Affective Interference: An Explanation for
Negative Attention
Biases in Dysphoria?
Greg J. Siegle
Western Psychiatric Institute and Clinic, VA Pittsburgh
Healthcare System
Rick E. Ingram
Southern Methodist University
San Diego State University
Georg E. Matt
San Diego State University
Running head: AFFECTIVE INTERFERENCE IN DEPRESSED
STATES
Key words: Depression, Information Processing,
Attention, Rumination
in press, Cognitive Therapy and Research
Thanks to Monica Barback who provided valuable
feedback on earlier versions of this manuscript, and Mark Shibley, Maureen
Flaherty, Sean Gyll, Danielle Grant, and Ivan Nepomunceno who aided in the
study's conceptualization and data collection.
Correspondence concerning this article should be
addressed to the first author: Greg Siegle, Ph.D., Western
Psychiatric Institute and Clinic, 3811 O’Hara St, Pittsburgh, PA 15213, e-mail:
gsiegle+@pitt.edu, fax: 412-365-5259.
This research was supported in part by NIMH grants
MH16804, MH55762 and the Veterans Administration.
Research
suggests that individuals with features of depression pay excessive attention
to negative information. Yet, it is unclear what aspects of negative
information are attended to by these individuals. Different answers to this
question suggest different roles for attention in the onset and maintenance of
depressive states. This study investigated aspects of emotional information to
which college students with and without features of depression attend. Research
participants completed an affective lexical decision task and an affective
valence identification task. Dysphoric
individuals were slow to identify the emotional valence of positive information, and nonemotional aspects of negative
information (the lexicality of negative words), but were not slow to identify
the emotional valence of negative words. To explain these results, aAn "affective-interference"
hypothesis is advanced to explain these results, which. suggests
that dDysphoric individuals are proposed to attend to the emotional
content of negative information at the expense of attending to other aspects of
the information. Results are related to
theories of ruminative coping with depression.
Affective Interference: An
Explanation for Negative Introduction
Attention Biases in Depressed States?
Attention
to emotional information is central to many theories of the onset and
maintenance of depression (e.g., Beck, 1967, 1974; Ingram, 1984; Ingram,
Miranda, & Segal, 1998). Support for these theories is found in literature
suggesting that dysphoric and depressed individuals disproportionately attend
to and remember negative information (Blaney, 1986; Matt, Vazquez, &
Campbell, 1992; Matthews & Harley, 1996; Williams, Mathews, & MacLeod,
1996). Yet, it is unclear to which aspects of negative information dysphoric
and depressed individuals attend. It is also unclear, and whether biased attention to
negative information occurs in the early stages of attention, having to do with
initial perceptions of information, or in late stages of attention, involving
retrieval of associations from memory (Macleod & Mathews, 1991). Clearly
specifying these aspects of the relationship between mood and attention to
emotional information may lead to a better understanding of the role of information
processing in affective psychopathology. .
This paper will focus specifically on attention biases
in n dysphoric individuals who are dysphoric (have sad mood states thought to
underlie depression) rather than , rather than individuals diagnosed
with clinical depression. Focussing on dysphoria, or sad mood states
thought to underlie depressive states, allows investigation of the
relationships between mood and attention without introducing to avoid confounding features associated
with clinical diagnoses (Persons, 1986). That is, since cognitive biases are expected to be
associated with prolonged sad mood, such biases may not be characteristic of
all clinically depressed individuals. Similarly, there is little reason to
believe that all people who have cognitive features of depression will meet
formal DSM IV criteria for major depression. Thus, it might be easier to detect
emotional information processing biases
in dysphoric people who do not necessarily meet criteria for depression than in
a less homogeneous group of clinically depressed individuals.
