In our interactions with the world, social cues such as gaze help us focus on what others are interested in and infer others’ intentions and actions (Kwon et al., 2016; Wahl et al., 2013). This gaze-triggered attention is an important cornerstone of high-level cognitive processing, such as working memory, and plays a crucial role in human communication (Gobel & Giesbrecht, 2020; Gregory & Jackson, 2016, 2019; Nie et al., 2018). Researchers usually used the gaze-cueing paradigm modified from Posner’s cueing paradigm (Posner & Cohen, 1984) to study this gaze-triggered attention. Typically, gaze cues are presented in the center of screen, and a faster attentional orienting was found to a gazed-at than to a non-gazed-at target—that is, gaze-cueing effect. A series of studies using the gaze-cueing paradigm have found a significant gaze-cueing effect, even when the gaze was uninformative (50% cue validity) or counterpredictive of the target location (Downing et al., 2004; Gregory & Jackson, 2016; Tipples, 2008), which suggested the reflexive property of gaze. In addition to gaze, uninformative nonsocial cues (e.g., arrows) can also trigger attentional orienting to a cued target (Bonmassar et al., 2019; Hommel et al., 2001). However, numerous studies have found that the gaze distinguishes from nonsocial cues in attention orienting. Some studies found that the cueing effect from gaze than from arrow cues was stronger (Friesen et al., 2004; Ristic et al., 2007), lasted longer, and was more stable in the time course (Yoxon et al., 2019). For example, Liu et al. (2021) explored the cueing effect triggered by uninformative gaze and arrow cues across consecutive trials and found a temporally stable gaze-cueing effect throughout the task, and a gradually decreased cueing effect triggered by arrows. What is more, some neuroimaging also demonstrated different neural activations in response to uninformative gaze and arrows (Joseph et al., 2014), suggesting that gaze and arrows have essential differences. In short, the contrast between gaze and nonsocial cues suggested that gaze may be a special cue and distinguish from nonsocial cues due to its sociality.

In the literature, the underlying process of gaze-triggered attention is still the focus of debate. Some researchers proposed that gaze-triggered attention is a kind of reflexive attention, which is not affected by consciousness and top-down cognitive processing (Bonmassar et al., 2019; Feng & Zhang, 2014; Nie et al., 2018; Ristic et al., 2007; Stevens et al., 2008). Some researchers found that uninformative or counterpredictive gaze cues can still trigger a cueing effect, and the gaze-triggered attention was indistinguishable from exogenous attention that is reflexively induced by overt cues in the time course (Liu et al., 2021). This findings suggested that gaze triggered a reflexive and automatic manner. In addition to the reflective attention orientation, there is evidence that gaze cues can also trigger voluntary attention, such as orienting to expected location regardless of the gaze-cued location (Chanon & Hopfinger, 2011; Friesen et al., 2004). What is more, the gaze-cueing effect can be influenced by social factors, such as self-esteem, personal goals, and emotional expressions, while arrows cannot (Dalmaso et al., 2020; Ishikawa et al., 2021), suggesting that gaze-triggered attention can be modulated by top-down factors.

Autistic trait was found to be an important factor in affecting social attention.  Many researchers suggested that there is a continuum of autistic traits distributed within the general population. Individuals with non-clinical but relatively high autistic traits are similar to individuals with autism spectrum disorder (ASD) in some genetic aspects and behavioral performance (Jones et al., 2013; von dem Hagen & Bright, 2017). Numerous studies have found that individuals with high autistic traits also show abnormal responses to gaze. They showed shorter fixation duration on the experimenter’s eyes (von dem Hagen & Bright, 2017) and showed shorter and less frequent saccades, as well as reduced visual exploration, than individuals with low autistic traits during real-time interactions with the surrounding environment (Vabalas & Freeth, 2016). The gaze-cueing effect was found to be weaker (de Araújo et al., 2020; Tajmirriyahi et al., 2016) and more transient under cross-modal conditions in individuals with high autistic traits (Zhao et al., 2015), which suggested the changes of gaze-triggered attention in individuals with high autistic trait. Researchers have tried to explore whether the attention changes in high autistic trait are specific to the social domain. Arrows are often used as a nonsocial stimulus in contrast to the gaze, and the speciality of gaze compared with arrows in individuals with low autistic traits was found to disappear in individuals with high autistic traits (Lin et al., 2020; Wang et al., 2019). For example, Lin et al. (2020) found that individuals with high autistic traits showed insufficient attention orientation triggered by gaze, but not by arrows. These findings suggested that the changes of social attention in individuals with high autistic traits seems to be specific to the social domain.

