You are on page 1of 17

Neuropsychological and Cognitive

Psychophysiological Substrates of Impulsive


Aggression
Ernest S. Barratt, Matthew S. Stanford, Thomas A. Kent, and Alan Felthous

The purpose of this study was to test whether subjects who commit impulsive vs non-impulsive
aggression differ on measurements of personali~, neuropsychology, and cognitive psycho-
physiology, and whether these differences can yield information regarding the etiology of
impulsive aggression. Subjects were two groups of prison inmates, distinguished by their
committal of impulsive or nonimpulsive aggression, and matched noninmate controls. All
inmates met DSM III-R criteria for an antisocial personality disorder but for no other
disorder. Impulsiveness, anger, and peak P300 latencies did not differ between the inmate
groups, but verbal symbol decoding and peak P300 amplitudes did. Impulsiveness and verbal
skills were inversely correlated. Impulsiveness was inversely correlated with, and verbal skills
positively correlated with P300 amplitudes. The results indicate that aggression is not
homogenous, even among antisocial persons, and that impulsive aggression is related to
neuropsychological and cognitive psychophysiological measures of information processing
beyond those factors related to criminality alone. © 1997 Society of Biological Psychiatry

Key Words: Impulsive Aggression, P300 ERPs, verbal deficits, prisoners, antisocial person-
ality disorder

BIOL PSYCHIATRY 1997;41 : 1045-1061

Introduction being contrasted as impulsive vs non-impulsive. Although


the labels for these two types of acts have varied among
Defining human aggression poses clinical and research authors (Berkowitz 1974; Heilbrun et al 1976; Linnoila et
nosological problems (Lion 1991; Bech and Mak 1995). al 1983; Coccaro et al 1989), construct and predictive
There is no universally accepted nosology of human validity studies (Heilbrun 1972; Heilbrun et al 1978;
aggression but there is a long history of aggressive acts Shoham et al 1989; Virkkunen et al 1989; DeJong et al
1992; Atkins et al 1993) support the distinction between
From the Department of Psychiatry and Behavioral Sciences (ESB. TAK. AFt. the
Department of Neurology (TAK). the Department of Pharmacology and
these two types of aggression. The general purpose of this
Toxicology (TAK), The University of Texas Medical Branch, Galveston. study is to test hypotheses related to information process-
Texas; and the Department of Psychology, University of New Orleans, New
Orleans, Louisiana (MSS).
ing differences between persons who commit impulsive
Address reprint requests to Ernest S. Barratt, PhD. Department of Psychiatry and and nonimpulsive aggressive acts.
Behavioral Sciences, University of Texas Medical Branch. 301 University Impulsive aggression is defined in this study within
Boulevard. Galveston, TX 77555-0443.
Received June 29. 1994; revised March 3. 1996. B arratt's (1991) nosology which includes three classes of

© 1997 Society of Biological Psychiatry 0006-3223/97/$17.00


Pll S0006-3223(96)00175-8
1046 BIOLPSYCHIATRY E.S. Barratt et al
1997;41:1045 1061

aggressive acts: 1) premeditated; consciously executed or been expressed. Lewis (1992) emphasized the need to look
planned aggressive acts; these acts are best described at multiple measurements of the concomitants of violence
within a psychosocial model (Bandura 1973; Tedeschi and noting that "...when impulsivity, hypervigilance, and
Felson 1994); 2) medically related: aggression as a symp- cognitive expressive deficits coexist, the psychophysiolog-
tom of a medical disorder as typified by its main use in ical stage is set for violence to occur" (p 388). Volavka et
DSM-IV (American Psychiatric Association 1994); 3) al (1992) reviewed psychobiological studies of aggression
impulsive: a hair trigger response to a stimulus that results and concluded "...that substantial progress in this field
in an agitated state and culminates in an aggressive act will be impossible without integrative multidisciplinary
during the agitated state; interpersonal communication is studies" (p 247).
often nonadaptive during the agitated state and informa-
tion processing appears to be inefficient (Elliot 1976, Personali~ Traits and Aggression
1992).
The need for a better understanding of human aggres- Self-report personality inventories have been used as both
sion including impulsive aggression has been noted (Eich- predictor (Barratt 1991; Novaco 1994) and criterion (Bech
elman 1992; Stein, 1994). Barratt et al (1995b) using and Mak 1995) measures of aggression. Further, a wide
Robins and Guze's (1970) criteria for a psychiatric syn- range of self-report personality questionnaires have been
drome, argued that impulsive aggression should be in- developed to measure aggression per se (Cloninger 1987;
cluded as a disorder in future Diagnostic and Statistical Buss and Durkee 1957). Two of the more commonly
Manuals of the American Psychiatric Association along studied personality traits in aggression research are anger/
with Intermittent Explosive Disorder (the only disorder in hostility (Tangney et al 1992; Novaco 1994) and impul-
DSM-IV that is exclusively an aggression disorder). Hol- siveness (Barratt 1994). Barratt (1991) suggested that
lander and Stein (1995) edited an extensive review of impulsiveness and anger/hostility characterize impulsive
impulsivity and aggression and noted that "remaining aggression. Lish et al (1996) noted that "anger is the
problems in the field of impulsivity and aggression in- emotional state that usually precedes an aggressive act" (p
clude.., the need for better definition of both dimensional 24). Mood states including anger can affect cognitive
and categorical variables" (p 4). functioning and information processing in adverse ways
The present study extends impulsive aggression re- (Taylor et al 1991; Rolls 1995). Likewise, impulsiveness
search by using: 1) a semistructured interview to classify has been related to information processing through pro-
aggressive acts as impulsive or premeditated rather than posed differences in attention (Dickman 1993), arousal
relying on indirect measures of potential aggressiveness; (Eysenck and Eysenck 1985; Revelle et al 1980; Revelle et
2) personality, cognitive psychophysiological, and neuro- al 1987), and/or maintaining a cognitive tempo or selected
psychological techniques to study information processing rate of information processing (Barratt 1983, 1987; Barratt
differences between persons who commit impulsive ag- and Patton 1983). Stanford and Barratt (in press) have
gressive vs premeditated aggressive acts. We propose that shown that measures of impulsiveness, selected verbal
nonadaptive interpersonal communication that is charac- skills including reading, continuous motor performance,
teristic of the highly impulsive aggressive person when and judging time intervals, define a second order factor of
agitated is related to a lack of impulse control as a result temporal information processing. Related to the latter,
in part of frontal lobe dysfunctioning, a high level of Atkins et al (1993) demonstrated that impulsive errors of
anger/hostility which leads to agitation when disinhibited, commission using a continuous performance task were
and verbal skill deficits which are related to both parietal correlated significantly with hostile aggression, which is
and temporal lobe dysfunctioning. Self-report personality similar to impulsive aggression as defined in this study.
trait measures will be used to measure impulsiveness and Irritability, a complex personality trait as measured by
anger/hostility. Neuropsychology techniques will be used most self-report personality questionnaires, includes both
to measure verbal skills and cognitive psychophysiologi- impulsive and anger personality traits (Caprara et al 1985).
cal techniques will be used to identify cortical dysfunc- Brown and Goodwin (1984) reported that impulsiveness
tioning. The a priori interrelationships of these constructs and irritability were significantly related to aggression.
as stated above provide the basic rationale for this study. Coccaro et al (1989) used a combination of self-report
To test this general proposal, specific hypotheses related to questionnaires including the Buss-Durkee Hostility Inven-
this rationale will be tested using each of three types of tory (BDHI, Buss and Durkee 1957) and the Barratt
measurements: personality, neuropsychology, and cogni- Impulsiveness Scale (BIS, Barratt 1985; Patton et al 1995)
tive psychophysiology. No studies of impulsive aggression to distinguish between the irritable-impulsivity and ag-
have been reported using these three types of assessment gression-assault components of aggression. This distinc-
on the same subjects but the need for such research has tion implies different information processing stages for
Impulsive Aggression BIOLPSYCHIATRY 1047
1997;41:1045-1061

initiating the aggressive act; the irritable impulsivity stage the P300 has been related to the neural resources available
involves a "lowered threshold for a response to a noxious to process a stimulus. Lower P300 amplitudes purportedly
stimulus" (Coccaro et al 1989) while the aggression- involve fewer neural resources and usually have been
assault stage comes later. associated with less efficient cognitive functioning and/or
It is hypothesized in this study that impulsive aggressive pathological states (Branchey et al 1988; Duncan et al
inmates (IAIs) will score higher than nonimpulsive ag- 1994; Pritchard 1989). The amplitude of the P300 has been
gressive inmates (NIAIs) and noninrnate controls (NICs) related to a wide range of cognitive processes including,
on self-report measures of impulsiveness and anger. for example, attention, stimulus probability, arousal, re-
ducing uncertainty, and memory context upgrading.
Neuropsychology Tests and Aggression Donchin (1981) noted that "we have shown that it is the
subjective probability of the stimulus, not its prior proba-
A wide range of neuropsychological techniques has been bility that determines the amplitude of the P300" (p 497).
used to study aggression (Stein et al 1995; Reiss et al Johnson (1986, 1989, 1993) included subjective probabil-
1994; Mirsky and Siegel 1994). The etiological implica- ity as part of his "triarchic model of P300 amplitude" and
tions of these findings are complexly intertwined with postulated multiple subcortical generator sources for the
developmental and environmental constructs (Reiss et al P300 in different sensory modalities and at different
1994; Nestor 1992). As noted, Barratt (1983, 1993) has topographical sites. Subjective probability will be manip-
proposed that the rate or tempo of information processing ulated in this study in a modification of the oddball
or (temporal information processing) is related to impul- paradigm which has been used traditionally to elicit P300s.
siveness and selected verbal information tasks including Although a large number of studies have related ERPs
reading (Stanford and Barratt, in press). Consistent with to psychopathy which is associated with a "proneness to
Barratt's proposal, dyslexia and dysphasia have been commit violent and aggressive crimes" (Mirsky and Siegel
related to temporal information processing (Tallal 1993). 1994), very few studies have related ERPs to violence and
Further, one of the more consistent neuropsychological aggression in general and none to impulsive aggression
findings found in both criminals and noncriminals is an per se. Drake et al (1988) reported longer latency P300
inverse relationship between impulsive aggression and ERPs in patients with "impulsive, aggressive, and violent
verbal abilities (Mungas 1988; Vitiello et al 1990; Reiss et behavior." Branchey et al (1988) reported lower amplitude
al 1994). Many of these studies have involved criminals or P300 peaks among aggressive alcoholics; they noted that
psychopaths and as Raine and Scerbo (1991) noted, a patients with histories of incarceration for crimes involv-
"...conceptual issue that has received little attention in ing physical violence had the lowest P300 amplitudes. In
violence research concerns the issue of whether the cor- topographical studies, the cortical areas most often related
relates of violence differ in any substantive way from the to aggression are the frontal lobes (Raine and Scerbo
correlates of crime..." (p 3). This study addresses this 1991) and the temporal lobes, especially left temporal
issue by comparing two groups of inmates who have (Wong et al 1994). As noted, ERPs will be used in the
exhibited different types of aggressive acts. present study to measure hypothesized brain correlates of
It is predicted in the present study that IAIs will score information processing deficits. Delays in the evaluation
lower than NIAIs and NICs on verbal skills including time of verbal symbols and a lack of mobilization of
reading. sufficient neural resources to make correct verbal discrim-
inations could be related to nonadaptive information pro-
cessing in interpersonal relationships as observed among
Cognitive Psychophysiology and Aggression
IAIs when agitated. It is predicted that IAIs in contrast to
Event-related potentials (ERPs), especially the P300 com- NIAIs and NICs will have lower P300 peak amplitudes
ponent, have a long history as measures of brain function- and longer latency P300 peak latencies in a visual discrim-
ing related to information processing (Donchin 1979; ination task involving different probabilities of stimuli.
Hillyard and Kutas 1983; Duncan et al 1994). The focus in Further, topographically these ERP group differences will
this study will be on the P300 as a measure of differences primarily involve the frontal lobes (based on predicted
in brain functioning related to information processing higher levels of impulsivity among the IAIs) and the left
among inmates and noninmate controls although other temporal and parietal lobe areas (based on predicted lower
components of the ERPs will be evaluated as possible verbal skill levels among the IAIs). It is not assumed that
sources of differences in brain functioning among the the proposed cortical loci differences are necessarily the
groups. The latency of the P300 is generally agreed to be primary or only loci for neural differences related to
a measure of stimulus evaluation time (Hillyard and Kutas aggression between the two inmate groups since informa-
1983; Scheffers and Johnson 1994) while the amplitude of tion processing is related to parallel processing at different
1048 BIOLPSYCHIATRY E.S. Barratt et al
1997;41:1045-1061