Theories
of emotional information processing from cognitive and physiological psychology
may help to resolve ambiguities regarding the roles of cognitive and emotional
information processing in dysphoric individuals. They suggest that emotional
aspects of information (e.g., whether it is positive or negative;, for example, the notion that
"birthday" is positive) and semantic (conceptual, nonemotional)
aspects of the same information (e.g., the notion that "birthday" is
the day on which one is born) can be processed in parallel by different highly
inter-connected physiological (LeDoux, 1996; Tucker & Derryberry, 1992) and
cognitive (Bower, 1981; Ingram, 1984) systems. Attention can be differentially allocated
to these aspects of information (Kitayama, 1990; Matthews & Harley, 1996).
If
emotional and non-emotional aspects of information are processed in parallel then, information-processing biases could
occur as a function of attention to either affective aspects of
information, or nonaffective aspects of information,, or feedback between systems
responsible for representing affective and nonaffective aspects of information.
Depending on how much feedback occurs between the systems,, attending to the affect associated
with information could interfere with normal semantic associations (e.g.,
Siegle, 1999). This phenomenon will be referred to as “affective interference.”
Affective interference might involve responding to the word “death” by
immediately recognizing that the stimulus is negative, and then thinking of other negative
things unrelated to death such as one’s low self-image.
Attention
to either the emotional and and nonemotional aspects aspects of information can be examined
separately by observing how quickly individuals respond to questions that direct their attention to relevant features regarding these aspects of a stimulus.
Reaction times have long been assumed to reflect the amount of effort an
individual pays to information., and tThus, longer reaction times are traditionally
associated with paying less attention to a task (Massaro, 1988). By comparing
individuals' reaction times to questions regarding emotional and non-emotional
aspects of a stimulus,, relative attentional allocation to
each of these components can be measured.
Two
reaction time tasks can thus be used to determine whether attending to
emotional aspects of information impairs attention to non-emotional aspects of
information. A “valence identification” task requires participants to identify
the emotion associated with positive, negative, and neutral information. It
thus directs participants’ attention towards the emotional aspects of stimuli.
Participants’ reaction times are presumably related to how much attention they
pay to emotional aspects of information. In contrast, an “affective lexical
decision” task requires that participants to judge whether a string of letters spells a
word that, which may be positive negative, or
neutral. This task directs participants’ attention towards non-emotional
aspects of information.
If
dysphoric individuals attend to,, and make associations with,, emotional aspects of negative
information at the expense of attending to its non-emotional aspects, they
would be expected to respond especially quickly to negative information on a
valence identification task. Because of affective interference, they would be expected to react
especially slowly to negative information on a lexical decision task. Feedback
between structures responsible for processing affective and nonaffective
features could exaggerate these biases. For example, a dysphoric individual who
attends to negative aspects of information may respond to the typically
positive stimulus “birthday” by associating it with the (semantic) belief “I am
getting old.” This thought may trigger the person’s recognition of negativity.
The dysphoric indivdualindividual may thus be especially slow to
identify positive information as positive, but, by the same token, be very
quick to identify negative information as negative. Put another way, research
has long suggested that the more meaningful information is to an individual,
the more “deeply” it is processed (e.g., Craik, Lockhart, & Tulving, 1977).
If depressed dysphoric individuals are especially
well-practiced at identifying negativity, or if depressed individuals find negative
aspects of information to be particularly meaningful, they may process
emotional aspects more deeply than semantic aspects. Such a strategy, and thus could allowing affective aspects of information
to interfere with perception of other (e.g., semantic) aspects of that
information.
While Though the combination of these tasks is
fairly novel,, both tasks have been examined in the
past. Valence identification tasks have been used with anxious individuals
(Hill & Kemp-Wheeler, 1989; Mathews & Milroy, 1994), and with threat words with depressed
populations (Hill & Kemp-Wheeler, 1989), but not with depressed or dysphoric
individuals, using positive, negative, and
neutral words as stimuli. In contrast, a number of researchers have employed
affective lexical decision tasks with nondysphoric, dysphoric, and depressed
individuals. Most of these researchers
predicted that depressed and dysphoric individuals would respond more quickly
to negative words than to non-negative words on this task (e.g., Challis &
Krane, 1988; Macleod, Mathews, & Tata, 1986; Matthews & Southall, 1991;
Ruiz Caballero & Bermudez Moreno, 1992). This hypothesis was not confirmed
for any of these studies.
