Foveal feedback supports peripheral perception of both object color and form

Evidence from neuroimaging and brain stimulation studies suggest that visual information about objects in the periphery is fed back to foveal retinotopic cortex in a separate representation that is essential for peripheral perception. The characteristics of this phenomenon has important theoretical implications for the role fovea-specific feedback might play in perception. In this work, we employed a recently developed behavioral paradigm to explore whether late disruption to central visual space impaired perception of color. First, participants performed a shape discrimination task on colored novel objects in the periphery while fixating centrally. Consistent with the results from previous work, a visual distractor presented at fixation ~100ms after presentation of the peripheral stimuli impaired sensitivity to differences in peripheral shapes more than a visual distractor presented at other stimulus onset asynchronies. In a second experiment, participants performed a color discrimination task on the same colored objects. In a third experiment, we further tested for the foveal distractor effect with stimuli restricted to a low-level feature by using homogenous color patches. These two latter experiments resulted in a similar pattern of behavior: a central distractor presented at the critical stimulus onset asynchrony impaired sensitivity to peripheral color differences, but, importantly, the magnitude of the effect depended on whether peripheral objects contained complex shape information. These results taken together suggest that feedback to the foveal confluence is a component of visual processing supporting perception of both object form and color.

4 64 250ms post-stimulus onset, as well as beyond 400ms post-stimulus onset, did not have the same 65 disruptive effect on discrimination sensitivity. Taken together, these studies suggest a form of 66 feedback that constructs a representation of objects removed from the associated visual input 67 and, further, that this feedback is behaviorally relevant.

68
To date, studies examining the foveal feedback phenomenon have largely employed a 69 relatively difficult behavioral task where the participants discriminate between briefly-presented 70 novel greyscale objects [9,10]. However, in Williams et al. [6], the authors included one 71 experiment where participants performed a color discrimination task on colored objects 72 presented in the periphery. In that experiment, unlike the shape discrimination task, the authors 73 did not find information about object form at the fovea, raising the possibility that foveal 74 feedback is related to the task at hand. The authors did not, however, test whether color 75 information could be decoded in foveal retinotopic cortex. Therefore, it is unknown whether 76 foveal feedback is limited to carrying general shape information of visual stimuli, or if it may 77 function for any one object characteristic related to the task being performed. 78 We have previously reported a behavioral measure of foveal feedback [10]. In brief, 79 participants perform a discrimination task on achromatic novel objects briefly presented 80 (~100ms) in their periphery while fixating centrally. An achromatic visual distractor presented at 81 fixation impairs discrimination sensitivity when it appears 117ms after target onset, after the 82 targets have disappeared from the display. This disruption in discrimination sensitivity at 83 +117ms post-stimulus onset reliably occurs when a central distractor is presented to the observer 84 at a time entirely disparate from the target presentation, and is more pronounced compared to 85 distractor onsets at other stimulus onset asynchronies (SOAs), including SOAs later in a trial 86 (e.g., more than 250ms). In a previous paper, we termed this temporally-specific disruption of 5 87 peripheral discrimination sensitivity the "foveal distractor effect". We [10] also demonstrated 88 spatial specificity of this effect: discrimination sensitivity was not similarly impaired when a 89 visual distractor was presented in the periphery at the critical SOA. This behavioral paradigm 90 demonstrates the spatial and temporal specificity of foveal feedback and is an efficient method 91 for investigating how feedback influences peripheral perception (see also 6, 7). 160 The targets were presented in these same locations throughout the training tasks and the 161 experiments (Fig. 1). Participants were instructed to maintain fixation on the central cross 162 throughout each trial and determine if the two targets in the array were different shapes or if they 163 were identical in shape as quickly and accurately as possible, while ignoring the color of the 164 targets. In "same" trials, the targets were always presented in the same orientation. In half of the 165 trials, these target stimuli were different shapes, chosen at random from the larger set of 16, and 166 in the other half they were identical shapes. The targets, regardless of whether they were the 167 same or different shapes, always differed in color by a hue angle of 60°. The degree of color 168 difference was selected based on pilot data, such that participants' performance on a shape 169 discrimination task would be similar to their performance in a color discrimination task using the 170 same stimuli (see Experiment 2). Participants had 2000ms to respond with their right index 171 finger or middle finger on the keyboard to indicate a "same" or "different" judgment, 172 respectively. Following each response, participants were given onscreen accuracy feedback.  194 Experimental procedure 195 The procedure for Experiment 1 was similar to the training procedure with two major 196 changes: there was a fixed target presentation duration of 117ms and a distractor object appeared 11 220 targets. There were 80 trials for each of the ten conditions (40 "same", 40 "different") for a total 221 of 800 trials in a session. All of the trial types were randomly intermingled, fully crossed, and 222 blocked so that participants would have a chance to rest every 100 trials.

