Event-related potential assessment of information processing after closed head injury

Transcription

1 Psychophysiology, 40 (2003), Blackwell Publishing Inc. Printed in the USA. Copyright r 2003 Society for Psychophysiological Research Event-related potential assessment of information processing after closed head injury CONNIE C. DUNCAN, a,b MARY H. KOSMIDIS, b AND ALLAN F. MIRSKY b a Clinical Psychophysiology and Psychopharmacology Laboratory, Department of Psychiatry, Uniformed Services University of the Health Sciences, Bethesda, Maryland, USA b Section on Clinical and Experimental Neuropsychology, National Institute of Mental Health, Bethesda, Maryland, USA Abstract We evaluated alterations in information processing after closed head injury as a function of task demands and stimulus modality. Visual and auditory discrimination tasks were administered to 11 survivors of a head injury and 16 matched healthy controls. In auditory tasks, compared with controls, the survivors had smaller N100s, smaller and later N200s, a more posterior scalp distribution of N200, and longer P300 and response latencies. Auditory N200 and P300 correlated highly with duration of unconsciousness. In contrast, in visual tasks, only a reduced N200 in the survivors differentiated the groups. Our results indicate that processing of auditory stimuli, including the perception and discrimination of stimulus features and the evaluation and categorization of stimuli, may be impaired after head trauma. Visual sensory processing may be spared, but higher-order visual processing involved in stimulus classification may be compromised. Descriptors: Closed head injury, Information processing, Event-related potentials, N200, P300, Reaction time The extensive cognitive deficits that follow closed head trauma have been well documented. There is consensus that information processing is slowed, perhaps permanently, following the injury (e.g., Gronwall, 1987; Spikman, Timmerman, van Zomeren, & Deelman, 1999). Deficits may also include impairments in memory and learning (e.g., Watt, Shores, & Kinoshita, 1999), difficulties in the planning, initiation, and execution of activities (e.g., Stuss & Gow, 1992), and decreased motivation (e.g., Prigatano, 1987). Perhaps the most commonly reported problem among head-injury survivors, however, is impaired attention and concentration (e.g., Binder, Rohling, & Larrabee, 1997; Mirsky, Anthony, Duncan, Ahearn, & Kellam, 1991; Spikman, van This research was supported by the Intramural Research Program of the National Institute of Mental Health. The opinions and assertions expressed herein are those of the authors and are not to be construed as reflecting the views of the Uniformed Services University or the U.S. Department of Defense. We thank the Maryland Head Injury Foundation and Edward Turner, MSW, for assistance in recruiting participants; Barry Richmond, MD, and Barbara Pendleton Jones, PhD, for clinical evaluations of headinjury survivors; Lisa Slade, PhD, for screening healthy control participants; John Ingeholm, PhD, for technical assistance; Audrey Weinberg, BS, for contributions to data analysis; and Adrienne Elliott, MA, for assistance with manuscript formatting. We also thank Francie Gabbay, PhD, for her helpful comments on an earlier draft of the manuscript. The present address of Mary Kosmidis is Department of Psychology, Aristotle University of Thessaloniki, Thessaloniki, Greece. Address reprint requests to: Connie C. Duncan, MSC 2615, 5415 W. Cedar Lane, Suite 203B, Bethesda, MD , USA. connie_duncan@nih.gov. 45 Zomeren, & Deelman, 1996; Stratton & Gregory, 1994; Stuss, Stethem, Hugenholtz, & Richard, 1989). Because survivors of a closed head injury may exhibit improvement over time, impaired test performance early in recovery may not be an accurate reflection of the long-term cognitive effects of brain damage (Binder, 1997; Lishman, 1973; Rutherford, Merrett, & McDonald, 1979). Whether due to cerebral reorganization or the development of behavioral compensatory mechanisms, many cognitive deficits resulting from closed head injury diminish over time. In fact, it appears that most of the cognitive recovery occurs within the first year following the injury (Levin, Grossman, Sarwar, & Meyers, 1981; MacFlynn, Montgomery, Fenton, & Rutherford, 1984; Spikman et al., 1999; van Zomeren & Brouwer, 1994). Nevertheless, some deficits may persist and be manifest in the manner and speed in which information is processed. Event-related brain potentials (ERPs), which reflect sensory and cognitive processing, provide a unique and powerful method to investigate alterations in information processing following traumatic brain injury. Early ERP components, such as N100 and P200, reflect primarily sensory processing (Na ätänen, 1990, 1992; Na ätänen & Picton, 1987), whereas later ERP components, including N200 and P300, are associated with cognitive processes, such as stimulus discrimination, evaluation, and categorization (Donchin & Coles, 1988; Duncan-Johnson, 1981; Picton, 1992; Pritchard, Shappell, & Brandt, 1991; Ritter, Simson, Vaughan, & Friedman, 1979). Investigations of the effects of closed head injury on sensory ERP components have yielded somewhat inconsistent results (Clark, O Hanlon, Wright, & Geffen, 1992; Ford & Khalil, 1996;

2 46 C.C. Duncan, M.H. Kosmidis, and A.F. Mirsky Nativ, Lazarus, Nativ, & Joseph, 1994; Rugg et al., 1988, 1993; Solbakk, Reinvang, Nielsen, & Sundet, 1999). With few exceptions (Curry, 1980; Curry, Woods, & Low, 1986; Heinze, Mu nte, Gobiet, Niemann, & Ruff, 1992), such studies have found that neither visual nor auditory N100 latency or amplitude differentiated survivors of a head injury from healthy controls, suggesting that sensory functions may remain intact after trauma. In contrast, the small number of studies that investigated P200 reported increased latency (Cremona-Meteyard & Geffen, 1994; Pratap-Chand, Sinniah, & Salem, 1988) or reduced amplitude (Campbell, Houle, Lorrain, Deacon-Elliott, & Proulx, 1986; Clark et al., 1992; Curry, 1980; Heinze et al., 1992) in headinjury survivors as compared with controls. These investigators speculated that P200 attenuation following a head injury might reflect difficulty with stimulus discrimination (Clark et al., 1992). They based this interpretation on the finding, in healthy individuals, of an inverse relationship between P200 amplitude and