Moreover,
when the studies are looked at together, as summarized in Table 1, a conclusion
consistent with the notion of affective interference emerges (see Siegle, 1996,
for a complete review of these data). A literature search revealed seven
lexical decision task studies of nondepressed individuals and four studies
involving depressed individuals, in which priming was not employed[1],, and in which reaction times to
negative and neutral words were reported. Whereas nNo studies observed a significant difference
in nondepressed individuals' reactions to negative and neutral words. In contrast,, all studies of depressed individuals
suggested that they were slightly slower to respond to negative than neutral
words (meta-analytic CI=2.26-32.72 ms). Thus, these studies, generally show the opposite pattern
from what is usually predicted. Depressed individuals react more slowly to
negative than to neutral words, whereas nondepressed individuals do
not. This finding is consistent with the idea that the affect associated with
negative information interferes with the recognition of non-affective aspects
of negative information.
---------------------------------------------------------------
Table 1
Negative rt - Neutral rt (D) for nondepressed
and depressed individuals, in milliseconds for published affective lexical
decision tasks
|
Study |
Nondepressed |
Depressed |
|
Bradley et al., 1994 |
5 |
3 |
|
Bray, 1984 |
10 |
|
|
Challis & Krane,
1988 |
-62 |
|
|
Macleod, Tata, &
Mathews, 1987 |
10 |
29 |
|
Matthews &
Southall, 1991 |
SOA=1500 ms: 18 SOA=240 ms: 29 |
SOA=1500 ms: 76 SOA=240 ms: -50 |
|
Matthews, Pitcaithly,
& Mann (1995) |
SOA=1500 ms: -13 SOA=240 ms: -14 |
|
|
Stip et al, 1992,
1994 |
-8 |
30 |
|
Williamson et al.,
1991 |
-46 |
|
|
Meta-analytically
derived mean effect size3[2] |
MD=2.07, SDD=4.21 |
MD 17.49, SDD=7.77 |
Insert Table 1 About Here
---------------------------------------------------------------
The
affective lexical decision and valence identification tasks are probably not
pure, uncorrelated measures of affective
and nonaffective processing. Rather, it is predicted that dysphoric individuals
will be faster to name the valence of negative than positive or neutral
information, but will be slower to name the lexicality of negative words, as a function ofdue to feedback between systems involved in
affective and semantic determination. To test these hypotheses, the current
experiment contrasted the performance of dysphoric and nondysphoric individuals
on a lexical decision and valence identification task. By analyzing the tasks
separately it can be determined whether the general patterns of interest are
present in the data. By correlating performance on the tasks, a determination of whether
increasing reaction times on one task are associated with decreasing reaction
times on the other can be made.
A
multiple gating procedure was used for selection of research participants to
ensure that they had a stable mood over a number of weeks. Research
participants were selected from a pool of over 3000 students in an introductory
psychology course at San Diego State University. Dysphoric individuals were
selected for having a Beck Depression Inventory (BDI; Beck, 1967) score of over
15 (a score indicating moderate depressive symptomatology)., and nNondysphoric individuals were selected for
having a BDI score under 6 at a mass screening at the beginning of the
semester. Participants received class credit for their participation. In
total, 138 students participated in the experiment,, at least two weeks after the mass
screening. In the dysphoric group of participants, who had received BDI scores over
15 at the screening, 30 research participants had BDI scores
over 16 when they came in for testing (M=22.8, SD=4.11) and were
used in subsequent analyses. The increased cut-off was adopted to reflect the
highest possible level of depressive severity upon testing that preserved an
adequate sample size, derived through power analyses. They Dysphoric participants included ten men and
twenty women, of whom seven identified as Asian,
fourteen as Caucasian, seven as Hispanic, and one as having another
ethnicity. In the nondysphoric group,, 46 of the students who had BDI
scores between 1 and 6 at the screening were still in this range at testing (M=3.00,
SD=1.63) and were thus used in further analyses. They included
twenty-one men and twenty-five women of whom one identified as African
American, six as Asian, twenty-two as Caucasian, eleven as Hispanic, and one as
having another ethnicity. Distributions of ethnicity were not significantly
different between the groups, C2(4)=2.85, p=.65.
