Brain Abnormalities in Murderers Indicated by

Positron Emission Tomography

Adrian Raine, Monte Buchsbaum, and Lori LaCasse

Murderers pleading not guilty by reason of insanity (NGRI) are thought to have brain
dysfunction, but there have been no previous studies reporting direct measures of both cortical
and subcortical brain functioning in this specific group. Positron emission tomography brain
imaging using a continuous performance challenge task was conducted on 41 murderer's
pleading not guilty by reason of insanity and 41 age- and sex-matched controls. Murderer's
were characterized by reduced glucose metabolism in the prefrontal cortex, superior parietal
gyrus, left angular gyrus, and the corpus callosum, while abnormal asymmetries of activity
(left hemisphere lower than right) were also found in the amygdala, thalamus, and medial
temporal lobe. These preliminary ftndings provide initial indications of a network of abnormal
cortical and subcortical brain processes that may predispose to violence in murderer's
pleading NGRI. © 1997 Society of Biological Psychiatry

Key Words: Violence, murder, positron emission tomography, prefrontal, amygdala, hip-
pocampus, thalamus, corpus callosum, angular gyrus, parietal, occipital

BIOL PSYCHIATRY 1997;42:495--508

Introduction possible to localize which brain areas in particular may be

dysfunctional in violent offenders.
It has long been suspected that generalized brain dysfunc- Clues do however exist with respect to the source of
tion may predispose to violent behavior. Studies using brain dysfunction predisposing to violence. It has long
electroencephalographic (EEG), neurological, neuropsy- been thought that dysfunction of the prefrontal cortex may
chological, and cognitive test techniques have repeatedly disrupt the regulation of aggression, and this notion has
shown that violent offenders have poorer brain functioning been supported by neurological studies of patients with
than normal controls (Eichelman 1993; Eysenck and damage to the prefrontal cortex (Damasio et al 1990;
Gudjonsson 1989; Elliott 1987; Lewis et al 1988; Moffitt Weiger and Bear 1988). Some neuropsychological and
1988; Raine 1993), but until recently it has not been psychophysiological studies on violent and forensic pop-
ulations have shown abnormalities in hemispheric asym-
metries of function (Convit et al 1991; Hare and McPher-
From the Department of Psychology, University of Southern California, Los son 1984; Raine et al 1990a) and reduced EEG
Angeles, California (AR, LL); and Department of Psychiatry, Mount Sinai
School of Medicine, New York, New York (MB). interhemispheric coherence (Flor-Henry et al 1991), which
Address reprint requests to Adrian Raine, Department of Psychology, S.G.M. may be linked to dysfunction of the corpus callosum
Building, University of Southern California, Los Angeles, CA 90089-1061.
Received November 2, 1995; revised July 22, 1996. (Nachshon 1983; Yeudall 1977), but this hypothesis has

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

PII S0006-3223(96)00362-9
496 BIOL PSYCHIATRY A. Raine et al
1997;42:495-508

not been tested using direct measures of callosal function- Methods

ing. Recent event-related potential mapping techniques
have implicated dysfunction in the left angular gyrus in
Subjects
violent offenders as indicated by reduced slow-wave MURDERERS. The experimental group consisted of 41
amplitudes (Barratt et al in press). Experimental animal subjects tried in the state of California (39 men, 2 women)
research together with neurological studies of patients with a mean age of 34.3 years (SD = 10.1) who had been
have further implicated limbic structures such as the charged with either murder or manslaughter (labeled
amygdala and hippocampus in modulating aggression below as "murderers" for ease of reference). Subjects were
(Bear 1991; Elliott 1992; Gorenstein and Newman 1980; referred to the University of California, h'vine (UCI)
Mirsky and Siegel 1994; Watson et al 1983a), while the imaging center to obtain evidence relating to a NGRI
thalamus also provides an important afferent source of the defense or to capability of understanding the judicial
hypothalamic-induced attack in cats (Mirsky and Siegel process (incompetence to stand trial), while some who had
1994). Nevertheless, such research on animals and humans been found guilty were referred to obtain information for
who have suffered brain insults, although of key impor- diminished capacity as an ameliorating circumstance in
tance, is one step removed from the question of whether the sentencing phase of the trial. Reasons for referral were
severely violent offenders have brain dysfunction local- very diverse and included schizophrenia (6 cases), history
ized to specific brain areas. of head injury or organic brain damage (23), history of
The advent of brain imaging research has recently made psychoactive substance abuse (3), affective disorder (2),
it possible for the first time to directly assess brain epilepsy (2), history of hyperactivity and learning disabil-
functioning in violent individuals. Initial research in this ity (3), and passive-aggressive or paranoid personality
area has again implicated frontal brain regions in addition disorder (2). In 7 of the above cases, there were also
to the temporal cortex (Goyer et al 1994; Volkow and unusual circumstances surrounding the crime that addi-
Tancredi 1987). These important initial studies support the tionally lead to the suspicion of some mental impairment.
notion of localized brain dysfunction in aggressive pa- Offenders were not receiving regulated psychoactive med-
tients, although inevitable limitations of such initial re- ication at the time of positron emission tomography (PET)
search include small sample sizes in hospitalized patients, scans, and were instructed to be medication-free for the
and a focus on aggressive personality as opposed to 2-week period preceding brain scanning. All subjects were
seriously violent behavior. in custody during this period, and penal authorities agreed
One particularly important group of violent offenders to refrain from administering medication. Urine screens at
in forensic psychiatry consists of those who commit the time of PET scanning were negative for every mur-
murder and plead not guilty by reason of insanity derer referred for study.
(NGRI). Although it is thought that such individuals
have localized brain impairments, there has been no CONTROLS. A control group was formed by matching
previous brain imaging research on this important each murderer with a normal subject of the sanae sex and
population to support or refute this notion. In a prelim- age who had been tested using identical PET imaging
inary report on a pilot sample of 22 such offenders procedures in the same laboratory. Six murderers (all men)
compared to 22 normals, we provided some initial had been diagnosed as schizophrenic by psychiatrists.
support for the notion of prefrontal dysfunction in this These 6 were individually matched on age and sex with 6
group (Raine et al 1994). In the present study the schizophrenics from a larger psychiatric sample tested
sample size is extended to 41 murderers and 41 con- under identical procedures at the Brain Imaging Center at
trois, and analysis of subcortical structures is now the University of California, Irvine (Buchsbaum et al
undertaken. To our knowledge, this is the largest 1990). The resulting 41 controls (39 men, 2 women) had a
sample of violent offenders assessed on functional brain mean age of 31.7 years (SD = 10.3), which did not differ
imaging. It is hypothesized that these seriously violent from murderers (p > .26). Normal controls had been
individuals have relatively localized brain dysfunction screened for health by physical exam, medical history, and
in the prefrontal cortex, angular gyrus, amygdala, hip- a psychiatric interview. No subject was taking any medi-
pocampus, thalamus, and the corpus callosum, brain cation, had a history of psychiatric illness, in self or
areas previously linked empirically or conceptually to first-degree relatives, or had current significant medical
violence. Conversely, no dysfunction is expected in illness. Subjects with a history of seizure disorder, head
other brain areas (caudate, putamen, globus pallidus, trauma, or substance abuse were excluded. Subjects par-
midbrain, cerebellum), which have been implicated in ticipated under protocols and consent forms approved by
other psychiatric conditions but which have not been the Human Subjects Committee of University of Califor-
related to violence. nia, Irvine.
Brain Abnormalities in Murderers BIOLPSYCHIATRY 497
1997;42:495-508

