‘Fat, four‐eyed and female’ 30 years later: A replication of Harris, Harris, and Bochner's (1982) early study of obesity stereotypes

Abstract

Objective

This study aimed to replicate Harris, Harris, and Bochner's (1982) early experiment on obesity stereotyping to examine whether negative obesity stereotypes persist and in what form.

Method

A sample of psychology students (N = 506) read a description of a target described as female or male, overweight or average weight, and wearing glasses or not, who they subsequently rated on 12 descriptors.

Results

Overweight targets were rated as significantly less active, assertive, athletic, attractive, happy, hardworking, masculine, popular, and successful than average weight targets. This negative stereotype effect of target weight was much larger than effects observed for sex or wearing glasses. There were no differences in effect sizes for target weight between this study and the original study.

Conclusions

It was concluded that the negative obesity stereotypes reported by Harris et al. have persisted over a 30‐year period, despite the fact that people who are overweight are no longer a minority. Efforts are needed to challenge negative stereotyping of this group. Future research could examine why stereotypes of overweight people are resistant to change.

Key words:

• anti‐fat attitudes
• obesity
• stereotypes
• weight stigma

What is already known about this subject?

• Anti‐fat attitudes and weight stigma are widespread.

• There is little evidence that ‘sizeism’ is waning, despite a reduction in some other forms of discrimination.

• Anti‐fat attitudes have persisted over time in samples such as children and health professionals.

What this topic adds

• This Australian replication study demonstrates that negative obesity stereotypes have persisted over a 30‐year period.

• The study also shows that obesity stereotypes are more negative than stereotypes of other physical characteristics

• There was no increase in the strength of obesity stereotypes over a 30‐year period, and there were no sex differences in the strength of obesity stereotypes

In the early 2000s, obesity was declared as a significant public health issue in Western society (Oliver, 2006). Subsequently, the ‘burden’ of obesity on health‐care systems and society has become a prominent topic among health professionals, politicians, and policy makers over the last decade (Oliver, 2006; Saguy & Riley, 2005). The physical risks associated with obesity are well known (see Proietto & Baur, 2004; Renehan, Tyson, Egger, Heller, & Zwahlen, 2008), but the psychological impact of weight stigma experienced by heavier individuals is often overlooked (Puhl, Heuer, & Brownell, 2010).

Negative attitudes and beliefs about people perceived as overweight or obese (‘anti‐fat attitudes’) and the social disapproval attached to fatness (‘weight stigma’) are not only widespread but socially accepted (Lewis & Van Puymbroeck, 2008) and even encouraged in popular media and the news media (Boero, 2007; Harrison, 2000; Kim & Willis, 2007; Latner, Rosewall, & Simmonds, 2007; Lin & Reid, 2009). Phrases such as ‘obesity outbreak’, which connote a strong undertone of contagion and something unwelcome, are frequently found in research papers, and similar themes have been popularised in the media (Saguy & Riley, 2005). These messages convey negative and derogatory information beyond broad reference to obesity prevalence. Campos (2004) and Oliver (2006) argued that such messages have led to an unreasonable fear of fatness and vilification of obese persons in Western society. Weight stigma is not unique to Western society; an investigation across 10 countries by Brewis, Wutich, Falletta‐Cowden, and Rodriguez‐Soto (2011) found evidence for increasing globalisation of anti‐fat attitudes and thin ideals.

A review of the weight stigma literature (Puhl & Heuer, 2009) indicated that obese persons are stigmatised by employers, health professionals, educators, romantic partners, family members, and friends. Obese persons experience social disadvantages in a range of domains, with less access to educational and career opportunities, worse treatment in health care, and a higher likelihood of being bullied, teased, or rejected than their thinner counterparts. Brochu and Morrison (2007) found that adults reported being less likely to socially engage with an obese peer compared with an average weight one, and similar findings have been reported for children (Bell & Morgan, 2000) and adolescents (Greenleaf, Chambliss, Rhea, Martin, & Morrow, 2006). Obese people have been rated as more disgusting than several other stigmatised groups including homeless people, homosexuals, politicians, welfare recipients, and people with mental illness (Vartanian, 2010a). Wang, Brownell, and Wadden (2004) argued that even overweight persons internalise negative societal messages about their weight.

Experienced weight stigma has been linked to body image dissatisfaction, low self‐esteem, depression and psychiatric disorders, including symptoms such as suicidal ideation (Eisenberg, Neumark‐Sztainer, & Story, 2003; Puhl & Heuer, 2009). Furthermore, the bulk of research demonstrates that weight stigma does not motivate weight loss (Li & Rukavina, 2009; Wott & Carels, 2010; an exception is Latner, Wilson, Jackson, & Stunkard, 2009).

There is little evidence that ‘sizeism’ is waning, despite a reduction in racism and sexism (Blaine & McElroy, 2002; Himes & Thompson, 2007). While the portrayal of other traditionally stigmatised groups in media such as television has become more positive in recent times, weight‐based stigmatisation continues to proliferate (Fikkan & Rothblum, 2012). In particular, current characterisations of women in the media may mean that heavy women are judged more harshly than heavy men. Thin female characters are over‐represented on prime‐time television and are portrayed as attractive and successful role‐models to be emulated, while fat women appear less frequently and are often depicted as targets of humour (Fikkan & Rothblum, 2012; Himes & Thompson, 2007). In contrast, overweight men receive little punishment or rejection, contributing to a thin ideal ‘double standard’ (Himes & Thompson, 2007).

