مردمک‌سنجی کاهش یافته واکنش پذیری عاطفی ناخودآگاه در اوتیسم را آشکار می سازد

Abstract Recent theoretical conceptualisations have suggested that emotion processing impairments in autism stem from disruption to the sub-cortical, rapid emotion-processing system. We argue that a clear way to ascertain whether this system is affected in autism is by measuring unconscious emotional reactivity. Using backwards masking, we presented fearful expressions non-consciously (subliminally) as well as consciously (supraliminally), and measured pupillary responses as an index of emotional reactivity in 19 children with autism and 19 typically developing children, aged 2–5 years. The pupillary responses of the children with autism revealed reduced unconscious emotional reactivity, with no group differences on consciously presented emotion. Together, these results indicate a hyporesponsiveness to non-consciously presented emotion suggesting a fundamental difference in emotion processing in autism, which requires consciousness and more time.

Introduction One of the most powerful ways to establish genuine interpersonal connections is through emotional communication (Rawlins, 1992). It is through the transfer of emotions and emotional arousal between people that relationships are built (Malloch and Trevarthen, 2009, Nummenmaa et al., 2012, Stern, 2010 and Tomasello et al., 2005). For example, sharing a laugh, smile or even a cry with someone can bring about a sense of social closeness. Although much of our social-emotional lives is conscious, feeling states can lie below the threshold for conscious awareness (Damasio, 1999), and unconscious emotion (both non-consciously triggered and non-consciously experienced) influences and shapes our everyday perceptions, behaviour, and social judgements (e.g., Berridge and Winkielman, 2003, Hall et al., 2007, Murphy and Zajonc, 1993 and Niedenthal, 1990). Emotional communication is a significant area of difficulty for individuals with Autism Spectrum Disorder (ASD), neurodevelopmental disorders defined by social-communicative difficulties and behavioural rigidity (American Psychiatric Association [APA], 2013). Research on individuals with ASD has indicated impairments in emotional reciprocity (Hobson, 1989 and Rogers and Pennington, 1991), which are likely to stem from difficulties recognising subtle and briefly presented emotional expressions, from atypical physiological responses to emotion, and from the ambiguous spontaneous expressions of own feeling states towards others (for review, see Nuske, Vivanti, & Dissanayake, 2013). Despite these deficits, individuals with ASD can recognise basic emotions from still images of prototypical facial expressions, and can imitate these when told to do so, at least in the laboratory setting (Nuske, Vivanti, & Dissanayake, 2013). Moreover, although individuals with ASD are often reported to have atypical neurophysiological and physiological responses during emotion processing tasks (e.g., reduced amygdala activation), evidence suggests they have fewer abnormalities in neural and autonomic responding during explicit emotion tasks (e.g., emotion labelling), compared to implicit emotion tasks (e.g., gender labelling of emotional facial expressions; Critchley et al., 2000, Hubl et al., 2003 and Kuchinke et al., 2011). Together, this pattern of results suggests more difficulty with implicit emotional appraisals than when attention is explicitly directed towards the emotional nature of the stimuli ( Nuske, Vivanti, & Dissanayake, 2013). This interpretation is consistent with suggestions that the emotion processing deficit in ASD stems from a fundamental problem with rapid, automatic or sub-cortically mediated emotion processing (e.g., Adolphs et al., 2001, Dawson et al., 2004, Johnson, 2005, McIntosh et al., 2006 and Oberman et al., 2009). Furthermore, a greater reliance on explicit processing of emotion in ASD is consistent with documented perceptual differences in this population, compared to the typical population. For example, studies using the face-inversion paradigm have found that whilst the emotion recognition performance of typically developing individuals is shown to be diminished by turning the face up-side-down, this is not the case for individuals with ASD (Gross, 2008 and Hobson et al., 1988). Other studies have found that people with ASD rely on featural or fine-grained details to process emotion (vs. using configural information or the gestalt, respectively; Deruelle et al., 2008 and Ozonoff et al., 1991). The results from these studies suggest an atypical approach to the processing of emotion from faces in ASD. Moreover, findings of greater activity in visual processing areas in ASD during explicit emotion recognition (Daly et al., 2012, Hadjikhani et al., 