Attentional control as a moderator of the relationship between posttraumatic stress symptoms and attentional threat bias

Abstract

Attentional threat bias (ATB) has been suggested as one factor leading to maintenance and exacerbation of posttraumatic stress symptoms (PTSS). In the present study, attentional processes (i.e., facilitated engagement, difficulty disengaging) underlying the association between ATB and PTSS were examined. Additionally, attentional control (AC) was examined as a moderator of this relationship. Participants (N = 97) completed a dot-probe task with two levels of stimulus-onset asynchrony (SOA: 150 and 500 ms). Higher PTSS were associated with ATB when SOA was longer (i.e., 500 ms), suggesting difficulty disengaging from threat stimuli. AC moderated the relationship between PTSS and ATB when SOA was shorter (i.e., 150 ms), with participants high in PTSS and high in AC having disengaged and shifted attention from threat stimuli using top-down AC when the emotional valence of threat stimuli was less salient (i.e., shorter presentation duration). Findings implicate AC as a buffering mechanism against prolonged attentional engagement with threat-related stimuli among those with high PTSS. Current PTSD interventions may benefit from incorporating attention-based components.

Introduction

Hypervigilance toward threatening information, a symptom of posttraumatic stress disorder (PTSD; Diagnostic and Statistical Manual for Psychiatric Disorders-Fourth Edition-Text Revision [DSM-IV-TR]; American Psychiatric Association [APA], 2000), has been implicated as one factor leading to the maintenance and exacerbation of PTSD symptoms (e.g., intrusive thoughts, flashbacks, physiological reactivity, avoidance behaviors, heightened arousal; Constans, 2005, Elzinga and Bremner, 2002). Therefore, among trauma and PTSD researchers, there has been an emphasis on understanding the attentional processing of threat- and trauma-related information among those with PTSD, with the idea that facilitated threat detection (i.e., hypervigilance) is synonymous with PTSD-related attentional threat biases.

Consistent with this conceptualization, a plethora of published journal articles has reported that individuals with PTSD exhibit a bias for attending to trauma-specific stimuli (e.g., Beck et al., 2001, Bryant and Harvey, 1997, McNally et al., 1996, McNally et al., 1990) and to a lesser degree, general threat stimuli (e.g., Litz et al., 1996). Thus, there has been a general consensus among researchers that an attentional bias to threat information in PTSD is a phenomenon with an overabundance of support in the extant literature (Buckley et al., 2003, Constans, 2005, McNally, 1998). However, a recent examination of peer-reviewed literature and dissertation abstracts has called robustness of this phenomenon into question (Kimble, Frueh, & Marks, 2009).

Attentional threat bias is most often conceptualized as being reflexive, automatic, and occurring outside of conscious awareness (Constans, 2005, Yiend, 2010). The modified Stroop task has been used as the primary experimental task to examine attentional biases in PTSD (Constans, 2005). In the traditional Stroop task, participants are asked to name the colors in which words are printed as quickly as possible while ignoring the meaning of the word, a task which becomes more difficult when the color of the ink differs from the meaning of the word (e.g., the word “green” written in red ink). Similarly, in the modified Stroop task, participants are presented with a series of one-word stimuli, each from a specific word category (e.g., trauma, positive, neutral, etc.), and asked to name the color of the word while paying no attention to the word's meaning. Slower responding to specific word stimuli is thought to occur when attention is briefly captured by the potency of word meaning, thus disrupting the processing of color information.

Kimble et al. (2009) conducted a review of peer-reviewed journal articles and dissertation abstracts for studies using the modified Stroop task to examine attentional bias for threat information in individuals with PTSD versus individuals without PTSD; both groups had experienced a traumatic event. Contrary to predictions, only 8% of dissertation abstracts and 44% of peer-reviewed journal articles showed slowed responding to threat words among individuals with PTSD. Kimble et al. (2009) suggest that a lack of empirical support for the modified Stroop effect in PTSD may indicate absence of attentional threat bias in PTSD. However, considerable debate exists about the nature of the modified Stroop task, with some suggesting that the task is inappropriate for use in determining information-processing biases (McKenna and Sharma, 2004, Phaf and Kan, 2007, Weierich et al., 2008).