It would seem that an increasingly overweight population has failed to buffer the stigma associated with obesity. Latner and Stunkard (2003) replicated Richardson, Goodman, Hastorf, and Dornbusch's (1961) early study of weight bias in which over 500 children ranked their preferences for six pictures of children that were uniform, with exception of one characteristic, such as obesity. The original study showed a clear preference for the child depicted as healthy, with the child depicted as obese consistently being ranked last. Latner and Stunkard confirmed these results, and found that the difference between the healthy child and the obese child was even more polarised than in the original study, suggesting an increase in weight bias over time. More recently, Tomiyama et al. (2015) assessed changes in implicit (unconscious) and explicit anti‐fat biases over time by comparing bias scores from ObesityWeek 2013 Conference attendees to those for 2001 conference attendees. Explicit anti‐fat attitudes were assessed using the specific stereotype matched pairs of lazy/motivated, stupid/smart, worthless/valuable, extremely bad/extremely good. Findings suggested that implicit anti‐fat attitudes decreased from 2001 to 2013 but explicit anti‐fat bias for general bad feelings and laziness increased. However, these studies focused on specific samples (children, health professionals) and further studies are needed. Studies by Andreyeva, Puhl, and Brownell (2008) and Puhl, Andreyeva, and Brownell (2008) concluded that the level of weight‐based discrimination is relatively close to gender and race discrimination, particularly for women. Andreyeva et al. (2008) noted that there was a statistically significant increase in weight‐based discrimination from 1995–1996 to 2004–2006.

Given the rising prevalence of obesity in Australia (Australian Bureau of Statistics (ABS), 2013) and the impact of weight stigma on obese persons (Eisenberg et al., 2003; Puhl & Heuer, 2009), it is important to understand current perceptions of obese persons in the Australian context. The current study replicated Harris, Harris, and Bochner's (1982) early experimental study of weight‐based stereotypes. This Australian study was selected for replication on the basis that it not only examined obesity stereotypes but also stereotypes of other physical characteristics, i.e., glasses wearing and sex, thus allowing for comparison of the presence and strength of stereotypes across physical characteristics.

Harris et al. (1982) found that overweight targets significantly differed from average weight targets on an a priori obesity stereotype composite (the linear combination of low ratings for active, intelligent, hardworking, attractive, popular, successful, athletic, and appropriately sex typed and high ratings for outgoing). Overweight targets received lower mean ratings across all descriptors, including ‘outgoing’, which was expected to be higher consistent with the ‘jolly fat person’ stereotype. However, only differences for active, attractive, and athletic reached statistical significance. In contrast, wearing glasses was associated with a mix of positive and negative characteristics. Glasses‐wearing targets and non‐glasses wearing‐targets significantly differed on an a priori glasses stereotype composite (the linear combination of high ratings on intelligent, hardworking, and successful, and low ratings on active, outgoing, attractive, popular, athletic, and sex appropriate). Means differed in the expected direction for all descriptors except appropriately sex typed, however, none of the single descriptors showed a statistically significant glasses main effect. Sex of target did not influence mean ratings for any of the characteristics apart from feminine and masculine, which differed in the expected direction for female and male targets. Taken together, the findings of Harris et al. demonstrated that being overweight was evaluated more negatively than wearing glasses or being female.

Note that Harris et al. (1982) did not find a significant interaction between target weight and target sex, suggesting that overweight women were not more negatively stereotyped than overweight men. However, subsequent findings on sex and weight bias have been mixed (see Roehling, 2012). The current trend in media representations of women described earlier (Fikkan & Rothblum, 2012; Himes & Thompson, 2007) may mean that sex differences in weight‐based stereotyping have emerged more recently.

The Current Study: Aims and Hypotheses

The aim of the current study was to replicate Harris et al.'s (1982) early experimental study of stereotyping based on weight, glasses wearing, or sex to examine whether there is a continuation of negative obesity stereotypes despite the fact that people with overweight are no longer a minority. The contribution of this replication study is in identifying whether obesity stereotypes persist 30 years later, and in what form, and whether such stereotypes remain more negative than stereotypes of other physical characteristics. It was hypothesised that the obesity stereotypes reported by Harris and colleagues in the early 1980s have persisted over time such that:

Hypothesis 1: Overweight targets are more negatively stereotyped than average weight targets.

Hypothesis 2: Overweight is more negatively stereotyped than wearing glasses or being female.

While Harris et al. provided an early perspective on the contemporary issue of weight bias, the study was limited to the extent that it did not report effect sizes for the differences between average weight and overweight targets. We computed effect sizes for both the current study and for Harris et al. to enable comparison of results across the two studies and across the three physical characteristics of weight, glasses wearing, and sex. We also computed effect sizes for differences in ratings of average weight targets across the two studies, and separate effect sizes for differences in ratings of overweight targets across the two studies. This allowed for examination of whether ratings of average weight targets and overweight targets had changed over the past 30 years.

