اضطراب سلامتی و تعصب توجه: دوره زمان بیداری و عدم مقابله در پرتو اطلاعات بیماری های تصویری
|کد مقاله||سال انتشار||مقاله انگلیسی||ترجمه فارسی||تعداد کلمات|
|35386||2011||7 صفحه PDF||سفارش دهید||محاسبه نشده|
Publisher : Elsevier - Science Direct (الزویر - ساینس دایرکت)
Journal : Journal of Anxiety Disorders, Volume 25, Issue 8, December 2011, Pages 1131–1138
Cognitive-behavioral models of health anxiety stress the importance of selective attention not only towards internal but also towards external health threat related stimuli. Yet, little is known about the time course of this attentional bias. The current study investigates threat related attentional bias in participants with varying degrees of health anxiety. Attentional bias was assessed using a visual dot-probe task with health-threat and neutral pictures at two exposure durations, 175 ms and 500 ms. A baseline condition was added to the dot-probe task to dissociate indices of vigilance towards threat and difficulties to disengage from threat. Substantial positive correlations of health anxiety, anxiety sensitivity, and absorption with difficulties to disengage from threat were detected at 500 ms exposure time. At an early stage (i.e., at 175 ms exposure time), we found significant positive correlations of health anxiety and absorption with orientation towards threat. Results suggest a vigilance avoidance pattern of selective attention associated with pictorial illness related stimuli in health anxiety.
From a cognitive perspective (Beck, 1976), health anxiety occurs because a particular internal stimulus (i.e., a body sensation) is misjudged to be a threat to health (i.e., a symptom of illness) and one's ability to prevent the feared illness is perceived as insufficient (Warwick, 1989). A cognitive-behavioral model of hypochondriasis and health anxiety that helps to explain how hypochondriasis develops and is maintained was proposed by Salkovskis and Warwick (2001). In this model, not only internal stimuli (i.e., bodily sensations) but also external stimuli (e.g., images, information about illness) act as triggering events in a vicious circle of body sensations, their catastrophic interpretation, and affective, attentional, behavioral, and physiological consequences that foster the detection of more body sensations (Warwick, 1989, p. 708). Additionally, not only external but also internal intrusive images seem to play an important role in health anxiety as Muse, McManus, Hackmann, Williams, and Williams (2010) and Wells and Hackmann (1993) concluded in studies with patients suffering from hypochondriasis. In this regard, Muse et al. (2010) reported that 72% of the patients with hypochondriasis experienced recurrent, intrusive images that were either a memory of an earlier situation or strongly related to such a memory. Regarding the distribution of health anxiety, the model of Warwick and Salkovskis (1990) assumes that health anxiety represents a continuous construct ranging from mild and transient symptoms to full-blown hypochondriasis. This continuum hypothesis of health anxiety has recently also been empirically confirmed by two independent taxometric analyses (Ferguson, 2008 and Longley et al., 2010). The model of Warwick and Salkovskis (1990) is very similar to other models of hypochondriasis and health anxiety, e.g., the conceptual hypochondriasis model by Abramowitz, Schwartz, and Whiteside (2002) and a cognitive development model by Williams (2004). All of these models agree on the assumption that an attentional bias towards both, internal and external health threat stimuli is one key aspect in the development and maintenance of health anxiety and hypochondriasis. It is therefore reasonable to assume that the attentional system of health anxious people is distinctively sensitive to and biased in favor of health threat-related stimuli (for a review of cognitive abnormalities in health anxiety see Marcus, Gurley, Marchi, & Bauer, 2007). Somatosensory amplification (Barsky, Wyshak, & Klerman, 1990) is an example for how symptom focused attention plays an important role in health anxiety. Somatosensory amplification refers to the tendency to experience weak somatic sensation as unusually intense and involves bodily hypervigilance to those sensations which are often interpreted as signs of a severe illness. This attentional bias operates as a confirmatory filter: i.e., it exaggerates sensations that confirm the hypothesis of an illness and suppresses contradictory sensory input (Barsky, 2001). 