There are two primary presentation formats of the modified Stroop task: (a) a blocked format which provides a measure of the combined effect of the slow and fast components of attention and (b) a random format which provides a measure of the fast component of attention (McKenna & Sharma, 2004). Contrary to the expectation of automaticity that the traditional Stroop task is founded on, McKenna and Sharma (2004) found no evidence for fast, automatic effects in the modified Stroop task. Furthermore, in a recent meta-analysis of 172 studies that used the modified Stroop task to examine threat-related attentional biases in anxiety, Bar-Haim, Lamy, Pergamin, Bakermans-Kranenburg, and van IJzandoorn (2007) found evidence of the modified Stroop effect, but only when a blocked presentation format was used. Thus, observed effects are likely the result of slower, higher order regulatory processing; therefore, the modified Stroop task, which does not provide an independent measure of these effects, may not be an appropriate experimental task for examining the processing of threat information in PTSD. The findings of Kimble et al. (2009) are important in advancing research in the area of information processing in PTSD. However, instead of providing contradictory evidence of an association between PTSD and attentional threat bias, Kimble et al.’s (2009) findings suggest that it is necessary to examine attentional threat bias in PTSD using more appropriate experimental tasks.

The dot-probe task is another stimulus-response task used to examine the association between attentional threat bias and anxiety (Salemink, van den Hout, & Kindt, 2007). In the traditional dot-probe task, two stimuli are presented side by side on a computer screen. The stimuli remain on the screen for a specified duration of time, after which a dot appears on the screen, replacing one of the two pictures. The participant presses a button that corresponds to the relative position of the dot on the screen, thus providing the researcher with a snapshot of the participant's attention allocation at that point and time. If the participant has a bias for attending to threat stimuli, s/he should respond faster when the dot appears in the spatial postion previously held by the threat stimuli. Additionally, if the time interval of stimulus presentation varies (e.g., 150 milliseconds [ms], 500 ms), the dot-probe task can provide a temporal examination of attention allocation. However, the majority of research incorporating the dot-probe task has used only one stimulus duration (i.e., 500 ms; Yiend, 2010), a duration of time which likely allows the participant to make multiple shifts in attention.

The only published study related to attentional bias in PTSD to incorporate a dot-probe task found that participants with PTSD showed a bias for attending to mild threat words when compared to participants with subclinical PTSD and participants with low anxiety (Bryant & Harvey, 1997). Interestingly, no attentional bias was observed for strong threat words. These findings may be the consequence of (a) using a modified version of the dot-probe task in which stimulus words were presented simultaneously with a word indicating the direction of the button that the participant was instructed to press (i.e., “left” or “right”) and (b) no variation in stimulus-onset asynchrony (SOA; i.e., the duration of time between the initial stimulus presentation and response option). As previously mentioned, the dot-probe task can provide a temporal examination of attention allocation when multiple stimulus presentation intervals are used, and thus, would be helpful in identifying specific attentional processes (i.e., fast [bottom-up], slow [top-down]) that may account for information-processing biases in PTSD.

The vigilance-avoidance model of anxiety-associated attentional threat bias presupposes facilitated threat engagement, or orienting of attention toward threat stimuli, which is followed by the subsequent avoidance of such stimuli (Weierich et al., 2008). In contrast, the attention-maintenance model is not based on the premise that there is faster orienting toward threat stimuli; instead, once threat stimuli are attended to, it is more difficult to disengage from such stimuli, especially at higher levels of anxiety (Weierich et al., 2008).

Related theories of attention (i.e., Corbetta and Shulman, 2002, Eysenck et al., 2007, Metcalfe and Mischel, 1999) describe attentional threat bias in anxiety in terms of a bottom-up, sensory driven attentional system and a top-down attentional control (AC) system. AC has been described as one's ability to use higher level executive functioning to regulate, or override, automatic emotional responses (Derryberry & Reed, 2002). In Metcalfe and Mischel's (1999) conceptualization, the “hot” system (bottom-up) is specialized for immediate responding and emotional processing, whereas the “cool” system (top-down) is specialized for reflective emotion regulation and control of impulsive tendencies.

The balance between these two attentional systems is thought to be disrupted when high levels of stress/anxiety impair the cool system and potentiate the hot system (Eysenck et al., 2007, Metcalfe and Mischel, 1999). The top-down attentional system, controlled by the central executive, is seen as being responsible for three main functions of AC: (a) the inhibition of dominant, automatic responses, (b) shifting back and forth between multiple task demands, and (c) updating working memory. Research has consistently shown that anxiety impairs two of these three functions: the inhibition of dominant, automatic responses and shifting back and forth between multiple task demands (Graydon and Eysenck, 1989, Lavie et al., 2004). Therefore, individuals in a heightened state of anxiety will be less successful at tasks in which the executive attentional mechanisms of inhibition and shifting are needed. Furthermore, when task-irrelevant stimuli increase participant distress, these stimuli will be attended to for a greater length of time as a result of anxiety's detrimental influence on inhibition and switching functions. However, at low levels of stress, the cool system allows for the inhibition of hot system dominant response tendencies (Metcalfe & Mischel, 1999). Additionally, Metcalfe and Mischel (1999) proposed that hot system activity is more likely to trigger cool system activation in individuals with better developed, more complex cool systems, with cool control eventually becoming an almost automatic response to hot system activation. Thus, individuals with high levels of PTSS who report experiencing high levels of anxiety and persistent arousal (i.e., hot system activation) may have greater difficulty disengaging from threat-relevant information due to impairment of the cool system (i.e., AC).