levels of circuitry in brain functioning (Weiger and Baer impulsive aggression (the results of the medication trials
1988; Marr 1982). will be published elsewhere).
The ethnic makeup of the three groups of subjects was
Criminali~ vs Aggression not significantly different from each other: 1) 1AIs: Afro-
American, 13; Hispanic, 7; white, 7; 2) NIAIs: Afro-
It is proposed that inmates will differ from noninmates on American, 15; Hispanic, 7; white, 8; 3) NICs: Afro-
the personality, neuropsychology, and cognitive psycho- American, 25; Hispanic, 10; white, 9 (×2 [4] = 1.19).
physiological measures in directions consistent with the There were also no significant differences among the three
results of past studies (Raine 1993). Inmates compared to groups for age in years (IAIs, M = 26.6, SD = 3.9; NIAIs:
NICs will have higher impulsiveness and anger/hostility M = 26.4, SD = 4.5; NICs: M = 26.1, SD = 4.2; F2Aoo
scores, poorer verbal skills, and longer latency and lower = 1.03) and highest grade level completed in school (IAIs:
amplitude P300 peaks. The present research will extend M = 9.9, SD --- 1.4; NIAIs, M = 9.9, SD = 1.4; NICs:
prior research by characterizing differences between in- M = 11.0, SD = 1.6; F2Aoo = 1.42). NICs were paid for
mates who display impulsive and nonimpulsive (premed- participating in the study.
itated) aggression. This is not a study of criminal behavior
per se, but as noted, it is predicted that inmates vs normal INCLUSION CRITERIA. Inmates were not selected for
controls will differ as in past studies. the study based on the crimes they committed to get into
prison but, rather, on the aggressive acts they committed
Methods while in prison. Formal disciplinary reports of the correc-
tional officers were used as entry measures for aggression.
Subject Selection Each inmate in TDCJ prisons is given a handbook of
Subjects (all male) were 57 inmates and 44 noninmate Disciplinary. Rules and Procedures. Known violations of
controls (NICs) matched with a random sample of inmates these rules can result in formal disciplinary reports being
for age, race, and education. The inmates were in the filed. An inmate may contest a formal report and a hearing
Texas Department of Criminal Justice (TDCJ) prisons. and possible dismissal of the charge if the hearing officers
The control subjects were employees of the University of agree with the inmate. No "dismissed" incidents were used
Texas Medical Branch (UTMB) or TDCJ correctional as data in this study.
officers. There were no differences between the two inmate
Inmates were referred by prison officials and were groups in the types of crimes committed that resulted in
identified as having problems controlling their aggressive imprisonment. Although no formal measure of self-re-
behaviors. All subjects had the research explained to them ported aggression was used to measure aggressive behav-
both orally and in written consent forms. The research and iors in early life, the Psychiatric Diagnostic Interview-
consent forms were approved by the UTMB Institutional Revised (PDI-R, Othmer et al 1989) has items related to
Review Board for human research. fighting before age 16 and fighting as adults as part of the
Of 566 inmates screened for the study, 309 inmates antisocial personality disorder interview; 98 percent of the
were eliminated by chart screening using the exclusion inmates responded "yes" to these items for both age levels.
criteria below. Of the remaining 257 inmates, 155 volun- Further, in the initial informal screening interview, in-
teered for the initial interview during which the study was mates were asked how long they had problems with
explained and 138 agreed to participate in the medication aggressive behaviors and all identified the problems as
phase of the study. Of the 17 who did not volunteer, 11 predating incarceration.
were not interested in the study, 3 did not want to take To enter the study, an inmate must have committed at
medication, and 3 denied aggressive behaviors. Additional least three documented aggressive acts as outlined in the
subjects were eliminated during ERP artifact screening (6 Handbook over the three-month period prior to entry into
inmates and 4 NICs). Data for left-handed subjects were the study. Most inmates had more incidents, with the
also not included in the analyses (1 NIC and 6 inmates). average inmate having about 6.2 incidents. The aggressive
Using a semi-structured interview (see below), the acts were evaluated for intensity and for impulsiveness/
remaining inmates were divided into two groups: 1) nonimpulsiveness. The intensity of each act was weighted
impulsive aggressive inmates (IAIs, n = 27); 2) non- according to its social significance within the prison on a
impulsive (premeditated) aggressive inmates (NIAIs, n = ten-point scale established with the help of prison officials.
30). This research also includes noninmate outpatients but Hitting a correctional officer or prison employee was rated
the n to date is too small to report meaningful results. The 10, for example, while hitting another inmate was rated 8.
data for this study were obtained during the baseline phase To classify the aggressive acts as impulsive or nonim-
of a study that related anticonvulsants to the control of pulsive, a brief semistructured interview was developed
Impulsive Aggression BIOLPSYCHIATRY 1049
1997;41:1045-1061

that was used along with the correctional officers' formal were taught to respond to cards that contained possible
reports to make the classification. An impulsive aggressive item responses.
act was defined as a hair-trigger, non-premeditated re-
sponse to a stimulus that results in an immediate aggres- NEUROPSYCHOLOGY AND READING TESTS. Wechsler
sive act or an agitated state that culminates in an aggres- adult intelligence scale (WAIS; Wechsler 1955). Three
sive act. An example of a nonimpulsive aggressive act was verbal (comprehension, similarities, vocabulary) and three
intentionally hitting an inmate because he was a member performance (digit symbol, block design, picture arrange-
of another gang. To establish the reliability of the inter- ment) subtests were administered. Prorated verbal (VIQ)
view, two independent raters classified 58 acts with a and performance (PIQ) IQs were calculated.
kappa of 0.83 (p < 0.0001). All aggressive acts cannot Wechsler Memory Scale--Revised (WMS-R; Wechsler
clearly be classified as impulsive or nonimpulsive. Only 1987). Verbal memory, visual memory, and attention/
subjects whose aggressive acts were clearly impulsive or concentration scores were used.
nonimpulsive were used in this study. The NICs had no Gray oral reading test (GORT; Gray 1963) provides a
history of repeated aggressive outbursts. grade-level score based on the "mechanics" of oral read-
ing. Word repetition, mispronunciations, and word substi-
EXCLUSION CRITERIA. Subjects were excluded for: 1) tutions or omissions are typical errors.
having a prorated verbal and performance WAIS IQ less
than 80; 2) having a current major DSM-III-R Axis I E R P RECORDING CONDITIONS AND ANALYSES, S u b -
disorder as measured by the PDI-R (Othmer et al 1989); j e c t s were seated in a comfortable chair in a semi-
(subjects were not eliminated for having an antisocial darkened room. They wore earphones that masked extra-
personality disorder (ASP); 3) presently taking medica- neous sounds with a continuous white noise in the 45-60
tion; or 4) having a neurological or other serious medical dB range (adjusted to the comfort of the subject). Nineteen
disorder. Neurological and medical screenings were based scalp electrodes (International 10-20 System) were at-
on life history data, the medical chart (includes a record of tached using an electro-cap. Fifteen electrode sites were
medical examinations while in prison), and a five-minute used in this study: F3, Fz, F4, T3, C4, Cz, C4, T4, T5, P3,
Pz, P4, T6, O1, O2. Impedance for each electrode was less
resting EEG under eyes open and closed conditions at the
than 5 Kohms. All ERP data were acquired using SCAN
beginning of the psychophysiological recording session.
software (Neurosofl 1993) and a Nihon Kohden (Model
The EEG recordings were visually scanned for any indi-
EEG-4221) polygraph with a bandpass of 0.05-30 Hz.
cations of slowing or other suggestions of a neurological
Electrodes were referenced to linked ears. Subjects were
disorder. This was not intended to be a clinical EEG but a
grounded via an electrode on the forehead. Electrooculo-
gross screen.
grams (EOGs) were recorded from electrodes placed 2.5
All of the inmates qualified for an antisocial personality
cm supraorbitally and 1.5 cm laterally from the left eye.
disorder and 55 percent of the inmates had a lifetime, but
EEGs and EOGs were digitized every 4 ms for a 1000 ms
not current, substance abuse disorder that was evenly
interval beginning 100 ms prior to stimulus onset. EOG
distributed among inmate groups. None of the NICs met
artifact contamination was removed using the method of
criteria for a psychiatric disorder.
Semlitsch et al (1986); a transmission coefficient was
computed based on the estimated covariance of the aver-
age potentials of the ocular channel with the EEG channels
Measurement Techniques and Instruments
in the frequency domain and this coefficient was sub-
SELF-REPORT PERSONALITY MEASURES. Barratt im- tracted from each ERP sweep point by point. All trials
pulsiveness scale (BIS; Barratt 1985; Patton et al 1995): a were averaged to identify peaks after this correction. Peak
total score of general impulsiveness is obtained by sum- amplitudes were measured using baseline levels of the
ming three subtest scores: motor (act without thinking), average amplitude of the 100 ms of recordings prior to a
cognitive (making-up one's mind quickly), and nonplan- stimulus presentation. Latencies were measured from
ning (not planning ahead). stimuli onset.
NEO-PI impulsiveness facet (Costa and McCrae 1985): Seven subjects (3 IAIs, 3 NIAIs, 1 NICs) were elimi-
A self-report measure of behavioral self-control. nated because of excessive artifacts under all conditions.
Anger expression scale (Spielberger 1988) measures After ERPs for trials in which response errors were
anger-in, anger-out, anger control, and total anger expres- eliminated, at least 96 percent of the sweeps for the
sion. remaining subjects were used in computing subject aver-
Since many of the inmates had poor verbal skills, the ages for all conditions and tasks. No trials were lost due to
questionnaire items were read to all subjects. Subjects ocular artifacts using the Semlitsch et al (1986) procedure.
1050 BIOL PSYCHIATRY E.S. Barratt et al
1997;41:1045-1061