1.1. Attentional biases in general anxiety and health anxiety As described in a recent meta-analysis (Bar-Haim, Lamy, Pergamin, Bakermans-Kranenburg, & van Ijzendoorn, 2007), a large amount of research exists demonstrating the presence of attentional bias across different types of anxiety disorders and with different experimental paradigms. Given these findings, it is surprising that there has not been much research concerning the role of attentional bias in health anxiety. Most studies investigating the relation between health anxiety and attentional bias used the emotional Stroop task (e.g., Lecci and Cohen, 2002, Lecci and Cohen, 2007 and Witthöft et al., 2008) which is based on the assumption that the emotional intrusion of salient stimuli leads to a slowing of color naming RTs. However, the emotional Stroop task falls short of dissociating the detailed processes involved in selective attention (e.g., processes of engagement and disengagement). Additional research on attentional bias in health anxious subjects stems from a study by Brown, Kosslyn, Delamater, Fama, and Barksy (1999) who compared hypochondriacal patients and controls using a degraded word task. They report an unexpected bias against reporting health-threat words in the hypochondriacal group. Other experimental paradigms as the spatial cueing task (Posner, 1980) or visual search tasks (e.g., Öhman, Flykt, & Esteves, 2001) have provided support for the presence of an attentional bias towards threatening information in anxious subjects (Cisler & Koster, 2010) but yet have not been applied to the study of health anxiety. To the best of our knowledge there exists only one study (Lees, Mogg, & Bradley, 2005) examining the relationship of attentional bias and health anxiety using a dot-probe paradigm with illness related word and picture stimuli. Nevertheless, there are studies that report a relationship between anxiety sensitivity (which is defined as the fear of anxiety symptoms, e.g., Reiss, 1987) and the dot-probe task featuring words as stimuli in both, non-clinical participants (Hunt et al., 2006, Keogh et al., 2001 and Roelofs et al., 2003) and a sample including controls and patients suffering from chronic headache (Asmundson, Carleton, & Ekong, 2005). There seems to be a close relationship between anxiety sensitivity and health anxiety for patients suffering from panic disorder (Otto, Pollack, Sachs, & Rosenbaum, 1992, p. 98) and depression (Otto, Demopulos, McLean, Pollack, & Fava, 1998). Stewart, Sherry, Watt, Grant, & Hadjistavropoulos (2008) studied the relationship between anxiety sensitivity and health anxiety in a student sample and also reported large correlations between both constructs. Nevertheless, they concluded that both constructs are better represented by two correlated but distinct traits compared to one single construct. Keogh et al. (2001) conducted a dot-probe task with physical-threatening words as stimuli (e.g., agony, harm, and hurt injury) and report a significant correlation of r = .28 for the dot-probe index (i.e., an attentional shift towards health threat words) with the physical facet of anxiety sensitivity in a sample of healthy psychology students. Lees et al. (2005) compared the attentional bias of participants who scored either high or low in the Illness Attitude Scales (IAS, Kellner, 1981). They only found a significant group effect in the dot-probe task with picture stimuli (instead of word stimuli) in the 500 ms exposure condition (instead of 1250 ms) after they reallocated the participants of the experimental groups by their scores on an anxiety sensitivity test. No effect was found for illness related word stimuli, which is supported by the findings of Bar-Haim et al. (2007) who reported a stronger attentional bias towards threat stimuli when using pictorial stimuli. In an attempt to replicate and extend the findings of Lees et al. (2005), we therefore decided to use a dot-probe task with pictorial stimuli and enlarged the number of trials instead of additional non-pictorial trials. A positive side effect of this procedure should be an enhancement of the reliability of the dot-probe task. The reliability of the dot-probe paradigm is an important and well known problem in both, clinical and non clinical-samples (e.g., Schmukle, 2005 and Staugaard, 2009). As Schmukle (2005) points out, the reliability of experimental measures as the dot-probe task is usually not reported. This is surprising because there are no technical reasons that hinder the calculation of such measures. 