Research has provided support for the hypothesis that anxious individuals are slower to disengage attention from threat stimuli. Amir, Elias, Klumpp, and Przeworski (2003) found that individuals with social phobia had difficulty disengaging attention from social threat words (e.g., embarrassed, humiliated) when compared with a control group. Fox, Russo, and Dutton (2002) found that individuals with high trait anxiety, in comparison to individuals with low trait anxiety, had greater difficulty disengaging attention from emotional faces (angry and happy), but not neutral faces. Derryberry and Reed (2002) used a spatial cueing task to examine attention to threat cues at high and low levels of trait anxiety. Additionally, participants completed a self-report measure of AC (i.e., the Attentional Control Scale [ACS]; described below). As expected, individuals high in trait anxiety were slower to disengage attention from threat cues. However, individuals high in trait anxiety and high in AC showed significantly faster disengagement from threatening cues than participants high in trait anxiety and low in AC. In addition to supporting a delayed disengagement model of attentional threat bias, these findings provide evidence that higher order regulatory processes can attenuate automatic responding at a relatively early stage of information processing; and furthermore, after accounting for the detrimental effects of anxiety on attention, AC can still benefit individuals with higher levels of anxiety.

In an examination of more distal effects of AC, Bardeen and Read (2010) examined the degree to which AC predicted recovery from trauma re-telling negative mood. Trauma exposed participants provided a baseline measure of negative affect and AC (i.e., ACS), gave a first person account of their most distressing traumatic event, and then completed two subsequent post-trauma re-telling measures of state affect. Trauma-exposed individuals with high AC recovered faster from trauma re-telling induced negative mood than individuals low in AC, suggesting that distress is reduced by disengaging and shifting attention from trauma information (Bardeen & Read, 2010). Furthermore, mild distraction may help to expedite fear reduction during exposure therapy (Johnstone & Page, 2004). Therefore, AC may aid individuals with PTSD during exposure therapy by decreasing their level of fear during treatment, thus making treatment tolerable and increasing the likelihood that the client will continue attending exposure sessions rather than exhibiting escape and avoidance behaviors. Consistent with this view, Foa and Kozak (1986) hypothesized that emotional processing, a necessary component of successful exposure therapy for PTSD, is unlikely when a client's level of fear is too high. Distraction may help to bring one's level of fear into a range in which emotional processing can occur.

To our knowledge, only two published studies have examined the specific attentional processes (i.e., facilitated engagement, difficulty disengaging) underlying attentional biases associated with PTSS. Pineles, Shipherd, Welch, and Yovel (2007) assigned Vietnam-era veterans (N = 57) into high- and low-PTSD groups based on self-reported PTSS and had all participants complete a visual search task with the goal of identifying an “odd-ball,” or discrepant target among a group of identical stimuli. Participants high in PTSS, compared to those low in PTSS, had difficulty disengaging from trauma stimuli, as represented by slower identification of neutral word targets embedded in a group of Vietnam-related threat words in comparison to trials in which a neutral word target was embedded in a group of neutral words (i.e., “interference condition”). No evidence of facilitated threat engagement was found (i.e., “facilitation condition”: faster identification of Vietnam-related threat word targets embedded in a group of neutral words). Pineles, Shipherd, Mostoufi, Abramovitz, and Yovel (2009) replicated these findings in a group of sexual assault survivors (N = 46), finding evidence of difficulty disengaging from sexual assault-related threat words (e.g., scream, struggle) among participants with high self-reported PTSS.

These findings are the first to identify difficulty disengaging from threat stimuli as the attentional process underlying attentional biases in those with higher PTSS and they provide evidence contradicting the vigilance-avoidance model, with neither study finding facilitated threat engagement among those with higher PTSS. Interestingly, task order effects were shown in both studies; that is, among participants in the high-PTSD group, only those who completed the interference condition first, followed by the facilitation condition, showed evidence of difficulty disengaging from trauma stimuli relative to those in the low-PTSD group. As discussed by Pineles et al. (2009), difficulty disengaging from trauma stimuli may not have been observed among those high in PTSS who completed the interference condition second, because participants had already been exposed to the trauma words many times in the first condition. In other words, a process of habituation may have occurred, with trauma stimuli losing their potency over many trials. However, the use of word stimuli, which require greater semantic processing (Pineles et al., 2009), and are more prone to subjective familiarity and frequency of use than pictorial stimuli (Bradley et al., 1997), may have increased error variance, thus obscuring potentially important effects.