However, gaze as a special cue, not only contains the social significance but also involves the low levels of iris or pupil movement, which can also trigger the cueing effect (Nie et al., 2018; Zhang et al., 2019). In previous studies of the comparative studies of gaze and nonsocial cues, it is worth noting that most studies about autistic traits used the arrows as nonsocial cues (Freeth et al., 2010; Lin et al., 2020), which ignore the effect of iris or pupil movement on cueing effect. In addition, some studies did not use nonsocial cues for comparison (Cooney et al., 2017), and some studies only used cartoon faces rather than real faces as social cues, which reduced the ecological validity of gaze (Wei et al., 2019; Zhao et al., 2015). Therefore, it is still necessary to further study whether the changes of gaze-triggered attention are domain general or social specific. We tried to avoid the above methodological problems in the current study and mainly manipulated the sociality of gaze through the comparison of dynamic gaze and dot cues with similar motion characteristics as gaze cues to further explore the effect of autistic traits on gaze-triggered attention in neurotypical individuals.

In addition, attention is not a static process but a dynamic change process. The dynamic changes of attention triggered by uninformative cues could not be directly measured with RT or inferred from RT patterns; for example, Liu et al. (2021) used a sliding window method, which calculates the running average of the cueing effect across consecutive trials to indirectly explore the time course of gaze-triggered attention. They found gaze-cueing effect was relatively stable compared with the cueing effect triggered by arrows. In contrast, eye-tracking technology is a good way to directly observe the attention triggered by gaze and nonbiological motion cues. Previous studies on autistic traits that revealed potential differences between gaze and nonsocial cues have focused on the RT to target (i.e., cueing effect), not considering the dynamic deployment of attention over time. Therefore, in our study, we used the eye-tracking technique to explore how such effects may manifest over time and whether autistic traits affect the dynamic changes of attention triggered by gaze and nonbiological motion cues.

In this study, we set two kinds of cues—uninformative gaze cues and dot cues. The gaze cues consist of only the eyes from pictures of real faces, to reduce the influence of other parts of the face and as the dynamic social cues. The dot cues consisted of a black line segment with two black dots on it and as the nonbiological motion cues. To better control the motion of cues, the line segment length of dot cues was consistent with the width of the gaze. Dot size and position of dot cues remained consistent with the pupil of the gaze cue. Due to the complexity of daily life, complex tasks compared with simple cueing paradigm have greater ecological validity. Therefore, we used the cueing paradigm within the visual matching task and the eye-tracking technique to investigate whether the changes of gaze-triggered attention in individuals with high autistic traits was due to the sociality of the gaze and how autistic traits affected gaze-triggered attention over time. We also explored whether the changes of gaze-triggered attention in individuals with high autistic traits would affect their subsequent visual processing. We hypothesized that individuals with high autistic traits would show changes in gaze-triggered attention, which is specific to social cues. In individuals with low autistic traits, the benefit from gaze is significantly greater than that from dot, but in individuals with high autistic traits, the difference between these two cues is not significant. According to the argument of reflective or volitional attention about gaze-triggered attention, we predicted that the eye movement results may show a relatively later effect of autistic traits on gaze-triggered attention. For the visual processing to the target, we hypothesized that individuals with high autistic traits would show a shorter fixation proportion to the cued target than would individuals with low autistic traits.

Methods

Participants

The AQ questionnaire (Zhang et al., 2016) was used to measure participants’ AQ scores. Higher scores indicate higher levels of autistic traits. A total of 220 AQ questionnaires were randomly distributed among undergraduates and postgraduates, and 206 valid questionnaires were recovered. The AQ score distribution in our study is shown in Fig. 1. The participants who scored in the top 20% and bottom 20% were selected. Finally, 46 participants were divided into the high AQ group (AQ score ≥124) and low AQ group (AQ score ≤108) with reference to previous studies (Burnett & Jellema, 2013; English et al., 2017; Whyte & Scherf, 2018). There was a significant difference between the two groups in AQ scores, but not in age and gender compositions (the participants’ characteristics were shown in Table 1). Informed consent was obtained according to procedures approved by the Anhui Normal University ethics committee, and written informed consent was obtained from all the participants.