Method

Participants

The participants were 506 undergraduate psychology students (24.3% male, 74.5% female, 2.2% not specified) at an Australian university (37.5% were online students). The average age was 25.66 years (SD = 9.55). The average body mass index (BMI) was 23.74 (SD = 4.90). The majority of participants (56.3%) were within the ‘healthy weight’ range for BMI; 20.8% were overweight and 8.5% were obese. A small proportion was underweight (8.1%) or did not specify weight and/or height. Over a third of participants (35.8%) reported a need for glasses. Participation was voluntary and there were no inducements. The original study (Harris et al., 1982) was based on 154 undergraduate psychology students at an Australian university, with a similar demographic profile.

Design

As per Harris et al. (1982), the experiment was based on a between‐subjects design with eight conditions: 2 (weight: overweight target vs average weight target) × 2 (glasses: target wears glasses vs glasses not mentioned) × 2 (sex: female target vs male target). Participants received a description of a target manipulated for weight, glasses, and sex, and subsequently rated the target on 12 descriptors relating to common stereotypes associated with obesity, glasses, or sex (see Measures). Given the inclusion of an online cohort, conditions were allocated to tutorial groups (first year psychology classes of approximately 22 students); on‐campus students received condition instructions and measures in hard copy; online tutorial groups received condition instructions and measures via an assigned survey link.

Measures

Stereotype ratings (Harris et al., 1982)

Participants in all conditions rated their target on twelve 7‐point semantic differential rating scales with the following descriptors as anchors: assertive to unassertive, active to passive, unintelligent to intelligent, hardworking to lazy, reserved to outgoing, happy to unhappy, feminine to unfeminine, unattractive to attractive, unpopular to popular, unsuccessful to successful, athletic to unathletic, masculine to unmasculine. As per Harris et al., order and polarity of descriptors was randomly determined; ratings were reversed scored so higher scores corresponded to higher ratings on positive attributes, e.g., intelligent, rather than unintelligent. Composite variables were derived from the ratings as per Harris et al. and included:

1. ‘Sex‐appropriate’: computed as masculine rating minus feminine rating for male targets or feminine rating minus masculine rating for female targets. Higher scores reflect masculinity for men and femininity for women.

2. ‘Masculine’: computed as masculine rating minus feminine rating. Higher scores reflect higher masculinity.

3. ‘Halo’: computed from the average of ratings for assertive, active, intelligent, hardworking, outgoing, happy, attractive, popular, successful, athletic, and sex appropriate. Higher scores reflect higher ratings on the composite of all positive descriptors.

4. ‘Obesity stereotype’: computed from the average of ratings for active, intelligent, hardworking, attractive, popular, successful, athletic, and sex appropriate, minus the rating for outgoing. Lower scores reflect endorsement of the a priori obesity stereotype from Harris et al. (1982), i.e., overweight targets are rated lower on all descriptors except outgoing.

5. ‘Glasses stereotype’: computed from the average of ratings for intelligent, hardworking, and successful minus the ratings for active, outgoing, attractive, popular, and athletic. Higher scores reflect endorsement of the a priori glasses stereotype from Harris et al.

6. ‘Maleness stereotype’: computed from the average of ratings for assertive, active, intelligent, and athletic minus the ratings for outgoing and successful. Higher scores reflect endorsement of the a priori maleness stereotype from Harris et al.

As the stereotype composites each contained different numbers of descriptors, with some descriptors added and others subtracted, composites were weighted as per Harris et al. to ensure consistency.

Motivation to control prejudiced reactions (Dunton & Fazio, 1997)

This scale measures individual differences in efforts to control the expression of prejudice. Respondents indicate the extent to which they agree or disagree with statements on a scale from −3 (strongly disagree) to +3 (strongly agree). Items are reverse scored so that higher scores reflect higher motivation to control prejudice. In the current study, we used one of two subscales from the total scale consisting of four items: ‘I always express my thoughts and feelings regardless of how controversial they might be’, ‘Going through life worrying about whether you might offend someone is just more trouble than it's worth’, ‘I think that it's important to speak one's mind rather than to worry about offending someone’, and ‘I'm not afraid to tell others what I think even when I know they disagree with me’. The intent was to measure socially desirable response style in the context of prejudice as research has shown that scores on commonly used measures of anti‐fat attitudes are significantly and inversely related to socially desirable response style (Latner, O'Brien, Durso, Brinkman, & MacDonald, 2008). Harris et al. (1982) did not control for socially desirable response set in their study but it is important to know the effect of target weight on stereotype ratings over and above the effect of possible covariates. The items used reflect concern with how one appears in the eyes of others (external/social motivation to control prejudice) and therefore provides a more clear‐cut measure of socially desirable response style than the other subscale, which includes concern with appearing prejudiced to oneself (internal motivation to control prejudice) and to others. Internal consistency reliability (Cronbach's alpha) for the subscale in the current study was .83.

