An Examination of the Westermarck Hypothesis and the Role of Disgust in Incest Avoidance Among Fathers

Abstract

From an evolutionary perspective, incestuous behavior is puzzling. The goal of this study was to assess the tenability of the Westermarck hypothesis (1891, 1921)—that people who live in close physical proximity with one another during childhood will develop a sexual indifference or aversion toward one another—and the mediating role of disgust as an incest avoidance mechanism in father–daughter relationships. A sample of fathers with daughters (N = 632) from Canada and the United States were recruited by Qualtrics—a survey platform and project management company—to complete an online survey. The results from this study did not support the viability of the Westermarck hypothesis as a mechanism that facilitates incest avoidance for fathers. Physical proximity was not associated with incest propensity or disgust toward incest. Less disgust toward incest, however, was found to be associated with more incest propensity. These results indicate that physical proximity may not be a reliable kinship cue used by fathers to inform incest avoidance, but that disgust toward incest may still be a proximate mechanism that facilitates incest avoidance among fathers using kinship cues other than physical proximity.

Stoltenborgh, van IJzendoorn, Euser, and Bakermans-Kranenburg (2011) conducted an international meta-analysis and found that the prevalence of intrafamilial child sexual abuse was 18% for girls and 8% for boys ¹ (see also Stoltenborgh, Bakermans-Kranenburg, van IJzendoorn, & Alink, 2013). The consequences associated with child sexual abuse can include poorer psychological well-being, higher rates of mental illness, and decreased life satisfaction (e.g., Fergusson, McLeod, & Horwood, 2013; Ratican, 1992; Roberts, O’Connor, Dunn, Golding, & ALSPAC Study Team, 2004). Research also suggests that the negative consequences associated with incestuous abuse by a relative is even greater than the negative consequences experienced by victims of child abuse committed by nonrelatives (Stroebel et al., 2012).

Despite its prevalence, incestuous sexual behavior is puzzling from an evolutionary perspective. Due to the effects of inbreeding depression—increased genetic homozygosity that increases morbidity and mortality in offspring (e.g., Charlesworth & Willis, 2009; Seemanovà, 1971)—offspring of close genetic relatives reduces reproductive fitness for both parents. This selection pressure opposes incestuous behavior. Despite this selection pressure, however, incestuous behavior still occurs.

Kinship Cues and Incest Avoidance

In order to facilitate the avoidance of incestuous behavior, incest avoidance mechanisms must be able to identify kin. Genetic relatedness is not a trait that can be directly observed. Therefore, selection pressures have promoted the retention of genes that expressed traits that allowed us to distinguish kin from nonkin. Prior work on kin detection in humans suggests that perceived relatedness is based on the presence or absence of cues that ancestrally correlated with genetic relatedness (Wolf, 2014) and that humans avoid incest according to these cues (Lieberman, Tooby, & Cosmides, 2007). Importantly, the evolution of these traits is required in order for incest avoidance mechanisms to function. Humans need to be able to identify kin to avoid sexual relationships with them. In the absence of these relatedness cues—or in the presence of cues that indicate relatedness uncertainty—the incest avoidance mechanism may fail to inhibit incestuous behavior.

A biological mother has the evolutionary advantage of being certain that her child shares 50% of her genes because (ancestrally) mothers carry their child throughout pregnancy and give birth. A father, however, is not certain of his genetic relatedness to his child, as his partner may have conceived the child with another man. Indeed, it has been conservatively estimated that approximately 2% of men in the general population are unknowingly raising a child who shares none of their genes (see Anderson, 2006). Selection pressures have promoted the evolution of traits that allowed fathers to identify cues of paternity in ancestral environments.

There are a variety of cues that fathers may use to assess paternity certainty. For example, phenotypic similarity refers to the extent to which relatives share a similar expression of phenotypic traits. There are a variety of phenotypic traits that have been investigated in human and animal models that have been shown to act as kinship cues. Among avian species, auditory vocal signals are often used as kinship cues (Beecher, 1988). In both humans and mammals, similarity in odor has been shown to distinguish kin from nonkin (e.g., Heth, Todrank, & Johnston, 1999). Phenotypic resemblance in physical features can also be used as a kinship cue. Indeed, fathers tend to favor children (i.e., provide more resources) who look more like them (Apicella & Marlowe, 2004; Burch & Gallup, 2000). In a hypothetical scenario, parents (particularly fathers) are more likely to adopt unfamiliar infants when the perceived level of physical resemblance between the parent and infant is high (Volk & Quinsey, 2002, 2007). In a college sample, DeBruine (2005) found that when presented with the faces of opposite-sex strangers who were either physically similar to or different from themselves, participants rated the similar strangers as less sexually attractive and more trustworthy.