degree of difficulty in a tone discrimination task (Lindholm & Koriath, 1985). Studies focusing on later, cognitive components have also yielded inconsistent findings. Investigations using visual tasks have produced conflicting N200 results (Cremona-Meteyard & Geffen, 1994; Curry, 1980; Heinze et al., 1992; Nativ et al., 1994). In a visual directed-cue paradigm, Cremona-Meteyard and Geffen found increased N200 latency but normal amplitude among survivors of moderate to severe head injuries tested more than one year post-trauma. Heinze et al. observed that N200 in a visual search task was reduced in amplitude but normal in latency. More specifically, attenuated fronto-central and posterior N200 was found among head-injured participants relative to controls, despite intact early and later ERP components. They proposed that this was due to an impairment in the ability to discriminate the relevant features of stimuli rather than a more global problem with information processing. Using a visual short-term memory task, Montirosso and coworkers reported a smaller and later N200 in their sample of head-injury survivors (Montirosso, Manfredini, Castelli, Pozzoli, & Reni, 1997). In contrast, Nativ et al. found no alterations in N200 elicited in a visual go/no-go task in survivors of a closed head injury. Investigations using auditory tasks have also yielded mixed results. Although one study using an auditory go/no-go task yielded a group difference in the scalp distribution of the N200 component (Clark et al., 1992), studies from a different laboratory using similar tasks yielded conflicting findings (Rugg et al., 1988, 1993). In the first of two studies, Rugg and his coworkers found that survivors of a severe closed head injury had later though larger N200s than healthy controls. They proposed that head-injured participants took longer to categorize the stimuli due to difficulties in processing their meaning. Rugg et al. attributed their findings of enhanced N200 to the head-injury survivors allocation of additional processing resources to the task. This additional allocation, it was suggested, was an attempt to compensate for their slower information processing. Other investigators have made similar proposals regarding level of cognitive effort or arousal in survivors of closed head injury (Campbell & de Lugt, 1995; Campbell, Suffield, & Deacon, 1990; van Zomeren & Brouwer, 1987). A later study (Rugg et al., 1993) appeared to replicate the findings from their previous study; however, the ERPs elicited by frequent nontarget stimuli were substantially more negative in the head-injured than in the control group. When this difference in the ERPs to the frequent tones was taken into account, the group differences in N200 disappeared. In contrast to the Rugg et al. studies, Solbakk et al. (1999) found smaller N200s and processing negativities in a dichotic listening task in head-injury survivors than in those with frontal lobe lesions or healthy controls. Abnormalities in P300 following closed head injury have been observed somewhat more consistently. Several studies reported that visual P300 was delayed, although not reduced in amplitude, in survivors of a head injury (Cremona-Meteyard & Geffen, 1994; Curry, Cummins, Eames, Rogers, & Chaudhry-Dijkerman, 1996; Heinze et al., 1992; Montirosso et al., 1997; Sangal & Sangal, 1996; Wright, Cremona-Meteyard, Geffen, & Geffen, 1994). In a series of studies, Campbell, Deacon, and coworkers found both longer latency and reduced amplitude P300s in visual and auditory oddball and feedback tasks (Campbell et al., 1986; Deacon & Campbell, 1991a, 1991b; Deacon-Elliott & Campbell, 1987). Similar findings were reported by Unsal and Segalowitz (1995) and Olbrich et al. (1986). Other investigators, using auditory tasks, have reported either only prolonged latencies (Keren, Ben-Dror, Stern, Goldberg, & Groswasser, 1998; Pratap-Chand et al., 1988), only attenuated amplitudes (Ford & Khalil, 1996; Rugg et al., 1988; Solbakk, Reinvang, & Nielsen, 2000; Wirse n, Stenberg, Rose n, & Ingvar, 1992), or an absence of P300 abnormalities (Clark et al., 1992; Rugg et al., 1993; Sangal & Sangal, 1996; Werner & Vanderzant, 1991). ERP investigations of closed head injury have thus yielded conflicting results. The inconsistent effects of head injury on N200 and P300 observed across studies may be due to one or more factors. Among the variables that could play a role are differences in characteristics of the head-injury survivors, such as the severity of injury and/or the time since injury; the type of stimulus and response processing required by the task; the modality and characteristics of the eliciting stimuli; techniques of data quantification; and task difficulty. Task difficulty has been a variable of particular interest in understanding the long-term effects of closed head injury on attention and information processing. Van Zomeren and Deelman (1976) found a disproportionate increase in choice versus simple reaction time in survivors of a closed head injury as compared with healthy controls. Moreover, there was an interaction between task demands and duration of post injury coma. This interaction between task demands and severity of injury on performance was still evident almost two years after concussion. The authors suggested that the relationship between performance on simple and choice reaction time tasks might be more informative than performance on choice reaction time tasks alone. The goal of the present study was to use ERPs to identify alterations in information processing following closed head trauma. Because previous research on head injury has shown an effect of task difficulty on performance, we manipulated the level of task demands in a series of stimulus discrimination tasks. In this manner, we sought to examine the performance of headinjury survivors on more and less demanding tasks in relation to healthy control participants. If the head-injury survivors had developed compensatory mechanisms for their deficits and could perform well on a minimally demanding task, these mechanisms might falter under more challenging conditions. Increasing task demands would thereby unmask impairments in information processing. We administered a battery of stimulus discrimination tasks to all participants. Identical stimuli were used in each task, but response requirements were manipulated to vary task demands.

