SENSORY NEURAL PATTERNS AND GUSTATION 211 



pattern.* According to the present hypothesis this pattern is what signals 

 the quahty of NH4CI at least as far as these 13 fibers are concerned. 



A very important test of the pattern theory of taste quality sensitivity 

 is that if the quality of a stimulus is actually signaled by this kind of 

 patterning, then stimuli which give similar patterns should taste somewhat 

 alike. Conversely, stimuli producing highly dissimilar patterns should 

 taste considerably different. In Fig. 4 we see that KCl produces a pattern 

 quite similar to that produced by NH4CI, at least as contrasted with the 

 NaCl pattern. A measure of the similarity of these patterns is given by 

 the correlation between the amplitudes of the responses produced by 

 these solutions in these fibers. The product-moment correlation coefficient 

 between the amplitudes of responses to NH^Cl and KCl for these 13 

 fibers is +0.83, indicating a close similarity between these patterns. The 

 pattern produced by NaCl, as indicated by the unconnected open circles 

 in Fig. 4, is not very similar to either the KCl or NH4CI patterns. In 

 Table 1 are shown correlation coefficients based on a larger number of 

 fibers. Thus, the neural patterns predict that for the rat, KCl and NH4CI 

 should have similar tastes, but neither of them should taste very much 

 like NaCl. 



(It is clear from Table 1, in which are presented correlations between the 

 amount of response of the fibers tested to a number of taste solutions, 

 that NH4CI, KCl and CaCU all produce similar neural patterns, and that 

 LiCl produces a neural pattern very similar to that produced by NaCl. 

 High correlations also suggest that these stimuli are relatively adjacent 

 on the stimulus continua. However, it can be demonstrated that low 

 correlations indicate little more than that the neural patterns produced 

 are dissimilar ; the relative placement of these pairs of stimuli on the 

 stimulus continua are indicated more by the form of the correlation 

 scattergram than by the degree of correlation.) 



Now let us turn to the behavior predicted from the neural data. Two 

 stimuli may be defined as perceptually similar to the extent that a response 

 learned to one of them will generalize to the other. Thus, if these neural 

 patterns are the basis for taste quality sensitivity, a response learned to 

 one stimulus should generalize to another stimulus producing a similar 

 neural pattern, but not as much to one giving a different pattern. If such 

 predictions are borne out, it would be indicated that such neural patterns 

 actually are the basis of taste quality sensitivity. 



Two behavioral tests were employed, both based on the generalization 

 of shock-based avoidance of drinking from one salt to others. In the first 

 test, three rats each learned to avoid drinking one of three salts, KCl, NH4CI 



* Some other aspect of the across-fiber pattern of neural activity more subtle than 

 number of impulses in the first second of evoked activity may prove more predictive of 

 behavior. 