1.2. Processes of engagement and disengagement and the time course of attentional biases While it is agreed upon that attentional bias towards threat exists, there is still an ongoing debate if it comprises facilitated attention to threat, difficulty in disengagement from threat or even both (Cisler et al., 2009, Cisler and Koster, 2010 and Fox et al., 2002). Traditionally, the dot-probe effect is calculated as the mean reaction time difference between the trials when the probe appears instead of the neutral stimulus and when it appears instead of the threat stimulus (MacLeod, Mathews, & Tata, 1986). Recent findings by Koster, Crombez, Verschuere, and Houwer (2004), Koster, Crombez, Verschuere, and Houwer (2006) suggest that one should add baseline trials to the dot-probe procedure that only include neutral stimuli to identify if the dot-probe effect is a result of vigilance towards threat or difficulty to disengage from the threat stimuli. The average RT difference between the baseline (only neutral pairs) and trials where the probe appears at the former position of the health-threat stimuli can be called orienting index with a positive value indicating an orienting towards the threat. Additionally, the average difference in RT between the mixed neutral-threat trials where the probe appears at the position of the neutral stimulus and the baseline is called disengagement index, and a positive value indicates difficulties to disengage from the threat stimuli ( Koster et al., 2004). In order to distinguish between disengagement from and orienting towards the threat, we added baseline trials to the original dot-probe procedure ( Lees et al., 2005 and MacLeod et al., 1986). Koster et al. (2004) reported evidence for a difficulty to disengage from threat as an explanation of the dot-probe effect in a student sample. This pattern also occurred when they compared preselected high and low trait anxious students ( Koster et al., 2006). Salemink, van den Hout, and Kindt (2007) examined the two indices in a study that compared high and low anxious participants. They also reported evidence for a difficulty to disengage from threat as the rationale underlying the dot-probe paradigm. Accordingly, we especially expect positive correlations between the disengagement index (i.e., slower disengagement) and the measures of health anxiety and anxiety sensitivity. Bar-Haim et al. (2007, p. 10) reported group differences between anxious and control participants in dot-probe experiments that were significant for the traditional exposure time of 500 ms ( MacLeod et al., 1986) and shorter presentations but failed to reach significance with longer exposure times (>1.000 ms). As pointed out by Bradley, Mogg, and Millar (2000), the exposure time of 500 ms in the dot-probe task is quite long and more than one shift of attention could occur in this timeframe which complicates the interpretation of the task. Thus, it can be regarded as a shortcoming of the dot-probe task itself ( Cooper & Langton, 2006) that it only offers information about the attentional process at one particular point in the time course of attention (e.g., 500 ms after threat exposure). This shortcoming can only be bypassed by the implementation of more than one exposure duration. Weierich, Treat, and Hollingworth (2008) explain that disengagement and orienting processes do not have to contradict each other because one could expect vigilance towards a stimulus at an early stage of presentation (e.g., 100–200 ms) and problems to disengage from a stimulus at longer presentation times (e.g., 200–800 ms). This view is also consistent with the so-called vigilance-avoidance hypothesis (e.g., Mogg and Bradley, 2002 and Mogg et al., 2004). Therefore, we decided to implement 175 ms and 500 ms as exposure times in our study. Given the theoretical considerations by Weierich et al. (2008) and the implications of the vigilance-avoidance hypothesis (e.g., Mogg & Bradley, 2002), we expect larger effects for the orienting index at the short exposure time (i.e., faster orienting towards threat pictures) and directly opposite results for the disengagement index (i.e., slower disengagement from threat pictures). 1.3. Potential moderators of attentional biases Besides measures of anxiety sensitivity and health anxiety we assessed individual amount of absorption (i.e., openness to absorbing and self-altering experiences which is a trait related to hypnotic susceptibility; Tellegen & Atkinson, 1974). Tellegen and Atkinson (1974, p. 274) originally defined the construct as “episodes of a special attentional object relationship which can be described by such terms as ‘absorption’ and ‘fascination’. These terms suggest a state of ‘total attention’ during which the available representational apparatus seems to be entirely dedicated to experiencing and modeling the attentional object, be it a landscape, a human being, a sound, a remembered incident, or an aspect of one's self.”