Neurophysiological Evidence of Cognitive Inhibition Anomalies in Persons with Major Depressive Disorder

ARTICLE IN PRESS

Clinical Neurophysiology xxx (2008) xxx–xxx www.elsevier.com/locate/clinph

Neurophysiological evidence of cognitive inhibition anomalies in persons with major depressive disorder

Heather E. McNeelya,b,*, Mark A. Laua,b,*, Bruce K. Christensena,b, Claude Alainc,d

a Centre for Addiction and Mental Health, 250 College Street, Toronto, Ontario, Canada, M5T 1R8 b Department of Psychiatry, University of Toronto, Toronto, Ontario, Canada, M5T 1R8 c Rotman Research Institute, 3560 Bathurst Street, Toronto, Ontario, Canada, M6A 2E1 d Department of Psychology, University of Toronto, Toronto, Ontario, Canada, M5T 1R8

Accepted 19 March 2008

Abstract

Objective: The neural correlates of inhibitory deficits for emotional semantic material in persons with major depressive disorder (MDD) were investigated. Methods: Individuals (n = 15) with a diagnosis of MDD or MDD in partial remission, and healthy controls (n = 14) underwent recording of event-related brain potentials (ERPs) while performing a computerized emotional Stroop task. Results: There were no group performance differences on the emotional Stroop task. However, the analysis of ERP waveforms revealed a larger negative wave peaking at about 170 ms over the left than the right hemisphere only in controls; a negative displacement (N450) at parietal sites for positive and negative words only for persons with MDD; in both groups, processing negative and positive words was associated with a positive displacement that peaked at about 450 ms and was larger over the left lateral frontal region; and, the N450 modulation correlated with negative automatic thinking and depressive symptoms. Conclusions: The electrophysiological data reveal early changes in neural activity associated with word processing as well as valence- related changes in the N450 component at parietal sites in MDD. Significance: This valence-related increase in N450 amplitude at parietal sites may reflect an automatic capture of attention by words with emotional valence. Ó 2008 International Federation of Clinical Neurophysiology. Published by Elsevier Ireland Ltd. All rights reserved.

Keywords: Depression; Event-related potentials; Cognitive inhibition; Emotional Stroop

1. Introduction Word Test (SCWT; Benoit et al., 1992; Everett et al., 1989; Lemelin et al., 1996); negative priming task (Linville, Depressed or dysphoric individuals demonstrate impair- 1996; MacQueen et al., 2000); dichotic listening paradigms ments on a variety of cognitive tasks that involve the inhi- (McCabe and Gotlib, 1993) and Continuous Performance bition of distracting elements such as the Stroop Colour Test (CPT; Buchsbaum et al., 1988; Goeleven et al., 2006). Many studies also point to a disproportionate impairment in depressed/dysphoric individuals’ ability to * Corresponding authors. Addresses: St. Joseph’s Healthcare, Hamilton, ignore distracting information when it is negatively valen- Centre for Mountain Health Services, 100 West 5th Street, Box 585, Suite ced (Goeleven et al., 2006; Gotlib et al., 2004, 2005; Lau E-223 H, Hamilton, Ont., Canada L8N 3K7. Tel.: +1 905 522 et al., 2007; Joormann, 2004; McCabe and Gotlib, 1993, 1155x36422; fax: +1 905 381 5635 (H.E. McNeely); BC Mental Health 1995; Segal et al., 1995). For example, the emotional & Addiction Services, 201-601 West Broadway, Vancouver, BC, Canada Stroop task has been widely used to measure the extent V5Z 4C2. Tel.: +1 604 707 6358; fax: +1 604 707 6399 (M.A. Lau). E-mail addresses: [email protected] (H.E. McNeely), to which persons can inhibit the automatic processing of [email protected] (M.A. Lau). depression-related words (Gotlib and McCann, 1984; Got-

Please cite this article in press as: McNeely HE et al., Neurophysiological evidence of cognitive inhibition anomalies ..., Clin Neu- rophysiol (2008), doi:10.1016/j.clinph.2008.03.031 ARTICLE IN PRESS

