Deficiency of object-based attention specific to the gaze cue is independent of top-down attentional strategies.

This study investigated whether modes of attentional selection (location-based or object-based) are modulated by the cue type, specifically social cues such as eye gaze and pointing fingers, or by a non-social cue, such as an arrow. Earlier studies have demonstrated that the object-based attention effect was found only with arrow cues when presenting a spatial cue at either end of a rectangle: gaze cues did not yield object-based facilitation. We examined whether this deficiency of object-based attention is generalized to social cues such as pointing fingers. We measured the reaction times to the target at each cued location, an opposite side of a cued location in the same object, or the location in a different object equidistant from the cued location for each cue. Results indicated that only the gaze cue weakened the object-based attention effect, even under the condition of participants' voluntary extension of their attentional focus. The pointing cue induced sufficient object-based facilitation, as did the arrow cue. These results suggest that the deficiency of object-based attention was observed only for the gaze cue, and that it would be caused by a factor that is unique to the gaze cue, which narrows the attentional focus.

Keywords: gaze-cueing effect; gaze perception; object-based attention; spatial cueing; visual attention

Sharing attention with others is crucially important for social interaction. We shift our attention to another person's direction with our eyes, or we point with a finger to share visual information with others. Furthermore, we can infer someone's intentions and feelings from their gaze direction ([2]). This ability to share our attention with others using gaze direction or pointing with fingers, known as joint attention ([49]), plays an important role in the inference of another's mental state. In fact, it is expected to support the development of the theory of mind in infants ([3]).

For about two decades, many researchers have reported that another person's eye gaze direction influences the observer's attention. The typical method of attentional orienting by gaze direction is the pre-cueing task paradigm ([40]). In this paradigm, participants are asked to localize or identify visual targets presented after the central gaze cue orients to the left or right. Results of such studies show that participants have a faster reaction to the target presented in the location indicated by the gaze cue than in the opposite location ([14]). Moreover, the response is faster than when a neutral cue (with direct gaze) is presented. [17] described this attentional orienting triggered by gaze: The gaze-cueing effect occurs even when the gaze direction is irrelevant to the task: the probability of validity between the cue and target is 50%. The gaze-cueing effect occurs even when the cue-target stimulus onset asynchrony (SOA) is very short (105 ms), unlike the attentional shift by typical central symbolic cues, such as an arrow cue or number.

Such uniqueness of orienting by unpredictive gaze cues is also supported by results of neuroimaging studies ([25]; [39]). Results of these studies suggest that gaze cues activate specific brain regions that are responsible for biological stimuli such as head or body orientation (i.e., superior temporal sulcus [STS]) and that gaze cues rely on different neurological structures than biologically irrelevant stimuli, such as arrow cues, do. Nevertheless, some reports of earlier studies have described that attentional orienting by the arrow cue has almost identical properties to those of a gaze cue ([18]; [26]; [42]). For instance, both the arrow and the gaze are regarded as similar in that they cause strong and automatic attentional orientation because they cannot be ignored, even in situations for which the target position can be predicted ([18]). In addition, both non-predictive gaze cues and arrow cues produce similar attentional effects with different SOAs ([42]). Therefore, shifts of attention by the gaze cue and the arrow cue are behaviorally indistinguishable in terms of their temporal properties. Moreover, at neural levels, arrows and eye gaze have been shown to involve similar brain areas and ERPs ([4]; [5]). A meta-analysis of 67 studies comparing gaze and arrow cues ([8]) showed no quantitative difference between gaze effects and arrow cue effects. This finding implies the difficulty of distinguishing the attentional orienting triggered by gaze and arrows, at least for tasks using the classical attentional shift paradigm ([40]) or a similar paradigm. However, several studies have examined qualitative differences in social cues such as gaze and nonsocial cues such as arrows (e.g., [23], [24]; [36]). [23] reported the influence of social and nonsocial cues on a working memory (WM) task that required participants to recognize a colored tile that was presented simultaneously with the cue. Results showed that WM performance improved in the condition in which the gaze cue was oriented toward the color tile, but not with the arrow cue. Moreover, in a later study ([24]), they showed that the WM performance was not improved when a barrier was placed between the gaze cue and the tile. These study results suggest that higher processing of others' intentions or mental states is included in the processing of gaze cues. It is particularly interesting that the times to locate the tile did not differ between those achieved with and without barriers: faster responses to targets presented in the gaze cue direction were observed irrespective of the presence or absence of barriers.