2009, Hubl et al., 2003, Kleinhans et al., 2010, Loveland et al., 2008 and Silani et al., 2008), indicate that these individuals may rely on rule-based, ‘disembodied’ explicit emotion processing strategies, possibly of visuo-perceptual origin (for a comprehensive review on the topic, see Winkielman, McIntosh, & Oberman, 2009). Such strategies may include noticing the widened eyes in fearful expressions, leading to recognition of the emotion and consequently triggering an emotional reaction, rather than a direct and more immediate internal simulation of the emotion. This notion of a relatively greater difficulty with implicit processing of emotion and the use of explicit, disembodied strategies for emotion tasks, is consistent with evidence of a temporal delay in emotion processing in ASD, as these explicit strategies are likely to take longer than implicit ones (e.g., Akechi et al., 2009, Bal et al., 2010, Korpilahti et al., 2007, Nuske et al., 2013, Nuske et al., 2014, Oberman et al., 2009 and Wong et al., 2008). Furthermore, difficulty with implicit processing and reliance on explicit cues for social information processing in ASD has been recently proposed outside of the realm of emotion (for eye gaze, theory of mind and imitation) (Senju, 2012), which perhaps suggests that all social cognition deficits in this disorder lie on the implicit level. Given that explicit attention and conscious awareness are “intimately bound together” (Crick & Koch, 1990, p. 269), if atypical emotional reactivity in ASD is due to the greater recruitment of disembodied, explicit strategies and a lesser reliance on embodied, implicit processes, one would expect a greater difficulty with unconscious, compared to conscious emotion processing in ASD, as the latter allows for explicit attention. 1.1. Unconscious emotion processing Some of the most convincing empirical work on unconscious emotion processing has been conducted with individuals who are cortically blind. Here, when emotionally inducing stimuli are presented to the individuals’ blind visual field, despite the absence of conscious registration of the stimuli, robust emotional reactions (facial, pupillary and neural) are triggered in these individuals (Morris et al., 2001, Pegna et al., 2005 and Tamietto et al., 2009). Likewise, studies on individuals with right-parietal lobe damage, who show spatial extinction (a pathological form of inattentional blindness for contralesionally presented stimuli presented simultaneously with stimuli to the ipsilesional visual field; Mack & Rock, 1998), have shown amygdala activation to non-consciously presented fearful faces ( Vuilleumier et al., 2002 and Williams and Mattingley, 2004). In typical development, unconscious emotion has been mainly studied through experimental manipulations that suppress conscious awareness of the emotional stimuli through backward masking (Esteves and Öhman, 1993 and Raab, 1961). Emotional faces are presented briefly (≈30 ms), and immediately followed by a ‘mask’ which is usually either a scrambled image of a face or an emotionally neutral face. Many studies have confirmed that emotional faces presented in such a way cannot be consciously recalled (e.g., Dannlowski et al., 2007, Dimberg et al., 2000, Morris et al., 1998, Murphy and Zajonc, 1993, Whalen et al., 1998 and Winkielman et al., 1997). Nevertheless, brief emotion exposure with such backward masking has been found to trigger neural (Morris et al., 1998, Morris et al., 1999, Pegna et al., 2008, Smith, 2012 and Whalen et al., 1998), facial reactivity (Bornemann et al., 2012, Dimberg et al., 2000, Rotteveel et al., 2001 and Tamietto et al., 2009) and skin conductance responses (Esteves et al., 1994 and Öhman and Soares, 1994). According to Le Doux (1996), these low-level emotional reactions can be triggered without consciousness due to the phylogenetically early design of the brain, which allows for the processing of emotional stimuli to bypass the primary visual cortex, through the rapid, sub-cortical, colliculo-thalamo-amygdala neural pathway, rather than being processed via the newer (and slower) cortical route (i.e., thalamus-sensory cortex-amygdala). 