RF Size Changes and Modulation of Gain

The main rationale of the following comparison is to investigate the relation of gain and width at half-maximum responses (RF size) irrespective of the change of the spatial RF shift. As reported above we find RF shrinkage with attention inside versus outside the RF. We also reported in section 3.2.2 (p. 68) that the response strength (the average maximum probe response) was not significantly different be-tween the inside and outside condition. Taken together, these findings might suggest a shrinkage in the absence of modulation of response gain. However, previous studies involving a single stimulus inside the RF suggest the opposite, namely that response gain changes in the absence of changes in the half-width of tuning functions, which in our case corresponds to the half width of spatial tuning, i.e. RF size. More specifically, previous studies have shown that attention acts on sensory responses consistent with a gain mechanism, which is evident in multiplicatively scaled (Gaus-sian shaped) tuning curves for motion direction in area MT and orientation tuning curves in area V4 (cf. section 1.2.3 p. 16). This multiplicative attentional influence reflects a change in the gain of neurons in the absence of changes of the ratio of the gain and half-height width, or, in other words, in the absence of a narrowing of the tuning selectivity. We therefore set out to test whether this finding also holds true for changes of gain and size of RFs when attention is directed inside versus outside the RF.

It should be explicitly noted that the analysis relies on the assumption of Gaus-sian shaped RF profiles which has been suggested to hold true for the majority of area MT neurons (Raiguel et al., 1995). Figure 3.18 A (p. 88) illustrates the hypothetical outcomes based on the noted assumption that the RFs follow a Gaus-sian shape. According to the multiplicative gain hypothesis we would expect that changes in response strength do not change the half-height width (i.e. the size) of RFs which is illustrated in the left panel of figure 3.18 A. In contrast, theRF shrink-age hypothesis predicts smaller RF-half height sizes either without or with changes in gain (middle and right panel in figure 3.18 A). For the analysis we computed the ratio of the maximum response in the attend inside condition and the attend outside condition ¹. From the analysis of response strength presented in section 3.2.2 (cf. figure 3.6, B, p. 70) we already know that there is no significant average change in response strength for this comparison. We then computed the ratio of RF

1Note that the ratio is a different measure than the attentional index used to measure the modulation of response strength and shown in figure 3.6, B, p. 70. The attentional index reflects a normalization to bring values in the range of ±1, while ratios, on the other hand, can result in arbitrarily large values. In particular, when we compare the average maximum probe response in the inside and outside condition we find an average, non-significantnegative index value of -0.03, corresponding to a geometric mean of 5.5% lower responses in theattend insidecondition. Taking the ratio of the same response values, however, results in an average non-significantpositive ratio of 1.01. The distribution of the ratio is plotted in figure 3.16, B.

B C

degree visual angle degree visual angle RF-attend outside

RF-attend inside

Figure 3.18: RF size changes and its relation to gain. A: Illustration of the relation of gain and width of a one-dimensional Gaussian model (as a simplification of the two dimensional nature of neuronal RF). The spatial tuning curves might change in gain without (left panel) or with (middle panel) a change in size. Otherwise, only the RF size could change without a concomitant gain change (right panel).

Obviously RF shrinkage should always involve changes in RF size. B Comparson of gain changes with changes in RF size based on the ratio of the inside versus outside condition for width (smaller values = shrinkage) and the ratio for gain (response strength). There is no systematic relation of changes in gain and width for the population of cells. (cf. text for details, and footnote 1). C Illustration

half-height in the inside and outside condition based on the values presented in the previous section. Figure 3.18, B shows the result, with a mean ratio of gain and width of 1.01 and 0.99, respectively. Size and gain ratios are distributed with no systematic relation of changes in gain and RF size (figure 3.18, C). In other words, this result on its own does neither support any sort of shrinkage hypothesis, nor a pure response-gain hypothesis, because neuronal RF profiles reveal only a marginal change in gain and size.

In summary, in contrast to the absence of clear and systematic gain and size changes the shift of spatial sensitivity remains the strongest significant effect on its own. This finding is illustrated in the shifted (red coloured, ) one-dimensional Gaussian RF model in the upper panel of figure 3.18, C. We find a shift in the absence of clear gain increase or decrease. The lower two panels in figure 3.18, C illustrates

possible outcomes of RF shifts of the same extent but with concomitant changes in gain. Neither of these possibilities reflect the average effect of our data: We did not find a RF shift that is brought about by a spatially specific gain increase, nor by a gain decrease. Rather, we report a RF shift towards attended stimulus positions without systematic changes in gain and RF size.