Anxiety, conscious awareness and change detection

Abstract

Attentional scanning was studied in anxious and non-anxious participants, using a modified change detection paradigm. Participants detected changes in pairs of emotional scenes separated by two task irrelevant slides, which contained an emotionally valenced scene (the ‘distractor scene’) and a visual mask. In agreement with attentional control theory, change detection latencies were slower overall for anxious participants. Change detection in anxious, but not non-anxious, participants was influenced by the emotional valence and exposure duration of distractor scenes. When negative distractor scenes were presented at subliminal exposure durations, anxious participants detected changes more rapidly than when supraliminal negative scenes or subliminal positive scenes were presented. We propose that for anxious participants, subliminal presentation of emotionally negative distractor scenes stimulated attention into a dynamic state in the absence of attentional engagement. Presentation of the same scenes at longer exposure times was accompanied by conscious awareness, attentional engagement, and slower change detection.

Introduction

It is well documented that the attentional behaviour of highly anxious individuals is unusually sensitive to emotionally negative or threatening information in the visual environment. Findings from several paradigms converge on the conclusion that anxious individuals show an attentional bias for emotionally negative, potentially threatening information, and are more distracted when such information is present in the visual environment (Bar-Haim et al., 2007, Cisler et al., 2009, Eysenck et al., 2007). However, one issue that remains unresolved is how these well documented biases might influence the way in which anxious individuals scan the visual environment (Cisler et al., 2009), and even whether such biases have any consequences at all for visual scanning in anxiety (Freeman, Garety, & Phillips, 2000). In the current study we addressed this issue by making use of a widely used technique for studying visual scanning, the change detection paradigm. Although this paradigm has been applied productively to a variety of theoretical issues in studies of attention and perception (Rensink, 2002, Simons and Rensink, 2005) to our knowledge it has not been recruited in previous research to investigate attentional scanning in anxiety. Here we made use of a change detection paradigm to study the efficiency of attentional scanning in anxious and non-anxious individuals in different task contexts. More specifically, we investigated the efficiency with which anxious and non-anxious participants detected changes in pairs of scenes when these were accompanied by task irrelevant, emotionally valenced information. By using a pattern masking procedure together with a variable exposure time we were able to manipulate whether this emotionally valenced, but task irrelevant information was available or unavailable to conscious awareness.

Bar-Haim et al. (2007) reported a meta-analysis of 172 studies which examined attentional behaviour in anxious and non-anxious individuals. They noted that three experimental paradigms (emotional Stroop task, dot-probe task, and emotional spatial cueing task) account for the bulk of published evidence on this issue. Evidence from all three paradigms supports the conclusion that in anxious individuals, visual attention shows a bias towards threat-related information, while non-anxious individuals display no such bias. In addition, these authors noted that meta-analysis of findings from a paradigm that assesses spatial attention, the dot probe task, indicated that subliminal presentation of emotional stimuli was associated with attentional effects that were almost twice as large as those elicited by supraliminal exposures.

In addition to the three paradigms reviewed by Bar-Haim et al. (2007) a fourth technique, the visual search task, has also been used in studies of attention and anxiety. In a typical visual search task participants attempt to detect a target item in a display containing several objects, and the time taken to either detect the target, or verify its absence is measured. Evidence from visual search is directly relevant to the issue addressed in the current study, because like the change detection task, it provides an assessment of attentional scanning. In a study by Byrne and Eysenck (1995) participants searched through arrays of twelve photographed faces. When participants were searching for an angry face among distractors bearing a neutral expression, anxious individuals detected the target more rapidly than the non-anxious control group. In contrast, when participants searched for a happy face among distractors bearing angry expressions, anxious individuals detected the target more slowly than the control group. These findings were interpreted as indicating that anxious individuals display an attentional bias towards potentially threatening stimuli, such as angry faces, and that this bias is manifest as improved performance when an angry face is the target, but impaired performance when angry faces serve as distractors. Thus attentional scanning in anxiety can be facilitated when searching for a threat-related target, but the same bias increases susceptibility to distraction when searching for a non-threat target among threatening distractors. Eysenck et al. (2007) suggest that both these effects derive from increased reliance on stimulus driven, bottom-up attentional control, at the expense of goal driven top-down attentional control (Corbetta & Shulman, 2002). According to this view, threat-related information captures attention, in a bottom-up fashion, to a stronger degree in anxious compared to non-anxious individuals. This would thus lead to improved search for threat-related targets, because the targets of goal-directed and stimulus driven attention coincide. However, performance would be worse in the presence of threat-related distractors as stimulus driven attentional capture distracts attention away from the target, and impairs goal driven search. Cisler et al. (2009) review the outcome of visual search studies of attention in anxiety, and conclude that, despite some inconsistencies, the pattern of results reported by Byrne and Eysenck (1995) has been replicated in more recent research.