3 Information processing after head injury 47 We predicted that survivors of a closed head injury would have longer reaction times and prolonged N200 and P300 components and, further, that these differences would be accentuated as task demands increased. Moreover, we used comparable visual and auditory stimulus discrimination tasks to assess the modalityspecificity of group differences. Methods Participants Individuals who had sustained a closed head injury were referred to our laboratory by local neurologists and by the Maryland Head Injury Foundation. All survivors had been unconscious and hospitalized at the time of injury. Based on a review of medical records, those with a previous history of alcohol or drug abuse, a premorbid psychiatric or neurological disorder or brain lesion, or manifest deficits in the upper limbs were excluded. Of the 327 head-injury survivors who were screened initially, 35 were interviewed for participation in this study using the lifetime version of the Schedule for Affective Disorders and Schizophrenia (Spitzer & Endicott, 1975). Of these, 24 were excluded for lack of documentation regarding the nature of the injury (i.e., medical records unavailable or missing), postinjury brain surgery, a history of previous head trauma, or a suicide attempt, yielding a sample of 11 survivors. In every case, the injuries were the result of deceleration impact (10 vehicular incidents, one fall) and were followed by a period of unconsciousness that averaged 19.2 days (SD ). All head-injury survivors were at least two years postinjury at the time of testing (M years, SD 5 3.1); thus, our sample could be considered stable at the time of assessment. Healthy control participants were recruited through newspaper advertisements or from listings for research volunteers at the National Institutes of Health. The control group was selected to match the head-injury group in terms of sex, race, age, educational level, and handedness. Control participants reported no current or past substance abuse or psychiatric disorders (as assessed with the lifetime version of the Schedule for Affective Disorders and Schizophrenia; Spitzer & Endicott, 1975). They also reported no history of psychiatric disorder among their firstdegree relatives. Controls were healthy, free of medications, and had no history of head injury or neurological disorders. A total of 11 survivors of a closed head injury and 16 matched, healthy controls participated in this study. Most participants were white; there was one Asian American in the survivor group and one African American in the control group. Table 1 provides a summary of demographic data for the headinjury and control groups. The two groups did not differ significantly on any of the demographic variables. All participants had normal or corrected-to-normal vision and were screened for hearing loss using a standard audiometric evaluation. All participants gave written informed consent and were paid for their participation. Stimuli and Procedure Two-stimulus visual and auditory discrimination tasks were used to elicit ERPs. Visual stimuli were the letters S and H presented for 100 ms on a CRT display placed cm from the participant s eyes. The letters subtended an angle of 1.01 vertically and 0.81 horizontally. The letters S and H were presented in random sequences with relative probabilities of.10 and.90, respectively. Auditory stimuli were tone bursts, either low (600 Hz) or high (1500 Hz) in pitch. The two tones were equated for duration (100 ms), loudness (50-dB sensation level), and rise-fall time (10 ms), and were delivered binaurally via stereo headphones over broad-band masking noise (60-dB sound pressure level). The low, p 5.10, and high, p 5.90, tones were presented in random order. To manipulate task demands across trial blocks, we varied the response required, while keeping the stimuli identical in the three discrimination tasks. In the simple reaction time task, we instructed participants to press a button with the thumb of the dominant hand at the onset of every stimulus (both the.10 and the.90 stimuli). Because the response was perfectly predictable in this task, we varied the timing of stimulus delivery in order to prevent anticipatory responses. The interval between stimuli varied randomly between 1,200 and 1,800 ms, with a mean of 1,500 ms. In the go/no-go reaction time task, we asked participants to press a button with the thumb of the dominant hand to the.10 stimulus and to withhold responding to the.90 stimulus. In the choice reaction time task, participants were instructed to press a button with one thumb in response to the.10 stimulus and to press a second button with the other thumb in response to the.90 stimulus. The buttons assigned to the two responses were counterbalanced across participants in each group. In the go/no-go and choice tasks, the stimuli were presented at a fixed interstimulus interval of 1,500 ms. We administered two blocks of 150 trials for each task in each modality. Additional blocks were administered if eye movement contamination or incorrect responses reduced the number of acceptable target trials in the first two blocks: Most of the Table 1. Participant Characteristics Healthy Head-injured (n 5 16) (n 5 11) M SD M SD Age (years) Education (years) Sex 8 female, 8 male 5 female, 6 male Handedness 15 right, 1 left 10 right, 1 left Unconsciousness interval (days) Time since injury (years)

4 48 C.C. Duncan, M.H. Kosmidis, and A.F. Mirsky head-injury survivors (8 of 11) required more than two blocks for at least one of the tasks, whereas only 6 of the 16 controls did. Each trial block was followed by a short rest period. The order of presentation of blocks of stimuli in the two modalities was counterbalanced across participants in each group, but the order of presentation of the tasks within each modality was the same for everyone (simple, go/no-go, and choice reaction time tasks). Testing took place in a sound-attenuated, dimly lit room, with the participant seated in a comfortable chair. We gave standard instructions for each task, administered practice trials to ensure that participants understood these instructions, and gave them an opportunity to ask questions. In each task, we instructed participants to respond as quickly and as accurately as possible. Data Acquisition The electroencephalogram was recorded with Ag/AgCl electrodes, affixed with collodion, from frontal (F3, Fz, F4), central (C3, Cz, C4), parietal (P3, Pz, P4), and temporal (T3, T4) scalp sites according to the International system, all referred to linked earlobes. We recorded the electrooculogram (EOG) between supraorbital and outer canthal positions of the left eye. A ground electrode was placed at Fp2. The EEG and EOG were amplified with a bandpass of 0.01 to 100 Hz ( 3 db points; Duncan-Johnson & Donchin, 1979) and digitized at 200 Hz for a period of 1,100 ms, beginning 150 ms prior to stimulus onset. Impedance of the EEG electrodes did not exceed 3 ko. Data Quantification Prior to averaging, we discarded trials with incorrect responses or eye movement contamination. A button press within 100 to 950 ms (simple reaction time task) or 175 to 950 ms (go/no-go and choice reaction time tasks) of the onset of each target stimulus was considered a correct response, whereas a button press at any other time was considered an error. We quantified two measures of performance on each task: accuracy and reaction time to the.10 stimuli. Reaction time was defined as the time from stimulus onset to the button press; accuracy was the percentage of correct button presses. We assessed eye movement artifact in the following manner: For each trial, we calculated, for the EOG, the absolute value of the difference between each point in the epoch and the mean of the 150-ms prestimulus baseline. If a specified number of these difference scores exceeded a preselected criterion value, the trial was rejected. The criterion value, determined in previous work (Duncan-Johnson & Donchin, 1977, 1982), led consistently to the rejection of trials with eye movement artifact. The EOG of the accepted trials was averaged to verify the success of the algorithm for each average ERP. For each participant, average ERPs were computed at all electrode