Rumination has been associated with increased vulnerability to, maintenance of, and delays in recovery from depression. Yet, different researchers define rumination differently. A number of experiments were conducted to examine the extent to which these definitions represent a single construct and thus, the extent to which they are associated with aspects of depression. First, seven rumination scales were administered to 189 undergraduates. Generalizability analysis suggested that the scales were not responded to consistently by participants, though each scale had adequate internal consistency. Follow-up examinations suggested that while the scales overlap to some extent they measure different dimensions of rumination. The second study involved giving the same measures to depressed and never-depressed individuals during information processing tasks designed to provoke sustained processing of emotional information. Reaction times, pupil dilation (a correlate of cognitive load) and functional magnetic resonance imaging (fMRI) were used to assess sustained brain activity during the tasks. Scores on many of the administered rumination scales were highly correlated with reaction time biases, sustained pupil dilation to negative information, and sustained amygdala activation. Results suggest that while measures of rumination may tap different phenomena, they have similar physiological correlates.

The contribution to results reported below by of a number of collaborators, including Rick Ingram, Eric Granholm, and Georg Matt, is gratefully acknowledged.

Rumination, construed loosely as having repetitive, intrusive, negative cognitions, has been associated with elevated and prolonged sad mood (e.g., Nolen-Hoeksema, 1991) well as vulnerability to, and maintenance of clinical depression (Just & Alloy, 1997).Rumination has also been linked to other cognitive aspects of depression, e.g., self-reflection (meta-cognitive processes; Papageorgiou & Wells, 1999a,b). Yet, many definitions of rumination have been considered (e.g., Alloy, 1988; Ingram, 1984; Martin & Tesser, 1989; Nolen-Hoeksema, 1987; Teasdale & Barnard, 1993; Wells & Matthews, 1994; Philippot & Rime, 1998). The extent to which these definitions refer to a single construct is unclear. Similarly it is unclear whether different types of rumination are related, in the same way, to aspects of depression. These questions translate directly to difficulty in assessing rumination and examining its clinical correlates.

This presentation therefore examines relationships between a number of potential operationalizations of rumination including self-report, behavioral, and physiological measures. The presentation is divided into two sections. The first section examines correspondences between popular self-report measures of depression, many of which were discussed in the current symposium. The second section examines relationships between responses to these measures and objective behavioral and physiological measures of sustained information processing, which might also index processes underlying rumination.

Just as there are numerous theories of the nature of rumination, self-report measures created to reflect these theories differ considerably. Self-report measures of rumination have nominally assessed thinking about depressive symptoms (Nolen-Hoeksema, Morrow, & Fredrickson 1993), intrusiveness of thoughts about a distressing event (Horowitz, Wilner, & Alvarez, 1979), process and metacognitive dimensions associated with naturally occurring depressive rumination (Papageorgiou & Wells, 1999a,b), the frequency and degree of distress associated with thoughts about recent negative events (Luminet, Rime, & Wagner, in prep), searching for meaning of negative experiences, and thinking about what can be done to change one’s situation in regard to negative events (Fritz, 1999). Similarly, a host of other measures have assessed rumination-like constructs that closely resemble theoretical definitions of rumination proposed in the literature, e.g., "rehearsal" (Roger & Najarian, 1989) and "reappraisal" (Wells & Davies, 1994) of negative thoughts.

It is not known whether these instruments index the same type of rumination. Thus, it is unclear whether investigators employing just one measure of rumination should phrase their research as referring to rumination in general, or rumination as measured by specifically by that instrument. Clinically, this issue is important to understanding the true nature of associations between rumination and depressive symptomatology. Specifically, it is unclear whether rumination, as measured by each of these instruments, is differentially associated with dysphoria and depressive symptomatology.

To clarify the nature of rumination, this study explored the extent to which self-report measures of rumination represent similar constructs, by administering them to a large number of undergraduate students. The primary question of interest involved understanding the extent to which variation in individuals' scores on the administered instruments was due to individual differences in a similar construct (presumably rumination), versus reliable differences between instruments, versus idiosyncratic patterns of response to the different questionnaires.

Generalizability theory (Cronbach, 1963, 1972) was used to partition variance into the components in question using an ANOVA framework (Matt, in press). Supporting techniques including internal consistency analysis, exploratory factor analysis, and correlation were then used to aid in understanding results of the generalizability theory analysis. To clarify the relationship of the administered scales to dysphoria, analysis of zero-order, semi-partial, and dependent correlations between rumination measures and a measure of depressive symptomatology (Beck Depression Inventory; Beck, 1967) were investigated.

Method

Participants. Participants included 182 undergraduates (56 male, 142 Caucasian, M_age=20.73, SD_age= 4.41, M_education=13.59, SD_education=0.53) from an introductory psychology course at the University of Pittsburgh. Participants received course credit for their participation.

Instruments. A packet of self-report measures that have been used to assess rumination and rumination-like constructs was administered. The packet is described in depth below so that other researchers may decide whether they would like to use these scales as well.

The Response Styles Questionnaire (RSQ; Nolen-Hoeksema, Morrow, & Fredrickson, 1993) is a 71 item inventory with rumination and distraction subscales. The 22 item rumination subscale assesses the frequency which individuals think about their symptoms of depression, when they feel sad or depressed on a four point scale from "almost never" to "almost always" Previous studies have shown the RSQ rumination scale to be internally consistent and factor-analytically derivable. Higher scores are related to more severe and longer episodes of depression (e.g., Nolen-Hoeksema, Morrow, & Fredrickson, 1993) as well as distorted interpretations of hypothetical life events (Lyubomirsky, Caldwell & Nolen-Hoeksema, 1998). In addition, the RSQ rumination scale has been shown to relate to the rate of recovery from depression (Siegle, Sagratti, & Crawford, 1999) as well as depressive information processing biases (Siegle, 1999).