The
Beck Depression Inventory (BDI) was used to assess depressive symptomatology.
The BDI is a 21 item self-report inventory.
Individual items are worded so that responses reflect increasing degrees
of severity and are given the values of 0-3. The total possible score ranges
from 0-63. The inventory is frequently used to assess
depressive symptoms in college populations. It has acceptable validity and
reliability (Beck, 1967, Beck, Steer, & Garbin, 1988). The BDI was chosen
rather than its successor,, the BDI-II, as the study was begun before the BDI II
was published.
Stimuli
for a lexical decision and valence identification task were displayed on an IBM
PC compatible 486 computer with a 14 inch14-inch color monitor. Research participants
sat approximately 28 inches from the bottom of the stimulus. Stimuli were drawn
in lowercase letters approximately 5/8 inches high on the monitor, subtending
approximately 1.21 degrees of visual angle.
For
both computer administered tasks, two sets of 10 positive, 10
negative, and 10 neutral words, balanced for normed affect, word
frequency, and word length were chosen from Siegle’s (1994) corpus. This corpus
was compiled from published studies that employed normed word lists and pilot
data collected for this experiment. Examples of positive words include “happy”
and “bliss”. Examples of negative words include “hopeless” and “ashamed”.
Examples of neutral words include “library” and “slope”. The positivity and
negativity of words on the first word list were rated by 530 undergraduates.
Analysis showed that the each of the words groups strongly and significantly
differed from each other in both positivity and negativity ratings. , and that dDysphoric and nondysphoric undergraduates did
not tend to differ in their valence ratings (Williams et al, 1998). In that
sample, negative words were rated farther
from neutral words than neutral words were from positive words,, on negativity.; Llikewise, positive words were rated
farther from neutral words than neutral words from negative, on positivity,
suggesting that valences differed in the expected direction and magnitude.2[3]
For the lexical decision task, 15 nonwords, consisting of single letter
perturbations of equal numbers of normed positive, negative, and neutral words,
were also included (e.g., "cousip", "mendion").
For
the lexical decision task, a row of eleven X’s was present on the screen at the
beginning of the each presentation. After 2000ms, the X’s in the middle of the
string were replaced by letters spelling a word or nonword, with the X’s on
either side remaining on the screen. After the stimulus duration, the letters
were again masked by X’s. The question “Is it a word?” appeared on the screen.
The same procedure was used for the valence identification task, except that the participants responded to the question “What’s the
Valence?”
To
increase the chances that internal feedback processes would be engaged during
stimulus recognition,, leading
to affective interference, it was decided to slightly obscure stimuli
were slightly obscured by using a short
stimulus duration. Just the same, shorter stimulus durations were assumed to
add more noise (i.e., random variation) to reaction times, due to incurred perceptual
difficulties. To balance these concerns, pilot data was collected using a
stimulus duration of 150ms, which revealed relevant effects; this duration was
thus chosen for the experiment. As a sensitivity check, two shorter durations
(100ms, 50ms) were also included, to make sure that there would be
some duration short enough to allow cognitive effort sufficient to reveal
biases. These conditions were only to be analyzed , with the intent to only analyze these conditions as
a way of understanding whether stimulus duration contributed to null effects,
if they were obtained.