1.1.1. Unconscious emotion processing in autism Few studies have measured unconscious emotion processing in individuals with ASD, despite this arguably being one of the clearest ways to ascertain whether basic, low-level, sub-cortical emotional processes are deficient in this group. To examine how unconscious emotion affects later judgements of individuals with ASD, Kamio, Wolf and Fein (2006) used a modified version of the subliminal priming task by Murphy and Zajonc (1993). They found that the liking ratings of participants with ASD about (previously unseen) Japanese ideographs were unaffected by preceding, subliminally-presented emotional faces, but that they were in typically developing participants. Likewise, Hall et al. (2007) found that the face friendliness ratings of children with ASD were less likely be influenced by subliminally presented emotional information, compared to those of matched controls. Only two studies have measured neurophysiological responses during the viewing of backwardly masked emotions in individuals with ASD. In measuring brain activation to fearful faces presented with backward masking, Kleinhans et al. (2011) found reduced activation in the superior colliculi, pulvinar (an area of the thalamus), and amygdala (as well as in the fusiform gyrus) in individuals with ASD, relative to controls. On the basis of the numerous reports, mentioned above, there may be less sub-cortical engagement in ASD during emotion processing, especially in the absence of explicit attention to, or conscious awareness of the emotion (e.g., Critchley et al., 2000). However, also using a backward masking task, Hall, Doyle, Goldberg, West, & Szatmari (2010) found that adults with ASD had a similar magnitude of amygdala response (although reduced fusiform response), compared to matched controls. Given the limited research to date on response to unconscious emotions in ASD, and the contradictory current findings, more research is needed to establish whether there are indeed abnormalities in automatic and implicit, unconscious emotional reactivity in individuals with ASD. 1.2. The current study Our aim was to determine whether abnormalities in unconscious emotional reactivity are present in young children with ASD, by measuring pupillary reactions (using eye-tracking technology) to emotional expressions presented subliminally and supraliminally (below and above the threshold for consciousness, respectively). To our knowledge, this is the first study to measure physiological (autonomic) reactions to non-consciously presented emotion in individuals with ASD. Eye-tracking pupillometry has a number of advantages. First, pupillary responses have been shown to be a reliable marker of emotional arousal (Bradley et al., 2008 and Partala and Surakka, 2003), and have been long known to be functionally linked to the amygdala (Applegate et al., 1983, Graur and Siegle, 2013, Urry et al., 2006, Ursin and Kaada, 1960 and Zbrozyna, 1963), allowing for comparisons with neuroimaging studies. Second, pupillary responses provide a measure of emotional reactions regardless of whether the participant is aware of such changes (Laeng, Sirois, & Gredebäck, 2012). Third, whilst movement-related artefacts are common in neurophysiological and physiological data due to the sensitivity of the techniques to motion (Patriquin et al., 2013 and Tyszka et al., 2013), advanced eye-tracking systems are less prone to this type of artefact (explained further in Section 2.2). This issue is particularly relevant to the study of children with (and without) ASD, as they often have difficulty with staying still and following instructions. Fourth, eye-tracking pupillometry is non-invasive, and thus circumvents issues surrounding the application of electrodes (for measuring ERPs, skin conductance responses or heart-rate) which may in itself cause elevated arousal in individuals with ASDs who commonly present with tactile sensitivities (Marco, Hinkley, Hill, & Nagarajan, 2011). Fifth, the above-mentioned decrease in movement-related artefacts and the non-invasiveness of this technology together make it well-suited for use with lower functioning children with ASD, who are too often excluded from such research (for a recent discussion on this issue, see Vivanti, Barbaro, Hudry, Dissanayake, & Prior, 2013). We chose to examine responses to fear, as this emotion has been found to produce large, detectable neurophysiological and physiological responses (Adolphs et al., 1994 and Ekman et al., 1983). Based on the findings discussed above, we predicted that children with ASD would show reduced pupillary responses to fearful facial expressions, relative to matched controls, particularly when the emotions were non-consciously presented, with more typical (but still reduced) pupillary responses on conditions where the emotion presentation was above the threshold for consciousness. The most typical (or approximating typical) response from children with ASD was expected to be on the longer (vs. shorter) conscious processing condition, where there would be more opportunity to employ disembodied, explicit processing strategies. We also aimed to examine whether everyday empathic behaviour was related to unconscious and conscious emotional reactivity in children with and without ASD, as well as to investigate the relationship between of unconscious and conscious emotional reactivity with autism severity.