The Revised Impact of Event Scale (RIES; Horowitz, Wilner, & Alvarez, 1979) is a 15 item inventory that assesses the degree of intrusiveness of thoughts about a particular distressing event over the past seven days. It has intrusion and avoidance subscales. The intrusion subscale was selected as a potential measure of rumination. The questionnaire begins by requesting that the participant consider a stressful life event. When the questionnaire was normed, all participants had recently experienced the same stressful event (dissecting a cadaver). For the current questionnaire, participants generated their own stressful life event.

A 19 item questionnaire I have labeled Rumination on Negative Thoughts (RNT; Papageorgiou & Wells, 1999a,b) assessed dimensions of depressive and anxious thoughts about recent negative events (e.g., duration, degree of belief, degree of associated distress). Previous research with the questionnaire suggests that participants respond similarly for different negative events. Depressive symptomatology has been related to lower confidence in problem solving ability and past-orientation of thoughts using the measure, while difficulty dismissing thoughts and "meta-worry" have been associated with anxiety. The questionnaire was originally administered as part of a diary in which anxious and depressive thoughts were rated shortly after they occurred. The questionnaire was modified to reference a single depressive thought chosen by the participant, which had occurred in the past two weeks. Past research with this questionnaire has not employed a total score for the items since it was intended to assess different dimensions of rumination. As such, questionnaire items were subjected to principal components analysis with varimax rotation. Five factors had eigan values greater than 1, accounting for 61% of item variance. Of these, three factors passed a scree test, and together, accounted for 49% of item variance. Factor scores were preserved for further analysis. The first factor accounted for 30% of the variance and comprised 10 items that assessed feelings related to or arising from the thought (e.g., whether it made the individual sad). It was thus considered an emotion-focused factor (RNT-EMOT). The second factor accounted for 10.5% of the variance and comprised four items, three of which asked about the temporal orientation of the thought (past, present, and future) and was thus considered a temporal specifier (RNT-TIME). As it was unclear whether higher scores on this factor represented more rumination, the factor was not used as a quantitative index of rumination, it was not used in subsequent analyses. The third factor accounted for 9.5% of the variance and comprised four items, three of which assessed aspects of acting on or trying to solve problems associated with the thought, and was thus considered an instrumental factor (RNT-INS).

A 10 item questionnaire I have labeled Rumination on a Negative Event (RNE; Luminet, Rime, & Wagner, in prep) assesses a number of aspects of ruminative thoughts about a recent negative event such as their frequency, voluntariness, suddenness, dismissability, and intrusiveness. The questionnaire has been shown to demonstrate adequate reliability, and is related to the frequency with which individuals tap a button to represent thoughts about a negative event. The measure was developed for use in an experimental protocol in which participants had observed a particularly disturbing movie the day before, and referenced that movie. The questionnaire was modified to reference a negative event, chosen by the participant, that had occurred in the past two weeks. Past research with this questionnaire has not employed a total score for the items since it was intended to assess different dimensions of rumination. PCA of the items with varimax rotation yielded three items with eigan values greater than one, accounting for 59% of the variance. The first factor accounted for 34% of the variance and comprised 7 diverse items and was thus considered a general rumination factor (RNE-GEN). The second factor accounted for 14% of the variance and comprised three items including two that involved voluntary production of the thoughts and being alone when they occurred (RNE-INV). The third factor accounted for 10% of the variance and comprised a single item regarding attempts to dismiss the thoughts (RNE-DIS). Though the last two scales comprised few items, they were not absorbed in analyses constrained to one or two factors, and decreased the scale’s alpha when included in a total score. They were thus kept as separate scales.

A 27 item questionnaire (Multidimensional Rumination Questionnaire (MRQ); Fritz, 1999) assessed three potential subtypes of rumination in response to a particular stressful event including: "emotion-focussed rumination" (11 items assessing thinking about symptoms of depression), searching for meaning of negative experiences (5 items), and "instrumental rumination" (6 items; thinking about what can be done to change one’s situation in regard to negative events). The instrumental rumination and emotion-focussed rumination scales have both been shown to demonstrate adequate internal consistency. Emotion focussed rumination has been associated with neuroticism. All of the rumination scales were strongly intercorrelated. Fritz (in prep) has also demonstrated that emotion-focussed rumination is associated with delayed recovery from a first coronary incident while instrumental rumination is associated with quicker recovery. The "searching for meaning" subscale was originally administered in interview form to allow for free-form responses. These items were converted to questionnaire format using the exact text of Fritz’s interview questions; modifications were discussed with Fritz to make sure that they were in keeping with the original style of the questionnaire.

The Emotion Control Questionnaire (ECQ; Roger & Najarian, 1989) is a 56 item personality inventory designed to assess the degree to which individuals inhibit unwanted thoughts. It includes factor-analytically derived scales that nominally assess rehearsal (i.e., tending to think about negative events over and over), emotional inhibition (e.g., "I seldom show how I feel about things"), benign control (e.g., "I seldom feel irritable"), and aggression control (e.g., "If someone pushed me I’d push back"). The subscales demonstrate adequate internal consistency and test-retest reliability. The rehearsal subscale of the ECQ was chosen as a potential measure of rumination. It has been shown to correlate with physiological measures of sustained attention (e.g., Roger & Jamieson, 1988) as well as trait anxiety (Roger & Najarian, 1989). Fourteen items from the ECQ had rehearsal scale factor loadings above .3 and loaded uniquely on that scale; these items were used for item-analyses. Total scale scores were computed based on factor loadings reported by Roger & Najarian (1989).