P E T Task Procedure Precentrol

I Postcentral
Full details of general PET scanning procedures and amarginal
Middle f ro~t~,,~.~ erior parietal Iobule
quantification may be found in Buchsbaum et al (1990).
Superior ~iiiiii~i~ r / :::::::::::::::::::::: ~.ngular gyrus
Briefly, the fluorodeoxyglucose (FDG) tracer was injected frontal \ ~ i i i : : i i : : : : ~ I lii!!::::~!~ii.::'~Jf/lll//-----"Jl~l lateral occipital
into the subject in the test room and taken up by the brain i~ F..iiiii~i~g,7//A:,.~!I ================================= ~J----18
as a tracer of brain metabolic rate for a 32-min period I I::::::::::::::::::::::::: ' lii::~:,i::::::ili::::~::~ii~i[¢"/A[ "~d__19
)'liii::::iY - l~'i::ii::i::iiiiiiN ~,1d--17
during which the subject completed the continuous per- i i ~ .J~Ti~/i!iiiiiiiiiiiii~ ~1~ 17
formance task (CPT; Nuechterlein et al 1983). A degraded iii!~...r~///~
kl[i::iiiiw/////~r liiii::::::iiii::::~V////////J[ - ~
stimulus version of the CPT was employed as the frontal
challenge task because it has been shown to produce
Inferior frontal
increases in relative glucose metabolic rates in the frontal ~ferior temporal
lobes in normal controls, in addition to increases in fight Superior temporal Posteri,or temporal
temporal and parietal lobes (Buchsbaum et al 1990). The
Figure 1. Lateral view of 10 stacked slices showing surface
key signal detection performance measure of d' reflects superior, middle, and inferior cortical prefrontal areas, precentral
target recognition accuracy across the 32-min period frontal cortex, and temporal, parietal, and occipital areas from
(Davies and Parasuraman 1982; Nuechterlein 1991). Split- cortical peel analysis. The top slice corresponds to slice #2, or
half reliability for the task is high (r = .843, p < .001). 80% of head height in the brain atlas of Matsui and Hirano
Full procedural details are reported in Buchsbaum et al (1978).
(1990).
Ten minutes before the FDG injection, subjects were values (averaged across slices) for each hemisphere
given practice trials on the CPT. Thirty seconds before were extracted: superior frontal gyrus, middle frontal
injection, the task was started so that initial task novelty gyms, and inferior frontal gyms (see Figure 1).
would not be FDG labeled. After 32 min of FDG uptake, Bilateral temporal (superior, middle, inferior, and
the subject was transferred to the adjacent PET scanner posterior), parietal (postcentral, supramarginal, supe-
room. An individually molded, thermosetting plastic head rior parietal lobule, and angular gyrus), and occipital
holder was used to hold the head still during the scan. Ten (area 19, area 17 superior, area 17 inferior, and area
slices at 10-mm intervals parallel to the canthomeatal line 18) measures averaged across slices were also taken
were obtained. Scans started at the level of 80% of head (see Figure 1).
height above the canthomeatal line (vertex to can- Box Technique (medial areas). Medial cortical and
thomeatal line, usually 12-14 cm) and step downward at subcortical regions of interest were located on PET
10-mm intervals. slices by reference to stereotaxic coordinates as
Brain regions were identified using two techniques as detailed in Buchsbaum et al (1989). A 3 × 3 pixel
follows: region of interest box was placed on cortical and
subcortical structures at each level, according to a
Cortical Peel Technique (lateral areas). Surface cortical
standard list (see Figure 2). As each pixel measured
regions of interest were measured using a modifica-
2 × 2 ram, the size of the region of interest box was
tion of the original cortical peel technique (Buchs-
approximately one full-width half-maximum. Pre-
baum et al 1990) with the four lobes and four
frontal measures extracted from each slice level
anatomical subdivisions of each identified stereotac-
(given as a percentage of the distance from the
tically (Buchsbaum et al 1989). This technique has
external auditory meatus to the top of the head)
been used by at least nine different PET groups, and
according to a brain atlas (Matsui and Hirano 1978)
a review of its advantages for facilitating intrasubject
(see Figure 2) were as follows: superior frontal gyrus
and intersubject differences may be found in Harris et
(average of 80%, 74%, 68%, and 61% slice levels as
al (1991). Absolute glucose values for each region of
shown in Figure 2), anterior medial frontal gyrus
interest were expressed as a measure relative to all
(68% level), medial frontal gyms (average of 61%,
other regions contained in that slice. Relative rather
54%, and 47% levels), and orbital gyrus (21% level).
than absolute metabolic rates were used because
relative rates are more widely reported, have the To assess stereotaxic error due to individual differences
advantages of removing whole brain metabolic rate, in structure location within the plane, we evaluated the
are more likely to be related to function in specific stereotaxic frame based on the brain outline. Stereotaxic
neuroanatomical systems (Fox and Mintum 1989), error could place boxes in the caudate into the ventricle,
and show greater reliability within subjects over time thereby diluting metabolic rates with cerebrospinal zero
(Bartlett et al 1991). The following three prefrontal rates, but confidence limits based on application of the
498 BIOLPSYCHIATRY A. Raine et al
1997;42:495-508

MFG
F.W.

PreC. "~ ¢'~¢J74"/. "4t~=r,7~ 68% "V,.J v 61% ~ v 54%

"~¢~)~ FLW~
O~ ~d~ FLWM-~. ~ [] "~'~ FLWM~ T u . ~ , ~ ,RG

\% ~Zll ~ ,~=~ A. Calc'~"-- :~ ~[[ ~ Fus. G ~ 4 ~ ~,.~/ "~"~2~%Y AM"

Figure 2. Transverse view of the 10 slices showing medial cortical prefrontal structures used in box analysis. Percentages refer to
percent of head height above the canthomeatal line. Key to abbreviations: A. Calc = anterior calcarine gyms, A. Cer = anterior
cerebellum, ACG = anterior cingulate gyms, A. Corp = anterior corpus callosum, AMFG = anterior medial frontal gyrus, Amyg =
amygdala, ARG = anterior rectal gyms, A. Thai = anterior thalamus, CG = cingulate gyms, CN = caudate nucleus, FLWM = frontal
lobe white matter, Fus. G = fuisform gyms, GP = globus pallidus, Hipp = hippocampus, I. Coil = inferior colliculus, L. Thal= lateral
thalamus, MCG = middle cingulate gyms, M. Corp = middle corpus callosum, MFG = medial frontal gyms, Midb -- midbrain, MTG
= medial temporal gyms, M. Thai = medial thalmus, OG = orbital gyms, OR = optic radiation, Para = paracentral lobule, P. Calc
= posterior calcarine gyms, P. Cer = posterior cerebellum, PCG = posterior cingulate gyms, P. Corp = posterior corpus callosum,
PMFG = posterior medial frontal gyms, P. Put = posterior putamen, P. Thai = posterior thalamus, Prec = precuneus, PRG = posterior
rectal gyms, Put = putamen, S. C o l l = superior colliculus, SFG = superior frontal gyms, Unc = uncus.

current system to magnetic resonance images confirm comparisons (t tests) are two tailed. Means and SDs for all
2-SD limits within the caudate (Buchsbaum et al 1992). brain areas are shown in Table 1.
Subcortical regions theorized to relate to violence were
as follows: corpus callosum (47% level, see Figure 2);
Cortical Regions
medial temporal lobe, including the hippocampus (average
of 34%, 28%, and 21%), amygdala (21%); thalamus PREFRONTAL. AS anticipated on the basis of the pre-
(41%); and cingulate (61%, 54%, 41%, and 34%). Sub- vious pilot data, the expanded group of 41 murderers had
cortical regions theorized not to be related to violence lower glucose metabolism relative to controls in both
were extracted as follows: caudate (average of 41% and lateral and medial prefrontal cortical areas (,;ee Table 1
34%), putamen (average of 41%, 34%, and 28%), globus and Figure 3). We repeated exactly the same analyses that
pallidus (34%), midbrain (21%), and cerebellum (21%). we had previously conducted on a smaller sample (Raine
et al 1994) and found from two separate group x hemi-
sphere M A N O V A s a main effect for both lateral IF(l,80)
Results = 5.6, p < .02] and medial IF(I,80) = 6.2, p < .02]
For both cortical and subcortical analyses, values were prefrontal areas, with no interactions for hemisphere (p >
averaged across slices and two-way group (murderers and .75).
controls) x hemisphere (left and right) repeated-measures A more detailed breakdown of prefrontal subregions
multivariate analyses of variance using the M A N O V A indicated that murderers had significantly lower glucose
approach (Vasey and Thayer 1987) were conducted. For metabolism for left and right medial superior frontal
some brain areas gyrus was added as a third factor in a cortex (t = 2.6, p < .02), left anterior medial cortex (t =
three-way MANOVA. All tests of significance for planned 3.1, p < .003), right orbitofrontal cortex (t = 2.1, p < .04),
Brain Abnormalities in Murderers BIOL PSYCHIATRY 499
1997;42:495-508