Aspects of the marital relationship may also act as cues of paternity, as they may signal risk of partner infidelity. This may include spousal relationship conflict and relationship quality prior to and around time of conception (see Tal & Lieberman, 2007, for a more detailed analysis). For example, Billingsley, Antfolk, Santtila, and Lieberman (2018) found that suspicions of partner infidelity and jealousy in the marital relationship were negatively associated with levels of sexual arousal that fathers experienced toward their daughters. Furthermore, Williams and Finkelhor (1995) conducted a study with a sample of Navy fathers and found that fathers who had committed a sexual offense against their daughters had lower levels of marital satisfaction compared to fathers who did not commit a sexual offense.

The Westermarck Hypothesis

Westermarck (1891, 1921) postulated that physical proximity was a kinship cue used to regulate incest avoidance in sibling relationships. Siblings who have close physical proximity during childhood are expected to experience sexual indifference toward one another. Prior researchers have found evidence for a “critical window” in which the effect of physical proximity is most influential for incest avoidance and is often specified as the first 6 years of life (Shepher, 1971), although some authors suggest younger (before age 3; Wolf, 1995) and some authors suggest older (before age 10; Bevc & Silverman, 2000) critical ranges. If early physical proximity does not occur, Westermarck proposes that the likelihood of incestuous behavior between siblings will increase because incest avoidance has not been activated. This theory has found empirical support in the incest avoidance literature. Multiple studies have found that genetically unrelated children who are raised together from birth are unlikely to develop sexual relationships (e.g., Shepher, 1971; Spiro, 1958; Talmon, 1964). Additionally, Bevc and Silverman (1993, 2000) found that siblings who were separated during the first few years of life were more likely to engage in incestuous behavior that could lead to reproduction (vaginal intercourse) compared to siblings who were not separated. This relationship was not found for incestuous behaviors that could not result in reproduction (e.g., kissing, fondling).

Can the Westermarck hypothesis also explain incest avoidance in the context of father–daughter relationships? A putative father who purposefully engages in close physical proximity with a child may be exhibiting signs of paternity confidence, exhibiting a willingness to expend resources on the child. Furthermore, close physical proximity with a young child gives a putative father time to gather more evidence about known kinship cues such as phenotypic resemblance. If a father is not present in his child’s life during the child’s early developmental years, then the underlying tenets of the Westermarck hypothesis would suggest that the incest avoidance mechanism will not be activated or will be weaker, making it more likely that they will commit a sexual offense against their child. Very little research, however, has been conducted to assess this hypothesis in father–daughter relationships. Parker and Parker (1986) found that biological and stepfathers who had committed a sexual offense against their daughter were less likely to be present in the home during their daughters’ early childhood compared to nonoffending biological and stepfathers (34% vs. 70%). However, there was an important confound: Fathers could be voluntarily choosing to spend less time with their children. As such, men who choose not to spend time with their children (e.g., because of a lack of parental solicitude—the degree of affection, care, and concern a parent experiences toward their child—or because of a high degree of antisocial tendencies) may also be more likely to engage in incest. Williams and Finkelhor (1995) controlled for this confound by examining physical proximity between fathers and their daughters in a sample of men in the Navy. These men were often deployed, and thus spent significant amounts of time away from their children, but these separations were not voluntary. These authors did not find a difference in the proportion of time spent away from home in the first 4 years of their daughter’s life between incestuous and nonincestuous fathers, but did find that a lower frequency of performing solicitous behaviors (e.g., feeding child, bathing child), along with higher levels of marital dissatisfaction, were related to an increased risk of incest. These results do not support the Westermarck hypothesis as a viable explanation for father–daughter incest, although they do suggest that other family dynamic factors that could influence cues of paternity (e.g., marital relationship conflict) could be important.