The Thought Control Questionnaire (TCQ; Wells & Davies, 1994) is a 30 item inventory that assesses how people generally cope with intrusive thoughts. It contains factor analytically derived scales for distraction, social interaction, worry, (self) punishment, as well as a reappraisal scale that assesses the degree to which individuals focus on and revise their thoughts about stressful emotional events. The scales demonstrate adequate internal consistency and low inter-scale correlations. The worry (6 items), punishment (6 items), and re-appraisal (6 items) scales were selected as potential measures of rumination. Worry and punishment have been shown to correlate with trait anxiety, impaired mental control, and neuroticism. Re-appraisal has been shown to relate to self-consciousness.

In addition, to control for severity of depressive symptoms, participants received the Beck Depression Inventory (Beck, 1967), a 22 item self-report measure of depressive severity. The BDI has been validated as an indicator of dysphoria in undergraduate students (Beck, Steer, & Garbin, 1988).

Procedure. Participants received the questionnaires in a packet during group testing sessions, attended by up to 30 individuals. After receiving a verbal introduction to the study and signing their informed consent, participants were asked to respond to the questionnaires. Responses were entered on scantron sheets, to facilitate scoring.

Results

Analyses were performed on subscales that were selected to index rumination-like constructs. These included the RSQ rumination scale (henceforth, RSQRUM), the RIES intrusiveness scale, the ECQ rehearsal subscale (ECQ REH), the TCQ worry (TCQ WORRY), punishment (TCQ PUN), and re-appraisal (TCQ REAPP) scales, the MRQ emotion focussed scale (MRQ EMOTS), instrumental rumination scale (MRQ INST) and MRQ searching-for-meaning scale (MRQ SRCH), and the factor-analytically derived scales from the measures developed by Papageorgiou and Wells (RNT EMOT, RNT TIME, RNT INS), and Luminet et al (RNE GEN, RNE INV, RNE DIS). Because some scales were not originally designed to be used with summative total scores, item-level analyses were performed along with scale-total analyses to be sure that results were not due to within-scale variations in item-responding.

Generalizability Theory Analysis. Generalizability Theory (e.g., Matt, in press) can be used to examine the extent to which multiple measures index the same construct. ANOVAs are performed in which items and scale scores are treated as repeated measures, and each individual’s responses are entered as separate cases. The standard ANOVA table is used to partition the variance associated with different sources, including scale or item scores (MS within), individuals (MS between), and other factors (residual variance; MS error). The extent to which scale or item z-scores explain variation yields a measure of reliable distinctions between measures. The extent to which individuals explain variation yields a measure of the extent to which individual differences in a similar construct can be quantified by the examined measures. Residual variance is a function of individuals answering differently to different scales or items (i.e., scale by individual interactions).

Generalizability analyses were performed on both scale and item level responses. Since a number of individuals left items blank (primarily items on the RNE as per instructions to discontinue in the absence of ruminative cognitions), 112 individuals were used for the current analyses. The ECQ rehearsal scale was reverse-scored since it had a strong negative item-total correlation. Table 1 presents results from generalizability tests on the scaled scores as well as all items comprised in the scales. Results were very similar when the RNE was eliminated from the analysis, allowing a sample of 130 participants.

Source of Variation	MS	Component Variance	% Variance
SCALE SCORE ANALYSIS
Individuals (i)	3.05	=.147	15%
Scale Scores (s)	N/A
Residual (e)	.846	=.846	85%
Total	.988	=.988	100%
INDIVIDUAL ITEM ANALYSIS
Individuals (i)	1.55	=.012	5%
Item Scores (s)	3.51	=.030	14%
Residual (e)	.178	=.178	81%
Total	.219	=.220	100%

Note. The MS column represents the Mean Square from repeated measures ANOVAs on scale z-scores and dichotomized item scores as described in the text. Calculation of variance for each component proceeded as described in Matt (in press): =(MS_persons-MS_residual)/#items, =(MS_items-MS_residual)/ #persons, =MS_residual. %variance=component variance / total variance. The analysis was performed for both rumination scale scores and for all items comprising the scales. Variation attributable to scale-scores was not analyzed since entering z-scores into the analysis equated scale means.

While power was low to examine item-level responses this analysis was performed to examine whether drastic differences could be expected from the scale-level analysis. All items were dichotomized so that some items were not scored on a wider range than other items. Item-level analysis was consistent with the scale level analysis. As in the scale-level analysis, dependable individual differences, representing similar responses by individuals to many items, accounted for 5% of observed variation. Dependable differences in responses to different items accounted for 13% of the variation, suggesting that individuals generally responded higher to some items than others. The large amount of residual variation suggests that individuals did not respond consistently to the presented items. Thus, the analysis suggests that individuals indicate different levels of rumination based on which scale or item they are responding to. Together, these results suggest that the different items and different scales measure different constructs or types of rumination, rather than a single type of rumination. Thus, it may be useful to better understand what different qualities the scales measure.