Demographics

Participants reported their age and sex. Participants’ weight status (as indicated by BMI; self‐reported weight in kilograms divided by self‐reported height in metres squared) and need for glasses were also measured. Group membership may affect stereotyping. For example, past research has shown that obese persons also show negative attitudes towards obese people, or that they fail to show in‐group favouritism (Wang et al., 2004), although the magnitude of this bias is smaller than it is among leaner people (Schwartz, Vartanian, Nosek, & Brownell, 2006). Thus, analyses were conducted to determine the need to include participant BMI, need for glasses, and sex as covariates in the models to be tested (see Results).

Procedure

Institutional ethics approval was obtained for the study. Participants completed the experiment in paper and pencil form during tutorials (on‐campus cohort) or via an online survey software platform (online cohort). Other than the delivery platform, all other aspects of the experiment were identical for on‐campus and online participant cohorts. Consent information statements were presented before the questionnaire, and consent was implied by return/completion of an anonymous questionnaire. Participants completed a questionnaire booklet or online survey entitled ‘Person Perception Study’ and read a description (adapted from Harris et al., 1982) of a fictitious target person (described as female or male, overweight or average weight, and wearing glasses or not) presented on the first page of the questionnaire booklet/online survey. Participants were told that Chris Mortin was a woman or man in her/his 20s and were then presented with the following information:

Chris Mortin has dark curly hair; has brown eyes; (wears glasses) (no mention of glasses); is of medium height; (is of average weight) (is overweight); has a straight nose; has a pleasant smile; prefers casual clothes; often prefers the colour blue.

Harris et al. (1982) used the name ‘Chris Martin’, while the current study used the name ‘Chris Mortin’ to account for the former name now being associated with a popular musician. After reading the description, participants rated the target on the descriptors and then completed the self‐report measures (in the order presented above). In addition, participants completed the Behavioural Intentions Index (Brochu & Morrison, 2007) and the Social Distance Scale (Pearl, Puhl, & Brownell, 2012) after the stereotype ratings, and the Willpower and Dislike subscales of Anti‐fat Attitudes Questionnaire (Crandall, 1994) before the demographics. As the focus here is on the replication of the Harris et al. study, results for these additional measures are omitted. In total, the questionnaire took approximately 30 min to complete. There was mild deception in that the participants were told they were participating in a ‘person perception study’ and that the researchers were ‘interested in finding out how much people can tell about another person simply by reading a description of that person’. Participants were presented with a debriefing statement at the conclusion of the experiment.

Data analytic approach

Preliminary analyses were conducted to identify any effects of cohort (on‐campus vs online) or possible covariates (participant sex, BMI, need for glasses, social desirability) on the dependent variables (DVs). The remaining analyses replicated those of Harris et al. (1982). As per Harris et al. (1982), a 2 (overweight vs average weight target) × 2 (glasses wearing vs non‐glasses wearing target) × 2 (female vs male target) multivariate analysis of variance (MANOVA) was conducted with the 12 descriptors as dependent variables (DVs). This was followed by univariate ANOVAs on the 12 descriptors and the six composites. A Bonferroni adjusted significance level of p < .0028 was used given the number of comparisons performed in each analysis (i.e., .05/18 comparisons). In addition, effect sizes were computed for (1) the main effects of target weight, target glasses wearing, and target sex in the current study and in the original study and for (2) weight condition, i.e., effect sizes for differences between ratings of the average weight target in the current study versus Harris et al., and separate effect sizes for differences in ratings of the overweight target in the current study versus Harris et al. To compare negative stereotyping across the characteristics of weight, glasses, and sex, we focused on the halo composite, which assesses the overall favourability rating for the target. Effect sizes for the main effects of weight, glasses, and sex on halo were compared.

Results

Preliminary analyses

A comparison of the on‐campus and online cohorts included in the sample was conducted. Separate ANOVAs were conducted for all 12 of the descriptors and the six composites with the three experimental factors of target weight, target glasses wearing, and target sex and cohort status as main effects. Including cohort as a factor in the ANOVAs resulted in only one main effect and two interactions none of which were significant at the .01 level. Thus, data were pooled across cohorts.

Next, the effect of participant sex on the hypotheses was examined by performing a 2 (overweight vs average target) × 2 (glasses wearing vs non‐glasses wearing target) × 2 (female vs male target) × 2 (female vs male participant) MANOVA predicting the vector of the 12 descriptors. There were no significant interactions between participant sex and target sex, F(12, 473) = 0.46, p = .94, or between participant sex and target weight, F(12, 473) = 1.29, p = .22.

Finally, analyses were conducted to determine the need to include BMI, glasses wearing, and social desirability response set covariates in the models to be tested. Correlations were computed between BMI, need for glasses, social desirability, and (1) the 12 descriptors, (2) sex appropriate, (3) masculine, (4) halo, and (5) the obesity, glasses, and maleness stereotypes. In general, there were few significant correlations between covariates and the descriptors or composites, and inclusion of covariates did not substantively alter the experimental findings. Thus, we did not include covariates in subsequent ANOVAs.

