Attentional differences between gaze and arrows processing: where vs what are eyes looking at [Annotation]

Review of “Attentional differences between gaze and arrows processing: where vs. what are eyes looking at.” Reviewer 2: Xiangyong, Yuan. • The submitted study attempted to explain the differential congruency effects of gaze and arrow. The authors divided the participants into a direction and a color group, and expected a reversed congruency effect of gaze and a standard congruency effect of arrows when participants were required to judge the direction. However, they expected both a standard congruency when participants had to judge the color of the frame being directed towards. The authors had summarized several relevant studies, and proposed their two working hypothesis (absence of a specific object to look at, or following the focus of interest of a social partner). The results replicated their previous studies and revealed something new. I appreciate the authors’ effort to preregister their study before data collection and conduct a Bayesian analysis to support their results. But I have some major concerns about the hypothesis and the experiment design that prevent me from accepting this version of manuscript as published. In the introduction, the first hypothesis the authors proposed is the eye contact, but it has been abandoned later by a previous study showing the reversed congruency effect disappeared when the eye color was judged. In the second hypothesis, gaze direction is seen as a sign that signals the intention of a social partner by his or her focus of interest. But it seems a little difficult to tell the exact differences between these two hypotheses. Both assumed the participants attend the eye regions to decode the intention under the gaze direction. Is this process another type of eye contact (e.g.,an active contact)? A further clarification about the second hypothesisis needed. Also, if the social intention of the gaze direction elicits the reversed congruency effect, it is read out from the gaze direction. Either in the task which required to judge where the eyes look at, or in the task which required to judge the color of the object they looked at, participants should attend the eye regions and discriminate the gaze direction as an initial step. The two tasks thus do not apparently differ with each other in nature. So it is not surprising a similar reversed congruency effect was found. By contrast, in Narganes-Pineda et al (2022) when judging the color of the eyes, the gaze direction is completely orthogonal to the task goal. Some additional discussions about the task requirement are necessary in the Discussion section. Authors response: Although we think that both hypotheses are based on different assumptions, they can be mixed up because of their similarities. We are sorry for not have previously explained more clearly the differences and hope it is now clearer. As we have mentioned in the introduction section, the eye-contact hypothesis was the first tentative explanation for the reversed congruency effect. The perception of a direct gaze (i.e., that the eyes on the screen were looking at the participant’s eyes), even incidentally, is known to influence a wide range of cognitive processes (Conty et al., 2016). This distinctive difference between eyes and arrows was thought to explain the dissociation in the congruency effect, i.e., the RCE. Narganes-Pineda et al. (2022) tested the hypothesis that color detection while being orthogonal to direction, was also subjected to the same putative eye contact on incongruent trials. However, the reversion was not found and was therefore discarded as an explanatory hypothesis. Perhaps some kind of eye contact is happening, but, at least, it does not explain the dissociation between arrows and gaze congruency effects, because no reversion was observed when having to discriminate color instead of the gaze direction. Our hypothesis is more closely related to attentional orienting processes. All directional stimuli could produce a shift in attention. Those shared orienting mechanisms would explain why eyes or arrows lead to equal cueing effects. In the spatial interference task, there is also an additional dimension: the stimulus location. This direction-location mismatch would also yield a similar spatial conflict for both gaze and arrows. All these common processes were not supposed to be affected by instructions, so that, they will be present both in the color and the direction group. As the reviewer points out, both groups would share that "initial step". We hypothesized that what accounts for the gaze-specific outcome (the RCE) has to do with some processes beyond attentional orienting, and that’s what we wanted to target with our manipulation. Given the key role of gaze direction in inferring intentions, an attentional shift could incidentally trigger additional social mechanisms such as searching for the potential focus of attention (i.e., if you look in that direction, there must be something there). Gaze would not only bias attention in one direction but could complete the attentional act by selecting (or trying to select) the looked-at item. To make our hypotheses and goals clearer and more understandable, we have expanded and modified some parts of the introduction (lines 71-82; 86-93). Moreover, we describe more