In a more recent examination of the Westermarck hypothesis among fathers, and using methods similar to the current study, Kresanov et al. (2018) found that the proportion of coresidence duration a father had with his daughter (biological or sociolegal) was not associated with the level of sexual arousal or disgust toward an imagined sexual encounter. Participants were asked how arousing and how disgusting they found three separate imagined sexual scenarios with their daughter: kissing, seductive touching, and intercourse. The level of sexual arousal/disgust, however, was collapsed across the three scenarios to compute a total score for arousal and disgust. The influence of inbreeding depression may only be applicable to incestuous behavior that could result in reproduction (i.e., penile–vaginal intercourse). Indeed, as previously discussed, Bevc and Silverman (1993, 2000) found that less coresidence between siblings was associated with an increased likelihood of incestuous behavior that could result in reproduction (i.e., penile–vaginal intercourse), but coresidence duration was not associated with nonreproductive sexual behavior (e.g., kissing, genital touching).

Disgust

While support has been found for the Westermarck hypothesis for sibling incest, this support still does not address the nature of the incest avoidance mechanism or if it is applicable to fathers. How does close physical proximity facilitate the development of incest avoidance? When asked to explicitly think about a sexual relationship with a close relative, most people report feelings of disgust. Indeed, prior research has shown that disgust is a response that triggers the avoidance of sexual behaviors that are harmful to fitness (see Tybur, Lieberman, & Griskevicius, 2009, for a review). It is therefore theoretically plausible that disgust is the proximate mechanism that ensures humans avoid incest. In the context of this study, physical proximity acts as a cue of genetic relatedness. Once kin recognition has been activated, a disgust response toward the idea of sex with that relative develops.

Specifically addressing the relationship between physical proximity and disgust toward incest, Lieberman, Tooby, and Cosmides (2003) found that early physical proximity with an opposite-sex sibling was significantly and positively correlated with disgust toward incest. This trend was not found for the relationship between physical proximity with a same-sex sibling and disgust toward incest. Additionally, the authors found that disgust toward sibling incest was significantly lower for participants who had opposite-sex siblings with whom they did not coreside during childhood compared to participants who had opposite-sex siblings with whom they did coreside during childhood. Lastly, these authors found that physical proximity is a stronger predictor of disgust toward incest than the actual degree of genetic relatedness between siblings, and once controlling for physical proximity, degree of genetic relatedness no longer predicts disgust toward incest. These results have been replicated by the same authors in a subsequent study (Lieberman et al., 2007; see also Lieberman, & Lobel, 2012). To the best of our knowledge, the relationship between physical proximity and disgust toward incest in a sample of fathers has only been examined in one previous study (Kresanov et al., 2018) with null results. However, as mentioned, this study did not differentiate between different types of sexual activity in the imagined scenarios. The current study addresses this limitation by distinguishing between intercourse and other sexual acts. Furthermore, this is the first study to assess the potential mediating effect that disgust toward incest has on the relationship between physical proximity and incest propensity among fathers.

Measuring Incest and Disgust

The most direct method for measuring incestuous behavior is to ask participants to self-report whether they have ever engaged in sexual behavior with their child. In nonforensic samples, though, this will likely evoke socially desirable responding, especially when such behavior is often illegal. Similarly, asking fathers how disgusting they find the idea of sex with their daughter may also evoke socially desirable responding and generate ceiling effects (i.e., most respondents rating their disgust very high, with little variability). One method for minimizing this bias and reducing ceiling effects is to ask participants to read a vignette describing third-party incestuous behavior, where the described characters are not themselves or their own child. Participants are then asked a series of questions about each vignette, aimed at measuring their propensity for the behavior described, as well as their level of disgust toward the behavior. This method is based on the premise of egocentric empathy, which is the tendency to view another person’s behavior from one’s own perspective; reactions are expected to be stronger in individuals who have a potential incest target, such as an opposite-sex sibling (Fessler & Navarrete, 2004). These third-party vignette methods have been successfully used in previous research with siblings to measure disgust toward incest (Antfolk, Lieberman, & Santtila, 2012; Lespiau & Kaminski, 2016).

Disgust toward incest may also be culturally influenced. There are very strong cultural incest taboos in all societies, although the degree of relatedness included in the taboo does vary (see Wolf, 2014, for a review). Research suggests that there is a connection between the emotion of disgust and moral judgments. When people make judgments about an object or situation, they rely on their affect (Schwarz & Clore, 1983). If a person feels disgust, they will judge a situation more severely than if they do not feel disgust. Asking an individual to rate how morally wrong they view a particular situation, therefore, should involve the feeling of disgust as salient information (e.g., Schnall, Haidt, Clore, & Jordan, 2008). Indeed, Fox (1980) argues that the incest taboo (a moral judgment about incestuous behavior) originates in the fact that humans have evolved an innate disgust mechanism to ensure that we avoid incest. Furthermore, Turner and Maryanski (2005) elaborate, suggesting that while this innate evolved mechanism was sufficient to ensure mother–sons and brother–sisters avoided incest, a cultural taboo of incest was necessary to ensure fathers do not engage in incest with their daughters, due to the nuclear family being relatively new in our ancestral history, thus less opportunity for natural selection to select for innate biological mechanisms to regulate incest avoidance in father–daughter dyads.