2 H.E. McNeely et al. / Clinical Neurophysiology xxx (2008) xxx–xxx lib and Cane, 1987; Segal and Vella, 1990; Segal et al., tion. In healthy individuals, one ERP modulation that 1995). In this paradigm, neutral, positive and negative has been shown to correlate with cognitive inhibition dur- words are shown in different ink colours. Participants must ing the standard Stroop colour-word task is the N450 wave ignore word meaning and name the ink colours as quickly (McNeely et al., 2003; Rebai et al., 1997; West and Alain, as possible. Depressed individuals are typically slower at 1999, 2000). The N450 reflects a phasic negativity over naming the ink colour of negative versus positive and neu- frontal and central scalp regions that peaks between 400 tral words, but there is generally no valence effect for and 500 ms after stimulus onset. It inverses in polarity over healthy controls. lateral and inferior frontal regions, and appears to arise Similarly, persons with Major Depressive Disorder from activity of the ACC (Liotti et al., 2000; McNeely (MDD) have shown difficulties in ignoring negative versus et al., 2003). The amplitude of the N450 is modulated by positive or neutral words embedded in a section of prose as the proportion of congruent to incongruent stimuli (West manifest by increased reading times (Lau et al., 2007). and Alain, 2000) and is significantly attenuated in persons Importantly, in this study, no performance deficits were with schizophrenia (McNeely et al., 2003) and older adults observed on the Stop Signal Task (Logan et al., 1997), a (West and Alain, 2000) indicating that it is sensitive to vari- measure of behavioural or motor response inhibition, sug- ations in cognitive control and inhibition. gesting that these deficits may be limited to the cognitive In the present study, we used a computerized emotional domain. Furthermore, valence-related increases in Stroop task to examine the effects of depression on behav- response times were not observed in a control group of ioural and electrophysiological indices of cognitive inhibi- individuals with mixed anxiety disorders, suggesting that tion for emotionally valenced semantic stimuli. Of most difficulties in inhibiting word valence were specific to interest in the behavioural data was the expected emotional depression and could not be accounted for by general psy- Stroop interference effect for negative trials, which would chopathology (Garber and Hollon, 1991). Importantly, be expressed as increased response latency and decreased prose reading time performance deficits were correlated accuracy. This was expected to occur selectively for nega- with increased negative cognition, as measured by the tive versus positive and neutral trials in individuals with Automatic Thoughts Questionnaire (Hollon and Kendall, depression relative to controls. Also, we expected to 1980); participants reporting the greatest degree of negative observe modulation of the N450 response associated with automatic thinking displayed the most difficulty inhibiting negative trials in persons with MDD, which would be task irrelevant, negative, emotional material. This finding is indicative of impaired neural processes underlying cogni- consistent with the hypothesis that disturbances of cognitive control deficits consistent with MDD. Finally, consis- tive control for affectively salient information may play a tent with Lau et al. (2007) we predicted that modulation of critical role in the preponderance of negative thinking in the N450 in participants with MDD would be correlated persons with MDD (Lau et al., 2007). with the increased frequency of negative automatic Cognition may serve an important role in the nexus thoughts. between clinical symptomatology and neurobiology as a more discrete, measurable, and easily manipulated interme- 2. Methods diary in psychopathology (Danion et al., 1996; Andreasen, 1997). Although few studies have been conducted to 2.1. Participants directly examine the links between specific behavioural or symptomatic features of MDD and neural abnormalities, Participants (ages 18–57 years) had normal or corrected- there is preliminary evidence to suggest that anomalies in to-normal vision, at least 12 years of formal education and the functioning of the anterior cingulate cortex (ACC) reported English as their primary language. Fifteen partic- may underlie MDD-related cognitive inhibitory dysfunc- ipants with depression (MDD, including four in partial tion. First, the ACC is an important component in the net- remission) were recruited from individuals seeking treat- work of structures implicated in the pathophysiology of ment at the Cognitive Behaviour Therapy (CBT) Unit at MDD (Lane et al., 1998; Mayberg, 1997; Mayberg et al., the Centre for Addiction and Mental Health, in Toronto, 1999). Second, research has demonstrated that the ACC Ontario, Canada or from the Toronto community through plays a fundamental role in cognitive inhibition among newspaper advertisements from August 2001 until August healthy individuals (see Bush et al., 2000, for a review). 2002. The Structured Clinical Interview for the DSM-IV In addition, in clinical populations (e.g., schizophrenia), Axis I disorders (SCID-I/P; First et al., 1998) was used observed inhibitory deficits are correlated with impaired to confirm a diagnosis of MDD and screen for co-morbid ACC function (Alain et al., 2002). Together, these lines diagnoses in the MDD group, and to exclude participants of evidence support a possible role for the ACC in inhibi- based on any Axis I disorder in the control group. Eight tory dysfunction associated with MDD; however, this has participants (53%) with MDD were taking medication at yet to be empirically evaluated. the time of the ERP recording (fluoxetine (Prozac), n =1; Scalp recording of event-related brain potentials (ERPs) bupropion (Wellbutrin), n = 2; sertaline (Zoloft), n =1; provide a powerful tool with which to investigate the effects venlafaxine (Effexor), n = 1; alprazolam (Xanax), n =1; of depression on the neural correlates of cognitive inhibi- lorazepam (Ativan), n = 1; clozapine (Clozaril), n =1;

