Preferring thin people does not imply derogating fat people. A rasch analysis of the implicit weight attitude

Authors


  • Disclosure: The authors declare no conflict of interest.

Abstract

Objective:

The meaning of the implicit weight attitude in individuals of different weight by distinguishing the contribution of positive and negative associations to the overall measure was investigated.

Design and Methods:

The implicit weight attitude was assessed using the Implicit Association Test. A Rasch model was used to identify which stimuli most affected the implicit measure. Explicit attitudes were assessed with a thin-fat preference scale, and thermometer scales for both thin and fat people. The sample consists of 510 individuals categorized according to their BMI.

Results and Conclusion:

A significant implicit preference for thin relative to fat people was observed in all weight groups. In normal weight and obese individuals, the preference was mostly affected by positive stimuli (more easily associated with thin than with fat people). In underweight individuals, the preference was mostly guided by positive (more easily associated with thin than with fat people) and negative (more easily associated with fat than with thin people) stimuli. In overweight individuals, all stimuli contributed to the preference in a similar way. In all weight groups, the implicit weight attitude correlated with the explicit preference and/or the thin thermometer, whereas it never correlated with the fat thermometer. A pro-thin bias was observed in normal weight and obese individuals, whereas both a pro-thin and an anti-fat bias were observed in underweight individuals. A clear preference for thin people relative to fat people was observed in overweight individuals. Therefore, uncritically interpreting the implicit preference for thin people as a sign of derogating fat people might be misleading.

Introduction

The Weight Implicit Association Test (Weight IAT) (1) is a computerized two-choice discrimination task in which stimuli have to be categorized as belonging to the object categories Thin People and Fat People, or to the attribute categories Good and Bad by pressing, as quickly and accurately as possible, one of two response keys. Categorizing the stimuli more quickly when Thin People shares the response key with Good (and Fat People with Bad) than when it shares the response key with Bad (and Fat People with Good) indicates an implicit preference for thin people compared to fat people.

Research with the Weight IAT has shed light on the pervasiveness of the weight bias against obese people. Such bias was found to be common in different cultures, and to increase in strength with the percentage of obesity (2). Strong negative implicit attitudes toward obese people were observed in the general population (3), and even in health-care professionals who specialize in the treatment of obesity (4). Unlike other minorities, overweight people themselves were found to hold negative implicit associations toward their own group, instead of showing ingroup favoritism (5, 6).

The implicit preference for thin relative to fat people observed in non-obese population has been systematically interpreted as an expression of anti-fat bias (3, 4, 7, 8), thus implying a negative evaluation of fat people. The same implicit preference observed in overweight and obese individuals has been interpreted as an expression of ingroup devaluation (5, 6). We argue that such interpretations might be misleading.

The article aims to investigate the implicit weight attitude and how its meaning changes with the weight of respondents. An analysis procedure is proposed, which disentangles the contribution of positive and negative associations to the overall measure. We argue that if positive (rather than negative) stimuli are categorized more quickly in the condition Thin-Good/Fat-Bad than in the condition Fat-Good/Thin-Bad, then the implicit preference toward thin relative to fat people should be interpreted as an expression of pro-thin bias (or thin favoritism) rather than anti-fat bias (or fat devaluation). The same implicit preference should be interpreted as an expression of anti-fat bias rather than pro-thin bias if the opposite occurs.

Methods and Procedures

Respondents

Respondents completed the Weight IAT available at: https://implicit.harvard.edu/implicit/italy. They were self-selected, and are not representative of any particular population.

From the respondents who visited the website, 510 (mean age = 27.06 ± 8.99 years; 64.14% female) provided complete and interpretable data1 (10). Fifty-three respondents were underweight (UW, BMI < 18.5), 331 were normal weight (NW, 18.5 ≤ BMI < 25), 83 were overweight (OW, 25 ≤ BMI < 30), and 43 were obese or extremely obese (OB, BMI ≥ 30).

Materials and procedure

The Weight IAT used the category labels Thin People, Fat People, Good, and Bad. Ten computer-generated faces of thin and fat people were used to unambiguously represent the categories Thin People and Fat People. The same face was morphed to appear thin and fat. Sixteen words were used to represent the categories Good (glory, happy, joy, laughter, love, peace, pleasure, wonder) and Bad (agony, despicable, evil, failure, horrible, nasty, pain, terrible). The standard IAT procedure (11) was employed.