The actual number of trials averaged across conditions two tasks but with the addition of 17 unexpected stimuli
were comparable across groups after all contaminated (letters other than "A" or "B") being presented randomly
trials were removed. During the 0.1R condition of the off along with the target and nontarget stimuli. The unex-
target task, only 17.5 trials were used to obtain the average pected stimuli were presented in 10 percent of the trials
ERPs for each subject. This is below the suggested (0.1 nonresponse condition; 0.1NR), "B" in 10 percent of
minimum of 20 trials for individual ERP averages; how- the trials (0.1 response; 0.1R), and "A" in 80 percent of the
ever, the error rates were relatively low and constant trials (0.8 condition). Letters were presented randomly
across groups and the number of trials was considered over 175 trials. Subjects used their dominant hand to
comparable to those in other conditions. respond to the 0.1R condition stimuli and their nondomi-
Subjects were run at the same time each day starting nant hand to respond to the 0.8 condition stimuli.
between 12:30 PM and 1:30 PM. The starting time was at
least one hour after their last meal. Subjects did not smoke RATIONALE FOR SELECTION AND ORDER OF PRESENTA-
during the day prior to being run. There were no group TION OF COGNITIVE PSYCHOPHYSIOLOGICAL TASKS. The
differences in the time of year that subjects were run since three tasks were presented as three separate experiments.
an attempt was made to run subjects in all groups during Subjects were instructed to respond to 0.2, 0.8, and 0.1R
different seasons (Polich and Geister 1991) (X2 [6 df] = stimuli to provide a behavioral measure of alertness. They
1.39). were not instructed to "respond quickly." Past research
Although the primary focus in this study was on P300, (Barratt 1993) indicated that impulsiveness is usually
the amplitudes and latencies for N 100, N200, P200, P300 inversely related to reaction times when there is a warning
and slow wave peaks were determined for each task and signal prior to the imperative stimulus and subjects are
stimulus condition. The time windows for these peaks instructed to "respond quickly." Thus, it was expected that
were: N100, 100-200 ms; P200 and N200, 200-300 ms; the IAIs would have longer reaction times than the NIAIs
P300, 300-600 ms; slow wave, 550-750 ms. Peaks were although probably not statistically significant different
identified initially by visual inspection and amplitude and latencies. It was also expected that both inmate groups
latency values were then assigned using computer soft- would have longer reaction times than the NICs.
ware (SCAN). A random sample of peak values was
identified by a second observer to check on the reliability
of the visual inspection.
Data Analyses
Personality and neuropsychology data were analyzed for
TASK DESCRIPTION AND INSTRUCTIONS. Subjects w e r e group differences using Kruskal-Wallis one-way analysis
seated in front of a 48.3 cm monochrome video screen of variance (ANOVAs) and the Wilcoxon signed ranks
with their eyes 54 cm (_ 2 cm) from the center of the test (Siegel and Castellan 1988). Performance errors were
screen. Stimuli were 8 mm square letters at the center of analyzed using chi square (×2) for independent samples.
the screen. A 1000 Hz tone one second in duration Within-task-group differences for ERPs and reaction times
sounded one second prior to the appearance of each letter. (RTs) were analyzed using ANOVAs and Dunn's multiple
Letters remained on the screen for 1 second with a comparisons procedure (Bonferroni "t" procedure) where
3-second interstimulus interval. Subjects were instructed appropriate (Kirk 1982). Multiple comparisons (e.g.,
to relax, fixate on the dot in the middle of the screen, and group X electrode x task) were analyzed using repeated
rest their arms on the chair. Subjects completed the measure ANOVAs with Greenhouse-Geisser corrections
following three two-choice response tasks with two min- for violations of the sphericity assumption and Bonferroni
utes between tasks. multiple comparison tests if appropriate (Kirk 1982; Har-
Odd-left: The letter "A" appeared on 80 percent (0.8 ris 1994).
condition) of the trials and the letter "B" appeared on 20 Electrode groupings for within-task analyses include: 1)
percent (0.2 condition) of the trials. Subjects pressed a midline electrodes (Fz, Cz, Pz); 2) left vs right hemi-
response key with the left forefinger when "B" appeared sphere; 3) selected groupings to test specific hypotheses
and the right forefinger when "A" appeared, Their fore- (e.g., frontal F3-F4, temporal T3-T4). An ANOVA was
fingers rested on the response keys prior to the press. The also computed for group differences among all electrodes.
letters "A" and "B" were presented randomly over 100 Before analyzing the ERP data, the raw score peak
trials. amplitudes and latencies were converted to normalized
Odd-right: The same procedure as odd-left except right scores using a procedure suggested by McCarthy and
and left hands were reversed for pressing the keys in Wood (1985). This involved determining, for example, the
response to the two letters. maximum and minimum amplitudes for all electrodes
Off-target: The general procedure is the same as the first within electrode groups for each group of subjects and
Impulsive Aggression BIOLPSYCHIATRY 1051
1997;41:1045-1061

Table 1. Group Comparisons of Impulsiveness and Anger


Non-impulsive Impulsive aggressive
Noninmate controls aggressive inmates inmates

Test M SD M SD M SD

Barratt impulsiveness scale 32.6 12.2 48.0 11. I 51.9 11.3


NEO impulsiveness facet score 16.3 4.1 18.7 3.9 18.6 3.8
Anger in 16.5 4.3 21.4 4.8 20.2 4.2
Anger out 16.0 4.1 20.8 4.6 21.1 4.6
Anger control 22,6 5.2 17.4 4.2 16.0 4,3
Anger expression 27.0 5.9 41.3 6.2 39.1 6.9

Scores for noninmate controls were all significantly different (p -< 0.(X)5) from both groups of inmates for all tests. The scores for the two inmate groups did not differ
significantly from each other for any test (Kruskall-Wallis one way analysis of variance by ranks and the Wileoxon signed ranks test),

then using the formula suggested by McCarthy and Wood Neuropsychology and Reading Tests
(1985) for normalizing the amplitudes for all subjects. The two groups of inmates differed significantly from
This procedure permits topographical comparisons that each other on four of the five individual verbal skills
were not affected by the absolute voltage levels of the measures. This is in contrast to scores on visual-spatial
ERPs. The vector method (McCarthy and Wood 1985) subtests where there were no significant differences be-
was also used to normalize the amplitudes for midline
tween the two inmate groups. In general, the NICs scored
electrode analyses as a check on the method used here.
significantly better than both inmate groups on all of the
The results of the ANOVAs using the vector method to
neuropsychology tests including the verbal subtests. Pic-
normalize the midline electrode amplitudes did not differ
ture arrangement scores, a WAIS performance subtest that
significantly from the results using the procedure de-
is often considered a nonverbal counterpart of verbal
scribed above.
comprehension (Lezak 1983), were not significantly dif-
Neuropsychology, cognitive psychophysiology, and
ferent among the three groups. The WAIS verbal compre-
personality measures were intercorrelated using Pearson
hension subtest scores were also not significantly different
product moment correlation coefficients (Harris 1994).
between the two inmate groups (Table 2).
Differences between correlations were tested for signifi-
cance using the Fisher r-to-z' transformation (Kirk 1990).

Behavioral and Cognitive Psvchophysiological


Data Presentation Behavioral Measures
The main hypotheses involve comparing two inmate REACTION TIMES. There were no significant differ-
groups who displayed two different types of aggression. A ences in reaction times between the two inmate groups and
noninmate control group was included to compare the between the NICs and NIAIs. The NICs and IAIs differed
inmate-noninmate results of this study with reported significantly on reaction times in three conditions: 1)
relationships between criminals and noncriminals (Raine
odd-left 0.8 (Fz,99 =- 3.37, p < 0.05, Dunn's test with
1993). Because of the extensive data analyses, all of the
alpha = 0.05); 2) odd-left, 0.2 (F2,99 = 4.02, p < 0.05;
results are not presented in detail for inmate-noninmate
Dunn's test with alpha = 0.05; 3) odd-right, 0.8 (F2,99 =
analyses and for all of the insignificant differences be-
3.28, p < 0.05, Dunn's test with alpha = 0.05.
tween inmate groups for the cognitive psychophysiologi-
Although the differences in reaction times were in
cal analyses. Rather, summary statements are included to
general not significant between groups, the inmates had
cover these selected analyses.
longer latencies than the NICs in all instances. This latter
result is consistent with past research which reported
inverse relationships between impulsiveness and reaction
Results
times, even though the subjects were not instructed to
Impulsiveness and Anger Self-Report Measures respond quickly in the present experiment (Table 3).
Contrary to predictions, the anger and impulsiveness
scores did not differ significantly between the NIAIs and P E R F O R M A N C E ERRORS. Performance errors included

the IAIs. The NICs scored significantly lower than both responding before a stimulus appeared, not responding
inmate groups on both anger and impulsiveness (Table 1). within 1000 ms, or responding to a letter other than "A" or
1052 BIOL PSYCHIATRY E.S. Barratt et al
1997;41:1045-1061

Table 2. Group Comparisons of Neuropsychology Test Scores

Inmates
Non-impulsive Noninmate Impulsive
aggressives controls aggressive Probability levels of differences between
(NIAI) (NIC) (IAI) groups~
Test M SD M SD M SD NIC vs NIAI NIC vsIAI NIAI vslAI
WAIS
Comprehension 10.61 2.55 8.41 3.19 6.33 2.56 <0.01 <0.01 NS
Similarities 12.39 2.60 10.76 2.07 9.11 2.61 <0.0l <0.01 0.01
Vocabulary 9.07 2.18 9.02 3.41 6.01 2.14 <0.01 <0.01 <0.05
Digit symbol 10.95 2.72 9.22 1.92 9.21 1.67 <0.01 <0.01 NS
Block design 11.60 1.99 9.91 3.00 9.56 2.85 <0.01 <0.01 NS
Picture arrangement 9.07 2.33 9.02 2.11 9.02 1.55 NS NS NS
VIQ (Prorated) 103.19 11.12 93.99 14.1 84.86 1(/.99 <0.01 <0.01 <0.01
PIQ (Prorated) 103.47 12.98 98.2 11.19 95.22 8.32 <0.05 <0.01 NS
WMS-R
Attention/concentration 94.81 16.00 87.66 l 3.04 83.74 18.49 NS <0.01 NS
Verbal memory 93.1 13.73 85.02 11.42 78.19 9.49 <0.01 <0.01 <0.05
Visual memory 102.88 18.16 95.99 13.69 89.41 17.24 NS <0.01 NS
GORT
Grade level 10.00 3.42 8.71 3.41 5.18 2,49 NS <0.01 <0.01
"Based on Kruskall-Wallisone way analysis of variance by ranks and the Wilcoxon signed ranks test.