Internal Consistency Analysis. To evaluate which scales were responsible for the observed inconsistencies in responding, internal consistency of the subscale scores was examined. Together the 14 z-scored subscales had a Cronbach’s alpha of .74, 95%CI=.67-.81 (.82, 95%CI=.76-.86 when total, rather than subscale scores from the RNT and RNE were used). Were subscales to represent the same construct, alpha would be expected to decrease when scales were eliminated from the total pool of scales, since longer scales are more reliable (Friedenberg, 1996). Alternately, if a scale measures something different from the construct assessed by other scales, alpha would be expected to increase upon its elimination. Table 3 presents the change in alpha associated with the deletion of each examined scale. As can be seen from the table, deletion of the RNT time distraction subscales, and the RNE temporal and instrumental scales increased internal consistency. Thus, to the extent that most scales measure a single construct, these scales are expected to measure different constructs from the rest of the scales. The magnitude of alpha change is subject to the range of scores on a scale and thus, is not interpreted.

Note. The first column represents the change in internal consistency (from .74) when the scale is removed. Were scales consistent, alpha would be expectd to decrease upon removal of subscales. The second column represents the correlation of the subscale score with the sum of all other subscales. The third column represents the internal consistency of each subscale.

Alpha for all items was .90 (.88 when the items were converted to dichotomous scores). As most of the scales were highly internally consistent, and scales had different lengths, this estimate of internal consistency is artificially inflated due to high representation of longer scales. Further analyses were thus conducted to examine ways in which scales and items tended to cluster.

Exploratory Principal Components Analysis (PCA) of Rumination-like Scale Scores To examine the degree to which items on rumination scales assessed the same or different constructs, each participant’s scores on the reliable administered rumination scales was entered into a principal components analysis (PCA), and were subjected to varimax rotation. This procedure groups items with high bivariate correlations together into maximally orthogonal factors. As such, if there is one type of rumination to which participants respond similarly, all items that measure rumination should load on a single factor. To the extent that multiple constructs are measured by the questionnaires, multiple factors would be assumed to emerge.

Three factors were extracted with eigen values greater than one, of which one passed a Scree test. The three factors accounted for 64% of the observed variation. Examination of factor loadings over 0.4 suggested each scale loaded on a single factor. Factor 1 represented all scales from the MRQ. Factor 2 included the RSQ rumination scale, the RNT emotion-focused scale, the RNE general rumination scale, the RIES intrusive thoughts scale, and the ECQ rehearsal scale, making it a general rumination scale. Factor three represented the TCQ worry and punishment scales. Thus, two factors represented individual scales while the other factor reflected a more general construct. Similarly, when all scales from the administered instruments rather than just the rumination scales were examined, scales from the same instruments tended to load on the same factors. When the larger sample who took all measures except the RNE was examined, two factors emerged. The first factor contained the MRQ subscales and the RIES intrusive thoughts scale. All other scales and the RIES intrusive thoughts scale loaded on the second factor, potentially suggesting a general rumination construct.

Exploratory Principal Components Analysis (PCA) of Rumination-Like Items. To address the potential heterogeneity among items within scales, a second exploratory factor analysis was performed in which items from each of the relevant scales was entered separately. While this analysis was underpowered given the current sample size, it was performed primarily to see whether items from different scales could be expected to cluster together in a larger sample. Eight factors were extracted with eigen values greater than one of which five passed a Scree test. A second analysis was thus performed restricted to five factors, which accounted for 37.8% of the total variance. Factor loadings from items on each of the scales are plotted in the figure below. Horizontal lines represent loadings greater than .4 and less than -.4. Examination of factor loadings above .4 and below -.4 suggested that scales generally loaded on different factors.

The first factor exclusively represented the majority of the items from the MRQ. The second factor contained primarily RSQ rumination items that dealt with thinking about feelings, and items from the ECQ rehearsal scale. It also included a few items from the RNT and RNE, potentially making it a general rumination-like scale. Factor 3 consisted entirely of items from the RNT. Factor 4 consisted of items from the RSQ rumination scale that named specific negative thoughts, and items from the TCQ worry, punishment, and reappraisal scales. Factor 5 consisted of all of the RIES intrusion items and the RNE items that did not reflect either active inhibition or social interaction. Using a loading threshold of .5 to determine factor membership lead to a more complete separation of factors by scales.

Relationship of rumination scales to depressive symptomatology. Pearson product-moment correlations and semipartial correlations between rumination scale scores and dysphoria, as measured by the BDI, are shown in Table 3. As can be seen from the table, depressive symptomatology was statistically significantly, though variably associated with all but one of the reliable subscales that was used to assess rumination (the TCQ re-appraisal scale showed the weakest relationship among reliable subscales). Dysphoria overlapped most highly with rumination as measured by the RSQ (tests of dependent correlations ranged from t_RSQ,TCQReapp(158)=5.45, p<.001 to t_RSQ,MRQemots(172)=1.77, p=.04), which, as shown by semi-partial correlation analysis, explained significant unique variation in dysphoria above and beyond all other scales. Though the TCQ-reappraisal scale explained little variance on its own, the variance in dysphoria it did explain was unique. Together, the rumination scales explained 38% of the variation in dysphoria (as compared to ~25% by the RSQ rumination scale).

Table 3: Correlations and semipartial correlations of rumination scores with depressive severity (BDI scores)
 

RSQ RUM

Rnt emot

Rnt inst

Rne gen

Rne inv

Rne dis

MRQ EMOTS

MRQ INST

MRQ SRCH

RIES INT

TCQ WORRY

TCQ PUN

TCQ REAPP

ECQ REH

Pearson Correlation

.543*

.400*

-.187*

.347*

-.023

.037

.429*

.245*

.269*

.253*

.222*

.193*

.142

.380*

Semi-partial corr

.19*

.09

-.03

.07

-.03

.02

.06

-.09

.08

.05

.08

-.05

-.21*

-.14

Preliminary conclusions

Analyses of both scale and item scores were consistent in revealing heterogeneity among measured constructs. Each scale appeared to measure some construct consistently (i.e., all scales had high alphas), yet a great deal more variation in responses was accounted for by inconsistent responding to the scales than by reliable individual differences among test-takers. Exploratory factor analyses similarly suggested that different scales reflected different constructs. It is thus suggested that future examinations of rumination employ more than one self-report measure, and that relationships among potentially different types of rumination be examined. Each scale was moderately to strongly correlated with depressive symptomatology. The scales accounted for different amounts of variation in depressive symptomatology, though the scales largely overlapped in the variance in depression for which they accounted.

