American Journal of Medical Genetics (Neuropsychiatric Genetics) 74:475–479 (1997)
Genetic Association Between Monoamine Oxidase
and Manic-Depressive Illness: Comparison of Relative
Risk and Haplotype Relative Risk Data
Abbas Parsian1* and Richard D. Todd1,2
1
Department of Psychiatry, Washington University School of Medicine, St. Louis, Missouri
2
Department of Genetics, Washington University School of Medicine, St. Louis, Missouri
There have been several conflicting reports
of association of monoamine oxidase (MAO)
A gene polymorphisms and bipolar affective
disorder. In order to determine the possible
role of the MAO region in susceptibility to
affective disorders in an independent
sample, we have genotyped 83 probands of
bipolar affective disorder families, 56 sets of
parents of bipolar probands, and 84 normal
controls for intronic simple sequence repeat
polymorphisms of the MAO-A and MAO-B
genes. For MAO-A there were no significant
differences in allele frequencies between bi-
polar and normal control groups for both
genders. However, for MAO-B there were
significant differences between groups for
both genders. In contrast, allele-wise haplo-
type relative risk analysis for the 56 bipolar
proband-parent trios found no significant
differences between transmitted and non-
transmitted allele frequencies for MAO-A or
B. These data do not support the association
of MAO-A or B with bipolar affective disor-
der but do demonstrate that undetected
population stratification can be an impor-
tant source of bias in case-control studies.
Am. J. Med. Genet. 74:475–479, 1997.
© 1997 Wiley-Liss, Inc.
KEY WORDS: manic depression; mono-
amine oxidases; relative risk;
haplotype relative risk;
population stratification
Contract grant sponsor: National Institute on Alcohol Abuse
and Alcoholism; Contract grant number: AA09515; Contract
grant sponsor: National Institute of Mental Health; Contract
grant number: MH 31302.
*Correspondence to: Dr. Abbas Parsian, Department of Psy-
chiatry, Washington University School of Medicine, 4940 Chil-
dren’s Place, St. Louis, MO 63110.
Received 27 November 1996; Revised 8 April 1997
© 1997 Wiley-Liss, Inc.
INTRODUCTION
Association studies of complex disorders with biologi-
cal and genetic markers are increasingly popular due
to the relative low cost and lack of need of specification
of transmission models. Unfortunately, traditional
case-control studies have several potential shortcom-
ings including the generation of false positive results
due to undetected population stratification. An alter-
native approach to the case-control (relative risk) de-
sign for genetic studies is the haplotype relative risk
method. Falk and Rubinstein (1987) proposed the con-
struction of a contrast or control group from the non-
transmitted parental alleles of matings resulting in an
affected proband. This method has the advantage that
both control and disease alleles are sampled from the
same gene pool. A disadvantage of this approach is the
requirement of sampling both parents for each proband
which could be problematic for disorders with late age
of onset or high mortality among relatives. In the pre-
sent study we apply both of these methods to the study
of bipolar affective disorder and the monoamine oxi-
dase (MAO) region of the X-chromosome.
The MAO-A and B genes, located on the short arm of
chromosome X (Xp11.23–p11.4), have been of interest
as candidate genes for several disorders due to the
function of the oxidases in terminating neurotransmit-
ter action. In 1972 Murphy and Wyatt reported that
platelet MAO activity in chronic schizophrenics is
lower than in controls. Subsequently, Wyatt et al.
(1973) suggested that low platelet activity is a genetic
marker in the pathogenesis of schizophrenia. Low
platelet MAO activity has also been associated with
alcoholism (Major and Murphy, 1978; Sullivan et al.,
1979) and subtypes of alcoholism (Sullivan et al., 1990;
von Knorring et al., 1991). Finally, there are several
reports regarding abnormal MAO activity in patients
with affective disorders (Samson et al., 1985; von Knor-
ring et al., 1985).