Table 1. Group Means and Standard Deviations (in Parentheses) for Murderers and Controls for Cortical and Subcortical Relative
Glucose Metabolism

Left hemisphere Right hemisphere

Control Murderer Control Murderer
Cortical
Lateral prefrontal" 1.12 1.09 1.14 1.11
(0.05) (0.06) (0.05) (0.06)
Medial prefrontal" 1.25 1.20 1.22 1.17
(0.09) (0.11) (0. I0) (0.12)
Parietal" 1.15 1.10 1.17 1.13
(0.10) (0.11) (0.10) (0.11)
Occipital° 1.09 1.12 1.11 1.15
([Link]) (0.11) (0.10) (0.10)
Temporal 0.90 0.90 0.93 0.94
(0.08) (0.10) (0.08) (0.08)
Cingulate 0.99 0.96 0.94 0.92
(0.12) (0.17) (0.12) (0.14)
Subcortical
Corpus callosuma 0.68 0.56 0.67 0.56
(0.12) (0.18) (0.12) (0.18)
Amygdalab 0.97 0.94 0.83 0.88
(0.14) (0.17) (0.14) (0.16)
Medial temporal lobe 0.95 0.91 0.93 0.96
and hippocampusb (.10) (0.10) (0.1 l) (0.09)
Thalamusb 1.09 1.09 1.09 1.15
(0.14) (0.12) (0.16) (0.14)
Caudate 1.19 1.18 1.27 1.27
(0.16) (0.15) (0.13) (0.12)
Putamen 1.22 1.21 1.26 1.28
(0.12) (0.13) (0.11) (0.10)
Globus pallidus 0.96 0.94 0.97 0.98
(0.18) (0.13) (0.15) (0.16)
Midbrain 0.74 0.75 0.76 0.80
(0.12) (0.13) (0.11) (0.12)
Cerebellum [Link] 1.05 1.03 [Link]
(0.17) (0.16) (0.16) (0.17)
a Maingroup effect.
b Group X hemisphereinteraction.

and lateral middle frontal gyri of both left (t = 2.1, p < TEMPORAL. Murderers were identical to controls on
.04) and fight (t = 2.8, p < .007) hemispheres. lateral temporal lobe glucose m e t a b o l i s m (see Table 1). A
group X hemisphere x gyrus M A N O V A revealed no
PARIETAL. Murderers had lower parietal glucose me- significant m a i n group effect (p > .86) or interaction
tabolism than controls, especially in the left angular gyrus i n v o l v i n g group (p > .20).
and bilateral superior parietal regions. A three-way
group X hemisphere x gyrus (angular, superior, supra- OCCIPITAL. Murderers were found to show signifi-
marginal, and postcentral gyri) M A N O V A indicated a cantly higher occipital lobe glucose m e t a b o l i s m than
m a r g i n a l m a i n effect for group, F(1,80) = 3.7, p < .06, normals (see Table 1). A group x hemisphere × area
but also a significant group X gyrus interaction, F(3,78) = M A N O V A revealed a m a i n effect for group, F(1,82) =
3.9, p < .02. As indicated in the lower half o f Figure 4, 6.8, p < .02, and a group × area interaction, F(3,78) =
murderers had significantly lower glucose specifically in 4.5, p < .006. A b r e a k d o w n o f this interaction indicated
both left (t = 2.5, p < .02) and right (t = 2.0, p < .05) that increased m e t a b o l i s m in murderers was especially
superior parietal gyri, with an additional trend for the left m a r k e d bilaterally in areas 17 (inferior) (t = 3.8, p <
a n g u l a r gyrus (t = 1.9, p < .06). No other effects .0001) and 18 (t = 3.4, p < .001). N o other interactions
i n v o l v i n g group were significant (p > .33). with group were significant (p > . 11).
500 BIOL PSYCHIATRY ,at. R a i n e et al
1997;42:495-508

LATERAL PREFRONTAL CORPUS CALLOSUM

CONTROLS 1 MURDERERS ~CONTROLS 1MURDERERS

R R
E 114 E o7-
L L
A 1.18 A
T T 065
I 112 I
V V
E 111 E 06
G G
L 11 L
U U o.e5
C 1.o9 C
O 0
S [Link] S o6
E LEFT RIGHT E LEFT RIGFT
HEMISPHERE HEMISPHERE

MEDIAL PREFRONTAL PARIETAL CORTEX

~CONTROLS 1MURDERERS ~CONTROLS 1MURDERER8

R R
E 1.25 E 1.22-
L L
A 1.23 A
T T 1.17i
I I
V 1.21 V
E E 1.12"
G 1.19 G
L L
U 1.o7
U 1.17
C C
O O
S 116 S 1.02-
LEFT RIGHT E POST-CENTRAL SUPRAMARGPNAL ANGULAR SUPERIOR
E
HEMISPHERE GYRUS
Figure 3. Relative glucose metabolic rates for murders and Figure 4. Relative glucose metabolic rates for murderers and
controls in lateral prefrontal cortex (above) and medial prefrontal controls in the corpus callosum and parietal cortex. Murderers
cortex (below). Murderers have significantly lower lateral (p < have lower activity in the corpus callosum bilaterally (p < .001),
.02) and medial (p < .02) prefrontal functioning in both in the superior parietal gyri bilaterally (p < .05), and also in the
hemispheres. left angular gyrus (p < .06).

Subcortical Regions activity, but relatively greater fight amygdala activity. A

CORPUS CALLOSUM. Murderers had bilaterally lower laterality coefficient (computed using the formula left -
glucose metabolism in the corpus callosum than controls fight/left + right) indicated that murderers had relatively
(see upper half of Figure 4). A group X hemisphere lower left than right amygdala activity (mean = 0.03,
M A N O V A indicated a main effect for group, F(1,80) = SD = .10) compared to controls (mean = 0.08, SD = .08)
11.6, p < .001, with no interaction effect (p > .10). (t = 2.5, p < .02).

