An evaluation of early and late stage attentional processing of positive and negative information in dysphoria

Transcription

1 COGNITION AND EMOTION 2007, 21 (4), An evaluation of early and late stage attentional processing of positive and negative information in dysphoria Matthew S. Shane and Jordan B. Peterson University of Toronto, Toronto, Ontario, Canada Depressive disorders may be characterised by hyperattention toward negative information, hypoattention toward positive information, or a combination of both processing biases. In two studies, a dot-probe task was utilised to better ascertain the specific direction and time-course of these biases. In both studies, the dysphoric group showed significantly less attentional allocation toward positive stimuli than the non-dysphoric group. In study two, the dysphoric group also showed greater attentional allocation toward depression-specific stimuli. Importantly, the bias toward depression-specific stimuli, and the bias away from positive stimuli, were uncorrelated with each other. It may be that both biases can act as sufficient, but not necessary, characteristics of dysphoric processing. An additional possibility is that the relative level of each bias type may best characterise dysphoric processing. Each of these possibilities is discussed in turn. A current controversy in research on information processing in depression concerns the existence and nature of attentional biases in depressed individuals. Cognitive models based on schemas (Beck, 1976) or associative networks (Bower, 1981) predict that individuals with depressive disorders should show mood-congruent processing biases at all stages of attentional processing (Ingram, Miranda, & Segal, 1998). Thus, according to these models, depressed individuals should manifest hypervigilant orienting towards negatively valenced information in their current environment (early stage attentional processing) as well as increased sustained attention and Correspondence should be addressed to: Matthew S. Shane, Clinical Cognitive Neuroscience Laboratory, The MIND Institute, 1201 Yale Blvd. NE, Albuquerque, NM 87131, USA. mshane@themindinstitute.org. This research was supported by the Social Sciences and Humanities Research Council of Canada. We gratefully acknowledge Chad Ebesutani and Laura Elinson for their help with data collection, and a number of anonymous reviewers for their excellent feedback. # 2007 Psychology Press, an imprint of the Taylor & Francis Group, an Informa business DOI: /

2 790 SHANE AND PETERSON rumination (later stage attentional processing) on such negative information (see Posner & Peterson, 1990, or LaBerge, 1995, for a discussion of early and late stage attentional processing). Empirical support for these hypotheses has been mixed, however. While a number of studies have identified such mood-congruent processing biases (Bradley, Mogg, & Lee, 1997; Gotlib & Cane, 1987; Ingram & Ritter, 2000; Mogg, Bradley, & Williams, 1995), others have failed to identify such biases (Hill & Dutton, 1989; MacLeod, Mathews, & Tata, 1986). Conversely, still other studies have reported that depression may be characterised by a lack of attentional bias towards positive information, rather than an active bias towards negative information (Gotlib, McLachlan, & Katz, 1988; McCabe & Gotlib, 1995; McCabe, Gotlib, & Martin, 2000). This inconsistency has made it difficult for researchers to form definitive conclusions regarding the nature of attentional processing in depressive disorders. A variety of methodologies have been adapted from cognitive psychology to investigate selective attention in depressive and anxious disorders, including the emotional Stroop task (Mathews & MacLeod, 1986), the deployment-of-attention task (DOAT; Gotlib et al., 1988), and the dotprobe task (MacLeod et al., 1986). The latter two have proven particularly useful because of their ability to distinguish between early and late stage attentional processing. In the dot-probe task, which has been utilised in the present study, pairs of words or pictures (e.g., one negatively valenced word and one neutrally valenced word) are presented on screen (e.g., for 500 ms or 1000 ms) followed by a dot probe in the location of one of the two preceding stimuli. Participants are required to press one of two buttons to indicate which side of the screen the probe appeared on. Prior research has determined that participants reaction times are faster when the probe appears in the position of the attended-to stimulus. Thus, attentional bias scores are calculated by subtracting probe detection speeds when the probe appears in the place of a negatively valenced stimulus from probe detection speeds when the probe appears in the place of a neutral stimulus. Short stimulus durations (e.g., up to 500 ms) have been used to measure early stage orienting processes, whereas longer stimulus durations (e.g., 1000/ ms) have been employed when later stage sustained attentional processing is being evaluated. The DOAT works in a similar fashion, but uses a forced-choice format rather than relying on reaction times, asking participants to decide which of two coloured bars appeared first (in reality they both appear simultaneously, but it has been found that the attended-to bar seems to appear first). Researchers utilising relatively short stimulus durations (up to 500 ms post-stimulus onset) have generally failed to find consistent biases in depressed individuals attentional processing. Mogg et al. (1995) found

