ELSEVIER Psychiatry Research 56 (1995) 59-70

PSYCHIATRY

RESEARCH

Abnormal processing of irrelevant information in attention deficit hyperactivity disorder

Cameron S. Carter*a, Penelope Krenerb, Marc Chaderjianb, Cherise Northcuttb, Virginia Wolfeb

‘University of Pittsburgh, School of Medicine, Wesiern Psychiatric instirure and Clinic, 381 I O’Hara Street, Pittsburgh, PA 15213-2593, USA

bUniversiry of California Davis, School of Medicine, Department of Psychiatry, Sacramento, CA 95817, USA

Received 15 July 1993; revision received 18 August 1994; accepted 21 September 1994

Abstract

The presence of a selective attention deficit in children with attention deficit hyperactivity disorder (ADHD) was investigated by administering a trial-by-trial version of the Stroop Color-Naming Task to children, aged 9-12, with ADHD (n = 19) and age-matched normal control children (n = 19). Performance was evaluated on both interference and facilitation components of the task. On the standard version of the task, with equal numbers of color words and

neutral words, children with ADHD showed increased Stroop interference (prolongation of color-naming times by color-incongruent stimuli) but normal amounts of facilitation (speeding of color naming by color-congruent stimuli). This finding suggests that children with ADHD show increased disruption of color-naming performance by task- irrelevant information, probably secondary to decreased attentional control over the interference process. In contrast

to findings of studies in adults, both groups of children failed to use an attentional strategy to reduce interference when they were administered blocks of trials that varied their expectancy for color word trials. This precluded a direct test of the diminished control hypothesis. There were no significant correlations between abnormal Stroop performance and impairment on the Continuous Performance Test or the Wisconsin Card Sorting Test or measures of IQ or reading

performance. The implications of these findings for our understanding of information-processing deficits in children with ADHD and of the neurobiological underpinnings of these deficits are discussed.

Keywords: ADHD; Selective attention; Stroop Color-Naming Task; Child psychiatry

1. Introduction

The existence of specific deficits in information processing in children with attention deficit

hyperactivity disorder (ADHD) has been the sub-

* Corresponding author, Tel: +1 (412) 624-0481; Fax: +1

(4 12) 624-9464.

ject of intensive investigation over the past decade (Douglas, 1988; Sergeant, 1988; Swanson et al., 1990). Recent studies that used more rigorous

selection criteria for ADHD subjects have produc- ed evidence supporting the existence of such deficits. The ability of children with ADHD to sus- tain attention, as measured by variations of the

Continuous Performance Test (Sykes et al., 1983;

0165-1781/95/$09.50 0 1995 Elsevier Science Ireland Ltd. All rights reserved

SSDI 0 165- 178 1(94)02509-H

60 C.S. Carter et al. /Psychiatry Research 56 (1995) 59-70

Schachar et al., 1988; van der Meere and Sergeant, 1988; Chee et al., 1989; Seidel and Joshko, 1990), has been most intensively investigated. It has also recently been reported, however, that ADHD children are impaired in their ability to selectively allocate attention to locations in the visual field (Swanson et al., 1991; Carter et al., 1993).

Along with identifying specific forms of atten- tional pathology in ADHD comes the challenge of linking these impairments to specific dysfunctional neural systems. Neuropsychological studies have linked cognitive deficits in ADHD to impaired frontal lobe functioning (Chelune et al., 1983; Gorenstein et al., 1989; Barkley et al., 1992). In particular, deficits in the performance of the Wisconsin Card Sorting Task and the Stroop Color-Naming Task have been consistently found in children with ADHD (for review, see Barkley et al., 1992). Both tasks are generally considered to be sensitive measures of frontal lobe functioning (Milner, 1963; Perret, 1974; Golden, 1976; Stuss and Bensen, 1986). Neuropsychological tasks are complex, however, and it is difficult to specify precisely the disrupted cognitive mechanisms im- plicated by impaired performance. For example, Barkley et al. (1992) have reported that children with ADHD showed impaired performance on a traditional version of the Stroop (1935) task in which colored stimuli are printed in columns on cards. In that study, impairment was present for color-word reading as well as for naming color- incongruent words (e.g., the word RED printed in blue, the Stroop interference condition). Thus, while the interference condition of the Stroop task has often been considered to be a measure of selec- tive attention, the pattern of impairment found could also reflect other deficits, such as in visual scanning, rapid naming of lists, or general reading fluency (Barkley et al., 1992). These authors stress- ed the need for detailed analyses with more precise measurement of reaction time in future studies directed toward relating cognitive impairment to disrupted functional neural systems in ADHD.

Although studies in children with ADHD have generally used the neuropsychological (or Stroop card) version of this task, most recent investiga- tions by cognitive psychologists have used trial-by- trial versions (Macleod, 1991). In this procedure,

stimuli (colored words) are presented one at a time on a computer screen or tachistoscope, and laten- ties for naming the color of each stimulus are measured with millisecond precision. This results in a highly simplified procedure in which stimulus color and semantic meaning are varied while other stimulus properties remain constant. When this form of the Stroop task is used, two reliable effects are found (Macleod, 1991). The first is Stroop in- terference, mentioned above. Relative to non- color neutral words (e.g., DOG), the latency for naming the color of color-incongruent words (e.g., RED printed in blue) is prolonged. The second effect is Stroop facilitation. The latency for nam- ing the color of color-congruent words (e.g., RED printed in red) is shorter than that with neutral words. These findings have been considered to reflect the effect of a relatively automatic but task- irrelevant process (word reading) upon color naming. Although these two effects have been demonstrated in adults, pilot studies in our labora- tory have confirmed that they are also reliably present in children as young as 7 years of age.

