The structure and function of pathogen disgust

1. Introduction

It is unlikely to be a coincidence that many of the stimuli that elicit the emotion of disgust in humans are also implicated in the transmission of infectious disease [1–3]. Human excreta, for example, are both a major source of pathogenic viruses, bacteria and helminths and an important elicitor of disgust. Similarly, saliva, sexual fluids, spoilt foods, ectoparasites and unhygienic behaviour are, at the same time, disgust elicitors and sources of risk of infection. With others, we have contended that the explanation for this overlap between the disgusting and the infective reflects the functional role (censu Tinbergen [4]) of the disgust system in preventing infectious disease [3,5–10]. According to the parasite avoidance theory of disgust, the disgust motive is an adaptive system instantiated in neural tissue that guides behaviours that serve to avoid risk of infection [11].

This theory predicts that behaviour that reduces contact with pathogens and parasites (terms used interchangeably in this paper; see [12]) will have been under strong selection pressure throughout the evolutionary history of free-living organisms, and hence will be ubiquitous in Animalia. Examples include the avoidance of infected conspecifics in lobsters and mice [13,14], of pathogenic wastes in nematodes and kangaroos [15,16] and hygienic behaviour in ants and birds [17,18]. Support for this position is found in an increasing volume of animal literature, including this issue (for an overview see Curtis [19]).

In humans, this explanation has not always been accepted. Early models were Freudian, with disgust serving as a defence mechanism against inappropriate desires, including incest [20]. Darwin described it as a feeling related to the avoidance of offensive foods [21] and Angyal suggested that it serves to prevent overindulgence [1]. More recently, a standard model of disgust as serving to protect us from awareness of our animal natures and our mortality, and for the preservation of the moral order became accepted [2,22–25]. In the past two decades, support for an alternative, disease-avoidance model has accumulated. There are a number of lines of argument for this. In 2001, Curtis & Biran [3] demonstrated a close mapping between the elicitors of disgust and sources of infectious disease. Later, large-sample cross-cultural experiments showed that stimuli with disease salience were rated as more highly disgusting than paired items without the disease connotation [26]. The disgust response is now known to be functionally integrated with the immune system, with disgust elicitation resulting in short-term pre-emptive immune upregulation [27,28]. Fessler et al. [5] found that during the first trimester of pregnancy—a period of particular vulnerability to infection—disgust sensitivity increases. There is also a trade-off between the potential benefits of avoiding disease and the costs of such behaviour in terms of time, energy and lost opportunity to feed or engage in social interaction. For example, when the costs of disease-avoidance behaviour increase, due to food deprivation or loss of mating opportunity, disgust sensitivity decreases [29–33].

(a) Types of disgust

If the disgust adaptive system did, indeed, arise to orchestrate behaviour that leads to the avoidance of infection, and as sources of infection are distinct in the kinds of behavioural problem they pose, then it follows that there should be regularities in the disgust response that correspond to different types of infection threat. The disgust adaptive system should have an internal structure that orchestrates different types of protective behaviour. This paper aims to elucidate that structure.

An early factor structure for disgust was suggested by the work of Rozin and colleagues [2]. The researchers generated a pool of items by asking university students to list disgusting items. Another sample then rated their disgust responses to each. Guided by a factor analysis and some prior theory, they argued that disgust has seven different kinds of input: death; hygiene; animals; body wastes; sex; sympathetic magic; body envelope violations and foodstuffs. The researchers argued that these kinds of stimuli elicit disgust because they act as aversive reminders of mortality, can contaminate and poison or threaten the moral order. This scale was widely used but attracted criticism for its theoretical basis and lack of internal consistency [10,34]. In a later, revised version of the scale (the disgust scale-revized (DS-R)), the complexity and number of items were reduced to give three statistically better-supported factors, which were labelled, somewhat arbitrarily, ‘core’, ‘animal’ and ‘contamination’ disgust [35,36]. Thus, the DS and DS-R imply that disgust has a factorial structure, and, although the items were not generated from a pathogen perspective, almost all of the items contain direct or indirect cues of threat from pathogens.