3 ATTENTIONAL PROCESSING IN DYSPHORIA 791 no evidence of attentional biases in depressed individuals when the probe appeared 500 ms post-stimulus. Bradley et al. (1997) reported a similar lack of attentional bias, both at 14 ms and 500 ms post-stimulus, in naturally dysphoric individuals, and Mathews, Ridgeway, and Williamson (1996) presented data suggestive of biases only after 1000 ms post-stimulus. The consistency of these null findings suggest that depressive disorders, unlike anxiety disorders (Broadbent & Broadbent, 1988; Mogg & Bradley, 1998), may not be characterised by hypervigilant orienting or pre-attention towards negative environmental stimuli (Bradley et al., 1997; Williams, Watts, MacLeod, & Mathews, 1997; see, however, Bradley, Mogg, & Williams, 1994, and Luecken, Tartaro, & Appelhans, 2004, for potentially contrary evidence demonstrating biases during subliminal presentations). In contrast, when longer stimulus durations have been utilised (e.g., 1000/ ms), some evidence for depression-related attentional biases has been found. Mogg et al. (1995) found that clinical depression predicted negative attentional biases at 1500 ms post-stimulus, and Bradley et al. (1997) and Mathews et al. (1996) both found similar biases at 1000 ms post-stimulus. In each of these studies, depressed individuals were found to manifest increased sustained attention towards the negative stimuli at these longer stimulus durations, despite not showing orienting mechanisms more characteristic of anxiety disorders. This pattern of attentional deployment has led Williams et al. (1997) to hypothesise that the locus of depressionrelated biases may rest in later stage, elaborative processes. This interpretation remains consistent with a rich literature base demonstrating a high degree of ruminative behaviour in depressed (Beck, 1976; Harrington & Blankenship, 2002; Watkins & Brown, 2002) and dysphoric (Roberts, Gilboa, & Gotlib, 1998) individuals. A number of additional studies have failed to demonstrate even later stage attentional biases in depression, however (e.g., Hill & Dutton, 1989; MacLeod & Chong, 1998; Neshat- Doost, Moradi, Taghavi, Yule, & Dalgleish, 2000) and alternative hypotheses have been raised. Murphy et al. (1999) have, for instance, suggested that depressed individuals may have a more difficult time shifting attention away from negative stimuli once attention has been allocated, thus causing them to remain focused on the negative information after initial orientation in that direction (see also Ellenbogen, Schwartzman, Stewart, & Walker, 2002; Rinck & Becker, 2004). It should be noted that these two hypotheses are not necessarily mutually exclusive. Difficulty disattending to negative stimuli could, for instance, play a role in the depressed individual s increased ruminative behaviour. To further cloud the issue, Gotlib et al. (1988) have conversely suggested that it may be normal, rather than depressed, individuals who show biased

4 792 SHANE AND PETERSON attentional processing. Utilising the DOAT, Gotlib et al. (1988) found that whereas dysphoric individuals attended equally to negatively, positively and neutrally valenced words, non-dysphoric individuals attended more often to positive than to negative or neutral words. Thus, whereas the nondysphoric participants were characterised by biased attention toward positive stimuli, the dysphoric participants showed a lack of such bias in their attentional processing. This suggests that attentional processing biased toward positively valenced information may act as a protective factor against the onset of depression and dysphoria (see Taylor & Brown, 1988, who have explicitly suggested that the use of positive illusions, which are cognitive filters that preferentially screen out negative information, is commonplace, is a sign of mental health, and is negatively correlated with the onset of depressive symptomatology). Gotlib and colleagues have replicated their findings a number of times, in both clinically depressed patients (McCabe & Gotlib, 1995) and previously depressed patients (McCabe et al., 2000; see, however, Gotlib, Krasnoperova, Yue, & Joormann, 2004, for a recent non-replication using the dotprobe task). Only one study (McCabe & Toman, 2000) has, to date, varied the stimulus durations in the DOAT, however (750 ms has been standard), making determination of early and late stages of attentional processing within this task difficult. Thus, the present set of studies was designed to further investigate the extent to which dysphoria is characterised by hypervigilance to negative information, hypovigilance to positive information, or a combination of both processing biases. To this end, dysphoric and non-dysphoric individuals (as determined by Beck Depression Inventory scores [BDI]; Beck & Steer, 1993), were asked to perform a modified version of the dot-probe task, during which three different stimulus pairings were presented: positive neutral, negativeneutral, and neutral neutral (filler). Reaction times on positive neutral trials were used to index biases toward positive stimuli while reaction times on negative neutral trials were used to index biases toward negative stimuli. Separation of positive and negative stimulus presentations was implemented quite purposely. Although evaluating the nature of dysphorics attentional biases within situations characterised by the simultaneous presentation of both positive and negative stimuli may hold certain advantages (e.g., parsimony), such simultaneous presentation makes it impossible to determine the individual influence of each stimulus type. In such mixed-trials, for instance, a bias away from the negative stimulus, or a bias toward the positive stimulus, would be difficult to differentiate through reaction-time measures. A major purpose of the present study was to investigate the specific nature of positive and negative

5 ATTENTIONAL PROCESSING IN DYSPHORIA 793 biases in dysphoria; thus, use of mixed positive negative trials would not have served our purposes well. 1 The present set of studies was also intended to investigate a number of specific issues regarding the nature of positive and negative biases. First, we sought to determine whether these biases constitute global biases in attentional processing, as predicted by cognitive models based on schemas (Beck, 1976) or associative networks (Bower, 1981), or conversely are specific to either early or late stages of processing. The time of onset of probes was thus varied across trials, and appeared either after a short duration (500 ms in study one, and 200 ms in study two) or a long duration (1500 ms in both studies). Evidence of attentional biases on the short-duration trials is often interpreted as indicative of biases in early stage attentional processing, whereas evidence of attentional biases on the long-duration trials is often interpreted as indicative of later stage attentional biases. Second, we wanted to evaluate the possibility that the positive and negative biases may exist as independent markers of dysphoria. Davidson, Pizzagalli, Nitschke, & Putnam (2002) have suggested that symptoms of depression related to positive and negative emotions may manifest independently, and may be underlain by distinct proximal causes. By using separate trials to calculate positive and negative biases, it allows for the independence of these biases to be evaluated in several ways. First, and most directly, the correlation between the positive and negative biases could be calculated. A low correlation could suggest independence between attention allocated toward positive stimuli and attention allocated toward negative stimuli. In addition, we were particularly interested in the possibility that dysphoric individuals may manifest both positive and negative biases, but that each bias may show a distinct temporal pattern. A further possibility is that dysphoria may be characterised by a relative bias toward negative information (in relation to positive information), rather than an absolute bias toward either bias type (see Siegle, Ingram, & Matt, 2002, for the suggestion that dysphorics negative biases may come at the expense of attention allocated to more positive information). Thus, a final aim of the present research was to evaluate the predictive value of bias 1 It is important to recognise the benefits and drawbacks of our approach of using positive/ neutral and negativeneutral trials, rather than integrated positive negative trials. As was pointed out by an anonymous reviewer, our use of separate positiveneutral and negative neutral trials could serve to minimise a true correlation between positive and negative attentional biases. While acknowledging this, given that not all situations need be characterised by the presence of both positive and negative stimuli, and because almost no situation (and perhaps no situation) could be characterised by the presence of only positive and negative stimuli, use of mixed positive-negative trials in the absence of any non-valent stimuli, could cause an equal enhancement of the true correlation. Future research would benefit from a careful analysis of the benefits and drawbacks of each approach.