" B " d u r i n g the o f f - t a r g e t task. T h e r e w e r e n o s i g n i f i c a n t o n d i f f e r e n c e s b e t w e e n the t w o i n m a t e g r o u p s as p r e v i -


d i f f e r e n c e s a m o n g g r o u p s in p e r f o r m a n c e errors ( T a b l e 4). o u s l y noted.

PEAK AMPLITUDES AND LATENCIES FOR N100, P200, P 3 0 0 PEAK LATENCIES. C o n t r a r y to p r e d i c t i o n s , t h e r e


N 2 0 0 , AND SLOW WAVE E R P s . There were no significant were no significant differences among the latencies of the
g r o u p d i f f e r e n c e s in p e a k l a t e n c i e s or a m p l i t u d e s for p e a k P 3 0 0 E R P s for a n y task or c o n d i t i o n ( F i g u r e s 1, 2, 3).
N 1 0 0 , P 2 0 0 , a n d N 2 0 0 E R P s for a n y task or p r o b a b i l i t y
c o n d i t i o n . T h e r e w e r e also n o s i g n i f i c a n t g r o u p differ- P 3 0 0 PEAK AMPLITUDES. There were no significant
e n c e s in o n s e t o f the s l o w w a v e c o m p o n e n t , h o w e v e r , d i f f e r e n c e s a m o n g g r o u p s in the P 3 0 0 p e a k a m p l i t u d e s for
t h e r e w e r e s i g n i f i c a n t d i f f e r e n c e s in a m p l i t u d e s o f the the 0.8 c o n d i t i o n a c r o s s tasks. (F2.98 = 0.99). T h e r e w e r e
slow w a v e c o m p o n e n t b e t w e e n the i n m a t e s a n d N I C s l i k e w i s e n o g r o u p d i f f e r e n c e s in p e a k P 3 0 0 a m p l i t u d e s
d u r i n g the o d d - l e f t 0.2 a n d the o d d - r i g h t 0.2 c o n d i t i o n s at b e t w e e n h e m i s p h e r e s for l a t e r a l i z e d e l e c t r o d e c o n f i g u r a -
s e l e c t e d e l e c t r o d e sites ( F i g u r e s 1, 2). T h e s e d i f f e r e n c e s tions (F4, C4, P4, T4, T6, 0 2 v s F3, C3, P3, T3, T5, O 1 )
w e r e m o s t o b v i o u s in the p a r i e t a l leads (P3, P4, Pz) a n d (F2,98 -- 0.02). T h e r e w a s a s i g n i f i c a n t p e a k P 3 0 0 a m p l i -
less so in c o n t i n g e n t a n t e r i o r areas (C3, Cz, C4). T h e t u d e d i f f e r e n c e a m o n g g r o u p s for all e l e c t r o d e s (F2,88 =
s i g n i f i c a n t d i f f e r e n c e s b e t w e e n i n m a t e s a n d N I C s a n d the 3.46, p < 0.05 u s i n g D u n n ' s test w i t h p = 0.05; the t h r e e
i n s i g n i f i c a n t results o n o t h e r t h a n P 3 0 0 p e a k E R P s are n o t g r o u p s w e r e s i g n i f i c a n t l y d i f f e r e n t f r o m e a c h o t h e r across
p r e s e n t e d in detail s i n c e the p r i m a r y f o c u s o f this s t u d y is all e l e c t r o d e s ) ( F i g u r e s 1, 2, 3; T a b l e s 5, 6).

Table 3. Group Comparisons of Reaction Times (Milliseconds) by Task and Stimuli ~'t'

Odd-left Odd-right Off-target


0.8 0.2 0.8 0.2 0.8 0.1
M SD M SD M SD M SD M SD M SD
NICs 437 83 479 86 439 81 484 83 444 87 514 80
NIAIs 453 64 483 77 460 65 509 81 455 61 541 62
IAIs 480 63 505 73 469 59 512 63 470 71 549 88
~The NICs and IAls differed significantlyon the odd-left0.8 condition(F2.too = 3.37. p < 0.05). the odd-left0.2 condition(F2.11n~= 4.02, p < 0.05), and the odd-right
0.8 condition(F2.~oo = 3.28. p < 0.05). (Dunn's test with alpha = 0.05 was used for individualgroup comparisons).
Vl'here were no significantdifferences in reaction times between the NIAIs and NICs and the IAls and NIAIs.
Impulsive Aggression BIOLPSYCHIATRY 1053
1997:41:1045-1061

Table 4. Performance Errors: Group Comparisons ~'b


Odd-left Odd-right Off-target

0.8 0.2 0.8 0.2 0.8 0.1

NICs 0.3 1.0 0.1 1.0 1.0 1,0


NIAIs 0.4 2.0 0.3 2.0 0.3 2,0
IAls 1.0 2.0 3.0 3.0 1.0 2,0

~Errors are expressed as percentages based on total number of trials per group
for task and stimulus conditions•
bThere were no significant group differences (Xe [10] = 2.13).

Figure 3. Group comparisons of P300 super averages for offtar-


TASK × GROUP × E L E C T R O D E MIDLINE P 3 0 0 AMPLI- get 0.1R probability condition. S = stimulus; NICs = noninmate
TUDES. There was a significant task-by-group-by-elec- controls; NIAIs = nonimpulsive aggressive inmates; IAIs =
impulsive aggressive inmates.
trode interaction in P300 peak amplitudes for the midline
electrodes (Fz, Cz, Pz) (F8,336 = 5.21, p = 0.0001, e --
0.7437). This difference was primarily related to differ- electrode site and both inmate groups had significantly
ences in the amplitudes of the NIAIs across tasks for the lower amplitude P300 amplitudes than the NICs.
Cz and Pz electrodes (Figs. 1, 2, 3; Table 5). For the
odd-left and off-target task, the NIAIs had significantly GROUP × TASK × CONDITION (0.2/0.1R) COMPARISONS
higher peak P300 amplitudes than the IAIs at electrodes OF P300 AMPLITUDES. Peak P300 amplitudes for 0.2 and
Cz and Pz. During the odd-fight task, there was no 0.1R conditions were also compared for pairs of lateral
significant difference between the NIAIs and IAIs at any homologous electrodes from anterior to posterior scalp
locations (Table 6). The NICs had significantly higher
peak amplitudes than both inmate groups at frontal elec-
trode sites for all three tasks. Contrary to predictions, the
two inmate groups did not differ significantly from each
other in peak frontal amplitudes. The IAIs when compared
with NIAIs and NICs had significantly lower P300 peak
amplitudes at all sites posterior to the frontal areas for the
odd-left 0.2 condition. For the odd-fight task, the NICs had
significantly higher peak P300 amplitudes than both in-
mate groups at all electrode sites except T3, T4, T5, T6,
.......... ~ ~ I~.~ • and O 1. The IAIs and NIAIs did not differ significantly in
ll~0ms
3 . . . . I~s
s
peak amplitudes on the odd-fight task. The differences in
peak P300 amplitudes among groups for the 0.1R condi-
Figure 1. Group comparisons of P300 super averages for oddleft tion during the off-target task were similar to the 0.2
0.2 probability conditions. S = stimulus; NICs = noninmate condition for the odd-left task except that there were no
controls; NIAIs = nonimpulsive aggressive inmates; IAIs = significant differences among groups for electrodes C4
impulsive aggressive inmates.
and O 1.

INTERRELATIONSHIPS OF SELECTED PERSONALITY,


NEUROPSYCHOLOGICAL, AND COGNITIVE PSYCHOPHYSIO-
LOGICAL MEASURES ACROSS TASKS FOR COMBINED
GROUPS. The main purpose of these analyses was to get
a global overview of the interrelationships among selected
neuropsychology-personality measures and midline P300
amplitudes. Measures were selected that were relevant to
the main hypotheses.
3 . . . . l~s
Impulsiveness and anger were significantly intercorre-
lated (Table 7) consistent with past research (Barratt et al
Figure 2. Group comparisons of P300 super averages for odd-
right 0.2 probability condition. S = stimulus; NICs = noninmate 1995a). Items on self-report anger scales often include
controls; NIAIs = nonimpulsive aggressive inmates; IAIs = "control" implications and therefore correlate significantly
impulsive aggressive inmates. with impulsiveness measures.
1054 BIOLPSYCHIATRY E.S. Barratt et al
1997;41:1045-1061