In contrast to the findings of Byrne and Eysenck (1995), recent research reported by Phelps, Ling, and Carrasco (2006) and Becker (2009) suggests that the presence of emotionally threatening information in the environment can improve visual processing and facilitate visual search. In the former study (Phelps et al., 2006), observers discriminated the orientation of sinusoidal gratings which were preceded by a briefly presented (75 ms) face bearing either a neutral or a fearful expression. Contrast sensitivity was improved when gratings were preceded by fearful faces, when compared with the neutral face condition. In the study of Becker (2009) observers performed a visual search task, in which both the target (a house) and distractor objects were emotionally non-threatening. Search arrays were preceded by a face bearing a fearful, neutral or happy expression. Search times were reliably faster when preceded by a fearful face. Although this evidence suggests that the presence of emotionally threatening information can facilitate perceptual processing and the efficiency of visual search, anxiety levels were not assessed in either study (Becker, 2009, Phelps et al., 2006). Thus, the question of how these effects might be moderated, or perhaps even reversed as a function of participant anxiety remains unknown.

However, the visual search task is not the only technique available to experimental psychologists for studying attentional behaviour. As noted earlier, the change detection task has been widely used in studies of perception and attention (Rensink, 2002), but has yet to be applied to the question of how attentional behaviour is affected by anxiety. The current study sought to remedy this by using a well documented form of the change detection paradigm, the flicker task (Rensink, O’Regan, & Clark, 1997) to study attentional scanning in anxiety. In the flicker task participants are presented with two versions of a visual scene, which are shown successively, separated by an interposed slide, which often comprises a uniform grey or white screen. The two versions of the scene, together with intervening slides are presented, in cyclical fashion, until observers detect a difference between them. A well known finding from studies using this task is that observers often fail to detect surprisingly large differences between the two scenes and display change blindness (Simons & Ambinder, 2005). If the intervening blank slides are removed from the presentation cycle, change detection becomes trivially easy. In this situation, a difference between the two scenes produces a local visual transient when they alternate, and this captures attention in a bottom-up fashion, leading to rapid detection of the change. In contrast, when the scenes alternate with blank or ‘mud-splash’ slides, massive transients occur across the entire scene so, attention is not drawn selectively to the location of the change. A central finding from research using this and similar techniques has been that there is a tight link between change detection and attention: observers need to pay attention to the appropriate location for changes to be detected (Rensink et al., 1997, Simons and Rensink, 2005). In a sense change detection represents a special case of the visual search task, where the target – a change – remains undefined with respect to any specific visual (e.g. size, shape, colour) or categorical (e.g. facial expression) attribute.