An alternative factor structure for disgust was proposed by Tybur et al. [37]. Their three domain disgust scale (TDDS) arose from their evolutionarily informed hypothesis that disgust should respond to three adaptive problems: preventing infection; optimizing mate choice; and regulating others' social behaviour through punishment and/or avoidance. Developed using methods similar to the DS above, American students and professors were asked to list things that they found generally disgusting, sexually disgusting and morally disgusting. Confirmatory factor analysis on disgust ratings indicated that differences between participants were best captured by three factors, corresponding to the three original item pools: pathogen disgust, sexual disgust and moral disgust [37]. In this scale, pathogen-related stimuli only form one seven-item input category to the TDDS and the authors have called for further exploration of this domain [37].

(b) The inputs to a disease-avoidance mechanism

What, then, would we predict a pathogen avoidance system in animals to be able to recognize and respond to adaptively? Pathogens are typically too small to be seen directly and so their presence must be inferred from observable cues that tend to co-occur with them. For example, aromatic compounds such as indole reliably predict the presence of the many pathogens that can be found in faeces. Although it is not, of itself, harmful, the smell of faeces elicits disgust and faeces avoidance behaviour in primates, including humans [3,38]. By capitalizing on co-occurring cues, the risk of contact with unseen pathogens can be mitigated.

Since epidemiology is the study of the risk of disease in populations [39], it should offer us a means of categorizing disease threats. Six main transmission pathways for human pathogens can be identified [40]: (1) direct interpersonal contact; (2) interpersonally through aerosolized droplets; (3) interpersonal sexual contact; (4) contact with a secondary host or vector, e.g. a rodent or insect; (5) ingestion of contaminated food or water; and (6) contact with a fomite (a pathogen-contaminated object). Note that, though derived largely from studies of twentieth century diseases, we assume that these categories also describe ancestral pathogen transmission routes, an assumption supported by consistency in transmission pathways across related species [41]. To test our initial hypothesis that the factor structure of pathogen disgust should reflect the routes of disease transmission, we generated a set of stimuli based on a set of cues derived from five diseases from each of the above six transmission categories.

Over 2500 participants rated these written descriptions for disgust, and exploratory factor analysis was used to identify patterns of co-variation.

2. Material and methods

(a) Measures

(i) Infection cue vignettes

Infectious diseases were selected at random by assigning each disease in an infectious disease handbook [40] a number, and then selecting numbers at random until we had identified five diseases from each of the six main transmission pathways. For each of these 30 diseases, cues that might be associated with disease transmission risk were extracted. To take one example, yaws (Frambesia tropica) was selected as an example of a disease that is directly transmitted from person to person through skin contact with an infective individual (route 1). It is characterized by papilloma (externally projecting tumours), periostitis (inflammation around bones, which affects gait), and, in later stages, major lesions causing disfigurement and further disability. These cues—skin abnormalities and difficulty in movement—were recorded. Applying this strategy, the following types of cue arose repeatedly: presence of, or contact with animal and insect disease vectors, genital lesions or risky sexual behaviour, people showing signs of infection including skin or surface irregularities (e.g. scabs, sores, lesions, pus), unusual body shape (e.g. swelling, anorexia, obesity, heavy scarring, deformity, missing limbs), behavioural cues (e.g. expressions of physical pain, scratching, ‘sickness behaviour’, irregular gait), contextual cues associated with greater disease prevalence (e.g. poverty, poor personal hygiene, old age), cues to airborne transmission (e.g. hearing wheezing), unfamiliar and spoilt food, fomites and contaminated environments, and interactions with infectious substances. Based around these 14 cue categories we generated a series of vignettes, resulting in the generation of 75 items. Where possible, disease-relevant items from the TDDS and DS were reused. This disease cue set is detailed in table 1.

(ii) Study design and participant recruitment

An online survey was developed using Web Experiment software [42]. It included an information sheet, a consent form, a demographic information page and the 72 disease items listed in table 1. The items were presented as the scenario in writing beside a sliding scale with 100 divisions, anchored by ‘no disgust’ at one end and ‘extreme disgust’ at the other. Participants clicked at a point along the sliding scale to indicate the strength of their disgust response. Item presentation was quasi-randomized; the items appeared in a fixed order on each of eight pages, with the order of these pages being randomized. Participants were recruited through advertisements on Facebook with variants of the text ‘Participate in a brief psychology study on what makes people disgusted!’. The study was also advertised on psychology websites where studies seeking participants are listed (e.g. ‘Psychology Experiments on the Web’). Ethical permission for the study was granted by the London School of Hygiene and Tropical Medicine's ethical review board (approval no. 5930).

