, Box 8185, University of Miami, Coral Gables, Florida 33124. Reprint requests should be addressed to Dr. Newman. Sound coding in verbal learning with and without restrictions on decoding WILL/AM E. FORRESTERI and NORMAN E. SPEAR, Rufgers University, New Brunswick, N.J. 08903 A transfer design was used to assess the hypothesis that Ss may encode response terms as pronounceable sounds in learninga paired-associates list. The results indicated that the facilitative effeets of sound coding are great~v diminished ifS must use different decoding rules in learning original and transfer lists. Underwood & Erlebacher (1965, Experiment 6) provided evidence that response items in a verbal-learning task may bc encoded as pronounceable sounds even though Ss are required to speil responses aloud at recall. The procedures used to demonstrate this coding process conformed 91 l\' J IranSki c1cSI~n In whu:h c:l.pclllllenlal Ss \\Iw L'(lJcJ first-Ir SI responses as sounds could mediate via the .:oded rtem in learning second-hst pairs_ Thus, ifS learned 3-AKE as a pair in list land coded the response as äk, the coded form 01' the item could be used as a media tor in learning the pair 3-ACHE in list ~. As Underwood and Erlebacher point out, interfercnce could arise from the difference in decoding processes required in producing first- and second-list responses. Their results indicated, nevertheless, that the facilitative effects of sound coding were sufficiently great to outweigh any interfererice at decoding and, thus, produce a small net positive-transfer effec!. In the present study, an attempt was made to assess the magnitude of the interference effect associated with the decoding process_ For this reason, a factorial design was used in which Ss in one condition were not required to decode at recall; that is, S was instructed to pronounce rather than to speil responses_ Ss in the other condition were required, as in Underwood & ErIebacher's study, to decode (i_e., speIl) responses. It was expected that positive transfer effects would occur in' both experimental conditions, relative to appropriate contrJll groups, but a greater effect was expected under conditions not requiring decoding_ In short, an interaction was predicted between experimentalcontrol and spell-pronounce factors. SUBJECTS AND MATERIALS A total of 72 male undergraduates at Rutgers volunteered to serve in the study_ All Ss were enrolled in introduct (...truncated)