3. Results Data were first analysed for skewness, kurtosis and outliers. As data were normally distributed, parametric tests were used in all analyses. For brevity, peak percentage change from neutral to fear stimuli will be referred to as ‘peak amplitude (N → F)’ and peak percentage change on neutral/fear will be referred to as ‘peak amplitude (N)’/‘peak amplitude (F)’. 3.1. Preliminary analyses 3.1.1. Visual attention Firstly, to ensure that the pupillary reactivity recorded was in response to the neutral and fearful facial expressions, it was important to check the visual attention of the children to these images. Also, to determine whether to control for visual attention in the main analyses (Section 3.2), we checked whether the groups differed in visual attention and whether visual attention was related to peak amplitude (N → F, N, F). As shown in Fig. 2, there were no group differences in fixation duration to the fearful face (eyes + mouth area of interest, or AOI) in the 300 ms exposure condition, nor to the mouth region across all duration conditions. However, the ASD group looked less (i.e., had a shorter fixation duration) to the face in the 30 ms and 2 s exposure conditions. Moreover, the ASD group had a shorter fixation duration to the eye region across duration conditions, compared to the TD group, which was driven by fixation duration to the eye AOI in the 2 s exposure condition. As depicted in Fig. 3, although there are similarities between the fearful face-scanning patterns of children with and without ASD at the group level, the ASD group appeared to have slightly more scattered attention, with more time spent scanning outside the core emotional features of the face (i.e., eyes and mouth). Fixation duration (in seconds) to fearful expressions for different exposure ... Fig. 2. Fixation duration (in seconds) to fearful expressions for different exposure durations and face areas. The typically developing (TD) group looked longer than the Autism Spectrum Disorder (ASD) group on the 30 ms and 2 s condition (A and C), but not on the 300 ms condition (B). Overall, the TD group looked more to the eye area of interest (AOI), than the ASD group, but the groups did not differ on the mouth AOI (D and E), and this was driven by fixation duration on the 2 s condition (F, G and H). Values for each AOI represent a sum of the mean fixation duration for each exposure condition (2 s + 300 ms + 30 ms = max. 2.303 s). Error bars represent standard error of the mean. * p < .10, ** p < .05, *** p < .01. Figure options Average visual scan pattern for the fearful expression 2s condition, depicted by ... Fig. 3. Average visual scan pattern for the fearful expression 2 s condition, depicted by heat maps, for each group. (A) Children with autism spectrum disorder, (B) typically developing children. Colour signifies the duration of visual attention (red > yellow > green). Two layers of fixations for each group were superimposed on top of each other (one for each testing order). (For interpretation of reference to color in this figure legend, the reader is referred to the web version of this article.) Figure options The groups did not differ in fixation durations to the neutral face during the 30 ms (p = .62) and 300 ms (p = .39) conditions. However, the ASD group looked less to the neutral face in the 2 s condition (p < .001), as well as overall to the eye (p = .02) and mouth (p = .04) regions (across duration conditions), which was driven by a shorter fixation duration in the 2 s condition to the eye (p = .03) and mouth (p = .07) regions (the groups were not different in their fixation duration to the eye and mouth regions in the 30 ms and 300 ms conditions; p's > .29). For the ASD group, fixation duration to the neutral face in the 30 ms condition was moderately related to peak amplitude (N → F) in this exposure condition, r = .40, p = .09, and fixation duration to the eye region of the fearful faces was related to peak amplitude (F) in the 300 ms exposure condition, r = .51, p = .02. For the TD group, peak amplitude (F) in the 300 ms exposure condition was related to fixation duration to the eye region, r = −.56, p = .01, and mouth region, r = .47, p = .04, of fearful faces. No other correlations between fixation duration and peak amplitude were significant for either group (r range = ±.003–.39, p range = .10–.99). Thus, taking a conservative approach to control for the potential confound of diminished attention to the stimuli in the ASD group, we controlled for fixation duration in our main analyses of peak amplitude (N → F, N, F) data (Section 3.2), using a composite score of fixation duration across all conditions (neutral 30 ms + neutral 300 ms + neutral 2 s + fearful 30 ms + fearful 300 ms + fearful 2 s). 