(b) Analysis

Common factor analysis was used to identify latent variables underlying co-variation in the dataset. The analysis was conducted on the correlation matrix. As we predicted correlations between latent variables, we used an oblique rotation (direct oblimin) procedure. Velicer's MAP (minimum average partial) analysis was used to determine the number of factors [43]. Participants' factor scores were created by averaging the five items that loaded most heavily on each factor. The participants’ total disgust score was created by averaging these factor scores. To assess the relationship between disgust, age and gender, we conducted secondary analyses, using ANOVA and multiple regression.

3. Results

(a) Participants

Of the 2742 participants, 63 either failed to complete the survey, responded no to the question ‘did you answer the questions honestly and accurately?’, reported an age less than 16 or greater than 99, or had a standard deviation of 10 or less across all disgust items (we assumed participants with low standard deviation were clicking without paying attention) and were excluded from the analysis. The majority of the participants were citizens of the UK (67%), the US (20%) or Canada (6%). The mean age was 28 years (s.d. 11.6) and 66% of participants were women. Students accounted for 44% of the sample.

(b) Factor analysis

Data were suitable for factor analytic procedures. In all of the analyses reported below the Kaiser-Meyer-Olkin Measure of Sampling Adequacy (KMO) was 0.97, Bartlett's Test of Sphericity was statistically significant, X² < 0.001, individual item KMO statistics, extracted from the diagonal of the anti-image correlation matrix, were satisfactory (minimum value = 0.94). The mean communality was 0.42, with a range of 0.22 to 0.63. Velicer's MAP analysis indicated that 6 factors should be extracted.

Three of the factors were robust (according to criteria specified by Costello & Osborne [44]), with more than five or more items with loadings of 0.5 or greater. Factors 2 and 4 (table 2) were marginally weaker, with five or more items with loadings above 0.45. Factor 6, however, was substantially weaker, with just one factor-loading above 0.4. Moreover, many of the items that loaded on factor 6 also loaded on factors 2 or 4. Given these weak loadings and cross loadings, we re-ran the analysis selecting a five-factor solution. Factor 6 items were then largely absorbed into factor 4. In this five-factor model, four of the five factors were robust [44] while the fifth factor had three items with loadings of 0.5 of higher. Below, we describe the content of these factors in both the five- and six-factor solutions.

(i) Six-factor solution

The items that loaded on factor 1 were varied but a common theme was poor hygiene behaviour (‘Listening to someone sniffle and snort continually’, ‘Watching a woman pick her nose’, standing ‘close to someone with body odour…’, ‘Feeling someone cough into your face’) or contamination of the environment resulting from poor hygiene (‘Seeing some snotty tissues left on the table’, ‘Walking through a city alleyway, you get a strong smell of urine’, ‘…you see some un-flushed excrement in a toilet’). This was labelled hygiene disgust. Factor 2 includes items about animals (raw chicken, slugs, worms, cockroaches, teeming insects) and is here termed animal disgust. Factor 3 is dominated by sexual behaviour, with items about prostitution and promiscuity. We labelled this factor sex disgust. Factor 4 comprises items related to irregular body shape (‘…man with a disfigured face’, ‘… someone missing a thumb’, ‘…obese woman sunbathing’, ‘… a hairless old cat…’), though poverty (touching a ‘homeless man’) and illness, (‘Hearing someone wheeze heavily’) also featured. As most of these items related to individuals looking or acting in unusual ways, we labelled it atypical appearance disgust. Factor 5 was dominated by problems with the skin and body surface, for example: ‘… a friend shows you a big, oozing lesion …’, ‘…the eye is almost fully sealed and weeps constantly’ or ‘Seeing pus come from a genital sore’. Because this factor included items about lesions and other problems with the body surface, including lesions on sexual organs, we name it lesion disgust. Factor 6, termed food disgust, includes items that describe the decay or deterioration of foodstuffs.