Potential
research participants were invited by phone to participate in an experiment
involving the perception of words. After obtaining informed consent, directions
for the lexical decision task and valence
identification tasks were shown to research participants on the computer for the lexical decision task and valence
identification tasks. Research participants were then shown the
directions again for the first task that they would complete. Participants were
given a practice session consisting of fourthree trials that were repeated until they
were correctly performed. After querying for any questions, participants
completed the first task. The identical
procedure was used for the second task. In both tasks, each stimulus (10 positive words, 10
neutral words, 10 negative words, 15 nonwords) was presented at stimulus
durations of 50ms, 100ms, and 150ms for a total of 135 trials.
The
lexical decision task was conducted in the following manner. The research
participant was asked to place his or her right palm below the keyboard with
the index finger poised above the space bar, equidistant from the “N” and “M”
keys. Each word and nonword were
presented to the participant as described previously, at each stimulus
duration. For each stimulus, the research participant pushed a
buttons for “Yes” or “No”, labels for which were placed over the “N” and “M”
keys on the keyboard. The research participant’s reaction time and response was
recorded on the computer for each stimulus.
The
valence identification task was conducted in the following manner. The research participant was asked to place
his or her right palm below the keyboard with the index finger poised above the
space bar, equidistant from the “z”, “x”, and “c” keys. Each word, at each stimulus duration was
presented at each stimulus duration to the participant
as described previously. In response to each word, the research participant pushed
buttons labeled “+”, “-‘’, or “N” (standing for “Positive”, “Negative”, or
“Neutral”, respectively), which were placed over the “z”, “x”, and
“c” keys. Different keys were used for
the different tasks so that all keys could be labeled at the beginning of the
experiment.
After
the completion of both tasks the BDI was administered on a computer and
participants were debriefed. The order in which tasks were completed, the order
in which buttons for responses on the keyboard were labeled for the tasks, and
the assignment of which word list was used for which task were counterbalanced
using a strategy that ensured a balanced sampling of all possible combinations
as the number of participants increased. The order of words within each task,
and the order in which a given word was shown in a given stimulus duration were
assigned pseudo-randomly by computer (i.e., random to the extent possible using
a computer-based random number generator) for each participant.
Mean
reaction times to each valence and stimulus duration on each task were
calculated. Reaction times below 150ms were discarded as outliers because
previous results suggest that reaction times in this range indicate that a
response was made without regard for the stimulus (Matthews & Southall,
1991). Similarly, responses over 5000ms were discarded as outliers under the, assumptioning that such long response latencies
indicated inattention to the trial. This relatively long cut-off was adopted
since the magnitude of valid delays in reaction times due to interference
effects was unclear a priori. Reaction times to stimuli that were
incorrectly identified on the lexical decision task,, and reaction times to stimuli whose
valence was identified as incongruent with the normed valence on the valence
identification task were not removed from the computation of means. Reaction times on the valence identification
task were not significantly skewed, skew=.40, SEskew=.28.
Though reaction times on the lexical decision task were more skewed, skew=1.90,
SEskew=.277, no normalization transformation was applied so
as to allow comparisons between effects on the tasks and with other relevant
literature. One nondysphoric participant’s valence identification task data
were corrupted, and were therefore not included in
analyses or the table. The performance of dysphoric and nondysphoric
individuals on the lexical decision and valence identification tasks are shown
in Table 2.
---------------------------------------------------------------
Table 2
Mean reaction times for the 150ms condition for
nondysphoric (Nlexical decision=46, Nvalence identification=45)
and dysphoric (N=30) individuals, in milliseconds
|
|
|
NonDysphoric |
Dysphoric |
||
|
Task |
Valence |
Mean |
St Dev |
Mean |
St Dev |
|
Lexical Decision |
Positive |
481 |
175 |
506 |
317 |
|
|
Negative |
457 |
178 |
543 |
294 |
|
|
Neutral |
503 |
149 |
509 |
249 |
|
|
Nonword |
642 |
235 |
685 |
324
|
|
Valence
Identification |
Positive |
677 |
252 |
826 |
314 |
|
|
Negative |
657 |
273 |
660 |
242 |
|
|
Neutral |
838 |
337 |
863 |
294
|
Insert Table 2 About Here
---------------------------------------------------------------
Planned
contrasts were performed to test the following hypotheses: 1) that dysphoric
people would respond faster to negative words than nonnegative words on the
valence identification task, 2) that dysphoric people would respond slower to
negative than nonnegative words on the lexical decision task, and 3) that
differences in reaction times to negative and non-negative words would be
greater for dysphoric than for nondysphoric people. Based on pilot studies, the
150ms condition was hypothesized to be fast enough to necessitate cognitive
effort in identifying its valence, but slow enough to be reliably
recognized. The 150ms condition was therefore used for all contrasts.