Replication of Harris et al. (1982) and comparison of results between the studies

A MANOVA with the 12 descriptors as DVs indicated significant multivariate results for the main effect of target weight, F(12, 487) = 32.38, p < .001; glasses, F(12, 487) = 5.33, p < .001; and sex, F(12, 487) = 3.38, p < .001. Next, univariate ANOVAs were conducted with weight, glasses, and sex as the IVs and (1) the 12 descriptors, (2) sex appropriate, (3) masculine, (4) halo, and (5) the obesity, glasses, and maleness stereotypes from the original study as the DVs. Table 1 summarises the main effects for the current study. Table 2 presents Cohen's d effect sizes (.20 = small, .50 = medium, .80 = large) for weight, glasses, and sex both in the current study and the original study by Harris et al. (1982) and includes p‐values for the differences in the effect sizes between the studies.

Table 1. Estimated marginal means and statistical significance of experimental main effects

DV	Weight			Glasses			Sex			Pooled SD
	Overweight	Average	p‐value	Glasses	No Glasses	p‐value	Female	Male	p‐value
	M	M		M	M		M	M
Assertive	3.56	4.03	<.001	3.70	3.90	.076	3.93	3.66	.016	1.20
Active	2.79	4.32	<.001	3.41	3.70	.006	3.61	3.50	.278	1.14
Intelligent	4.85	5.01	.097	5.10	4.76	<.001	4.97	4.88	.326	0.99
Hardworking	3.65	4.56	<.001	4.17	4.04	.238	4.16	4.05	.278	1.14
Outgoing	3.85	4.23	.002	3.73	4.35	<.001	3.98	4.10	.318	1.28
Happy	4.52	5.15	<.001	4.71	4.95	.026	4.85	4.81	.689	1.11
Attractive	3.64	4.61	<.001	4.00	4.25	.011	4.12	4.13	.953	1.06
Popular	3.94	4.46	<.001	3.94	4.46	<.001	4.07	4.33	.018	1.11
Successful	4.33	4.91	<.001	4.75	4.49	.003	4.61	4.63	.889	0.91
Athletic	2.19	4.02	<.001	2.92	3.28	.001	3.08	3.13	.616	1.15
Sex appropriate	0.34	0.65	.118	0.32	0.66	.085	0.09	0.90	<.001	2.04
Feminine	3.54	3.50	.757	3.57	3.47	.402	3.76	3.28	<.001	1.24
Masculine	3.73	4.12	.001	3.73	4.12	.001	3.67	4.18	<.001	1.21
Masculine minus feminine	0.19	0.62	.028	0.16	0.65	.012	−0.09	0.90	<.001	2.04
Obesity stereotype	3.14	3.80	<.001	3.47	3.46	.777	3.51	3.42	.075	0.50
Glasses stereotype	3.53	3.15	<.001	3.49	3.19	<.001	3.33	3.35	.660	0.49
Maleness stereotype	3.48	4.17	<.001	3.86	3.80	.213	3.87	3.78	.063	0.51
Halo	37.00	44.80	<.001	40.10	41.69	.007	41.33	40.47	.144	6.06

Notes. Means are estimated marginal means from the factorial analysis of variance (ANOVA) without covariates. Pooled SD is the pooled standard deviation from the factorial ANOVA without covariates. For the obesity stereotype only, lower scores actually indicate higher consistency with the stereotype since lower scores correspond to lower ratings for positive attributes, i.e., less active, less attractive, less athletic, etc. Significant differences at the Bonferroni adjusted .0028 level (i.e., .05/18 comparisons) are shown in bold.

Table 2. Comparison of effect sizes between current study and the Harris et al. (1982) study

DV	Overweight			Glasses			Female
	Current d	Harris d	p‐value	Current d	Harris d	p‐value	Current d	Harris d	p‐value
Assertive	−0.39	−0.18	.270	−0.17	0.03	.285	0.23	0.13	.624
Active	−1.35	−1.42	.728	−0.26	−0.26	.963	0.10	0.36	.160
Intelligent	−0.16	−0.16	.983	0.34	0.23	.556	0.09	−0.17	.161
Hardworking	−0.80	−0.42	.052	0.11	0.17	.778	0.10	−0.69	.000
Outgoing	−0.30	−0.13	.382	−0.49	−0.54	.773	−0.09	−0.67	.002
Happy	−0.56	−0.31	.188	−0.22	−0.08	.473	0.04	−0.10	.472
Attractive	−0.92	−1.03	.573	−0.24	−0.41	.366	−0.01	0.05	.762
Popular	−0.47	−0.26	.270	−0.47	−0.42	.812	−0.23	−0.11	.502
Successful	−0.63	−0.24	.041	0.29	0.01	.145	−0.02	−0.12	.598
Athletic	−1.59	−1.37	.296	−0.31	−0.18	.488	−0.04	0.07	.546
Sex appropriate	−0.15	−0.40	.180	−0.17	0.19	.060	−0.40	−0.35	.816
Feminine	0.03	−0.09	.519	0.08	0.25	.367	0.39	1.06	.000
Masculine	−0.32	−0.02	.102	−0.32	−0.24	.661	−0.42	−1.17	.000
Masculine minus feminine	−0.21	0.04	.173	−0.24	−0.29	.795	−0.48	−1.33	.000
Obesity stereotype	−1.31	−1.30	.954	0.02	0.18	.393	0.18	−0.54	.000
Glasses stereotype	0.78	1.22	.024	0.61	0.49	.522	−0.04	0.02	.744
Maleness stereotype	−1.36	−1.10	.203	0.12	0.22	.601	0.18	0.12	.754
Halo	−1.29	−1.20	.673	−0.26	−0.21	.772	0.14	0.00	.460