It has recently been shown that when normal adults are presented with blocks of stimuli in which they have a high expectancy that a stimulus will be a color word, they are able to selectively reduce the amount of interference associated with color-incongruent words, while the amount of facilitation remains constant (Tzelgov et al., 1992; Henik et al., 1993). This suggests that the use of the trial-by-trial version of the Stroop task can reveal two dissociable operations, one of which (interference) is amenable to top-down control, and one of which (facilitation) is more automatic (Tzelgov et al., 1992). A task with dissociable com- ponents renders it particularly useful for studies with neuropsychiatric patients (Posner et al., 1988), because separable components provide a built-in control for nonspecific deficits such as generalized slowing of responses.

The precise nature of the information-process- ing mechanisms revealed by the Stroop task re- mains controversial (Macleod, 1991), but the potential usefulness of this procedure in evaluating the efficiency of selective attention in ADHD is compelling. The increased interference suggested by earlier Stroop card studies suggests that these

C.S. Carter et al. /Psychiatry Research 56 (1995) 59-70 61

children are more susceptible to the disruptive ef- fects of task-irrelevant information. The purpose of the current study is to test the following two hypotheses with a single trial version of the Stroop task: (1) That children with ADHD demonstrate increased Stroop interference; specifically, that latencies of color-naming responses to color- incongruent color words are longer than those to non-color-related words. (2) That ADHD children show a deficit in the control of selective attention as indicated by a failure to reduce interference under conditions of high expectancy for a color- word stimulus; specifically, when expectancy for a color-word stimulus is high, latencies of response to color-incongruent words will be selectively decreased.

Since there is evidence in adults that the amount of interference can be controlled in a top-down manner under conditions of high expectancy for color-word stimuli, we included an expectancy manipulation in our procedure. We also evaluated the relationship of other aspects of the children’s cognitive and intellectual functioning to their Stroop performance. This allowed us to examine the question of whether deficits found in Stroop performance were related to more general deticien- ties in intellectual functioning.

2. Methods

2.1. Subjects The ADHD subjects were outpatients in a Uni-

versity Psychiatric Center and ranged in age from 9 to 12 years. All patients met DSM-III-R criteria for ADHD (American Psychiatric Association, 1987). Diagnoses were made by clinical interview of both parents and children conducted by a child psychiatrist (PK. and V.W.). Other inclusion cri- teria were a score more than 1.5 standard devia- tions above the mean for the subject’s age and gender on the ADHD Rating Scale (Barkley, 1991) and T scores above 70 on the Hyperactivity Index of the Connors Teacher Rating Scale (Connors, 1973). On the Connors Scales, T scores below 50 are in the normal range, and T scores above 70 are considered to be clinically significant. A score that is 1.5 standard deviations above the mean on the

Home Situations Questionnaire (Barkley, 1991) confirms the clinical impact of ADHD symptoms.

A total of 20 patients and 20 control subjects were recruited for the study. Control subjects were recruited from the same schools as the ADHD subjects, were age- and gender-matched to the ADHD subjects, and were paid $50 for participa- tion. All control children also had a psychiatric evaluation to rule out psychiatric disorders. Both subjects and their parents gave written informed consent to participate in the research. Children who met criteria for comorbid mood or anxiety disorders, conduct disorder, or oppositional de- fiant disorder were excluded from the study group. Color blindness was also a reason for exclusion. In addition to the clinical evaluation, the Harter De- pression Inventory (Harter and Kowakiwski, 1987) was administered to the children. All chil- dren and parents were given respective versions of the revised Manifest Anxiety Scale (Reynolds and Richmond, 1978). Subjects were excluded from study if they scored in symptomatic ranges on these measures. The Achenbach Child Behavior Checklist (Achenbach and Edelbrook, 1986) was also administered, and children were excluded if they had T scores over 80 on more than one inter- nalizing scale. All children received a psychiatric evaluation and were tested with the Wechsler In- telligence Scale for Children-Revised (WISC-R; Wechsler, 1974). A Full-Scale IQ score below 80 was an exclusion criterion for both ADHD and control subjects. Cognitive testing was also done with the Test of Variables of Attention, a non- verbal Continuous Performance Test (CPT; Greenberg, 1991), and with a computerized ver- sion of the Wisconsin Card Sorting Test (WCST). All patients and all but two control subjects also completed the reading section of the Wide Range Achievement Test (WRAT-R). Cognitive testing was performed when children had been withdrawn from psychostimulant medication (methylpheni- date or dextroamphetamine) for at least 3 days.

Control subjects were matched individually to the ADHD subjects for age and sex. Control sub- jects were excluded if they met criteria for any mental disorder or had any first-degree family his- tory of ADHD. One male control subject was found to have color blindness on the day of testing

62 C.S. Carter et al. /Psychiatry Research 56 (1995) 59-70

and therefore did not perform the Stroop proce- dure. The data for one female patient were lost due to a technical error. Deletion of these two subjects resulted in experimental and control groups con- sisting of 19 subjects each: 4 girls and 15 boys in the ADHD group, and 5 girls and 14 boys in the control group.

Table 1 summarizes the age and gender of the two groups, as well as their performance on the WCST and the CPT. The groups were closely mat- ched for age (t = -0.31, df= 36, P > 0.90) and sex (x2 = 0.15, df= 1, P > 0.71). Means for the mea- sures of ADHD were as follows: ADHD Rating Scale - patients 3 1.5, control subjects 4.15; Con- nors Parent Rating Scale - patients 77.65, control subjects 43.85; Hyperactivity Index Home Situa- tions Rating Scale number of items-patients 7.5, control subjects 1.75; Home Situations Rating Scale severity on a scale of 1 to 5 - patients 4.5, control subjects 1.75.