Convergent evidence suggests sustained processing of emotional information seconds after it is presented shares features with the longer sustained processing associated with rumination (e.g., Luminet et al, 2000; Siegle, 1999a). Like rumination, emotional information processing biases are associated with elevated and prolonged sad mood as well as vulnerability to, and maintenance of clinical depression (e.g., MacLeod & Mathews, 1991), suggesting that it may share roles with those proposed for rumination.

To examine relationships between rumination and sustained processing of emotional information it is useful to present individuals with tasks that could spark such sustained processing. Based on computational models of interactions among relevant brain structures, Siegle (1999b) has suggested that a task in which individuals’ attention is directed towards emotional features of information may perform this function, and can be accomplished using an affective valence-identification task (e.g., Hill & Kemp-Wheeler 1989; Mathews & Milroy 1994; Siegle, 1999a, Siegle et al, 2000a,b), in which individuals are asked to name whether a word is positive, negative, or neutral. By allowing individuals to generate stimuli that are personally relevant and emotional for them, Siegle (1999b) suggests, the likelihood that sustained activity will be apparent on the tasks is maximized. Thus, tasks reported below used a combination of normed and personally relevant, idiosyncratically generated stimuli.

As a first attempt at this type of research, Siegle (1999a) suggested that rumination may contribute to interference effects on emotional information processing tasks. For example, Siegle (1999; Siegle et al, 1998) has shown that depressed and dysphoric individuals are slow to name the emotional valence of positive information, potentially as a result of ruminative cognitions during which a negative interpretation is put on traditionally positive material. A final potential conceptualization of rumination involves the extent to which individuals continue to devote cognitive resources to emotional information long after they have behaviorally reacted to it. To examine correspondences between these measures and self-report measures of rumination, 25 clinically depressed and 25 never depressed individuals’ reaction times were examined in response to an emotional valence identification task. The same individuals also received the Response Styles Questionnaire (RSQ; Nolen-Hoeksema et al, 1993). Rumination scores on the RSQ were related to delays in naming the valence of positive information r²=.27, F(1,20)=16.94, p<.0005, even after depression was controlled for, sr².076, F_change(1,44)=4.63, p_change=.037.

Yet, to more directly assess sustained processing, it is useful to examine whether rumination is related to what happens after individuals have responded to a task. Physiological measurement has been used to detect sustained processing of emotional information over non-emotional information up to five seconds after the onset of a stimulus (e.g., Cuthbert et al, 2000; Deveney et al, 1999). Sustained cognitive activity can be measured easily and non-invasively by tracking individuals’ pupil dilation during emotional information processing tasks. Many studies have demonstrated pupil dilation to be a reliable correlate of cognitive load, in that the pupil dilates more under conditions of higher attentional allocation, memory use, or interpretation of more difficult material (see Beatty 1982a, Steinhauer & Hakerem 1992, for reviews). As individuals are asked to remember larger numbers of digits, for example, their pupils reliably dilate (e.g., Kahneman & Beatty 1966; Granholm et al. 1996). Pupil dilation has also been observed to increase with the difficulty of mental arithmetic (Ahern & Beatty 1979; Hess & Polt 1964). Of particular interest is the finding that pupils remain dilated during conditions of sustained cognitive load (Beatty 1982b). Siegle (2000, in press) has shown that depressed individuals show higher levels of sustained processing of emotional information than never-depressed individuals, on multiple information processing tasks.

Measurement of areas active during cognitive and emotional processing can be accomplished using functional magnetic resonance imaging (fMRI). fMRI has, in the past, been shown to be sensitive to changes in processing of both nonemotional information (Carter et al, 2000) and emotional information (Davidson, 1998). Activation in brain areas associated with emotional processing has, in the past, frequently been observed using neuroimaging techniques in response to presentation of emotional information. Positron emission tomography (PET) has been used to detect amygdala and hippocampal activation (Reiman et al, 1997) as well as higher activation of the amygdala in response to negative than positive information (Paradiso et al, 1999), and higher activation of the insula in response to negative versus positive information (Lane et al, 1997). Cahill et al (1996) found correlations of amygdala and orbitofrontal activity during perception of emotional information with later recall of that information, but not for neutral information, suggesting that the amygdala is important in encoding affective aspects of information. fMRI has also detected amygdala activation in response to emotional stimuli versus neutral stimuli (e.g., Irwin et al, 1996). Teasdale et al (1999) find that emotional pictures paired with congruent emotional statements elicited insular, dorsolateral prefrontal, and anterior cingulate activity. Orbitofrontal activity has been found in response to presentation of negative, compared to neutral, stimuli (Northoff et al, 2000). New techniques for analyzing event related designs (e.g., Carter et al, 2000) allow examination of the time course of brain activation to individual trials, thus allowing differential examination of brain areas involved in early and late processing of emotional information.