H.E. McNeely et al. / Clinical Neurophysiology xxx (2008) xxx–xxx 3 one missing medication data). All medicated participants 2.2. Materials and procedure were on a single medication, with the exception of one who was on a combination of two medications. Control Participants were tested individually in two sessions fol- participants were 14 volunteers recruited from the Toronto lowing the SCID-IV screening session. In the first session, community through newspaper advertisements who had all participants completed a medical history questionnaire neither current Axis 1 disorder nor a lifetime history of and the following clinical and neuropsychological mea- depression. sures: selected subtests from the Wechsler Adult Intelli- Exclusion criteria for both groups included: substance gence Scale-III (WAIS-III; Wechsler, 1997); the reading abuse within the last 6 months; lifetime substance depen- subscale of the Wide Range Achievement Task (WRAT- dence; significant self-reported medical conditions that 3; Wilkinson, 1993); the Edinburgh Handedness Inventory could have had a significant impact on cognitive function (Oldfield, 1971); the Beck Depression Inventory-II (BDI-II; (e.g., hepatitis C); self-reported neurological illness or his- Beck et al., 1996); and, the Hamilton Rating Scale for tory of head trauma with the loss of consciousness exceed- Depression (HRSD; Hamilton, 1960). Participants with ing 20 min; and, English language acquired after age 5. depression also completed the Automatic Thoughts Ques- Participants with depression were also excluded based on tionnaire (ATQ; Hollon and Kendall, 1980). co-morbid diagnoses of any other Axis I disorder [with The second session was conducted on a separate day the exception of co-morbid social anxiety (n = 2) and gen- within 1 week of the first experimental session, at which time eralized anxiety disorder (n = 2) with one depressed partic- the ERPs were recorded while participants performed an ipant having both diagnoses]. All participants provided emotional Stroop task. Participants were seated approxi- written informed consent for participation in this study mately 24 in. in front of a 17-in. computer monitor in a dimly after the nature of the procedure(s) had been fully lit room. Emotional Stroop stimuli consisted of negative explained. The study protocol was approved by the (e.g., MOURN), positive (e.g., CHEERY) and neutral Research Ethics Board of the Centre for Addiction and (e.g., GLIMPSE) words between 4 and 7 letters in length Mental Health. Participants were paid a modest stipend and equated for frequency of usage in the English language to help cover expenses (e.g., parking, travel). (Kucera and Francis, 1967). Word stimuli were taken from Individuals with depression were similar to healthy a previous ERP emotion study (McNeely et al., 2004) and controls with respect to demographic characteristics, were rated for degree of pleasantness–unpleasantness by a but reported more depressed symptomatology (see Table separate sample (n = 5) of persons with a SCID-confirmed 1). Two participants with MDD and one control partic- diagnosis of MDD. The words were presented blocked by ipant were left-handed, while the remaining were right- valence to increase the emotional Stroop effect, and handed. pseudo-randomly (i.e., randomized colour order was super- ceded if more than two stimuli in a row were the same colour in order to prevent response habituation) by font colour Table 1 within blocks to ensure an equal number of words per font Sociodemographic and clinical characteristics for depressed and healthy colour. Stimuli appeared on the screen against a black back- groups ground for 400 ms, followed by a blank screen for 2600 ms. Variable Depressed Healthy The task was programmed using STIM software (Compu- M (SD)/n (%) M (SD)/n (%) medics Neuroscan, 2007). Participants were asked to identify Age (years) 38.5 (8.7) 34.9 (10.5) the font colour by pressing one of four colour-coded (red, Gender (%) blue, green, yellow) response buttons on a key-pad using Male 4 (27) 6 (43) their middle and index fingers of their right and left hands. Female 11 (73) 8 (57) Years of education 16.9 (2.8) 16.4 (2.4) If the word was presented in light gray, participants were WAIS-III estimated IQ 119.6 (10.7) 118.2 (11.2) instructed to read this word silently. Word-reading trials WAIS-III digit symbol* 10.4 (2.5) 12.6 (2.6) were included to increase the tendency to attend to word WRAT-3 reading SS 108.5 (6.9) 108.1 (7.5) ** information (and not simply glaze over, making it easier to BDI-II 29.9 (6.8) 1.3 (2.6) respond to colour), and thereby enhancing the magnitude HRSD** 14.8 (5.1) 0.4(0.6) ATQ degree of belief 90.8 (23.0) – of the Stroop effect (West and Alain, 2000). Task instructions ATQ frequency 90.8 (22.5) – emphasized the importance of speed and accuracy equally. Note: Depressed, n = 15; healthy, n = 14. WAIS-III, Wechsler Adult As in our previous work with the standard single-trial Intelligence Scale-III (Weschler, 1997); WAIS-III estimated IQ based on 2- Stroop (McNeely et al., 2003), the emotional Stroop task subtest version using Vocabulary and Matrix Reasoning (Sattler and was divided into a colour-key acquisition phase, a practice Ryan, 1999); WRAT-3, Wide Range Achievement Task (Wilkinson, phase, and a test phase. The colour-key acquisition phase 1993); WRAT-3 reading SS = scaled score; age corrected norms used for consisted of 100 trials with each of the four colours (i.e., WAIS-III, WRAT-3; BDI-II, Beck Depression Inventory (Beck et al., 1996); HRSD, Hamilton Rating Scale for Depression (Hamilton, 1960); not including gray) presented in a random order 25 times ATQ, Automatic Thoughts Questionnaire (Hollon and Kendall, 1980). as a series of X’s. Following this, a 50 trial practice block * p < .05; healthy > depressed. was completed in which neutral words were shown in all ** p’s < .001; depressed > healthy. colours including word-reading trials in light gray. The

4 H.E. McNeely et al. / Clinical Neurophysiology xxx (2008) xxx–xxx practice phase allowed participants to become familiar with Table 2 the task and ensured that they understood the instructions. Mean response times (and standard deviations), in milliseconds, for The test phase consisted of 3 blocks of 130 trials each that negative, neutral and positive trials in depressed and control groups included 100 emotional target stimuli (25 per colour, Variable Depressed Healthy pseudo-randomly distributed) and 30 word-reading stimuli M (SD) M (SD) pseudo-randomly distributed throughout the test trials. Negative 763 (107) 722 (95) Stimuli were blocked by valence, including valence congru- Neutral 744 (100) 722 (105) ent word-reading trials. Block order was counter-balanced Positive 747 (95) 720 (116) across participants. Participants were given rest breaks between each learning and test block. three valence conditions. The data were submitted to a 2 Behavioural data (response time and accuracy) were (group) by 3 (valence: negative, positive, neutral) ANOVA. analyzed using separate mixed-design repeated measures There were no significant main effects (p’s > .10) and the analysis of variance (ANOVA) with clinical group as the group by valence interaction was not significant between-subjects factor and valence condition (negative, 2 F(2,54) = 1.29, p = .269; gp ¼ :046. The response times positive, neutral) as the within-subjects factor. of participants with depression were no different from that of controls when responding correctly across valences. 2.3. Electrophysiological recording and analysis

Continuous electroencephalograph (EEG) activity 3.1.2. Accuracy data (bandpass 0.05–30 Hz), digitized at 250 Hz, was recorded Table 3 contains the mean (SD) number of errors and from an array of 32 electrodes based upon an extended misses for depressed and control groups for the three 10–20 system. Ocular movements were recorded from elec- valence conditions. As for the response time data, there trodes placed lateral to and below both eyes. During were neither significant main effects nor interactions for recording, all channels were referenced to the central mid- either analysis, indicating that both groups made a similar line electrode (Cz). For data analysis, the ERPs were re-ref- number of errors and missed responses. erenced to an average reference. ERP epochs were extracted off-line and included a 200- 3.2. Electrophysiological data ms prestimulus baseline and 1400 ms of post-stimulus activity. Trials contaminated by eye movement or other The group average mean ERPs from all channels super- artifacts in excess of ±50 lV at electrodes not adjacent to imposed are presented in Fig. 1. In both groups, visual the eyes were discarded before averaging. ERPs were aver- stimuli elicited a large negative wave peaking on average aged for correct trials, as a function of valence (i.e., nega- at 168 ms after stimulus onset (N170) at inferior parietal tive, positive, neutral). ERP averages were digitally sites (i.e., P7 and P8), occipital (O1, O2) and cerebellar lowpass filtered to attenuate frequencies above 20 Hz. (CB1, CB2) electrodes. There was no difference in N170 For each individual average, the ocular artifacts (e.g., peak latency between individuals with MDD and age- blinks and lateral movements) were corrected by means matched controls, nor was the effect of stimulus type signif- of ocular source components (Picton et al., 2000). icant, F < 1 in both cases. An ANOVA on the mean ampli- For the analysis of the ERP data, electrode position was tude between 140 and 180 ms yielded a significant added as an additional within-subjects factor. Statistical group hemisphere interaction, F(1,27) = 4.75, p < .05. analyses were performed on the mean voltages over various While controls showed greater N170 amplitude over left intervals of interest in the ERP waveform similar to those hemisphere (F(1,13) = 5.68, p < .05), individuals with identified in previous work (West and Alain, 1999; McNe- MDD generated N170 with comparable amplitude over ely et al., 2003). Significant results of omnibus ANOVAs were further examined using pairwise comparisons. Type Table 3 I errors associated with non-homogeneity of variance were Mean number (and standard deviations) of errors and misses for negative, controlled for using the Huynh–Feldt (Huynh and Feldt, neutral and positive trials in depressed and control groups 1976) procedure. The probability estimates are based on Variable Depressed Healthy reduced degrees of freedom, but the original degrees of M (SD) M (SD) freedom are reported. Negative Errors 2.5 (1.6) 2.5 (1.6) 3. Results Misses 2.3 (4.5) 1.9 (3.4) Neutral 3.1. Behavioural data Errors 3.0 (1.4) 3.3 (1.1) Misses 2.1 (3.5) 1.8 (2.7) 3.1.1. Response time Positive Table 2 contains the mean (SD) response times on cor- Errors 2.7 (2.1) 2.9 (1.6) rect trials (hits) for depressed and control groups for the Misses 2.7 (3.9) 2.9 (4.6)