Children with ADHD had significantly more errors of omission on the CPT (t = 2.37, df = 36, P < 0.03), but not errors of commission (t = 1.43, df = 36, P > 0.17), than control subjects, indi- cating that they may have been a primarily inattentive rather than an impulsive group (Green- berg, 1991). Children with ADHD also showed

trends toward making significantly more perseverative errors (t = 1.92, df = 36, P c 0.06) and sorting fewer categories (t = - 1.75, df = 36, P < 0.09) than the control subjects on the WCST.

There was a nonsignificant trend for WISC-R Full-Scale scores to be lower in the ADHD sub- jects than in the control subjects (t = -1.96, df = 36, P < 0.06). The ADHD children scored significantly worse on the Performance subscale of the WISC-R (t = -2.7, df = 36, P < 0.01). The two groups did not differ on the Verbal subscale of this measure (t = - 0.59, df = 36, P > 0.60). The ADHD group scored significantly worse on the reading section of the WRAT (t = -3.6, df = 34, P < 0.001).

2.2. Procedures Stimuli were presented using Microexperimental

Laboratory (Schneider, 1988) software and an IBM PC-286 compatible computer (Hewlett- Packard Vectra ES) with VGA color monitor (Hewlett-Packard, Sunnyvale, CA). The latencies of subjects’ spoken responses were collected to millisecond accuracy through a microphone and a voice-operated relay (Gerbrands Model 1341T, Arlington, MA) connected to the PC game port. The microphone was placed 48 cm from the

Table 1 Mean demographic, cognitive, and psychoeducational characteristics of children with ADHD and normal control subjects (SD)

ADHD

(n = 19)

Control

(n = 19)

Age Sex F:M

CPT % commission errors

CPT % omission errors

WCST categories sorted

WCST % perseverative errors

WISC-R Full-Scale IQ

WISC-R Verbal IQ WISC-R Performance IQ

WRAT reading

CPT reaction time (ms)

CPT variability

10.58 (1.26)

4.15

10.47 (8.04)

IO.6 (11.4)

3.9 (2.0)

21.7 (11.4)

94.3 (13)t

94.9 (15.8)t

96.2 (12.2)t 87.4 (14.9)t

598 (142)

279 (I 32)

10.63 (1.21)

5.14

7.37 (5.00)

3.7 (5.6)

4.8 (1.1)

16.1 (5.5)

102.5 (13)

91.6 (23.6) 108.2 (13.7)

103.4 (12.5)

537 (143)

188 (85)

=P > 0.90

bP > 0.71

“P > 0.17

aP < 0.025’

aP < 0.09

aP < 0.06

=P < 0.06

aP > 0.60

aP < 0.01; aP < 0.003’

aP > 0.19

=P < 0.02*

aUnpaired t test.

b~2 test.

*Significant at P < 0.05 level, 2-tailed comparison.

tl8 subjects.

C.S. Carter et al. /Psychiatry Research 56 (1995) 59-70 63

monitor screen, and subjects sat directly in front of it, though their heads were not restrained from moving. Word stimuli were formed from capital letters of the system’s standard character set; char- acters were 5 mm high x 3 mm wide, and words subtended a visual angle that was approximately 0.5” high and between 1” and 2” wide.

Stroop stimuli were formed from the words DOG, BEAR, TIGER, MONKEY, RED, BLUE, GREEN, and YELLOW presented in the colors red, blue, green, and yellow. Stimuli were presented one at a time, and there were three types of trials presented during the experiment. On color-congruent trials, colors matched words (e.g., the word GREEN written in green), on color- incongruent trials, colors mismatched words (e.g., the word RED written in green), and on color- neutral trials, colors were unrelated to words (e.g., the word DOG written in green). Three blocks of 192 trials were formed, each representing different proportions of neutral trials. Standard or no- expectancy (50% neutral, 50% color words) blocks consisted of 96 color-neutral trials, 48 color- incongruent trials, and 48 color-congruent trials. High-color-word expectancy (25% neutral, 75% color words) blocks consisted of 48 color-neutral trials, 72 color-incongruent trials and 72 color- congruent trials. Low-color-word expectancy (75% neutral, 25% color words) blocks consisted of 144 color-neutral trials, 24 color-incongruent trials, and 24 color-congruent trials. Rest periods were given every half-block of 96 trials, and a set of 32 practice trials preceded each block. Rest periods ranged from 30 seconds to 3 minutes, as needed by the child. Children were instructed about expec- tancy by a message on the screen that informed them of a change in proportions of neutral words when a change occurred. The message indicated that half of the words would be animal names, while half would be color names (standard Stroop block); that most of the words would be animal names (low-expectancy block); or that most of the words would be color names (high-expectancy block). The experimenter drew attention to the in- struction screen and repeated or paraphrased it, including a reminder to the child that the task was always the same: to name the colors and ignore the words. The order of presentation of data blocks

was counterbalanced within each experimental group.

At the outset of a data session, subjects were given standardized instructions informing them that a word would appear in the center of the com- puter screen in front of them, that it would be printed in one of four different colors, and that they were to try to ignore what the word said and name the color it was written in as quickly and ac- curately as they could. They were further told that words would sometimes be color words and other times be animal words; information about how often words of each type would appear was given before each practice set, as well as reminders that subjects were to ignore the word and name the color. A block of trials was initiated by an ex- perimenter key press; a trial began with presenta- tion of a Stroop stimulus, which remained on the screen until the subject made a verbal response. The experimenter tapped a key on the keyboard to code the subject’s response and to begin the next trial. An effort was made to keep a steady pace of 1 to 2 seconds between trials. The entire procedure took 25 to 35 minutes.