Measurement of both fMRI and pupil dilation are useful in investigating rumination since fMRI provides excellent spatial localization, but sparse temporal resolution (one brain image every two to four seconds) while pupil dilation does not provide spatial localization but provides excellent temporal resolution (60 measurements per second). Thus, pupil dilation provides detailed information regarding the time-course of cognitive load (e.g., for investigating relationships between early and later processing biases); fMRI provides information regarding its location. The current study therefore examined individuals who are and are not depressed on an affective valence identification during measurement of both pupil dilation and fMRI. It was expected that individuals who ruminate will show sustained processing of emotional information, as evidenced by sustained pupil dilation and sustained brain activity in brain areas that process emotion, specifically, the amygdala. Based on the behavior of Siegle’s (1999b) computational model, it was further suggested that these effects should be strongest for negative words, especially if they are personally relevant.

Sustained pupil dilation in depression.

Depressed individuals demonstrate sustained processing. Siegle et al. (2000b) found that compared to 25 never-depressed individuals, 25 unmedicated depressed individuals showed sustained processing in response to an affective valence identification task, as measured by sustained pupil dilation seconds after the individuals had responded to task stimuli, as shown in the figure on the right. The same work demonstrated effects of emotional valence and interactions of valence and personal-relevance of stimuli on pupil dilation. Dilation was similar for depressed and non-depressed individuals on a non-emotional processing (cued-reaction time) task suggesting that the measure was sensitive and specific to emotional processing. These results suggest individuals with major depression process emotional information after other people have stopped processing it.

Sustained activity predicts interference with other processing. Recently (Siegle & Steinhauer, 2000), I have examined the extent to which emotional information processing biases interfere with subsequent non-emotional processing. Pupil dilation was measured in ten never-depressed and ten depressed individuals on a cued reaction time task, valence identification task (half of the words were generated from a normed corpus (Bradley & Lang, 1999) and half were idiosyncratically generated to be personally relevant), and emotional sentence rating task. Trials alternated between these tasks and Sternberg memory trials in which three numbers were presented followed by a target. Participants were instructed to say whether the target was in the set of three numbers. These data show that participants display sustained cognitive load on both the emotional sentence-rating task and the valence identification task. In this sample, depressed participants showed more sustained processing than controls, particularly for personally relevant words, F(1,16)=6.1, p=.025, as shown in the figure on the right. Moreover, sustained dilation for the valence identification task predicted an average of 38% of variance in sustained dilation on subsequent Sternberg trials after subtracting baselines (all average R² values and statistics calculated on R²s were computed on Fisher’s z-transformed correlations). Additional variation in Sternberg rt's (18% total) and pupil dilation (53% total) was explained by interactions of the previous trial’s pupil dilation and emotional valence, upon aggregating within-subject hierarchical regressions. Thus, task-relevant sustained-processing of emotional aspects of information is measureable up to 15 seconds after a stimulus is presented. Only one subject showed a significant relationship of the words’ arousal ratings to sustained dilation, suggesting that obtained results are not due to arousal but rather to some other aspect of emotional processing. Moreover, the word’s valence and its interaction with dilation on the valence identification trials accounted for 10% more variation in depressed individuals’ subsequent Sternberg trial dilation than in controls, t(19)=2.4, p=.025, suggesting that emotional content is associated with more interference for depressed than other individuals.

Relationship of sustained dilation to depressive rumination. Relationships between rumination, assessed using a variety of self-report measures, and pupil dilation were examined in my recent data. Biases were defined as late pupil dilation for personally relevant negative words minus the same measure for non-personally relevant positive words, as this contrast was hypothesized to engender the largest difference in processing for ruminators. Some rumination measures were strongly correlated (r>.5) with pupil dilation biases both 4 seconds after the stimulus, and 4 seconds after the subsequent Sternberg memory trial stimulus. Depressive symptomatology accounted for 17% of the variation in late pupil dilation. Fritz's (1999) multi-dimensional rumination measure accounted for 20% more variation (38% total) in late pupil dilation. Similarly, the "worry" scale on Wells and Davies' (1994) Thought Control Questionnaire accounted for 7% of variation in pupil dilation above and beyond depressive symptomatology. Thus, to the extent that these relationships replicate, rumination may be considered an interesting clinical correlate of sustained processing in depression. Correlations of late dilation on each trial, and each subsequent Sternberg trial are shown below.

Brain activation associated with sustained processing of emotional information in depression

 Depressed individuals demonstrate sustained amygdala processing for negative information. To examine brain structures involved in sustained processing in depression, the same tasks (alternation of Sternberg trials with valence identification, emotional-sentence rating, and cued reaction time) were re-administered to six depressed and five never depressed individuals on whom pupil dilation was measured. In this session fMRI data was collected (twenty-six coronal 3.8mm slices every 4 seconds acquired perpendicular to the AC-PC line using a 2-interleave spiral pulse sequence, TR=2000ms, TE-35ms, FOV=24cm, flip=70 on a 1.5T GE scanner). Multiple analyses lead to convergent findings.