Some earlier studies have further examined spatial properties of attentional orienting by gaze compared to arrows (Chacón-Candia et al., 2020; [35], [34]). These approaches rely on the finding that the unit of attentional selection is based also on objects, not solely on spatial location ([15]; [16]). [16] presented two rectangles placed horizontally or vertically parallel to the object stimulus. After a pre-cue presented at one end of one of the rectangles, a target was presented at any of three locations at the end of the rectangles. The target could therefore be presented either in the same location as the pre-cue (valid trials), the opposite location of the cued object (same-object trials), or in a horizontally or vertically different location on the un-cued object (different-object trials). It is noteworthy that targets in two spatially invalid cue trials (same-object trials and different-object trials) were equidistant from the cued location. Results indicate faster detection of the targets in the valid trials than in the two invalid cue trials (location-based attention). Moreover, participants detected objects faster in the same-object trials than in the different-object trials, although the two conditions were equivalent in terms of distance from the cued location. These results suggest that visual attention activates processing for whole objects as well as a specific spatial location. [35] investigated whether the object-based effect of visual attention would occur also with gaze cues. They used the same paradigm as that used for the study of [16], with arrow cues and gaze cues. Their results clarified that the object-based effect was found only when arrow cues were presented. By contrast, when gaze cues were presented, no object-based effect occurred: only the location-based attentional effect was found. They argued that the deficiency of object-based attention, even when the location-based attention was sufficiently induced in the gaze cue condition, is mediated by the theory of mind mechanisms. In other words, this absence of object-based effects in the gaze-cueing of attention is related to the ability to understand another's intention: eye gaze jointly orients our attention to a specific position rather than to the whole object. Actually, we use gaze direction as a tool for interpreting others' intentions and mental states.

Several studies have used finger-pointing as an attentional cue ([13], [12]; [21]; [22]). These studies have examined clinical groups and infants to assess the orienting of attention by pointing finger, gaze, and an arrow. Results showed that these types of cues produce attentional orienting in a specific manner. For example, anorexic patients can show attentional cueing by pointing fingers at short SOA (100 ms; [12]). Patients with schizophrenia showed no attentional effect with gaze only, which is probably related to abnormal function in the STS region ([13]). Another study of 3–10-year-old children ([21]) indicated that children younger than five years old have difficulty in saccading from a gaze cue to the target. That study also demonstrated that the orientation of attention emerged at 8–10 years of age. [22] report that socio-biological cues (gaze and pointing finger) automatically recruit the oculomotor system, but non-biological cues (arrows) do not. However, it is noteworthy that these cues show similar patterns of orienting attention on simple cueing tasks, as measured by reaction times (RT) of healthy control subjects ([12], [13]; [42]).

If the deficiency of object-based attention by the gaze cue is based on systems such as joint attention or theory of mind mechanisms by which the gaze cue emphasizes only the specific location of a part of the object, then that phenomenon might be generalized to socio-biological cues. Developmental research has shown that pointing gestures, as well as gaze direction, function as tools to interpret the intentions and mental states of others for preverbal infants ([11]). Specifically, the use of nonverbal cues such as gaze direction and pointing gestures plays an important role in infants' understanding of language ([50]). Results of another study suggest that pointing with an index finger shifted an observer's attention more automatically than pointing with other fingers, a thumb, a little finger, or a clenched fist ([1]). In other words, pointing with an index finger strengthens attentional orienting more than pointing with any other finger or fingers. Although it is open to discussion whether the influence of these pointing gestures on visual attention reflects higher cognitive systems such as the theory of mind mechanisms, pointing with the index finger, even by adults, is an important tool for referring to the intentions and actions of others.