More recently, Brunner et al. (1993a) reported a
large pedigree in which several males were affected by
a syndrome of borderline mental retardation and ex-
hibited abnormal behavior, in particular aggressive
and violent behavior. The results of linkage analyses
were positive (LOD 4 3.69) for the region Xp11.23–
476 Parsian and Todd
p11.4 which contains the MAO genes. Urine analysis
showed a marked disturbance of monoamine metabo-
lism but platelet MAO-B activity was normal suggest-
ing a disturbance in MAO-A. In a later study (Brunner
et al., 1993b) a selective MAO-A enzyme deficiency was
demonstrated in affected males. Amplification and se-
quencing of the eighth exon of MAO-A gene detected
the presence of a C to T mutation at the nucleotide 936
in affected males and in carrier females but not in un-
affected males. The two point linkage analysis between
the clinical phenotype and the mutation in the MAO-A
gene resulted in a LOD score of 3.55 without recombi-
nation. These investigators concluded that there was a
complete and selective deficiency of MAO-A resulting
in illness in this family.
Based on the two reports mentioned above Lim et al.
(1994) argued that some of the behavioral characteris-
tics of the affected males are similar to symptoms ex-
hibited by bipolar patients during the manic phase of
their illness. Therefore, they hypothesized that either
the point mutation reported by Brunner et al. (1993b)
or other MAO-A mutations might contribute to bipolar
affective disorder. A sample of 57 unrelated bipolar pa-
tients and 59 matched normal controls of western Eu-
ropean extraction was genotyped with a microsatellite
repeat polymorphism in the MAO-A gene (Black et al.,
1991). The overall association between alleles of
MAO-A and affective disorder was significant (x2 4
24.63, P 4 0.0018). Allele A2 was less frequent and
allele A5 was more frequent in patients than controls
(x2 4 4.5, P 4 0.034; x2 4 3.58, P 4 0.058, respec-
tively). Kawada et al. (1995) repeated the Lim et al.
(1994) study in a Japanese population. The frequencies
of MAO-A alleles were compared between 58 bipolar
patients and 68 controls. There was an overall differ-
ence in allele frequencies (seven in their sample) be-
tween the two groups (x2 4 14.03, df 4 6, P 4 0.029).
However, in their sample, allele A4 was more frequent
and allele A5 less frequent in the bipolar group than
controls (x2 4 4.14, P 4 0.042; x2 4 6.31, P 4 0.012,
respectively). However, the differences were not sig-
nificant after correction for multiple tests. It was sug-
gested that bipolar disorder is in linkage disequilib-
rium with the (CA)n polymorphism of the MAO-A locus
but that the associated marker alleles have reversed in
the two ethnic populations (Kawada et al., 1995). Ru-
binsztein et al. (1996), studying a small sample of bi-
polar and matched neurological controls, reported an
excess of alleles 5 and 6 and a deficiency of allele 2 in
East Anglians of English origin.
Nöthen et al. (1995) used the haplotype relative risk
approach in a sample of 82 unrelated German bipolar
patients and their parents. No significant overall asso-
ciation was observed for MAO-A in their sample but
allele A5 showed a trend toward overrepresentation in
the patient alleles. They concluded that the positive
findings of Lim et al. (1994) and Kawada et al. (1995)
are due to chance or to undetected population stratifi-
cation. In a case-control study of Western Europeans
Craddock et al. (1995), using the three polymorphisms
of the MAO-A gene, also could not detect any evidence
for allelic association between any of the markers and
bipolar disorder for males or females.
However, Lim et al. (1995) repeated their study us-
ing additional polymorphisms in the MAO-A gene with
the same sample. Significant associations were found
only in female bipolars and were suggestive that
MAO-A may contribute to susceptibility to bipolar dis-
order in a subgroup of individuals. Similarly, in the
Rubinsztein et al. (1996) case-control meta-analysis,
significant associations were observed for female bipo-
lars only.
In order to evaluate the possible role of the MAO
region in susceptibility to affective disorders in an in-
dependent sample, we have genotyped bipolar patients
of North American non-Hispanic white origin with mi-
crosatellites repeats at both the MAO-A and MAO-B
genes. To overcome methodological problems of previ-
ous studies, we have used both normal control and pa-
rental alleles for comparison groups and have incorpo-
rated Monte Carlo simulations of allele frequency dif-
ferences to estimate significance values.
