[328 HANDBOOK OF PHVSIOLOGV ^ NEUROPHVSIOLOGV II 



FIGS. 4 AND 5. Diagrams of the multiple object problem showing examples of the three and 

 the seven object situations. Food wells are indicated by dashed circles, each of which is assigned a 

 number. The placement of each object over a food well was shifted from trial to trial according to 

 a random number table. A record was kept of the object moved by the monkey on each trial, only 

 one mosc being allowed per trial. Trials were separated by lowering an opaque screen to hide from 

 the monkey the objects as they were repositioned. 



describing the performance of the control and pos- 

 teriorly operated groups are similar whether total 

 repetitive errors (fig. 4) or search errors (fig. 7) are 

 plotted. In spite of the increasing complexity of the 

 succeeding situation, the curves appear little different 

 from those previously reported to describe the forma- 

 tion of a discrimination in complex situations (8, 130). 

 Although one might a priori expect the number of 

 repetitive responses to increase monoionically as a 

 function of the number of objects in the situation, this 

 does not happen. Rather, during one or another phase 

 of the discrimination, the number of such responses 

 increases to a peak and then declines to some asymp- 

 totic level (8, 130). Analysis of the data of the present 

 experiment has shown that these peaks or 'humps' can 

 be attributed to the performance of the control and 

 posteriorly operated groups during the initial trials 

 given in any particular (e.g. 2, 3, 4 • ■ • cue) situation 

 — i.e. when the monkey encounters a novel object. 

 The period during which the no\cl and familiar ob- 

 jects are confused is reflected in the 'hump' (fig. 8). 

 The importance of experience as a determinant of the 

 discriminability of objects has been emphasized by 

 Lawrence (75, 76). His formulation of the "acquired 

 distinctiveness' of cues is applicable here. In a progres- 

 sively more complex situation, sufficient familiarity 

 with all of the objects must be acquired before a novel 



object is sufliciently distinctixe to be readih' discrimi- 

 nated. 



But there is a difference between the control and 

 the posteriorly operated groups as to when the con- 

 fusion between itovel and familiar objects occurs. The 

 peak in errors for the group with posterior lesions lags 

 behind that for tlie controls — a result which forced 

 attention because of the paradoxically "better per- 

 formance" of this group throughout the five- and six- 

 cue situations (in an experiment which was originally 

 undertaken to demonstrate a relation between num- 

 ber of objects in the situation and the discrimination 

 "deficit" previously shown by this group). 



These paradoxical results are accounted for by a for- 

 mal treatment based on mathematical learning theory: 

 on successive trials the monkeys had to 'learn" which 

 of the objects now covered the peanut and which ob- 

 jects did not. At the same time they had to 'unlearn,' 

 i.e. extinguish, what they had previously learned — 

 under which object the peanut had been and under 

 which objects it had not been. Both neural and formal 

 models have been invoked to explain the results ob- 

 tained in such complex discrimination situations. 

 Skinner (130) postulated a process of neural induction 

 to account for the peak in errors — much as Sherring- 

 ton had postulated 'successive spinal induction' to 

 account for the augmentation of a crossed extension 



