Look at them and they will notice you : Distractor-independent attentional capture by direct gaze in change blindness

Full Terms & Conditions of access and use can be found at https://www.tandfonline.com/action/journalInformation?journalCode=pvis20 Visual Cognition ISSN: 1350-6285 (Print) 1464-0716 (Online) Journal homepage: https://www.tandfonline.com/loi/pvis20 Look at them and they will notice you: Distractorindependent attentional capture by direct gaze in change blindness Pessi Lyyra, Piia Astikainen & Jari K. Hietanen To cite this article: Pessi Lyyra, Piia Astikainen & Jari K. Hietanen (2018) Look at them and they will notice you: Distractor-independent attentional capture by direct gaze in change blindness, Visual Cognition, 26:1, 25-36, DOI: 10.1080/13506285.2017.1370052 To link to this article: https://doi.org/10.1080/13506285.2017.1370052 © 2017 The Author(s). Published by Informa UK Limited, trading as Taylor & Francis Group Published online: 19 Sep 2017. Submit your article to this journal Article views: 802 View related articles View Crossmark data Citing articles: 4 View citing articles Look at them and they will notice you: Distractor-independent attentional capture by direct gaze in change blindness Pessi Lyyra a,b , Piia Astikainen b and Jari K. Hietanen a a Human Information Processing Laboratory, Faculty of Social Sciences/Psychology, University of Tampere, Tampere, Finland; b Department of Psychology, University of Jyväskylä, Jyväskylä, Finland ABSTRACT Humans have shown a detection advantage of direct vs. averted gaze stimuli in visual search tasks. However, instead of attentional capture by direct gaze, the detection advantage in visual search may depend on attention-grabbing potential of the distractor stimuli to which the target needs to be compared. We investigated attentional capture by direct gaze using the change blindness paradigm, in which successful detection does not require comparison between the target and the distractor items. Participants detected a masked gaze direction change in one of four simultaneously presented schematic faces. The distractor gaze directions were systematically varied across three experiments. Changes resulting in direct gaze were detected more efficiently than those resulting in averted gaze, independently of distractor gaze directions. This finding suggests that the detection advantage is specifically due to attentional capture by direct gaze, not properties of distractor items. ARTICLE HISTORY Received 31 August 2016 Accepted 8 August 2017 KEYWORDS Gaze perception; change detection; change blindness; stare-in-the-crowd effect Gaze perception is one of the foundations of nonverbal human communication, along with the perception of facial expressions, body posture, and biological motion. Gaze conveys information specifically about other persons’intentions, immediate interests, and direction of attention. During face-to-face social interaction, humans particularly attend to each other’s eye areas (George & Conty, 2008), and they are naturally adept at judging each other’s gaze direction (Gale & Monk, 2000; Symons, Lee, Cedrone, & Nishimura, 2004). At the neural level, this is enabled by subcortical and cortical mechanisms dedicated to gaze processing (e.g., Baron-Cohen, 1995; Nummenmaa & Calder, 2009; Perrett, Hietanen, Oram, Benson, & Rolls, 1992). Within gaze perception, the perception of direct gaze enjoys a particular relevance. Infants prefer direct gaze to averted gaze (Farroni, Csibra, Simion, & Johnson, 2002; Farroni, Menon, & Johnson, 2006). Mutual gaze of the newborns and their immediate caregivers supports the earliest interaction and enables formation of attachment relations (Argyle & Cook, 1976; Reddy, 2003). For adult humans, receiving direct gaze typically marks the beginning of an interaction and the perception of direct gaze automatically initiates a set of preparatory cognitive and physiological processes specific to approach behaviour (e.g., Conty, George, & Hietanen, 2016). These processes also include attentional ones: direct gaze has been shown to attract and grab the perceiver’s attention. In the visual search paradigm, detection times are shorter for deviant direct gaze targets among averted gaze distractors than for averted gaze targets among direct gaze distractors, a phenomenon dubbed the “stare-in-the-crowd”effect (Conty, Tijus, Hugueville, Coelho, & George, 2006; Doi, Ueda, & Shinohara, 2009; Senju, Hasegawa, & Tojo, 2005; Shirama, 2012; von Grünau & Anston, 1995). Moreover, even a task-irrelevant direct gaze facilitates visual search for facial expressions (Doi & Shinohara, 2013), and targets presented at the location of a direct gaze are detected faster than those presented at the location of averted gaze (Böckler, van der Wel, & Welsh, 2014; Miyazaki, Ichihara, Wake, & Wake, 2012). However, the speed of detection of direct gaze