AMYGDALA. Murderers had relatively reduced left and MEDIAL TEMPORAL LOBE INCLUDINGTHE HEPPOCAMPUS.
greater right amygdala activity relative to controls. A Murderers had relatively reduced left and greater fight
group X hemisphere M A N O V A revealed no main group activity. A group X hemisphere M A N O V A revealed no
effect (p > .83), but instead showed a significant group x main group effect (/7 > .93), but insteadi showed a
hemisphere interaction, F(1,80) = 6.8, p < .02. As significant group X hemisphere interaction, F(1,80) =
indicated in Figure 5, murderers showed an abnormal 8.4, p < .005. As indicated in Figure 5, murderers showed
asymmetry consisting of relatively reduced left amygdala an abnormal asymmetry consisting of relatively reduced
Brain Abnormalities in Murderers BIOLPSYCHIATRY 501
1997;42:495--508

THALAMUS THALAMUS. Murderers had relatively greater right tha-

lamic activity relative to controls. A group X hemisphere
~ - OOHTROLS ~ MURDERER,~
M A N O V A revealed no main group effect (p > .25), but
-

R instead showed a significant group x hemisphere interac-

E 1,15
L
A ~.~3
tion, F(1,80) = 4.4, p < .04. As indicated in Figure 5,
T murderers showed an abnormal asymmetry consisting of
I
V 111 relatively greater right thalamic activity. A laterality co-
E
G loe efficient indicated that murderers had relatively lower left
L than right thalamic activity (mean = - 0 . 0 3 , SD = 0.07)
U "i.o7
C compared to controls (mean = 0.0, SD = 0.06) (t = 2.0,
O
S lOe I I
p < .05).
LEFT RIGHT
E
HEMISPHERE

CINGULATE. Murderers did not differ fi'om controls on

cingulate glucose metabolism (see Table 1). Both the
AMYGDALA
group main effect and the group X hemisphere interaction
were nonsignificant (p > .29).
- 6- CONTROLS ~ MURDERERS

CAUDATE, PUTAMEN, GLOBUS PALLIDUS, MIDBRAIN, AND

1

cl
CEREBELLUM. TO assess specificity of subcortical find-
og5 ings, groups were compared on glucose metabolic activity
in the above structures, which have characterized other
0 0
mental disorders (see Discussion), but which have not
been theorized to relate to violent crime. Means and SDs
0,85
"u
are given in Table 1. Group main effects for the caudate,
0,8
i i putamen, globus pallidus, and midbrain were all nonsig-
LEFT RIGHT
HEMISPHERE nificant (p > .25), as were all interactions involving group
(p > .17). A trend was observed for murderers to have
slightly higher (not lower) cerebellar glucose metabolic
activity than normals (t = 1.7, p < .10). All interactions
MEDIAL TEMPORAL LOBE / HIPPOCAMPU8 involving group were nonsignificant (p > .30).
-~'CONTROL$ ~MURDERERS

R
Behavioral Performance on the CPT
E [Link]
L Groups did not differ on any aspect of behavioral perfor-
A
T o,oe mance on the CPT. Averaged means and SDs (in paren-
I
v theses) for the two groups (controls and murderers respec-
E o.04 tively) were as follows: d': 3.63 (0.73), 3.:55 (0.74), t =
G
L 0.5, p > .64; true positives (correct hits): 37.1 (5.6), 36.7
U 092
c (4.5), t = .79, p > .42; false negatives (eJrrors of omis-
O , i sion): 3.6 (5.0), 3.3 (4.5), t = .24, p > .80; false positives
s 0~ LEFT RIGHT
E (errors of commission): 4.5 (8.4), 2.3 (3.9), t = 1.5, p >
HEMISPHERE

Figure 5. Significant group X hemisphere interactions for rela- .15; true negatives (correct misses): 114.8 (8.4), 117.7
tive glucose metabolic rates for murderers and controls in the (3.9), t = 1.3, p > .20.
thalamus (p < .04), medial temporal lobe/hippocampus (p <
.005), and amygdala (p < .02).
Effects of Handedness, Head Injury, and Ethnicity
Although subjects were matched on gender, age, and
schizophrenia, it was not possible to simultaneously match
left medial temporal/hippocampal activity, but relatively them on handedness, head injury, and ethnicity. Six of the
greater right activity. A laterality coefficient indicated that murderers were left-handed. These were compared to
murderers had relatively lower left than right activity right-handed murderers and the analyses described earlier
(mean = - 0 . 0 3 , SD = 0.06) compared to controls for PET variables that produced significant group differ-
(mean = 0.01, SD = 0.07) (t = 2.8, p < .006). ences were repeated. All such analyses were nonsignifi-
502 BIOLPSYCHIATRY A. Raine et al
1997;42:495-508

cant, with the exception that left-handed murderers tended 1989; Mirsky and Siegel 1994; Weiger and Bear 1988).
to have higher (not lower) medial prefrontal activity (p < The amygdala, hippocampus, and prefrontal cortex make
.08), and had a significant less abnormal amygdala asym- up part of the limbic system governing the expression of
metry (p < .002) than right-handed murderers. Results emotion, while the thalamus relays inputs from subcortical
indicate therefore that greater rates of left-handedness in limbic structures to the prefrontal cortex (Fuster 1989;
the murderer group relative to controls cannot account for Mirsky and Siegel 1994). The hippocampal formation is
reduced prefrontal activity and the abnormal amygdala thought to modulate aggression in cats through its action
asymmetry. on the lateral hypothalamus via the lateral septal area
Fourteen of the murderers were nonwhite. Analyses (Mirsky and Siegel 1994; Siegel and Flynn 1968), and
comparing them to white murderers on PET measures together with the septal area and prefrontal cortex forms
were nonsignificant (p > . 14) in all cases, indicating that the neurobiological basis of the behavioral inhibition
ethnic status did not influence findings. system of Gray (1982), which is theorized to be dysfunc-
Twenty-three murderers had a history of head injury. tional in violent and psychopathic individuals (Gorenstein
They did not differ from murderers without a history of and Newman 1980). The amygdala is believed to act on
head injury on PET measures (p > .27), with the one the medial hypothalamus through at least two pathways in
exception of a trend for head-injured murderers to have the modulation of aggression in animals (Watson et al
lower activity in the corpus callosum (p < .08) than 1983b). The hippocampus, amygdala, and thalamus are
non-head-injured murderers. Although analyses suggest also of critical importance to learning, memory, and
that history of head injury cannot account for most attention; abnormalities in their functioning may relate to
findings, the possibility that they account for group dif- deficits in forming conditioned emotional responses and
ferences in the corpus callosum cannot be ruled out. the failure to learn from experience displayed by criminal
and violent offenders (Cleckley 1976; Raine 1993). The
amygdala additionally plays a role in the recognition of
Discussion
affective and socially significant stimuli (Nishijo et al
Key Findings 1988), with destruction of the amygdala in animals result-
The key findings from this preliminary study are that ing in a lack of fear (Bear 1991) and in man in a reduction
murderers pleading NGRI are characterized by a) reduced in autonomic arousal (Lee et al 1988); thus, abnormalities
glucose metabolism in bilateral prefrontal cortex, the in the amygdala could be relevant to a [Link] theory
posterior parietal cortex (bilateral superior gyrus and left of violence based on psychophysiological findings of
angular gyrus), and the corpus callosum, and b) abnormal reduced autonomic arousal in offenders (Raine et al
asymmetries of activity (left hemisphere lower than right) 1990b; Raine 1993).
in the amygdala, thalamus, and medial temporal gyrus The posterior parietal cortex (including superior and
including the hippocampus. These data both confirm angular gyri) is centrally involved in the integration of
deficits in the prefrontal cortex from our earlier pilot sensory input and the formation of abstract concepts (Kolb
study, and also yield new findings. These in turn provide and Wishaw 1990), and in conjunction with its reciprocal
both some general support for preexisting biological the- connections with the dorsolateral prefrontal cortex (Gold-
ories of violence, and also suggest new perspectives for man-Rakic et al 1983) may contribute to the cognitive and
understanding the type of brain dysfunction that may social information processing deficits observed in violent
predispose to violence in this specific group of offenders. offenders (Dodge and Crick 1990; Moffitt and Silva
1988). Reductions in glucose metabolism in the left
angular gyrus have been correlated with reduced verbal
Biosocial Pathways from Brain Deficits to Violence ability (Gur et al 1994), while damage to the left angular
A key question concerns how these multisite deficits can gyrus has been linked to deficits in reading anti arithmetic.
translate into violence via neuropsychological, psycholog- Such cognitive dysfunction could predispose to educa-
ical, cognitive, social, and situational pathways. Regarding tional and occupational failure, which in turn predispose to
prefrontal deficits, damage to this brain region can result crime and violence. Learning deficits have been found to
in impulsivity, loss of self-control, immaturity, altered be common in violent offenders who also haw~ low verbal
emotionality, and the inability to modify behavior, which IQs (Quay 1987; Raine 1993). One caveat here is that the
can all in turn facilitate aggressive acts (Damasio 1985; finding for the left angular gyrus was a trend using a
Damasio et al 1994; Moffitt and Henry 1991; Stuss and two-tailed test (p < .06), although the effect is significant
Benson 1986; Weiger and Bear 1988). Regarding limbic on a one-tailed test (p < .03) and was predicted on the
deficits, the amygdala has been repeatedly associated with basis of event-related brain potential data from Barratt et
aggressive behavior in both animals and humans (Bear al in press. In contrast to these posterior parietal areas,
Brain Abnormalities in Murderers BIOLPSYCHIATRY 503
1997;42:495--508