Table 5. Peak P300 Amplitudes across Groups, Conditions, and Tasks


Odd-left Odd-right Off-target

N1Cs NIAIs 1Als NICs NIAIs IAls NICs NIAIs IAls

0.8 0.2 0.8 0.2 0.8 0.2 0.8 0.2 0.8 0.2 0.8 0.2 0.8 0.1 0.8 0.1 0.8 0.1

F3 5.9 10.7 3.3 8.9 4.9 6.4 5.1 8.9 4.6 4.9 5,0 5.4 5.1 8.3 3.1 6.0 4.0 5.0
Fz 6.7 10,5 3.4 7.2 5.3 6.1 5.3 9.9 5.3 6.7 5.1 6.6 5.3 9. l 3.1 5.4 4.4 4.8
F4 8.0 13,4 4.7 6.3 5.3 5.2 6.0 10.2 4.7 6.7 5,2 6.9 6.3 7.5 4.5 5.3 5.6 5.3
T3 4.9 6.9 3.5 6.5 3.1 4.1 4.7 6.0 3.0 4.5 3.6 4.8 4.7 5.4 2.9 4.5 3.6 3.0
C3 7.5 11,9 4.4 9.9 5.1 6.5 7.8 10.9 5.6 8.5 5.7 6.3 6.8 9.9 4.7 8.9 5.3 5.7
Cz 8.(1 11.1 5.0 9.6 5.3 6.2 8.5 13,4 4.4 7.7 4.9 7.4 5.7 9.1 4.0 8.9 5.0 5.6
C4 7,6 9.9 6.1 8.6 5.6 6.7 7.2 13.7 5.0 8.0 4.8 7.5 8.2 8.6 5.2 7.8 6.0 5.4
T4 4.4 6,3 3.4 4.0 3.4 4.0 1.8 4.9 3.2 3.4 4.0 4.8 5.4 5.6 4.4 5.4 4.3 5.4
T5 5.4 7.8 4.5 7.7 3.4 3.6 5.2 8.4 4.1 6.9 3.3 5.0 5.5 5.9 4.1 5.8 3.2 3.1
P3 8.1 13.6 6.3 12.3 5.3 7.3 8.6 13,4 6.7 9.8 5.7 8.2 8.4 10.5 6.0 9.3 5.8 6.0
Pz 9.8 14.1 6.5 10.8 5.7 7.2 11.1 14.7 7.2 9.5 6.3 8.5 9.5 11.0 5.9 9.8 6.0 6.5
P4 8.1 11.4 6.1 10.2 5.8 7.2 10.3 13.4 6.4 9.9 5.6 7.9 8.7 10.4 5.6 9.6 5.l 6.1
T6 4.1 7.1 3.6 6.6 2.9 4.1 5.2 7.2 3.3 5.0 4.1 4.6 4.9 6.1 3.4 5,9 2.5 3.7
O1 4.4 7.4 3.8 7.0 3.1 3.7 3.6 7.3 4.3 6.1 3.4 4.7 5.1 5.4 3. t 5,2 2.4 2.7
02 7.8 12.1 6.9 10.1 4.5 5.9 6.7 11.6 5.4 9.6 4.5 6.6 8.3 9.5 7.1 7,1 3.1 3.4

Both impulsiveness and anger were inversely related to while the neuropsychology measures were positively re-
the neuropsychology measures with impulsiveness (and lated (Table 8). The correlations of the personality-
not anger) correlations being significant (Table 7). Impul- neuropsychology measures with the P300 amplitudes var-
siveness was correlated significantly higher with VIQ than ied across tasks. During the odd-left task, the BIS had a
with the other neuropsychology tests. The V1Q had the significantly higher negative correlation with the P300
highest average correlation among the intercorrelations of amplitude at Fz than at Pz. The reverse was true for the
the neuropsychology tests with an especially high corre- VIQ and GORT measures with the Pz P300 amplitude
lation with the GORT. correlation being significantly higher at Pz than at Fz.
Consistent with the relationship of the personality mea- During the odd-fight task, the correlations of BIS, VIQ,
sures with the neuropsychology measures, impulsiveness and GORT with P300 amplitudes at each of the midline
and anger were negatively related to P300 amplitudes electrodes were not significantly different, although they

Table 6. P300 Peak Amplitude Differences across Groups and Tasks for the 0.2 and 0.1R Probability Conditions for Selected
Electrode Configurations
Odd-left Odd-right Off-target
(0.2 Condition) (0.2 Condition) (0.1R Condition)

Electrode ANOVAs F2.9~ Individual electrodes ANOVAs F2.9,~ Individual electrodes ANOVAs F2,99 Individual electrodes
configuration Ip) (Group differences) "'h IT) /Group differences)"J' (p) (Group differences) ~'t'
Fz, Cz, Pz 4.61 Fz 1>2,3 7.90 Fz 1>2,3 4.01 Fz 1>2,3
(<0.05) Cz 3 < 1,2 (dO.() 11 Cz 1> 2,3 ( <(I.05 ) Cz 3<1,2
Pz 3< 12 Pz 1>2,3 Pz 3<1,2

F3, F4 6.21 F3 I >2,3 5.65 F3 1>2,3 5.60 F3 1>2,3


(<0.01) F4 1>2,3 ( <0.01 ) F4 1>2,3 (<0,01) F4 1>2,3
C4, C3 6.94 C4 3< 1,2 7.92 C4 I >2,3 8,25 C3 3<1,2
1<0.01 ) C3 3< 1,2 (<0.(11) C3 I >2,3 (<0,01)
P4, P3 5.00 P4 3< 1,2 5.01 P4 l >2,3 6,30 P4 3<1,2
(<0.01) P3 3< 1,2 /<0.01 ) P3 1>2,3 (<0.01) P3 3<1,2
T3, T4 1.09 1.06 1.02
(NS/ (NSi (NS)

T5, T6 3.27 T5 3< 1,2 2.82 2.83


(<0.05) T6 3< 1,2 iNS) (NS)
O1, 02 7.19 Ol 3< 1,2 3.91 02 3< 1,2 3.49 02 3<1,2
(<0.01 ) 02 3< 1,2 (<0.05) (<0.05)

"Group I = noninmate controls (NICs); Group 2 = nonimpulsive aggressive inmates IN1Als): Group 3 = impulsive aggressive inmates (lAls).
~'Dunn's test with alpha = 0.05 was used for individual group comparisons.
Impulsive Aggression BIOLPSYCHIATRY 1055
1997-41:1045-1061

Table 7. Intercorrelations of Personality, Neuropsychology and Reading Tests for Combined Groups"
BIS ANG VIQ PIQ Verbal mem Visual mem GORT

Barratt impulsiveness scale (BIS)


Anger out (ANG) 0.51 --
V1Q (prorated) -0.52 -0.19 I

PIQ (prorated) -0.33 - 0.10 0.54


WMS-R verbal memory (Verb Mere) -0.29 -0.14 0.43 0.24 --
WMS-R visual memory (Vis Mem) -0.33 ,0.13 0.50 0.56 0.35 --
Gray oral reading test (GORT) -0.35 -0.19 0.67 0.40 0.42 0.47

~n = 100; r.o~ = 0.25.

were significantly different from zero (Table 8). Also, the personality disorder (American Psychiatric Association
positive-negative pattern of the correlations was the same 1994). It would be expected that two groups with this
as during the odd-left task. disorder would not differ significantly on impulsiveness
During the off-target task, the anterior-posterior gradi- and anger, especially impulsiveness.
ent of the BIS correlations with the midline electrodes was
the same as during the odd-left task (Table 8), however, Neuropsychology/Reading
the VIQ and GORT correlations did not differ signifi- In general, the IAIs scored significantly lower than the
cantly across electrodes as they had during the odd-left NIAIs on verbal skill tests as predicted. The one exception
task, although these correlations were also significantly was the WAIS comprehension subtest score that was not
different from zero. significantly different between the two inmate groups.
Related to this finding, there were no significant differ-
Discussion ences among the three groups on the WAIS picture
arrangement subtest, a test that is considered to be a
Impulsiveness and Anger nonverbal counterpart of the comprehension subtest. Thus,
The IAIs did not score significantly higher than the NIAIs all three groups are not different in the ability to compre-
on self-report measures of impulsiveness and anger as had hend using nonverbal symbols, the NICs comprehend
been predicted. There are several possible explanations for significantly better than both inmate groups using verbal
this. First, the self-report measures of anger and impul- symbols, and the two inmate groups do not differ on verbal
siveness may not discriminate among subjects who have comprehension. The items on the WAIS comprehension
very high levels of both traits (Barratt et al 1995a). These subtest provide a context, often social, for the responses.
measures do discriminate between persons with high and Words per se may not be understood but often can be
low trait scores as evidenced by the NICs scoring signif- interpreted within the context of the overall text for the
icantly lower than both inmate groups as was expected. item. This is in contrast to the verbal tests that were
A second and probably more plausible explanation significantly different between the inmate groups that
relates to both inmate groups meeting the criteria for an either did not provide a context for interpreting words per
antisocial personality disorder. Impulsiveness and irrita- se (WAIS similarities and vocabulary subtests; WMS-R
bility (anger) are both potential criteria for an antisocial verbal memory subtest) or were scored for errors in the

Table 8. Personality, Neuropsychology and Reading Related to Midline P300 Amplitudes (0.2 and 0.1R Conditions) across Tasks
for Combined Groups
Odd-left~ Odd-fight" Off-target"

Fz Cz Pz Fz Cz Pz Fz Cz Pz

Barratt impulsiveness scale (BIS) -0.44 -0.34 -0.27 -0.31 -0.41 -0.41 -0.37 -0.30 -0.21
Anger out (ANG) -0.31 -0.22 -0.17 -0.21 -0.20 -0.24 -0.22 -0.16 --0.07
WAIS VIQ (prorated) 0.28 0.35 0.46 0.28 0.40 0.42 0.36 0.40 0.34
WAIS PIQ (prorated) 0.24 0.23 0.29 0.17 0.18 0.22 0.30 0.28 0.30
WMS-R verbal memory (Verb Mem) 0.24 0.35 0.36 0.36 0.37 0.30 0.27 0.30 0.18
WMS-R visual memory (Vis Mem) 0.15 0.22 0.25 0.16 0.26 0.19 0.33 0.35 0.26
Gray oral reading test (GORT) 0.20 0.26 0.37 (/.23 0.37 0.34 0.40 0.42 0.35

aStimulus probability - 0.2. hStimulus probability = 0.1.


n - 98; rm = 0.25.
A difference of 0.16 between correlations is significant using the Fisher r-to-z transformation.
1056 BIOLPSYCHIATRY E.S. Barratt et al
1997;41 : 1045-1061