Fig. 1
figure 1

The AQ score distribution in the current study

Full size image
Table 1 Participant characteristics of age, gender, and AQ score in two groups
Full size table

Experimental materials and instruments

AQ questionnaire

The AQ questionnaire was developed by Baron-Cohen et al. (2001). Zhang et al. (2016) translated and revised the Chinese version. The questionnaire includes five dimensions: social skills, attention shifting, attention to detail, communication, and imagination, with 50 items in total. Each item has four options: “definitely agree,” “slightly agree,” “slightly disagree,” and “definitely disagree,” which are scored using a 4-point Likert scale. The internal consistency coefficient of the revised Chinese version was 0.81, the test–retest reliability was 0.89. In this study, the Cronbach’s alpha was 0.70.

Materials and apparatus

With reference to the experiment by Nie et al. (2018), six images of real faces were processed by FaceGen Modeler 3.4 and Photoshop CS4 to form gaze cues. The nonsocial dot cues consisted of a black line segment with two black dots on it. The line segment length was consistent with the width of the gaze. Dot size, and position remained consistent with the pupil of the gaze cue. The memory and probe items were selected from six irregular polygons (subtending 1.5° × 1.5° each; Alvarez & Cavanagh, 2004). Two irregular polygons were randomly presented on the memory interface at a time, located to the left and right of the screen, both 3.5° from the center. The probe could not be the same as the mirror image of the memory items.

The stimuli were presented on a 19-inch monitor (resolution: 1,024 × 768, refresh rate: 75 Hz) through E-Prime2.0. The Hi-Speed eye tracker manufactured by SensoMotoric Instrument (Germany) was used to record the eye movement trajectories of the participants’ dominant eye. The sampling frequency was 1250 Hz.

Design

The experiment was a 2 × 2 × 2 mixed design, with autistic traits (High AQ vs. Low AQ) as a between-participant variable, and cue sociality (gaze cues vs. dot cues) and cue validity (valid vs. invalid) as within-participant variables.

Procedure

The participants sat approximately 72 cm away from the screen. Experimental sessions started with 9-point eye tracker calibration. Participants received instructions on the computer screen and was told that the cue validity was 50%. After the practice, they proceeded to the formal experiment. The experimental procedure is shown in Fig. 2. First, a fixation point was presented for 500ms. Then, two irregular polygons were presented simultaneously on the screen for 250 ms. After a blank screen interval of 500 ms, the eyes or dots were presented in the center of the screen for 500 ms. Then the eyes or dots shifted 0.8° to the left or right and presented for 500ms as gaze or dot cues. After a 500 ms interval, a polygon appeared on the left or right side of the screen for 3,000 ms as a probe. The participants were asked to judge whether the probe was the same as the previous item presented at the same location as accurately and quickly as possible. The formal experiment consisted of 192 trials: two blocks (gaze cues and dot cues) and 96 trials in each block. The order of the two blocks was counterbalanced across the participants.

Fig. 2
figure 2

Experimental paradigm

Full size image

Data collection and analysis

Preliminary analysis

One participant was discarded due to high error rate (55.80%, more than 50%). Incorrect responses (19.6%) and responses exceeding three standard deviations of the individual’s mean reaction time (1.7%) were excluded. For the eye-tracking analysis, four participants with a tracking ratio of less than 60% and one participant with no data in the regions of interest (ROI) due to the excessive amplitude of head movement were excluded. Finally, 40 participants were included in the eye-tracking analysis (21 in the High AQ and 19 in the Low AQ). All data were analyzed using SPSS 19.0, Greenhouse–Geisser correction was used to adjust the degrees of freedom for the ANOVAs that did not pass the test of sphericity.

Eye-tracking data indicators

With reference to previous studies (Falck-Ytter et al., 2012; Navab et al., 2011; Zhang et al., 2016), and the locations where the memory items and probes appeared, we set two equally sized regions of interest (ROI) which occupied 6.50% of the screen, and we also select five indicators of different interfaces: (1) Fixation count on cue interface, which is the average number of fixations to each ROI under each condition. It is the indicator of fixation preference. (2) The fixation count on postcue. (3) The percentage of first gaze shifts on probe was calculated by dividing the number of trials in which the participant’s first saccade was made to the cued/uncued ROI by the total number of usable trials. (4) Latency of first looks to the target is the time it takes for participants to first fixate the target, which is the indicator of attentional shift, the shorter latency indicates a faster attentional shift. (5) Fixation proportion to the target, representing the extraction and processing of the target.