. It has been previously demonstrated that this construct is modestly correlated with self reported somatic symptoms (Gick et al., 1997 and Watten et al., 1997), anxiety sensitivity (Lilienfeld, 1997), and hypochondriacal concerns (McClure & Lilienfeld, 2002). Given these findings and the proposed relationship of absorption to attentional states by Tellegen and Atkinson (1974), we expect associations between absorption and the indices of orienting and disengagement in the dot-probe paradigm. 1.4. Aims and hypotheses of the present study In sum, our study aims at investigating the time course of selective attention towards pictorial illness related threat cues in health anxiety. By including a shorter (175 ms) and a longer (500 ms) stimulus probe onset asynchrony and a neutral condition for the orienting and disengagement effects, we hypothesized to detect faster orienting in the brief presentation condition and slower disengagement in the longer presentation condition for illness related stimuli associated with higher levels of health anxiety. A priori, we expect a relationship between general health anxiety as measured by the sum scores of the instruments we are going to apply and the dot-probe indices. Since health anxiety is currently regarded as a multidimensional construct (Longley, Watson, & Noyes, 2005) including affective (i.e., illness worries), cognitive (i.e., illness convictions), perceptual (i.e., somatosensory attention), and behavioral components (e.g., reassurance and safety-seeking behavior), the current study also aims at exploring in detail, which facets of health anxiety reveal the strongest association with attentional abnormalities. As previous studies have detected significant associations between anxiety sensitivity and an attentional bias towards (pictorial) illness cues (Keogh et al., 2001 and Lees et al., 2005), we also expect associations between different attentional bias indices and anxiety sensitivity. We expect similar but not identical correlational patterns between our dot-probe indices and measures of anxiety sensitivity and health anxiety because we regard both construct as related but still distinct. In line with Keogh et al. (2001), we expect the largest relationships with the physical facet of anxiety sensitivity. Finally, a possible association between the personality trait of absorption that has been demonstrated to be linked to health anxiety, and the attentional bias scores will be explored. In order to be able to compare our results to the initial study of Lees et al. (2005) and given the dimensional structure of health anxiety as demonstrated in taxometric analyses (Ferguson, 2008 and Longley et al., 2010), we decided to choose a correlational design using a sample of college students. Moreover, besides the work of Lees et al. (2005), this is the first attempt to study the relationship of attentional bias and health anxiety by means of a pictorial dot-probe task.
نتیجه گیری انگلیسی
1. Data preparation First of all, erroneous responses (2.6%) were excluded from statistical analyses. For the remaining responses, RTs shorter than 200 ms or longer than 2000 ms were removed from the data (5 trials). We defined individual outliers as response latencies deviating more than three SDs from the individuals mean RT in the six outcome variables that are used to calculate the dot-probe effect (baseline, neutral–threat with probe behind threat, neutral–threat with probe behind neutral; each at 175 ms and 500 ms exposure time). The resulting number of outliers (N = 196) per participant ranged from 0 to 6 (M = 2.39, SD = 1.18). All outliers were excluded from further data analysis. A chi-square test of independence was performed and showed no relationship between trial type and number of outliers (χ2(5, N = 196) = 4.27, p = .51). Discarded error and outlier RTs accounted for 3.7% of the entire data. To test our main hypotheses, we calculated six indices: (a) traditional dot-probe score (175 ms or 500 ms exposure time), as mean RT of neutral–threat trials when the probe appeared at the same position as the threat stimulus subtracted from neutral–threat trials with the probe at the position of the neutral stimulus; (b) orienting index (175 ms or 500 ms exposure time), as the mean RT of neutral–threat trials when the probe appeared at the position of the threat subtracted from mean RT in neutral–neutral trials; (c) disengagement index (175 ms or 500 ms exposure time), as the mean RT of neutral–neutral trials subtracted from the mean RT in neutral–threat when the probe appeared at the position of the neutral stimuli. We applied the Shapiro-Wilk test to the six indices and only found a significant deviation from normal distribution for the orienting index at 175 ms exposure time (W = .91, p < .01). This index showed two significant correlations with the questionnaires. One was slightly smaller (.19 instead of .23) and the other larger (.30 instead of .24) compared to a Spearman rank correlation. Because of these small differences and the robustness of the Pearson correlation coefficient (e.g. Edgell and Noon, 1984 and Zimmerman, 1986) we decided to apply the latter. As estimates of reliability, we calculated odd even split-half correlations ( Gulliksen, 1950) for the six dot-probe indices. 