MATERIAL AND METHODS
Subjects
The subjects of this study consist of probands from
multiple incidence bipolar affective disorder pedigrees
and psychiatric normal controls. The identification, as-
certainment, and evaluation of this sample have been
detailed in a previous report (Parsian et al., 1995a). All
subjects are non-Hispanic white. The bipolar affective
disorder sample consists of 83 probands who met the
Research Diagnostic Criteria (RDC) and DSM-III-R di-
agnostic criteria. The probands are from families with
at least one other first degree relative with a diagnosis
of bipolar I, bipolar II, or schizoaffective/mania. The
normal group consists of 84 unrelated individuals, as-
certained through the Epidemiologic Catchment Area
(ECA) study, who met no DSM-III-R criteria for affec-
tive disorder, alcoholism, schizophrenia, or other psy-
chotic or drug use disorders. DNA from parents was
also available for 56 of the bipolar probands.
Genotyping
The genotyping method is detailed in previous pub-
lications (Parsian et al., 1995b) and briefly described
here. For the MAO-A simple sequence repeat polymor-
phism (SSRP), the polymerase chain reaction (PCR)
reactions were preheated to 94°C for 5 min, followed by
22 cycles each for 30 sec at 94°C, 45 sec at 61°C, 45 sec
at 72°C, and a final extension for 5 min at 72°C. For the
MAO-B SSRP, the PCR reactions were preheated at
94°C for 1.5 min, followed by 25 cycles each for 1 min at
94°C, 1 min at 58°C, 1 min at 72°C, and a final exten-
sion for 7 min at 72°C. The sizes of the fragments for
both MAO-A and MAO-B were identical to those pre-
viously reported (Black et al., 1991; Konradi et al.,
1992). In addition, we detected three new alleles for
MAO-B gene in the bipolar and normal control groups
(Table II). All genotypes were independently scored by
two investigators without knowledge of clinical group
status. To insure that allele scoring was identical for
all groups, selected samples from each group were co-
electrophoresed.

Statistical and Genetic Analyses
Relative-risk allele frequency comparisons were per-
formed using x2 or Fisher’s exact tests. Male and fe-
male allele frequencies were tested separately and to-
gether. For simultaneous comparisons of the frequen-
cies of multiple alleles (where many values are less
than five) the x2 probability distribution was estimated
using Monte Carlo simulation (Roff and Bentzen, 1989)
as implemented by a computer program developed by
George Carmody (Carleton University, Ottawa, On-
tario). The advantage of this computational approach is
that x2 simulations with estimated standard errors can
be completed in a matter of minutes without collapsing
cells with small numbers of observations. Similarly,
contingency table analyses were used to test for link-
age disequilibrium between alleles of the MAO-A and
MAO-B loci. This assessment was restricted to males
where chromosome X haplotypes were unambiguous.
Haplotype relative risk (HRR) analyses, were per-
formed on the entire sample and separately for males
and females as described by Falk and Rubinstein
(1987) except that individual alleles, rather than geno-
types, were contrasted (Parsian et al., 1995a). Power
calculations were completed as described in Cohen
(1988) for contingency table analyses.
RESULTS
Tables I and II present the allele frequencies for the
MAO-A and B genes for 83 bipolar and 84 normal in-
dividuals. To compare all alleles together, Monte Carlo
simulations of the x2 probability distribution were per-
formed (Roff and Bentzen, 1989) using a computer pro-
gram developed by Carmody (1995). For MAO-A the
overall comparisons between bipolar and normal con-
trol groups for both genders were not significant. For
MAO-B the overall comparisons for both genders were
highly significant (P < 0.003). To determine which al-
leles contribute to the MAO-B group difference, similar
analyses were conducted for each allele separately. For
the male group, no allele was significantly more fre-
quent in the bipolar group. There was a trend for a
decreased frequency of allele B2 and an increased fre-
quency of allele B7 (P < 0.1, corrected for multiple com-
parisons). For females, the frequencies of alleles B2
and B3 were significantly decreased in the bipolar
group (P < 0.008).
As described above, in case-control (relative risk)
analyses there are always questions about group
TABLE I. Allele Frequencies for MAO-A Gene in Bipolar Affective Disorder and Normal Control Groups Categorized by Gender
Group 1 2 3 4
Bipolar
Male 4 2 2 0
Female 3 16 2 7
Control
Male 3 8 3 0
Female 7 15 3 5
a
Allele sizes: 126-112 bp (1–8). There were no within group significant differences between males and females. Comparing all alleles together, the
differences between the two groups were not significant (males x26 4 5.90, P 4 0.333 ± 0.005; females x27 4 7.38, P 4 0.439 ± 0.005; combined x27 4 6.91,
P 4 0.439 ± 0.005). MAO-A, monoamine oxidase-A.