Three items were used to assess explicit attitudes toward fat and thin people. One item asked respondents to indicate their preference for thin or obese people on a 5-point scale from “I strongly prefer fat people to thin people” (1) to “I strongly prefer thin people to fat people” (5). Two items asked respondents to rate how cold or warm they felt toward fat and thin people on an 11-point scale from “Extremely cold” (0) to “Extremely warm” (10).

Data analysis

The IAT data were analyzed through a Many-Facet Rasch Measurement (MFRM) (12), by using the software Facets (13). Facets were: (a) respondents, (b) BMI categories of respondents, (c) associative conditions, and (d) attribute stimuli. The data were prepared for the analysis according to the procedure described in the literature (9, 14, 15). A parameter γ was estimated for each associative condition indicating the ease of the condition, and a parameter δ for each attribute stimulus indicating its speed of categorization. Greater estimates mean greater speed and greater ease. The MFRM provides the analysis of interactions between elements of different facets (13). We investigated whether the ease of the associative conditions changes according to the BMI category of respondents (differential condition functioning), and whether the speed of categorization of the attribute stimuli changes according to both the BMI category of respondents and the associative condition the stimuli are presented in (differential stimulus functioning).

One-sample t tests were used to test the significance of the explicit measures of attitude. A MANOVA with the three items as dependent variables and the BMI category of respondents as factor was used to test differences between the groups in the explicit attitudes.

Pearson's correlation coefficients among implicit and explicit attitudes were computed.

Results

Implicit attitude

The condition Thin-Good/Fat-Bad (TGFT) was easier than the condition Fat-Good/Thin-Bad (FGTB; γTGFB = 0.38; γFGTB = −0.38; χ2(1) = 1104.4, P < 0.001). The distance d = 0.76 between the two conditions represents the size of the IAT effect. It is significantly different from 0 (z = 53.74, P < 0.001), meaning that, taken all together, respondents implicitly preferred thin to fat people. A significant preference toward thin relative to fat people is observed in respondents of each BMI category (dUW = 0.80, z = 11.31; dNW = 0.78, z = 27.58; dOW = 0.76, z = 13.44; dOB = 0.50, z = 6.40, P < 0.001 for all). The IAT effects observed in underweight, normal weight, and overweight respondents do not differ from each other, but are significantly greater than that observed in obese respondents (zUW-OB = 2.86; zNW-OB = 3.38; zOW-OB = 2.69, P < 0.05 for all). The stimuli were categorized with different speeds (δ from −0.35 to 0.31, χ2(15) = 148.8, P < 0.001).

In normal weight and obese respondents, the IAT effect computed on positive stimuli is significantly stronger than that computed on negative stimuli (dpos = 0.94, dneg = 0.64, zpos-neg = 5.30, P < 0.001 for normal weight; dpos = 0.77, dneg = 0.24, zpos-neg = 3.26, P < 0.01 for obese). Thus, their implicit preference for thin relative to fat people is driven most by associating positive stimuli with thin people rather than negative stimuli with fat people. Table 1 provides differential stimulus functioning that shows the contribution of each stimulus to the implicit measure. Compared with their overall speed of categorization, the positive stimuli joy, pleasure, and happy were categorized by normal weight respondents faster in the condition Thin-Good/Fat-Bad and slower in the condition Fat-Good/Thin-Bad. These stimuli were associated more easily with thin than with fat people. Compared with their overall speed of categorization, the negative stimuli agony and terrible were categorized by normal weight respondents faster in the condition Fat-Good/Thin-Bad and slower in the condition Thin-Good/Fat-Bad. These stimuli were associated more similarly to thin and fat people. Obese respondents associated happy more easily with thin than with fat people, and associated failure, evil, and terrible more similarly to both thin and fat people.

Table 1. Speed of categorization of the stimuli in the two associative conditions for each BMI category
 Thin-Good/Fat-BadFat-Good/Thin-Badtdf
StimulusORSMSRSEORSMSRSE
  • ORS = observed raw scores; MSR = measure. The t values test the hypothesis that the difference between the measures is equal to zero.