sites, for each combination of modality, task, and probability. Prior to quantification, average ERPs were smoothed by a low-pass, zero-phase-shift digital filter with a cutoff frequency ( 3 db attenuation) of 12.4 Hz. Measures of ERP components used in the statistical analyses were derived from each participant s ERPs elicited by the.10 stimuli in each task, using standard baseline-to-peak measurements (Picton et al., 2000). The latencies of the components were defined as the time from stimulus onset to the point of peak amplitude. Peak selection was done using a computer algorithm, and the selected peak latencies were checked visually to verify their accuracy. Amplitudes were calculated by subtracting the average voltage in the 150-ms prestimulus baseline from the peak voltage in specified latency ranges. Visual P200 and N200 and auditory N100, P200, and N200 were quantified at Cz, where they were maximal in amplitude, whereas visual N100 was quantified at Pz. The latency windows for the visual N100, N200, and P300 components were , , and ms, respectively. The latency windows for the auditory N100, N200, and P300 components were , , and ms, respectively. P200 was defined as the maximum positive peak between N100 and N200. Because the amplitude of N200 can be reduced by overlap with P200, we quantified N200 in the.10 minus.90 difference waveforms at the three midline electrode sites. P300 amplitude was measured at all of the electrode sites, to allow an assessment of group differences in scalp topography. We selected P300 data at Cz for the primary analyses because P300 was large in amplitude at this location in all individuals, and had less overlap with other ERP components, especially slow wave, at Cz than at other electrode sites (Duncan et al., 1994). Mean slow wave activity was quantified at the midline electrode sites (Fz, Cz, and Pz) by summing the voltages from 555 to 750 ms for visual ERPs and from 505 to 700 ms for auditory ERPs and dividing by the number of points sampled (i.e., 40). Data Analysis A mixed design repeated-measures analysis of variance was used to evaluate the effects of group (head-injury survivors, healthy controls) and task (simple, go/no-go, and choice reaction time) on ERP and performance data. Because of evidence of modalityspecific P300 activity (Barrett, Neshige, & Shibasaki, 1987; Duncan et al., 1990; Duncan, Morihisa, Fawcett, & Kirch, 1987; Egan et al., 1994; Johnson, 1989a, 1989b, 1993; Johnson, Miltner, & Braun, 1991), we conducted separate analyses of the visual and auditory data. For this reason, it was not necessary to exclude from the analyses participants for whom we had complete data for only one modality. The Greenhouse Geisser epsilon correction was used to evaluate F ratios for repeated measures involving more than one degree of freedom. Uncorrected degrees of freedom are reported with the epsilon-corrected p values. Level of significance was set at.05 for all statistical analyses. Group differences in the topographic distribution of ERP components were also evaluated with repeated-measures analysis of variance for mixed designs. For N200 and slow wave, there was one between-subjects variable, group, and two withinsubject variables, task and midline anterior-posterior axis (Fz, Cz, Pz). The analysis of the scalp topography of P300 comprised one between-subjects variable, group, and the within-subject variables of task, hemisphere (left, right), and anterior-posterior axis (frontal, central, parietal). Greenhouse Geisser corrections were applied where appropriate. To evaluate group differences in scalp topography independently of overall amplitude differences between groups, interactions involving group and hemisphere and/or anterior-posterior axis were reevaluated after the data were averaged over task and normalized (McCarthy & Wood, 1985). When an interaction was statistically significant, followup analyses of variance were used to clarify the interaction. Bonferroni tests were used for post hoc contrasts. Pearson product-moment correlation coefficients (two-tailed tests) were used to assess the relationship between the severity and recency of injury and measures of ERPs and performance. We considered the duration of unconsciousness as an indication of severity and time since injury as the measure of recency.

5 Information processing after head injury 49 Results Performance Data Task effects. Figure 1 depicts mean reaction time to the visual, p 5.10, and auditory, p 5.10, stimuli in the three tasks, averaged over the entire sample of participants. Reaction time increased as task demands increased, for both the visual, F(2, 44) , po.0001, e 5.94, and the auditory, F(2,46) , po.0001, e 5.86, tasks. Post hoc comparisons indicated that in both modalities, reaction time was shorter in the simple reaction time task than in the go/no-go and choice tasks, which did not differ from each other. In Figure 2, mean accuracy is plotted as a function of task, averaged over the pooled sample of participants. There was a significant effect of task demands on the accuracy of performance on tasks in both modalities (visual: F(2,44) , po.0001, e 5.55; auditory: F(2,46) , po.003, e 5.74). Post hoc comparisons confirmed that in both stimulus modalities, accuracy was lower in the choice reaction time task as compared with the other two tasks. Accuracy did not differ between the simple and go/no-go tasks. The pattern of results for the two measures of performance suggests that the manipulation of task demands in each modality was successful. Thus, task demands were minimal in the simple reaction time task (in which responses were the fastest and more accurate than in the choice task), intermediate in the go/no-go task (in which responses were slower than in the simple reaction time task but equally accurate), and maximal in the choice reaction time task (in which responses were slower than in the simple task and least accurate). These effects were the same for both the head-injury survivors and the healthy controls. Figure 1. Mean reaction times to visual (left) and auditory (right) stimuli presented with a probability of.10 in the simple, go/no-go, and choice reaction time tasks, averaged over the entire sample of participants. Note that reaction time increased as task demands increased. healthy control participants for each modality, task, probability, and electrode location. Visual ERPs Figure 3 displays the grand-mean ERPs elicited by the.10 stimulus in the three visual discrimination tasks. The waveforms for the two groups are superimposed for each of the scalp electrode sites and EOG. In each task, the ERPs to the.10 stimuli are characterized by small negative (N100) and positive (P200) components, peaking at an average of 140 and 205 ms, Group effects. Mean performance data for the head-injury and control groups are presented in Table 2. There was only one significant difference between groups in task performance: The head-injury survivors had longer reaction times than the healthy controls on all tasks with auditory stimuli, F(1,23) , po.025. This difference between groups failed to reach significance for tasks with visual stimuli, F(1,22) , po.07. There were no significant group differences in the accuracy of performance, nor any Group Task interactions, on either performance measure in either modality. ERP Data We computed grand-mean ERPs for each group by averaging the data acquired from the 11 closed head-injury survivors and 16 Figure 2. Mean accuracy of performance to visual (left) and auditory (right) stimuli presented with a probability of.10 in the simple, go/no-go, and choice reaction time tasks, averaged over all participants. Accuracy of performance was reduced in the most demanding, choice reaction time, task in both modalities. Table 2. Mean Reaction Time and Accuracy for Each Modality, Group, and Task Visual Auditory Healthy Head-injured Healthy Head-injured Task (n 5 14) (n 5 10) (n 5 16) (n 5 9) Reaction time (ms) Simple 245 (52) 255 (68) 277 (84) 358 (84) Go/no-go 401 (66) 476 (109) 374 (96) 436 (109) Choice 435 (63) 498 (96) 378 (49) 457 (108) Accuracy (%) Simple 99 (2) 97 (6) 96 (9) 95 (7) Go/no-go 100 (1) 98 (3) 98 (4) 97 (6) Choice 80 (11) 73 (20) 89 (16) 82 (22) Note: Values are means (SD).