"mechanics" of reading (GORT; scored for word omis- IAIs in the topography of the peak P300 amplitudes in
sions, mispronunciations, obvious letter reversals, etc.). electrodes posterior to the frontal areas. These differences
Errors on these latter subtests are more likely to be related varied with tasks and involved the aforementioned differ-
to decoding problems in reading than to the ability to ences in the P300 peak amplitudes for the NIAIs across
comprehend or understand. Although the two inmate electrodes and tasks. During the odd-left task, the IAIs
groups differed significantly on WMS-R verbal memory, compared to NIAIs had significantly lower P300 peak
there was no significant difference between them on the amplitudes posterior to frontal leads in the midline, bilat-
WMS-R visual memory. Thus, the differences between eral sensory-motor strip (C3, C4), bilateral posterior tem-
NIAIs and IAIs are not attributable to memory differences poral leads (T5, T6), and the striate cortex (O1, 02).
per se. Except for the difference at electrode 02, the differences
A distinction is often made between reading compre- between NIAIs and IAIs were not present during the
hension and reading problems related to the grapheme- odd-right task. During the off-target task, the differences
phoneme linkage (decoding) in reading (Gaddes 1980; in P300 peak amplitudes between IAIs and NIAIs were
Njiokiktjien 1993). Many persons who do not read well similar to the differences recorded during the odd-left task
because of decoding problems can comprehend well. except for electrodes T5, T6, and O1. The pattern of
These decoding problems in reading are usually labeled differences in P300 peak amplitudes between the NIAIs
developmental dyslexia. The GORT is a reading test that and IAIs during the odd-left and off-target task were more
is scored primarily for decoding errors. The NICs and in line with predicted differences than those recorded
NIAIs did not differ significantly on the GORT but the during the odd-right task. It is not clear why the P300 peak
IAIs scored significantly lower than both NICs and NIAIs amplitudes for the NIAIs would vary across tasks while
on this test. the amplitudes for the other two groups remained rela-
In summary, the neuropsychology/reading data suggest tively stable. The off-target task differs in stimuli pre-
that the two inmate groups differ significantly on deficits sented and total number of trials from the odd-right and
involving a particular type of verbal information process- odd-left tasks. The odd-right and odd-left tasks are similar
ing, namely, developmental dyslexia (Duncan et al 1994). in stimuli presented. It may be that NIAls and NICs were
They do not differ in the ability to comprehend except as more aroused to relatively new task demands than were
decoding of the reading symbols enters directly into verbal the IAIs but the NIAIs did not maintain arousal as high as
information processing related to comprehension and/or the NICs when tasks become redundant. Thus, the odd-left
no context is present for interpreting single words. Mem- task was presented first and was "something different," the
ory per se and attention-concentration do not appear to be odd-right task redundant, and the off-target task was again
related significantly to the verbal skill differences between "different." From the viewpoint of arousal, the NIAIs were
the two groups. less aroused for the odd-right than odd-left task while they
were rearoused for the off-target task. In contrast, the level
of arousal for the IAIs remained relatively low across tasks
Cognitive Psychophysiology and remained relatively high for the NICs for all three
Significant group differences in ERPs were present only tasks. These results suggest that arousability (Zuckerman
for the peak P300 amplitudes for the 0.2/0.1R stimulus 1983, 1991) varies among the three groups which could
conditions. The relationships between the P300 amplitudes lead to selective attention differences among the groups
for the NICs and IAIs were relatively constant across tasks although the WMS-R attention/concentration scores were
with higher amplitudes for the NICs suggesting a greater not significantly different between IAIs and NIAIs. Barratt
neural involvement in the processing of the 0.2/0.1R (1987) and Barratt and Patton (1983) reported that persons
stimuli. The peak P300 amplitudes for the 0.2/0.1R con- with high levels of impulsiveness have different threshold
ditions for the NIAIs varied across tasks with their ERPs levels for arousability. Persons with high levels of trait
showing higher amplitudes during the odd-left and off- impulsiveness in comparison to persons with average or
target task and lower amplitudes during the odd-right task. low levels often do not respond to less intense levels of
Contrary to predictions, there were no significant P300 stimulation or to changes in stimuli. This finding can be
peak amplitude differences between IAIs and NIAIs in misinterpreted as persons with high impulsivity not being
frontal electrodes. This result is consistent with a lack of a able to respond to changes in stimuli. As Barratt and
significant difference in impulsiveness between the two Patton (1983) report, persons with high impulsivity can
inmate groups since frontal lobe dysfunctioning has been respond to stimuli which are of interest to them. For
implicated in impulsiveness (Barratt 1987, 1993; Stein et example, male subjects with high levels of impulsiveness
al 1995). responded physiologically to sexually provocative pictures
There were group differences between the NIAIs and although they did not respond to tones or randomly
Impulsive Aggression BIOLPSYCHIATRY 1057
1997;41:1045-1061

designed visual shapes in a laboratory setting. Thus, were significantly related in general to the P300 ampli-
although the NIAIs and IAIs did not score significantly tudes across all tasks, however, the degree of relationship
different on self-report impulsiveness measures and on the between impulsiveness/verbal tests and midline P300
WMS-R attention/concentration subtest in this study, the amplitudes varied across tasks. During the odd-left and
NIAIs in comparison to IAIs may have a lower arousabil- off-target tasks, impulsiveness was more significantly
ity threshold for responding to new or different stimuli in related to frontal than parietal P300 amplitudes with no
terms of the global stimulus probability context described differences during the odd-right task. VIQ and GORT
by Johnson (1993). As noted, the IAIs can probably be were more highly related to parietal lobe than frontal
aroused to positively or negatively cathected stimuli which activity during the odd-left task but not during the other
would explain their getting agitated and becoming aggres- two tasks. To the extent that arousal is related to P300
sive in selected situations. Polich and Kok (1995) re- amplitudes as discussed earlier, these different patterns of
viewed the potential role of arousal as a "unifying theo- relationships of impulsiveness and verbal skills to P300
retical mechanism" in interpreting the bases of P300 amplitudes would suggest that at least two arousal (or
amplitude and latency differences. They noted that "it is arousability) systems are involved in determining these
not unreasonable to conclude that arousal level does affect relationships.
the P300 component in various task situations" (p 136). A number of authors have suggested that there are two
The topographical differences in P300 amplitudes be- or more arousal systems that influence cortical activity
tween IAIs and NIAIs involved primarily the posterior (Broadbent 1971; Eysenck 1982; Routtenberg 1968;
temporal lobe, the parietal lobes, the sensori-motor strip, Polich and Kok 1995; Zuckerman 1991), although the
and the striate cortex. These areas have all been implicated exact descriptions of the systems vary. Barratt (1985) and
in verbal information processing, especially in reading (Barratt and Patton (1983) have implicated the reticular
(Rosen et al 1993; Njiokiktjien 1993; Caplan 1993). The activating system as the basis for arousal related to
differences in P300 amplitudes in the parietal-occipital impulsiveness. This would be a general nonspecific type
areas would be especially important in decoding of verbal of arousal and that which Eysenck (1982) refers to as
symbols (LaBerge 1995) and dysfunctioning in these areas passive arousal or natural arousal in Polich and Kok's
would be consistent with the suggestion of dyslexia in (1995) classification. Verbal discrimination tasks are more
IAIs. These cortical areas have a rich multisensory input related to specific arousal systems (Eysenck 1982) or
from the pulvinar (LaBerge 1995) and could be related to environmental arousal (Polich and Kok 1995). Polich and
the grapheme-phoneme linkage problems often seen in Kok (1995) note that " . . . the connection between the
dyslexia (Njiokiktjien 1993). P300 component and its underlying neurophysiological
Temporal information processing involved in timing mechanism(s) is not yet clear, so that a direct association
and rhythm behaviors are also related to sensory-motor between arousal and ERPs may be difficult to demonstrate
and parietal-occipital cortical areas (Merzenich et al because the former typically reflects tonic changes while
1993; Brooks 1986). Reading has been related to temporal the latter are phasic in origin" (p 133). In the present study,
information processing (Tallal et al 1993) and Stanford it appears that different arousal systems are involved with
and Barratt (in press) have shown that impulsiveness, impulsiveness and verbal skills but the exact brain loci or
verbal skills, and continuous performance task measures processes for the two systems are not obvious from these
define a second-order temporal information processing data.
task. It is possible that a cluster of symptoms which One theoretical interpretation of P300 amplitude is that
include antisocial behaviors, impulsiveness, poor reading it relates to working or immediate memory in updating the
skills, and aggressive behaviors may be related in part to mental model of the stimulus environment (Polich and
brain dysfunctioning related to temporal information pro- Kok 1995; Donchin 1981). In the present study, verbal and
cessing. visual immediate memory performance were not compa-
rably related to midline P300 amplitudes. Verbal memory
had a higher average relationship with the P300 ampli-
Personality, Neuropsychology, and Cognitive tudes, especially during the odd-left and odd-right tasks. In
Psychophysiological Measures Interrelated contrast, visual memory had a higher relationship with
Impulsiveness and neuropsychology tests were inversely P300 amplitudes during the off-target task. These results
correlated (Table 7). Further, impulsiveness was inversely would indicate that immediate or working memory tasks
related to midline P300 amplitudes while the neuropsy- are not related to P300 amplitudes in the same way and
chology tests were positively related (Table 8). Both would suggest a modification of the memory/context
impulsiveness and verbal tasks (e.g., VIQ and GORT) updating theory.
1058 BIOLPSYCHIATRY E.S. Barratt et al
1997:41:1045-1061

Impulsive vs Non-impulsive Aggression plained by two alternative explanations based on brain


