Gender Differences in the Relationship between Personality Dimensions and Relative Body Weight
Obesity Research Center, St. Luke's-Roosevelt Hospital Center, Columbia University College of Physicians and Surgeons, 1090 Amsterdam Avenue 14th Floor, New York, NY 10025. E-mail firstname.lastname@example.org
Objective: The relationship between body mass index (BMI: kg/m2) and personality seems to differ for men and women, although these effects may be driven by the extremes of the BMI distribution. It is unclear whether these associations exist for most individuals in the relatively normal range of BMI scores, excluding the thinnest and heaviest extremes in the population. We tested the association of BMI with neuroticism, extraversion, and psychoticism with a trimmed BMI sample.
Research Methods and Procedures: Using a cross-sectional design, we tested the association of BMI with the aforementioned psychological variables in a British population-based sample. Participants were 7889 adult men and women (30 to 50 years old) selectively sampled from four counties in west England. Participants reported their height and weight and completed the Eysenck Personality Inventory (EPQ). We tested the association of BMI with the EPQ subscales among individuals with BMI ≥19.16 kg/m2 and ≤37.78 kg/m2, i.e., the approximate 5th and 95th percentiles.
Results: Despite elimination of extreme BMIs, different associations between BMI and EPQ subscales emerged for men and women. Among women, increasing BMI was significantly associated with increased neuroticism and reduced extraversion. Among men, increasing BMI was associated with increased extraversion and psychoticism. In all cases, the magnitude of the association was very small.
Discussion: Increasing BMI was associated with potentially poorer adjustment among women but better adjustment among men. These findings are consistent with recent reports and, taken together, suggest that these patterns are not accounted for solely by the extremes of the BMI distribution.
Obesity is an increasingly prevalent public health problem worldwide that is associated with numerous medical complications and increased all-cause mortality (1). In contrast, an understanding of the psychological correlates, causes, and consequences of obesity remains unclear (2). On balance, obesity does not seem to be characterized by a distinct psychological profile or drastically elevated psychopathology (3, 4, 5). However, several caveats must be considered when interpreting the literature (2). First, many studies have relied on small, convenience, or clinical samples—the results of which may not generalize to the broader population. Second, relatively few studies have made use of psychometrically well-validated instruments or measures of psychopathology. Thus, strong conclusions that there are no psychological traits related to obesity may be unfounded.
Certain community- or population-based studies suggest that gender may moderate the relationship between psychological characteristics and body mass index (BMI: kg/m2) (6, 7, 8). Thus, greater body mass may show a covariance with slightly poorer adjustment in women but not necessarily in men. However, this finding comes from few studies and awaits replication. Moreover, this finding emerges when analyzing the full spectrum of BMI scores—including the thinnest and heaviest individuals in the population. Because there often seems to be an exponential increase and/or step-function for the deleterious effect of increased BMI on certain variables (9), it is unclear whether these associations exist for most individuals in the normal range of BMI scores, that is, in anything less than the extremes of the distribution.
Psychological well-being is clearly a complicated construct. Measures of personality or temperament have been shown to be robustly heritable and to have clear associations with psychopathology. This study tested the association of BMI with neuroticism, extraversion, and psychoticism among women and men. The study addressed the aforementioned issues by studying a large population-based sample, using an extensively validated assessment tool, and trimming the tails of the BMI distribution (see Research Methods and Procedures section). Because the association between BMI and psychological variables may differ for men and women (2), we tested the relationship between BMI and three Eysenck Personality Inventory (EPQ) subscales (i.e., neuroticism, extraversion, and psychoticism) in gender-specific analyses. We predicted that increased BMI would be associated with poorer psychological adjustment in women but not in men. We anticipated that the magnitude of effects would be small based on previous studies.
Research Methods and Procedures
In the United Kingdom, practically all the names and addresses of the British population are contained in the lists of general practitioners (10). We used this resource to contact a large initial cohort (N = 100,000), between 30 and 50 years of age, through the patient registers of general practices in four counties in west England. Data from general practitioner registers were acquired electronically and were used to establish a database from which EPQ forms with the appropriate addresses were generated. Subsequently, EPQ data from returned forms and the names and addresses of siblings were entered into a database. The study was approved by relevant local and regional committees for research ethics.
We received responses from a total of 88,141 individuals, including 20,427 unique sibships. Most of the sibships consisted only of the proband and one additional sibling (15,259 of 20,427 = 75%) but larger sibships of sizes 3, 4, 5, and 6 were also identified (N = 4146, 838, 151, and 23, respectively). The total sample, thus, consisted of 34,580 sib-pairs in 20,427 sibships. Forty-six percent of the total respondents provided no information about other relatives.