3. Results

Table 2 presents the mean reaction times for correct responses and error rates for the two groups under each of the three expectancy condi- tions. Table 3 shows the significant main effects and interactions.

3.1. Stroop interference and facilitation The amounts of interference and facilitation

demonstrated by the two groups were compared under conditions of no expectancy (i.e., the condi- tion in which 50% of the stimuli were color words and 50% were neutral words). This condition gives us a measure of baseline Stroop performance. Zn- terference is defined as the difference in reaction times between incongruent and neutral stimuli, while facilitation is defined as the difference be- tween reaction times for congruent and neutral stimuli.

Interference: In the analysis of variance (ANOVA) of Stroop interference under the stan- dard no-expectancy condition, with group as the

64 C.S. Carter et al. /Psychiatry Research 56 (1995) 59-70

Table 2

Mean color-naming latencies and error rates for ADHD and control groups for each word type, under each expectancy condition (SD)

Block Word Type

ADHD (n = 19)

RT (ms) % errors

Control (n = 19)

RT (ms) % errors

Standard (50% color words,

50% neutrals)

Low expectancy (25% color

words, 75% neutrals)

High expectancy (75% color

words, 25% neutrals)

Neutral

Incongruent

Congruent

Neutral

Incongruent

Congruent

Neutral

Incongruent

Congruent

1025.3 (166.1) 3.4

1172.1 (241.8) 13.9

882.8 (175.1) 1.4

995.0 (169.7) I .8

1142.2 (317.6) 16.7

861.5 (181.0) 0.2

1121.1 (246.9) 1.8

1207.4 (347.2) 2.6

910.8 (210.7) 1.6

939.7 (282.8) 1.7

1013.1 (301.9) 12.7

7.38.6 (202.3) 0.5

968.1 (286.6) 1.7

1114.6 (362.6) 15.8

806.2 (205.3) 0

1060.2 (383.8) 1.3

1079.5 (310.5) 3.3

832.6 (218.2) 0.3

between-subjects factor x word type (incongru- ent, neutral), the main effect of group was not significant (P= 2.3; df = 1, 36; P > 0.13). As expected, there was a significant main effect of

Table 3

Means for significant main effects and interactions from reac-

tion time analysis

Baseline

Interference Word type

Word

type x group (ADHD)

(Control) Facilitation Word type

Incongruent Neutral 1093 983

1172 1025

1013 939

Congruent Neutral 883 983

Expectancy effects

Interference Incongruenl Neutral Word type 1136 1036

Expectancy 1055 (25% color I1 I7 (75% color

words) words)

Expectancy 1128 982

(25% color)

x Word type 1143 1091

(75% color)

Expectancy effects Facilitation Word type

Expectancy

Expectancy

(25% color)

x Word type (75% color)

Congruent 850

908 (25% color

words)

834

866

Neutral IO36 978 (75% color

words)

982

1091

word type (F = 48; df = 1, 36; P < .OOOl), which indicated that incongruent words produced inter- ference. The critical group x word type inter- action was also significant (F = 5.3; df = 1, 36; P < 0.03). Children with ADHD showed more sen- sitivity to interference effects than did normal control subjects in naming the colors of color- incongruent stimuli.

Facilitation: When the effect of Stroop facilita- tion was analyzed under the no-expectancy condi- tion, the main effect of group was again not significant (F = 1.9; df = 1, 36; P > 0.17). The main effect of word type was significant (F = 66.5; df = 1, 36; P < O.OOOl), a finding which suggested that color-congruent words facilitated color- naming times. The group x word interaction was not significant (F = 0.14; df = 1, 36; P > 0.71). The groups did not differ on Stroop facilitation.

Fig. 1 shows the magnitude of interference and facilitation during standard single trial Stroop task performance for children with ADHD and normal control subjects. These results suggest that ADHD children demonstrate a specific deficit in process- ing disruptive task-irrelevant semantic informa- tion. This deficit consists of increased sensitivity to interference effects in naming the colors of color- incongruent words.

3.2. Expectancy effects Fig. 2 shows the mean reaction times for each of

the two expectancy conditions (high and low). The effects of expectancy on interference and facilita- tion were evaluated by comparing the two groups’ performance on these components of the task

C.S. Carter et al. /Psychiatry Research 56 (1995) 59-70 65

300 H ADHD

250 - 0 Control T T

Interference Facilitation

Fig. I. Interference and facilitation scores for ADHD and con-

trol subjects. Interference scores represent the difference be-

tween reaction times (RTs) for incongruent vs. neutral stimuli.

Facilitation scores represent the difference between RTs for

neutral vs. congruent stimuli. Bars indicate SDS. * indicates P < 0.03, baseline Stroop interference ANOVA.

under conditions of low expectancy (25% color words, 75% neutral words) and high expectancy (75% color words).

Interference: An ANOVA that compared the two groups’ color-naming times for neutral and in- congruent stimuli under high- and low-expectancy conditions revealed a main effect of word type (F= 18.4; df= 1, 36; P c 0.0001) and a main effect of expectancy (F = 7.3; df = 1, 36; P < 0.01). The main effect of group was not significant (F = 0.43; df = 1, 36; P < 0.52). The significant main effects indicate that incongruent stimuli

---- ADHD

- Control

A Incongruent words

0 Neutral words

0 Congruent words

c 2 900

! -0

___--- ___---

1; 3 a 800

L I I

Low High

Expectancy

Fig. 2. Mean reaction times of ADHD and control subjects for

each of the 3 word types (neutral, incongruent, and congruent)

under conditions of low (25% of stimuli) and high (75% of

stimuli) expectancy for color words.

resulted in interference and that reaction times were generally slower under conditions of high expectancy. There was also a significant expectan- cy x word type interaction (F = 4.2; df = 1, 36; P < 0.05), which indicated a reduction of interfer- ence under the high-expectancy condition. No interactions involving group were significant (P > 0.15 in all cases).