targets is shown to depend on the gaze directions of the distractor crowd (Palanica & Itier, 2011). There is © 2017 The Author(s). Published by Informa UK Limited, trading as Taylor & Francis Group This is an Open Access article distributed under the terms of the Creative Commons Attribution-NonCommercial-NoDerivatives License (http://creativecommons.org/licenses/by-nc-nd/ 4.0/), which permits non-commercial re-use, distribution, and reproduction in any medium, provided the original work is properly cited, and is not altered, transformed, or built upon in any way. CONTACT Pessi Lyyra [email protected] Human Information Processing Laboratory, Faculty of Social Sciences/Psychology, University of Tampere, Tampere, Finland; Department of Psychology, University of Jyväskylä, Jyväskylä FI-40014, Finland VISUAL COGNITION, 2018 VOL. 26, NO. 1, 25–36 https://doi.org/10.1080/13506285.2017.1370052 also recent evidence from studies using the visual search paradigm suggesting that direct gaze is not necessarily prioritized by detection alone, but the findings more likely reflect the attention-grabbing potential of the distractor stimuli across different search conditions (Cooper, Law, & Langton, 2013). For example, when the distractor context was identical— uniform or heterogeneous—the detection of direct and averted gaze deviants were equally efficient. In other words, the context comprising the distractor stimulus array can both disrupt and facilitate target detection. Cooper et al. (2013) found no evidence for the stare-in-the-crowd effect in the visual search paradigm when these contextual confounding factors were controlled for. It is possible that this paradigm is particularly prone to this problem, since it is based on deviance detection, i.e., the target is the sole deviant item in the search array, and deviant detection is based on comparison of that item to all of the distractor items. Consequently, visual search may not be the optimal way to study the attentional capture by direct gaze because of the attention-grabbing potential of the simultaneously presented distractor faces. Alternatively, the effect that attention has in the breaking of direct gaze stimuli to consciousness can be investigated with the so-called change blindness paradigm. Ordinary change detection in humans is highly efficient: if two versions of an image are successively presented to observers trying to detect changes from one version (S1) to another (S2), changes are immediately detected (Rensink, O’Regan, & Clark, 1997). Consequently, however, this setup cannot be used to study the detection of different change types because of a ceiling effect. In a change blindness paradigm, by contrast, automatic change detection is eliminated by a mask, for example, a blank screen (“flicker”) interspersed between S1 and S2 (Rensink et al., 1997). The mask eliminates the change transient that would normally summon attention to its location, and automatically bring the changed stimulus into consciousness. In the change blindness paradigm, it may take up to several seconds for the observer to notice even a large change. The delayed detection rate thus allows for the emergence of detection differences between different stimulus change types, and allows for studying differential detection of direct and averted gaze changes (Yokoyama, Ishibashi, Hongoh, & Kita, 2011). The change blindness paradigm is typically used to study attentional effects of the stimuli, since successful change detection depends on attention (e.g., Rensink et al., 1997). The inability to perceive the change typically triggers an active search period guided by top-down attention. Only focal attention to the change—and consolidation to the sensory short-term memory—allows for transient perception of change to emerge into consciousness across the unstimulated periods (e.g., Jensen, Yao, Street, & Simons, 2011; Rensink et al., 1997). Evidence has been provided that bottom-up preconscious perception of the changed features can aid the explicit detection of the changes (Smilek, Eastwood, & Merikle, 2000). In particular, socially and biologically relevant contents of the change have been shown to facilitate recovery from change blindness, such as facial vs. non-facial stimuli (Kikuchi, Senju, Tojo, Osanai, & Hasegawa, 2009; Lyyra, Mäkelä, Hietanen, & Astikainen, 2014; Ro, Russell, & Lavie, 2001) and threatening vs. non-threatening facial stimuli (Lyyra, Hietanen, & Astikainen, 2014). The facilitated recovery probably stems from an attentional bias created by the bottom-up signal of the implicitly perceived changes. Thus, it would be expected that direct gaze as a socially and biologically salient stimulus would have a similar bottom-up attentional influence, and, therefore, it should enhance the detection of direct gaze stimuli relative to averted gaze stimuli. This finding would be akin to corresponding