Manic Depression and Monoamine Oxidase 477
matching. Falk and Rubinstein (1987) developed an al-
ternative approach termed haplotype relative risk
(HRR) analysis in which both the parents of the pro-
band and the proband are genotyped and non-
transmitted parental alleles are considered as controls.
A modified haplotype relative risk analysis was per-
formed for the total available sample of bipolar pro-
bands and parents (56 proband-parent trios) and sepa-
rately by proband gender. Of the total 56 trios there
were 16 male probands and 18 female probands with
fully informative parents for MAO-A and B. The sum-
mary of these HRR analyses for MAO-B is depicted by
allele in Table III. There were no significant differences
between transmitted and non-transmitted alleles for
MAO-B. There were also no significant differences in
allele frequencies of bipolar probands used in the hap-
lotype relative risk analyses and the total bipolar pro-
band group (x2 4 1.876, P 4 0.934 ± 0.003 for MAO-B).
There were fewer transmitted allele 3s for MAO-B (6
vs. 8) but this difference was not significant. Given the
large effect size of the MAO-B case-control study (W 4
0.42 for females and W 4 0.51 for males), the combined
HRR sample had 80 percent power to detect a differ-
ence in frequency between transmitted and non-
transmitted alleles at a significance level of a 4 0.05
(Cohen, 1988). There were also no significant differ-
ences in transmitted and non-transmitted allele fre-
quencies for MAO-A.
The MAO-A and MAO-B genes are only 50 Kb (kilo-
base) apart, but are usually reported to be in linkage
equilibrium (Parsian et al., 1995b). There was no sug-
gestion of linkage disequilibrium between MAO-A and
MAO-B for both the normal control sample (P 4 0.79)
and the bipolar sample (P 4 0.89). Unambiguous hap-
lotypes were available for males in both the normal
control (n 4 42) and the bipolar (n 4 22) groups. There
were no significant differences between haplotype fre-
quencies in the two groups.
DISCUSSION
In this study, we used simple sequence repeat poly-
morphisms of the MAO-A and MAO-B genes to exam-
ine the relationship of these genes to bipolar affective
disorder. The polymorphism in the MAO-A gene is due
to a (CA)n repeat near the second exon (Black et al.,
1991). The second intron of MAO-B gene contains a
(GT)n repeat which produces a highly informative poly-
morphism (Konradi et al., 1992). We compared the al-
Allelesa
5 6 7 8 Total
3 19 2 0 32
19 52 2 1 102
8 16 4 0 42
9 37 4 0 80
478 Parsian and Todd

TABLE II. Allele Frequencies for MAO-B Gene in Bipolar Affective Disorder and Normal Control Groups Categorized by Gender
Allelesa
Group 1 2 3 4 5 6 7 8 12 Total
Bipolar
Male 0 0 4 9 5 6 8 0 0 32
Female 0 0 17 23 20 18 17 4 3 102
Control
Male 0 8 14 9 6 3 2 2 0 44
Female 1 8 31 10 16 8 5 1 0 80
a
Allele sizes: 184-160 bp (1–12). Within groups, there were no significant differences between males and females. Comparing all alleles together, the
differences between the two groups were highly significant (males x26 4 18.82, P 4 0.003 ± 0.001; females x28 4 31.64, P < 0.000 ± 0.000; combined
x28 4 47.01, P < 0.000 ± 0.000).

lele frequencies of these two genes between a bipolar
affective disorder group (n 4 83) and a normal control
group (n 4 84). The results of this relative-risk analy-
sis were highly significant for MAO-B but not for MAO-
A. The two study groups were unrelated non-hispanic
whites from the same geographical areas.
In order to further explore this result, we performed
haplotype relative risk (HRR) analysis on a subset of 56
bipolar probands on whom parent DNA was available.
For this analysis 34 out of 56 pairs of parents were fully
informative. The power of HRR method is identical to
that of RR method for equal group sizes. The differ-
ences between overall transmitted and non-
transmitted alleles were not significant for both
MAO-A and B. To advance our analysis further, we
compared haplotypes in males (because their phase is
known) for both bipolar and normal control groups. For
both groups the MAO-A and MAO-B loci were in link-
age equilibrium and there were no significant group
differences in haplotype frequencies. Given the differ-
ence in MAO-B results using the two methods, the
most parsimonious explanation is that we have de-
tected population stratification at the MAO-B locus for
case-control selected from outbred North American
whites of non-Hispanic origin.