For this study, we selected a subset of respondents to ensure that the sample consisted only of subjects in the nonextreme or common range of BMI, measurement errors due to transcription or computer-reading errors were eliminated or minimized, and only one sibling per sibship was used to ensure independent observations. To achieve this, subjects meeting any of the following conditions were excluded: missing data for height or weight, a height above 183 cm or below 151 cm, or a weight below 51 kg or above 108 kg. These numbers represent approximately the 5th and 95th population percentiles for these variables. We restricted our analyses to only one child per family by selecting the first sibling for whom complete information was available. This yielded a final sample of 7889 individuals, among whom ∼52% were women.
The full 90-item revised EPQ (11) was sent to all study participants. The EPQ scales of neuroticism, extraversion, and psychoticism were constructed from the individual items according to questionnaire guidelines. According to Eaves et al. (12), traits associated with the neuroticism scale include anxiety, depression, guilt feelings, low self-esteem, tension, shyness, and being emotional; traits associated with the extraversion scale include sociability, liveliness, activeness, assertiveness, sensation seeking, dominance, and being adventuresome; and traits associated with the psychoticism scale include aggressiveness, coldness, egocentricity, being impersonal, impulsiveness, antisocial, creative, and tough-mindedness. The EPQ was used not only for convenience but also for its long-standing tradition in the psychological and psychiatric literature. The instrument has been extensively validated and widely used around the world to measure these personality constructs. Angular transformations were applied to the neuroticism and extraversion scales (12), and square-root transformations were applied to the psychoticism scale scores.
All subjects were asked to report their height and weight, which was used to derive BMI. Although self-reported heights and weights have some small but systematic biases, they are quite reliable and tend to correlate with their measured counterparts (r > 0.95) (13).
Partial correlation coefficients tested the age-adjusted association between BMI and the three EPQ subscales. To test the independent association of BMI with each EPQ subscale (i.e., controlling for the other subscales), multiple regression analyses were performed. Specifically, we regressed BMI onto the following predictors: age, neuroticism scale, extraversion scale, and psychoticism scale. Separate regression models were performed for men and women. (When testing interactions between gender and EPQ subscales [data not presented], significant interactions emerged with the neuroticism and extraversion subscales. These findings provided additional justification for conducting gender-specific analyses.)
Descriptive statistics for the subjects are shown in Table 1. Partial correlations between BMI and EPQ subscales are presented in Table 2. Results suggested distinct gender differences, although all effect sizes were small. Among women, greater BMI was associated with increased neuroticism (r = 0.09) and decreased extraversion (r = −0.06; p < 0.0001 in both cases). Among men, greater BMI was associated with increased extraversion (r = 0.07) and psychoticism (r = 0.07; p < 0.0001 in both cases). For all three EPQ subscales, correlations were significantly different between men and women (p < 0.01).
Table 1. Sample characteristics
|Neuroticism||12.94 (5.62)||10.22 (5.91)|
|Extraversion||11.90 (5.42)||11.59 (5.42)|
|Psychoticism||2.30 (2.03)||3.29 (2.46)|
|Age (years)||40.02 (6.25)||40.38 (6.18)|
|BMI (kg/m2)||25.47 (3.88)||25.16 (3.24)|
|Height (cm)||165.91 (6.61)||176.08 (5.56)|
|Weight (kg)||69.95 (10.34)||77.98 (10.53)|
Table 2. Partial correlations of BMI with EPQ subscales (adjusting for age)
The results of the partial correlation analyses were confirmed by multiple regression analyses (Table 3). Among women, increasing BMI was associated with increased neuroticism (p < 0.0001) and reduced extraversion (p = 0.02). Among men, increasing BMI was associated with increased extraversion (p < 0.0001) and psychoticism (p < 0.0001).
Table 3. Results of multiple regression analyses
|Model statistics||F = 15.67, p < 0.0001, R2 = 0.02||F = 18.83, p < 0.0001, R2 = 0.02|
The relationship between relative body weight and dimensions of personality was distinctly different for men and women in this population-based study from England. This finding is consistent with recent reports in the literature (6, 7) and was observed despite the fact that extreme BMI scores were excluded from analyses. Thus, as Friedman and Brownell (2) intimated, gender seems to be a consistent moderator of BMI–psychological functioning relationships. In contrast, the observed effects were very small. These results suggest that in people who are not too thin or too heavy, variability in BMI has a small association with variability in neuroticism, extraversion, and psychoticism, and this seems to vary by gender. One avenue for future research is to map out causal pathways and underlying mechanisms for these associations. A variety of potential mechanisms, including genetic, physiological, and psychosocial, have been suggested (14). The results of this study caution against haphazardly pooling men and women and suggest the potential benefits of developing gender-specific models when testing mechanisms.
This work was supported in part by Wellcome Trust and National Institutes of Health Grants K08MH01530, R01DK51716, P30DK26687, and P30DK56336.