This study specifically examines how spatial characteristics of attention triggered by socio-biological cues (eye gaze and pointing finger) differ from those by typical central cues (arrows). Few reports have described studies examining the gaze cueing effect while particularly addressing the spatial characteristics of visual attention (e.g., [35], [34]). Therefore, it remains unclear whether the deficiency of object-based attention is a phenomenon that is unique to the gaze cue or one that generalizes to socio-biological cues, including the pointing finger. If pointing cues do not exhibit a preference for objects, and thereby shows the same tendency as that found for gaze cues, then results support the account that the formation of object representations is mediated by a common foundation that processes socio-biological cues. The candidate responsible for the processing includes mechanisms that are related closely to socio-biological cues such as the theory of mind or joint attention, for instance. In contrast, when pointing cues generate object-based attention, as opposed to gaze, properties common to socio-biological cues do not affect object-based attention, but rather suggest differences in processing at lower perceptual dimensions (differences in the developmental period of corresponding mechanisms or in the priority in terms of attention to faces or gaze).

Therefore, to ascertain whether determination of the mode of attention (location-based or object-based) involves complex processes such as understanding the intentions of others, which is recruited by sociobiological cues, we followed the following two important points of inquiry. First, we tested whether a difference exists in the mode of attentional selection between the arrow and the gaze cue ([35]). To elucidate this point, we made some improvements based upon a report of an earlier study ([35]), specifically to the paradigm based on a typical object-based attention study ([19], [20]; [33]). We conducted two experiments separately in a manner closely resembling the work of [35] to ascertain whether an uninformative arrow, a schematic gaze (Additional experiment A in Appendix), a gaze stimulus of a real face photograph, and a pointing finger cue (Additional experiment B in Appendix) each generate an object-based attention effect. Although we attempted to reproduce earlier findings, at least an arrow cue recruits an object-based effect sufficiently, all the central cues failed to produce object-based facilitation. [33] demonstrated that central symbolic cues, which can provoke a voluntary (endogenous) shift of attention such as an arrow, produced no object-based allocation effect, unlike peripheral cues that cause exogenous attention. [19] reported that this difference is attributable to differences in the initial focus of attention to pre-cue: peripheral cues focus participants' attention on the object itself, whereas symbolic cues specifically examine the central cues themselves. Based on this attentional focusing hypothesis, they confirmed that even endogenous cues produce the object-based attentional effect under conditions in which observers are encouraged to extend their attention to the whole object by the instruction. In other words, bottom-up factors such as the type of attentional cue and top-down factors attributable to the difference in strategy for the task both presumably mediated the object-based attention. Based on findings obtained from those studies, the initial focus of attention on cues can explain that gaze cues do not cause object-based attention. Therefore, we instructed participants to extend their attention to the entire object. If this manipulation of instruction does not promote the object-based attentional effect of gaze cues, then the deficiency of object-based attention by gaze cues might be independent of strategic factors. In addition to maximizing the object-based attentional effect, we approached the methodology employed for earlier typical object-based attentional studies (e.g., [16]). Specifically, we placed the objects vertically or horizontally. Then we assigned higher validity to the spatially valid trials in which the target appeared at the cued location compared to the two spatially invalid trials (same-object trials and different object trials) that were evenly distributed. Under these conditions, we investigated whether arrows and gaze cues produce an object-based attentional effect.

Second, we investigated whether the pointing finger cue also showed the same tendency as that of the gaze cue. Therefore, we used pointing fingers as the attentional cues in addition to the arrow cues and the gaze cues. If the object-based attentional effect is not obtained when the gaze cue is presented, then our primary interest is to ascertain whether pointing cues also promote the object-based attentional effect. Whether the pointing finger cue prioritizes attentional selection towards an object reveals whether a mechanism of social attention is involved in the mode of attentional selection. If the pointing finger indicates only a specific spatial location, similarly to a gaze cue, then the results are expected to be compatible with the idea that the theory of mind mechanisms mediates this phenomenon ([35]).

All 32 university student participants (19 women and 13 men; mean age = 21.9 years, SD = 5.64 years) had normal or corrected to normal vision. All were unaware of the experiment purpose. The experiment was approved by the ethical committee of the Department of Chiba University. Written informed consent was obtained from all participants.