In addition, the relatively small sample size in both studies (i.e., Study 1: N = 57, Study 2: N = 46) suggests the possibility of insufficient power to detect effects. For example, Pineles et al. (2009) found no association between group status and a bias for attending to general threat-related words, however, it may be the case that there was insufficient power to detect an association between general threat stimuli and attentional disengagement among individuals with higher PTSS. Furthermore, due to the small sample size, Pineles et al. (2009) did not examine the extent to which the association between PTSS and difficulty disengaging from trauma stimuli was accounted for by state anxiety. Building on the work of Pineles and colleagues, future research examining associations between attentional threat bias and PTSS would benefit from (a) the use of pictorial stimuli, which are thought to induce emotional valence at a more visceral level than word stimuli (Bradley et al., 1997), (b) increasing the sample size in order to account for state levels of anxiety in analysis, (c) the use of an experimental paradigm which can dileneate the temporal sequence of attentional processes, and (d) examining potential associations between AC, attentional disengagement difficulties, and PTSS.

The fear network model of PTSD (Foa & Kozak, 1986) explains the automaticity of threat processing in PTSD by describing a network of interconnected trauma memory nodes which are hypothesized to be linked by a process of spreading activation. The triggering of one node in the network, presumably by threat-relevant stimuli in one's environment, causes spreading activation across the network and a subsequent triggering of all other nodes, thus resulting in a response consistent with the physiological symptoms in PTSD (e.g., reexperiencing, hyperarousal).

The fear network model, in its present form, suggests that exposure therapy is effective in treating PTSD because pathological associations in the fear network are weakened when exposure to threat stimuli is coupled with disconfirmation of feared outcomes, a process of unlearning (Foa, Huppert, & Cahill, 2006). However, recent evidence has provided a different conceptualization of the underlying mechanisms in exposure therapy. This evidence suggests that extinction is not a matter of unlearning, but rather new learning; that is, extinction does not occur through a disintegration of associations in the network, but by establishing new associations that inhibit pathological responses to conditioned stimuli (Bouton, 2002, McNally, 2007, Myers and Davis, 2002). Because executive attentional mechanisms have been shown to inhibit dominant reflexive responses (e.g., eye blink responses to startle probes; Gyurak & Ayduk, 2007), the effects of attentional training on one's ability to use higher order executive attention should presumably aid in the extinction process, thus reducing PTSS.

The fear network model has provided us with an important conceptualization of bottom-up threat processing in PTSD. However, to maintain consistency with current research, which has shown that top-down AC can attenuate bottom-up attentional processing relatively quickly, and to maintain consistency with other empirically supported models (i.e., Corbetta and Shulman, 2002, Eysenck et al., 2007, Metcalfe and Mischel, 1999), fear-network models of information processing would be well served by incorporating higher-level executive processes into an overarching model.

As previously mentioned, mild distraction may help to expedite fear reduction during exposure to threat stimuli (Johnstone & Page, 2004). Therefore, the ability to intentionally disengage from threat-related stimuli may help to attenuate distress, thereby allowing one to remain in and learn from the environment rather than feeling the need to escape from an environment where the threat remains salient. Consistent with this proposition, shifting attention to safe or novel stimuli can help to reduce negative affect (Harman et al., 1997, Nolen-Hoeksema and Morrow, 1993). On the other hand, it could be argued that shifting attention away from threatening information is maladaptive (i.e., experiential avoidance), eventually leading to problematic avoidance behaviors (e.g., social phobia, agoraphobia); however, research to date suggests that this form of attentional avoidance is adaptive (e.g., Eisenberg, Fabes, Guthrie, & Reiser, 2000).

Attention regulation capabilities have been shown to be protective against negative emotionality and externalizing behaviors and positively associated with social adaptation (Eisenberg et al., 2000). In college students, better attentional disengagement and shifting abilities, measured via a covert attentional orienting task, predicted less negative affect after a laboratory induction of negative mood (Compton, 2000) and self-reported AC, measured using the ACS, has been shown to predict less negative affect following trauma re-telling (Bardeen & Read, 2010). AC, measured via the ACS, is positively associated with indices of positive emotionality, such as extraversion, and inversely related to aspects of negative emotionality, such as trait anxiety (Derryberry & Reed, 2002).

In light of the role of AC in the experience and modulation of negative affect, it stands to reason that AC may be an important individual difference factor in post-trauma adaptation and the maintenance of PTSS. Therefore, among individuals with PTSD, an inability to disengage attention from threat-relevant information may help to maintain and prolong the rumination and reexperiencing seen in PTSD; better AC abilities may facilitate recovery among individuals with PTSD.