findings in interocular suppression, in which a target stimulus is presented to one eye, and conscious perception of the target is delayed by a dynamic high contrast mask presented to the other eye. This delay is found to be shortened for direct gaze, i.e., it breaks to consciousness faster than averted gaze (Akechi et al., 2014; Chen & Yeh, 2012; Stein, Senju, Peelen, & Sterzer, 2011). In interocular suppression, however, it is not clear which cognitive functions contribute to the quickened breakthrough of direct gaze to consciousness, and the role of attention in this phenomenon has not been studied directly (see, e.g., Sterzer, Stein, Ludwig, Rothkirch, & Hesselmann, 2014). Unlike in interocular suppression, enhanced detection of direct gaze stimuli in a change blindness condition could show that attention can contribute to this process, since attention to the changes is necessary for change detection in change blindness: If direct gaze captures attention more efficiently than an averted gaze at the preconscious level, it should lead to an attentional bias to direct gaze stimuli, as found for other socially relevant stimuli. 26 P. LYYRA ET AL. Importantly, in change blindness, the changed item does not have to differ from the other items in S2, and there is no need to compare the changed item to the other items in the scene. Additionally, the distractor stimuli are not necessarily present in the response phase, potentially decreasing their distractor value. Therefore, attentional capture by target stimuli in a search condition may be investigated more effectively in the change blindness paradigm than in the visual search paradigm. In a previous study using the change blindness paradigm, changes to direct gaze were indeed more efficiently detected than those to averted gaze (Yokoyama et al., 2011). However, the composition of the distractor gaze directions was similar throughout theexperiments inthatstudy.Thus, basedonthatstudy, it was impossible to show unequivocally that the direct gaze detection advantage in change blindness was not influenced by attention-grabbing by the distractor gaze directions or by a comparison between the distractors and the target, as in visual search. The purpose of the present study was to systematically investigate the direct gaze advantage and the effects of the background distractor gaze information on this advantage in a change detection condition. To this end, we conducted a series of three change blindness experiments in which we systematically manipulated the gaze distractor background, similarly to Cooper et al.’s(2013) study, which used the stare-inthe-crowd visual search paradigm. In Experiments 1– 3, we used a uniform background with distractor faces displaying an opposite gaze direction to the target (direct or laterally averted gaze, Experiment 1), an identical distractor background for both conditions (all distractors displaying an upward or downcast gaze, Experiment 2), and a heterogeneous background (two distractor faces displaying a direct and two displaying a laterally averted gaze, Experiment 3). Changes were either from averted to direct gaze offering the possibility for eye contact (below referred to as a “change to direct gaze”), or from direct (or from upward/downcast) to averted gaze (below referred to as a “change to averted gaze”), as depicted in Figures 1–3.Inall experiments, the participants’task was to detect changes across two scenes consisting of an array of four faces with an occasional and randomly located gaze direction change. We assumed that changes to direct gaze are detected more efficiently than changes to averted gaze, independent of gaze directions of the distractor face stimuli, since change detection does not depend on perception of distractor stimuli. As change detection in the change blindness paradigm relies on attention to the changes, enhanced Figure 2. Detection efficacies in changes to direct gaze (black bar) and to averted gaze (grey bar) conditions in Experiments 1–3. Error bars indicate standard errors of the mean. The asterisks indicate significance level of p< .05. Figure 1. Stimulus setups for the change to direct gaze and change to averted gaze conditions in Experiment 1. Changes in S2s are indicated with dotted circles. VISUAL COGNITION 27 detection of changes to direct gaze would indicate that direct looking faces attract attention more efficiently than averted looking faces. If the change detection advantage remained similar despite manipulation of the distractor gaze composition, it would mean that change detection in a stare-in-the-crowd is relatively immune to the effect of distractor gaze directions. The change blindness paradigm could thus inform us about whether faces displaying direct gaze attract attention more strongly than those displaying averted gaze, and consequently enhance the conscious detection of faces with