6 50 C.C. Duncan, M.H. Kosmidis, and A.F. Mirsky Figure 3. Grand-mean ERP waveforms elicited by the p 5.10 visual stimulus in the simple, go/no-go, and choice reaction time tasks. The ERPs for the head-injury survivors (dashed lines) and the healthy controls (solid lines) are superimposed for each of the 11 electrode sites and EOG. An 1,100-ms epoch is shown for each waveform. An S on the time scale indicates stimulus onset. Positivity of the scalp electrode with respect to the reference electrodes is shown as a downward deflection in this and subsequent figures. ERP components are identified on the Cz waveform in the choice task. respectively, followed by larger N200 (280 ms) and P300 (425 ms) components. Slow wave activity overlapped the P300 and was positive at all scalp sites (Duncan-Johnson & Donchin, 1982). Task effects. Inspection of Figure 3 shows that there was a dramatic increase in visual P300 from the simple to the choice reaction time task. This observation was confirmed by statistical analysis: P300 amplitude increased monotonically with increasing task demands, F(2,44) , po.0001, e Figure 4 depicts visual P300 amplitude as a function of task, averaged over all participants. Post hoc tests confirmed that visual P300 was larger in the choice task than in the go/no-go task; P300 in both tasks was larger than that elicited in the simple reaction time task. The latency of visual P300 did not vary systematically with task demands. The amplitude of the visual slow wave component also varied with task demands, F(2,44) , po.001, e Post hoc tests indicated that slow wave was larger in the choice and go/no-go tasks than in the simple task but did not differ between the choice and go/no-go tasks. No other components of the visual ERP varied systematically with task demands (Tables 3 and 4). Group effects. Figure 5 presents the visual grand-mean difference waveforms obtained by subtracting the ERPs to the.90 stimulus from the ERPs to the.10 stimulus, separately for each task. Because the N100 and N200 components elicited in

7 P300 Amplitude (uv) Information processing after head injury Visual auditory stimuli elicited N100 (M ms), P200 (M ms), and N200 components (M ms), followed by P300, peaking at an average of 355 ms. P300 was overlapped by slow wave activity, which was positive at central and parietal sites, and negative at frontal sites (Duncan-Johnson & Donchin, 1977; Squires, Squires, & Hillyard, 1975) Simple Go/No-Go Choice Reaction Time Task Figure 4. Mean amplitude of visual P300, elicited by the p 5.10 stimulus and recorded at Cz, in the simple, go/no-go, and choice reaction time tasks, averaged over the entire sample of participants. Note that visual P300 increased as task demands increased. the simple reaction time task were too small to quantify reliably (see Figures 3 and 5), they were excluded from the analyses. As shown in Figure 5, visual N200 was reduced in the head-injury survivors as compared with the controls, F(1,22) , po.03. That is, in the more demanding, go/no-go and choice tasks, in which N200 was elicited, it was markedly attenuated in amplitude in the head-injury survivors. Although there is an apparent delay in the latency of visual P300 in the survivor group (cf. Figure 3), this delay was not statistically significant, po.09. There were no significant main effects or interactions involving group on visual N200 latency or on N100, P200, or P300 (Tables 3 and 4). Auditory ERPs The grand-mean ERPs elicited by the.10 stimulus in the three auditory discrimination tasks are displayed in Figure 6. The.10 Task effects. Task demands had a significant effect on the amplitude of auditory N200, F(2,46) , po.004, e Post hoc tests indicated that N200 was larger in amplitude in the choice and go/no-go tasks than in the simple reaction time task. N200 in the choice and go/no-go tasks did not differ. As shown in Figure 7, the effect of task on P300 was similar to the effect on N200, with P300 amplitude larger in the choice task than in the go/no-go or simple reaction time tasks, F(2,44) , po.0001, e However, in contrast to N200, the latter two tasks did not differ in terms of P300 amplitude. There was a trend for auditory slow wave to vary as a function of task demands, F(2,46) , po.07, e There were no systematic task effects on the amplitude or latency of other auditory ERP components (Tables 3 and 4). Group effects. The auditory grand-mean (.10.90) difference waveforms are presented in Figure 8. The apparent group differences in the latency and amplitude of N200 were confirmed by the analyses: N200 was later and smaller in the head-injury group than in the control group, F(1,23) , po.002, and F(1,23) , po.05, for latency and amplitude, respectively. Moreover, N100 was significantly reduced in amplitude in the head-injury group, F(1,22) , po.05. The latency of auditory P300 was longer in the head-injury group as compared with the control group, F(1,23) , po.01. Although it appears in Figure 6 that the head-injury survivors had smaller P300s than did the healthy controls, this group difference did not reach statistical significance, F(1,22) , po.15. There were no group effects on the latency of auditory N100 or on P200 (Tables 3 and 4). Table 3. Mean Amplitude of ERP Components for Each Modality, Group, and Task Visual Auditory Healthy Head-injured Healthy Head-injured ERP component (n 5 14) (n 5 10) (n 5 16) (n 5 9) Simple task N100 F F 9.7 (3.6) 8.7 (3.7) P (4.8) 6.0 (2.9) 1.3 (5.9) 3.1 (3.9) N200 a F 6.7 (6.4) 1.0 (4.9) P (6.4) 11.8 (6.3) 13.6 (7.2) 7.9 (6.5) Go/no-go task N (3.2) 2.1 (2.9) 12.2 (3.6) 8.0 (2.9) P (5.2) 6.4 (5.0) 1.2 (6.1) 3.1 (3.3) N200 a 6.0 (6.6) 1.2 (3.1) 10.1 (9.9) 4.8 (3.6) P (6.4) 14.6 (10.3) 14.1 (9.3) 10.1 (8.6) Choice task N (4.5) 1.7 (2.9) 11.2 (3.9) 8.8 (3.7) P (5.4) 7.5 (3.7) 1.9 (4.7) 4.2 (4.9) N200 a 7.8 (7.5) 2.7 (4.6) 10.4 (7.2) 5.4 (4.1) P (9.7) 18.8 (8.2) 21.4 (8.3) 17.0 (9.1) Note: All components were quantified at Cz except visual N100, which was quantified at Pz. Values (in microvolts) are means (SD). Dashes indicate that the component was too small to be quantified reliably. a N200 data are based on (.10.90) difference waveforms.