Both the IAIs and NIAIs had higher levels of anger and circuitry: 1) brain circuitry related to arousability to new
impulsiveness than NICs and satisfied the criteria for an situations is at a higher threshold for the IAIs than NIAIs
antisocial personality disorder. There were significant and arousal may affect cortical areas differently; 2) brain
circuitry involved with decoding verbal symbols (graph-
differences between the IAIs and NIAIs in verbal skills
eme/phoneme linkage) does not function as efficiently
consistent with developmental dyslexia (Duncan et al
among IAIs as NIAIs beyond arousal influences. Both
1994; Siegel 1994). The NIAIs and IAIs also differed in
inmate groups had poorer verbal skills and suggested
the functioning of selected central and posterior cortical
lower levels of arousability in frontal lobe functioning
areas with the IAIs having fewer neural resources involved
than the NICs that could have led to their becoming
in processing the visual stimuli which were presented at
criminals. The consistently lower threshold of arousability
different probability levels. The differences between the
along with developmental deficits in brain maturation in
NIAIs and IAIs in brain activity suggested that the NIAIs
more posterior cortical areas among the IAIs in compari-
responded to new or novel situations with arousal levels
son to NIAIs could have led to stressful life experiences
that were different than the IAIs. The NIAIs' ERPs
that fostered the development of impulsive aggression.
suggested arousability (or arousal) levels more sensitive to
From a developmental viewpoint, it is highly probable
new or different stimuli in comparison to IAIs but the
that the reading deficits present in the impulsive aggres-
ERPs for the NIAIs also varied across tasks with lower
sive inmates were there from the beginning of their school
amplitudes during a redundant task (odd-right). To the
experiences. The inability to easily discriminate visual
extent that arousal is a valid explanation for changes in the
information in social situations could have led to their
P300 amplitudes, it appears, as noted above, that different
being frustrated and to impulsive aggressive acts. This
cortical areas have different levels of arousability or there
conclusion would be consistent with the studies of Dodge
are different arousal systems involved in the functioning and colleagues who showed that aggressive children when
of frontal and parietal lobes related to impulsiveness and compared to nonaggressive children used fewer environ-
verbal skills. Impulsiveness and verbal skills are inversely mental cues to mediate behavior (Dodge and Newman
related as reported by Stanford and Barratt (in press) and, 1981), interpreted the behavior of peers as more hostile
further, both of these measures are significantly related to (Dodge 1980), and were less capable at generating poten-
P300 amplitudes, although in opposite directions and with tial responses to conflict situations (Richard and Dodge
different topographical patterns in the relationships. The 1982).
data suggest that "arousability" related to performance on Verbal skill deficits could influence impulsive aggres-
sequential tasks may vary in cortical areas as a function of sive behavior psychosocially in other ways. For example,
task demands. Further research is needed to clarify the the incorporation of mores and laws into memory to guide
effects of arousal or arousability on cortical functioning social behavior could be impaired (Chandler and Moran
related to the role of impulsiveness/verbal skills in impul- 1990). Further, covert "stop-orders" for overt behavior are
sive aggression. often verbal and a deficit in verbal skills could limit this
Both inmate groups in comparison to NICs had higher inhibitory process (Camp 1977; Camp et al 1977; Berk
levels of anger and impulsiveness, poorer performance on and Potts 1991). The two inmate groups did not differ in
most neuropsychology tests including the verbal tests, and general in their ability to reason but the IAIs showed a
less neural involvement in frontal cortical areas as well as decreased ability in information processing related to
selected other areas. These results are typical of past decoding words. Learned aggressive behaviors could re-
research on the etiology of becoming a criminal (Raine sult from the frustrations related to these verbal skill
1993). Impulsive aggression is related to neuropsycholog- deficits in everyday situations, especially in early school
ical and cognitive psychophysiological measures of infor- experiences (Hirschi 1969; Berkowitz 1974).
mation processing beyond those which are related to In summary, this study has shown that aggression is not
becoming a criminal. The inability to decode verbal a homogeneous construct, even among antisocial persons.
symbols well in reading (dyslexia) may be related to Impulsiveness and verbal skills were inversely correlated.
differences in the functioning of selected cortical areas Both inmate groups scored significantly higher than the
that in combination with frontal lobe differences are noninmate controls on impulsiveness. Verbal skills but not
related to being impulsively aggressive. Frontal lobe impulsiveness were significantly different between the
dysfunctioning and high levels of anger and impulsiveness aggressive and nonimpulsive aggressive inmates. Differ-
may be necessary for impulsive aggression but are not ent arousal systems appear to be related to impulsiveness
sufficient. and verbal skill measures. Developmental deficits in
The differences between NIAIs and IAIs can be ex- cortical areas involved with the mechanics of reading (e.g.,
Impulsive Aggression BIOL PSYCHIATRY 1059
1997;41:1045-1061

grapheme/phoneme linkage) could result in stressors (e.g., impulsively aggressive. Impulsiveness per se appears to be
poor academic performance) that influence everyday life necessary but not sufficient to cause impulsive aggression.
adjustments. Thus, it is concluded that the developmental
deficits in selected verbal skills interacting with differ- This research was supported by the Health Foundation, The Rogosin
ences in arousability thresholds result in a tendency to be Institute, New York Hospital--Cornell Medical Center.

References
American Psychiatric Association (1994): Diagnostic and Sta- Berkowitz L (1974): Some determinants of impulsive aggres-
tistical Manual of Mental Disorders: Fourth edition. Wash- sion: Role of mediated associations with reinforcements for
ington DC: American Psychiatric Press. aggression. Psychol Rev 81 : 165-176,
Atkins MS, Stoff DM, Osborne ML, Brown K (1993): Distin- Branchey MH, Buydens-Branchey L, Lieber CS (1988): P3 in
guishing instrumental and hostile aggression: Does it make a alcoholics with disordered regulation of aggression. Psychi-
difference? J Abnorm Child Psychol 21:355-365. atry Res 25:49-58.
Bandura A (1973): Aggression: A social learning analysis. Broadbent DE (1971): Decision and Stress. London: Academic
Englewood Cliffs, NJ: Prentice Hall. Press.
Barratt ES (1983): The biological basis of impulsiveness: The Brooks VB (1986): The Neural Basis of Motor Control. New
significance of timing and rhythm disorders. PersonaliO" and York: Oxford University Press.
Individual Differences 4:387-391. Brown GL, Goodwin FK (1984): Diagnostic, clinical and per-
Barratt ES, Patton J (1983): Impulsivity: Cognitive, behavioral sonality characteristics of aggressive men with low 5-HI AA.
and psychophysiological correlates. In Zuckerman M (ed), Clin Neuropharrnacol 7:$408-$409.
Biological Bases of Sensation Seeking, Impulsivity and Anx- Buss AH, Durkee A (1957): An inventory for assessing different
iety. Hillsdale, NJ: Lawrence Earlbaum Associates, pp 77- kinds of hostility. J Consult Clin Psychol 21:343-349.
122,
Camp BW (1977): Verbal mediation in young aggressive boys. J
Barratt ES (1985): Impulsiveness subtraits: Arousal and infor- Abnorm Psychol 86:145-153.
mation processing. In Spence JT, Izard CE (eds), Motivation,
Emotion and Personality. New York: Elsevier Science Pub- Camp BW, Zimet SG, van Doorninck WJ, Dahlem NW (1977):
lishers, pp 137-146. Verbal abilities among young aggressive boys. J Educational
Psychol 69:129-135.
Barratt ES (1987): Impulsiveness and anxiety: Information pro-
cessing and electroencephalographic topography. J Res Per- Caplan D (1993): Language: Structure, Processing, and Disor-
sonality 21:453-463. ders. Cambridge: The MIT Press.
Barratt ES (1991): Measuring and predicting aggression within Caprara GV, Cinanni V, D'Imperio G, Passerini S, Renzi P,
the context of a personality theory. J Neuropsychitao, Clin Travaglia G (1985): Indicators of impulsive aggression:
Neurosci 3:$35-$39. Present status of research on irritability and emotional sus-
ceptibility scales. Personality and Individual Differences
Barratt ES (1993): lmpulsivity: Integrating cognitive, behavioral, 6:665-675.
biological, and environmental data. In McCown WG, Johnson
JL, Shure MB (eds), The Impulsive Client: Theory,, Research, Chandler M, Moran T (1990): Psychopathy and moral develop-
and Treatment. Washington DC: American Psychological ment: A comparative study of delinquent and nondelinquent
Association, pp 39-56. youths. Dev Psychopathol 2:227-246.
Barratt ES (1994): Impulsiveness and aggression. In Monahan J, Cloninger CR (1987): A systematic method for clinical descrip-
Steadman HJ (eds), Violence and Mental Disorder: Develop- tion and classification of personality variants: A proposal.
ments in Risk Assessment. Chicago: University of Chicago Arch Gen Psychiatry 44:573-588.
Press, pp 61-79. Coccaro EF, Siever LJ, Klar HM, Maurer G, Cochrane K,
Barratt ES, Kent TA, Stanford MS (1995a): The role of biolog- Cooper TB, Mobs RC, Davis KL ( 1989): Serotonergic studies
ical variables in defining and measuring personality. In Ratey in patients with affective and personality disorders: Correlates
J (ed), Neuropsychiatry of Personality Disorders. Oxford: with suicidal and impulsive aggressive behavior. Arch Gen
Blackwell Science. Psychiatry 46:587-599.
Barratt ES, Kent TA, Tourian KA (1995b): Impulsive aggres- Costa PT, McCrae RR (1985): The NEO Personality Inventory:
sion: Symptom or syndrome? Paper presented at the Annual Manual. Odessa, FL: Psychological Assessment Resources.
Meeting, American Psychiatric Association, Miami. DeJong J, Virkkunen M, Linnoila M (1992): Factors associated
Bech P, Mak M (1995): Measurements of impulsivity and with recidivism in a criminal population. J Nerv Ment Dis
aggression. In Hollander E, Stein DK (eds), Impulsivity and 180:543-550.
Aggression. New York: John Wiley and Sons, Ltd, pp 25-41. Dickman SJ (1993): Impulsivity and information processing. In
Berk LE, Potts MK (1991): Development and functional signif- McGown WG, Johnson JL (eds), The Impulsive Client.
icance of private speech among attentional deficit hyperac- Washington DC: The American Psychological Association.
tivity disordered and normal boys. J Abnorm Child Psychol Dodge KA (1980): Social cognition and children's aggressive
19:357-377. behavior. Child Dev 51 : 162-170.
1060 BIOL PSYCHIATRY E.S. Barratt etal
1997;41:1045-1061