H.E. McNeely et al. / Clinical Neurophysiology xxx (2008) xxx–xxx 5

Fig. 1. (A) Butterﬂy plots showing the group mean ERPs averaged over stimulus type in controls and patients with MDD. Each gray line represents the evoked response recorded at a particular electrode. The ERPs recorded over the left parietal (P7) and midline parietal site (Pz) are shown in black. (B) Isocontour maps showing the topographic distribution (view from the back) of the N170 and late positive complex (LPC).

both hemispheres, F < 1 (see also Fig. 2). The main effect of revealed a main effect of stimulus type, F(2,28) = 8.51, stimulus type was not significant nor was the interaction p < .005. Pairwise comparisons revealed more pronounced between group and stimulus type, F < 1 in both cases. negative-going ERPs for negative and positive words than for neutral words, p < .01 in both cases. The difference in 3.2.1. Cognitive inhibition and ERPs: N450 ERP amplitude between negative and positive valence Fig. 2 shows the group mean ERPs elicited by negative, words was not significant, p = .91. In controls, as expected, neutral, and positive words over the left and right inferior the main effect of stimulus type was not significant, parietal scalp sites (i.e., P7 and P8) in both controls and F(2,26) = 1.03. However, there was a significant interac- MDD. In patients with MDD, positive and negative words tion between stimulus type and hemisphere, were characterized by enhanced negativity, which inverted F(2,26) = 3.65, p < .05, reflecting greater valence-related in polarity over the left lateral frontal site (i.e., F7, see Figs. changes in ERPs amplitude over the left than the right 4 and 5). The positive wave peaked on average at 375 ms hemisphere. Moreover, separate ANOVAs for each stimu- post-stimulus. The ANOVA on the peak latency measured lus type showed that individuals with MDD generated between 300 and 500 ms yielded a main effect of stimulus more negative ERPs than controls during both negative type, F(2,54) = 9.61, p < .001, with latencies being longer and positive valence conditions, F(1,17) = 4.42 and 4.76, for negative than neutral or positive words, p < .005 in p < .05 in both cases. There was no difference between both cases. There was no difference in latency between neu- the two groups for the neutral condition, F(1,27) = 2.15, tral and positive words. The effect of valence on ERPs was p = .15. These valence-related differences in ERP amplitude between controls and patients are best illustrated in quantified for the 300–500 ms interval in a 2 (group) 3 the difference waves between the ERPs elicited by negative (stimulus: negative, neutral, positive) 2 (hemisphere: left, or positive words minus those elicited by neutral words right) 3 (electrode: cerebellar (e.g., CB1), inferior parietal (e.g., P7), and occipital (e.g., O1)) ANOVA. This analysis (Fig. 3). The enhanced negativity over the inferior pari- yielded a main effect of stimulus type (F(2,54) = 6.57, eto-occipital regions could be related to patients allocating p < .005), and significant interactions between group and more attention to the emotionally meaningful words than stimulus type (F(2,54) = 3.31, p = .05), and between stimu- did controls. lus type and hemisphere (F(2,54) = 3.79, p < .05). To gain The processing of emotional words was also associated a better understanding of these interactions, we performed with a modulation over the lateral left and right frontal separate ANOVAs on each group. In MDD, the ANOVA areas in both controls and MDD (Fig. 4). In both groups,

6 H.E. McNeely et al. / Clinical Neurophysiology xxx (2008) xxx–xxx

Fig. 2. Group mean ERPs recorded over the left and right inferior parietal regions (i.e., P7 and P8, respectively). Note the asymmetry in N170 amplitude in controls, which was not present in patients. Moreover, note the valence-related changes in ERP amplitude peaking at about 400 ms in patients. negative words were also characterized by enhanced nega- after stimulus onset. Its latency was comparable in both tivity at midline frontal and central sites (data not shown), groups (F < 1), and was not affected by word valence, which inverted in polarity over the left lateral frontal site F < 1. The effect of valence on ERPs was quantified for (i.e., F7). This modulation peaked on average at 375 ms the 300–500 ms interval in a 2 (group) 3 (stimulus: nega-

Fig. 3. (A) Diﬀerence wave between ERPs elicited by either negative (top) or positive (bottom) words and those evoked by neutral words in both groups. The traces reﬂect activity recorded over the left inferior parietal (i.e., P7). (B) Isocontour maps showing the amplitude distribution (view from the back) at 400 ms after sound onset.