(ii) Five-factor solution

The hygiene, sex, atypical appearance and lesion disgust factors were largely unchanged by the changed factor solution. The difference lay in the treatment of food disgust items. All five of these items were absorbed into the animals factor, although the loadings of these items on this factor were low (<0.4).

(iii) One-factor solution

One further solution involved extracting a single factor. This general factor accounted for most of the item covariances; indeed, in a one-factor solution, only one item had a loading below 0.40. Table 3 displays the one- and five-factor solutions.

(iv) Univariate analysis of sex and age differences across six disgust factors

We generated factor scores for each participant by averaging their scores on the five items that loaded most strongly on the associated factor. (While we used the six-factor solution to calculate these scores, creating scores from the five-factor model would have led to identical factor scores, save the omission of the extra ‘food disgust’ factor. This is because the five highest-loading items on the five common factors were the same.) A general disgust score was generated by averaging each participant's scores across these six factors.

We carried out t-tests to examine if men and women differed in their reaction to the disease stimuli as has been found in many previous studies [45]. As figures 1 and 2 show, while women rated all six categories of disease stimuli more disgusting than men, the magnitude of these sex differences varied. Women rated sex disgust (t₂₃₂₀ = 15.89, p < 0.001) and animal disgust (t₂₃₂₀ = 18.57, p < 0.001) items as about 18 points more disgusting than men, and effect sizes were large: d = 0.70 and 0.82 respectively. Effect sizes for hygiene disgust (t₂₃₂₀ = 10.86, p < 0.001, d = 0.48), food disgust (t₂₃₂₀ = 10.11, p < 0.001, d = 0.44) and lesion disgust (t₂₃₂₀ = 7.15, p < 0.001, d = 0.31) were moderate, and sex differences were small for atypical appearance items (t₂₃₂₀ = 3.68, p < 0.001, d = 0.16).

• Download figure
• Open in new tab
• Download powerPoint

• Download figure
• Open in new tab
• Download powerPoint

There was no correlation between hygiene disgust sensitivity and age (r = 0.04, p = 0.06). All other disgust variables were negatively associated with age: sex, r = −0.17, p < 0.001; lesionr = −0.26, p < 0.001; foodr = −0.21, p < 0.001; animalsr = −0.12, p < 0.001; atypical appearancer = −0.30, p < 0.001; and total, r = −0.23, p < 0.001.

(v) Multivariate analysis of age and sex differences across six disgust factors

Because male participants were, on average, older, we also performed multivariate analysis exploring age and sex differences across the six disgust factors using multiple regression. The results are displayed in table 4. These replicate the univariate analysis with one exception: older age predicted higher hygiene disgust.

Correlation between the six disgust factors. The relationship between the six disgust factors was examined using Pearson's correlations; see table 5. The correlations between the six different factors were intermediate, ranging from 0.31 to 0.57, suggesting that these six factors do capture distinct facets of disgust. Disattenuated for unreliability, the correlations ranged from 0.39 to 0.76, suggesting a greater degree of coherence in people's responses to the items, consistent with the one-factor solution presented in table 3.

4. Discussion

A number of attempts have been made to explain the multifarious objects, events, behaviours and individuals that occasion disgust. In this study we took the view that the most parsimonious explanation is likely to lie in the disgust system's functional role in motivating the avoidance of debilitating or life-threatening infection. We therefore took a different tack from previous explorations of the structure of disgust, generating a set of stimuli derived from the epidemiology of disease risk in humans and then examining patterns of co-variation of reported disgust levels in a large international sample. We assumed that clusters of stimuli that correlated strongly with each other would reflect categories of input to the disgust system. Several plausible domain structures emerged from the analysis. The MAP analysis pointed to a six-factor solution. A one-factor solution had some statistical support, and a five-factor solution had fewer cross-loadings and weak factors. The five- and six-factor models differed only in their treatment of food items (see below). These analyses indicate that disgust takes as input the following kinds of cues:

• — Hygiene: displays of, or physical evidence of, unhygienic behaviour.

• — Animals/insects: such as mice and mosquitoes that represent disease vectors.

• — Sex: behaviour pertaining to promiscuous sexual activities. (However, symptoms of sexually transmitted diseases on genitalia loaded on the lesion factor.)