Family-wise alpha was controlled at 0.05 using a Bonferroni adjustment for each
family of contrasts. Contrasts
for each task were analyzed as representing different families of tests. Exact p values are reported to allow
interpretation of significance using the applied or other significance
thresholds.
On
the lexical decision task although dysphoric individuals reacted approximately
30ms slower to negative than to neutral words, this difference was not
statistically significant, t(29)=-1.6, p=.12. Nondysphoric
individuals reacted approximately 40ms faster to negative than to neutral
stimuli. The difference between the dysphoric and nondysphoric participants’
differences in means (D) was 70ms, which was significant F(1,74)=7.37, p=.008,
h2=.09, supporting the affective interference
hypothesis. Differences were smaller but in the same direction when reactions
to positive words were compared to negative words, F(1,74)=1.99, p=.162,
h2=.03.
On
the valence identification task,, dysphoric individuals’ reaction times
were 200ms faster to negative than to neutral stimuli, t(29)=5.6, p<.001.
Because nondysphoric individuals also reacted 180ms faster to negative than to
neutral stimuli,, the 20ms difference in biases
between dysphoric and nondysphoric individuals was not significant. The
greatest difference between the reactions times of dysphoric and nondysphoric
individuals was with regard to positive stimuli on the valence identification
task. Dysphoric individuals were significantly slower to react to positive than
negative words, D=166ms, in comparison to nondysphoric individuals, D=20 ms, F(1,73)=6.45,
p=.013, h2=.08, supporting the notion that depressed
individuals attend more quickly to negative than positive information.
Omnibus
ANOVAs. To be sure that results of
relevant planned contrasts were not qualified by unexpected interactions
omnibus group (dysphoric, nondysphoric) x valence (positive, negative, neutral)
multivariate split-plot ANOVAs, controlled using a Bonferroni correction at each
level, were performed on the 150ms stimulus duration condition. On the lexical
decision task, a significant Valence x Group interaction, F(2,72)=3.65, p<=.0532, h2=.09, was observed. Simple effects analysis (Maxwell
& Delaney, 1990) yielded no significant effect of valence within the
depressed group, F(2,28)=1.25, p>.05=.30, h2=.082,
or within the control group, F(2,43)=2.75, p>.05=.08, h2=.11,
suggesting the interaction was motivated by the previously reported complex
contrast. On the valence identification task, a significant Valence x Group
interaction, F(2,72)=3.39, p<.05=.039, h2=.09,
was observed. There was a significant effect of valence within the depressed
group, F(2,28)=15.04, p<.00105, h2=.52, which
was driven by the previously reported faster responses to negative than
positive or neutral words. Similarly, in the control group, the effect of valence was also
significant, F(2,43)=9.23, p<.00105, h2=.30, andwhich was driven by faster responses
to positive and negative words than to neutral words.
Robustness
of results to gender and stimulus duration. To examine robustness of the results to the extra noise introduced by
decreasing stimulus duration, a 2 (Dysphoric, Nondysphoric) x 3 (Positive,
Negative, Neutral Valence) x 3 (50ms, 100ms, 150ms stimulus duration) mixed
ANOVA on reaction time was performed for each task. No significant interactions
with stimulus duration were observed for either task. The analysis revealed
significant main effects of stimulus duration on the lexical decision task, F(2,73)=103.8,
p<0.001, and valence identification task, F(2,72)=84.1, p<0.001,
explained by longer reaction times in the shorter stimulus duration condition.