Lim et al. (1994) reported an overall x2 value of 24.63
(P 4 0.0018) when comparing MAO-A allele frequen-
cies in 57 unrelated bipolar patients with 59 normal
controls of Western European extraction. It is interest-
ing that Kawada et al. (1995) detected a similar differ-
ence in a Japanese population. In the first study, allele
A5 was more frequent but in the second study allele A4
was more frequent. However, allele A5 is less frequent
in the Japanese groups. Lim et al. (1995) subsequently
reported a positive association between female bipolars
and two additional polymorphisms in MAO-A gene.
This study is not a replication, however, for at least two
reasons. First, their bipolar and control groups are the
same in both studies. Second, as stated in their report,
the two additional polymorphisms are in linkage dis-
equilibrium with the (CA) n repeat polymorphism.
Therefore, we would expect to see the same trend of
association for all three polymorphisms. More interest-
ingly, the study by Craddock et al. (1995) used a larger
sample (84 bipolars and 84 controls) from the same
general population as Lim et al. [1994, 1995] (United
Kingdom) and found no association with any of the
three polymorphisms of the MAO-A gene. Collapsing
results across three studies Rubinsztein et al. (1996)
reported an association of MAOs and bipolar disorder
for females only. Nöthen et al. (1995) is the only study
in which the HRR method was used for MAO-A and
their negative results are similar to ours.
In summary, our current evidence does not support
the association of alleles of the MAO-A or MAO-B loci
with bipolar affective disorder but does suggest that
population stratification is an important source of bias
among non-Hispanic whites from North America. Cau-
tion should be used in interpreting the results of case-
control genotyping studies when confirmation is lack-
ing from studies using different theoretical approaches.

TABLE III. Haplotype Relative Risk Analysis for MAO-B

Male probanda Female probandb Totalc
Maternal Maternal Maternal Maternal Maternal Maternal
transmitted nontransmitted transmitted nontransmitted transmitted nontransmitted
Allele number (n 4 16) (n 4 16) (n 4 18) (n 4 18) (n 4 34) (n 4 34)
3 2 3 4 5 6 8
4 7 7 4 4 11 11
5 3 2 3 5 6 7
6 2 0 2 1 4 1
7 2 2 3 2 5 4
8 0 0 0 1 0 1
9 0 1 0 0 0 1
10 0 1 0 0 0 1
12 0 0 2 0 2 0
a
Maternal transmitted vs. nontransmitted (n 4 32); x26 4 4.400, P 4 0.766 ± 0.004.
b
Maternal transmitted vs. nontransmitted (n 4 36); x26 4 4.144, P 4 0.787 ± 0.004.
c
Total maternal transmitted vs. nontransmitted (n 4 68); x28 4 7.274, P 4 0.558 ± 0.005.

ACKNOWLEDGMENTS
We thank Dr. Ted Reich for providing additional bi-
polar samples, Lorienne Fisher for technical support,
and Jan Konrad for manuscript preparation. This work
was supported in part by the National Institute on Al-
cohol Abuse and Alcoholism (AA09515) and the Na-
tional Institute of Mental Health (MH 31302).
REFERENCES
Black GCM, Chen Z-Y, Craig IW, Powel JF (1991): Dinucleotide repeat
polymorphism at the MAO-A locus. Nucleic Acid Res 19:689.
Brunner HG, Nelen MR, van Zandvoort P, Abeling NGG, van Gunnip AH,
Wolters EC, Kuiper MA (1993a): X-linked borderline mental retarda-
tion with prominent behavioral disturbance: phenotype, genetic local-
ization, and evidence for disturbed monoamine metabolism. Am J Hum
Genet 52:1032–1039.
Brunner HG, Nelen M, Breakfield XO, Ropers HH, van Oost BA (1993b):
Abnormal behavior associated with a point mutation in the structural
gene for monoamine oxidase A. Science 262:578–580.
Carmody G (1995): Personal communication.
Cohen J (1988): Statistical Power Analysis for the Behavioral Sciences, 2nd
ed. Lawrence Erlbaum Associates, Hillsdale, NJ.