Dodge K.A, Newman JP (1981): Biased decision-making pro- Kirk RE (1990): Statistics: An Introduction (Third Edition).
cesses in aggressive boys. J Abnorm Psychol 90:375-379. Chicago: Holt, Rinehart and Winston, Inc.
Donchin E (1979): Event-related potentials: A tool in the study LaBerge D (1995): Computational and anatomical models of
of human information processing. In Begleiter H (ed), Evoked selective attention in object identification. In Gazzaniga MS
Brain Potentials and Behavior. New York: Plenum Press, pp (ed), The Cognitive Neurosciences. Cambridge: MIT Press.
13-88. Lewis DO (1992): From abuse to violence: Psychophysiological
Donchin E (1981): Surprise!...Surprise! Psychophysiology 18: consequences of maltreatment. J Am Acad Child and Adolesc
493-513. Psychiatry. 31:383-391.
Drake ME, Jr, Pakalnis A, Brown ME, Hietter SA (1988): Lezak MD (1983): Neuropsychological Assessment, 2nd ed. New
Auditory event related potentials in violent and nonviolent York: Oxford University Press.
prisoners. Eur Arch Psychiatry Clin Neurosci 238:7-10. Linnoila M, Virkkunen M, Scheinin M, Nuutila A, Rimon R,
Duncan CC, Rumsey JM, Wilkness SM, Denckla MB, Ham- Goodwin FK (1983): Low cerebrospinal fluid 5-Hydroxyin-
burger SD, Odou-Potkin M (1994): Developmental dyslexia dolacetic acid concentration differentiates impulsive from
and attention dysfunction in adults: Brain potential indices of nonimpulsive violent behavior. Life Sci 33:2609-2614.
information processing. Psychophysiol 31:386-401. Lion JR (1991): Pitfalls in the assessment and measurement of
Eichelman B (1992): Aggressive behavior: From laboratory to violence: A clinical view. J Neuropsychiatry 3:$40-$43.
clinic (quo vadit). Arch Gen Psychiatry 49:488-492. Lish JD, Kavoussi RJ, Coccaro EF (1996): Aggressiveness. In
Elliot FA (1976): Neurological factors in violent behavior (the Costello CG (ed), Personality Characteristics of the Person-
dyscontrol syndrome). Bull Am Acad Psychiatry Law 4:297- ality Disordered. New York: John Wiley and Sons, Inc. pp
315. 2 4 - 47.
Elliot FA (1992): Violence: The neurologic contribution. Arch Marr D (1982): Vision. San Francisco: Freeman.
Neurol 49:595-603.
McCarthy G, Wood CC (1985): Scalp distributions of event-
Eysenck HJ, Eysenck M (1985): Personality and Individual related potentials: An ambiguity associated with analysis of
Differences. New York: Plenum Press. variance models. Electroencephalogr Clin Neurophysiol 62:
Eysenck M (1982): Attention and Arousal: Cognition and Per- 203-208.
formance. Berlin: Springer-Verlag. Merzenich MM, Schreiner C, Jenkins W, Xiaoqin W (1993):
Gaddes WH ( 1980): Learning Disabilities and Brain Function: A Neural mechanisms underlying temporal integration, segmen-
Neuropsychological Approach. New York: Springer-Verlag. tation, and input sequence representation: Some implications
Gray D (1963): Gray Oral Reading Test: Manual. New York: for the origin of learning disabilities. In Tallal P, Galaburda
Bobbs Merrill. AM, Llinas RR, von Euler C (eds), Temporal Information
Processing in the Nervous System: Special Reference to
Harris RJ (1994): ANOVA. Itasca, IL: F.E. Peacock Publishers. Dyslexia and Dysphasia. New York: New York Academy of
Heilbrun AB, Knopf IJ, Bruner P (1976): Criminal impulsivity Sciences, pp 1-22.
and violence and subsequent parole outcome. Br J Criminol Mirsky AF, Siegel A (1994): The neurobiology of violence and
16:367-377. aggression. In Reiss AJ, Jr, Miczek KA, Roth JA (eds),
Heilbrun AB, Heilbrun LC, Heilbrun KL (1978): Impulsive and Violence: Biobehavioral Influences, Vol 2. Washington DC:
premeditated homicide: An analysis of subsequent parole risk National Academy Press, pp 59-172.
of the murderer. J Criminal Law Criminol 69:108-114. Mungas D (1988): Psychometric correlates of episodic violent
Heilbrun AB (1972): Self regulatory behavior in skid row behavior: A multidimensional neuropsychological approach.
alcoholics. Quarterly Journal of Studies on Alcohol 33:990- Br J Psychiatry 152:180-187.
998. Nestor PG (1992): Neuropsychological and clinical correlates of
Hillyard SA, Kutas M (1983): Electrophysiology of cognitive murder and other forms of extreme violence in a forensic
processing. Annu Rev Psychol 34:33-61. psychiatric population. J Nervous Mental Dis 180:418 - 423.
Hirschi T (1969): Causes of Delinquency. Berkeley: University Neurosoft Inc. (1993): SCAN: Manual. NEUROSCAN, INC.
of California Press. Herndon, VA.
Hollander E, Stein DJ (1995): lmpulsivi~ and Aggression. New Njiokiktjien C (1993): Neurological arguments for a joint devel-
York: John Wiley and Sons. opmental dysphasia-dyslexia syndrome. In Galaburda AM
Johnson R, Jr (1986): A triarchic model of P300 amplitude. (ed), Dyslexia and Development. Cambridge: Harvard Uni-
Psychophysiology 23:367-384. versity Press.
Johnson R, Jr (1989): Auditory and visual P300s in temporal- Novaco RW (1994): Anger as a risk factor for violence among
lobectomy patients: Evidence for modality-dependent gener- the mentally disordered. In Monahan J, Steadman, JH (eds),
ators. Psychophysiology 26:633-650. Violence and Mental Disorder. Chicago: University of Chi-
Johnson R, Jr (1993): On the neural generators of the P300 cago Press, pp 21-59.
component of the event-related potential. Psychophysiology Othmer E, Penick EC, Powell B J, Read MR, Othmer SC (1989):
30:90 -97. Psychiatric Diagnostic Interview--Revised: Manual. Los An-
Kirk RE (1982): Experimental Design: Procedures for the geles: Western Psychological Services.
Behavioral Sciences (2nd Edition). Belmont, CA: Brooks/ Patton JH, Stanford MS, Barratt ES (1995): Factor structure of
Cole Publishing. the Barratt impulsiveness scale. J Clin Psychol 51:768-774.
Impulsive Aggression BIOL PSYCHIATRY 1061
1997;41:1045-1061

Polich J, Geisler MW (1991): P300 seasonal variation. Biol ogy and Cognitive Rehabilitation. Hillsdale, NJ: Lawrence
Psychol 32:173-179. Earlbaum Associates, pp 379-400.
Polich J, Kok A ( 1995): Cognitive and biological determinants of Spielberger CD (1988): State-Trait-Anger Expression Inventory
P300: An integrative review. Biol Psychol 41:103-146. (Research Manual Edition) Odessa, FL: Psychological As-
Pritchard W (1989): P300 and EPQ/STPI personality traits. sessment Resources.
Personality and Individual Differences 10:15-24. Stanford MS, Barratt ES (1996): Verbal skills, finger tapping,
Raine A, Scerbo A (1991): Biological theories of violence. In and cognitive tempo define a second-order factor of temporal
Milner JS (ed), Neuropsychology of Aggression. Boston: information processing. Brain Cogn 31:35-45.
Kluwer Academic Publishers, pp 1-25. Stein DJ (1994): Is impulsive aggression a disorder of the
Raine A (1993): The Psychopathology of Crime: Criminal individual or a social ill? A matter of metaphor. J Biol
Behavior as a Clinical Disorder. New York: Academic Press. Psychiatry 36:353-355.
Reiss AJ, Jr, Miczek KA, Roth JA (eds) (1994): Understanding Stein DJ, Towey J, Hollander E (1995): The neuropsychiatry of
and Preventing Violence, Vol 2. Washington DC: National impulsive aggression. In Hollander E, Stein DJ (eds), lmpul-
Academy Press. sivity and Aggression. New York: John Wiley and Sons, pp
91-105.
Revelle W, Humphreys MS, Simon L, Gilliland K (1980): The
interactive effect of personality, time of day, and caffeine: A Tallal P, Galaburda AM, Llin~s RR, Von Euler C (1993):
test of the arousal model. J Experimental Psychoh General Temporal Information Processing in the Nervous System.
109:1-31. New York: New York Academy of Sciences.
Revelle W, Anderson KJ, Humphreys MS (1987): Empirical Tangney JP, Wagner P, Fletcher C, Gramzow R (1992): Shamed
tests and theoretical extensions of arousal based theories of into anger? The relation of shame and guilt to anger and self
personality. In Strelau J, Eysenck HJ (eds), Personali~, reported aggression. Personality and Social Psychology 62:
Dimensions and Arousal. London: Plenum Press, pp 17-36. 669-675.
Richard BA, Dodge KA (1982): Social maladjustment and Taylor SL, O'Neal EC, Langley T, Butcher AH (1991): Anger
problem solving in school-aged children. J Consulting Clin- arousal, deindividuation, and aggression. Aggressive Behav
ical Psychol 50:226-233. 17:193-206.
Robins E, Guze SB (1970): Establishment of diagnostic validity Tedeschi JT, Felson RB (1994): Violence, Aggression, and
in psychiatric illness: Its application to schizophrenia. Am Coercive Actions. Washington DC: American Psychological
J Psychiatry 126:107-111. Association.
Rolls ET (1995): A theory of emotion and consciousness, and its Virkkunen M, DeJong J, Bartko J, Goodwin FK, Linnoila M
application to understanding the neural basis of emotion. In (1989): Relationship of psychobiological variables to recidi-
Gazzaniga MS (ed), The Cognitive Neurosciences. Cam- vism in violent offenders and impulsive fire setters: A
bridge: The MIT Press, pp 1091-1106. follow-up study. Arch Gen Psychiatry 46:600-603.
Rosen GD, Sherman GF, Galaburda AM (1993): Dyslexia and Vitiello B, Behar D, Hunt J, Stoff D, Ricciuti A (1990):
brain pathology: Experimental animal models. In Galaburda Subtyping aggression in children and adolescents. J Neuro-
AM (ed), Dyslexia and Development. Cambridge: Harvard psychiatry Clin Neurosci 2:189-192.
University Press, pp 89-111. Volavka J, Martell D, Convit A (1992): Psychobiology of the
Routtenberg A (1968): The two arousal hypothesis: Reticular violent offender. J Forensic Sci 37:237-251.
formation and limbic system. Psychol Rev 75:51- 81. Wechsler D (1955): Wechsler Adult Intelligence Scale: Manual.
Scheffers MK, Johnson R, Jr (1994): Recognition memory and New York: The Psychological Corporation.
search for attended letters: An event-related potential analy- Wechsler D (1987): Wechsler Memory Scale---Revised: Manual.
sis. J Psychophysiol 8:328-347. New York: The Psychological Corporation.
Semlitsch HV, Anderer P, Schuster P, Presslich O (1986): A Weiger WA, Baer DM (1988): An approach to the neurology of
solution for reliable and valid reduction of ocular artifacts, aggression. J Psychiatr Res 22:85-98.
applied to the P300 ERP. Psychophysiology 23:695-703. Wong MTH, Lumsden J, Fenton GW, Fenwick PBC (1994):
Shoham SG, Askenasy JM, Rahav G, Chard F, Addi A (1989): Electroencephalography, computed tomography and violence
Personality correlates of violent prisoners. Personality and ratings of male patients in a maximum-security mental
Individual Differences 10:137-145. hospital. Acta Psychiatr Scand 90:97-101.
Siegel S, Castellan NJ (1988): Nonparametric Statistics for the Zuckerman M (1983): Comments. In Zuckerman M (ed), Bio-
Behavioral Sciences, 2nd ed. New York: McGraw-Hill. logical Bases of Sensation Seeking, Impulsivity and Anxiety.
Siegel LS (1994): Phonological processing deficits as the basis of Hillsdale, NJ: Lawrence Earlbaum Associates, pp 117-112.
developmental dyslexia: Implications for remediation. In Zuckerman M ( 1991): Psychobiology of Personality. New York:
Riddoch M J, Humphreys GW (eds), Cognitive Neuropsychol- Cambridge University Press.

You might also like