H.E. McNeely et al. / Clinical Neurophysiology xxx (2008) xxx–xxx 7

Fig. 4. Group mean ERPs recorded over the left and right lateral frontal regions (i.e., F7 and F8, respectively). Note that the polarity of the ERPs is inverted relative to those recorded over the midline frontal and central scalp areas. tive, neutral, positive) 2 (hemisphere: left, right) 2 difference in ERP amplitude between negative and positive (electrode: lateral–frontal (e.g., F7), frontal (e.g., F3)) valence words was not significant, p = .62. The stimulus ANOVA. This analysis yielded a main effect of stimulus type hemisphere interaction was significant, type (F(2,54) = 6.62, p < .01). Pairwise comparisons F(2,54) = 5.57, p < .01, reflecting greater valence-related revealed more positive ERPs for negative and positive changes in ERPs recorded over the left than the right hemi- words than for neutral words, p < .05 in both cases. The sphere. However, the group stimulus interaction was not

Fig. 5. (A) Diﬀerence wave between ERPs elicited by either negative (top) or positive (bottom) words and those evoked by neutral words in both groups. The traces reﬂect activity recorded over the left lateral frontal cortex (i.e., F7). Note that the polarity of the ERPs is inverted relative to those recorded over the midline frontal and central scalp areas. (B) Isocontour maps showing the amplitude distribution at 400 ms after sound onset (bird’s eye view).

8 H.E. McNeely et al. / Clinical Neurophysiology xxx (2008) xxx–xxx

Fig. 6. Group mean ERPs recorded over the left and right fronto-polar regions (i.e., FP1 and FP2, respectively). Note the asymmetry in sustained activity in controls for the negative valence condition, which was not present in patients. Moreover, note the long lasting valence-related changes in ERP amplitude beginning at about 130 ms and lasting several hundreds of milliseconds in patients. significant (F(2,54) < 1), nor was the group stimulus reach significance, (n = 13; r = .410, p = .165). The ampli- type hemisphere interaction (F(2,54) < 1), indicating tude of the difference wave between ERPs elicited by nega- comparable valence-related changes in ERP amplitude tive and neutral words recorded over the left frontal areas over the left frontal sites (Fig. 5). correlated significantly with level of depressive symptoms In addition to changes in ERP amplitude during the as measured by the HRSD (n = 13; r = .447, p = .109). 400–500 ms interval, processing words of different valence This finding suggests that increased neurophysiological was paralleled by changes in ERPs that lasted several hun- activity to salient negative stimuli is associated with higher dreds of milliseconds over the frontal and fronto-polar levels of depression among those with MDD. (i.e., FP1, FPz, FP2) scalp region in patients with MDD (Fig. 6). The ANOVA on the mean amplitude between 4. Discussion 1000 and 1400 ms recorded at fronto-polar sites yielded a main effect of group, F(1,27) = 6.23, p < .05. However, This study examined three predictions with respect to the interaction between group and stimulus type was not the behavioural and neurophysiological correlates of cogni- significant, F(2,54) = 1.34, p = .27. tive inhibition within the context of an emotional Stroop task in individuals with MDD versus healthy controls. 3.3. Relationship between neural indices of cognitive Contrary to our first prediction, there were no group differ- inhibition and clinical measures ences in response times or accuracy on the emotional Stroop task. Our second prediction was partially supported Relationships between neurophysiological measures of in that major depression yielded larger N450 over the pari- cognitive inhibition and clinical measures of automatic etal region for positive and negative words relative to neu- negative thinking and depressive symptoms were assessed tral words, which may reflect a deficit in inhibiting word using simple bivariate correlations. Among persons with valence or attentional capture by the positive and negative depression, the amplitude of the difference wave between words. Finally, our third prediction was also partially sup- ERPs elicited by positive and neutral words recorded over ported, as there were significant correlations between the the parieto-occipital region of the right hemisphere corre- N450 modulation over the left prefrontal cortex and the lated significantly with frequency of automatic negative right parietal scalp regions and clinical symptom measures thoughts as measured by the ATQ, (n = 13; r = .559, among persons with depression. p < .05), suggesting that those allocating more attention In controls, all classes of stimuli generated an N170 that to positive words also reported less frequent negative was larger over the left than the right hemisphere. This thinking. For negative words, the correlation failed to hemispheric difference in N170 elicited by words is consis-