On the valence identification task, a significant valence x group interaction
was also observed, F(2,72)=4.6, p<=0.01. Simple effects analyses
controlled using a Bonferroni correction at each level, revealed that the
planned contrasts, reported previously, largely accounted for the Dysphoria x
Valence interaction. When gender was included in the same ANOVAs as a factor,
no significant main effects or interactions with gender were observed. The only
significant effects were those present when gender was not included. Thus, the
reported effects do not appear to be qualified by gender differences or differences
in stimulus duration.
Robustness
to outliers. The following strategy was adopted tTo be sure that the current results
were robust to outliers, the following strategy was adopted.
Relevant effects were determined to be captured by a significant Dysphoria x
Task x Valence interaction that accounted for approximately 11.5% of the
variation in reaction times, Hotelling’s F(2,69)=4.48, p<.05=0.015. The Dysphoria x Task x Valence
interaction was examined using a number of data aggregation methods suggested
by Ratcliff (1993). These included examination of a range of temporal thresholds for outlier removal other than the
5000ms threshold used in other analyses. Examined values ranged from 2000ms to
no rejection. Similarly, measures of
central tendency for reaction times that have been shown to correct for
violations of normality in individual subjects’ data including medians and
harmonic means were explored. In each case, the Dysphoria x Task x Valence
interaction was significant,, and the effect size was comparable
to that for the analyses reported above.
Results were also qualitatively similar when incorrect responses were
removed from the computation of means., and iIn fact, effect sizes for all contrasts
increased under this restriction; in most cases, signal detection rates were high for
both dysphoric and nondysphoric individuals on both tasks. Mixed ANOVAs
entering counterbalanced conditions (button order, task order, wordlist),
stimulus duration, and dysphoria as independent variables yielded no main
effects or interactions with the counterbalanced variables on reaction times in
separate analyses for each task.
Reliability
of reaction times. To examine whether
apparent differential response latencies were tempered by low reliability of
reaction times, the internal consistency of reaction times for each valence,
for each word set, on both tasks was calculated. In the reported sample,
Cronbach’s alpha was >.7 for positive and negative words on each word set,
on both tasks, and for neutral words on the valence identification task. Alpha
was .68 for one word set, and .61 for the other, for neutral words on the
lexical decision task.
Results
presented thus far suggest that dysphoria may be associated with attention to
emotional aspects of stimuli, at the expense of attention to nonemotional
aspects of stimuli. If true, a large portion of variation in reaction times on
the lexical decision task should be accounted for by reaction times on the
valence identification task, at least for dysphoric individuals. A first
analysis of this relationship involved examining correlations between relevant
effects on the tasks. Differences in reaction times to negative and neutral
words on the lexical decision task were significantly correlated with
differences in reaction times to positive and negative words on the valence
identification task, r=.23, p<.05=.043. To investigate this
relationship in a more sophisticated fashion, hierarchical multiple regressions
were performed on the differences in reaction times to each of the valences.
Variables representing differences in reaction times to negative and neutral
words, positive and neutral words, and positive and negative words on the
valence identification task were entered on the first step of the regression.
Dysphoria, scored dichotomously as in previous analyses, was entered on the
second step. Terms representing the interaction of dysphoria and valence
identification biases were entered on the third step. This technique allowedIn this way, examination of the unique contribution of
valence-mediated reaction-time biases above and beyond dysphoria, as well as
the role of dysphoria in moderating affective interference could be examined.
Biases
on the valence identification task accounted for 5.7% of the variation in the
difference in lexical decision reaction times to negative and neutral words, which was marginally significant, F(2,72)=2.19,
p=.12. Dysphoria accounted for an additional 6.3% of the variation, FD(1,71)=5.09, p<.05=.03. Interactions of dysphoria and
valence identification biases were not statististically significant, DR2=.016,
FD(2,69)=.64, p>.05=.53. In contrast, 10.1% of the
variation in the difference in reaction times to negative and positive words on
the lexical decision task was accounted for by biases on the
valence-identification task, F(2,72)=4.04, p<.=.052. Quicker reaction times to negative
versus neutral words on the valence identification task were associated with
slower reaction times to negative versus positive words on the lexical decision
task. Neither dysphoria nor interactions of dysphoria with valence
identification biases explained significantly more variation, DR2=.04,
p>>.025. This behavior is consistent with
the notion that interference on the lexical decision task is due to
preoccuption with negative aspects of the stimuli.