Initially, based on an earlier study of object-based attention ([20]), we assumed that 20 participants were sufficient and then recruited 22 participants (14 women and 8 men; mean age = 21.0 years, SD = 6. 22 years). Because no power analysis was performed, we calculated the minimum effect size (partial eta squared; ηp2) for a 3 × 3 interaction (cue type and cue-target validity), using MorePower 6.0.4 ([6]). In addition, based on the assumed large effect size, ηp2 = 0.14 ([9]), we calculated the statistical power (1−β) of the two-way interaction for the 22 participants. Setting α = 0.05, power = 0.95, the minimum effect size was ηp2 = 0.189. The actual effect size was ηp2 = 0.208 for a significant 3 × 3 interaction with 22 participants [F (4, 84) = 5.52, p <.001]. However, the statistical power (1−β) was slightly smaller, 0.83. Therefore, we again calculated the required sample size, which indicated that 30 participants provided 0.95 power using ηp2 = 0.14, with α = 0.05. Based on this analysis, we added 10 new participants. Therefore, a total of 32 participants took part in this experiment. The effect size was assumed to be large (ηp2 = 0.14 transformed from f = 0.40, [6]; [9]) from an earlier study examining the object-based attention effect (ηp2 = 0.16; [20]; ηp2 = 0.25 calculated from F (1, 23) = 7.84 interaction of Experiment 1; [35]).

Because we performed our analysis with data collected in two stages, we regarded it as appropriate to conduct an interim analysis and corrected for p-values ([29]). The interim analysis is actually an a posteriori analysis, but it assesses robustness across two stages of data collection. Although such sequential procedures are fundamentally designed to terminate sampling before collecting the planned sample size, the effect-size estimates can be more precise when the sequential procedure is terminated at a later stage (i.e., when the planned sample size has been collected) compared to non-sequential procedures ([51]). The first stage involved 22 samples. The second stage involved 10 new samples. When we hypothesized that at least arrow cue produces the object-based attention (RT difference between same-object and different-object condition), the bounds of the t value for each stage (i.e., before and after adding samples) were ± 2.12, ± 2.17. The respective adjusted p-values were.034,.016. At both stages, the object-based attention effect by the arrow cue was significant at these levels (t (21) = 3.27, adjusted p <.01 at the first stage), supporting the overall significance of our results. Subsequent analyses were based on the adjusted p-values.

Stimuli were presented on a 17-inch color VGA monitor (E173FPc, 1280 × 1024 pixel screen resolution, 60 Hz refresh rate; Dell Inc.). A PC (Dell Inc.) running PsychoPy (ver. 1.82.01; [38]) controlled the presentation of the stimuli and data collection. Responses were given using a standard keyboard. A chin rest was used to fix the observation distance.

Stimuli consisted of object stimuli, pre-cue stimuli, and target stimuli. Two white rectangles with a black outline subtending 10.5 deg × 3 deg of visual angle, the same properties as those used in [35], were presented horizontally or vertically as object stimuli in all trials. In the arrow cue condition, a linear stimulus with an arrowhead (1.8 deg × 1 deg) was presented. In the gaze cue condition, a schematic face (3.5 deg × 3.5 deg) used by [17] was presented. The lengths of the two ends of the eye contour in the gaze cue were identical to the length of the arrow cue and the finger cue. In the finger cue condition, a grayscale of a hand gesture with the index finger extended (1.8 deg × 1 deg; the finger of a male or female) was presented. Each type of pre-cue (arrow, gaze, and finger) was presented on the center of the display indicating upper right, upper left, lower right, or lower left (Figure 1). Target stimuli of "O" or "X" (0.9 deg × 0.9 deg) were presented at a position approximately 5 degrees from the center of the display. Before the pre-cue was presented, the object stimuli and a fixation point (0.9 deg × 0.9 deg; the shape of "+") were presented at the center of the display. All stimuli were presented on a white background (luminance = 155.0 cd/m2).

Graph: Figure 1. Cue types used in the main experiment. (a) Arrow cue, (b) gaze cue, and (c, d) finger cues. See the text for the accurate sizes for these cues.

