To ensure that attention was deployed consistently during the viewing of the stimuli, subjects were instructed to detect subtle changes in the contrast of the stimulus texture that occurred randomly on 20% of the trials. In this task, subjects demonstrated a high level of accuracy, correctly identifying 94% of the contrast decrement targets. It can therefore be inferred that subjects were consistently vigilant and attentive to the presentation of the segmenting stimuli.
Sensor-Space Visual Evoked Potential results
In the present experiment we compare the visually evoked potentials to four types of 1 Hz, texture-modulating stimuli. Our description of these results proceeds first with a ‘sensor-space’ analysis (voltage as a function of electrode location) by describing the waveform morphology and global-field power (GFP) for the orientation-defined form and then each of the configuration and cue contrasts, as well as their double difference. We then show source analyses of waveforms derived from individually defined visual area ROI's (V1, V3A, V4, MT, and LOC) that are each sufficiently separated to be resolvable by our inverse method 
Waveform Morphology and Global Field Power
For all subjects, prominent responses were present after each of the two texture transients in the stimulus cycle. Grand average evoked responses are shown for the orientation-defined form, changing segmentation condition in . Here one cycle (1 second) of the grand average waveforms for all 128-channels are shown superimposed on top of each other and are presented above the global field power (GFP) of the response. As seen in both the waveforms and the GFP, the appearance and disappearance of a segmented figure results in a highly asymmetric response. Responses to the appearance of the segmented figure in the first half of the stimulus cycle (0–500 ms) are larger and more protracted than those to the disappearance of the figure, in the second half of the cycle (500–1000 ms). In particular, the appearance produces three high-amplitude peaks of activity, while the disappearance produces responses of lesser signal amplitude with only two peaks. In both cases, the evoked responses produce bilateral, occipital distributions that differ in polarity at successive peaks (see ).
Grand average waveforms and global field power for the orientation-defined form response.
Sensor-space configural selectivity contrasts.
Sensor-space differences due to configuration
For both cue types, cortical responses evoked by the changing and constant segmentation stimuli differed after the image updates at 0 ms, but not after those at 500 ms (e.g. the times of appearance and disappearance of the figure in the changing segmentation condition, respectively). As shown in for the orientation-defined textures and for the phase-defined textures, the changing-segmentation stimuli evoked responses of greater amplitude than responses to the corresponding constant segmentation stimuli for protracted periods following the onset of the segmented figure at 0 ms (solid and dashed curves, respectively). Configural selectivity, as determined by run-length corrected permutation tests comparing the changing versus the constant segmentation stimuli, was present from 148–262 ms for the orientation-defined textures (panel A) and from 143–365 ms for the phase-defined (panel B) textures (see black bars). For reference the uncorrected, sample-by-sample p-values are depicted by the yellow-red color map overlays for both configuration comparisons.
The two cue types produced qualitatively different topographic distributions after figure onset as shown in by the three topographic distributions plotted below the GFP traces. Whereas the onset of a phase-defined figure produced an initial bilateral component that peaked at 159 ms, the orientation-defined figure produced a lower amplitude difference that was more medial occipital in its initial focus. Scalp topographies at later time points (228 and 288 ms) did not differ substantially as a function of the defining cue type. These difference-wave distributions are explored further below in an ROI-based source space analysis, below.
Sensor-space differences due to cue type
Cortical responses evoked by the changing and constantly segmenting stimuli also differed as a function of the defining cue type. As shown in , the GFP evoked by the orientation-defined stimuli (blue) produced greater amplitude responses that those evoked by the phase-defined stimuli (red) both when these textures supported segmentation, and not. Statistically significant differences were present between the two changing segmenting textures from 119–253 ms following the appearance of the figure. Significant differences in cue selectivity were present to the constantly segmented stimuli from 119–174; 205–322; and 406–463 ms during the segmented phase of the cycle, and also from 636–689 ms of the uniform phase of the response (i.e. 136–189 ms following the return to the uniform state). In general, texture-cue differences were present during both segmented and uniform phases of the response cycle as indicated by cluster of high, uncorrected significance levels between 600 and 750 ms.
Sensor-space cue selectivity contrasts.
Source-Space Visual Evoked Potential results
In order to quantitatively assess the differences in configuration- and cue-selectivity in the brain, we performed a region-of-interest (ROI) analysis on the time-averaged responses, focusing on the first half of the response cycle where selectivity was greatest in the sensor-space results (i.e., 0–500 ms). Source current density reconstructions were computed for five ROIs centered on well-separated visual areas. These regions; LOC, hMT+, V4, V3a, and V1 are depicted for each individual subject in , are located on the lateral surface, the dorsal surface, the ventral surface, and the posterior pole, and can be defined reliably in all subjects.