Notes. Cohen's d was calculated as (Mean 1–Mean 2)/SD, where SD was square root of the MS error in the factorial analysis of variance from the current study without covariates. Overweight d was overweight minus average weight; glasses d was glasses minus no glasses; female was female minus male. The p‐value assesses the null hypothesis that effect sizes are the same for the two studies. Significant differences at the Bonferroni adjusted .0028 level (i.e., .05/18 comparisons) are shown in bold.

The overweight target was rated as significantly less active, assertive, athletic, attractive, happy, hardworking, masculine, popular, outgoing, and successful than the average weight target. Effect sizes were large for active, athletic, attractive, and hardworking. Ratings for intelligent did not significantly differ between overweight and average weight targets, nor did ratings for feminine or sex appropriateness. Although Harris et al. (1982) recorded fewer significant differences between ratings for overweight and average weight targets, effect sizes did not significantly differ between the studies for any of the descriptors.

With respect to the composite measures, overweight targets and average weight targets differed significantly and in the expected direction on the a priori linear combination of nine adjectives representing the ‘obesity stereotype’. In addition, overweight targets scored significantly lower than average weight targets on ‘halo’. The effect sizes for these composites were large but did not significantly differ from those for Harris et al. (1982).

Glasses wearing targets were rated as significantly more intelligent and successful but significantly less active, athletic, outgoing, masculine, and popular than non‐glasses wearing targets. In addition, targets described as wearing glasses were rated significantly higher than targets without glasses on the a priori combination of measures representing the ‘glasses stereotype’ and significantly lower on halo. Effect sizes for glasses wearing ranged from small to medium, with the largest effect size recorded for the glasses stereotype. Notably, despite a considerably larger number of significant differences between glasses wearers and non‐glasses wearers in the current study, there were no differences in effect sizes for this study and Harris et al. (1982).

Male targets were seen as significantly more masculine and significantly less feminine than female targets, thus validating the target sex manipulation. However, effect sizes for the femininity and masculinity differences were much greater in Harris et al. (1982) than in the current study. As per the original study, there were no differences between female and male targets on the a priori combination of measures representing the ‘maleness stereotype’. There were no significant interactions between target weight and target sex for any of the characteristics. This finding is consistent with Harris et al. (1982).

As can be seen from the effect sizes in Table 2, the negative effect of being overweight on halo (the overall favourability rating) was very large, yet the negative effect of glasses on halo was very small, and there was no significant difference between men and women on halo. The effect of being overweight on halo was significantly larger than the effect of wearing glasses on halo (p < .001) and the effect of sex on halo (p < .001).

In addition to testing whether overall weight effects have persisted, we compared ratings of average weight targets in the current study and the Harris et al. (1982) study, and ratings of overweight targets between the two studies. While Table 2 above suggested that overall weight effects have not changed in the current study compared to Harris et al., as shown in Table 3, there were some general differences in mean responses in the samples. Specifically, the current sample generally rated both average weight and overweight participants more negatively (e.g., lower halo).Targets in both groups also received ratings more consistent with the obesity stereotype.

Table 3. Standardised mean differences between the current study and Harris's study for obesity conditions

DV	Average weight		Overweight
	d (Current—Harris)	p‐value	d (Current—Harris)	p‐value
Assertive	0.16	.116	−0.05	.352
Active	−0.33	.007	−0.26	.027
Intelligent	0.15	.133	0.15	.126
Hardworking	0.04	.377	−0.33	.006
Outgoing	−0.12	.188	−0.28	.017
Happy	0.18	.089	−0.07	.294
Attractive	−0.24	.033	−0.13	.159
Popular	−0.39	.002	−0.60	<.001
Successful	0.23	.044	−0.16	.107
Athletic	−0.25	.031	−0.47	<.001
Sex appropriate	−0.54	<.001	−0.29	.015
Feminine	−0.10	.233	0.02	.428
Masculine	0.18	.083	−0.12	.174
Masculine minus feminine	0.17	.103	−0.09	.253
Obesity stereotype	−0.89	<.001	−0.90	<.001
Glasses stereotype	1.14	<.001	0.69	<.001
Maleness stereotype	0.01	.476	−0.25	.027
Halo	−0.48	<.001	−0.56	<.001

Note. The Cohen's d value was calculated by subtracting the Harris et al. study mean from the current study mean and then dividing by the current study SD. Thus, a positive d indicates that the mean in the current study was higher than in the Harris et al. study. Significant differences at the Bonferroni adjusted .0028 level (i.e., .05/18 comparisons) are shown in bold.

Discussion

This study examined whether obesity stereotypes have changed over time by comparing obesity stereotypes in 2012 to published data from 1982. The study replicated Harris et al.'s (1982) early experiment on obesity, glasses, and sex stereotypes. It was hypothesised that the obesity stereotypes observed by Harris et al. in the early 1980s have persisted over time such that being overweight is still negatively stereotyped and more negatively than wearing glasses or being female.