6 794 SHANE AND PETERSON differentials (BD), which were calculated by subtracting participants biases to negative information from their biases to positive information. By simultaneously taking into account the magnitude of both bias types, BDs may prove capable of identifying smaller effect sizes than either absolute bias. Theoretically, if BDs show superior prediction than positive and negative biases it would imply that dysphoric individuals need not consistently show strong attentional bias away from negative stimuli, nor strong bias toward positive stimuli. Rather, any combination of attentional processing resulting in a relative preference for negative stimuli may serve as the crucial characteristic underlying dysphoric processing. Method STUDY ONE Participants. Ninety-seven undergraduate psychology students at the University of Toronto volunteered for the study, in partial fulfilment of course credit. Of these participants, 66 were women and 29 were men. Their ages ranged from 1831 (M /20.43, SD /2.12). Assessment of dysphoria. Participants completed the Beck Depression Inventory (BDI; Beck & Steer, 1993), which has been shown in numerous studies to be a reliable and valid measure of severity of depressive/dysphoric symptomatology in college (Oliver & Burkham, 1979), community, and patient samples (for a review see Beck, Steer, & Garbin, 1988). Those scoring 6 and below on the BDI were classified as non-dysphoric, whereas those scoring 10 and above were classified as dysphoric. By removing those scoring between 7 and 9, potential difficulty regarding categorising slightly dysphoric individuals was avoided. These cut-off scores are consistent with previous research utilising the BDI to establish research groups (e.g., Tennen, Hall, & Affleck, 1995), and is in line with Kendall, Hollon, Beck, Hammen, and Ingram s (1987) recommendations regarding use of the BDI in non-clinical samples. Only those participants who met these cutoffs (N /73) were included in the final analyses. In addition to the BDI, participants also completed the short-form of the Taylor Manifest Anxiety Scale (TMAS; Bendig, 1956), because prior research has demonstrated depressive and anxious disorders to be highly comorbid. The TMAS is a 20-item, true/false questionnaire that has been shown to correlate highly with other measures of trait anxiety and negative affectivity (Watson & Clark, 1984). Description of affective stimuli. Stimuli were taken from the International Affective Picture System (Lang, Bradley, & Cuthbert, 1995), which

7 ATTENTIONAL PROCESSING IN DYSPHORIA 795 contains normative ratings of affective valence for each picture on a 9-point scale from unpleasant (1) to pleasant (9). Forty positive, 40 negative and 120 neutral pictures were chosen, based largely on these valence ratings. Negative pictures were selected in accordance with their relevance to sadness and threat, and included depictions of illness, disaster, drug injections and homelessness. Positive pictures included puppies, sunsets, sweet desserts and flowers. Neutral pictures were largely depictions of household items, such as lamps or chairs. Each emotional picture (positive or negative) was paired with a neutral picture, matching as closely as possible within each pair for colour and shape of the focal object. In total, there were 40 positive neutral, 40 negativeneutral, and 20 neutral neutral trials. The latter trial type acted as filler, and was prepared so that emotionally valent stimuli were not displayed on each trial, and to decrease the opportunity for participants to determine the nature of the task. Sexually explicit scenes from the IAPS were not included in the stimulus sets, as their influence on general arousal processes remains somewhat unclear. Procedure. After obtaining informed consent, participants completed the BDI and the short-form of the TMAS. Participants next completed the dot-probe task (designed using E-Prime software; PSInet Corporation) on a Pentium II computer and a 15-inch monitor, in a quiet room in the laboratory. The dot-probe task consisted of 10 practice, and 100 experimental trials (comprising the 40 positive neutral, 40 negativeneutral and 20 neutral neutral picture pairs). The practice trials were formed by using neutral neutral picture pairs not included in the experimental trials. Each trial started with a central fixation cross for 500 ms, followed by a picture pair, with half the pairs being displayed for 500 ms and half for 1500 ms. The positive or negative pictures appeared on the right or left side of the screen with equal frequency, as did the probe. In addition, the probe appeared in the place of the emotionally valent pictures on half of the trials, and in the place of the neutral picture on half the trials. For the neutral neutral trials, the probe appeared on the right or left an equal number of times. The trials were presented in a different random order for each participant. The dot probe was displayed immediately following the presentation of the picture pair, and participants were required to press one of two keys to indicate which side of the screen the probe appeared on. The left index finger was used to hit the Z key for probes located on the left side, and the right index finger was used to press the / (slash) key for probes located on the right side. Participants were informed to respond as quickly as they could, while remaining accurate. The probe remained displayed until the participant responded, or for a maximum of 10 seconds. The intertrial interval was 1000 ms. Upon completion of the dot-probe task, participants were awarded course credit, debriefed, and allowed to leave.