8 52 C.C. Duncan, M.H. Kosmidis, and A.F. Mirsky Table 4. Mean Latency of ERP Components for Each Modality, Group, and Task Visual Auditory Healthy Head-injured Healthy Head-injured ERP component (n 5 14) (n 5 10) (n 5 16) (n 5 9) Simple task N100 F F 100 (12) 107 (9) P (26) 206 (34) 177 (31) 181 (16) N200 a F F 208 (21) 234 (35) P (48) 421 (79) 358 (47) 379 (47) Go/no-go task N (31) 141 (33) 103 (9) 106 (14) P (33) 213 (23) 167 (9) 176 (17) N200 a 286 (34) 284 (33) 208 (19) 227 (19) P (47) 457 (60) 334 (35) 359 (46) Choice task N (36) 134 (37) 101 (8) 105 (14) P (27) 205 (35) 158 (15) 177 (21) N200 a 287 (25) 266 (31) 196 (17) 229 (32) P (36) 438 (64) 324 (25) 388 (68) Note: All components were quantified at Cz except visual N100, which was quantified at Pz. Values (in milliseconds) are means (SD). Dashes indicate that the component was too small to be quantified reliably. a N200 data are based on (.10.90) difference waveforms. P300 att3 and T4 Because head trauma often compromises the temporal lobes (Holbourn, 1943; Ommaya & Gennarelli, 1974), we also quantified visual and auditory P300 at T3 and T4. (These waveforms are displayed in Figures 3 and 6). There was no main effect of group on the amplitude of visual P300 at these electrode sites, but there was a significant Group Hemisphere interaction, F(1,22) , po.03. Post hoc tests revealed a T34T4 difference Figure 5. Grand-mean difference waveforms for the visual simple, go/no-go, and choice reaction time tasks derived by subtracting the.90 ERPs from the.10 ERPs. The difference ERPs for the head-injury survivors (dashed lines) and the healthy controls (solid lines) are superimposed for the three midline electrode sites and EOG.

9 Information processing after head injury 53 Figure 6. Grand-mean ERP waveforms elicited by the p 5.10 auditory stimuli in the simple, go/no-go, and choice reaction time tasks. The ERPs for the head-injury survivors (dashed lines) and the healthy controls (solid lines) are superimposed for each of the 11 electrode sites and EOG. ERP components are identified on the Cz waveform in the go/no-go task. in the head-injury group; an apparent T44T3 difference in the control group was not statistically significant. There were no group differences in visual P300 latency at T3 and T4. Auditory P300 recorded at the temporal sites showed a different pattern, namely, no difference in amplitude but a significant group difference in latency: P300 latency at T3 and T4 was longer in the head-injury group as compared with the control group, F(1,23) , po.005. Scalp Topographies Visual ERP components. We evaluated group differences in the distribution of N200 along the midline anterior posterior axis. Analysis of variance of normalized visual N200 revealed no group differences in scalp topography: In both groups, visual N200 had a Pz 5 Cz4Fz distribution, F(2,44) , po.0001, e We also evaluated group differences in the scalp topography of P300. Analysis of visual P300 revealed a significant Group Hemisphere Anterior-posterior axis interaction, F(2,44) , po.02, e Follow-up analyses of variance were used to evaluate the scalp distribution of P300 within each group: The Hemisphere Anterior-posterior interaction was highly significant for the controls, F(2,26) , po.0001, e 5.99, but not the survivors, F(2,18) , po.07, e Post hoc tests in the controls revealed that the interaction was due to a right4left asymmetry of P300 at the frontal, p 5.05, electrodes, which reversed (left4right) at the parietal, po.002, electrodes. Both groups showed the same anterior-posterior gradient of visual P300 (parietal 5 central4frontal), F(2,44) , po.0001, e There was no overall group difference in the amplitude of visual slow wave. There was, however, a Group Anterior-

10 54 C.C. Duncan, M.H. Kosmidis, and A.F. Mirsky (Pz 5 Cz4Fz), slow wave did not differ significantly between groups at any of the midline electrode locations. Figure 7. Mean amplitude of auditory P300, elicited by the p 5.10 stimulus and recorded at Cz, in the simple, go/no-go, and choice reaction time tasks, averaged over the entire sample of participants. Note that auditory P300 was enhanced only in the most demanding, choice reaction time, task. posterior axis interaction, F(2,44) , po.01, e Although the distribution of visual slow wave amplitude along the midline appeared to be more posterior in the head-injury survivors (Pz4Cz4Fz) as compared with the controls Auditory ERP components. In contrast to visual N200, there was a significant interaction of Group Anterior-posterior axis on normalized auditory N200 amplitude, F(2,46) , po.015, e Follow-up analyses within each group indicated that in the head-injury survivors, auditory N200 had a Cz 5 Pz4Fz distribution; whereas in the control participants, it had a Cz4Pz 5 Fz distribution. That is, auditory N200 had a central distribution in the control group but a more posterior, centroparietal distribution in the head-injury group. Post hoc tests on each electrode site revealed that the apparent sources of the interaction were the Fz and Cz electrode sites, where N200 was significantly reduced in the head-injury group. Analysis of auditory P300 revealed no interactions of group with the hemisphere or anterior-posterior axis variables. Thus, unlike visual P300, both groups had the same scalp distribution of auditory P300. There were significant main effects of hemisphere (right4left), F(1,22) , po.05, and anteriorposterior axis (parietal4central4frontal), F(2,44) , po.0001, e 5.63, and a significant Hemisphere Anteriorposterior interaction, F(2,44) , po.0002, e Post hoc comparisons indicated that the interaction was due to a right4left asymmetry of auditory P300 at frontal and central sites, which reversed (left4right) at the parietal sites. The parietal asymmetry, however, did not reach significance by post hoc test, po.09. Figure 8. Grand-mean difference waveforms for the auditory simple, go/no-go, and choice reaction time tasks derived by subtracting the.90 ERPs from the.10 ERPs. The difference ERPs for the head-injury survivors (dashed lines) and the healthy controls (solid lines) are superimposed for the midline electrode sites and EOG.