direct gaze—a question left unaddressed by the visual search and interocular suppression studies. Experiment 1 As in Cooper et al. (2013), Experiment 1 was aimed as a repetition of the original stare-in-the-crowd effect (Experiment 1; von Grünau & Anston, 1995) to confirm that our stimulus setup works similarly to previous studies. We used a corresponding composition of gaze direction stimuli as in the experiment by von Grünau and Anston (1995), but adapted to the change blindness paradigm. The crowd in the S1 displays was composed of either faces with exclusively averted gaze or direct gaze. One of the faces in S2 occasionally changed gaze direction, thus appearing like one of the faces had turned its gaze. The change direction was from averted to direct (“change to direct gaze”), or from direct to averted gaze (“change to averted gaze”). The faces were schematic faces as in Öhman, Lundqvist, and Esteves (2001; cf. Yokoyama et al., 2011), but were manipulated to differ slightly in terms of the size of facial elements and their relative distances (see Figure 1) to render the change detection task more challenging. The presentation times of S1 and S2 were kept very short (i.e., 30 ms) for the task performance to reflect change detection, not deviance detection in S2 (Lyyra, Hietanen et al., 2014). The main hypothesis was that changes to direct gaze would be detected more efficiently than changes to averted gaze, corresponding to the stare-in-thecrowd effect in visual search. Material and methods Participants Sixteen healthy individuals (8 males, mean age ± SD = 23.1 ± 3.6), participatedinthestudy,andtheywerecompensated with movie tickets. All participants gave their written informed consent before the experiment. The study conformed to the ethical standards of the American Psychological Association (APA) and The Code of Ethics of the World Medical Association (Declaration of Helsinki). According to Finnish regulations (Act on Medical Research and Decree on Medical Research 1999, amended 2010), specific ethics approval was not necessary for this study. A written informed consent was obtained from all participants. The sample size was estimated based on previous studies (Lyyra, Hietanen, et al., 2014; Yokoyama et al., 2011). To investigate whether this sample size has sufficient power to reveal the detection difference between changes to averted and to direct gaze, an a priori power analysis for repeated measures ANOVA with one two-level within-subjects factor was conducted using G*Power 3.1 software (Faul, Erdfelder, Figure 3. Stimulus setups for the change to direct gaze and change to averted gaze conditions in Experiment 2. Changes in S2s are indicated with dotted circles. 28 P. LYYRA ET AL. Lang, & Buchner, 2007), with the parameters of 95% power, expected effect size of at least 0.2–0.3 ( h 2 p), an alpha level of .05, the default within-subjects measurement correlation of .5, and non-sphericity correction value (ε) of 1. The expected effect size was estimated based on previous corresponding change blindness studies with reported effect size (e.g., Lyyra, Hietanen, et al., 2014; Yokoyama et al., 2011). The calculation suggested sample sizes of 10–16 participants. Stimuli and procedure The stimuli were schematic faces with a direct gaze or an averted gaze (see Figure 1), and were presented in an S1–S2 change blindness paradigm. In S1 (30 ms), four faces were simultaneously presented in an array of 2 × 2 faces (see Figure 1). Each face covered an area of 5° × 5°, and the whole array subtended 17.9° × 13.3° of visual angle horizontally and vertically, respectively. Half of the S1s contained an array of four faces all with direct gaze, and half of S1s four faces all with laterally (randomly left or right) averted gaze. S1s were followed by a blank screen for 800 ms after which S2 (30 ms) was presented. The stimulus setup was designed to result in approximately 50% change detection rate and consequently an equal amount of trials for change detection and change blindness trials. On one-third of the trials, S2 was identical to S1. On another third of the trials, in those with an S1 of four faces with averted gaze, the gaze of one face changed from averted to direct (change to direct gaze). On the remaining third of the trials, those with an S1 of four faces with direct gaze, the gaze of one face changed from direct to randomly left or right laterally averted (change to averted gaze). The location of the change was randomized in both conditions. The participants’task was to detect whether a change between S1 and S2 occurred or not. There were no restrictions imposed on eye movements. Participants did not need to recognize the direction of the change, but it sufficed to sense clearly where the change transient occurred. They were discouraged to report having detected the change if they felt unsure about the location of the change. A