Results

The results of RT and accuracy

A 2 (autistic traits: High AQ vs. Low AQ) × 2 (cue sociality: gaze cues vs. dot cues) × 2 (cue validity: valid vs. invalid) repeated-measures analysis of variance (ANOVA) was performed for RT and accuracy. Figure 3 presents the mean RT and accuracy in the High AQ and Low AQ.

Fig. 3
figure 3

RT (left) and accuracy (right) of the high and low AQ group by cue sociality (gaze cues/dot cues) and cue validity (valid/invalid). High-dot = the High AQ in the dot cues; High-gaze = the High AQ in the gaze cues; Low-dot = the Low AQ in the dot cues; Low-gaze = the Low AQ in the gaze cues. Error bars represent the 95% confidence interval and the full dots represent individual observer data. *p < .05, **p < .01, ***p < .001.The full dots represent individual observer data

Full size image

For RT, the results indicated that the main effect of cue validity, F(1, 43) = 13.70, p = .001, ηp2 = 0.24, 95% CI [−70.66, −20.82], and the three-way interaction was significant, F(1, 43) = 6.26, p = .02, ηp2 = 0.13. The simple effects test found that the Low AQ showed significantly lower RT for valid gaze cues compared with invalid gaze cues (p < .001, 95% CI [−133.47, −51.51]), and a marginally significant between valid and invalid dot cues (p = .05). The differences were not significant in all other conditions (p > .05).

Using the RT difference between invalid and valid cue (invalid RT − valid RT) as the dependent variable to further analyse the differences between the benefits from gaze and dot cues. A 2 (autistic traits: High AQ vs. Low AQ) × 2 (cue sociality: gaze vs. dot) repeated-measures ANOVA was performed. The results showed a signifcant interaction between autistic traits and cue sociality, F(1, 43) = 6.26, p = .02, ηp2 = 0.13, the benefits from gaze in the Low AQ was larger than that in the High AQ (p = .02, 95% CI [13.47, 128.11]), and was larger than that from dot cues (p = .004, 95% CI [17.97, 89.75]). The differences were not significant in all other conditions (p > .05). The accuracy did not show any significant main effects or interactions.

Eye-tracking results

Eye-tracking indicators for the cue interface

For the fixation count, a 2 (autistic traits: High AQ vs. Low AQ) × 2 (cue sociality: gaze vs. dot) × 2 (cue congruence: cued vs. uncued ROI) repeated measures ANOVA was performed. The results showed that the main effect of autistic traits, F(1, 38) = 4.02, p = .05, ηp2 = 0.10, 95% CI = [−0.001, 1.89], and the main effect of cue congruence, F(1, 38) = 3.14, p = .085, ηp2 = 0.08, 95% CI [−0.005, 0.07], was marginally significant. The main effect of cue sociality was significant, F(1, 38) = 9.64, p = .004, ηp2 = 0.20, 95% CI [−0.015, 0.07], with a significantly larger fixation count to ROIs in dot cues than gaze cues. The differences were not significant in all other conditions (p > .05).

We also used the validity effect (cued ROI − uncued ROI) as the dependent variable to further analyse the differences between gaze and dot cues. A 2 (autistic traits: High AQ vs. Low AQ) × 2 (cue sociality: gaze vs. dot) repeated-measures ANOVA was performed. The results showed no significant main effects or interactions.