Participants were seated approximately 57 cm from the monitor to perform the task in a dimly lit and soundproof room. Participants' observation distance and head movement were fixated by the chin rest. For each trial, participants started the trial by pressing the space key in response to a screen display prompt of "Ready?" in the center of the display. When the space key was pressed, two objects arranged horizontally or vertically with the fixation point were presented for 1000 ms. Then a pre-cue indicating one of the four directions was presented for 400 ms. After the pre-cue, targets of "O" or "X" were presented at either end of the two objects: one of the four positions on the screen. Participants responded by pressing the "C" key or the "M" key on the keyboard according to the target. The target was presented for 2000 ms. Participants were told to orient their attention voluntarily in the direction of the indicated pre-cue and to divide their attention to both objects, responding to the target as quickly and as accurately as possible. We confirmed that all participants understood the instructions before and after the experiment trials. The condition of pre-cue type (arrow, gaze, and finger) was divided by blocks (Figure 2). Deviation was found in the predictability of pre-cues indicating the target location: 66.7% of all trials were valid trials. The remaining trials were divided equally into two invalid trials: same-object trials, for which the target was presented at the opposite location of the same object that the pre-cue indicated, and different-object trials, for which the target appeared at the near end of the opposite object.

Graph: Figure 2. Illustration of experimental procedures. (a) Presentation rate of the target for each condition and (b) schematic illustration of a trial sequence in the arrow cue condition (same-object trial).

Participants performed 3 blocks of 192 experimental trials each. Each was preceded by 32 practice trials. Therefore, there were 672 trials in all. Participants were free to take a rest during the trial. The order of blocks executed and the positions of reaction keys were counterbalanced. The conditions of each block (pre-cue direction, target letters, and object arrangement) were selected randomly for each block.

The experiment consisted of two factors of repeated measure design: cue type, and cue-target validity. The cue types were the arrow cue, the gaze cue, and the finger cue. The cue-target validities were valid location trials, same-object trials, and different-object trials. We applied analysis of variance (ANOVA) for mean correct RT and error rates. Multiple comparisons using modified sequentially rejective Bonferroni method (Shaffer, 1986) was conducted when necessary.

Mean correct RT and error rates were calculated for each participant for each cue type and cue-target validity (Table 1). Incorrect responses (2.76%), and responses given faster than 100 ms, slower than 3,000 ms, or 3 SD away from the mean RT were excluded from the RT analyses. The total trials excluded were 4.31% of the trials (including incorrect responses).

Graph

Table 1. Mean reaction times (RT in ms) and error rates as functions of the cue type and cue-target validity.

A cue type (3) × cue-target validity (3) repeated measures of ANOVA was applied to participants' error rates using factors of cue type and cue-target validity. The main effect of cue-target validity was found to be significant [F (2, 62) = 9.73, p <.01, ηp2 = 0.239]. Multiple comparison revealed that the rate of incorrect response for valid location trials (M = 2.25%) was smaller than for same-object trials (M = 3.26%) or for different-object trials (M = 4.33%) [valid-location trials and same-object trials—t (31) = 2.96, adjusted p <.01; valid-location trials and different-object trials—t (31) = 3.58, adjusted p <.01]. However, no significant difference was found between the rates of incorrect response in same-object trials and different-object trials [t (31) = 2.31, adjusted p =.027]. The main effects of cue type [F (2, 62) = 0.09, p =.910, ηp2 = 0.003] and two-way interaction [F (4, 124) = 0.80, p =.524, ηp2 = 0.025] were not significant. These results of ANOVA suggest that the participants' reactions are more accurate for valid location trials than for two spatially invalid trials. Moreover, the results imply that no consistent speed–accuracy tradeoff exists.

A cue type (3) × cue-target validity (3) repeated measures of ANOVA was applied to participants' RT using factors of cue type and cue-target validity (Figure 3). The main effect of cue-target validity was found to be significant [F (2, 62) = 68.21, p <.001, ηp2 = 0.687], revealing that RT was shorter in the valid location trials (M = 419.0 ms) than in the same-object trials (M = 465.2 ms; t (31) = 8.08, adjusted p <.001) and different-object trials (M = 471.2 ms; t (31) = 8.80 adjusted p <.001). In addition, RT in the same-object trials was significantly shorter than in the different-object trials (t (31) = 3.01, adjusted p <.01). Moreover, the interaction was significant [F (4, 124) = 5.64, p <.001, ηp2 = 0.154]. The main effect of cue type was not significant [F (2, 62) = 1.96, p =.149, ηp2 = 0.060].

Graph: Figure 3. Mean reaction times (RT) for the respective cue type conditions as a function of cue-target validity. Error bars represent the within-subject standard errors of the mean.