11 Information processing after head injury 55 The groups did not differ in terms of the scalp topography of auditory slow wave, which showed the same Pz4Cz4Fz distribution in both groups, F(2,46) , po.0001, e Relationship of Clinical Variables to ERP and Performance Measures We used Pearson correlations to assess the relationship between clinical variables (severity and recency of the injury) and N200, P300, and measures of performance in the head-injury survivors. To reduce the number of correlations, we pooled data from the choice and go/no-go tasks in each modality and excluded data from the simple reaction time task. The severity and recency of the injury were not significantly correlated. Visual Tasks On tasks with visual stimuli, there were no significant correlations between either the severity or recency of injury and measures of ERPs or performance. Auditory Tasks Performance data. Neither of the clinical measures was significantly correlated with performance on tasks with auditory stimuli. N200. The severity of injury was significantly correlated with the latency and the amplitude of N200 in the auditory tasks, r(9) 5.80, po.01, and r(9) 5.80, po.01. The longer the duration of unconsciousness, the longer the latency of N200 and the smaller (less negative) its amplitude. Recency of injury was not significantly correlated with N200. P300. Severity of injury showed a significant positive correlation with the latency of auditory P300, r(9) 5.86, po.003, with longer periods of unconsciousness associated with later P300s. Discussion Our results indicate that sensory processing of visual stimuli may be spared after head trauma, but aspects of higher-order processing are compromised. Consistent with previous studies, we observed no effects of closed head injury on the visual N100 or P200 components, suggesting that sensory processing was intact in the visual modality. We did, however, find that the amplitude of visual N200 was reduced in the head-injury survivors, suggesting that they allocated fewer cognitive resources than controls to the discrimination of visual stimuli. The absence of group differences in performance on visual tasks suggests that the attenuation of visual N200 was not due to reduced motivation or effort in the survivors. The only other aspect of visual processing that differentiated the groups was a more symmetrically distributed P300 at the frontal electrodes in the survivors than the controls. In contrast, auditory information processing was substantially altered by a closed head injury. In the ERPs elicited by auditory stimuli, both N100 and N200 were smaller, and N200 and P300 were later, in the head-injury survivors than in the controls. It is noteworthy that auditory reaction time was longer in the head-injury group despite accuracy rates that were comparable to those of the control group. Auditory slow wave did not distinguish the groups. Our findings are consistent with those of other studies (e.g., Campbell et al., 1986; Pratap-Chand et al., 1988; Unsal & Segalowitz, 1995): A major consequence of head trauma appears to be slowing of higher-order auditory processes, presumed to be the result of injury at multiple loci within the information processing system (Campbell & de Lugt, 1995; Rugg et al., 1993). The more pronounced effects seen in auditory, as compared with visual, tasks suggest that the brain system supporting processing of auditory stimuli is more vulnerable to trauma than that supporting visual processing. The reduction in auditory N100 could reflect damage to cerebral generators in auditory cortex or auditory association areas. It is also conceivable that the reduced N100 is an attentional effect, resulting from compromise of a thalamocortical gating mechanism (Na ätänen & Picton, 1987). The longer latency and smaller amplitude of the auditory N200 component also have implications for the information processing capacities of head-injury survivors. Previous studies have described N200 as a reflection of the controlled, taskrelevant processing that occurs during voluntary attention (Na a tänen & Picton, 1986; Pritchard et al., 1991). More specifically, it is thought to reflect detection and discrimination of stimulus features (Heinze et al., 1992; Novak, Ritter, Vaughan, & Wiznitzer, 1990; Ritter et al., 1979) and their categorization (Rugg et al., 1988, 1993). N200 thus appears to depend on retaining and comparing memory traces of stimuli, a process that leads to categorization of the current stimulus and selection of the appropriate response. Our N200 findings are consistent with the notion that the head-injury survivors allocated fewer cognitive resources than controls to processes involved in stimulus discrimination. Previous findings regarding N200 in survivors of head trauma have been inconsistent. The discrepant findings across studies may be due to differences in tasks and sample characteristics, as well as to variations in methods of data quantification. In contrast to our findings of attenuated visual and auditory N200, some investigators have reported enhanced N200s relative to those of healthy controls in the presence of comparable reaction times (Rugg et al., 1988). These investigators, however, did not replicate their own finding of enhanced N200 amplitude in survivors of a closed head injury (Rugg et al., 1993). Clark et al. (1992) reported that closed head injury affects the scalp distribution of auditory N200, with an equipotential distribution along the anterior-posterior axis for head-injury survivors versus a frontal distribution for healthy controls. We had a similar finding: Auditory N200 was centro-parietally distributed in the head-injury group but centrally distributed in the control group. The more posterior distribution of N200 in the group of head-injury survivors we studied may have resulted from injury to anterior cerebral structures: Almost all of our head-injury survivors had documented frontal damage, characteristic of deceleration trauma. The topography of N200 observed in the survivors may thus reflect the action of compensatory neural mechanisms used to maximize task performance. Group differences in scalp topography indicate that the combination of brain sources underlying the component must also differ between groups (Ruchkin, Canoune, Johnson, & Ritter, 1995). In our survivors, head injury appears to have affected the neural generators of N200, leading to altered processing in the discrimination of auditory stimuli. Based on the number of significant group differences in ERP and performance measures, it appears that auditory information processing in the survivors was more impaired than visual processing. The latency of N200 was significantly longer in the head-injury survivors than in the controls, whereas auditory