which are dysfunctional in murderers, the more anterior for processing socially relevant information (Brothers and
parietal regions are involved in more basic somatic sen- Ring 1993), and functions in parallel with the object
sations and perceptions and are unaffected in murderers, recognition system of the hippocampus and the spatial
indicating some specificity of dysfunction within the recognition system of the posterior parielLal cortex (Kolb
parietal region. and Wishaw 1990). Disruption of such a system could in
Although there have been speculations for many years part relate to the socially inappropriate behavior shown by
that dysfunction to the corpus callosum may be a neuro- some violent individuals (Cleckley 1976) and the mis-
biological predisposition to violence (e.g., Nachshon recognition and misappraisal of ambiguous stimuli in
1983; Yeudall 1977), until now there has been no direct social situations that have potential for violent encounters
evidence to support such a contention. Although white (Dodge et al 1990; Nachshon and Rotenberg 1977).
matter metabolic values are only approximately 50% of Findings of this study suggest that the neural processes
grey matter values (thus biasing toward floor effects and underlying violence are complex and cannot be simplisti-
nonsignificant results), we obtained our strongest group cally reduced to single brain mechanisms causing violence
difference for this region. Callosal dysfunction and the in a direct causal fashion. Instead, violent behavior prob-
consequent lack of interhemispheric integration could ably involves disruption of a network of multiply interact-
contribute to the abnormal asymmetries of function and ing brain mechanisms that predispose to violence in the
reduced interhemispheric integration previously observed presence of other social, environmental, and psychological
in antisocial and violent groups (Hare and McPherson predispositions (Eichelman 1992; Earls 1991; Lewis et al
1984; Flor-Henry et al 1991; Raine et al 1990a). We have 1988). Nevertheless, attempts to "network" findings from
previously hypothesized that the reduced lateralization for the individual brain sites in this study must proceed
processing linguistic information observed in violent cautiously, because there are brain mechanisms relevant to
groups may arise from a reduction in the normal neuro- aggression (e.g., septum and hypothalamus) that could not
developmental processes of hemispheric specialization, a be imaged in this study. For this reason, this study cannot
process that may in part be accounted for by dysfunction provide a complete account of the neurophysiology of
of the corpus callosum (Raine et al 1995). violence in this specific and selected subgroup of violent
Another potential implication of poor interhemispheric offenders, although it is felt both that it does provide
transfer is that the right hemisphere, which has been preliminary evidence that murderers pleading NGRI have
implicated in the generation of negative affect in humans different brain functioning compared to normals, and also
(Davidson and Fox 1989), may experience less regulation that it gives initial suggestions as to which specific neural
and control by left hemisphere inhibitory processes (Cook processes may predispose to their violent behavior.
1986; Flor-Henry 1987), a factor that may contribute to the
expression of violence in predisposed individuals. In
Potential Confounds
animals, rats who are stressed early in life are right
hemisphere dominant for mice-killing (Garbanati et al We do not believe that these results reflect merely chance
1983). Severing the corpus callosum in these rats leads to findings for five reasons. First, the sample ,;ize (41 in each
an increase in muricide (Denenberg et al 1986), indicating group) is not small for PET research, and is substantially
that the left hemisphere acts to inhibit the right hemisphere larger than other imaging studies of violent populations.
mediated killing via an intact corpus callosum. The fact Second, the strength of effects were not trivial, with a
that both Sperry (1974) and Dimond (1979) commented mean effect size of 0.55 (range = 0.36-0.80), which is
on the inappropriate nature of emotional expression and viewed as medium (Cohen 1988). Third, areas were
the inability to grasp long-term implications of a situation selected for analysis on the basis of prior l:heorizing, and
in split-brain patients may also give pointers to the all but one of these produced significant effects. Fourth, to
inappropriate emotional expression of violent offenders help limit the possibility of type I errors, overall MANO-
and their lack of long-term planning (Cleckley 1976). VAs were conducted and two-tailed tests used throughout.
Nevertheless, findings from animal research cannot be Fifth, brain areas that have not been theoretically linked to
directly extrapolated to humans. Furthermore, callosal violence but that have been linked to other mental disor-
dysfunction per se is unlikely to cause aggression; instead ders (caudate, putamen, globus pallidus, raidbrain, cere-
it may contribute to violence in those with concurrent bellum) did not yield group differences; this double
limbic and cortical abnormalities. dissociation lends some support to the relalively differen-
Findings of group differences in glucose metabolism in tial nature of the brain deficits in terms of both anatomy
the posterior parietal cortex, amygdala, and medial tem- and mental condition. Nevertheless, it shoald be empha-
poral lobe including the hippocampus may not be unre- sized that some effects were marginal (e.g., left angular
lated. The amygdala has been viewed as part of a system gyrus), and results must be treated cautiously, particularly
504 BIOL PSYCHIATRY A. Raine et al
1997;42:495-508