Eye-tracking indicators for the postcue interface

For fixation count, 2 (autistic traits: High AQ vs. Low AQ) × 2 (cue sociality: gaze vs. dot) × 2 (cue congruence: cued vs. uncued ROI) repeated-measures ANOVA was performed. The results showed that the main effect of cue congruency, F(1, 38) = 22.10, p < .001, ηp2 = 0.37, 95% CI [0.11, 0.29], and an interaction of cue sociality by cue congruency, F(1, 38) = 4.36, p = .04, ηp2 = 0.10, was significant. The fixation count was significantly greater for gaze-cued ROIs than gaze-uncued ROIs (p < .001, 95% CI [0.14, 0.35]), and for dot-cued ROIs than dot-uncued ROIs (p = .001, 95% CI [0.07, 0.24]). The fixation count for the gaze-cued ROIs was also significantly greater than that for dot-cued ROIs (p = .035, 95%CI [0.006, 0.14]). The three-way interaction was marginally significant, F(1, 38) = 3.19, p = 0.08, ηp2 = 0.08. For gaze-uncued ROIs, fixation count was significantly higher in the High AQ than that in the Low AQ (p = .003, 95% CI [0.06, 0.28]), while for gaze-cued ROIs, the difference between two groups was not significant (p = .60). The High AQ had a significantly higher fixation count for dot-cued ROIs than dot-uncued ROIs (p = .048, 95% CI [0.001, 0.24]) and a trend of larger fixation count in gaze-cued ROIs than gaze-uncued ROIs (p = .075, 95% CI [−0.01, 0.28]). While the Low AQ had a significantly higher fixation count for gaze-cued ROIs than gaze-uncued ROIs (p < .001, 95% CI [0.21, 0.52]), and for dot-cued ROIs compared with dot-uncued ROIs (p = .005, 95% CI [0.06, 0.31]). The Low AQ also showed a significantly greater fixation count for gaze-cued ROIs than dot-cued ROIs (p = .003, 95% CI [0.06, 0.28]). The differences were not significant in all other conditions (p > .05).

Taking validity effect (cued ROI − uncued ROI) as the dependent variable, a 2 (autistic traits: High AQ vs. Low AQ) × 2 (cue sociality: gaze vs. dot) repeated-measures ANOVA was also performed. The results showed that the main effect of cue sociality was significant, F(1, 38) = 4.36, p = .04, ηp2 = 0.10, 95% CI [0.003, 0.18], with a significantly greater validity effect from gaze cues than dot cues. The interaction of autistic traits by cue sociality was marginally significant, F(1, 38) = 3.19, p = .08, ηp2 = 0.08. The Low AQ showed a larger validity effect from gaze than the High AQ (p = .04, 95% CI [0.01, 0.44]) and a larger validity effect from gaze cues than that from dot cues (p = .01, 95% CI [0.04, 0.31]). The differences were not significant in all other conditions (p > .05).

Eye-tracking indicators for the probe interface

For the percentage of first gaze shifts, the three-way interaction was significant, F(1, 38) = 6.53, p = .015, ηp2 = 0.15. Therefore, we directly take the validity effect (cued ROI − uncued ROI) as the dependent variable and a 2 (autistic traits: High AQ vs. Low AQ) × 2 (cue sociality: gaze vs. dot) repeated-measures ANOVA was performed. The results showed that the main effect of cue sociality was significant, F(1, 38) = 5.20, p = .03, ηp2 = 0.12, 95% CI [0.01, 0.20], with a significantly greater validity effect from gaze cues than that from dot cues. The interaction of cue sociality by autistic traits, F(1, 38) = 6.53, p = .015, ηp2 = 0.15, was significant. The Low AQ showed a larger validity effect from gaze than the High AQ (p = .02, 95% CI [0.03, 0.41]) and a larger validity effect from gaze cues than that from dot cues (p = .002, 95% CI [0.09, 0.35]). The differences were not significant in all other conditions (p > .05).

For the latency of first looks to the target, the three-way interaction was significant, F(1, 38) = 6.41, p = .02, ηp2 = 0.14. The validity effect (invalid-valid) was calculated as the dependent variable and a 2 (autistic traits: High AQ vs. Low AQ) × 2 (cue sociality: gaze vs. dot) repeated-measures ANOVA was performed. The results showed that the main effect of cue sociality was significant, F(1, 38) = 5.37, p = .03, ηp2 = 0.12, 95% CI [4.71, 70.05], with a significantly greater validity effect from gaze than that from dot cues. The interaction between autistic traits and cue sociality was significant, F(1, 38) = 6.41, p = .02, ηp2 = 0.14, The Low AQ showed a larger validity effect from gaze than the High AQ (p = .01, 95% CI [19.72, 135.76]) and a larger validity effect from gaze than that from dot cues (p = .002, 95% CI [30.91, 125.59]). The differences were not significant in all other conditions (p > .05). Figure 4 showed the results of eye-tracking results from cue, postcue and probe interfaces.