Facilitation: The ANOVA that compared the groups’ color-naming times under the two expec- tancy conditions also revealed main effects of word type (F = 77.2; df = 1, 36; P < 0.0001) and expectancy (F = 15.4; df = 1, 36; P < 0.0001). Compared with neutral words, color-naming times of color-congruent words were speeded. Color- naming times were slower under the high- expectancy condition. A significant word type x expectancy interaction (F = 7.8; df = 1, 36; P < 0.01) indicated that facilitation was increased under high- relative to low-expectancy conditions. As was the case for the expectancy analysis of in- terference, there were no significant interactions involving group (P > 0.36 in all cases).

The above analyses suggest that children reduce interference and increase facilitation under condi- tions of high expectancy. However, inspection of Fig. 1 suggests that these effects are not due to faster response times for incongruent and con- gruent stimuli, but instead are mostly derived from slower responses to neutral words. This interpreta- tion was confirmed by a post hoc analysis that used paired t tests to compare color-naming times for each word type under the two expectancy con- ditions (incongruent mean difference = 15.6 ms, SD = 252.0, t = 0.37, df = 35, P > 0.71; con- gruent mean difference = 34.6 ms, SD = 138.7, t = 1.5, df = 35, P > 0.16; neutral mean differ- ence = 113.0 ms, SD = 143.9, t = 4.7, df = 35, P < 0.001). Although we found the predicted in- creased interference in baseline Stroop perfor- mance in children with ADHD, the results of the expectancy effects did not indicate that the ADHD group had a deficit that controlled this compo- nent. No differences were found between the two groups’ performance with this manipulation. However, the change in the pattern of perfor- mance seen in both groups of children with this manipulation was quite different from that seen in normal adults (Tzelgov et al., 1992; Henik et al.,

66 C.S. Carter et al. / Psychiarry Research 54 (1995) 59-70

1993; Carter et al., unpublished data). Adults type x group interaction (F < 0.02, P > 0.89). reduce interference but not facilitation with expec- There was a trend for a significant expectancy x tancy. In contrast, both groups of children word type interaction (F = 3.86; df = 1, 36; P c reduced interference and increased facilitation. 0.06), which did not vary with group (F = 0.39; However, instead of improving color-naming df = 1, 36; P > 53). Errors associated with in- times under incongruent or congruent conditions, congruent trials decreased more under high- children showed prolonged color-naming times expectancy conditions (12.9% from 16.3%) than under the high-expectancy condition, particularly errors associated with neutral trials (1.6% from for neutral stimuli. Neither ADHD subjects nor 1.8%). It is likely that this difference reflects the control subjects appeared to use an attentional very low overall rate of color-naming errors strategy to control interference when they had a associated with neutral words, which limits the high level of expectancy that a stimulus would be degree to which performance can improve on this a color word. parameter.

3.3. Error analysis Interference: An ANOVA that compared color-

naming error rates in the two groups for neutral and incongruent trials under the baseline no- expectancy condition indicated a main effect of word type (F = 0.89; df = 1, 36; P < O.OOOl), but neither a main effect of group (F = 0.89; df = 1, 36; P > 0.35) nor a group x word type interac- tion (F = 0.05; df = 1, 36; P > 0.81). Although color-naming errors were significantly greater for incongruent words, the two groups did not differ in their overall error rates or the distribution of errors according to stimulus type.

Facilitation: The ANOVA for color-naming error rates for neutral and congruent words under the no-expectancy condition revealed the expected main effect of word type (F = 5.94; df = 1,36; P < 0.03), indicating that error rates were lower for congruent than for neutral words. There was no main effect of group (F = 2.44; df = 1, 36; P > 0.12) or group x word type (F = 0.46; df = 1, 36; P > 0.50). The two groups did not differ in error rates or in the distribution of color-naming errors between the two word types.

Facilitation: In the evaluation of the effect of ex- pectancy on error rates for neutral and congruent words, the expected main effect of word type was present (F = 19.3; df = 1, 36; P < O.OOl), in- dicating that fewer color-naming errors were made during congruent than during neutral trials (0.5% vs. 1.6%, respectively). There was neither a signifi- cant main effect of group (F = 2.21; df = 1, 36; P > 0.14) nor of expectancy (F = 1.07; df = 1, 36; P > 0.30). A significant expectancy x word type in- teraction was present (F = 6.16; df = 1, 36; P < 0.02). This suggests that while error rates decreas- ed under high expectancy for neutral words (from 1.75% to 1.5%), error rates for congruent words went from 0.1% under the low-expectancy condi- tion to 0.8% under the high-expectancy condition. No interactions involving group reached signifi- cance (all P > 0.2). With error rates in the range of 0 to 1.8%, however, it is difficult to interpret either the lack of expectancy effects or the interac- tion between expectancy and word type; with per- formance at near perfect levels, there is little scope for improvement in error rates. The significance of the very small changes observed with the expectan- cy manipulation for facilitation trials is uncertain.