those regarding subcortical laterality effects and increased ously reported in psychiatric patients. For example,
occipital functioning, which were not predicted a priori. whereas altered functioning has been found in schizo-
It does not appear that the findings are a function of phrenics for the lateral temporal cortex (Buchsbaum et al
group differences in age, gender, schizophrenia, handed- 1990; DeLisi et al 1989), basal ganglia (Buchsbaum 1990;
ness, ethnicity, or history of head injury. Groups were Early et al 1987; Gur and Pearlson 1993), cingulate gyrus
matched on age, gender, and ethnicity. Analyses compar- (Siegel et al 1993), caudate (Siegel et al 1993), and
ing left-handed with right-handed murderers, white versus cerebellum (Volkow et al 1992), these structures were
nonwhite murderers, and murderers with and without a unaffected in murderers. Similarly, there i,; a growing
history of head injury do not support the view that the consensus that affective disorder involves dysfunction to
greater rates of left handedness, head injury, and non- both frontal and temporal lobes (Baxter et al 1989;
whites in the murderer group account for overall murderer Cummings 1993; George et al 1993). In contrast, although
versus control group differences. One caveat to this murderers have widespread bilateral reductions in prefron-
conclusion is that there was a trend (p < .10) for tal glucose utilization, they did not show the lateral
murderers with a history of head injury to have reduced temporal deficits that have been observed in schizophren-
glucose metabolism in the corpus callosum. This would be ics using the same methodology (Buchsbaum et al 1990),
consistent with the notion that sheering of white nerve while depressives tend to have dysfunction lateralized to
fibers during closed head injuries could contribute to the left hemisphere (Baxter et al 1989; Bench et al 1993;
damage to the corpus callosum (McAllister 1992). In Drevets et al 1992) and to the left dorsolateral prefrontal
addition, because we did not have more extensive neuro- region in particular (Baxter et al 1989; Bench et al 1993),
logical and medical data to assess history of head injury, in contrast to the bilateral prefrontal findings for murder-
we cannot definitively rule out prior head injury as a ers. Furthermore, depressives have been reported to show
possible contribution toward reduced brain activity in the additional involvement of the caudate nucleus (Cummings
murderers. 1993) and cingulate gyrus (George et al 1993), brain areas
The fact that groups did not differ in behavioral perfor- unaffected in murderers. Obsessive-compulsives show
mance on the CPT suggests not only that difference in higher, not lower, glucose levels in orbitofrontal cortex
brain functioning is not easily accounted for by motiva- (Baxter et al 1988; Benkelfat et al 1990), while symptom
tional or attentional deficits in the murderers, but also that intensity in this group is associated with higher (not lower)
the significantly greater occipital activity (visual areas 17 functioning in the hippocampus and thalamus (McGuire et
and 18) in murderers may possibly represent compensa- al 1994). With respect to substance abuse, acute cannabi-
tion for the reduced activity in the prefrontal cortex, an nol administration affects cerebellar functioning (Volkow
area which is critical for the execution of this challenge et al 1991), whereas murderers showed normal cerebellar
task (Buchsbaum et al 1990). Cognitive parity between activity. Detoxified alcoholics show increased (not de-
groups cannot be claimed, because no IQ data were creased) brain metabolism during detoxification, with
available on the subjects. Nevertheless, we do not believe persistent low metabolic levels being shown for the basal
that lower IQ in the murderer group can account for ganglia (Volkow et al 1994), a structure unaffected in
findings of reduced glucose metabolism, because low IQ murderers. Whereas cerebellar hypometabolism and de-
has been associated with higher, not lower, cerebral generation has been observed in chronic alcoholics (Gil-
glucose metabolism (Haier et al 1988). man et al 1990), murderers showed nonsignificantly
higher, not lower, cerebellar activity.
Reduced prefrontal activity does not seem to be specific
Specificity of Findings to severe violence, as this finding has been observed in a
The question of whether comorbid psychiatric conditions variety of psychiatric conditions. On the other hand, to the
in the murderers could account for PET findings needs to authors' knowledge there have been no previous reports in
be considered. The most important psychiatric condition in any psychiatric condition of left lower than right asymme-
murderers consists of schizophrenia. We controlled for tries in the amygdala, thalamus, and hippocampus coupled
this by matching 6 schizophrenic murderers with age- and with dysfunction of the corpus callosum and left angular
sex-matched hospital schizophrenics. We do not believe gyrus. For example, although there have been variable
that other forms of psychiatric comorbidity can easily reports of either increased, decreased, or normal thalamic
account for our findings, because differences in brain activity in schizophrenia (Buchsbaum et al 1987; Resnick
functioning in murderers show a different pattern to that et al 1988; Siegel et al 1993), the left lower than right
observed in other mental disorders. Psychiatric patients asymmetries for these structures in murderers have not
show abnormalities in brain structures not found in the been previously reported. While prefrontal dysfunction
murderers, while murderers have abnormalities not previ- may represent a deficit common to many fo~mas of psy-
Brain Abnormalities in Murderers BIOLPsYCmATRY 505
1997 ;42:495-508

chopathology, additional dysfunction to these other brain by multiple obstacles to research. Despite some limitations
structures may lead to a pathway toward violence as in the research, we nevertheless feel it appropriate to
opposed to other conditions. In drawing comparisons report these findings, because they constitute the first to
across imaging studies, it must be borne in mind that some document multiple but selective brain deficits assessed
studies have used exactly the same imaging methodology using PET in a group of severely violent offenders who are
employed in the present study (e.g., Buchsbaum et al of particular importance in forensic psychiatry, and be-
1990; Siegel et al 1993; DeLisi et al 1989), whereas others cause they provide both theoretical directions and a critical
have employed different methodologies (e.g., Baxter et al empirical base upon which future brain imaging studies of
1989; Volkow et al 1992). As such, strict comparisons violent offenders may build. At the same time, the need for
across studies are not possible. caution in interpreting findings due to the preliminary
Although coexisting psychopathology may contribute to nature of the findings and the need for independent
violence and should not be discounted as unimportant, it replication must be reemphasized.
does not seem that such pathology per se can account for Although the study has strengths, it is critically impor-
the specific network of brain dysfunction observed in this tant to document what this study does and does not
violent group. Nevertheless, although subjects constituted indicate. First, these findings cannot be taken to demon-
a relatively specific subgroup of violent offenders (all had strate that violence is determined by biology alone;
committed homicide and were pleading NGRI), it must be clearly, social, psychological, cultural, and situational
acknowledged that they do not constitute a homogenous factors also play important roles in predisposing to vio-
clinical group. Specifically, heterogeneity would contrib- lence (Eichelman 1992; Elliott 1987, 1988). Second, these
ute to type II error and the failure to observe significant data do not demonstrate that murderers pleading NGRI are
group differences in some brain regions of interest. As not responsible for their actions, nor do they demonstrate
such, it must be emphasized that these initial findings must that PET can be used as a diagnostic technique. Third, our
be viewed with caution and due circumspection. findings cannot speak to the issue of the cause (genetic or
environmental) of the brain dysfunction, nor do they
Strengths, Limitations, and Conclusions establish causal direction. Fourth, findings cannot be
generalized at the present date from NGRI murder cases to
As with all initial findings, the current study has limita- other types of violent offenders. Fifth, specificity to
tions, including relatively modest spatial resolution rela- violence as opposed to crime per se has not been estab-
tive to the most advanced present-day PET techniques, the lished, as this requires the inclusion of a nonviolent
lack of standardized diagnostic and neuropsychological criminal control group, which was not available. What
assessments, and the use of the canthomeatal line for slice these initial findings do document, however, is that as a
placement, which has a variable orientation to brain group, murderers pleading NGRI have statistically signif-
landmarks, and which can lead to significant variability icant differences in glucose metabolism in selected brain
across subjects in the anatomical localization of regions of regions compared to normals. They also suggest, but do
interest. Limitations such as the absence of psychiatric not conclusively demonstrate, that reduced activity in the
control groups have also characterized the first brain prefrontal, parietal, and callosal regions of the brain,
imaging studies of other conditions such as schizophrenia together with abnormal asymmetries of activity in the
as well as some current studies. In addition, it must be amygdala, thalamus, and medial temporal lobe including
reiterated that findings apply only to a select subgroup of the hippocampus, may be one of many predispositions
severely violent offenders and cannot be generalized at toward violence in this specific group. As with all initial
this stage to violence per se. Furthermore, findings for findings in the field, future independent replication, refine-
subcortical asymmetries and the occipital cortex were not ment, and extension to less select populations of violent
predicted a priori and need to be replicated in an indepen- offenders are greatly needed.
dent study.
Balancing these limitations, it is felt that the study also
has a number of strengths. These include by far the largest
sample of seriously violent offenders ever imaged, match- This research was supportedin part by a ResearchScientistDevelopment
Award from NIH (1 KO2 MH01114-01) to the first author, and a grant
ing for age, sex, and schizophrenia, ruling out confounds to the first author from the Southern California ]injury Prevention
of handedness, ethnicity, and head injury, and establishing Research Center (Center for Disease Control, grant R49/CCR903622).
group equivalence on behavioral performance of the chal- We thank Jill Stanley for research assistance, Steven Lottenbergand
lenge task and psychopharmacologic control over medica- Leonard Abel for help in data collection, Joseph Woo for discussions on
tion and illegal drug use in weeks prior to scanning. Such alternative methods of data analysis, and Keith Nuechterlein for helpful
advice on the continuous performancetest.
strengths are not common in this field, which is hampered
506 BIOL PSYCHIATRY A. Raine et al
1997;42:495-508