Empirical evidence for the connection between disgust and moral judgments has been found using many different research designs. For example, there is substantial neurological overlap between areas of the brain that are activated when feeling disgust and when making a moral judgment (Moll et al., 2005). Additionally, disgust is known as one of the six universal facial expressions (Ekman et al., 1987), and facial expressions of disgust elicited by aversive tastes or viewing contaminants are very similar to facial expressions produced by moral transgressions (Chapman, Kim, Susskind, & Anderson, 2009). Schnall, Haidt, Clore, and Jordan (2008) conducted a series of experiments in which they found that artificially eliciting a disgust response provoked participants to rate a number of situations more severely in regard to moral wrongness than individuals in the control conditions when disgust was not provoked. This was a linear relationship: The more disgust that was evoked, the more severely they regarded the moral transgression. Additionally, the effect was specific to the induction of disgust, in comparison to sadness, suggesting that the effect is not explained by any negative affect. Based on this literature, disgust toward incest can be measured through third-party descriptions of incestuous behavior, as well as with judgments of moral wrongness toward the incestuous behaviors described. Indeed, judgments of moral wrongness toward third-party incest have also been successfully used in previous research of sibling incest to operationalize disgust toward incest (Lieberman, Tooby, & Cosmides, 2003; Lieberman et al., 2007).

Antisociality

There is a solid foundation for understanding the causes of child sexual abuse. Two broad factors that have been implicated in both the initiation and persistence of child sexual abuse are atypical sexual interests and antisociality. Atypical sexual interests include sexual interest in prepubescent children (pedophilia) or pubescent children (hebephilia), as well as excessive sexual preoccupation (often referred to as hypersexuality). Antisociality refers to a constellation of traits including impulsivity, risk-taking, callousness, and offense-supportive attitudes. These two factors have been highlighted in theoretical models of the onset of child sexual abuse (e.g., Quinsey, 1986; Seto, 2019) and are also two of the strongest predictors of sexual and nonsexual recidivism in sexual offenders (Hanson & Morton-Bourgon, 2004, 2005), indicating these variables play important roles in the persistence of these behaviors as well.

Incest offenders should have more forces inhibiting their behavior compared to those who offend against unrelated children—specifically, incest avoidance and incest taboo. It would follow from contemporary models of child sexual abuse that individuals who engage in incest would be more problematic in atypical sexual interests and/or antisociality, factors known to be associated with sexual offending, in order to overwhelm these inhibitions. Meta-analytic results, however, have found that incestuous sexual offenders against children are less likely to display atypical sexual interests or antisociality compared to sexual offenders against unrelated children (Seto, Babchishin, Pullman, & McPhail, 2015), suggesting that these factors alone are not adequate to explain the crimes committed by incest offenders.

Furthermore, another meta-analysis (Pullman, Sawatsky, Babchishin, McPhail, & Seto, 2017) examined the distinction between biological and sociolegal incest offenders who were primarily fathers who committed a sexual offense against their biologically or sociolegally (e.g., step, adopted) related child, respectively. Pullman et al. found that while there were no differences between biological and sociolegal fathers in the prevalence of atypical sexual interests, sociolegal incest offenders were more problematic than biological incest offenders on some indicators of antisociality. These results suggest that traditional theories of child sexual abuse that highlight antisociality as a facilitating factor (e.g., Seto, 2019) may be more applicable in explaining child sexual abuse committed by sociolegal fathers than biological fathers.

Antisociality is also an important confounding factor; antisocial fathers may choose to be away from their children (i.e., have less physical proximity) and may also be more likely to engage in incest. Furthermore, antisocial individuals may exhibit attentional biases toward certain kinship cues. For example, correlational research has identified a relationship between aggression (Archer & Webb, 2006) and impulsivity (Costa, Sophia, Sanches, Tavares, & Zilberman, 2015) with sexual jealousy, suggesting that individuals high in antisociality may overestimate paternity uncertainty.