3.2. Data analysis From a pure statistical point of view, there are almost no reasons to dichotomize a continuous variable because this will result in a severe loss of statistical power (Cohen, 1983, MacCallum et al., 2002 and Preacher et al., 2005) and can produce very misleading results (e.g., Humphreys, 1978). We therefore decided to focus on the results of correlation analyses. All p-values reported throughout the article are two-tailed. Because we do not test the same hypothesis multiple times and we state no general null hypothesis that is valid for all comparisons (e.g., H0: There are no correlations between the dot-probe task and any questionnaire), the requirements for a so-called Bonferroni-adjustment are not met (e.g., Perneger, 1998, Rothman, 1990 and Savitz and Olshan, 1995). 3.3. Dot-probe effects The Spearman–Brown corrected split-half correlations (Gulliksen, 1950) for the six dot-probe indices ranged from r = −.07 (orienting at 500 ms) to r = .49 (disengagement at 175 ms). The four remaining indices yielded split-half correlations of r = .05 (traditional index at 500 ms), r = .39 (traditional index at 175 ms), r = .41 (orienting index at 175 ms) and r = .43 (disengagement index at 500 ms). Table 1 summarizes the results of the analyses concerning the relationship between the dot-probe indices and measures of health anxiety, anxiety sensitivity, and absorption. We are going to report the results of the 175 ms exposure condition first, followed by the 500 ms condition. Table 1. Pearson correlations of dot-probe indices with questionnaire measures. 175 ms 500 ms Orienting Disengagement Traditional Orienting Disengagement Traditional WI (sum score) .21 −.28** −.06 −.28* .20 −.04 WI health anxiety .18 −.15 .04 −.30** .22* −.04 WI illness beliefs .08 −.22* −.13 −.12 .07 −.03 MIHT (sum score) .23* −.12 .11 −.25* .24* .03 MIHT affective .16 −.15 .02 −.17 .18 .03 MIHT behavioral .23* .05 .26* −.24* .23* .03 MIHT cognitive .22 −.13 .08 −.14 .17 .06 MIHT perceptual .20 −.17 .04 −.28* .22* −.01 ASI physical .23* −.12 .01 −.22* .12 −.07 ASI social .19 −.16 .03 −.23* .19 .00 ASI cognitive .07 −.02 .05 −.17 .23* .09 TAS .24* −.07 .16 .07 .12 .19 Note. N = 83 for all correlations. WI = Whitley Index; MIHT = Multidimensional Inventory of Hypochondriacal Traits; ASI = Anxiety Sensitivity Index; TAS = Tellegen Absorption Scale. * p < .05. ** p < .01. Table options 3.3.1. Results of the 175 ms condition Significant positive correlations with the orienting index at 175 ms exposure time occurred for the MIHT sum score (r = .23, p < .05), the MIHT behavioral (r = .23, p < .05) and cognitive (r = .22, p = .05) scales, the physical facets of anxiety sensitivity (r = .23, p < .05) and for the TAS (r = .24, p < .05), suggesting that facets of health anxiety, anxiety sensitivity and absorption were positively associated with a faster orientation to threat under the 175 ms condition. No significant positive relationship between the disengagement index at 175 ms and any questionnaire was found. Concerning the traditional dot-probe indices ( MacLeod et al., 1986), we found a significant positive relationship with the MIHT behavioral scale at 175 ms exposure time (r = .26, p < .05), suggesting that the behavioral facet of health anxiety is associated with an attentional bias towards illness related pictures at 175 ms. 3.3.2. Results of the 500 ms condition At 500 ms exposure time, significant positive relationships with the disengagement index appeared for several measures of health anxiety and anxiety sensitivity. Especially the MIHT sum score (r = .24, p < .05), the MIHT behavioral (r = .23, p < .05) and perceptual (r = .22, p < .05) scales, the WI health anxiety scale (r = .22, p < .05) and the cognitive facet of anxiety sensitivity (r = .23, p < .23) yielded significant positive correlations to this index. In all those cases, except for the ASI subscale, the effects on the corresponding orienting index at 500 ms exposure time were significant (p < .05) in the opposite direction (negatively signed). This means that, the tendency for facilitated response to threats compared to the