References
Barratt ES, Stanford MS, Kent TS, and Felthous, AR (1997): in the EEG of persistently violent psychiatric inpatients. Biol
Neuropsychological and cognitive psychophysiological sub- Psychiatry 30:363-370.
states of impulsive aggression. Biol Psychiatry 41:1045- Cook ND (1986): The Brain Code. London: Methuen.
1061.
Cummings JL (1993): The neuroanatomy of depression. J Clin
Bartlett EJ, Baouche F, Bodie JD, et al (1991): Stability of Psychiatry 54:14-20.
resting deoxyglucose metabolic values in PET studies of
schizophrenia. Psychiatry Res Neuroimaging 40:111-20. Damasio A (1985): The frontal lobes. In Heilman KM, Valen-
stein E (eds), Clinical Neuropsychology. New York: Oxford
Baxter LR, Schwartz JM, Mazziotta JC, et al (1988): Cerebral University Press, pp 339-375.
glucose metabolic rates in non-depressed patients with obses-
sive-compulsive disorder. Am J Psychiatry 145:1560-1563. Damasio AR, Tranel D, Damasio H (1990): Individuals with
sociopathic behavior caused by frontal damage fail to respond
Baxter LR, Schwartz JM, Phelps ME (1989): Reduction of autonomically to social stimuli. Behav Brain Res 41:81-94.
prefrontal cortex glucose metabolism common to three types
of depression. Arch Gen Psychiatry 46:243-250. Damasio H, Grabowski T, Frank R, Galaburda AM, Damasio AR
(1994): The return of Phineas Gage: Clues about the brain
Bear D (1989): Hierarchical neural regulation of aggression: from the skull of a famous patient. Science 264:1102-1105.
Some predictable patterns of violence. In Britzer DA,
Crowner M (eds), Current Approaches to the Prediction of Davidson RJ, Fox NA (1989): Frontal brain asymmetry predicts
Violence. Washington, DC: American Psychiatric Press, pp infants' response to maternal separation. J Abnorm Psychol
85-100. 98:127-131.
Bear D (1991): Neurological perspectives on aggressive behav- Davies DR, Parasuraman R (1982): The Psychology of Vigilance.
ior. J Neuropsychiatry 3 (suppl):S3-S8. London: Academic Press.
Bench CJ, Friston KJ, Brown RG, Frackowiak RS, Dolan RJ DeLisi LE, Buchsbaum MS, Holcomb HH, et al (1989): In-
(1993): Regional cerebral blood flow in depression measured creased temporal lobe glucose use in chronic schizophrenic
by positron emission tomography: The relationship with patients. Biol Psychiatry 25:835- 851.
clinical dimensions. Psychol Med 23:579-590. Denenberg VH, Gall JS, Berrebi A, Yutzey DA (1986): Callosal
Benkelfat C, Nordahl TE, Semple WE, et al (1990): Local mediation of cortical inhibition in the lateralized rat brain.
cerebral glucose metabolic rates in obsessive-compulsive Brain Res 397:327-332.
disorder. Arch Gen Psychiatry 47:840-848. Dimond SJ (1979): Disconnection and psychopathology. In
Brothers L, Ring B (1993): Mesial temporal neurons in the Gruzelier JH, Flor-Henry P (eds), Hemisphere Asymmetries
macaque monkey with responses selective for aspects of of Function in Psychopathology. Amsterdam: Elsevier, pp
social stimuli. Behav Brain Res 57:53-61. 35-46.
Buchsbaum MS (1990): The frontal lobes, basal ganglia, and Dodge KA, Crick NR (1990): Social information processing
temporal lobes as sites for schizophrenia. Schizophr Bull bases of aggressive behavior in children. Pers Soc Psychol
16:379-389. Bull 16:8-22.
Buchsbaum MS, Wu JC, DeLisi LE, et al (1987): Positron Dodge KA, Price JM, Bachorowski JA (1990): Hostile attribu-
emission tomography studies of basal ganglia and somato- tional biases in severely aggressive adolescent,;. J Abnorm
sensory cortex neuroleptic drug effects: Differences between Psychol 99:385-392.
normal controls and schizophrenic patients. Biol Psychiatry Drevets WC, Videen TO, Price JL, Preskorn SH, Carmichael ST,
22:479-494. Raichle ME (1992): A functional anatomical study of unipo-
Buchsbaum MS, Gillin JC, Wu J (1989): Regional cerebral lar depression. J Neurosci 12:3628-3641.
glucose metabolic rate in human sleep assessed by positron Earls F (1991): Not fear, not quarantine, but science: Preparation
emission tomography. Life Sci 45:1349-1356. for a decade of research to advance knowledge about causes
Buchsbaum MS, Nuechterlein KH, Haier RJ (1990): Glucose and control of violence in youths. J Adolesc Health 12:619-
metabolic rate in normals and schizophrenics during the 629.
continuous performance test assessed by positron emission Early TS, Reiman EM, Raichle ME, Spitznagel EL (1987): Left
tomography. Br J Psychiatry 156:216-227. globus pallidus abnormality in never-medicated patients with
Buchsbaum MS, Potkin SG, Marshall JF, and Lottenberg S schizophrenia. Proc Nat Acad Sci USA 84:561-.563.
(1992): Effects of clozapine and thiothixene on glucose Eichelman B (1992): Aggressive behavior: From laboratory to
metabolic rate in schizophrenia. Neuropsychophatrnacology clinic. Arch Gen Psychiatry 49:488-492.
6:155-163. Eichelman B (1993): Bridges from the animal laboratory to the
Buchsbaum MS, DeLisi LE, Holcomb HH (1994): Anteroposte- study of violent or criminal individuals. In Hodgins S (ed),
rior gradients in cerebral glucose use in schizophrenia and Mental Disorder and Crime. Newbury Park, CA: Sage, pp
affective disorders. Arch Gen Psychiatry 41:1159--1166. 194-207.
Cleckley HC (1976): The Mask of Sanity 5th ed. St. Louis: Elliott FA (1987): Neuroanatomy and neurology of aggression.
Mosby. Special issue: Treatment of aggressive disorders. Psychiatr
Cohen J (1988): Statistical Power Analysis for the Behavioral Ann 17:385-388.
Sciences, 2nd ed. Hillsdale, NJ: Erlbaum. Elliott FA (1988): Violence: A product of biosocial interactions.
Convit A, Czobor P, Volavka J (1991): Lateralized abnormality Bull Am Acad Psychiatry Law 16:131-143.
Brain Abnormalities in Murderers BIOL PSYCHIATRY 507
1997;42:495-508