• — Atypical appearance: infection cues in other people including abnormal body shape, deformity, behaviour such as wheezing or coughing and contextual cues related to high risk such as homelessness.

• — Lesions: stimuli related to signs of infection on the body surface such as blisters, boils or pus.

• — Food: food items that show signs of spoilage. (This factor was weak and in a five-factor model, these items loaded with the animal/insects factor.)

These results partially supported our initial prediction that different kinds of disgust would reflect the different transmission routes of infectious disease, but they also departed from our prediction in an interesting way. It appears that cues to infectious disease threats are not categorised following the abstract biomedical categories of disease transmission risk recognized in the literature (interpersonal by contact, sexual activity or droplet, vectors, ingestion, fomite), but, rather, as categories of recognizable cues as to what to avoid. These include potentially contaminated objects such as bodily fluids, infected lesions, spoilt foodstuffs, and animals that vector disease, practices such as those who run the risk of contracting sexually transmitted diseases, and people who display visible signs of disease or poor hygiene. This six-factor categorization makes sense from the point of view of a pathogen detection system that could not ‘see’ microscopic pathogenic microbes and parasites. Instead selection operated on behavioural avoidances of specific categories of people, practices and objects.

We expected our domain structure to differ from that suggested in the existing psychology literature, because we began from a functional perspective grounded in infectious disease risk avoidance. Our findings are, to some extent, consistent with the three factors of the DS-R (core, animal reminder and contamination) and also with the three domains of the TDDS (pathogen, sexual and moral disgust). Our food/animal domain clearly overlaps with ‘core disgust’ in the DS-R (example DS-R item: ‘Seeing a cockroach in someone else's house doesn't bother me’). Our one-factor solution is consistent with the pathogen disgust domain of the TDDS, and our five- and six-factor solutions provide more granularity within it. Our sexdisgust domain has similar content to that of the TDDS. The hygiene input category, which was one of the largest in our model (in the factor model, 40% of our disease items loaded on this factor; table 2) was largely missing from both the TDDS and DS-R.

The DS-R and the TDDS models make different claims about the kinds of things that elicit disgust. It is important to note, however, that these measures can only characterize inputs that were included in the original item pool. Both scales were originally generated by asking a small number of students and professors to generate lists of items they found disgusting. However, self-reported lists of disgusting items may be biased both by limited experience and by normative pressures. People are unlikely to mention things that they rarely encounter. Stimuli associated with most infectious diseases such as sanitation, animal husbandry and ectoparasites are no longer part of many peoples'—especially American students’ and professors’—everyday life [46]. Further, social norms proscribe the open discussion of many disgust- and disease-relevant subjects. The fact that people may be unwilling to mention deformity or sexual behaviour may explain why the DS-R is missing items that relate to cues of sources of infection such as sexual promiscuity, and atypical appearance, which emerged as factors in our model.

Because we began with a prediction that the majority of disgust elicitors would play a role in infectious disease transmission, we did not focus on moral disgust. We have elsewhere suggested that moral disgust may have arisen as an extension of hygiene disgust [11]. In effect, a system that causes one to distance oneself from conspecifics who represent a possible source of infection could come to be used to distance oneself from those who perpetrate other social infractions. Expressions of disgust and a refusal to engage in social interaction are a cheap form of punishment that does not carry the likely costs associated with angry displays of aggression [47–49].

We assumed that this work would uncover different categories of input to the disgust processing system. However, one could argue that these six domains better represent categories of output; i.e. categories of desirable behavioural responses to a putative pathogen threat. While a general avoidance response is appropriate for all six types of infection cue, more specific types of behaviour are called for to tackle each of these types of threat. For example, the lesion response should specifically limit behaviour that brings one into direct contact with wounds and sores, the sexdisgust response limit promiscuity and contact with the promiscuous, the hygiene and the atypical appearancedisgust factors promote social distancing behaviour, fooddisgust limit the ingestion of suspect foodstuffs and animal disgust limit contact with potential disease vectors. In effect, we propose that these different categories of avoidance behaviours represent the phenotype that underwent selection pressure from infectious disease, producing this specific architecture in the disgust adaptive system.