  • *

    P < 0.05;

  • **

    P < 0.01;

  • ***

    P < 0.001.

Underweight (53)
 Horrible1290.450.2193−0.190.202.2*103
 Happy1310.540.2296−0.080.202.11*103
 Pleasure1310.540.22980.000.191.85103
 Joy1280.400.21980.000.191.4103
 Wonder1200.070.2095−0.120.200.66103
 Glory1230.190.201020.150.190.14103
 Failure1190.030.20990.040.19−0.03103
 Despicable1190.030.201010.110.19−0.30103
 Evil1250.270.211080.370.19−0.35103
 Laughter1270.360.211110.480.19−0.44103
 Love1260.310.211100.450.19−0.46103
 Pain1260.390.211130.560.19−0.58102
 Nasty1210.110.201060.300.19−0.68103
 Peace1210.110.201070.330.19−0.82103
 Terrible1230.190.201100.450.19−0.92103
 Agony111−0.280.191040.220.19−1.83103
Normal weight (331)
 Joy7870.450.08587−0.060.084.4***659
 Pleasure7600.260.08585−0.070.082.96**659
 Happy7610.270.08594−0.020.082.52*659
 Love7930.490.086390.280.081.83658
 Horrible7440.160.086080.070.080.77659
 Peace7530.220.086200.160.080.54658
 Laughter7270.050.086010.030.080.20659
 Glory7370.110.086140.110.080.04659
 Nasty7200.010.085980.010.08−0.03659
 Despicable687−0.200.08568−0.170.08−0.21658
 Wonder666−0.330.08549−0.300.08−0.26658
 Evil7260.040.086220.160.08−1.04659
 Failure7220.020.086190.140.08−1.1659
 Pain698−0.120.086080.070.08−1.7658
 Terrible717−0.010.086340.230.08−2.22*659
 Agony685−0.210.086100.090.08−2.66*659
Overweight (83)
 Happy1900.320.17144−0.140.171.96163
 Terrible1810.080.16136−0.360.171.86163
 Joy1910.350.17147−0.060.161.74163
 Nasty1840.160.171490.000.160.69163
 Glory177−0.030.16143−0.170.170.58163
 Horrible1870.240.171540.130.160.48163
 Love1870.240.171560.180.160.25163
 Laughter1840.160.171570.210.16−0.22163
 Pleasure171−0.190.16146−0.080.16−0.45163
 Wonder165−0.340.16141−0.220.17−0.53163
 Evil168−0.260.16144−0.140.17−0.55163
 Peace177−0.030.161550.160.16−0.81163
 Despicable168−0.260.16147−0.060.16−0.91163
 Failure171−0.190.161520.080.16−1.15163
 Agony159−0.500.16143−0.170.17−1.44163
 Pain168−0.260.161550.160.16−1.84163
Obese (43)
 Happy970.340.2266−0.450.252.38*83
 Pleasure910.050.2266−0.450.251.5283
 Pace88−0.100.2271−0.170.230.2383
 Horrible87−0.140.2271−0.170.230.0983
 Glory87−0.140.2272−0.120.23−0.0883
 Love940.190.22810.330.22−0.4683
 Laughter83−0.330.2271−0.170.23−0.5183
 Despicable82−0.380.2271−0.170.23−0.6583
 Nasty82−0.380.2274−0.010.23−1.1683
 Wonder71−0.930.2365−0.520.25−1.2283
 Pain77−0.620.2272−0.120.23−1.5783
 Joy72−0.880.2368−0.340.24−1.6283
 Agony80−0.470.22770.140.22−1.9683
 Terrible79−0.520.22790.240.22−2.44*84
 Evil72−0.880.2374−0.010.23−2.67**83
 Failure78−0.570.22800.280.22−2.75**83

In both underweight and overweight respondents, the IAT effect computed on positive stimuli does not significantly differ from that computed on negative stimuli. However, it is worth noting that, whereas in overweight respondents all stimuli contributed to the preference for thin relative to fat people in a similar way, in underweight respondents there were two stimuli, one positive (happy, more easily associated with thin than with fat people) and the other negative (horrible, more easily associated with fat than with thin people), that contributed the most.