response window followed S2, in which the participant reported on having detected the change or not by pressing of a left or right button assigned to the given response on a two-button response panel. The assigned buttons for yes/no responses were counterbalanced across participants. Participants completed as many trials as could manage in 40 minutes (M= 320.84 ± SD = 78.92). Detection rates, for both change conditions (change to direct gaze, change to averted gaze) were measured for each participant. The proportion of reports of change despite their absence was very small (1%). The no change trials served as catch trials, and the low false alarm rate indicated that the participants were not guessing but following the task instructions correctly. To investigate that the results were not due to a response bias, as the general performance level was eventually below chance (31.82%, see Figure 2), the discriminability index d’was calculated for each participant. The d’, according to Signal Detection Theory (e.g., Green & Swets, 1988), shows, in terms of standard deviations, the distance between the distributions of successful detection and false alarms. The d’s were high, M= 2.06, SD = 0.12, 95% CI [1.83, 2.30], showing that the participants were following the task instructions correctly and detecting changes successfully. Apparatus Stimulus presentation and data acquisition were controlled by E-Prime software (Psychology Software Tools). The stimuli were presented on a 23”CRT monitor (screen resolution: 1280 × 1024 pixels; refresh rate: 75 Hz) at a distance of 100 cm from the participant. Data analysis Change detection was measured as the ratio (percentage) of detected/all changes (Yokoyama et al., 2011). Mean detection rates of each participant were subjected to a repeated measures ANOVA with Gaze Direction (change to direct gaze, change to averted gaze) as a within-subject factor. An alpha level of .05 was used in all the analyses. Partial eta squared ( h 2 p) represents effect size estimates for ANOVA. Results and discussion Change detection performance Expectedly, a detection advantage for changes to direct gaze relative to changes to averted gaze emerged (M change to direct gaze = 35.38%, SD = 17.02; M change to averted gaze = 28.25%, SD = 14.21, see Figure 2), F(1, 15) = 7.35, p= .016, h 2 p=.329. VISUAL COGNITION 29 Changes from averted to a direct gaze were detected more efficiently than changes from direct to an averted gaze. Direct gaze, thus, seems to break into consciousness faster than averted gaze also when considering change detection. These results repeat the original stare-in-the-crowd effect observed for visual search (e.g., Cooper et al., 2013; von Grünau & Anston, 1995) for change blindness. As in the original version of the visual search paradigm, in which the stare-in-the-crowd phenomenon was first recognized, the present setup had averted gaze distractors for changes to direct gaze and direct gaze distractors for changes to averted gaze. This kind of a stimulus setup has recently been criticized as the effect of the distractor context, or the “crowd”, can confound this result: the enhanced detection rates of faces with direct gaze compared to those with averted gaze may be due to the distraction effect of the crowd being comprised of attention-grabbing direct looking faces (Cooper et al., 2013; see Frischen, Eastwood, & Smilek, 2008 for a similar problem concerning detection advantage for threatening faces). Therefore, it is possible that the more efficient detection of changes to direct than to averted gaze is due to a distracting, attention engaging effect of the direct looking distractor faces in the stimulus array of the change to averted gaze condition. To eliminate the possible confounding effect of distractor context in change conditions, another experiment was conducted using the same context for both the change to averted gaze and the change to direct gaze conditions. Experiment 2 The same motivation formed the foundation for Experiment 2 and it was devised in the same fashion as Experiment 3 in the visual search study by Cooper et al. (2013): to control for the effect of distractor context. The context was unitary and it was held similar for both gaze direction change conditions. Because of this, no differences in attention-grabbing distractor context were possible between the change conditions. We expected, again, that detection efficacy would be greater for changes to direct gaze relative to changes to (laterally) averted gaze. Such a result would confirm that the differences between the conditions result from the nature of the changes themselves—for changes to direct gaze presumably from their particular relevance for attention. The setup for Experiment 2, including the instructions to participants for the experimental task, was kept as similar as possible to that of Experiment 1. Only