3.1.2. Cognitive ability The groups differed on cognitive ability (see Table 1) with the ASD group functioning at a lower level than the TD group. To determine whether to control for cognitive ability in the main analyses, we examined whether cognitive ability (the MSEL standard score) was related to peak amplitude (N → F, N, F) or to the EmQue scores. In the ASD group only, cognitive ability was related to peak amplitude (F) in the 30 ms, r = .52, p = .02, and the 300 ms exposure conditions, r = .46, p = .049. No other correlations were significant (r range = ±.001–.23, p range = .10–.997). Thus we controlled for cognitive ability in the main analysis including peak amplitude (F) (Section 3.2.2). Other analyses were conducted without covarying the MSEL standard score. (For comparison, the results covarying the MSEL standard score in these analyses are also given in the footnotes.) 3.2. Main analyses 3.2.1. Peak amplitude (N → F) To determine whether the groups differed in response to subliminally and supraliminally presented fearful expressions, a two-way ANCOVA (2 groups × 3 exposure durations) was conducted on peak amplitude (N → F), controlling for visual attention (fixation duration) to the stimuli. However, the effect of the visual attention covariate was non-significant, F(2,34) = .23, p = .80, η2 = .01. Therefore we re-ran this analysis without the visual attention covariate. (For comparison, the results with the visual attention covariate are also given in the footnotes.) 3 The main effect of Exposure was not significant, F(2,35) = .39, p = .68, η2 = .02, but there was a significant main effect of Group, F(1,35) = 19.11, p < .001, η2 = .35. However, as shown in Fig. 4, this appeared to be driven by the Group × Exposure interaction, F(2,35) = 4.29, p = .02, η2 = .20. Peak percentage change from neutral to fear (peak amplitude N→F), for 30ms ... Fig. 4. Peak percentage change from neutral to fear (peak amplitude N → F), for 30 ms (unconscious exposure), 300 ms and 2 s (short and longer conscious exposures) for each group (unadjusted means). Error bars represent standard error of the mean. Figure options Between-group pairwise comparisons showed a group difference on the 30 ms condition (p < .001, η2 = .33) and the 300 ms condition (p = .002, η2 = .23), with positive values in the TD group and negative values in the ASD group. This indicated that whilst the TD group responded more to the fear expressions that then neutral expressions in the 30 ms and 300 ms conditions, the ASD group showed the opposite pattern, responding more to the neutral faces than the fearful faces. The groups did not differ on the 2 s condition (p = .93, η2 < .001). Within-group pairwise comparisons showed that, for the TD group, peak amplitude (N → F) was not different across the exposure conditions (p's > .18). By contrast, for the ASD group, peak amplitude (N → F) in the 30 ms condition was significantly different to the 2 s condition (p = .01), and the 300 ms was marginally different from the 2 s condition (p = .09). There was no difference between the 30 ms and 300 ms condition (p = .15). 4 3.2.2. Peak amplitude (N, F) As the peak amplitude (N → F) analyses revealed that the ASD group had larger pupillary reactions to neutral faces than fearful faces, we next analysed the neutral and fearful conditions separately with a three-way ANCOVA (2 Groups × 3 Exposures × 2 Emotions) on peak amplitude (N, F), controlling for visual attention (fixation duration) to the stimuli and cognitive ability (MSEL standard score). Neither the effect of the visual attention covariate, F(1,34) = 1.04, p = .32, η2 = .03, nor the cognitive ability covariate were significant, F(1,34) = .48, p = .50, η2 = .01. Therefore, we re-ran this analysis without these covariates. (For comparison, the results with the visual attention and cognitive ability covariates are also given in the footnotes). 