Additional analysis for the interaction revealed the simple main effect of cue type in valid-location trials [F (2, 62) = 9.07, p <.001, ηp2 = 0.226] as significant. Also, RT in the gaze cue trials was longer than in the arrow cue trials (t (31) = 3.14, adjusted p <.01) or in the finger cue trials (t (31) = 3.81, adjusted p <.01). A simple main effect of cue type in the same-object trials [F (2, 62) = 1.88, p = 161, ηp2 = 0.057], and in the different-object trials was not significant [F (2, 62) = 0.54, p =.582, ηp2 = 0.017]. The simple main effect of cue-target validity was found to be significant in all cue type conditions [arrow cue—F (2, 62) = 73.38, p <.001, ηp2 = 703; gaze cue—F (2, 62) = 40.82, p <.001, ηp2 = 0.568; finger cue—F (2, 62) = 47.90, p <.001, ηp2 = 0.607]. Multiple comparison revealed that RT in the valid-location trials was significantly shorter than in the same-object trials (arrow cue—t (31) = 8.57, adjusted p <.001; gaze cue—t (31) = 6.74, adjusted p <.001: finger cue—t (31) = 7.02, adjusted p <.001), or in the different-object trials (arrow cue—t (31) = 9.81, adjusted p <.001; gaze cue—t (31) = 7.01 adjusted p <.001; finger cue—t (31) = 7.41, adjusted p <.001) for cues of all types. The object-based effect (RT difference between the same-object condition and different-object condition) was significant in the arrow cue condition (t (31) = 2.63, adjusted p =.013) and in the finger cue condition (t (31) = 3.04, adjusted p <.01). By contrast, in the gaze cue condition, no difference was found between RT in the same-object trials and the different-object trials (t (31) = 0.78, adjusted p =.44).

We also conducted Bayesian analysis to assess the null hypothesis that there was no object-based effect (RT difference between the same-object and different-object trials) for each cue type. Using computer software (JASP; [28]), we found that RT data in the gaze cue trials is about four times more likely under the null hypothesis than under the alternative hypothesis (BF01 = 3.99, prior = 0.5). In contrast, the same Bayesian analysis for the arrow cue and finger cue trials supported the alternative hypothesis that the data are more likely to be significant under the alternative hypothesis than under the null hypothesis (arrow cue—BF10 = 3.52, prior = 0.5; finger cue—BF10 = 8.34, prior = 0.5).

Consistent with earlier reported findings ([35]), results of the present study indicated the object-based effect occurred under the arrow-cue condition but not under the gaze-cue condition. Results suggest that the gaze cue was unaffected by top-down factors because of manipulation of the strategy for the task. Remarkably, a sufficient object-based effect was found even in the finger cue condition. These results imply that common characteristics of socio-biological cues such as eye gaze and pointing finger are unlikely to mediate the mode of visual attention (location-based or object-based).

A salient result of this study was that, even under conditions in which central cues were likely to elicit object-based attention using a top-down strategy that encouraged extension of their attentional focus ([19]), gaze cues did not produce the object-based attention effect. We also obtained data for a condition in which observers were not instructed about the attentional strategy (Additional experiment A in the Appendix), but the results show that neither gaze cues nor arrow cues elicited any object-based attention. [35] reported that non-predictive arrow cues generate sufficient object-based attention. In Additional experiment A, addition of an object arrangement (horizontal and vertical, as well as diagonal arrangement) might have reduced object-based attention, but the results of Additional experiment B suggest that object placement had no effect. Consequently, the modifications of the procedure used by [35] are unlikely to be related to the deficit of object-based attention in the arrow cue.