12 56 C.C. Duncan, M.H. Kosmidis, and A.F. Mirsky N100 and P200 were not prolonged. The processing delay thus appears to begin during the discrimination of stimulus features. The group difference in latency increased from an average of 20 ms at N200 to 37 ms at P300; the group difference in reaction time was 74 ms. This increasing difference suggests that the later P300s and slower responses observed in head-injury survivors were not due solely to receiving a delayed input, but also to taking longer to discriminate and categorize the stimuli, and to select and execute a response. The ERP data therefore indicate that the longer reaction times on auditory tasks were due to a general slowing of information processing that begins after sensory input and registration. However, because the delay in reaction time was twice as long as the delay in P300, prolonged stimulus evaluation and categorization cannot account fully for the slower responses of the survivors of a head injury (Campbell et al., 1986; Deacon & Campbell, 1991b; Reinvang, 1998, 1999). Deacon and Campbell used feedback to show that reaction time can be decreased in head-injury survivors, thereby suggesting that their slow responses are not due to impaired motor performance. It has been proposed that response selection in those who have sustained a head injury is characterized by uncertainty, a checking and rechecking of the identity of a stimulus before responding (Campbell & de Lugt, 1995; van Zomeren, 1981; van Zomeren & Deelman, 1976). Thus, the response strategy adopted by head-injury survivors, which emphasizes accuracy of performance at the cost of speed, appears to contribute to their long reaction times. The latency, but not the amplitude, of P300 distinguished the groups, suggesting that slowness, rather than impaired attention, characterizes survivors of a closed head injury (Spikman et al., 1996). This conclusion must, however, be tempered with the observation that the amplitude of auditory and visual P300 appeared to be attenuated in the head-injury survivors. Moreover, ERP analyses of other information processing tasks (i.e., the Continuous Performance Test) administered to our sample yielded significant reductions in P300 amplitude (Kosmidis, Duncan, & Mirsky, 1993). The data from the two experiments suggest that tasks that tap different aspects of information processing may yield different patterns of results in survivors of a head injury (e.g., Cooper, Curry, & Cummins, 1984; Deacon & Campbell, 1991b; Heinze et al., 1992; Mu nte & Heinze, 1994; Solbakk et al., 1999). These findings thus underscore the importance of using multiple tasks and stimulus modalities to characterize information processing deficits in head-injury survivors. The healthy controls exhibited a right-greater-than-left asymmetry in P300 at frontal sites in the visual and auditory discrimination tasks. Although P300 in the head-injury survivors showed the same right-greater-than-left asymmetry in the auditory tasks, it was distributed symmetrically over frontal electrode locations in the visual discrimination tasks. That is, visual P300 was reduced in amplitude over right frontal areas in the survivors as compared with the controls. It is not evident why such a difference in brain organization would be reflected in visual but not auditory P300; however, there is ample evidence that there are independent generators of visual and auditory P300 (e.g., Duncan et al., 1987, 1990, 1994; Johnson, 1989a, 1989b, 1993). Closed head injury may, therefore, have differential effects on visual and auditory processing. Hemispheric asymmetries of P300 are thought to depend on task demands that require specialized hemispheric processing resources (Friedman & Polson, 1981). Amplitude may be enhanced over the hemisphere that has major responsibility for attending to and discriminating the stimuli. Previous studies in healthy participants have reported right-greater-than-left asymmetries of P300 in visual (Alexander et al., 1995) and auditory (Alexander et al., 1996; Duncan et al., 1994; Kayser, Tenke, & Bruder, 1998) discrimination tasks. The sources of these asymmetries remain to be elucidated, but are thought to reflect, in part, asymmetric cognitive processes governing allocation of attentional resources to incoming stimuli. A possible interpretation of the asymmetry difference in the head-injury survivors is that in the healthy individual, more attentional resources are recruited in the right than the left hemisphere during the processing of the visual stimuli; and this difference is reflected in the right-greater-than-left asymmetry of visual P300. According to this formulation, the head-injury survivors cannot allocate resources in this way because of the postulated damage to right hemisphere structures supporting visual P300. The absence of the expected asymmetry at frontal sites is a manifestation of this damage. It should be noted, however, that asymmetries in the hemispheric distribution of P300 do not necessarily reflect asymmetries in the allocation of processing resources (Rugg, Kok, Barrett, & Fischler, 1986). A primary goal of the current investigation was to reveal subtle deficits in information processing in head-injury survivors. For this purpose, we chose to administer a battery of tasks in which the performance requirements were increasingly demanding. We expected that compensatory mechanisms would be more likely to fail under conditions of increased challenge. The increasing reaction times and decreasing accuracy rates with increasing task demands provide evidence that the manipulation of task demands was successful in both modalities. The cognitive components of the ERP were also affected by manipulations of task demands. For the sample as a whole, both visual and auditory P300, as well as auditory N200 and visual slow wave, increased in amplitude with increases in task demands, suggesting that the tasks invoked processing resources proportional to the level of difficulty. The latencies of these components, however, were stable across tasks. That is, it did not appear that participants required more time to identify, evaluate, and categorize the stimuli in the more demanding tasks. We cannot exclude the possibility, however, that any latency differences were too small to detect. However, despite the successful manipulation of task demands, differences between groups in measures of ERPs or performance did not increase with increasing difficulty. It may be that the processing load of our tasks varied over too narrow a range, and that further increases would have elicited greater variations in ERPs and performance between groups. Alternatively, it is likely that our tasks tapped only some of the deficits in information processing that characterize head-injury survivors. In our sample of survivors, recency of injury was not correlated with indices of ERPs or performance. The lack of correlation with recency may reflect the fact that a minimum of two years had elapsed since the injury, and all survivors may have reached a plateau of recovery of function. Correlations with recency of injury may be present only in the immediate postinjury epoch. In contrast, severity of injury was highly correlated with cognitive components of auditory, although not visual, ERPs. Specifically, longer periods of unconsciousness were associated with later and smaller N200s and later P300s. The significant