3.4. Expectancy effects Interference: In the evaluation of the effects of

expectancy on color-naming errors for in- congruent and neutral trials, a main effect of ex- pectancy was present (F = 5.4; df = 1, 36; P < 0.03). Error rates were lower under the high- expectancy condition (7.25% vs. 9.5%). The ex- pected main effect of word type was also present (F = 110.6; df = 1, 36; P < O.OOOl), but there was neither a main effect of group nor a word

To summarize the analysis of color-naming er- rors, the two groups did not differ significantly in the number of color-naming errors overall or in the difference in errors associated with in- terference or facilitation. Likewise, they did not differ in the numbers of errors associated with in- creasing expectancy for color words. For in- congruent trials, where error rates were large enough to show substantial decreases in perfor- mance, both groups made fewer errors under the high-expectancy condition. Thus, the increase in

C.S. Carter et al. /Psychiatry Research 56 f 1995) 59-70 61

Table 4 Pearson’s correlation coefftcients and probability values for the relationship between Stroop interference scores and perfor- mance on key components of the CPT, the WCST, and psychoeducational performance

% omission errors WCST categories WCST % perseverative

errors WISC-R Full-Scale IQ WISC-R Verbal IQ WISC-R Performance IQ WRAT reading

r P

0.05 0.78 -0.02 0.92

0.13 0.44

-0.19 0.27 -0.08 0.62 -0.20 0.24 -0.11 0.52

color-naming time under the high-expectancy con- dition may reflect the use of an immature strategy by both groups of children; specifically, a speed- accuracy trade-off.

3.5. Cognitive correlational analyses Table 4 summarizes the results of the planned

correlational analysis between Stroop interference and performance on the CPT, on the WCST, and on WISC-R Full-Scale and Verbal subscale scores. There is no evidence of any correlation between in- terference and errors of omission on the CPT (r = 0.05, P > 0.78). There is also no evidence of any significant correlation between Stroop in- terference and the number of categories sorted (r = -0.01, P > 0.92) or the percent perseverative errors (r = -0.13, P > 0.44) on the WCST. There were no significant correlations between in- terference scores and WISC-R Full-Scale scores (r = -0.19, P > 0.27) or the Verbal or Perfor- mance subscales (r = -0.08, P > 0.62; r = -0.20, P > 0.24, respectively). Similarly, although the ADHD children were poorer readers as measured by the reading subsection of the WRAT, there was no significant correlation between performance on this measure and the interference score (r = -0.11, P > 0.52).

4. Discussion

Our findings confirm the hypothesis that chil- dren with ADHD are more vulnerable to the ef- fects of disruptive task-irrelevant information in

the trial by trial version of the Stroop Color- Naming Task than are normal children. Their vulnerability is evident from the increased Stroop interference seen in the analysis of color-naming latencies under conditions where no expectancy for color words is demanded by the task. Our sec- ond hypothesis was that ADHD children would be selectively impaired in their ability to reduce in- terference under conditions of high expectancy. However, neither normal nor ADHD children ap- peared to use the attentional strategy that we had predicted based on results from studies of normal adults, who selectively reduce interference under high-expectancy conditions. Rather, both groups of children decreased interference and increased facilitation for color-naming trials, but they did this by increasing color-naming times overall and for neutral words in particular. The longer re- sponse latencies under the high-expectancy condi- tion were accompanied by decreased error rates. This finding suggests the use of an age-appropriate strategy by our young subjects under these condi- tions, but it precludes a direct test of attentional control over Stroop interference.

Relative to neutral stimuli, ADHD children took longer to name the colors of color- incongruent words. However, they did not make more color-naming errors than normal children. The response (the color of the conflicting stimulus) was as likely to be given correctly by children in either group. This finding suggests that the in- creased interference evident from the reaction time data was not due simply to impulsive responding to the irrelevant dimension of the stimulus by the ADHD children but reflected a specific information-processing deficit.

It has been suggested that the increased Stroop interference previously reported in ADHD with Stroop card versions of the procedure may have been an artifact of an overall slowing of responses or of other nonspecific problems in scanning lists of words (Barkley et al., 1992). However, our tind- ings suggested that children with ADHD did have increased interference for color naming by color- incongruent stimuli. Overall, color-naming times were not significantly longer in children with ADHD than in control subjects; therefore, general slowing was unlikely to have accounted for the in- creased interference observed. In the current pro-

68 C.S. Carter et al. /Psychiatry Research 56 (1995) 59-70

cedure, subjects were not required to scan an array of stimuli, as in the card version of the Stroop task. Rather, stimuli were presented one at a time at a central location on the computer screen, Thus, scanning deficits or problems in dealing with mul- tiple stimuli would not account for the perfor- mance of the ADHD group. Only one of the two dissociable components of this procedure was ab- normal in the ADHD group. As is evident in Fig. 1, while interference scores were significantly larger in the ADHD group, facilitation scores were of very similar magnitude in both groups. Therefore, we interpret the finding of increased Stroop interference as indicating that children with ADHD have a specific information-pro- cessing abnormality. This deficit is characterized by greater disruption of color naming by irrele- vant, conflicting semantic information. The ab- sence of any correlational relationship between Stroop interference scores and impairments on the CPT or the WCST further supports the specificity of this deficit.

Previous studies of Stroop performance in child- ren have linked the development of Stroop interference to reading ability (Comalli et al., 1962; Schiller, 1966; Fournier et al., 1975). The better the child’s reading ability, the greater the magnitude of Stroop interference. Comparisons of our subjects’ scores on the reading section of the WRAT suggest that the ADHD children were significantly worse readers than control children. Since the ADHD group were worse readers, they would be expected on this basis to show smaller than normal amounts of interference, but they showed increased interference. Therefore, differ- ences in reading ability are unlikely to account for the differences in interference between the two groups in the current study. There were also no significant correlations between WISC-R total or subscale scores, or WRAT reading scores, and the amount of Stroop interference. The lack of signifi- cant correlations further supports the interpreta- tion that the increased interference shown by the ADHD group reflects a specific information- processing deficit.