the discrepant methodological details are described in the following section. Material and methods Participants Eighteen healthy individuals (9 males, mean age of all participants ± SD = 23.56 ± 2.22), participated in the study after having given written informed consent, and were compensated with movie tickets. The sample size for Experiment 2 was increased slightly because of smaller change magnitude relative to Experiment 1 (see Figures 1 and 3). None of the participants had participated in Experiment 1. The study conformed to the ethical standards of the American Psychological Association (APA) and The Code of Ethics of the World Medical Association (Declaration of Helsinki). According to Finnish regulations (Act on Medical Research and Decree on Medical Research 1999, amended 2010), specific ethics approval was not necessary for this study. Stimuli and procedure The stimuli in S1 were four schematic faces either all with an upward or all with a downcast gaze (see Figure 3). In S2, the gaze of one face shifted to a completely direct or to a slightly averted gaze. The magnitude of the displacement of the pupil from upward/ downward gaze to laterally (left or right) averted gaze and to direct gaze was held as similar as possible (about three millimetres on the computer screen for both change conditions, see Figure 3). A couple of amendments to the stimulus procedure were made relative to Experiment 1. As in Experiment 1, we aimed for presentation times during which all faces could not be searched through in S2, and for the detection levels for changed scenes to be as close to 50% as possible. The duration of S1 and S2 was set to 500 ms, and that of the blank screen to 250 ms. The task was also made shorter and less wearisome for the participants than in Experiment 1, and the participants completed as many trials as they managed in 20 minutes (M= 254.00 ± SD = 40.43). As in Experiment 1, the false alarm rate was low (1%), and the d’-values were high, M= 3.19, SD = .21, 95% CI [2.80, 3.60], indicating that the participants were performing the task as requested. 30 P. LYYRA ET AL. Data analysis The data were analysed as in Experiment 1 by conducting a repeated measures ANOVA with gaze direction (change to direct gaze, change to averted gaze) as a within-subject factor. To investigate the effect of background on change detection more closely, an additional factor of gaze prior to change (downcast gaze, upward gaze) was included. An alpha level of .05 was used in all the analyses. Partial eta squared ( h 2 p) represents the effect size estimates for ANOVA. Results and discussion Change detection performance The changes to direct gaze were detected more often than changes to averted gaze (M change to direct gaze = 54.50%, SD = 13.35; M change to averted gaze = 50.72%, SD =12.86,seeFigure 2), F(1, 17) = 4.53, p= .048, h 2 p=.21. Change detection was not influenced by gaze prior to change, F(1, 17) = 4.53, p= .048, h 2 p=.21. The interaction between the main effects was not significant. The main hypothesis of Experiment 2, that changes to direct gaze are detected more efficiently than changes to averted gaze, was supported by the results of Experiment 2, thus repeating the results of Experiment 1, even though the stimulus context was identical for both change type conditions. This finding seems to confirm that the detection advantage for direct gaze observed in Experiment 1 was probably not due to contextual effects. More likely, the particular relevance of direct gaze for attention underlay the detection advantage for changes to direct gaze in both experiments. It is also possible that the detection of direct gaze in this paradigm resembles that of the visual search paradigm, as the changed face is the sole deviant in the S2 displays. Even though the observers could not have been browsing through all the stimuli in the array of S2 because of short presentation times, some kind of deviance detection at the implicit level may have contributed to explicit change detection. To counteract even a theoretical possibility of this, the changed face should not be the sole deviant face in the S2. This way the detection would be based solely on change and not on deviance detection. To assess this, a third experiment was devised. The results from Experiments 1 and 2 also do not completely exclude the possibility that the distractor context could exert some effect on gaze direction change detection rates. In both experiments, the context was unitary, and all of the faces present in a single view had equal attention-grabbing potential. An experiment with a non-unitary context is required to settle whether context can also exert some influence on change detection efficacy. For example, if the crowd consisted of both direct and averted gaze distractors, attention could be grabbed more efficiently by faces with direct gaze than by faces