Statistical analyses used software developed locally through the Human Brain Project. fMRI data was cross-registered to (i.e., warped to conform to the shape of) the same space. Voxels in each subject’s cross-registered fMRI data were subjected to mixed model ANOVAs with subject as the random factor and scan, valence, and personal relevance as fixed factors. Random effects analysis permits generalization of results at the population level and hence, is well suited to clinical studies. Voxels were identified in which effects of congruence were detectable at p<.05, corrected for multiple comparisons. An amygdala particle responded more strongly to personally relevant than non-personally relevant words in depressed individuals. In this group, sustained activation was highest for negative words. In contrast, in control participants, sustained activation was highest in response to positive words. To more carefully examine activation in specific structures, the structures were traced on each subject's structural MRI’s. Principal Components Analysis was used to identify clusters of voxels in the structures with similar time-courses. The first extracted factor was interpreted as a general index of processing. Plots of factor scores for each valence, at each scan were examined. On the valence-identification task, for 4/6 depressed participants, sustained activation (at least 6 scans, or 24 seconds), as indexed by high scores on the first extracted factor, was observed for negative, but not positive or neutral words. This pattern was not present in any of the controls. In contrast, 2/3 controls demonstrated sustained processing of positive, information; this pattern was not present in any of the depressed individuals. Activation to negative stimuli in the amygdala was subjected to a hierarchical regression in which activation to positive stimuli was entered on the first step, accounting for 12% of the variation, and group was entered on the second step, accounting for 17% more variation. Thus, both exploratory and confirmatory analyses suggest depressed individuals show greater sustained amygdala activation for negative than positive words compared to healthy controls. Time courses for the group and representative single subjects are shown on the right.

Sustained amygdala processing is related to rumination. Similar to pupil dilation data, self-reported rumination was strongly related to sustained amygdala activation. Specifically, in the individuals for whom fMRI assessment was performed, 17% of variation in the amygdala particle’s response to negative versus positive words on scan 5 was accounted for by depression. An additional 75% of variation (92% total) was accounted for by adding the emotional coping scale of Fritz’s (1999) multidimensional rumination measure and Luminet’s (submitted) measure of rumination on a negative thought.
 
 

These results suggest many constructs have been labeled rumination. They overlap but preserve unique variance. Thus, different types of rumination may differ in their contribution depression. Measuring cognitive and physiological processes associated with rumination may help to relate information processing biases to the maintenance of depression. It is suggested that future studies that empirically examine rumination clearly define the type of rumination being examined, attend to construct validity in operationalizing this definition, and discuss aspects of depression with which it is theoretically expected to be associated.

References

Ahern, S.K., & Beatty, J. (1979). Physiological signs of information processing vary with intelligence. Science, 205, 1289-1292.

Beatty, J. (1982a). Task-evoked pupillary responses, processing load, and the structure of processing resources. Psychological Bulletin, 91, 276-292.

Beatty, J. (1982b). Phasic not tonic pupillary responses vary with auditory vigilance performance. Psychophysiology, 19, 167-172.

Beck, A. T. (1967). Depression: Clinical, experimental, and theoretical aspects. New York: Hoeber.

Beck, A., Steer, R., & Garbin, M., (1988). Psychometric properties of the Beck Depression Inventory: Twenty-five years of evaluation. Clinical Psychology Review, 8, 77-100.

Bradley, M. M. & Lang, P. J. (1997) Affective Norms for English Words (ANEW): Technical Manual and Affective Ratings. Gainsville, FL. The Center for Research in Psychophysiology, University of Florida.

Cahill L., Haier R.J., Fallon J., Alkire M.T., Tang C., Keator D., Wu J., McGaugh J.L. (1996). Amygdala activity at encoding correlated with long-term, free recall of emotional information. Proceedings of the National Academy of Sciences of the United States of America, 93, 8016-21.

Carter C.S., Macdonald A.M., Botvinick M., Ross L.L., Stenger V.A., Noll D., Cohen J.D. (2000). Parsing executive processes: strategic vs. evaluative functions of the anterior cingulate cortex. Proceedings of the National Academy of Sciences of the United States of America, 97, 1944-8.

Cronbach, L.J., Gleser, G.C., Nanda, H., & Rajaratnam, N. (1972). The dependability of behavioral measurements: Theory of generalizability for scores and profiles. New York: John Wiley.

Cronbach, L.J., Nageswari, R., & Gleser, G.C. (1963). Theory of generalizability: A liberation of reliability theory. The British Journal of Statistical Psychology, 16, 137-163.

Cuthbert, B. N., Schupp, H. T., Bradley, M. M., Birbaumer, N., & Lang, P. J. (2000). Brain potentials in affective picture processing: Covariation with autonomic arousal and affective report. Biological Psychology, 52, 95-111.

Davidson, R. J. (1998). Affective style and affective disorders: Perspectives from affective neuroscience. Cognition & Emotion. 12(3), 307-330

Deveney, C. M., Deldin, P. J., Kim, A. S., Best, J. L. (1999). Memory processing biases in mood disorders: A slow wave ERP investigation. Presentation at the meeting of the Society for Physiological Research.

Friedenberg, L. (1995). Psychological testing: Design, analysis, and use. Needham Heights, MA: Allyn & Bacon.

Fritz, H. (1999). The role of rumination in adjustment to a first coronary event. Unpublished doctoral dissertation, Carnegie Mellon University, Pittsburgh, PA.

Granholm E, Asarnow RF, Sarkin AJ, Dykes KL (1996): Pupillary responses index cognitive resource limitations. Psychophysiology 33:457-461.

Hess, E.H. & Polt, J.H. (1964). Pupil size in relation to mental activity during simple problem solving. Science, 143, 1190-1192.

Hill AB & Kemp-Wheeler SM (1989): The influence of anxiety on lexical and affective decision time for emotional words. Personality and Individual Differences 10:1143-1149.

Horowitz, M. J., Wilner, N., Alvarez, W. (1979). Impact of Event Scale: A measure of subjective stress. Psychosomatic Medicine, 41(3), 209-218.

Irwin W,, Davidson R.J., Lowe, M.J., Mock, B.J., Sorenson, J.A., & Turski, P.A. (1996). Human amygdala activation detected with echo-planar functional magnetic resonance imaging. Neuroreport, 7, 1765–1769.

Kahneman D, Beatty J (1966): Pupil diameter and load on memory. Science 154:1583-1585.