223
Participants were given 3s to respond after the completion of the trial before the next trial 224 automatically commenced. As in the training task, participants used their right index finger to 225 indicate a "same" judgment or their right middle finger to indicate a "different" judgment.
226 Following each response, participants were given onscreen accuracy feedback.

227
As in training, participants were instructed to maintain fixation on the central cross 228 throughout each trial and respond as quickly and accurately as possible. The eye-tracker was 229 unavailable for six participants. However, given the short duration of the target display as well as 230 the disparate peripheral target locations, any eye-movements towards the peripheral stimuli are 231 likely to have impaired behavioral performance on the task, as only a single target would be able 232 to be fixated (if that) during the display, which would make the second target further from 233 fixation, making it more difficult to compare the two stimuli. In the cases of eye-tracked 234 participants, we had to discard only 0.08% of completed trials from analysis due to eye-235 movements. Participants were able to complete the experimental task in ~45 minutes.

236
We did not include a non-distractor condition in the main experiment because the training 237 task was effectively the discrimination task without a distractor. Additionally, a non-distractor 238 condition differs from the experimental distractor-present condition. Thus, a 'no-distractor' 239 condition would not be a good baseline as performance could be better due simply to practice or 240 the other changes. Instead, we used performance in the -267ms SOA condition as a baseline for 241 comparison as it is matched the experimental conditions in all key aspects with the only 242 difference being the onset time of the distractor.

244
Our dependent variable was d' as a measure of discrimination sensitivity for comparing the 245 targets. The hit rate was defined as the proportion of correct "same" responses on "same" trials, 246 and the false alarm rate was defined as the proportion of "same" responses on "different" trials 247 (see Table in S1 Table). We ran a two-way repeated measures ANOVA on d' with the factors of 248 SOA (-267ms, -117ms, 0ms, +117ms, +267ms) and distractor type (grey, colored). We applied a our key SOA of +117ms, discrimination sensitivity (d') was impaired compared to our relative 267 baseline -267ms (p < 0.001), as well as compared to -117ms (p < 0.001), 0ms SOA (p < 0.001) 268 and +267ms SOA (p < 0.001; Fig. 3). The only other significant difference was that 269 discrimination sensitivity was significantly lower at 0ms SOA than -267ms SOA (p < 0.001). No 270 other comparisons approached significance after correction (p > 0.005; see Table in S2 Table). The stimuli and apparatus were the same as in Experiment 1. Prior to taking part in the 292 experiment, participants were trained on a basic discrimination task similar to the training for 293 Experiment 1, except that in Experiment 2, participants discriminated between the colors rather 294 than the shapes of the objects. The shapes of the target objects in Experiment 2 were always 295 different, randomly chosen from the set of 16 exemplars. Participants were instructed to ignore 296 the shapes of the targets and make a judgement on whether the colors of the targets were 297 identical or different. In each trial, one color was chosen at random between the hue angles of 0° 298 and 200°. In "same" trials, the objects' colors were identical. In "different" trials, the second 299 target's color always differed by a hue angle of 60°. The degree of difference was determined 300 based on pilot data such that participants would be able to discriminate between the two colors 301 with a similar accuracy as when doing the shape task described in Experiment 1, and the range of 302 colors was chosen to complement the variability in the shapes of the exemplars. The parameters 303 of the training task were the same as in Experiment 1 (see Fig. 1). Participants were trained until 304 they were able to make at least 70% correct discriminations when the target array was displayed 305 for 117ms, while maintaining fixation throughout the block.

306
Experiment 2 was carried out in a similar way to Experiment 1 (see Fig. 2), but the 307 required task was different: participants were asked to judge whether the two colored objects 308 were the same color while ignoring their shapes.