While this work was carried out in humans, we suspect that psychological/behavioural disease-avoidance systems are as ancient as the threat of parasitism, and that the different types of pathogen disgust diverged long before humans inherited its basic components. Hence, we should expect to see analogues of this structure across species. Behavioural responses to infectious threat should prove fertile for further research, both in humans and other animals. For example, we would predict behavioural tradeoffs to differ across the disgust domains. The sex disgust response should be supressed as arousal increases [50], fooddisgust diminish with increasing hunger and atypical appearancedisgust diminish as the nurture motive increases with increasing kinship. In humans and other social animals we would expect hygienedisgust to be supressed as the affiliation motive increases and in humans only, we would expect to see hygienedisgust covary with levels of moral outrage [51]. For each of these tests, only the specific type of disgust should be implicated, with other kinds being less affected. Trade-offs would be expected to differ with age and gender: for example, a stronger female nurture motive might explain why the well-known male and female disgust differences [10] did not here apply to the atypical appearance domain. In infants, the curiosity motive may trump hygiene disgust during potty play.

While the item pool used in this study was derived exclusively from cues related to infectious disease avoidance, it is likely that other selection pressures have shaped the evolution of the emotion. For instance, sex disgust—factor 3 in the above model—has patterns of between- and within-person variation that suggest that it plays a role in the avoidance of inbreeding [52–55]. Further work on pathogen and inbreeding avoidance behaviour and mechanisms in other species could help to reconstruct the relative importance of these selection pressures during the phylogeny of disgust.

What is the point of drawing yet more fine distinctions about the structure and function of disgust? There are several reasons. First, if ancestral disease avoidance is the ultimate purpose of the disgust adaptive system then any scales that we employ to measure it should reflect this. Second, teasing out the various facets of disgust helps to generate new hypotheses, such as on the mechanics of how trade-offs between motives gain control of behaviour, or whether disgust is oral in its origins, as some contend [25] (and this work suggests it is not, as food was our weakest factor). Third, a better understanding of the structure of the brain's disease-avoidance systems can assist those working in public health. By understanding the ways in which we respond to disease threats, we can work with and sometimes counteract those responses—for example, to avoid the harm that can be done when we mistakenly see atypical appearance as a cue to infection. Fourth, it is likely that malfunctioning of the disgust system's subdomains may give rise to specific clinical manifestations; for example, ‘heebie-jeebies’ [56] and trypophobia (Kupfer & Fessler, this issue) [57] may relate to the skin lesion and animal/insect subdomains of disgust respectively. Finally, we need to better understand and measure the subdomains of disgust for the purposes of research into animal disease-avoidance behaviour.

5. Conclusion

We have presented evidence that the disgust motive has a factor structure that reflects the different tasks that human ancestors have had to accomplish to avoid falling prey to infectious disease. These were to avoid objects, skin lesions, spoilt foods and animal vectors, and individuals with poor hygiene and of atypical appearance and promiscuous sexual practices. This six-factor categorization is likely to reflect a pathogen detection system that could not ‘see’ microscopic pathogenic microbes and parasites directly, but could only evolve behavioural responses to categories of perceptible cues as to what to avoid: the people, practices and objects that have tended to co-occur with infectious disease. This study provides further evidence that disgust does indeed serve to prevent infectious disease, among other functions. It adds weight to the suggestion that parasite avoidance theory should be at the centre of future development of disgust science, for example, in the examination of the behavioural tradeoffs that have to be made to meet multiple needs while avoiding disease, and suggests the need for further work to develop new scales for clinical and psychological investigation. Disgust studies have great potential to elucidate the structure, function and emergence of emotions [10].

Data accessibility

The datasets supporting this article have been uploaded to de Barra, Mícheál (2017): Disgust responses to disease related stimuli. See http://dx.doi.org/10.6084/m9.figshare.5481139.

Authors' contributions

M.d.B. and V.C. designed the study. M.d.B. collected and analysed the data. V.C. and M.d.B. wrote the manuscript. Both authors gave final approval for publication.

Competing interests

We declare we have no competing interests.

Funding

We received no funding for this study.

Footnotes

One contribution of 14 to a Theo Murphy meeting issue ‘Evolution of pathogen and parasite avoidance behaviours’.