8 796 SHANE AND PETERSON Results Group characteristics. Women (M /8.30, SD/5.72) showed slightly higher BDI scores than men (M /7.92, SD /5.95), however, this difference did not prove significant, t(71) /0.07, p /.79. Subsequent data was thus collapsed across gender. As expected, the dysphoric group (M/14.10, SD/3.84) was characterised by significantly higher levels of dysphoria than the nondysphoric group (M/4.05, SD /2.01; t(71) / /14.55, d /3.47, p B/.001). In addition, there was a trend toward the dysphoric group manifesting higher anxiety than the non-dysphoric group (dysphoric group: M/29.73, SD/4.18; non-dysphoric group: M /27.98, SD/3.90; t(71) /1.84, d /0.43, p/.07). Attentional biases. Data from two participants (one dysphoric, one nondysphoric) were not included in the data set due to complications during the administration of the dot-probe task (the computer crashed in the middle of the task). Latency data from trials with errors (1.25% of all trials) were excluded from the analyses, as were latencies below 200 ms (2 trials) and greater than 700 ms (2.21% of all trials) after inspection of box and whisker plots. These cutoff points, as well as the percentage of trials excluded based on those cut-offs, are comparable with that of previous research utilising the dot-probe paradigm (Bradley, Mogg, Falla, & Hamilton, 1998; Mogg, Millar, & Bradley, 2000). Mean RTs were collected for each group (see Table 1) and attentional bias scores were calculated for each trial type by subtracting the mean RTs when probes were in the same location as the emotional pictures from the mean RTs when they were in the opposite location (Bradley et al., 1998; Mogg et al., 2000). Thus, bias to negative stimuli was calculated from participant s RTs on negative neutral trials, and bias to positive stimuli was calculated from participant s RTs on positiveneutral trials. Positive values of each bias score reflect attentional vigilance toward the valenced picture, whereas negative values reflect attentional avoidance of the valenced picture. These two bias scores were found to be uncorrelated with one another, r/.07, ns. These attentional bias scores were entered into a 2 /2/2 mixed-design ANOVA, with valence and duration as within-subject variables and dysphoria as a between-subject variable. 2 A significant main effect of 2 Before calculating attentional bias scores, we entered the raw RT data into a 2/2/2/2/ 2 ANOVA with valence (positive, negative), duration (500 ms, 1500 ms), picture location (left, right), and probe location (left, right) as within-subject variables, and dysphoria (high, low) as a between-subject variable. Significant main effects of picture location, F(1, 70)/7.97, p B/.01, duration, F(1, 70)/178.25, p B/.001 and valence, F (1, 70)/88.30, p B/.001 were identified. Of particular note, the predicted four-way interaction of dysphoria / valence / picture location / probe location reached significance, F(1, 70)/4.83, p/.03, which provided the basis for the calculation of the attentional bias scores used in the remainder of the analyses. The five-way interaction involving duration did not reach significance, F(1, 70)/1.46, p /.2.

9 ATTENTIONAL PROCESSING IN DYSPHORIA 797 TABLE 1 Mean response latencies to probes (in ms): Study one Picture valence Level of Exposure Picture Probe dysphoria duration location location Positive Negative Low 500 ms Left Left (56.34) (62.14) Left Right (61.07) (59.62) Right Left (66.66) (63.58) Right Right (59.14) (58.18) 1500 ms Left Left (49.77) (56.67) Left Right (46.17) (62.17) Right Left (54.31) (61.41) Right Right (50.15) (50.10) High 500 ms Left Left (53.05) (67.51) Left Right (67.89) (65.70) Right Left (58.59) (69.51) Right Right (63.27) (60.31) 1500 ms Left Left (61.32) (67.68) Left Right (62.74) (73.28) Right Left (63.49) (60.96) Right Right (57.22) (50.05) duration was revealed, F(1, 70)/8.59, h 2 /.11, pb/.01, as was a significant interaction between dysphoria and valence, F(1, 70)/4.11, h 2 /.06, pb/.05. The three-way dysphoria /valence/duration interaction did not reach significance, F(1, 70) /1.28, p /.20. We also performed a similar ANOVA including participants TMAS scores as a covariate to consider the possibility that the observed effects may be due to higher levels of anxiety in the dysphoric group. In this analysis, the main effect of duration disappeared (p /.50), however the predicted dysphoria/valence interaction remained significant, F(1, 69)/3.65, h 2 /.05, p/.05. Furthermore, the anxiety/valence interaction was nonsignificant, F(1, 69)/.19, p/ Dissection of the significant dysphoria /valence interaction allowed for an initial investigation into the influence of positivity and negativity biases in dysphoria. Figure 1a depicts this interaction in graphical form, with early and late biases collapsed. As can be seen, non-dysphoric individuals showed a slight vigilance toward positive stimuli, whereas dysphoric individuals showed a more pronounced avoidance of positive stimuli. Planned comparisons confirmed that these attentional biases differed significantly from 3 Interpretation of these covariate analyses need be made with caution. Miller and Chapman (2001) have pointed out that using a covariate that is highly correlated with the measure of interest may regress out substantial variance, and may leave a variable with little in common with the original construct. With appropriate caution, we believe these covariate analyses remain useful, however, we also believe it important to note these issues of interpretation.