H.E. McNeely et al. / Clinical Neurophysiology xxx (2008) xxx–xxx 9 tent with prior studies showing left hemisphere dominance domains or whether it is specific to cognitive inhibitory in visual word processing (Rossion et al., 2003), represent- dysfunction. ing an important step in visual word recognition likely The present findings may reflect abnormal activity in related to a prelexical orthographic processing. In the pres- cortical and subcortical neural networks thought to play ent study, individuals with MDD generated comparable a role in MDD (Mayberg, 1994, 1997; Mayberg et al., N170 over the left and right hemisphere. This disparity in 1999) and emotional functioning (Lane et al., 1998; Dre- N170 amplitude distribution between controls and individ- vets and Raichle, 1998; Whalen et al., 1998). According uals with MDD may reflect differences in early word pro- to Mayberg’s model, depression involves dysfunction in cessing with controls relying on a left lateralized network a network of brain regions that coordinate aspects of involved in orthographic processing. Further research is attention and cognition with somatic and mood pro- needed to better understand this difference in orthographic cesses. The regions of the brain thought to be important processing. in the cognitive aspects of depression include the dorsal The N450 response in participants with depression, anterior cingulate, dorsal lateral prefrontal cortex, infe- although not entirely consistent with a selective processing rior parietal cortex and the striatum (Mayberg, 1997). deficit for negatively valenced information is, to our knowl- In regional cerebral blood flow neuroimaging investiga- edge, the first demonstration of an ERP inhibition anomaly tions of mood states in both healthy and depressed per- in these individuals. These findings are similar to reports of sons, induced sadness and MDD have both been an ERP inhibitory response in other clinical groups known associated with decreased blood flow in dorsal and ros- to have inhibitory deficits, including elderly adults (West tral anterior cingulate and connected cortical and subcor- and Alain, 2000), and patients with schizophrenia (McNe- tical regions (Lane et al., 1998; Mayberg et al., 1999). ely et al., 2003). Thus, these findings suggest disrupted Interestingly, altered glucose metabolism and blood flow neurophysiological processes associated more generally in the rostral anterior cingulate may mediate response to with cognitive inhibition in persons with MDD in partial antidepressant treatment (Mayberg, 2003). Moreover, remission or mild to moderate MDD. increases in dorsal ACC activity have been observed with Interestingly, inhibition has been implicated as a recovery from depression (Mayberg et al., 1999) provid- mechanism central to a number of different cognitive ing further evidence for a link between alterations in operations (in addition to selective attention) such as ACC activation and MDD. behavioural control (Harnishfeger, 1995; Nigg, 2000), Activity of the ACC is also implicated in cognitive which, as a result, has made empirical operationaliza- inhibition of goal-irrelevant information in healthy adults tions elusive. Nevertheless, inhibition is now considered (see Bush et al., 2000 for a review). Increased activation by most as consisting of several separate but related pro- in the ACC has been consistently observed during inhibi- cesses (Harnishfeger, 1995; Nigg, 2000). Harnishfeger tion of conflicting colour-word information in the (1995) draws on the distinction between behavioural, or Stroop, using both ERP (Liotti et al., 2000; McNeely response, versus cognitive inhibition; the former is et al., 2003; West and Alain, 1999, 2000) and cerebral defined as the inhibition of an overt motor response blood flow methods (Bench et al., 1993a,b; Drevets and and the latter as the suppression of cognitive contents Raichle, 1998; Leung et al., 2000; MacDonald et al., or processes. Given this context, it is interesting that 2000). A recent single cell recording study in healthy our findings are convergent with Kaiser et al. (2003) adults also confirmed the involvement of the caudal who demonstrated a specific response inhibition deficit ACC in Stroop inhibition at the cellular level (Davis on the Go/Nogo task associated with a reduction in an et al., 2005), suggesting that a sub-population of ACC early fronto-temporal positivity in the N2 time window neurons may function to detect salient information of the ERP waveform in depressed versus healthy con- including conflict, targets, or emotionally meaningful trols. These two results are consistent with a model of material. Taken together, the current findings are consis- an inhibition deficit in depressed participants across a tent with the notion of a neurobiological link between number of different cognitive operations. However, depression and deficient cognition inhibition. inconsistent with this model is the demonstration of a Contrary to previous reports (Gotlib and McCann, dissociation between cognitive and response inhibitory 1984; Gotlib and Cane, 1987; Segal and Vella, 1990; Segal dysfunction in depressed participants using the Stop Sig- et al., 1995; Williams and Broadbent, 1986), there were no nal task (Logan et al., 1997) and the prose distraction significant group differences in behavioural performance on task (Connelly et al., 1991) to measure response and cog- the emotional Stroop task despite significant between nitive inhibition, respectively (Lau et al., 2007). It may group differences in the amplitude of the ERP inhibitory be that the results of Kaiser et al. (2003), using auditory response for emotionally meaningful as compared to neu- stimuli in the Go/Nogo task, reflect deficits at an earlier tral words. While the depressed participants did demon- stage of processing than that measured by Lau et al. strate slower reaction times for negative versus neutral or (2007) who used visual semantic stimuli. Future research positive information and an effect size for the group by 2 is required to determine whether depression is associated valence interaction ðgp ¼ :046Þ, the most likely explanation with an inhibitory deficit across different psychological for the lack of significant effects is the lack of power (calcu-