Hypothesis 1 was supported; overweight targets were rated much more negatively on a range of attributes. Of the various descriptors, large effect sizes were recorded for ‘active’, ‘athletic’, ‘attractive’, and ‘hardworking’. This is broadly consistent with public health messages that emphasise that fatness is undesirable and unhealthy and can be managed through physical activity. Large effect sizes were also observed for the obesity stereotype composite and overall favourability rating, confirming that obesity was negatively stereotyped. There was no significant difference between overweight and average weight targets for the descriptor ‘intelligent’, suggesting that while obesity stereotypes of ‘undesirable’ and ‘lazy’ were observed, the stereotype of overweight persons as ‘stupid’ was not present. Past research has shown that people make negative inferences about the competence of obese persons (see Puhl & Brownell, 2001). Perhaps competence is tied to the perception of obese persons as unattractive (e.g., for visible jobs) or lazy (e.g., for physical jobs) rather than unintelligent. There were no differences in ratings of overweight and average weight targets for feminine or sex appropriate, suggesting that heaviness was not associated with sex in the present study (the effect size for differences on masculine was small).

There were a greater number of significant differences between overweight and average weight targets in the present study compared with Harris et al. (1982), but the present study had a larger sample size and was thus able to estimate effect sizes with greater precision. Thus, for any given population effect size, the present study had greater statistical power. Importantly, there were no significant differences in effect sizes between the studies suggesting that there has been little change in obesity stereotypes over the past 30 years. It is noteworthy that both average weight and overweight targets were rated significantly more negatively in the present study compared with Harris et al. It may be that people have become more critical in their judgements of other people over time.

The finding that obesity stereotypes have persisted but have not increased is inconsistent with studies that have suggested an increase in weight bias over time (Latner & Stunkard, 2003; Tomiyama et al., 2015). Latner and Stunkard's study focused on changes in ranked preferences for an obese target relative to five other targets by comparing mean rankings over time. Therefore, their study essentially measured change in the overall evaluation of obese persons on a dimension of favourability (i.e., general attitudes) as opposed to beliefs about the characteristics of obese persons (i.e., stereotypes). Perhaps the stereotypes themselves have not changed, but overall dislike for obese persons has increased. Future research could compare changes in both stereotypes and favourability ratings in similar samples over time.

Tomiyama et al. explored explicit anti‐fat bias on a limited range of dimensions and found increases only for ‘extremely bad’ and ‘lazy’. Furthermore, they compared mean difference scores (thin people minus fat people) from 2001 and 2013 but did not report effect sizes around differences in ratings of thin versus fat targets within time periods, or compare effect sizes over the two time periods. The computation of effect sizes in the current study provides a more rigorous assessment of change over time. Furthermore, the use of the terms ‘thin’ and ‘fat’ in the Tomiyama et al. study might have increased perceived differences between the targets (c.f Brochu & Esses (2011). To the extent that weight bias is relative to actual weight, the term ‘overweight’ may have less severe implications than terms such as ‘fat’ or ‘obese’, and people who are described as fat may be more negatively stereotyped. Past research has confirmed that weight labels make a difference in assessing weight bias, suggesting that specific labels can affect study outcomes and interpretations. Brochu and Esses (2011) questioned whether it is perceived body size or what the terms themselves convey that matters. They found that the weight label ‘fat’ evoked more negative stereotypes than ‘overweight’, even though the labels were used to refer to the same weight target group suggesting that ‘fat’ conveys negative information beyond weight status. Vartanian (2010b) found that the weight label ‘obese people’ evoked more negative evaluations than the weight label ‘fat people’. Future research should assess whether stereotype content and severity differs across BMI category labels.

Consistent with Harris et al. (1982), glasses‐wearing targets were rated as having a mix of positive and negative characteristics; more intelligent and successful but less appealing and sociable. There were no differences in effect sizes from the original study and the effect size for the glasses stereotype composite was medium to large, suggesting that there has been little change in stereotypes for this physical characteristic over 30 years. For target sex, the general pattern was for smaller sex effects in the present study compared with Harris et al. Although male and female targets differed significantly on ratings for masculine and feminine (suggesting that the sex manipulation was successful), there were small effect sizes for sex in the present study suggesting that stereotypes based on sex have decreased over time.

Taken together, the results for weight, glasses, and sex supported Hypothesis 2; on the whole, overweight targets were stereotyped much more negatively than those wearing glasses or targets of either sex, with larger effect sizes for differences in overall favourability between overweight and average weight targets than between glasses wearers and non‐glasses wearers or men and women.

There was no interaction between target weight and target sex, and overweight men were rated just as negatively as overweight women. This finding is inconsistent with the notion that heavy women are more stigmatised than heavy men (see e.g., Fikkan & Rothblum, 2012). Harris et al. also failed to find an interaction between target weight and target sex. Perhaps the dimensions on which overweight men and women are differentiated were not captured in the Harris et al. descriptors. For example, Tiggemann and Rothblum (1988) found that female obese targets were rated as less self‐confident than male obese targets but this attribute was not assessed here.