10 SHANE AND PETERSON (a) 10 Dysphoric Nondysphoric PB NB (b) PB E NB E (c) PB L NB L Figure 1. Level of bias toward/away from positive and negative stimuli on (a) all trials (b) shortduration trials, and (c) long-duration trials. Note : PB/bias to positive stimuli; NB/bias to negative stimuli; PB E /bias to positive stimuli on early trials; NB E /bias to negative stimuli on early trials PB L /bias to positive stimuli on late trials; NB L /bias to negative stimuli on late trials.

11 ATTENTIONAL PROCESSING IN DYSPHORIA 799 each other, t(70) /2.13, d /.50, p /.04. However, neither bias differed from zero [dysphoric: t(29) / /1.55, d /.20, p /.13; non-dysphoric: t(41) /1.33, d /.28, p /.19]. In contrast, both dysphoric and non-dysphoric individuals showed tendencies to avoid the negative stimuli, and this avoidance reached significance for the non-dysphoric participants, t(41) / /2.20, d /.33, p/.03. Temporal characteristics. Because this study represented one of the first investigations of temporal differences between biases toward positive and negative stimuli, we examined the pattern of biases on short- and longduration trials separately, despite the nonsignificant three-way interaction involving duration (see Figures 1b and 1c). On short-duration trials, the dysphoric group showed pronounced avoidance of the positive pictures that differed from zero, t(29) / /2.35, d /.42, p /.03, and from the nondysphoric group, t(70) /1.76, d /.41, p/.08, and unbiased processing of the negative pictures. The non-dysphoric group, in contrast, showed the opposite pattern of processing: pronounced avoidance of the negative pictures that differed from zero, t(41) / /3.59, d /.54, p B/.001, but not from the dysphoric group, t(70) / /1.63, d /.39, p /.11, and equal processing of the positive pictures. On long-duration trials, dysphoric individuals showed no evidence of biases, while the non-dysphoric individuals showed pronounced bias toward positive stimuli, t(41) /2.52, d/.39, p/.02. Bias differentials. Finally, we wanted to investigate the possibility that the relative magnitude of individuals biases toward positive and negative stimuli may provide a more robust indicator of dysphoric processing than either absolute bias. To this end, we calculated bias differential (BD) scores for each participant by subtracting their bias toward negative stimuli from their bias toward positive stimuli. Positive BDs thus indicate an overall bias toward positive stimuli, whereas negative BDs indicate an overall bias toward negative stimuli. BDs were found to correlate significantly with biases toward both positive and negative stimuli, and also correlated significantly with BDI scores, r//.23, p/02. Furthermore, the nondysphoric group was characterised by highly positive BDs (M/13.13, SD/ 31.27) that differed significantly from baseline, t(41) /2.72, d/.42, p/.01, and from the dysphoric group, t(70) /1.93, d /.46, p/.05. In contrast, the dysphoric group manifested slightly negative BDs (M//3.34, SD/41.04), which did not differ from zero, t(29) /0.45, p B/.60. Thus, non-dysphoric individuals showed overall attentional patterns skewed toward positive stimuli, while the dysphoric individuals showed a mild, but nonsignificant, skewing toward negative stimuli. To investigate the possibility that BDs added to the prediction of dysphoria over and above either absolute bias

12 800 SHANE AND PETERSON TABLE 2 Regression models using absolute biases and bias differentials to predict BDI: Study one Predictor Standardised coefficient (b) p-value Model p-value Model 1 Bias to positive stimuli / Model 2 Bias to positive stimuli / Bias differential / Model 1 Bias to negative stimuli Model 2 Bias to negative stimuli / Bias differential / score, a series of regression analyses were conducted in which the absolute biases toward positive and negative stimuli were entered in block one, and BD was entered in block two. 4 Results demonstrate that BD successfully entered into both regression models in block two, while knocking the absolute bias scores out of the model (see Table 2). Finally, early and late BDs were considered separately. Non-dysphoric individuals showed positive BDs at both early and late stages of attentional processing (early: M /3.92, SD /47.09; late: M/12.87, SD/53.68), while dysphoric individuals showed negative BDs at both stages of processing (early: M / /24.31, SD /72.13; late: M/ /14.78, SD/90.82). A repeated measures ANOVA with early and late BDs as within-subjects variables failed to reach significance (p /.95). Discussion In the present study, only non-dysphoric individuals showed significant avoidance of negative stimuli, while dysphoric individuals showed significantly less attention to positive stimuli than non-dysphoric individuals. On the whole, these findings are consistent with a myriad of previous research identifying one or the other of these biases in dysphoric and depressed populations (Bradley et al., 1997; Gotlib et al., 1988; Ingram & Ritter, 2000; Mogg et al., 1995). Our results do differ from previous findings in a number of ways, however. First, most previous work has found that individuals in depressive states show pronounced vigilance toward negative stimuli. In contrast, we report unbiased processing in dysphoric individuals, and pronounced avoidance 4 For these continuous analyses, all participants including those with BDI scores between 6 and 10 were included in the analyses. Results remained similar when excluding this group, however, the nature of continuous analyses required inclusion of the full sample of participants.