10 H.E. McNeely et al. / Clinical Neurophysiology xxx (2008) xxx–xxx lated as .20 for the current between-group behavioural for stories containing negative distractor words were asso- comparison). Alternatively, there are a number of method- ciated with increased rumination and frequency of negative ological differences between the current and previous para- automatic thinking only for depressed individuals (Lau digms that may also have contributed to discrepancies et al., 2007). These correlations were significant when con- between current and previous behavioural results. The trolling for level of depressive symptoms, whereas in the stimuli presented in this study differed from previous stud- current study there was a significant association between ies, in that, although a separate small group of persons increased neurophysiological activity to salient negative with depression rated the emotional valence of the included stimuli and higher levels of depression. However, the corre- stimuli, these stimuli were not specifically self-referent lations reported in the current study must be interpreted adjectives as used, for example, in Segal et al. (1995) but with some caution since similarly lateralized ERP effects included more generally negative stimuli such as fear and were not found among controls. Thus, whether the demon- disgust related words in addition to depressive words. strated associations between neurophysiological activity Finally, given a concern that repeating a small set of highly and negative automatic thinking are moderated/mediated self-referent adjectives could confound ERP inhibition by the level of depressive symptomatology remains to be results by eliciting an ERP recollection or familiarity effect determined. (see McNeely et al., 2004), we used a large number of In conclusion, the results of this study provide novel words in order to eliminate repetition and maximize neurophysiological evidence of anomalies in cognitive inhi- N450 ERP effects. However, repetition of word stimuli bition to salient emotional material in the face of intact may be required to detect effects in diagnostic groups. In behaviour during the inhibition of emotional stimuli order to test this hypothesis, we have since reduced the among persons with mild to moderate depression com- number of word stimuli in subsequent research. pared to healthy controls. The field of psychiatry has until Nevertheless, our Stroop results are similar to George recently relied on subjective data for diagnostic procedures. et al. (1997) who did not demonstrate significant perfor- The current findings suggest that ERPs may provide a mance differences between depressed versus age- and sex- valuable tool for objectively indexing illness onset, severity matched control participants on a standard and sad and potentially, treatment response. (consisting of four sad words) Stroop task, likely as a result of a small sample size as well. Also similar to Acknowledgements the current study was that in spite of no group differences on the behavioural measure, there were group dif- The authors thank Carol Garson for her assistance in ferences in regional cerebral blood flow; that is, a relative recruiting and screening participants. We thank He Yu hypoactivation of the right anterior cingulate gyrus in for technical assistance. Portions of this article were pre- depressed participants. Relatedly, McNeely et al. sented at the 45th annual meeting of the Society for Psy- (2003), demonstrated that 13 patients with schizophrenia chophysiological Research, Lisbon, Portugal, September performed the standard Stroop task as well as controls 21–24, 2005. This research was supported by grants from (when general response slowing was controlled for) while the Ontario Mental Health Foundation, the Canadian Psy- demonstrating significantly attenuated ERP inhibition chiatric Research Foundation, and the Centre for Addic- effects. Given that all three studies had relatively small tion and Mental Health Foundation. sample sizes, the most likely explanation for the dissociation between behavioural and neuroimaging results is References that larger sample sizes may be required to detect differences at the behavioural level that are more readily Alain C, McNeely HE, He Y, Christensen BK, West R. Neurophysio- apparent at the neurophysiological level. This also sug- logical evidence of error-monitoring deficits in patients with schizo- gests that ERPs may be more sensitive to subtle between phrenia. Cereb Cortex 2002;12:840–6. group differences than behaviour alone, as patients may Andreasen NC. Linking mind and brain in the study of mental illnesses: a project for a scientific psychopathology. Science have adopted a down-stream compensatory strategy to 1997;275(14):1586–93. produce normal behavioural responses in the face of Beck AT, Steer RA, Brown GK. Beck depression inventory – second edition anomalous early neurophysiological responses; despite manual. San Antonio, TX: The Psychological Corporation; 1996. an MDD sample with average to above average intelli- Bench CJ, Friston KJ, Brown RG, Frackowiak RS, Dolan RJ. Regional gence (as demonstrated by their neuropsychological cerebral blood flow in depression measured by positron emission tomography: the relationship with clinical dimensions. Psychol Med scores) we were still able to demonstrate the hypothe- 1993a;23:579–90. sized ERP differences. Bench CJ, Frith CD, Grasby PM, Friston KJ, Paulesu E, Frackowiak The demonstrated association between the N450 modu- RSJ, et al. Investigations of the functional anatomy of attention using lation over the right parietal scalp regions and negative the Stroop test. Neuropsychologia 1993b;31:907–22. automatic thinking is consistent with the notion that those Benoit G, Fontin L, Lemelin S, Laplante L, Thomas J, Everett J. L’attention selective dans la depression majeure: ralentissement allocating less attention to positive words also reported clinique et inhibition cognitive. Can J Psychol 1992;46:41–52. more frequent negative automatic thinking. This is conver- Buchsbaum MS, Lee S, Haier R, Wu JC, Green M, Tang SW. Effects of gent with a previous finding that increased reading times amoxapine and imipramine on evoked potentials in the Continuous