5 None of the main effects were significant: Group, F(1,36) = .66, p = .42, η2 = .02, Exposure, F(2,35) = 2.44, p = .10, η2 = .12, and Emotion, F(1,36) = .31, p = .58, η2 = .01, but two interaction effects were evident. The first was the Group × Emotion interaction, F(1,36) = 21.28, p < .001, η2 = .37, which, as shown in Fig. 5, was driven by the second interaction, the Group × Exposure × Emotion interaction, F(2,35) = 3.66, p = .04, η2 = .17. Peak percentage change from the last 100ms of the pre-scrambled image to the ... Fig. 5. Peak percentage change from the last 100 ms of the pre-scrambled image to the neutral and fearful faces (peak amplitude N, F), for the 30 ms (unconscious exposure), 300 ms and 2 s (short and longer conscious exposures), for each group (unadjusted means). Error bars represent standard error of the mean. Figure options Between-group pairwise comparisons showed a group difference on the 30 ms condition, for both the neutral (p = .046, η2 = .11) and the fearful faces (p = .01, η2 = .17). However, in the 300 ms condition, the groups were not different on the neutral (p = .11, η2 = .07) or the fearful faces (p = .16, η2 = .05); likewise, in the 2 s condition, the groups were not different on the neutral (p = .12, η2 = .07) or the fearful faces (p = .26, η2 = .04). Within-group pairwise comparisons showed that, for the TD group, peak amplitude (N and F) were similar across the exposure conditions (p's > .17). Likewise, for the ASD group, peak amplitude (N) was not different across the exposure conditions (all p's = 1.00). In contrast, for the ASD group, peak amplitude (F) in the 30 ms condition was significantly smaller than in the 300 ms condition (p = .04) and the 2 s condition (p = .02), with no difference between the 300 ms and 2 s conditions (p = 1.00). Finally, as reflected in the peak amplitude (N → F) analyses in Section 3.2.1, whilst the TD group had larger peak amplitudes to fearful faces than to neutral faces in the 30 ms, p = .003, and the 300 ms conditions, p = .01 (with no difference on the 2 s condition, p = .60), the ASD group showed the opposite pattern. The ASD group had larger peak amplitudes to the neutral faces compared to the fearful faces in the 30 ms condition, p = .001 (with no difference on the 300 ms, p = .12, or 2 s conditions, p = .95). 3.2.3. Everyday empathic behaviour As can be seen in Table 2, parents rated children in the ASD group significantly lower on the EmQue total score, as well as on the Attention to Others’ Feeling and Prosocial Actions EmQue sub-scales. The groups were not differentiated on the EmQue Emotional Contagion sub-scale. Table 2. Group differences on everyday empathic behaviour, as measured by the EmQue. ASD group M (SD) TD group M (SD) t-value p-value Cohen's d EmQue total 13.83 (6.72) 21.16 (3.83) 4.10 <.00 1.39 Emotional contagion 3.22 (2.26) 3.68 (1.98) .66 .51 .22 Attention to others’ feelings 7.50 (2.94) 10.42 (2.06) 3.52 <.00 1.19 Prosocial actions 3.11 (2.74) 7.05 (1.22) 5.70 <.00 1.93 Table options 3.3. Correlation analyses 3.3.1. Associations of everyday empathic behaviour with emotional reactivity and visual attention To test the hypothesis that everyday empathic behaviour is related to emotional reactivity and attention, Pearson's correlations between EmQue scores and peak amplitude (N → F), as well as with fixation duration were computed (shown in Table 3). With regard to emotional reactivity, in the ASD group, the Emotional Contagion EmQue sub-scale was strongly and negatively associated with peak amplitude on the 30 ms condition, and the Attention to Others’ Feelings sub-scale was moderately, negatively related to peak amplitude on the 300 ms condition. By contrast, in the TD group, peak amplitude in the 2 s condition was related to the overall EmQue score, which seem to be driven by the strong, negative correlation between this and the Emotional Contagion EmQue sub-scale. Furthermore, although the relationship in the ASD group between the Emotional Contagion sub-scale and peak amplitude in the 2 s condition was diluted once the effect of fixation duration was controlled, the direction of the association remained positive (not negative, as for the TD group). With regard to emotional attention, in both groups, the Emotional Contagion sub-scale