13 ATTENTIONAL PROCESSING IN DYSPHORIA 801 of negative information in non-dysphoric individuals. One reason for these divergent results may pertain to the use of IAPS slides in the present study. Recent work suggests that the use of depression-specific stimuli may be preferable to the use of broadly negatively-valenced stimuli, which can include a combination of threat-related, depression-related, and disgustrelated stimuli, not all of which may be expected to elicit depressionrelated biases (Gotlib et al., 2004; Westra & Kuiper, 1997). The IAPS slides used in the present study may then have constituted too heterogeneous a group of negative pictures to appropriately evaluate the processing biases present in dysphoria. This could have had two effects. First, the extent to which the IAPS slides elicited biases within the dysphoric individuals may have been somewhat reduced. Second, the inclusion of anxiety- and disgust-related stimuli may have unintentionally tapped alternate processing biases that could have been present within the non-dysphoric population. In study two, then, we ran a similar version of the dot-probe task utilised in study one, but presented depression-specific words, instead of negatively valenced IAPS pictures. Thus, three word-pair types were utilised in study two: positiveneutral, depression neutral, and neutral neutral. Positive neutral trials were used to index participant s bias toward positive stimuli, and depressionneutral trials were used to index participant s bias toward depression-specific stimuli. Our predictions were similar to study one, in that we predicted the dysphoric group to manifest increased processing of the depression-specific words. In addition, based on the results of study one, we anticipated that the dysphoric group may also show increased avoidance of positive stimuli. A second difference between the results of the present study and those of previous research pertains to the temporal characteristics of the attentional biases. Previous research has fairly consistently, although not unanimously, reported a late-stage bias toward negative stimuli in depressed or dysphoric individuals, and, conversely, has failed to show the corresponding early stage bias (e.g., Suslow, Junghanns, & Arolt, 2004). The present study, in contrast, found no significant differences between early and late stage processing when evaluating the relevant repeated-measures ANOVA, and subsequent exploratory analyses, if anything, revealed stronger biases in the dysphoric group during early stage processing. To this we have two comments. First, the present findings are not without precedent, as a number of recent studies have reported increases in early stage*and possibility even preconscious*susceptibility to stimuli in depression. Lundh, Wikstrom, Westerlund, & Ost (1999), for instance, found that subliminal presentations of panic-related words correlated significantly with level of trait depression. Luecken et al. (2004) have shown similar effects of subliminal stimuli in a dot-probe task not unlike the

14 802 SHANE AND PETERSON presently devised task. These studies evaluated attentional patterns earlier, even, than the 500 ms timeframe utilised in the present study and together with the present findings suggest the possibility of early stage biases in dysphoric processing. Second, it should be noted that the dysphoric group showed early stage processing biases not on negative neutral trials, but on positive neutral trials. Previous work demonstrating late stage biases in depression and dysphoria have reported such biases with regard to negative information. We are, however, unaware of any previous work evaluating the temporal patterning of biases to positive stimuli in depression or dysphoria. It may be that biases to positive information have a different temporal pattern than biases to negative information. In this regard, it is interesting to note that the dysphoric group only showed a bias toward positive information on shortduration trials. Another possibility, however, is that the early stage duration utilised in the present study (500 ms) may not have sufficiently tapped early stage attentional processing. Recent evidence has suggested that individuals are capable of shifting their attention within ms of stimulus onset (Chun & Wolfe, 2001; Duncan, Ward, & Shapiro, 1995), and thus by 500 ms, we may have already been evaluating later stages of attentional processing. This could explain why we did not obtain a significant interaction with duration in study one, and may further explain why we found such pronounced biases during early stage processing. To this end, study two reduced the probe onset time in the early stage trials to 200 ms, in order to ensure that we were accurately indexing the initial orienting tendency of all participants. Method STUDY TWO Participants. Eighty-one undergraduate psychology students at the University of Toronto volunteered for the study, in partial fulfilment of course credit. Of these participants, 51 were women and 30 were men. Their ages ranged from 1829 (M /20.20, SD /1.74). Assessment of dysphoria. As in study one, participants completed the BDI, as well as the TMAS. Again, as in study one, those scoring 6 and below on the BDI were classified as non-dysphoric, whereas those scoring 10 and above were classified as dysphoric (consistent with Tennen et al., 1995, and in line with Kendall et al. s, 1987, recommendations). Sixty-six participants (39 non-dysphoric, 27 dysphoric) met this final criterion and were included in the subsequent analyses.

15 ATTENTIONAL PROCESSING IN DYSPHORIA 803 Description of affective stimuli. Words were utilised in study two rather than pictures because previous studies have utilised depression-specific words with considerable success (Bradley et al., 1997; Yovel & Mineka, 2005), and because creating a collection of depression-specific pictures seemed a somewhat daunting task. Three word lists were thus developed: a positive word list (40 words, e.g., success, party, joy), a depression-related word list (40 words, e.g., gloomy, bleak, failure), and a neutral word list (100 words, e.g., range, vary, directly). Each word list was matched for both word length and word frequency (Kuchera & Francis, 1967), and three independent judges rated each word on 7-point Likert scales for its level of positivity, anxiety-relatedness, depression-relatedness, and arousability. To this end, each depression-related word was rated as more related to depression than to anxiety. In addition, the positivity and arousal ratings of the positive words were matched to the depression-relatedness and arousal ratings of the depression-related words. Thus, differences in the degree of positivity/ depression-relatedness or the degree of arousal could not be used to provide alternative explanations of the findings. Procedure. After obtaining informed consent, participants were seated in front of a computer with a 15 inch monitor, and completed the dot-probe task. The task was identical to that described in study one, with the following four exceptions. First, words were used as stimuli, rather than pictures. Second, each word pair was presented twice, once for each duration-type. Of these, half of the word-pairs appeared for the first time on the short-duration trials, while the other half appeared for the first time on the long-duration trials. Third, since each word pair was presented twice, the number of trials was doubled from that of study one. Thus, 10 practice trials (utilising 20 unique neutralneutral word pairs) were followed by 200 experimental trials (comprising 80 positiveneutral, 80 negativeneutral and 40 neutralneutral word pairs). This was done in order to increase the number of observations used to calculate each bias score, and thus to increase the reliability of the task. Fourth, the short-duration trials had a probe onset of 200 ms, rather than 500 ms as was utilised in study one. The probe onset for long-duration trials remained at 1500 ms. The final difference between the procedure in study two was that participants completed the BDI and TMAS after completing the dot-probe task, rather than before. Mild concern was raised regarding the possibility that completing these scales before performing the dot-probe task may have led to subtle priming of attention toward the negative stimuli. In order to ensure this did not influence the data for study two, we reversed the order of administration in the second study.