The descriptors used to evaluate targets in the present study were a replica of those used by Harris et al. (1982), and it is possible that current obesity stereotypes incorporate descriptors that were not included here. Tiggemann and Rothblum (1988) found that Australian students rated obese targets as less attractive and happy and lazier than non‐obese targets. However, obese targets were also rated as less self‐disciplined and self‐confident and more friendly, self‐indulgent, and warm than non‐obese targets suggesting that the obesity stereotype is broader than the range of characteristics included in the Harris et al. study. Thus, the use of alternative or more detailed assessment of the ‘obesity stereotype’, as established in more recent studies (e.g., Puhl, Schwartz, & Brownell, 2005), may reveal additional information about how overweight persons are perceived, or perhaps differences in how overweight men and women are perceived. For example, descriptors relating to health status such as ‘fit’, ‘healthy’, ‘thriving’, and ‘well’ could be important in establishing current public views of obesity, given recent public health messages that focus on the implications of obesity for health status e.g., ‘Chubby kids may not outlive their parents’ (Freeman, 2011). A qualitative investigation of the impact of public health messages (Lewis et al., 2010) found that many overweight and obese persons felt that public health messages focussed too much on the physical health risks associated with being ‘too fat’, and that such messages amplified guilt, blame, and shame and were disempowering.

Another limitation of the current study is that it focused on stereotypes of obesity in a university sample. Thus, the findings may not reflect the strength of obesity stereotypes in the general population. For example, university students are likely to be less diverse in characteristics such as age and BMI, and psychology students may be less prejudiced on average than the general population due to their education around social issues. Accordingly, it is important to emphasise that the finding of no change in obesity stereotypes over time refers to university students only; broader conclusions about obesity stereotypes cannot be drawn from the data presented here, or in Harris et al. (1982).

The lack of random assignment of individual participants to conditions is a further limitation of the study because assigning participants in the same tutorial to the same condition introduces non‐independence in the observation (independence of observations is an assumption of ANOVA). To account for the dependencies in the data, the analyses could be performed using multilevel modelling; however, the intention here was to replicate the analyses of Harris et al.

It will be important for future studies to examine the real world outcomes of obesity stereotyping for persons with obesity. While the current research examined the content of stereotypes, the impact of such stereotypes on behaviour towards obese persons was not examined. Furthermore, there are a number of factors, other than stereotypes, which may account for weight‐based discrimination in real‐world contexts such as job selection (e.g., rational bias due to social norms, perceived costs of obesity; see Grant & Mizzi, 2014).

While the current study had some limitations, it also had a number of strengths that enhance confidence in the finding of a persisting negative perception of overweight persons relative to average weight persons across a broad range of characteristics. This study provided effect size calculations to qualify and compare differences between overweight and average weight targets and across physical characteristics (weight, glasses, sex). The addition of effect sizes also allowed for direct comparison with the original study, both in terms of the weight main effect (average weight vs overweight) and the effect of average weight and overweight considered separately. The fact that the sample was comparable to the sample for the original study (Australian undergraduate psychology students in their early 20s with a similar profile) further strengthens the replication and comparison. There were a number of large effect sizes in the current study, which underscores both the extent of negative obesity stereotypes in this sample and their persistence over time given that there were no differences in effect size when compared with the Harris et al. study in 1982.

Consistent with Harris et al. (1982), the current study focused on psychology students. In both studies, psychology students showed a negative bias against overweight targets suggesting that psychology courses may benefit from more education and training around recognising weight bias as a modern form of prejudice, identifying personal anti‐fat attitudes and behaviour, and interacting with persons who are overweight or obese in respectful and sensitive manner.

Dispelling the content of negative obesity stereotypes is an important step in improving the daily living experiences of overweight and obese persons by helping to reduce weight stigma and its negative consequences. Non‐stereotypical portrayals of obese individuals in outlets such as the media may help reduce stereotypical beliefs about obesity. An experiment by Pearl et al. (2012) showed that individuals who viewed stereotypical, negative media portrayals of obese persons endorsed more negative attitudes towards obese persons and stronger social distance than participants who viewed non‐stereotypical, positive media portrayals. In particular, current public health campaigns may miss the mark when it comes to reducing stereotypes and supporting these individuals. A study of public perceptions of major obesity‐related public health media campaigns (Puhl, Peterson, & Luedicke, 2013) from the USA, the United Kingdom, and Australia, found that participants responded most favourably to messages that focused on health behavioural changes without reference to obesity or body weight and least favourably to messages that contained stigmatising content. Stigmatising messages such as Australia's ‘Childhood obesity is child abuse’ campaign, reinforce negative stereotypes that link weight with negative assumptions about character or abilities and shame and blame, and may actually lower motivation for behaviour change (Puhl et al., 2013). The current landscape of public health campaigns about obesity needs to be reviewed to address negative stereotypes and incorporate information about the mental health risks associated with weight stigma, particularly given that both obesity and weight stigma are persisting issues. In particular, this study shows that negative obesity stereotypes have persisted over a 30‐year period.