baseline (i.e., a positive orienting index) is associated with lower anxiety values on the questionnaires and a slower engagement with threat stimuli compared to the baseline (i.e., a negative orienting index) is associated with higher anxiety values on the questionnaires (the index ranges from −60.27 to 37.18, M = −5.88, SD = 19.41). There were no significant relationships with any questionnaire for the traditional dot-probe index at 500 ms exposure time. 4. Discussion The primary aim of the current study was to investigate the time course of an attentional bias towards health threatening picture cues associated with health anxiety. Since it is currently debated whether health anxiety and hypochondriasis are not better understood as an anxiety disorder (compared to a somatoform disorder; Olatunji, Deacon, & Abramowitz, 2009), one would expect anxiety-like abnormalities in attention allocation (e.g., a faster orienting and a slower disengagement from threat stimuli). Only one study has yet explored the process of attention allocation using the dot-probe task in the realm of health anxiety (Lees et al., 2005). One might critically object that picture stimuli are almost irrelevant in the pathogenesis of health anxiety and hypochondriasis since bodily sensations have been suggested as the primary focus of concerns in this condition. However, clinical observations and prior research suggest that attention allocation to external representations of health threat (e.g., media reports, articles) and mental images of severe illnesses (Muse et al., 2010) play a major role in the triggering of health anxiety episodes. In concert with Lees et al. (2005), we therefore consider the detailed investigation of attention allocation processes in health anxiety as relevant for better understanding the mechanisms of etiology and pathogenesis of severe health anxiety. Compared to the pioneer study by Lees et al. (2005), we observed attentional bias effects not only to anxiety sensitivity but also to dimensions of health anxiety, especially the behavioral facet that comprises safety- and reassurance-seeking behavior. Larger effects (i.e., positive correlations) occurred for the disengagement index at 500 ms exposure time, compared to 175 ms (mostly negative correlations). A different pattern emerged for the orienting index which showed larger effects at 175 ms exposure time instead of 500 ms. The fact that the orienting index tends to show significant correlations in the opposite direction as the disengagement index is not surprising as the orienting and disengagement indices, given the same exposure time, rather contradict each other on theoretical grounds (Koster et al., 2004 and Salemink et al., 2007). To find out if anxiety sensitivity and health anxiety accounted for distinct variance components of the disengagement index at 500 ms exposure time, we entered the ASI cognitive scale (which was the only ASI scale that showed a significant relationship to the disengagement index at 500 ms) in one step and each of the health anxiety measures that previously showed significant zero order correlations to this index (WI health-anxiety, MIHT sum score, MIHT behavioral, MIHT perceptual) in the second step. No significant improvements in the prediction of the index emerged (p > .05) and no matter which construct was entered first, the second one did not significantly improve the prediction of the index (p > .05). The same results occurred for the orienting index at 175 ms exposure time with the ASI physical scale and both the MIHT sum score and the MIHT behavioral scale. Thus, it seems to be impossible to distinguish the effects of health anxiety and anxiety sensitivity in our study. It is well known that these constructs share a lot of common variance ( Stewart et al., 2008) which is also the case in our dataset. We observed the largest correlations between the ASI physical subscale and the MIHT facets (all r > .50, p < .01), followed by the ASI cognitive scale (all r > .30, p < .01) and finally the ASI social scale (all r > .20). We do not want to focus on the comparison of the ASI and MIHT (e.g., Otto et al., 1992, Otto et al., 1998 and Stewart et al., 2008) because it is beyond the scope of this article. Nevertheless, we have reasons to conclude that in the domain of attentional bias with pictorial stimuli, both constructs probably account for similar variance components of the experimental measures. One possible reason for difficulties to find similar relationships between the dot-probe task and measures of health anxiety as we did could lie in the independent measures used previously. This becomes particularly clear with the WI sum score which