was moderately, negatively related to visual attention to the face (eye AOI + mouth AOI), which in both groups appeared to be driven by attention to the eye region. The overall pattern of associations remained when controlling for cognitive ability (and visual attention for peak amplitude variables). Table 3. Correlations of empathy with peak amplitude (PA) N → F and visual attention (VA). EmQue Total Emotional contagion Attention to others’ feelings Prosocial actions ASD group PA a 30 ms −.35 (−.34) (−.35) −.55** (−.57**) (−.57 **) −.18 (−.16) (−.16) −.22 (−.21) (−.22) 300 ms −.36 (−.31) (−.26) −.33 (−.28) (−.28) −.45* (−.42*) (−.37) −.12 (−.09) (−.02) 2 s .24 (.16) (.23) .40* (.31) (.33) .37 (.31) (.41) −.13 (−.18) (−.13) VA b, c Eyes −.25 (−.33) −.40*(−.34) −.20 (−.32) −.07 (−.20) Mouth −.24 (−.25) −.28 (−.25) −.20 (−.23) −.14 (−.18) Face −.33 (−.41 *) −.47**(−.42*) −.27 (−.39) −.13 (−.27) TD group PA a 30 ms <.01 (.05) (.05) −.09 (<−.01) (<−.01) .19 (.21) (.21) −.18 (−.18) (−.20) 300 ms −.13 (−.13) (−.13) <−.01 (<.01) (<.01) −.29 (−.29) (−.29) .09 (.09) (.09) 2 s −.46** (−.42*) (−.42) * −.62*** (−.57*) (−.58 **) −.13 (−.12) (−.12) −.24 (−.24) (−.22) VA b, c Eyes −.03 (−.03) −.29 (−.29) .31 (.31) −.15 (−.15) Mouth −.14 (−.15) <−.01 (<−.01) −.31 (−.31) .07 (.06) Face −.23 (−.23) −.40*(−.40 *) −.02 (−.02) −.04 (−.05) a For peak amplitude (PA) from neutral to fear (N → F), the r-values for partial correlations, controlling for the effect of visual attention (fixation durations) are in bold and in parentheses, controlling for the effect of visual attention and cognitive ability (MSEL standard score) are in italics and parentheses (not bold), and r-values outside parentheses are without controlling for either of these effects. b For visual attention (VA), the r-values for partial correlations, controlling for the effect of cognitive ability (MSEL standard score) are in italics and parentheses, and r-values outside parentheses are without controlling for either of these effects. c Neutral + fearful condition composite score, measured by fixation duration. Significant/marginally significant correlations shaded in grey. * p < .10. ** p < .05. *** p < .01. Table options 3.3.2. Associations of autistic symptoms with emotional reactivity and visual attention To determine whether autism symptoms are related to emotional reactivity and attention, Pearson's correlations between ADOS scores and peak amplitude (N → F), as well as with fixation duration were computed. In the ASD group, the Play Algorithm of the ADOS was negatively related to peak amplitude (N → F) in the 2 s exposure condition (r = −.49, p = .03), indicating that children who responded more in the longer conscious emotion exposure condition had fewer deficits in the area of functional and symbolic play. No other ADOS sub-scales (the Social Communication Algorithm nor the Restrictive and Repetitive Behaviours Algorithm) were related to peak amplitude (N → F) (p range = .24–.89). The ADOS sub-scales were not related to visual attention (fixation duration) to the eye AOI, mouth AOI, nor the face (eye AOI + mouth AOI) on the emotional expressions (p range = .11–.87). 3.3.3. Associations between unconscious and conscious emotion processing For the ASD group only, within-group differences in peak amplitude (N → F) were found between the 30 ms and 2 s exposure conditions (with an intermediate peak amplitude in the 300 ms condition), suggesting that different processing strategies may be employed for unconscious (30 ms) and longer conscious (2 s) emotion exposures in this group. To test this hypothesis, we computed Pearson's correlations between the exposure conditions. As evident in Table 4, whilst the 30 ms and 300 ms conditions were positively associated, associations with the 2 s condition were negative, with a significant, moderate and negative correlation between the 300 ms condition and the 2 s conditions. These associations remained when controlling for visual attention and cognitive ability. Table 4. Inter-condition N → F associations for the ASD group. PA 30 ms & 300 ms PA 30 ms & 2 s PA 300 ms & 2 s Simple correlations .50** −.20 −.47** Partial correlations Controlling for visual attention .50** −.19 −.45* Controlling for visual attention and cognitive ability .52** −.18 −.53** * p < .10. ** p < .05.