Neuroelectric correlates of conditioning

R. Hall
1976 Science  
al. (1) reported changes in multiple-unit responses in the rabbit's medial geniculate body associated with the acquisition and reversal of a discriminative conditioned avoidance response. In both stages of the experiment the positive conditional stimulus (CS +), which was followed by shock, purportedly evoked larger neural responses than the CS-, which was not followed by shock. The conditioned discrimination and its reversal were regarded as adequate conditions for "producing unambiguous
more » ... ative effects." The controls employed by Gabriel et al. are appropriate for one kind of nonassociative effect, specifically, effects that are not correlated with conditioned changes in behavior. They are not adequate, however, for nonassociative effects that actually depend on conditioned changes in behavior. Suppose, for example, that an increase in level of arousal leads to an increase in neural activity evoked by a CS. Such an increase in arousal can result from the presentation of a noxious unconditional stimulus (UCS) like electric shock. A general increase in arousal might occur in an aversive conditioning situation and have little or no relationship to the conditioned changes in behavior. It might, nevertheless, enhance the neural activity evoked by the CS as long as the heightened arousal is maintained by the repeated presentation of shock in the conditioning procedure. Such a change in evoked activity would be revealed as nonassociative by the discrimination and reversal controls employed by Gabriel et al. If, on the other hand, the increase in arousal were itself conditioned together with, say, an instrumental avoidance response, one might expect the conditioned arousal to lead secondarily to a "conditioned" increase in the CSevoked response. Is this an associative change in the evoked response? Only in a trivial sense, because it is not unique to the conditioning operation and throws little light on the conditioning process. It is, nevertheless, correlated with condi-246 tioned changes in behavior, and one would expect it to remain so throughout discrimination and reversal learning. It is possible that the effects reported by Gabriel et al. are of this kind, although that is not the only possibility. Other behavioral changes, for example, in orientation toward conditional stimuli, may also be confounded with conditioned changes in behavior, and one needs assurances that adequate measures have been taken to eliminate such possibilities. Level of arousal seemed the most appropriate variable to illustrate the argument, however, for we (2) and others (3) have shown that changes in late components of evoked activity in primary afferent pathways during conditioning can reflect mainly conditioned arousal or fear responses. Such changes in the later components of evoked activity remain a strong possibility in the experiment by Gabriel et al. To show that modifications in evoked activity are in some way unique to a conditioning process or are primary changes not dependent upon behavioral modifications has become a demanding task. This is not to argue that such modifications cannot be found; there is some evidence for them (4), but very little considering the numerous claims. Ad hoc arguments that the data have not been compromised are encountered more often than adequate controls. My main intention in this note has been to call attention to conceptual difficulties in the study by Gabriel et al. which are not, however, peculiar to this study. There are, however, several technical shortcomings in their report. Consider, for example, the data in the lefthand column of their figure 1, which presumably support the conclusion that in the final stage of acquisition the geniculate responses to the CS + were larger during the first 40 msec than the responses to the CS-. This appears to be the case for subject 44; but the opposite relationship is seen in the data from subject 42, and it is difficult to distinguish any systematic differences between the curves for the other three subjects. There is no statistical evaluation of this difference. Moreover, the arbitrary selection of different points on the curves of individual subjects as illustrative of significant differences is at variance with accepted statistical practices; and the standard deviations used as the measure of those differences have no relevance to statistical decisions about the differences between the curves for the CS+ and CSconditions. The reversal data in the right-hand column are only a little less disturbing, especially in view of a confounding difference between the initial (preconditioning) amplitudes of the responses to the two CS's (which Gabriel et al. were preparing to explain in a later publication) and a statistical evaluation based on not just the early ( 5 to 40 msec) activity, but on the complete response. Hall agrees that differential conditioning and reversal of short-latency medial geniculate nucleus (MGN) neuronal activity (1) is an associative neuronal effect. However, he argues that the effect may be associative "only in a trivial sense." Hall's judgment of triviality seems to us to be based on his belief that our effect "depend[s] on conditioned changes in behavior." We presume Hall expects a trivial neuronal effect to be "correlated with conditioned changes in behavior . . . and to remain so [correlated] throughout discrimination and reversal learning." We interpreted Hall's phrase, "depend[s] on conditioned changes in behavior," to mean either (or both) of the following: (i) the associative neuronal responses were mediated by prior conditioned behavioral activity; (ii) the effects were positively correlated with conditioned behavioral activity in all stages of conditioning and reVersal. Neither assertion is descriptive of the data. The latencies of the differential neuronal responses (5 to 40 msec) were too brief for those responses to have been mediated by prior conditioned stimulus (CS) related behavioral responses. In fact, it is unlikely that the briefest latencies of neu-SCIENCE, VOL. 193 on
doi:10.1126/science.935868 pmid:935868 fatcat:3za4go6kjfflndhowstotgwjea