unexpectedly showed no significant positive relationships with any measure of the dot-probe indices. An effect only occurred after empirically founded subscales (Schwarz et al., 2007) were used. Initially, we expected a significant relationship between general measures of health anxiety (MIHT and WI sum scores) and our dot-probe task. In the next step, we explored which facets account for the variance of the indices. While the perceptual facet showed one significant relationship, the cognitive facet of health anxiety did not show any relationships. The MIHT behavioral scale, which is the only facet of health anxiety that showed relationships with dot-probe indices at 175 ms and 500 ms exposure time, is supposed to assess “reassurance seeking used to allay illness fears” (Longley et al., 2005, p. 5). Put in a more general way, all of its items concern the behavior or reaction of the respondent in situations that are in some way connected to health issues or worries. Our version of the dot-probe task requires the subjects to respond to health-threatening stimuli, which is also a situation that is related to health issues. This is not the case for the other scales of the MIHT which contain items like “I am aware of my body position” or “I wish other took my health complaints more seriously” (Longley et al., 2005, p. 5). Therefore, it seems plausible that this scale shows substantial relationships with an experiment that requires (fast) reactions to a threatening stimulus. It is quite interesting that only the orienting index (and in one case the traditional index) showed significant positive relationships at an exposure time of 175 ms while at 500 ms it was the disengagement index that yielded significant positive correlations. Thus, it seems to be necessary to keep the exposure times of the stimuli in mind to explain the nature of the dot-probe effect. This is also what Cooper and Langton (2006, p. 1325) concluded in their dot-probe task that included a baseline (neutral faces) similar to ours. Contrary to our findings, they report problems to disengage from threat (angry faces) at 100 ms (significant, p < .05) and an orienting process at 500 ms (not significant, p > .05) exposure. This leads to another aspect concerning the nature of the dot-probe effects which is the actual eye movement of participants in the task. Mogg, Miller, and Bradley (2000) who measured RTs and eye movements in a dot-probe task with angry, sad, and happy faces (1000 ms exposure time) reported that individuals with generalized anxiety disorder were more likely to look at the threatening faces first compared to normal controls and subjects suffering from depressive disorder. They also reacted more quickly towards the threat faces. However, some questions concerning the time course of eye movements still remain because the precise course of eye movements was not assessed. Stevens, Rist, and Gerlach (in press) compared patients with social phobia and controls in a dot-probe task that featured a 200 ms and a 600 ms exposure condition. They report significantly more eye movements during the 600 ms condition and propose the use of eye movement measures at long exposure times. Stevens et al. (in press) conclude that presentation times below 200 ms in a visual dot-probe study prevent overt shifts of attention. Thus, at an exposure time of 175 ms it seems rather unlikely that the eyes focus one target (e.g., the threat stimuli) first, switch to another target (e.g., the neutral stimulus) and finally switch back again to the initial location. Therefore, and because the orienting index showed significant relationships to the questionnaires at 175 ms exposure time, one could assume that the focus of attention of the health anxious (anxious sensitive) subjects lies on the threat stimuli, which causes the faster RTs compared to the baseline trials. Although Young and Stark (1963) found out that most saccades are separated by about 200 ms, many researchers observed much shorter saccadic intervals (for an overview, see Bahill, Bahill, Clark, & Lawrence, 1975). McPeek, Keller, and Nakayama (1999) summarized several experiments and concluded that the saccadic system is capable of simultaneously programming two movements towards different (competing) targets which is probably the reason of shorter inter-saccadic intervals. Thus, it is not even totally clear which locations are focused in the 175 ms condition. Therefore, we think that our results have to be taken with care and a replication of the study would be important. At 500 ms exposure time it is even more difficult to make assumptions about the actual eye movements because, without doubt, multiple saccades could