One possibility is that differences in participants' experience, knowledge, and habitual use of language might account for the differences between the results of the two Additional experiments reported in the Appendix and in an earlier study ([35]). The incorrect rates of our two Additional experiments were 2.53% and 2.18%, respectively, which were slightly lower than those of the three experiments reported by [35] (Experiment 1, 7.0%; Experiment 2, 3.7%; and Experiment 3, 5.0%). In addition, the overall means of RTs in our two Additional experiments were 432.9 and 448.2 ms, which were shorter than reported by [35] (Experiment 1, 493.9 ms; Experiment 2, 554.8 ms; and Experiment 3, 546.7 ms). Although these are not purely comparable, if the participants in our experiment performed the similar task more rapidly and more accurately than the participants in the earlier study, then the object-based attention might consequently not have been detected as a sufficient RT difference. Such higher task performance of the participants in this study might be attributable to the fact that the Japanese participants showed less interference from language and experience, which is demonstrated by the correspondence between the target "O" or "X" and the response key "C" or "M" ("Correct" and "Miss"). It remains unclear whether this interference was greater in the earlier study ([35]), but it might affect at least the speed and accuracy of the task. In any case, we have not yet identified what can account for in this discrepancy between the results of the present study and earlier reported findings ([35]). Consequently, further investigation must be conducted to examine this issue.

Results of earlier research have indicated that symbolic central cues such as arrows usually do not induce object-based attention and that attentional effects emerge through the instruction of attentional extension ([19]; [33]; Appendix). Consequently, the arrow cue results obtained from the present study are consistent with earlier findings obtained from a typical study of object-based attention. By contrast, gaze cues lacked the object-based attention effect despite the instruction of task strategy. These results suggest that gaze cues have features that suppress both the extension of attention and the activation of object representations. According to the attentional focus hypothesis ([19], [20]), the manner in which attention is allocated (location or object) depends on top-down factors such as task strategy, as well as the visual saliency of the object stimulus ([20]), which affects the initial spatial distribution of attention. In fact, [20] reported that object-based attention does not occur under conditions of low stimulus discriminability. Therefore, differences in the discriminability of gaze cues and arrow cues might affect the occurrence of object-based attention. However, as described by [20], an attentional cue with low discriminability did not reduce location-based attention but mediated object-based attention. The present study demonstrated that the RT values found from valid location trials for the gaze cue condition were longer than those for the other two cues. Results of several studies have indicated that when gaze versus arrows is used to measure social attention, RTs are longer than arrows for both gaze cues ([25]) and gaze targets ([36]), compared to arrows. Reportedly, face stimuli capture attention longer than other object stimuli do ([41]; [48]), which can be expected as one factor affecting exogenous attention ([30]; [31]). Given those findings, it seems possible that the complexity or uniqueness of the processing of gaze and face cues, rather than visual discriminability, influences the initial distribution of attention. The fact that the human face is attractive in the observer's visual field simultaneously implies the possibility that a system that is sensitive to where others are looking (Eye Direction Detector; Baron-Cohen, 1995b) is involved. Earlier reports of some studies have described that a direct gaze rather than a deviated gaze better captures an observer's attention ([45], [44]; [47]). Although direct gaze is not part of the task examined for this study, it is not surprising that observers devote more attentional resources to discerning whether others are looking at them or at some other spatial location. The present study showed no means of manipulating the visual discriminability constant between the gaze cue and the other cues, but the sizes of cues used for arrows and gaze were identical (edge-to-edge of the eye contour). At least, unlike the other cues, the gaze cue narrowed the initial distribution of attention without being affected by top-down strategic factors.

To ascertain whether the gaze cue simply has lower discriminability than the arrow cue, it will be interesting to compare results obtained for typically developed individuals with groups with those from people with autism spectrum disorder (ASD). Adults and children with ASD often show different patterns of gaze cueing of attention than those associated with typically developed people, but differences are not attributable to abnormal gaze direction perception. Research conducted with the ASD group ([52]) has indicated that individuals with autism showed sufficient attentional orientation to the gazed-at direction of a robot. Another study of children with ASD ([43]) revealed evidence that children with autism equate gaze cues with arrow cues: typically developed control children showed a difference of attentional shift in the counter-predictive gaze and the arrow, but autistic children showed the same cueing effect for cues of both types. Therefore, if individuals with ASD show object-based attention by the gaze cues, similar to an arrow, then the narrowing of attentional focus by the gaze cue, at least for typically developing individuals, will suggest that the basis of social cognitive mechanisms unique to the gaze is involved rather than the visual discriminability stimuli.