Fig 4
figure 4

Fixation patterns of the high and low AQ groups at different interfaces. High-dot = the High AQ in the dot cues; High-gaze = the High AQ in the gaze cues; Low-dot = the Low AQ in the dot cues; Low-gaze = the Low AQ in the gaze cues. Error bars represent the 95% confidence interval and the full dots represent individual observer data. *p < .05, **p < .01, ***p < .001.a Fixation count of two groups by cue sociality (gaze/dot) and cue congruence (cued/uncued ROI) at cue interface. b Fixation count of two groups by cue sociality (gaze/dot) and cue congruence (cued/uncued ROI) at postcue blank screen interface. c Percentage of first gaze shifts of two groups by cue sociality (gaze/dot) and cue congruence (cued/uncued ROI) at probe interface. d The latency of first looks to the target of the high and low AQ groups by cue sociality (gaze/dot) and cue validity (valid/invalid)

Full size image

For the fixation proportion to the target, 2 (autistic traits: High AQ vs. Low AQ) × 2 (cue sociality: gaze vs. dot) × 2 (cue validity: valid vs. invalid) repeated-measures ANOVA was performed. The results only showed a significant main effect of autistic traits, F(1, 38) = 4.71, p = .04, ηp2 = 0.11, 95% CI [0.51, 14.74], whereby the fixation proportion to the target was significantly greater for the High AQ than for the Low AQ.

Discussion

Our study is the first to study the gaze-triggered attention in individuals with high autistic traits from a temporal-stability perspective. The present study used eye-tracking technique to primarily investigate whether the changes of gaze-triggered attention in individuals with high autistic traits is due to the sociality of gaze and how autistic traits affected gaze-triggered attention over time. By comparing the gaze and nonbiological motion dot cues, we further excluded the influence of iris or pupil movement of gaze. The Low AQ showed a significant gaze-cueing effect and a marginally significant dot-cueing effect, and the gaze-cueing effect is significantly larger than the dot-cueing effect. However, no significant cueing effect was found in the High AQ, either in gaze or in dot cues. Eye movement results showed autistic traits affected the validity effect after 1,000 ms of cue presentation, but not within the 500 ms of cue presentation and within 500–1,000 ms after cue presentation. The results of this study are therefore two-fold: (1) the changes of gaze-triggered attention in the High AQ was social-specific; (2) autistic traits affected the gaze-triggered attention in a relatively late stage.

The findings of our study support our hypothesis that the changes of gaze-triggered attention in the High AQ was social specific. Compared with previous studies using arows as nonsocial cues (Bonmassar et al., 2019; Lin et al., 2020; Marotta et al., 2019), our study further ruled out the effect of low levels of iris or pupil movement. Both RT and eye movement data of probe interface showed that the Low AQ had a significantly greater benefits from the gaze than that from dot, which suggested the specificity of gaze-triggered attention in the Low AQ. However, the High AQ did not showed significant differences between benefits from these two cues, suggesting that the High AQ may treat gaze as the dot cues and ignore the social significance of the gaze. The Social Motivation Theory suggests that individuals with ASD fail to form a representation of the reward value of social stimuli, and the value of social stimuli is less for them than for individuals with TD. The reduced reward value of social stimuli in individuals with ASD would reduce their motivation and their willingness for social maintenance and social interaction. Therefore, they show deficiency in the processing of social stimuli and abnormalities in social functioning (Bottini, 2018; Lönnqvist et al., 2017). Some researchers also have found that, similar to individuals with ASD, the reward value of social stimuli was reduced in individuals with high autistic traits (Cox et al., 2015; Williams & Cross, 2018). Therefore, the changes of gaze-triggered attention in individuals with high autistic traits may be attributed to their inherent reduced value of social stimuli.

Our findings also suggested the attention triggered by gaze is more temporally stable than that triggered by dots. The data of postcue interface, the probe interface and RT all showed the significant benefits from gaze cues, however, the benefits from dot cues disappeared at the indicator of the latency of first looks to the target in the probe interface and RT data. This finding was consistent with the results of Liu et al. (2021), who used the sliding window method and found cueing effect from gaze last longer than that from arrows. Previous studies also found that compared with other cues, the gaze-cueing effect lasted longer and the inhibition of return (IOR) occurred later (Driver et al., 1999; Frischen et al., 2007). For example, Frischen and Tipper (2004) found that the IOR was exhibited only when the SOA was 2400ms under gaze cues. Yoxon et al. (2019) also found that when the SOA was 250 ms, 400 ms, 700 ms, and 1,000 ms, significant gaze-cueing effect was still observed. In contrast, there were no gaze-cueing effect in the 100, 1,700, and 2,400 ms SOA. These findings suggested that gaze as a special cue, which essentially distinguish from nonsocial cues due to its sociality.