Elliott FA (1992): Violence: The neurologic contribution; An HF (1988): Changes in autonomic responsiveness following
overview. Arch Neurol 49:595-603. bilateral amygdalectomy in humans. Neuropsychiatry Neuro-
Eysenck HJ, Gudjonsson GH (1989): The Causes and Cures of psychol Behav Neurol 1:119-129.
Criminality. New York: Plenum. Lewis DO, Pincus JH, Bard B, et al (1988): Neuropsychiatric,
Flor-Henry P (1987): Cerebral aspects of sexual deviation. In psycho-educational, and family characteristics of 14 juveniles
Wilson GD (ed), Variant Sexuality: Research and Theory. condemned to death in the United States. Am J Psychiatry
London: Croom Helm, pp 49-83. 145:584-589.
Matsui T, Hirano A (1978): An Atlas of the Human Brain for
Flor-Henry P, Lang RA, Koles ZJ, Frenzel RR (1991): Quanti-
Computerized Tomography. Tokyo: Igaku-Shoin.
tative EEG studies of pedophilia, lnt J Psychophysiol 10:253-
258. McAllister TW (1992): Neuropsychological sequelae of head
injuries. Psychiatr Clin North Am 15:395-413.
Fox PT, Mintum MA (1989): Noninvasive functional brain
mapping by change-distribution analysis of averaged PET McGuire PK, Bench C J, Frith CD, Marks IM. Frackowiak RS,
images of H2-150 tissue activity. J Nucl Med 30:141-149. Dolan RJ (1994): Functional anatomy of obsessive-compul-
sive phenomenon. Br J Psychiatry 164:459-468.
Fuster JM (1989): The Prefrontal Cortex: Anatomy, Physiology,
and Neuropsychology of the Frontal Lobe, 2nd ed. New Mirsky AF, Siegel A (1994): The neurobiology of violence and
York: Raven Press. aggression. In Reiss AJ, Miczek KA, Roth JA (eds), Under-
standing and Preventing Violence. Vol. 2. Biobehavioral
Garbanati JA, Sherman GF, Rosen GD, Hofmann M J, Yutzey Influences. Washington, DC: National Academy Press, pp
DA, Denenberg VH (1983): Handling in infancy, brain 59-172.
laterality and muricide in rats. Behav Brain Res 7:351-359.
Moffitt TE (1988): Neuropsychology and self-reported early
George MS, Ketter TA, Post RM (1993): SPECT and PET delinquency in an unselected birth cohort. In Moffitt TE,
imaging in mood disorders. J Clin Psychiat~ 54:6-13. Mednick SA (eds), Biological Contributions to Crime Cau-
Gilman S, Adams K, Koeppe RA, et al (1990): Cerebellar and sation. New York: Martinus Nijhoff, pp 93-120.
frontal hypometabolism in alcoholic cerebellar degeneration Moffitt TE, Henry B (1991): Neuropsychological studies of
studied with positron emission tomography. Ann Neurol juvenile delinquency and juvenile violence. ]in Milner JS (ed),
28:775-785. Neuropsychology of Aggression. Boston: Kluwer Academic
Goldman-Rakic PS, Isseroff A, Schwartz ML, Bugbee NM Publishers.
(1983): The neurobiology of cognitive development. In Mus- Moffitt TE, Silva PA (1988): IQ and delinquency: A direct test
sen P (ed), Handbook of Child Psychology: Biology and of the differential detection hypothesis. J Abnorm Psychol
Infancy Development. New York: Wiley, pp 281-344. 97:227-240.
Gorenstein EE, Newman JP (1980): Disinhibitory psychopathol- Nachshon I (1983): Hemisphere dysfunction in psychopathic and
ogy: A new perspective and a model for research. Psychol behavior disorders. In Myslobodsky MS (ed), Hemisyn-
Rev 87:301-315. dromes: Psychobiology, Neurology, Psychiatry. New York:
Goyer PF, Andreason PJ, Semple WE, et al (1994): Positron- Academic Press, pp 389-414.
emission tomography and personality disorders. Neuropsy- Nachshon I, Rotenberg M (1977): Perception of violence by
chopharmacology 10:21-28. institutionalized offenders. J Crim Law Criminol 68:454-
457.
Gray JA (1982): The Neuropsychology of Anxie~: An Enquiry
into the Functions of the Septo-Hippocampal 5~stem. Oxford: Nishijo H, Ono T, Nishino H (1988): Single neuron responses in
Oxford University Press. amygdala of alert monkey during complex sensory stimula-
tion with affective significance. J Neurosci 8:3570-3583.
Gur RE, Pearlson GD (1993): Neuroimaging in schizophrenia
research. Schizophr Bull 19:337-353. Nuechterlein KH (199l): Vigilance in schizophrenia and related
disorders. In Steinhauer SR, Gruzelier JH. Zubin J (eds),
Gur RC, Ragland JD, Resnick SM, et al (1994): Lateralized Handbook of Schizophrenia, Vol. 5: Neuropsychology, Psy-
increases in cerebral blood flow during performance of verbal chophysiology, and Information Processing. New York:
and spatial tasks: Relationship with performance level. Brain Elsevier Science Publishers.
Cogn 24:244-258.
Nuechterlein KH, Parasuraman R, Jiang Q (1983): Visual sus-
Haier RJ, Siegel BV, Nuechterlein KH, et al (1988): Cortical tained attention: Image degradation produces rapid decrement
glucose metabolic rate correlations of abstract reasoning and over time. Science 220:327-329.
attention studied with positron emission tomography. Intelli-
Quay HC (1987): Intelligence. In Quay HC (ed), Handbook of
gence 12:199-217.
Juvenile Delinquency. New York: Wiley, pp 106-117.
Hare RD, McPherson LM (1984): Psychopathy and perceptual Raine A (1993): The Psychopathology of (?rime: Criminal
asymmetry during verbal dichotic listening. J Abnorm Psy- Behavior as a Clinical Disorder. San Diego: Academic Press.
chol 93:141-149.
Raine A, O'Brien M, Smiley N, Scerbo A, Chan CJ (1990a):
Harris GJ, Links JM, Pearlson GD, Camargo EE (1991): Cortical Reduced lateralization in verbal dichotic listening in adoles-
circumferential profile of SPECT cerebral perfusion in Alz- cent psychopaths. J Abnorm Psychol 99:272-277.
heimer's disease. Psychiatry Res Neuroimaging 40:167-180.
Raine A, Venables PH, Williams M (1990b): Relationships
Kolb B, Wishaw IQ (1990): Fundamentals of Neuropsychology, between central and autonomic measures of arousal at age 15
3rd ed. New York: WH Freeman. years and criminality at age 24 years. Arch Gen Psychiatry
Lee GP, Arena JG, Meador KJ, Smith JR, Loring DW, Flanigan 47:1003-1007.
508 BIOL PSYCHIATRY A. Raine et al
1997;42:495-508

Raine A, Buchsbaum MS, Stanley J, Lottenberg S, Abel L, Volkow ND, Tancredi L (1987): Neural substralEes of violent
Stoddard S (1994): Selective reductions in prefrontal glucose behavior: A preliminary study with positron emission tomog-
metabolism in murderers. Biol Psychiatry 36:365-373. raphy. Br J Psychiatry 151:668-673.
Raine A, Lencz T, Scerbo A (1995): Antisocial behavior: Volkow ND, Gillespie H, Mullani N, et al (1991): Cerebellar
Neuroimaging, neuropsychology, neurochemistry, and psy- metabolic activation by delta-9-tetra hydrocannabinol in the
chophysiology. In Ratey JJ (ed), Neuropsychiatry of Person- human brain: A study with positron emission tomography and
ality Disorders. Cambridge: Blackwell Science, pp 50-78. 18F-2-fluro-2-deoxyglucose. Psychiatry Res Neuroimaging
Resnick SM, Gut RE, Gur RC, Reivich M (1988): Positron 40:69 -78.
emission tomography and subcortical glucose metabolism in Volkow ND, Levy A, Brodie JD, et al (1992): Low cerebellar
schizophrenia. Psychiatry Res 24:1-11. metabolism in medicated patients with chronic schizophrenia.
Siegel A, Flynn JP (1968): Differential effects of electrical Am J Psychiatry 149:686-688.
stimulation and lesions of the hippocampus and adjacent Volkow ND, Wang GJ, Hitzemann R, et al (1994): Recovery of
regions upon attack behavior in the cat. Brain Res '7:252-267. brain glucose metabolism in detoxified alcoholics. Am J
Siegel BV, Buchsbaum MS, Bunney WE, et al (1993): Cortical- Psychiatry 151:178-183.
striatal-thalamic circuits and brain glucose metabolic activity Watson REJ, Edinger HM, Siegel A (1983a): An analysis of the
in 70 unmedicated male schizophrenic patients. Am J Psychi- mechanisms underlying hippocampal control of hypothalami-
atry 150:1325-1336. cally-elicited aggression in the cat. Brain Pes 269:327-
Sperry RW (1974): Lateral specialization in the surgically 345.
separated hemispheres. In Schmitt FO, Worden FG (eds), The Watson EEJ, Troiano R, Poulakos J, Weiner S, Block CH, Siegel
Neurosciences: Third Study Program. Cambridge, MA: MIT A (1983b): A 14C-2-deoxyglucose analysis of the functional
Press. neural pathways of the limbic forebrain in the rat. 1. The
Stuss DT, Benson DF (1986): The Frontal Lobes. New York: amygdala. Brain Res Rev 5:1-44.
Raven Press. Weiger WA, Bear DM (1988): An approach to the neurology of
Vasey MW, Thayer JF (1987): The continuing problem of false aggression. J Psychiatry Res 22:85-98.
positives in repeated measures ANOVA in psychophysiol- Yeudall LT (1977): Neuropsychological assessment of forensic
ogy: A multivariate solution. Psychophysiology 24:479-486. disorders. Can Ment Health 25:7-16.