16 804 SHANE AND PETERSON Results Group characteristics. Once again, women (M/9.82, SD/9.21) showed slightly higher BDI scores than men (M /8.83, SD /8.88), however this difference was not significant, t(64) /.47, p/.64. Thus, as in study one, subsequent analyses were conducted with data collapsed across gender. As expected, the dysphoric group (M /19.15, SD /9.65) was characterised by significantly higher levels of dysphoria than the non-dysphoric group [M / 3.38, SD /1.63], t(64) / /8.40, d /2.83, p B/.001, and also showed significantly higher levels of comorbid anxiety [dysphoric: M /31.37, SD / 4.07; non-dysphoric: M /26.94, SD /3.77], t(64) / /4.47, d /.17, p B/.001. Attentional biases. As in study one, latency data from trials with errors (B/1% of all trials) were excluded. Similarly, after inspection via box and whisker plots, latencies below 200 ms or greater than 700 ms (1.73% of all trials) were also excluded from analyses. These exclusion criteria remain consistent with those utilised in study one, as well as with the extant dotprobe literature (Bradley et al., 1998; Mogg et al., 2000). Mean RTs were collected for each group, and are displayed in Table 3. As in study one, attentional bias scores were calculated for each trial type by subtracting the mean RTs when probes were in the same location as the emotional pictures from the mean RTs when they were in the opposite location (Bradley et al., 1998; Mogg et al., 2000). Thus, positive neutral trials were used to calculate biases toward/away from positive stimuli, and depressionneutral trials were used to calculate biases toward/away from depression-specific stimuli. As in study one, these two biases were found to be largely uncorrelated, r/.09, ns. These attentional bias scores were entered into a 2 /2/2 mixed-design ANOVA, with valence and duration as within-subject variables, and dysphoria as a between-subject variable. 5 Neither the main effect of duration, F(1, 63) /0.12, p /.73, nor valence, F(1, 63) /1.83, p /.18, reached significance. The predicted dysphoria / valence interaction was highly significant, F(1, 63)/7.49, h 2 /.11, pb/.008. As in study one, the three-way interaction involving duration did not reach significance, F(1, 63) /0.003, p/.96. A second similar ANOVA was 5 Once again, before calculating attentional bias scores, we entered the raw RT data into a 2/2/2/2/2 ANOVA with valence (positive, negative), duration (200 ms, 1500 ms), picture location (left, right), and probe location (left, right) as within-subject variables, and dysphoria (high, low) as a between-subject variable. Significant main effects of duration, F(1, 63)/479.80, p B/.001, and valence, F(1, 63)/15.01, p B/.001 were identified. The predicted four-way interaction of dysphoria/valence/picture location / probe location reached significance, F (1, 63)/9.93, p/.002, which provided the basis for the calculation of the attentional bias scores used in the remainder of the analyses. The five-way interaction involving duration did not reach significance, F(1, 63)/1.66, p/.20.

17 ATTENTIONAL PROCESSING IN DYSPHORIA 805 TABLE 3 Mean response latencies to probes (ms): Study two Picture valence Level of Exposure Picture Probe dysphoria duration location location Positive Negative Low 200 ms Left Left (22.12) (23.92) Left Right (37.79) (37.71) Right Left (25.23) (18.08) Right Right (31.26) (32.04) 1500 ms Left Left (18.72) (40.30) Left Right (28.09) (39.28) Right Left (22.73) (38.28) Right Right (36.70) (47.46) High 200 ms Left Left (20.57) (20.59) Left Right (34.17) (32.55) Right Left (19.95) (22.26) Right Right (40.00) (36.69) 1500 ms Left Left (29.57) (27.89) Left Right (32.01) (28.04) Right Left (25.51) (33.26) Right Right (26.88) (46.87) conducted with participants TMAS scores included as a covariate, and this analysis yielded a similar valence/dysphoria interaction, F(1, 62) /8.96, h 2 /.13, p/.004. In addition, the valence/anxiety interaction was nonsignificant, F(1, 62) /1.59, p/.21. Thus, as in study one, the results appear to have been carried predominantly by level of dysphoria, rather than comorbid anxiety. Figure 2a depicts the dysphoria /valence interaction. Non-dysphoric individuals showed a slight, although nonsignificant, vigilance toward the positive stimuli, t(38) /1.22, p/.23, while dysphoric individuals showed a more pronounced avoidance of the positive stimuli, t(25) / /2.51, d/.55, p /.02. Planned comparisons confirmed these attentional patterns to differ significantly from one another, t(63) /3.03, d /.77, p /.004. In addition, dysphoric participants manifested a bias toward the depression-specific stimuli, t(25) /2.17, d /.39, p /.04. Planned comparisons confirmed that the two groups dysphoric-specific bias scores differed significantly from each other, t(63) / /1.97, d /.51, pb/.05. Temporal differences. Despite the nonsignificant three-way interaction involving duration, we once again examined the data for early and late duration trials separately (Figures 2b and 2c). These analyses revealed two separate effects. First, the dysphoric group showed significant avoidance of the positive stimuli on short-duration trials, t(25) / /2.23, d/.45, p/.03,

