We found that reporting a high self-perceived weight, among persons who actually were overweight or obese, was more frequent in women than men, in obese than overweight persons, and in persons of high vs. low SES. The association between appropriate perception of one's own weight and SES was fairly similar for all three SES indicators (i.e. education, occupation or income) when these SES indicators were considered in isolation, but the association was no longer significant for income when all three indicators were considered together. This suggests that any SES indicator is a useful marker when information is not available on the other SES indicators but that income does not add substantial information when a person's education and occupational status are known. Correspondingly, the OR for the association between weight perception and SES ranged between 2.7 and 4.3 when education, income or occupation were considered in isolation and reached 6.1 based on a score combining education and occupation and 6.9 based on a score combining these three SES indicators.
Among individuals with a normal BMI, the majority had an appropriate weight perception, i.e. they perceived their weight as being 'good'. It is worth noting however, that high SES individuals more frequently overestimated their weight (i.e. reported a high self-perceived weight when their actual weight was normal). This is consistent with previous reports showing that high SES persons, particularly women, tend to be less satisfied with their weight , and are more likely to overestimate their weight and/or underestimate what a normal weight should be [1, 3, 5].
Obese participants perceived their weight as too high more often than overweight participants (81% vs. 46%, respectively). These results are in line with previous studies showing that obese individuals are less likely to misclassify their weight status as compared to overweight individuals [5, 6, 8]. Consistent with previous reports [2, 6, 8], women were more likely than men to appropriately report a high self-perceived weight, regardless of SES. These associations of appropriate weight perception with sex (women vs. men) and weight status (obese vs. overweight) can also underlie different body ideals in women than men (as women have smaller body ideals than men) [1, 25] and larger readiness to acknowledge excess weight among obese than overweight persons [5, 6].
We found that overweight/obese individuals of high SES were more likely to have an appropriate perception of their excess weight. The association between appropriate self-perception of weight and high SES has been previously documented [1, 3, 5, 6, 9], and may be attributed to differences in defining 'normal' or 'ideal' body weight across SES groups [11, 19, 26]. Moreover, individuals with high SES tend to have greater access to health information that promotes healthy lifestyles, thus rendering these individuals more weight-conscious  and more prone to recognize excess weight along the standard overweight and obesity categories. This suggests that interventions that aim to address individuals' weight perceptions can be specifically targeted at low SES groups. Interestingly, both men and women had a more appropriate perception of their weight if they were of high than low SES. The fact that the prevalence of obesity is higher in men of high than low SES suggests, however, that factors other than weight perception underlie the direct obesity-SES relationship among men in the Seychelles. Inversely, the prevalence of obesity is higher among women of low than high SES in the Seychelles, which is consistent with the social gradient in weight perception. More generally, the social pattern in the prevalence of obesity in men and women may also be consistent with a trend for women to increasingly value a lean weight in developing countries, while men may value a heavy weight as a sign of physical dominance and prowess . These observations emphasize the potential role of beliefs and values related to one's weight when assessing social trends in obesity in populations [9, 18].
While previous reports have shown that the SES-obesity relationship is more apparent when using education and occupation as SES indicators , there remains some controversy as to which of the three SES indicators is the most strongly associated with appropriate weight perception. One previous study showed that education was a stronger contributor to body dissatisfaction than occupation  (of note, body dissatisfaction is not fully equivalent with appropriate body perception: a person (e.g. a man) can be aware of being too heavy from a health perspective but still be satisfied with a heavy weight). In our study, we found that any of the three considered SES indicators, i.e. education, occupation, and income, were fairly similarly associated with appropriate weight perception. This means that for practical purposes, any of the SES indicators can be useful to anticipate appropriate perception of a person's own weight, if only one such indicator is available. However, analysis including all three indicators together showed that only education and occupation were independently associated with appropriate weight perception, suggesting that information on income is not necessary if education and occupation are known. This message is also conveyed by the finding that the association was much stronger based on a score combining all three single SES components (OR = 6.9), or a score based on education and occupation (OR = 6.1), than based on any single SES variable (OR between 2.7 and 4.3). Of note, the three SES indicators identified largely different persons and only 16% of individuals placed in the 'high' SES category based on any of the three SES indicators had a high level for all three indicators. It can be observed that quintiles or sixtiles of the two considered overall SES scores imply smaller numbers in the outer score categories as compared to the numbers of persons in the outer categories of the trichotomized scores based on one SES component (education, occupation, or income). The use of more stringent categories in the case of the overall scores vs. the one-component scores may explain the higher ORs in the former than the latter scores. Since our data do not allow us to generate scores of education and occupation that are composed of more than 3 categories, we cannot simulate the ORs that would arise from having narrower categories for these single component scores. Overall, our figures suggest that the one-component scores perform well (particularly the scores based on education and occupation) but factoring knowledge from all three indicators may possibly slightly improve the prediction.
The finding that weight perception was more strongly associated with education and occupation vs. income may reflect differences in health literacy across educational groups . Individuals with a high education may be more able to interpret and use information related to 'healthy' weight and weight control measures  compared to individuals with a lower education.
Our findings provide further evidence on phenomenological mechanisms that can fuel the obesity epidemic in the population in this region, and clues to guide interventions to prevent and control overweight and obesity. At a clinical level, our data suggest that health professionals should systematically clarify their patients' beliefs related to their own weight and address the identified related misbelieves. At a population level, our findings suggest that it is important to gather information on weight perception in populations according to various dimensions (gender, SES; etc) in order to guide information campaigns and other culturally sensitive interventions related to a healthy weight.
Strengths of this study include the population-based design and the availability of three SES indicators reflecting three main domains, education, occupation and income. Moreover, weight and height were actually measured, in contrast to a number of similar studies that have relied on self-reported values. On the other hand, the cross-sectional design of this study limits inference on the direction of the associations (i.e. whether low SES leads to poor weight perception or whether poor weight perception -possibly a marker of other poor cognitive skills- leads to poor SES outcomes). Also, as we did not have data on ideal body size, health awareness or cognitive skills (e.g. abstraction skills), we could not disentangle whether differences in appropriate weight perception corresponded to differences in cognitive skills, healthy weight awareness, or body size ideals.