While these results confirm that ADHD child- ren are more susceptible to the disrupting effects of conflicting task-irrelevant information on color-

naming performance, we are unable to draw de- finitive conclusions about the precise nature of the attentional deficit revealed by this study. As Macleod (1990) has pointed out, no single atten- tional mechanism captures the wealth of deter- minants of Stroop performance evident from the very extensive literature on this subject. An early filtering deficit, a failure to modulate spreading ac- tivation in semantic networks (Seymour, 1974; Tzelgov et al., 1992) a failure to inhibit prepotent response tendencies (Douglas, 1983; Barkley et al., 1992), or even a more widely distributed distur- bance affecting multiple levels of processing could account for this effect. Variations of the single trial Stroop task may be implemented in future studies to test each of these hypotheses systematically.

Neuropsychological studies have suggested that impaired performance on the Stroop task reflects disruption of frontal lobe functioning (Anderson et al., 1973). The combination of impulsiveness, poor self-monitoring, and inattention seen in ADHD children has led to the speculation that these children have frontal-lobe impairment (Ben- son 1991). Preliminary observations using positron emission tomography (PET) in adult patients with ADHD support this hypothesis (Zametkin et al., 1991). The anterior cingulate cortex, another structure linked to selective attention (Posner and Peterson, 1990), has been implicated in the in- terference effect in PET activation studies of nor- mal adults (Pardo et al., 1990; Bench et al., 1993). The results of the current study suggest that the Stroop task may be a useful tool for incorporation into functional imaging studies that seek to iden- tify the neural substrates of selective attention dysfunction in children with ADHD.

In a parallel distributed processing (PDP) model of Stroop performance described by Cohen and Servan-Schreiber (1992), attentional control was decreased and interference increased when the gain associated with activating connections be- tween nodes in the network was decreased. This manipulation was intended to simulate the effects of a physiological reduction of catecholaminergic neuromodulatory effects in actual neural net- works. It has been proposed that a reduction of catecholamines in frontal-striatal neural systems underlies the attentional pathology in children

C.S. Carter et al. /Psychiatry Research 56 (1995) 59-70 69

with ADHD (Levy, 1991). Further light on the physiological role of catecholamines in the im- paired information processing in ADHD may be shed by studies that combine cognitive modeling with empirical studies of the effects of phar- macological agents on cognitive performance in ADHD.

Acknowledgements

The authors are grateful to Michelle Myers for her assistance with the psychoeducational testing, Michelle Muncell for her administrative assis- tance, and Jane Rachford and Erin Petrelle for assistance with the manuscript. The work reported was supported, in part, by a Young Investigator Award to Cameron S. Carter, M.D., from the Na- tional Alliance for Research in Schizophrenia and Affective Disorders.

References

Achenbach, T.M. and Edelbrook, C. (1986) Manual for fhe Child Behavior Checklist and Revised Child Behavior Profile.

University of Vermont, Burlington, Vt. American Psychiatric Association (1987) DSM-III-R: Diagnos-

tic and Statistical Manual of Mental Disorders. 3rd rev. edn. APA, Washington, D.C.

Anderson, R.P., Halcombe, C.G. and Doyle, R.B. (1973) The measurement of attention deficits. Except Child 39,

534-539.

Barkley, R.A. (1991) Attention Deficit Hyperactivity Disorder.

A Clinical Workbook. Guilford Press, New York, pp. 46-67.

Barkley, R.A., Grodzinsky, G. and Dupaul, G.J. (1992) Fron- tal lobe functions in attention deficit disorder with and without hyperactivity: a review and research report. J Ab-

norm Child Psycho1 20, 163-188. Bench, C.J., Frith, C.D., Grasby, P.M., Friston, K.J., Paulesu,

E., Frackowiak, R.S.J. and Dolan, R.J. (1993) Investiga- tions of the functional anatomy of attention using the Stroop test. Neuropsychologia 3 I, 907-922.

Benson, D.F. (1991) The role of frontal dysfunction in atten- tion deficit hyperactivity disorder. J Child Nemo1 6(Suppl.), S9-S12.

Carter, C.S., Krener, P., Chaderjian, M., Northcutt, C. and Wolfe, V. (1993) Attentional asymmetry in ADHD: evi- dence for a right hemispheric deficit. Biol Psychiatry 33,

l08A.

Chee, P., Logan, G., Schachar, R., Lindsay, P. and Wachsmuth, R. (1989) Effects of event rate and display time on sustained attention in hyperactive, normal, and control children. J Abnorm Child Psychoi 17, 371-391.

Chelune, G.J., Ferguson, W., Koon, R. and Dickey, T.O. (1983) Frontal lobe disinhibition in attention deficit disor- der. Child Psychiatry Hum Dev 16, 221-234.

Comalli, P.E., Wagner, S. and Werner, H. (1962) Interference effects of Stroop color word test in childhood, adulthood and aging. J Genet Psycho/ 100, 47-53.

Cohen, J.D. and Servan-Schreiber, D. (1992) Context, cortex and dopamine: a connectionist approach to behavior and biology in schizophrenia. Psycho1 Rev 99, 45-77.

Connors, C.K. (1973) Rating scales for use in drug studies with children. Psychopharmacol Bull 9, 24-84.

Douglas, V.I. (1983) Attention and cognitive problems. In: Rutter, M. (Ed.), Dev Neuropsychiatry. Guilford Press, New York, pp. 280-329.