18 SHANE AND PETERSON (a) 20 Dysphoric Nondysphoric (b) 20 PB DSB PB E DSB E (c) PB L DSB L Figure 2. Level of bias toward/away from positive and depression-specific stimuli on (a) all trials (b) short-duration trials, and (c) long-duration trials. Note : PB/Bias to positive stimuli; DSB/bias to depression-specific stimuli; PB E /bias to positive stimuli on early trials; DSB E /bias to depressionspecific stimuli on early trials; PB L /bias to positive stimuli on late trials; DSB L /bias to depressionspecific stimuli on late trials.

19 ATTENTIONAL PROCESSING IN DYSPHORIA 807 which differed from the non-dysphoric group, t(63) /3.15, d/.76, p/.002. Second, the dysphoric group showed significant vigilance toward dysphoricspecific stimuli on long-duration trials, t(25) /2.37, d/.47, p/.03, which differed from the non-dysphoric group, t(63) / /2.36, d/.61, p/.02. The non-dysphoric group showed no evidence of either bias at either stimulus duration. Bias differentials. Finally, bias difference (BD) scores were once again calculated for each participant by subtracting their bias toward positive stimuli from their bias toward depression-specific stimuli. Analyses indicated that the non-dysphoric group was characterised by somewhat positive BDs (M /6.74, SD/32.68), while the dysphoric group manifested highly negative BDs (M / /22.62, SD/36.47). These BDs differed significantly from each other, t(63) /3.39, d/.85, p B/.001, but only the dysphoric group s negative BD differed significantly from zero [dysphoric: t(25) /3.16, d /.62, p /.004; non-dysphoric: t(38) / /1.29, p/.20]. Thus, similar to study one, dysphoric individuals manifested considerably more negative BDs than non-dysphoric individuals, a result that may have been further enhanced by the use of depression-specific content. Regression analyses entering the absolute bias scores in block one, followed by BD in block two, indicated that BD significantly entered into the model in which depressionspecific bias was entered in block one (specific bias: b / /.206, p/.26; BD: b/.355, p/.05), but failed to enter the model when positivity bias was entered in block one (see Table 4). It should be noted, however, that BD did knock the bias to positive stimuli out of the regression model when entered in block two. However, neither predictor remained significant in this model. Finally, early and late BDs were considered separately. Non-dysphoric individuals showed neutral or positive BDs at both early and late stages of attentional processing (early: M / /.87, SD /49.26; late: M /2.63, SD / 46.78), while dysphoric individuals showed negative BDs at both stages of processing (early: M/ /20.27, SD /46.89; late: M / /15.79, SD/49.36). A repeated measures ANOVA with early and late BDs as within-subjects variables failed to reach significance, p /.95. Discussion The results of study two extend the findings from study one in a number of important ways. First, whereas study one demonstrated that dysphoric individuals manifested a bias away from positive stimuli at 500 ms poststimulus onset, the results of the second study indicate that a similar bias away from positive stimuli can be identified as early as 200 ms post-stimulus onset. Two-hundred milliseconds was chosen as the stimulus duration for the

20 808 SHANE AND PETERSON TABLE 4 Regression models using absolute biases and bias differentials to predict BDI: Study two Predictor Standardised coefficient (b) p-value Model p-value Model 1 Bias to positive stimuli / Model 2 Bias to positive stimuli / Bias differential / Model 1 Bias to depression-specific stimuli Model 2 Bias to depression-specific stimuli / Bias differential / second study because research demonstrates that ms is generally required in order to perform a voluntary disengagement and re-engagement of attentional gaze (Chun & Wolfe, 2001; Duncan et al., 1995). Thus, at 200 ms we can be relatively assured that the direction of the participants initial orienting in response to the dual-word display was being evaluated, and further that dysphoric individuals show initial orienting responses away from positive stimuli. Research on orienting responses often claim that such responses occur automatically. Although we are hesitant to characterise this dysphoric bias as automatic, the results of the present studies do suggest that they are likely the result of particularly early stage processes. This notion is consistent with a handful of recent studies which have also revealed evidence of early stage abnormalities in depressives attention to negative stimuli (Luecken et al., 2004; Lundh et al., 1999). Second, by using depression-specific stimuli in study two, in contrast to the more general category of negatively valent stimuli utilised in study one, a strong bias toward the depression-related stimuli was revealed within the dysphoric group. A number of recent investigations have suggested that depressive and dysphoric individuals manifest a specific bias toward depression-specific information, which may not generalise to other negatively valent stimuli (Gotlib et al., 2004; Westra & Kuiper, 1997). The results of study two are consistent with these observations, and suggest that the inability to demonstrate a significant bias toward the negative stimuli in study one may have been due to the more heterogeneous collection of IAPS slides used. The present study thus confirms the importance of utilising depression-specific stimuli when investigating processing biases in dysphoria, and provides further evidence to indicate that dysphoria is characterised by biased attention toward such depression-related stimuli. The depression-specific nature of this bias is intriguing. A number of cognitive and neural models have been proposed to explain how early, and perhaps pre-attentive, systems may preferentially process threat-related stimuli. We are unaware of any similarly conceptualised system for the early