Lane R.D., Reiman, E.M., Ahern, G.L., Schwartz, G.E., & Davidson, R.J. (1997). Neuroanatomical correlates of happiness, sadness, and disgust. American Journal of Psychiatry, 154, 926–933.

Luminet, O., Rime, B., & Wagner, H. (in prep). Intrusive thoughts in the laboratory and their long-lasting consequences.

Lyubomirsky, S., Caldwell, N. D,, & Nolen-Hoeksema, S. (1998). Effects of ruminative and distracting responses to depressed mood on retrieval of autobiographical memories. Journal of Personality and Social Psychology, 75, 166-177.

Matt, G. E. (in press). Generalizability Theory. In N.J., Smelser, & P.B. Baltes, (Eds).
International Encyclopedia of Social and Behavioral Sciences. Amsterdam: Pergamon/Elsevier.

Mathews A, & Milroy R (1994): Processing of emotional meaning in anxiety. Cognition and Emotion 8:535-553.

Nolen-Hoeksema, S. (1991). Responses to depression and their effects on the duration of depressive episodes. Journal of Abnormal Psychology, 100, 569-582.

Nolen-Hoeksema, S. Morrow, J. & Fredrickson, B. L. (1993). Response styles and the duration of episodes of depressed mood. Journal of Abnormal Psychology, 102, 20-28.

Northoff, G., Richter, A., Gessner, M., Schlagenhauf, F., Fell, J., Baumgart, F., Kaulisch, T., Koetter, R., Stephan, K. E., Leschinger, A., Hagner, T., Bargel, B., Witzel, T., Hinrichs, H., Bogerts, B., Scheich, H., Heinze, H. (2000). Functional dissociation between medial and lateral prefrontal cortical spatiotemporal activation in negative and positive emotions: A combined fMRI/MEG study. Cerebral Cortex, 10, 93-107.

Papageorgiou, C. & Wells, A. (1999). Process and Meta-Cognitive Dimensions of Depressive and Anxious Thoughts and Relationships with Emotional Intensity. Clinical Psychology and Psychotherapy, 6, 152-162.

Paradiso, S., Johnson, D. L., Andreasen, N. C., O'Leary, D. S., Watkins, G. L., Boles Ponto, L. L., Hichwa, R. D. (1999). Cerebral Blood Flow Changes Associated With Attribution of Emotional Valence to Pleasant, Unpleasant, and Neutral Visual Stimuli in a PET Study of Normal Subjects. American. Journal of Psychiatry 156, 1618-1629.

Reiman, E.M., Lane, R.D., Ahern, G.L., Schwartz, G.E., Davidson, R.J., Friston K.J., Yun, L.S., Chen, K. (1997). Neuroanatomical correlates of externally and internally generated human emotion. American Journal of Psychiatry, 154, 918-925.

Roger, D. & Najarian, B. (1989). The construction and validation of a new scale for measuring emotion control. Personality & Individual Differences, 10 (8), 845-853.

Siegle G.J. (1999a). Cognitive and Physiological Aspects of Attention to Personally Relevant Negative Information in Depression. Unpublished Doctoral Dissertation. San Diego State Univeristy, University of California, San Diego. This document is also available on the World Wide Web at www.sci.sdsu.edu/CAL/greg/dissert/.

Siegle, G. J. (1999b). A neural network model of attention biases in depression. In J.A. Reggia, E. Ruppin, D. L. Glanzman (Eds.), Disorders of brain, behavior, and cognition: The neurocomputational perspective. (pp. 415-441). NJ: Elsevier.

Siegle, G. J. (2000, submitted). Convergence and divergence in self-report measures of rumination and its relationship to depressive symptomatology.

Siegle, G. J., Ingram, R. E., & Matt, G. E. (2000a, in press). Affective interference: Explanation for negative information processing biases in dysphoria? Cognitive Therapy and Research.

Siegle, G. J., Granholm, E., Ingram, R. E., & Matt, G. E. (2000b, in press). Pupillary response and reaction time measures of sustained processing of negative information in depression. Biological Psychiatry.

Siegle, G. J., Sagratti, S., & Crawford, C. (1999). Effects of rumination and initial severity on response to cognitive therapy for depression. Meeting of the Association for the Advancement of Behavior Therapy. Toronto, ON.

Siegle, G. J. & Steinhauer, S. R. (2000). Pupillary and reaction-time assessment of interference of emotional and non-emotional information on subsequent processing. Presentation at the meeting of the Society for Psychophysiological Research, San Diego, CA.

Steinhauer S.R. & Hakerem G. (1992): The pupillary response in cognitive psychophysiology and schizophrenia. In D. Friedman & G. E. Bruder (Eds), Psychophysiology and Experimental Psychopathology: A Tribute to Samuel Sutton. (pp. 182-204). New York: New York Academy of Sciences.

Teasdale, J. D., Howard, R. J., Cox, S. G., Ha, Y., Brammer, M. J., Williams, S. C. R., & Checkley, S. A. (1999). Functional MRI study of the cognitive generation of affect. American Journal of Psychiatry, 156, 209-215.

Wells, A. & Davies, M. I. (1994). The thought control questionnaire: a measure of individual differences in the control of unwanted thoughts. Behaviour Research and Therapy 32, 871-878.
 
 

This research was supported by NIMH grant MH30914 through the UCSD Mental Health Clinical Research Center, NIMH training grant MH16804, MH55762, the Veterans Administration, UCSD Academic Senate grant 444972-RW178M, and a Sigma Xi Grant in Aid of Research to Greg Siegle.