Other results of this study were that the pointing stimuli, as well as symbolic arrows, produced the object-based attention effect. This result indicates that the cognitive processes involved in orienting attention in terms of eye gaze differ from those involved in orienting attention in terms of a pointing finger. In other words, a pointing finger functions as a symbolic cue acquired through daily experience and learning such as an arrow, rather than functioning as a socio-biologically cue, such as a gaze cue. [1], who examined the influence of pointing with a finger on visual attention, demonstrated that reflexive orientation of attention triggered by pointing with the index finger was found, using hand gestures as attentional cues. They argue the possibility that the rapid visual processing of socio-biologically relevant stimuli, that is, pointing with the index finger, might be part of a hard-wired, automatic response that has traditionally been regarded as limited to low-level processing. It might form part of a primitive orienting reflex. In addition, because of the possibility that it is not exclusive from another, they discussed the possibility of intensive over-learning established in early childhood. The experiment conducted for this study used a double rectangle paradigm ([16]), which indicated that rapid visual processing associated with a pointing finger reflects intensive over-learning rather than socio-biological relevance, as in the case of eye gaze. Indeed, interest in the gaze direction was observed at approximately 3 months of age ([27]). Pointing is acquired in infants at approximately 12 months ([32]). Consequently, these results might reflect the earlier establishment of gaze direction as a cue than the establishment of pointing with a finger. From a perspective of cross-cultural study, pointing with hand gestures is not necessarily universal, with at least some communities showing a preference for non-hand pointing, such as the nose or head orientation ([10]). However, as indicated by [35], it remains to be investigated how the orientation of attention to a specific location rather than an object triggered by gaze reflects social processing, or not. For instance, the finger-pointing cue used for our experiment was in the range of the wrist to the fingertip. In other words, it remains questionable whether the finger cue communicated the intention of orienting attention to a specific location. If an agency, specifically the intentions of others, is important, then eye gaze itself need not be available as an attentional cue ([37]). Additional studies must be conducted to examine whether the mental state of others mediates this phenomenon. In any case, the deficiency of object-based attention was presumably not mediated, at least by the system common to socio-biologically related cues such as gaze direction and pointing gestures.

This study has several limitations. First, the results obtained from this study were obtained using a typical object-based attention paradigm rather than the experiment-oriented approach used by [35]. For example, to be closer to typical methods for object-based attention (e.g., [16]), the object arrangement was changed (from diagonal to horizontal and vertical). Also, the conditions for the cue and the target were changed. The fact that object-based attention was not observed in the preliminary experiments (Appendix) must be investigated further. Second, we were unable to confirm that participants in this experiment actually followed the instructions of the strategy to spread their attentional focus. Therefore, further investigation must be undertaken to assess whether the gaze cue actually narrows the attentional focus. However, it would have been interesting if object-based attention triggered by the gaze cue was not observed even in the modified conditions of the present study, which is a typical paradigm in the context of object-based attention. The phenomena were sufficiently observed at least with the arrow cues and finger cues.

In summary, this study has produced two conclusions: First, the deficiency of object-based attention by the gaze cue is unaffected by top-down factors such as the manipulation of attentional strategies, especially in a typical paradigm of object-based attention (e.g., [16]; [19]). This finding suggests that the gaze cue itself prevents activation of object representations by strategies of extending the attentional focus and consequently modulating the initial allocation of attention. Second, this phenomenon is not generalizable to socio-biologically relevant cues such as pointing with a finger. The results of this study demonstrated that the process responsible for the pointing gesture and the gaze direction to understand another person's intention or mental state, which play important roles in early development, do not mediate object-based attention. In other words, the results of this study indicate that the unique properties of gaze might modulate the mode of attentional orienting, rather than the involvement of higher cognitive processing such as the theory of mind mechanisms. In future research, using this paradigm with pointing and gaze as an attentional cue, it will be interesting to ascertain whether the mode of attention in autistic individuals differs from the mode of a group of individuals with typical development. Although many earlier studies have specifically examined quantitative differences in attentional shifts triggered by social and non-social cues, whether such differences from typically developing people also lead to qualitative differences in attention-related spatial characteristics is an important point that must be assessed.

Graph: Supplemental material, sj-docx-1-pec-10.1177_03010066231168570 for Deficiency of object-based attention specific to the gaze cue is independent of top-down attentional strategies by Hirokazu Eito and Akio Wakabayashi in Perception

The authors wish to thank Makoto Ichikawa and two anonymous reviewers for their helpful comments on earlier versions of this manuscript.