The
main findings from this experiment suggest that compared to nondysphoric
participants, dysphoric research participants were slower to name the valence
of positive than negative or neutral words, and were , and slightly, though significantly,
slower to identify the lexicality of negative than neutral words. Being quick
to name the negativity of words was associated with being slow to name their
lexicality. In general, these findings are consistent with the idea the idea of
"affective interference," i.e., that dysphoric individuals attend to
the emotional aspects of stimuli at the expense of attending to the
non-emotional aspects of stimuli. Additionally, in support of the idea of
affective interference, responding slowly to the lexicality of
positive, compared to negative words, was associated with quick identification
of the emotionality of negative words. Together, these findings suggest that
depressed individuals may more easily recognize emotional qualities of negative
than other information, and that such biases could be associated with
difficulty responding to any other aspects of negative information.
An
explanation for the obtained data, which that is consistent with the affective
interference hypothesis, suggests that dysphoric individuals often
“ruminate” or think repeatedly about information that is salient to their depression.
Nolen-Hoeksema has suggested that individuals possessing a ruminative cognitive
style tend to focus on their symptoms of depression at the expense of other
tasks they are given to do (e.g., Nolen-Hoeksema, Parker, & Larson, 1994;
Nolen-Hoeksema & Morrow, 1993). Using a broader definition of rumination
(e.g., Ingram's (1984) notion that rumination involves feedback between the
mental representations of non-emotional and emotional associations with
information) ruminating individuals might be hypothesized to reveal biases on
tasks assessing both emotional and nonemotional information processing, when
they are dysphoric. Specifically, such ruminative coping might lead
dysphoric individuals to quickly recognize and ruminate upon negative
information, yielding to quick judgments of that information as negative. In
contrast, positive information would provide the least possible grounds for
either quick recognition or ruminative association of the information with
relevant memories, leading to quick recognition of its lexicality but slow
recognition of its emotional features.
Appeals
to rumination might also predict that people who ruminate on their own
dysphoria (or negative information that is relevant to them) would be most
delayed in identifying negative information not specifically relevant to their
dysphoria; when presented with negative information that is not personally
relevant, a dysphoric individual might think about negative information that is
personally relevant (e.g., a loss event), and thus respond slowly to the
presented non-personally relevant negative information. This hypothesis may
explain why the difference in response times to negative and neutral words was
larger for dysphoric individuals than nondysphoric individuals on the lexical decision
task. It also may suggest a need to assess the personal relevance of
experimental stimuli.
Yet,
relationships between differential lexical decision reaction times to negative
and neutral words and valence identification biases were weak and not replicated
under different types of analysis. If failure to detect this effect is not due
to low power it could reflect the presence of two groups of dysphoric
individuals. One group was slow to respond to negative words on the lexical
decision task; these individuals may also have been slow to respond to negative
words on the valence-identification task. Such individuals might be said to
have a generalized bias away from negative information, perhaps as a result of
heightened perceptual defense to negative information (Powell & Helmsley,
1987). A second group of dysphoric
individuals were primarily slow to respond to positive words on the valence
identification task, a phenomenon which was very weakly associated with being
slower to respond to negative than to neutral words on the lexical decision
task. This group of individuals might experience affective interference.
An
alternate explanation for discrepancies between performance on the tasks is
that the lexical decision and valence identification tasks might require
different levels of processing resources. In this case, the lexical decision
task stimuli might not have been processed to the same extent as the valence
identification stimuli , and thus, biases were not as large.
To exami