Time Course of Spatial RF Shift

The analysis of the spatial sensitivity modulation so far was based on the neuronal response to the RF probes averaged over an interval from 60 to 200ms relative to the onset of the RF probe. Averaging spiking activity assumes that the attentional effect on the spatial sensitivity is stable throughout this time period of the presence of the flashed probe stimulus. However, it could well be that the spatial modulation of sensitivity is not present throughout the trial but due to the transient capture of attention by the abrupt luminance and motion onset of the probe stimulus (Yantis and Jonides, 1990; Yantis, 1998; Egeth and Yantis, 1997). According to this view, probe stimuli close to the attended target stimulus would capture attention more effectively than probe onsets more distant to the attentional focus. As a consequence of such an attentional capture, neuronal responses would be affected more strongly for probe locations close to the attended target stimulus. This would create a spatial shift of response strength (or sensitivity) similar to the one that we observe. The two interpretations should be distinguishable on the basis of a temporal analysis of the spatial shift. A putative transient, abrupt onset effect of attentional capture should be evident in a spatial shift of the RF stimulus that is restricted to the later part of the probe response. The on-response phase of the two probe presentations should be less modulated than the sustained, tonic response period, because the attentional effect would be initiated only at the time of probe onset and will be instantiated only with a temporal delay. Such a delayed effect contrasts to a sustained effect of spatial attention which would affect already the first spikes in response to the probe and will also be evident in the earlier on-response time.

We therefore analyzed the time course of the spatial sensitivity shift in response to the RF probe stimuli. For this we computed the RFs for successive 20ms intervals following probe onset and averaged the activity orthogonal to the axis of S1 and S2, i.e. orthogonal to the direction of the attentional shift between the inside condition (the method is indicated in figure 3.20, A). Rather than considering the average stimulus-evoked response to the probe stimulus, we applied a modified version of reverse correlation of the spike times which was triggered on the onset time of the RF probe stimulus, rather than with regard to the mere presence of the probe stimulus as would be used for conventional reverse correlation. The spatial reverse correlograms are computed in analogy to the spike-triggered averages applied for the direction tuning (cf. figure 2.4, p. 55), but with the stimulus vector reflecting the two spatial dimensions (x and y) of the position of the probes, rather than the single dimension (direction) in the direction tuning correlograms. Similar methods with different visual stimulation regimen are widely applied in the analysis of striate and extrastriate visual response properties (cf. e.g. De Angelis and Ohzawa 1993;

260

% shift of RF center (rel. to S1-S2 distance)% shift of RF center (rel. to S1-S2 distance)

n: 57

n: 57

Figure 3.20: Spatial reverse correlation analysis of the time course of the spatial RF shift. A: Two dimensional RFs were obtained for successive 20ms bins following probe onset (i.e. at 0msec). Activity was then averaged orthogonal to the X-axis to obtain a X-T plot. Stimuli inside the RF always lay along the X-dimension of the plot. The center position of the RF at each time slice was compared between the conditions with attention inside the RF (S1 and S2). B: The spatial RF shift of theattend inside conditions (attend S1 versus attend S2) for different times aligned to the onset of the probe. Positive values indicate a shift towards the attended stimulus. Error bars are the 0.95 confidence limits of the spatial difference of the centroid of the two conditions for the set of 57 cells. The graph shows that RFs are shifted towards the attended stimulus already 80ms following probe onset and remain shifted until 240ms from stimulus onset. The horizontal bar in the upper right reflects the time of probe presentation. C: Time course of the spatial shift as inB but aligned to the time of the peak response of the correlograms of individual cells. The graph shows that the maximum RF shift is found at the time of the maximum response to the probe stimulus, which shows that it is evident already in the transient on-response to the probes.

De Angelis et al. 1999; Livingstone and Tsao 1999; Livingstone, Pack, and Born 2001).

Figure 3.20 illustrates the resulting X-Y RF maps for an example neuron for successive time intervals from probe onset. The spatial layout of the RF maps were always rotated to align the X-dimension parallel to the axis of stimuli S1 and S2 inside the RF. In other words, the shift of attention from S1 towards S2 between the attend inside conditions always ocurred along the x-dimension. In order to

reduce noise of the correlograms we then averaged activity of the Y-dimension, i.e.

orthogonal to the axis of the attentional shift, i.e. of the S1-S2 axis. As a result we obtained an X-versus-Time (X-T) plot representing the probability of observing spikes in response to a probe stimulus at a particular x-position at a particular time from probe onset. The shift of the RF at different times of the X-T plots were obtained in the same were as in the analyis of the RF slices: After obtaining the center of the RF at each time, we computed the spatial difference of the RF center between the two conditions with attention to either S1 or S2, with positive spatial differences indicating that the RF center lies closer to the attended stimulus at that time point. We find that the spatial shift becomes statistically significant already 80ms after probe onset and is continuously present until 240ms following probe onset (cf. fig. 3.20, B). This time course mimics the sensory latency of the probe response and suggests that the RF shift is not due to a transient effect induced by the abrupt onset of the probe since this would require an additional delay because of the reaction time to the probe. This is also supported by an additional analysis of the RF shift aligned to the time of the maximum (peak) response to the probe (cf. fig. 3.20, C). The RF shift is strongest at the time of the maximum probe response, again indicating that the transient on-response of the probes is already spatially modulated by attention.