ROI locations for all subjects color coded and shown from posterior and inferior perspectives.
Source-space differences due to configuration
The appearance of both orientation- and phase-defined stimuli evoked differential activity for the changing versus constant segmentation stimuli across many of the measured ROIs. These patterns of selectivity for the two cue-types both shared early run-length corrected activity in V1 and later selectivity in the LOC ROI. The dynamics of configural selectivity however, differed substantially across other ROIs and latencies (see for specific latencies).
Start and stop times for the run-length corrected permutation tests (depicted by the black bars in & ).
For the orientation-defined stimuli (), run-corrected differences emerged in V1 from 133–183 ms after figure onset. Orientation-defined stimuli also produced selective activity that didn't reach run-length-corrected significance levels in the V4 and MT ROIs over this latency range. Early configural selectivity for the phase-defined stimuli () was present from 143–183 ms, according to run-length corrected criteria in V1, but also at uncorrected levels in the V3A and LOC ROIs at these early latency ranges. For both cues, this early configuration selectivity occurs on the rising phase of the activity evoked by the segmenting texture transient. As seen in the first difference topography for each contrast in (150 and 159 ms), this difference is more lateral and negative in polarity for the phase-defined contrast, and more central and positive for the orientation-defined segmentation, and is in agreement with the underlying pattern of ROI activations.
Source-space configural selectivity contrasts.
At the time of the second GFP peak (228 ms), the scalp topography for both cue types comprises bilateral negative occipital potentials. Over this time range differential activity is present at the corrected criterion in the V4 ROI (212 to 248 ms) for the orientation-defined stimuli. In contrast, at the same latency no ROIs reached run-length corrected criteria for the phase-defined stimuli, but moderate levels of uncorrected selectivity were present in the V3A and hMT+ ROIs.
Later activity depicted by the 3rd topographic map for each contrast produced run-length corrected selectivity in the LOC ROI from 255–339 ms and 281–332 for the orientation and phase stimuli, respectively. Configural selectivity was also present in the V3A ROI over this same period, reached corrected significance levels for the phase stimuli (291–355 ms), and was robust but not significant for the orientation-defined stimuli. In addition, there were sporadic periods of differential activity in the V1 ROI over this latency range that did not meet corrected criteria. Overall, the phase-defined configuration exhibited more periods of differential segmentation-related activity than did the orientation-defined configuration, including activity in the V3A and hMT+ ROIs that was not present with the orientation-defined stimuli.
Source-space differences due to cue type
As seen in the sensor-space data, ROI differences to the two cues types occurs earlier than selectivity to the stimulus configuration. The earliest latency at which differential run-corrected activity is present is 114 ms in the V1 ROI for the constantly segmented stimuli (), and at 124 ms for the changing segmentation stimuli (). By contrast, the earliest significant configural selectivity occurs in V1 at 133 ms. Differences between cues are widespread in the early cortical areas for both the changing and constant segmenting contrasts. For the changing segmentation contrast (solid lines), there are early un-corrected differences present in all the ROIs between ~125 and 175 ms and later run-corrected differences in V3A and LOC. In comparison to the changing segmentation differences, cue selectivity for constantly segmented textures evokes relatively less early contribution from V4 and hMT+, but is otherwise quite similar.
Source-space cue selectivity contrasts.
In our previous work, we showed that several different texture cues led to a similar pattern of cortical activation, indicating that a certain degree of cue-invariance is present in the evoked response from lateral cortex 
. The overall pattern of activity seen with the present stimuli is similar to our previous findings in that both stimuli produce prominent responses in lateral cortex. To quantitatively compare the responses to the different cues, we computed the difference response between the two tsVEPs (orientation- and phase-defined).
This double difference () revealed that while the separate tsVEPs for each type alone produced significant differences across early latencies (e.g. 130–200 ms) there was no differences between these effects at these latencies. Similarly, no differences were seen at any latency in the V1, hMT+ and V3A ROI's. Significant differences were present, however, between 220 and 290 ms in the LOC ROI and between 250 and 300 ms in the V4 ROI. Cue invariant tsVEPs are thus present between 130 and 200 ms and in the V1, V3A and MT+ ROIs, but not at later time-points in the LOC and V4 ROIs.
Source-space cue-configuration double difference.