a relatively greater proportion of disappearances of nonfixation points. In group M, for instance, the mean proportion of fixation-point disappearances was only 0.19 for figure B and 0.16 for figure C. The same pattern of results was obtained when the reported dimmings (rather than disappearances) were analyzed. In no category were the mean proportions for dimmings significantly different from those for disappearances. Hart specifically noted that none of his subjects reported the disappearance

more »

... of point d, one which was never fixated. In the present study, however, 7 of 12 subjects in group R and 15 of the 24 subjects in group M reported at least one disappearance of that point. Even though our data for group M do not reveal a preponderance of fixation-point disappearances, a fixationpoint effect is evident when only the disappearances of points a, b, and c are considered. When the proportion of fixation-point disappearances was calculated for each subject, only these three points being considered, the mean proportions were 0.54 for figure B and 0.60 for C. Both proportions are significantly greater (p < .01) than 0.33, the expected proportion under the assumption that the fixation point does not affect disappearances. Our data also differed from Hart's in the frequency of reported disappearances. Considering only the total number of disappearances reported by each subject, the means for our replication were 17 for figure B and 20 for C, whereas the respective values calculated on Hart's original data were 35 and 30. Only if the reports of dimmings are pooled with those of disappearances do the values for group R compare to those of Hart. However, Hart's preponderance of fixation-point disappearances can not be related in any simple way to this difference in frequency, since the comparable values for our group M were 71 and 72, more than twice the values Hart obtained. Obviously, the modified instructions not only increased the proportion of lines reported, but also significantly (p < .01) increased the overall rate of reporting as well. In general, our evidence supports McKinney (5) and Clarke and Evans (6) who suggested the possibility of two kinds of fragmentation: point-offixation disappearances related to receptor distribution on the retina, and more structured fragmentations having a central neural basis. Fixation-point dis-appearances are not the most likely to be reported, however, unless the experimenter inadvertently induces a response bias. Consequently, the fragmentation problem should continue to have important implications for current theories of perceptual organization. Science 145, 1070 (1964)] compares two methods for training rats to avoid shock in a lever-pressing apparatus. In both methods shock could be avoided by pressing the lever during the 7.5-sec interval between the onset of a light signal (conditioned stimulus) and the beginning of shock (unconditioned stimulus). In the first method failure to avoid resulted in a train of 0.2-sec shocks spaced at intervals averaging 13 sec. This shock was described as inescapable because a response made after the train of shocks had begun did not terminate a shock pulse if one happened to be on at the time, although it did terminate the series and turn off the light. In the second method, failure to avoid resulted in a continuous shock which had a maximum duration of 10 sec. This shock was described as escapable because a response during the shock terminated it and the light simultaneously. Hurwitz presents convincing evidence that the rats learned to avoid the inescapable but not the escapable shock. In attributing this result to the difference in escape contingency, the