In the study reported here we adapted the flicker task (Rensink, O’Regan, & Clark, 1997), to investigate how attentional scanning might be affected by the presence of emotionally valenced information. The design of our modified task is illustrated in Fig. 1 below. As this figure shows, rather than interposing a single blank screen in between the original and changed scene, two further screens were interposed. The first of these contained an additional scene that we refer to below as the ‘distractor scene’. Participants were presented with distractor scenes that were either emotionally negative in nature (e.g. war scene, traffic accident, mutilation), or emotionally positive (e.g. romantic scene, attractive natural scene, smiling baby; see Appendix A for a full list). The second screen contained a pattern mask. In the short duration – subliminal condition, the distractor screen was presented for 33 ms, and the mask was presented for 197 ms, while in the long duration – supraliminal condition the distractor scene was presented for 150 ms and the mask was presented for 80 ms. The aim of this was to present the distractor scenes at durations whereby the content of the scene was either available to conscious awareness (long duration), or unavailable to conscious awareness (short duration). Previous work has shown that when the delay between onset of a visual stimulus and onset of a pattern mask (stimulus onset asynchrony – SOA) is very brief, access of the first stimulus to conscious awareness is limited or absent. In their wide ranging review of this literature, Breitmeyer and Ogmen (2006) conclude that an SOA of between 30 ms and 100 ms “produces a total or nearly total suppression of the perception of the target’s contrast, colour, and contour” (Breitmeyer & Ogmen, 2006, p. 38). By using an SOA of 33 ms in the short duration – subliminal condition we aimed to ensure that the content of the vast majority of distractor scenes was unavailable to conscious awareness. Conversely, in the long duration condition (SOA 150 ms) the content of distractor scenes was available to conscious awareness. As a check on the effectiveness of this manipulation, participants reported verbally their perception of the distractor scene on each trial (see Section 2.5). An additional aim of the study was to discover whether effects of distractor scene valence and exposure time on change detection latency would be modulated by the emotional context in which the distractor scenes were presented. Thus, the emotional valence of the scenes that changed were also either negative or positive (see Appendix B for full list). We refer to these scenes below as the ‘change scenes’.

According to attentional control theory (Derakshan and Eysenck, 2009, Eysenck et al., 2007) attention in anxiety is characterised by an altered balance of attentional control, in which the allocation of attention tends to be dominated by bottom-up, stimulus driven attention, at the expense of top-down, goal directed attention. A central prediction of this theory is that in anxiety the ability to control goal-directed movements of attention will be compromised in situations where there is scope for attention to be captured by task irrelevant stimuli – especially when the task irrelevant stimuli are threat-related or perceptually salient. In such situations anxious individuals will show greater distractibility than non-anxious controls. Attentional control theory contends that the latter group maintain top-down control over attention more effectively, enabling them to avoid the distracting effects of task irrelevant stimuli. Thus, attentional control theory predicts that overall change detection latencies will be slower in anxious compared to non-anxious participants. This is because during change detection participants must perform a goal-directed search of a visually complex scene, directing attention in a systematic way to multiple candidate locations. For this to be effective, attention must not be monopolised by perceptually salient features or by emotionally salient objects in the scene, but needs to be directed to a variety of less salient features and relatively marginal scene objects. The changes that we implemented in the change scenes generally involved alterations that were low in perceptual salience, and involved objects or features that were of marginal rather than central interest in the scene (Rensink, O’Regan, & Clark, 1997). According to attentional control theory, anxious participants will perform this task less efficiently, because for these participants goal-directed scanning of attention will tend to be ‘hijacked’ by bottom-up attentional capture, directing attention to visually salient features or objects with emotional significance. Furthermore, the degree of impairment in goal-directed attentional scanning experienced by anxious participants should be especially marked in emotionally negative visual contexts. Thus, change detection in anxiety should be especially slow when participants scan scenes that are emotionally negative and when an emotionally negative distractor scene is interposed between the two versions of the scene. In both cases attentional control theory predicts that for anxious individuals, features of the emotionally negative scenes will be more likely to capture attention and disrupt goal-directed scanning. Moreover, the meta-analysis of Bar-Haim et al. (2007), indicating increased attentional effects with subliminally presented emotional stimuli, leads to the further prediction that the disruptive effect of interposed distractor scenes should be stronger at brief – subliminal exposure times, which prevent access of the scene to conscious awareness.

On the other hand, the recent work of Phelps et al. (2006) and Becker (2009) suggests that a very different pattern of results could be observed. This work, described above, showed that task irrelevant, emotionally negative information can facilitate perceptual processing. Although Becker’s (2009) finding with visual search might be expected to generalise to the change detection task employed here, as noted earlier, it is currently unknown whether this effect is moderated, or perhaps even reversed as a function of participant anxiety.