Explicit attitudes

Respondents of different BMI categories preferred thin to fat people (MUW = 3.75, t(52) = 7.91; MNW = 3.77, t(329) = 18.24; MOW = 3.63, t(82) = 6.66; P < 0.001 for all), and felt warm toward thin people (MUW = 7.04, t(52) = 7.76; MNW = 6.96, t(330) = 20.61; MOW = 6.69, t(82) = 8.35; MOB = 6.09, t(42) = 3.85; P < 0.001 for all). Only obese respondents did not significantly prefer thin to fat people. No respondents of any BMI category felt warmth or coldness toward fat people.

Underweight and normal weight respondents preferred thin to obese people significantly more than obese respondents (PUW-OB < 0.05; PNW-OB < 0.01), and felt warmer toward thin people than obese respondents did (PUW-OB < 0.05; PNW-OB < 0.05).

Correlations

In respondents of all BMI categories, the explicit preference scale positively correlated with the thin thermometer (r = 0.37, 0.49, 0.53, 0.76 for underweight, normal weight, overweight and obese, respectively, P < 0.01) and negatively correlated with the fat thermometer (r = −0.41, −0.39, −0.41 for normal weight, overweight, and obese, respectively, P < 0.01; r = −0.33 for underweight, P < 0.05). The two thermometers did not correlate.

The implicit attitude correlated with the explicit preference scale (r = 0.13, 0.27, 0.31 for normal weight, overweight and obese, respectively, P < 0.05) and/or with the thin thermometer (r = 0.35, P < 0.01 for underweight; r = 0.12, .30, P < 0.05 for normal weight and obese, respectively), and it did not correlate with the fat thermometer.

Discussion

Consistently with the literature (4-8), we have found that individuals of all BMI categories implicitly preferred thin to fat people. Moreover, we have observed that positive and negative stimuli did not contribute to the implicit measure in a similar way, and that differences were present among individuals of different weights. This result suggests that the meaning of the implicit preference for thin people differs with the weight of respondents. In normal weight and obese respondents, the implicit preference was mostly an expression of pro-thin bias rather than anti-fat bias because the stimuli that mostly contributed to it were positive stimuli. The implicit preference observed in underweight respondents was an expression of both pro-thin and anti-fat biases because the stimuli that mostly influenced it were positive and negative. Only the IAT measure observed in overweight respondents could be correctly considered the expression of implicit preference for thin relative to fat people because all stimuli, positive and negative, contributed to it in a similar way. In none of the BMI categories did the implicit preference exclusively express an anti-fat bias. Weak to moderate associations were found between the implicit attitude and the explicit preference, the thin thermometer or both. An association between the implicit attitude and the fat thermometer was never found. The small number of items used to assess the explicit weight attitude represents a limitation of the present study.

Our results have an important theoretical implication for the definition of implicit weight attitude. Previous work systematically interpreted the preference toward thin relative to fat people observed in individuals of different weight as a sign of pervasive fat devaluation and anti-fat bias (3-8). Such a view implies a negative evaluation of fat people. Our results suggest that the meaning of the implicit preference for thin people differs with the weight of respondents, and uncritically interpreting it as an expression of fat derogation might be misleading.

Results of the study allow a new reading of findings present in the literature. The implicit preference for the outgroup observed in obese individuals was interpreted as evidence that they internalize the negative attitude and the stigma against their own group that exist in society (5, 6). According to our results, the same preference could express the internalization of positive attributes that society associates with being thin.

The strong implicit preference toward thin relative to obese people observed in professionals in obesity treatment was explained by arguing that health care professionals are exposed to the same social messages about obese individuals that the general population is (4, 7). Our results suggest an alternative explanation. Professionals, who treat and study obesity and who understand that it is due to genetics and not simply to individual behavior, are expected to develop a positive attitude toward thin people rather than a negative one toward fat people. The Weight IAT actually measures the former.

Acknowledgements

The authors thank the principal investigators of Project Implicit for their permission to report data.

  1. 1

    Part of this data set was used in Anselmi, Vianello and Robusto (9). However, the purpose of the present article and the analyses are new.

Ancillary