H.E. McNeely et al. / Clinical Neurophysiology xxx (2008) xxx–xxx 11

Performance Test in patients with affective disorder. Neuropsychologia Lemelin S, Baruch P, Vincent A, Laplante L, Everett J, Vincent P. 1988;20:15–22. Attention disturbance in clinical depression: deficient distractor Bush G, Luu P, Posner MI. Cognitive and emotional influences in anterior inhibition or processing resource deficit? J Nerv Ment Dis cingulate cortex. Trends Cogn Sci 2000;4:215–22. 1996;184:114–21. Connelly SL, Hasher L, Zacks RT. Age and reading: the impact of Leung H-C, Skudlarski P, Gatenby JC, Peterson BS, Gore JC. An event- distraction. Psychol Aging 1991;6(4):533–41. related functional MRI study of the Stroop color word interference Danion JM, Weingartner H, Singer L. Is cognitive psychopathology task. Cereb Cortex 2000;10:552–60. plausible? Illustrations from memory research. Can J Psychiatry Linville P. Attention inhibition: does it underlie ruminative thought? Adv 1996;41(Suppl. 1):S5–S13. Soc Cogn 1996;9:121–34. Davis KD, Taylor KS, Hutchison WD, Dostrovsky JO, McAndrews MP, Liotti M, Woldorff MG, Perez III R, Mayberg HS. An ERP study of the Richter EO, et al. Human anterior cingulate cortex neurons encode temporal course of the Stroop color-word interference effect. Neuro- cognitive and emotional demands. J Neurosci 2005;25(37): 8402–6. psychologia 2000;38:701–11. Drevets WC, Raichle ME. Reciprocal suppression of regional cerebral Logan GD, Schachar RJ, Tannock R. Impulsivity and inhibitory control. blood flow during emotional versus higher cognitive processes: Psychol Sci 1997;8:60–4. implications for interactions between emotion and cognition. Cogn MacDonald AW, Cohen JD, Stenger VA, Carter CS. Dissociating the role Emotion 1998;12:353–85. of dorsolateral prefrontal and anterior cingulate cortex in cognitive Everett J, Laplante L, Thomas J. The selective attention deficit in control. Science 2000;288:1835. schizophrenia: limited resources or cognitive fatigue?. J Nerv Ment Dis MacQueen GM, Tipper SP, Young LT, Joffe RT, Levitt AJ. Impaired 1989;177:735–8. distractor inhibition on a selective attention task in unmedicated, First MB, Spitzer RL, Gibbon M, Williams JBW. The structured clinical depressed subjects. Psychol Med 2000;30:557–64. interview for DSM-IV Axis I disorders. New York: Biometrics Mayberg HS. Frontal lobe dysfunction in secondary depression. J Research; 1998. Neuropsychiatry Clin Neurosci 1994;6:428–42. Garber J, Hollon SD. What can specificity designs say about causality in Mayberg HS. Limbic–cortical dysregulation: a proposed model of psychopathology research? Psychol Bull 1991;110:129–36. depression. J Neuropsychiatry Clin Neurosci 1997;9:471–81. George MS, Ketter TA, Parekh PI, Rosinsky N, Ring HA, Pazzaglia PJ, Mayberg HS. Modulating dysfunctional limbic-cortical circuits in depres- et al. Blunted left cingulated activation in mood disorder subjects sion: towards development of brain-based algorithms for diagnosis during a response interference task (the Stroop). J Neuropsychiatry and optimized treatment. Br Med Bull 2003;65:193–207. Clin Neurosci 1997;9:55–63. Mayberg HS, Liotti M, Brannan SK, McGinnis S, Mahurin RK, Jerabek Goeleven E, De Raedt R, Baert S, Koster HW. Deficient inhibition of PA, et al. Reciprocal limbic-cortical function and negative mood: emotional information in depression. J Affect Disord converging PET findings in depression and normal sadness. Am J 2006;93:149–57. Psychiatry 1999;156:675–82. Gotlib IH, Cane DB. Construct accessibility and clinical depression: a McCabe SB, Gotlib IH. Attentional processing in clinically depressed longitudinal approach. J Abnorm Psychol 1987;96:199–204. subjects: a longitudinal investigation. Cogn Ther Res 1993;17:359–77. Gotlib IH, Krasnoperova E, Yue DN, Joormann J. Attentional biases for McCabe SB, Gotlib IH. Selective attention and clinical depression: negative interpersonal stimuli in clinical depression. J Abnorm Psychol performance on a deployment of attention task. J Abnorm Psychol 2004;113(1):127–35. 1995;104:241–5. Gotlib IH, McCann CD. Construct accessibility and depression: an McNeely HE, Christensen BK, West R, Alain C. Changes in neurophysi- examination of cognitive and affective factors. J Pers Soc Psychol ological correlates of conflict processing precede behavioral distur- 1984;47:427–39. bance in patients with schizophrenia. J Abnorm Psychol Gotlib IH, Yue DN, Joorman J. Selective attention in dysphoric 2003;112:679–88. individuals: the role of affective interference and inhibition. Cogn McNeely HE, Dywan J, Segalowitz SJ. ERP indices of emotionality and Ther Res 2005;29:417–32. semantic cohesiveness during recognition judgements. Psychophysiol- Hamilton M. A rating scale for depression. J Neurol Neurosurg ogy 2004;41:117–29. Psychiatry 1960;23:56–62. Nigg JT. On inhibition/disinhibition in developmental psychopathology: Harnishfeger KK. The development of cognitive inhibition: theories, views from cognitive and personality psychology and a working definitions, and research evidence. In: Dempster FN, Brainerd CJ, inhibition taxonomy. Psychol Bull 2000;126:220–46. editors. Interference and inhibition in cognition. New York: Academic Oldfield RC. The assessment and analysis of handedness: the Edinburgh Press; 1995. p. 175–204. Handedness Inventory. Neuropsychologia 1971;31:347–53. Hollon SD, Kendall PC. Cognitive self-statements in depression: devel- Picton TW, Bentin S, Bergp, Donchin E, Hillyard SA, Johnson RJR, et al. opment of an automatic thoughts questionnaire. Cogn Ther Res Guidelines for using human event-related potentials to study cogni- 1980;4:383–95. tion: recording standards and publication criteria. Psychophysiology Huynh H, Feldt LS. Estimation of the Box correction for degrees of 2000;37:127–52. freedom from sample data in the randomized block and split-plot Rebai M, Bernard C, Lannou J. The Stroop’s test evokes a negative brain designs. J Educ Stat 1976;1:69–82. potential, the N400. Int J Neurosci 1997;9:85–94. Joormann J. Selective attention in dysphoria: the role of inhibitory Rossion B, Joice CA, Cottrell GW, Tarr MJ. Early lateralization and processes. Cogn Emotion 2004;18:125–47. orientation tuning for face, word, and object processing in the visual Kaiser S, Unger J, Kiefer M, Markela J, Mundt C, Weisbrod M. cortex. Neuroimage 2003;20:1609–24. Executive control deficit in depression: event-related potentials in a Sattler JM, Ryan JJ. Assessment of children: revised and updated third Go/Nogo task. Psychiatry Res 2003;122:169–84. edition WAIS-III supplement. San Diego: Jerome M. Sattler; 1999. Kucera H, Francis WN. Computational analysis of present-day American Segal ZV, Gemar M, Truchon C, Guirguis M, Horowitz LM. A priming English. Providence: Brown University Press; 1967. methodology for studying self-representation in Major Depressive Lane RD, Reiman EM, Axelrod B, Yun L-S, Holmes A, Schwartz GE. Disorder. J Abnorm Psychol 1995;104:204–13. Neural correlates of levels of emotional awareness: evidence of an Segal ZV, Vella DD. Self-schema in major depression: replication and interaction between emotion and attention in the anterior cingulated extension of a priming methodology. Cogn Ther Res 1990;14: cortex. J Cogn Neurosci 1998;10:525–35. 161–76. Lau MA, Christensen BK, Hawley LL, Gemar MC, Segal ZV. Inhibitory STIM Software. Compumedics Neuroscan, El Paso, TX, USA; 2007. deficits for negative information in Major Depressive Disorder. Wechsler D. Wechsler Adult Intelligence Scale. 3rd Rev. ed. San Anto- Psychol Med 2007;37:1–11. nio: Psychological Corporation; 1997.

12 H.E. McNeely et al. / Clinical Neurophysiology xxx (2008) xxx–xxx

West R, Alain C. Event-related brain activity associated with the Stroop resonance imaging probe of the anterior cingulated aﬀective division. task. Brain Res 1999;8:157–64. Biol Psychiatry 1998;44:1219–28. West R, Alain C. Age-related decline in inhibitory control contributes to Wilkinson GS. The Wide Range Achievement Test-3. San Antonio: The the increased Stroop eﬀect observed in older adults. Psychophysiology Psychological Corporation; 1993. 2000;37:179–89. Williams JMG, Broadbent K. Distraction by emotional stimuli: use of a Whalen PJ, Bush G, McNally RJ, Wilhelm S, McInerney SC, Jenike MA, Stroop task with suicide attempters. Br J Clin Psychol 1986;25: et al. The emotional counting Stroop paradigm: a functional magnetic 101–10.

Please cite this article in press as: McNeely HE et al., Neurophysiological evidence of cognitive inhibition anomalies ..., Clin Neu- rophysiol (2008), doi:10.1016/j.clinph.2008.03.031