Douglas, V.I. (1988) Cognitive deficits in children with atten- tion deficit disorder with hyperactivity. In: Bloomingale, L.M. and Sergeant, J. (Eds.), Attention Deficit Disorder:

Criteria, Cognition, Intervention. Pergamon Press, New York.

Fournier, P.A., Mazzarella, M.M., Ricciardo, M.M. and Fingeret, A. (1975) Reading level and the locus of in- terference in the Stroop color word task. Percept Mot Skills

41, 239-42.

Golden, C.J. (1976) Identification of brain disorders by the Stroop color and word test. J Clin Psycho1 32, 654-58.

Gorenstein, E.E., Mammato, C.A. and Sandy, J.M. (1989) Per- formance of inattentive-overactive children on selective measures of prefrontal type functioning. J Chn Psycho/ 45,

619-632.

Greenberg, L.R. (1991) The Test of Variables of Attention

(TOVA). Universal Attention Disorders, Inc., Los Alamitos, Calif.

Hatter, S. and Kowakiwski, M. (1987) Manual for the Dimen-

sions of Depression Profire for Children and Adolescents.

University of Denver Press, Denver, Colo. Henik A., Singh J., Beckley D.J. and Rafal, R. (1993)

Disinhibition of automatic mental word reading in Parkin- son’s disease. Correx 29, 589-599.

Levy, F. (1991) The dopamine theory of attention deficit disor- der (ADHD). Aust NZ J Psychiatry 25, 277-283.

Macleod, CM. (1990) Half a century of research on the Stroop effect: an integrative review. Psycho1 Bull 109, 163-203.

Milner, B. (1963) Effects of different brain lesions on card sort- ing. Arch Neural 9, 90-100.

Pardo, J.V., Pardo, P.J., Janer, K.W. and Raichle, M.E. (1990) The anterior cingulate cortex mediates processing selection in the Stroop attentional conflict paradigm. Proc Nat1 Acad

of Sci USA, 87, 256-259.

Perret, E. (1974) The left frontal lobe of man and the suppres- sion of habitual responses in verbal categorical behavior. Neuropsychologia 12, 323-330.

Posner, MI., Early, T.S., Reiman, E., Pardo, P.J. and Dhawan, M. (1988) Asymmetries of attentional control in schizo- phrenia. Arch Gen Psychiatry 45, 814-821.

Posner, M.I. and Peterson, S.E. (1990) The attention system of the human brain. Annu Rev Neurosci 13, 25-42.

Reynolds, C.R. and Richmond, B.O. (1978) What I think I feel:

70 C.S. Carter et al. /Psychiatry Research 56 (1995) 59-70

a revised measure of children’s manifest anxiety. J Abnorm Child Psycho1 6(2) 27 I-280.

Schachar, R., Logan, G., Wachsmuth, R. and Chajczyk, D.

(1988) Attaining and maintaining preparation: a com-

parison of attention in hyperactive, normal and disturbed

control children. J Abnorm Child Psycho1 16, 361-378. Schiller, P. (1966) Developmental study of color-word

interference. J Exp Psycho1 72, 105-108. Schneider, W. (1988) Microexperimental Laboratory: an inte-

grated system for P.C. compatibles. Behav Res Methods, Instruments and Computers 30, 206-2 17.

Sergeant, J.A. (1988) In search of processing deficits of atten-

tion in ADD-H children. In: Bloomingale, L.M. and

Sergeant, J. (Eds.), Attention Deficit Disorder: Criteria, Cognition, Intervention. Pergamon Press, New York.

Seidel, W.T. and Joshko, M. (1990) Evidence of difftculties in

sustained attention in children with ADHD. J Abnorm Child Psycho1 18, 217-229.

Seymour, P.H.K. (1974) Stroop interference with response,

comparison, and encoding stages in a sentence-picture com-

parison task. Mem Cognit 2, 19-26. Stroop, J.R. (1935) Studies of interference in serial verbal reac-

tions. J Exp Psycho1 18, 643-662.

Stuss, D.T. and Benson, D.F. (1986) The Frontal Lobes. Raven

Press, New York.

Swanson, J.M., Posner, M.I., Potkin, S., Bonforte, S., Youpa,

D., Fiore, C., Cantwell, D. and Crinella, F. (1991) Activat-

ing tasks for the study of visual-spatial attention in ADHD

children: a cognitive anatomic approach. J Child Neural 6(Suppl.), Sll7-Sl25.

Swanson, J.M., Shea, C., McBurnett, K., Potkin, S.G., Fiore,

C. and Crinella, F. (1990) Attention and hyperactivity. In:

Enns, J.T. (Ed.) The Development of Attention: Research and Theory. ElsevierMorth-Holland, Amsterdam.

Sykes, D.H., Douglas, V. and Morgenstern, G.L. (1973) Sus-

tained attention in hyperactive children. J Child Psycho1 Psychiatry 14, 213-220.

Tzelgov, J., Henik, A. and Berger, J. (1992) Controlling Stroop

effects by manipulating expectations for color words. Mem

Cognit 20, 727-735. van der Meere, J. and Sergeant, J. (1988) Controlled processing

and vigilance in hyperactivity: time will tell. J Abnorm Child Psycho1 16, 641-655.

Wechsler, D. ( 1974) Manual for the Wechsler Intelligence Scale for Children - Revised. The Psychological Corporation,

New York.

Zametkin, A.J., Nordahl, T.G., Semple, W.E., King, A.C.,

Rumsey, J., Hamberger, S. and Cohen, R.M. (1991) Brain

metabolism in hyperactive adults with childhood onset. N

Engl J Med 323, 1361-1366.