Subject-controlled intertrial intervals in concept learning

,wilLhing ;lI1d cOllformers shouid not hcsitatc tu :Idlllit that they did switch. Studics in which public, but not privatc, l'llnfnrmity has brcn ["und havr USl'd cither stimuli In which Ihere was an uhjcci ivcly currect answer (Asch, 1956) ur have empluyed attitude statements as stimuli (Feslingcr, 1957). In the former ca,e, group IIlembers except a singk naive S givc ~n ohviously incorrect answer, and in the lalter group members express an opinion opposed to that of the naive S. Both conditions produce considerable ego-involvement. The conforming individual, when placed in such a situation, may publicly agree with the group decision for fear of ostracism, while privateIy retaining his initial opinion. an REFERENC"ES ASCH. S. L Studie, of independence and submission to group pressure: I. A minority of onr again"it a unanimous majority. Psyc'hologkal ~onograph, 1956.70, No. 9. DivrST A, I·. J. Hfrds of confidcncc and motivation on susceptibility of informational sodal intluence. Journal of Abnormal & Social Psychology. 1959,59.204-209. I LSTINGER, L. A theory o} cognitil'e JissonQnce. Evanston, 111: Row, "ctorson, 1957. HOCHBAüM, G. M. The relationship between group members' self-confidence and their reaetion to group pressure to conformity. American Social Review, 1954. 19, 678-687. HOLLANDER, E. P. Competence and eonformity in the aeeeptanee of intluence. Journal of Abnormal & Sodal Psyehology, 1960,61,365-369. KELMAN, H. C. Hfeets of sueeess and failure on "suggestibility" in the autokinetic situation. Journal of Abnormal & Social Psychology, 1950,45,267-285. ROSENBERG, L. A. Conformity as a fundion of confidenee in partner. Human Relations, 1963, 16,131-139. SAMELSON, F. Conforming behavior under two conditions of contlict in the eognitive field. J oumal of Abnormal & Social Psychology, 1957,55,181-187. SHERIF, M. A study of some sodal faetors in pereeption. Archives of Psychology, 1935,27, No. 187. Subject-controlled intertrial intervals in concept learning l IIERBERT Wt-LLS, University of North Carolina, Greensboro, N.e. 27412 (1966) has suggested that Ss are less efficient in extracting information from errors precisely because the inferential 5s wal' allowed to control their own process involved is temporally more in tatrial in terval (ITI) in a two-response demanding. cUllcept-leaming task. The mean ITI was It is known empirically that a longer longer following e"ors than following intertrial interval (lTI) facilitates co"ect responses and decreased over trials. concept-Ieaming performance (Boume & The results are compatible with measures Bunderson, 1963; Boume, Guy, Dodd, & of response latency" on trials following Justesen, 1965). As Boume (1966, p. 72) e"ors and co"ect responses, but appear to has suggested, hypothesis-testing models conflict with results ubtained when the would therefore prediet more efficient length (lf the ITI is made contingent on the performance when errors are followed by a co"ectness uf 5's response. long ITI and correct responses by a short ITI, than when the contingency is reversed. Erickson and his colleagues (Erickson, Boume, Dodd, Guy, & Justesen (1968) Zajkowski, & Ehmann, 1966; Eriekson & recently tested this prediction and found Zajkowski, 1967) have shown that in a no difference between the two concept-Ieaming task, response latencies on response-contingent interval conditions. trials following errors are longer than those But suppose that the S himself is following correct responses. These results allowed to determine the duration of his provide support for hypothesis-testing ITI on each trial in a self-paced procedure. models of concept leaming, whieh assign Would he perform in the way that the an important role to error trials as theoretieal rcasoning above suggests is most occasions for processing information and efficient? The experimen ts of Erickson selecting new hypotheses (e.g., Restle, et al (1966, 1967~ suggest that the process 1962). It is c1ear, particularly from of hypo thesis selection after an error Levine's work using the blank-trials task continues into the next trial when the ITI (Levine, 1966) that Ss also profit from is held constant. Conversely, if a rather information conveyed by correct trials. But brief, constant time limit is imposed for the selection of a new hypothesis remains the response on each trial, would Ss select an error-related event. Indeed, Levine a longer !TI following errors? Psychon. Sei., 1970. Vol. 19 (2) METHOD The concepts to be leamed had one relevant dimension. There were two levels of stimulus complexity: three irrelevant dimensions (ID) and five ID. Stimuli for the three-ID problems were geometrie figures composed of four trinary dimensions: color (red, green, blue); shape ( triangle , cirele, square); size (smalI, medium, large); and shading within the figure (dots, stripes, crosshatching). For the five-ID problems, there were two additional trinary dimensions: the orientation of a black bar in the upper left comer of the slide (vertieal, horizontal, diagonal) and the number of black dots in the upper right (one, two, three). At each complexity level, two different problems were used: one with size relevant ("Iarge figures are A"), the other with shading relevant ("stripes are A"). Equal numbers of positive (A) and negative (B) stimuli were presented in random order, but with the restriction that no more than three A or B stimuli occur consecutively. The stimuli were photographed on 35-mm slides and shown on a rear-projection screen through a Carousel projector with external timing controls. A panel in front of the screen contained two response levers, labeled A and B, with a red information feedback (IF) light above each lever. The S was fully instructed conceming the stimuli and the kind of concept to be leamed. On each trial, the stimulus was presented and 4 sec later the IF signal was lighted for 1 sec. The S was instructed to respond on each trial before the onset of the IF signal or an error would be counted. There was, on the average, fewer than one such error per S, The Ss were run to a criterion of 16 consecutive correct responses. The Ss were provided a distinctive button which, when pressed, activated the projector to present the next trial. The elapsed time between the end of the IF signal and the button press was recorded manually from an electric stopclock. Neither the stimulus nor the IF signal was visible during the ITI. Sixty introductory psychology students at the University of North Carolina-Greensboro served as a course requirement. An equal number of Ss was assigned to each of the four (two complexity levels by two relevant dimensions) problems. RESULTS AND DISCUSSION Each S's precriterion trials were divided into a first and a second half. The mean posterror and postcorrect ITI within earh half was computed for each S and then averaged across all Ss. These data are 109 Table I Mean ITI (in Seconds) for the First and Second Halves 01' the Precrit (...truncated)