Craddock N, Daniels J, Roberts E, Rees M, McGuffin P, Owen MJ (1995):
No evidence for allelic association between bipolar disorder and mono-
amine oxidase A gene polymorphisms. Am J Med Genet 60:322–325.
Falk CT, Rubinstein P (1987): Haplotype relative risks: an easy reliable
way to construct a proper control sample for risk calculation. Ann Hum
Genet 51:227–233.
Kawada Y, Hattori M, Dai XY (1995): Possible association between mono-
amine oxidase A gene and bipolar affective disorder. Am J Hum Genet
56:335–336.
Konradi C, Ozelius L, Breakfield XO (1992): Highly polymorphic (GT)n
repeat sequence in intron II of the human MAO-B gene. Genomics
12:176–177.
Lim LCC, Powell JF, Murray R, Gill M (1994): Monoamine oxidase A gene
and bipolar affective disorder. Am J Hum Genet 54:1112–1124.
Lim LCC, Powell J, Sham P, Castle D, Hunt N, Murray R, Gill M (1995):
Evidence for a genetic association between alleles of monoamine oxi-
dase A gene and bipolar affective disorder. Am J Med Genet 60:325–
331.
Manic Depression and Monoamine Oxidase 479
Major LF, Murphy DL (1978): Platelet and plasma amine oxidase activity
in alcoholic individuals. Br J Psychiatry 132:528–554.
Murphy DL, Wyatt RJ (1972): Reduced platelet monoamine oxidase activ-
ity in chronic schizophrenia. Nature 238:225–228.
Nöthen MM, Eggerman K, Albus M, Borrman M, Rietschel M, Körner J,
Maier W, Minges J, Lichtermann D, Franzek E, Weigelt B, Knapp M,
Propping P (1995): Association analysis of the monoamine oxidase A
gene in bipolar affective disorder by using family-based internal con-
trols. Am J Hum Genet 57:975–977.
Parsian A, Chakraverty S, Todd Rd (1995a): Possible association between
the dopamine D3 receptor gene and bipolar affective disorder. Am J
Med Genet 60:234–237.
Parsian A, Suarez BK, Tabakoff B, Hoffman P, Ovchinnikova L, Fisher L,
Cloninger CR (1995b): Monoamine oxidases and alcoholism: I. Studies
in unrelated alcoholics and normal controls. Am J Med Genet 60:409–
416.
Roff DA, Bentzen P (1989): The statistical analysis of mitochondrial DNA
polymorphisms: Chi-square and the problem of small samples. Mol Biol
Evol 6:539–543.
Rubinsztein DC, Leggo J, Goodburn S, Walsh C, Jain S, Paykel ES (1996):
Genetic association between monoamine oxidase A microsatellite and
RFLP alleles and bipolar affective disorder: analysis and meta-
analysis. Hum Mol Genet 5:779–782.
Samson JA, Gudeman JE, Schatzberg AF, Kizuka PP, Orsulak PJ, Cole
JO, Schildkraut J (1985): Toward a biochemical classification of de-
pressive disorders VIII. Platelet monoamine oxidase activity in sub-
types of depressions. J Psychiat Res 19:547–555.
Sullivan JL, Cavenar JO, Maltbie AA, Lister P, Zung WWK (1979): Famil-
ial biochemical and clinical correlates of alcoholics with low platelet
monoamine oxidase activity. Biol Psychiatry 14:385–394.
Sullivan JL, Baenziger JC, Wagner DL, Rauscher FP, Nurnberger JI,
Holms JS (1990): Platelet MAO in subtypes of alcoholism. Biol Psy-
chiatry 27:911–922.
von Knorring L, Perris C, Oreland L, Eisemem M, Holmgfren S, Perris H
(1985): Morbidity risk for psychiatric disorders in families of probands
with affective disorders divided according to levels of platelet MAO
activity. Psychiatry Res 15:271–279.
von Knorring A-L, Hallman J, von Knorring L, Oreland L (1991): Platelet
monoamine oxidase activity in type I and type II alcoholism. Alcohol
26:409–416.
Wyatt RJ, Murphy DL, Belmarker R, Cohen S, Donnelly CH, Pollin W
(1973): Reduced monoamine oxidase activity in platelets: A possible
genetic marker for vulnerability to schizophrenia. Science 179:916–
918.