occur during this timeframe. Thus, it is not clear if the correlations between the questionnaires and the disengagement index at 500 ms exposure time occurred because more health anxious (anxious sensitive) participants focused the position of the threat for the whole presentation time and if this, in turn, is the reason for their difficulties to disengage from the threat. One possible explanation for our findings is that a minimum of stimulus exposure time somewhere beyond 175 ms is necessary for a disengagement process to happen but further research is necessary to shed some light on this issue. Future studies could systematically vary the presentation duration of the stimuli in the time frame between, 50 ms and 600 ms keeping all other factors constant. This could help to identify more precisely when the dot-probe response shifts from an orientation to a disengagement process. The reason for the significant correlation of the TAS with the orienting index at 175 ms exposure time is not totally clear. We calculated a hierarchical multiple regression (Cohen, Cohen, West, & Aiken, 2003) and entered the MIHT behavioral scale first, followed by the TAS. The increase in multiple R from .23 to .28 was not significant (F(1, 80) = 2.48, p = .12). The same was true when the TAS was entered first, followed by the MIHT behavioral scale (F(1, 80) = 2.11, p = .15). This means that the correlation of the TAS with the dot-probe effect could also be explained by its correlation with the MIHT behavioral scale of r = .35 (N = 83, p < .05). Our study is only the first step and showed at least some evidence for a relationship between health anxiety, anxiety sensitivity, and attentional bias in the dot-probe paradigm. In this article we observed correlations between r = −.30 and .26. Obviously, these are no large effects but still their size can be regarded as approximately medium ( Cohen, 1992). In line with various authors (e.g., Pernice, van der Veer, Ommundsen, & Larsen, 2008), we consider the use of student samples as adequate in early stages of research. Of course, it would be interesting to replicate our findings in a study with people suffering from full-blown hypchondriasis. This would also likely result in larger effect sizes that in turn would permit even more precise hypotheses and the inclusion of additional moderators (age, sex, etc.). A dot-probe version with baseline measures could also be used to evaluate the effect of attention training procedures in the treatment of hypochondriasis (Papageorgiou & Wells, 1998). In our study we found low reliability estimates of the dot-probe task which are comparable to the ones reported by Schmukle (2005) and Staugaard (2009). The reliabilities we found for the six dot-probe indices ranged from rather low estimates (r = .49) to even negative values (r = −.07) which means that one can hardly expect larger correlations with any questionnaire than we found. The lowest estimates emerged for the orienting and traditional index, both at 500 ms exposure time. It has to be pointed out that these results are in line with other studies which examined the reliability of the task itself and found even lower reliabilities ( Schmukle, 2005 and Staugaard, 2009). Given these results, it seems very unlikely that other dot-probe tasks yield a much higher reliability than ours. According to classical test theory ( Gulliksen, 1950), correlations (or covariances) are only based on true (i.e., reliable) scores and thus our results probably depict a lower bound estimate of the real relationship between attentional bias and health anxiety. Moreover, as Williams and Zimmerman (1977, p. 683) showed before, the ordinary equation (e.g., Gulliksen, 1950) that depicts the reliability of a difference score tends to underestimate the real reliability if the error terms of the scores are correlated. Since the two reaction time scores that are used to form the dot-probe indices are highly correlated (due to the dominance of general mental speed variance) reliabilities of such difference almost necessarily represent an underestimation of their true reliability. Nevertheless, as Schmukle (2005) and Staugaard (2009), we conclude that the reliability of the dot-probe task is still a crucial issue and one should attempt to improve it by further modifications of the procedure. A promising approach to deal with the rather low reliability of the dot-probe task emerges from the domain of latent variable modeling ( Loehlin, 2004). It relies on the variances and covariances of the true scores which makes the results less dependent of the low task reliability but one drawback is that is requires a large sample size ( Loehlin, 2004).