The eye-tracking data in three interfaces suggested a specific temporal course of autistic traits on gaze-triggered attention. The relationship between autistic traits and cue sociality in validity effect changed over time. In the cue interface, there was no significant main or interaction effect, that is, neither the High AQ nor the Low AQ showed an attention preference for gaze-cued location or dot-cued location. However, the interaction of autistic traits and cue sociality was marginally significant in the postcue interface and was significant in the probe interface. The dynamic change of attention orienting suggested that autistic traits affected the gaze-triggered attention in a relatively late stage. Researchers have debated whether gaze-triggered attention is a form of reflexive attention or top-down voluntary attention (Feng & Zhang, 2014; Green et al., 2013; Stevens et al., 2008). Rapid, automatic reflexive attention usually occurs in the early stage of the cue presentation. In contrast, voluntary attention appears later, lasts longer and more likely to be influenced by top-down factors (Green et al., 2013). In our study, autistic traits had no effect on validity effect in the cue interface but had a marginally significant effect in the postcue interface and a significant effect in the probe interface. This result may imply that in the early stage of the cue presentation, the gaze-triggered attention is indistinguishable from dot cues and appear to be reflexive attention, but in the later stage, the gaze-triggered attention can be modulated by autistic traits and is more likely to be top-down voluntary attention.

The above temporal course of attention triggered by these two cues also suggested the gaze may intrinsically distinguish from nonsocial cues in attention orienting. It was found that uninformative gaze and arrow induced nearly identical response patterns in cueing effect, eye movement and event related potentials, which suggested these two kinds of cues triggered a similarly reflexive attention (Bonmassar et al., 2019; Hietanen et al., 2008; Stevens et al., 2008). In contrast, there are evidence to support the diverging behaviors in attention orienting triggered by gaze and nonsocial cues, which may rely on qualitatively different underlying processes (Joseph et al., 2014; Liu et al., 2021; Marotta et al., 2018; Marotta et al., 2019; Nie et al., 2018). For example, gaze were found to not only trigger reflexive orienting to gazed-at location but also induce voluntary attention (Chanon & Hopfinger, 2011; Friesen et al., 2004). In our study, the different effects of autistic traits on attention triggered by gaze and dot cues in the later stages suggested that attention orienting triggered by gaze and dots may rely on different underlying processes.

The fixation proportion to the target only showed a significant main effect of autistic traits, suggesting that the uninformative gaze did not affect the subsequent processing of the targets. In addition, the accuracy of visual matching task showed that there is no significant difference between the two groups under various conditions. This finding is inconsistent with previous studies, which have found that autistic traits affect the visual working memory (Richmond et al., 2013; Takahashi & Gyoba, 2012). In our study, the accuracy of visual matching task under each condition is about 80%, which is relatively low, suggesting that the task in our study may be difficult for participants. What is more, here, two irregular objects are only presented for 250 ms in the current design, participants may only encode one object within the 250 ms, which limit the participants’ working memory ability and resulted in no difference in task performance between the two groups. In the further, we can set various levels of difficulty to discuss the further influence of gaze-triggered attention. It is worth noting that the cueing effect and visual matching task were confused in some extent during the experimental task, which may affect the experimental results. In our study, although cues was uninformative and the visual matching task was quite difficult, the significant gaze-cueing effect still exists. What’s more, The RT data but not accuracy also showed similar trend as the eye tracking data in the probe interface. Therefore, due to the experimental design, we mainly focused on the attention triggered by these two cues and did not discuss performance of visual matching task too much.

In conclusion, the present study implicated that the changes of gaze-triggered attention in individuals with high autistic traits might be due to an insufficiency of voluntary attention in the late stage. These results are critical as they therefore suggest a dynamic process of gaze-triggered attention in individuals with high autistic traits over time, which provide inspiration for the study of social attention in individuals with ASD.

Conclusions

The present study was to investigate whether the changes of gaze-triggered attention in individuals with high autistic traits is social-specific and how autistic traits affected gaze-triggered attention over time. We used a cueing paradigm within a visual matching task and recorded individuals’ eye movements. Through the comparison of dynamic gaze and nonbiological motion cues, the results clearly demonstrated that the specific pattern of gaze-triggered attention in individuals with high autistic traits is due to the sociality of gaze rather than the motion of gaze in a relatively late stage. These results provide strong evidence for the changes of social attention in individuals with high autistic traits from a temporal-stability perspective and have important implications for the prevention and intervention in social attention changes in autism.