thoroughly the above-mentioned experiment of Narganes-Pineda et al. (2022) testing the eyecontact hypothesis (lines 68-70). • Furthermore, another factor may somewhat reduce this reversed congruency effect and made it marginally significant in statistic. In the experiment, the left and right border of the bicolor frame, where the gaze or arrows were displayed, are not equal in distance from the eyes located at the left or right side. The border incongruent with the gaze direction is always farther away than the border that is congruent. For instance, the eyes located in the right side looked at the right side of border in congruent trials, but looked at the left border in incongruent trials. Since RT in the incongruent trials probably increased when the border was farther from the eye, this may reduce the reversed congruency effect and amplify the congruency effect. Authors response: As the reviewer pointed out, the distance of colors is an added and potentially relevant variable to consider in the Color group since the color is farther in incongruent than in congruent trials. This difference should affect both gaze and arrows increasing the standard congruency effect on Color group. However, we found no interactive effect between Congruency and Group for neither gaze, F(1,48) < 1, p = .78, BF = 0.4, nor arrows, F(1,48) < 1, p = .85, BF = 0.76. In fact, there was no difference between Direction and Color groups on congruent, t(48) = 2.22, pholm = .18, or incongruent trials, t(48) = 2.29, pholm = .15, when responding to arrows. Responding to gaze, there was a difference for both congruent, Mdif = 46.5, t(48) = 2.68, pholm = .04, and incongruent trials, Mdif = 49, t(48) = 2.83, pholm =.03, but in the opposite direction to what would be expected if there were increased difficulty in incongruent trials for the Color group: there were faster responses in the Color than in the Direction group. Moreover, it does not seem that this potential difficulty detecting color in incongruent trials was affecting the gaze’s results as we found the reverted effect. • I believe that the inversed congruency effect may not be caused by an empty space in which no specific objects are searched for. But the current design which added a bicolor frame is not a straightforward and optimal solution to test a standard congruency effect (as it will amplify this effect). This may not be a severe problem after an inversed congruency effect have been obtained, however, if the congruency effect was expected as the authors hypothesized, this is not that ideal. Probably, to put a real object, e.g., a ball or simply a colored square, just at the left or right side of the eyes with equal distance? I would like to see the authors conduct a new stricter experiment, or at least discuss about the limitation in their manuscript. Authors response: We agree with the reviewer that this topic requires further research— probably through some variations in the standard paradigm—to fully understand the nature of the reverse congruency effect. Adding real objects instead of colors is definitely something to test in subsequent studies, which we will be testing in our lab. Also, the location of the objects and their distance from the stimuli are relevant variables to consider. Certainly, the modification of the task proposed by the reviewer would solve the problem regarding the distance of the objects but it would change the basic tenets of the paradigm. If the objects appear on one side of the screen, on either side of the stimulus, there would be no locationdirection interference. The stimuli would become the center of this new spatial configuration. In this case, the identification of the objects would be much more like a cueing task, which we already know is insufficient to disentangle gaze-specific effects. Taking into account the reviewer's observations, we extended the discussion sections adding some limitations and possible modifications to consider in future studies (lines 248-253). • Minor o L111, 118, the size of the stimuli, and their distance from the fixation should be provided in a unit of visual degree. Authors response: We added the suggested content to the manuscript (line 122). Based on this recommendation and to enhance replicability we have also added some missing data, such as the monitor resolution (Apparatus and stimuli section, line 112). o For gaze, in incongruent condition, there were faster RT but higher errors compared with congruent condition. Is there a speed-accuracy trade-off in this inversed congruency effect? Authors response: We appreciate this interesting observation. Indeed, even though the gaze’s congruency effect was numerically reverted in terms of reaction time, the errors followed a standard congruency pattern. Nevertheless, error data should be interpreted with caution because of a putative ceiling effect. As in other studies using a similar spatial task (Ishikawa et al., 2022) the overall accuracy was really high (M = 0.97, SD = 0.12) with no participant having an average error rate two standard deviations below the group mean. In fact, although the frequentist analysis suggested a significant standard congruency effect for errors, F(1,48) = 3.99, p = .05, ηp2 = .08, there was no conclusive Bayesian evidence BFincl = 1.25. The error analysis section has been updated including these data (lines 184 to 192).