with averted gaze, thus facilitating change detection in direct gaze faces (i.e., changes to averted gaze). Experiment 3 In Experiment 3, the distractor crowd consisted of faces with both direct and averted gaze. As in Experiment 2, the context (S1) in Experiment 3 was held similar for both change conditions. However, the context was not unitary as in Experiment 2, but consisted of two faces with direct gaze and two with laterally averted gaze. Change conditions were similar to those in the previous two experiments, averted gaze changing to a direct gaze (i.e., change to direct gaze) or direct gaze changing to averted gaze (i.e., change to averted gaze, see Figure 4). This kind of a setup can only reflect change detection and not deviance detection as in the visual search paradigm. In S2, the changed face is similar to two other faces while the sole deviant face in the scene remains unchanged relative to S1. With this setup, we expected that the direct gaze advantage would still hold. It was also ensured that the participants were not able to browse through all the faces during the presentation of S2. We designed the stimulus setup so that only two or three faces could be searched through in S2, but the detection levels for changed scenes would be close to 50%. Beyond this the setup for Experiment 3, including the instructions for the experimental task, was kept as similar as possible to those in Experiments 1 and 2. Material and methods Participants Sixteen healthy individuals (8 males, mean age of all participants ± SD = 27.81 ± 6.25), participated in the study and were compensated with movie tickets. This sample was independent of those in Experiments VISUAL COGNITION 31 1 and 2. The study conformed to the ethical standards of the American Psychological Association (APA) and The Code of Ethics of the World Medical Association (Declaration of Helsinki). According to Finnish regulations (Act on Medical Research and Decree on Medical Research 1999, amended 2010), specific ethics approval was not necessary for this study. Stimuli and procedure The stimulus paradigm was similar to those in Experiments 1 and 2. As in Experiment 2, the backgrounds of the change were kept identical between the change conditions. S1 consisted of four schematic faces, two with direct gaze and two with laterally averted gaze (left or right). The locations of the four faces were randomly assigned to the four locations in the stimulus matrix (see Figure 4). In S2, the gaze of one face with a direct gaze shifted to right or left, or one with laterally averted gaze shifted to direct gaze. The duration of S1 and S2 was 750 ms and that of the blank screen was 250 ms. Participants completed as many trials as could manage in 20 minutes (M= 223.32 ± SD = 43.38). The false alarm rate remained very low also for this experiment (below 2%). Accordingly, the d’-values were high, M= 2.37, SD = .24, 95% CI [2.27, 3.17], showing that the participants were performing the task successfully and as instructed also in Experiment 3. Data analysis The data were analysed as in Experiments 1 and 2 by conducting a repeated measures ANOVA with gaze direction (change to direct gaze, change to averted gaze) as a within-subject factor. An alpha level of .05 was maintained in all analyses. Partial eta squared ( h 2 p) represents effect size estimates for ANOVA. Results and discussion Again, changes from averted to direct gaze were detected more efficiently compared to those from direct to averted gaze (M change to direct gaze = 60.42%, SD = 13.79; M change to averted gaze = 52.26%, SD = 17.06, see Figure 2), F(1, 15) = 9.13, p= .008, h 2 p=.381. The results showed a similar detection advantage for changes to direct gaze relative to changes to averted gaze as those in the previous two experiments. The direct gaze detection advantage observed here indicates that the change blindness paradigm reflects change detection and not deviance detection in S2. In terms of effect size, the detection difference between the gaze directions remained at the same level throughout all experiments despite distractor context differences (Experiment 1: h 2 p=.33; Experiment 2: h 2 p=.21; Experiment 3: h 2 p=.38). This seems to suggest that the distractor context does not interfere with change detection, or with the attentional bias created by the socially relevant changes to direct gaze. An additional ANOVA with a between-subjects factor of experiment (Experiment 1, Experiment 2, Experiment 3) and a within subjects factor of gaze direction (direct, averted) showed a main effect of gaze direction (M change to direct gaze = 50.28%, SD = .18, M change to averted gaze = 44.02%, SD = .18), F(1, 47) = 20.98, p< .001, h 2 p=.309. Figure 4. Stimulus setups for the change to direct gaze and change to averted gaze conditions in Experiment 3. Changes in S2s are indicated with dotted circles. 32 P. LYYRA ET AL.