Hypotheses

Very few studies have examined the Westermarck hypothesis in father–daughter relationships (Kresanov et al., 2018; Parker & Parker, 1986; Williams & Finkelhor, 1995), and no study has examined the mediating role of disgust in the relationship between physical proximity and incest propensity among fathers. The aim of this study is to examine the tenability of the Westermarck hypothesis in the context of both biological and sociolegal father–daughter relationships and assess the mediating role that disgust toward incest has in the relationship between physical proximity and incest avoidance, while simultaneously controlling for the confounding influence of antisociality.

The primary hypothesis for this study is that the relationship between physical proximity and incest propensity will be mediated by disgust toward incest. Less physical proximity with a child is expected to be associated with less disgust toward incest, which in turn is associated with increased incest propensity. Predictions were also made about factors that may moderate the strength of this mediated relationship. For example, while it is expected that sociolegal fathers will have a higher level of incest propensity than biological fathers (due to the nature of nonnuclear families and how this would impact physical proximity with a child), because humans rely on kinship cues like physical proximity to provide information to our incest avoidance mechanisms as opposed to relying on actual genetic relatedness, relationship type is not expected to influence the strength of this indirect relationship. Additionally, given that the Westermarck hypothesis is based on the possibility of inbreeding depression, the strength of the indirect effect is expected to be larger when modeling incest propensity for reproductive sexual behavior (penile–vaginal intercourse) than nonreproductive sexual behavior (e.g., genital fondling).

Materials and Methods

Participants

Inclusion criteria for this study specified that all participants had to be 18 years old or older, be proficient in English, and be a father with at least one daughter (biological, step, common-law, adopted). A total of 983 fathers began this study. Of these, 148 participants withdrew before the end of the survey. Additionally, 2 participants were excluded for completing the survey too quickly (less than one third the median length), and a further 181 participants were excluded for failing one or more validity/consistency criteria, which suggested inattentive responding. Lastly, 20 participants were excluded for refusing to answer 10% or more of the questions in the survey (see the Supplemental Materials for more details regarding this participant selection procedure). This resulted in a final sample size of 632 participants. Of the participants included in the final sample, 74% lived in the United States and 26% lived in Canada. On average, they were over 40 years old (M = 42.7, SD = 11.6) and had completed at least some college or university training (71%). These participants had, on average, 3.3 children (M_daughter = 2.1, SD = 1.3). The proportions of the sample who were biological (45%) or sociolegal (55%) fathers were similar, due to the recruitment procedure employed (see the Procedure section for recruitment details). See Table 1 for more descriptive characteristics of this sample.

Table 1.

Demographic and Descriptive Characteristics of Fathers.

Characteristic	%/M (SD)
Age	42.74 (11.60)
Education
Less than elementary school  Elementary school  Some high school  Completed high school  Some college/university  Completed college/university  Graduate school	0.0% 0.1% 5.7% 23.6% 30.1% 31.3% 9.2%
Incomea
$0 to $10,000  $11,000 to $20,000  $21,000 to $30,000  $31,000 to $40,000  $41,000 to $50,000  $51,000 to $60,000  $61,000 to $70,000  $71,000 to $80,000  $81,000 or more	10.0% 8.8% 10.4% 10.6% 13.3% 11.2% 7.9% 8.5% 19.3%
Country of residence
Canada  USA	25.6% 74.4%
Current marital status (N = 629)
Married  Not married	65.3% 34.7%
Number of childrenb	3.27 (2.03)
Number of female childrenb	2.08 (1.26)
Type of father (N = 631)
Biological father  Sociolegal father	44.7% 55.3%
Incest propensityb (N = 610), range = 12–120	31.73 (12.06)
Disgust toward incestb (N = 623), range = 8–80	70.83 (12.77)
Physical proximity (N = 624), range = 0–1	0.64 (0.37)

Note. Unless otherwise noted, the sample size was based on N = 632.

^aIncome was expressed as $CAD or $USD, depending on where the participant lived.

^bOutliers were identified by visual inspection of histograms, followed by review of z-scores (z > 3.29). Outliers were reduced in magnitude but maintain their rank order.

Materials

While the survey was extensive and asked questions about a wide range of topics for the purpose of a larger program of research, only those measures that are pertinent to the current study will be described herein. These included measures of childhood abuse history, sexual behavior history, sexual interests, parental solicitude, parent–child relationship quality, and marital relationship characteristics.

Demographic characteristics

Participants were asked to report a number of demographic characteristics about themselves (e.g., age, sex, education), as well as characteristics about their children (e.g., age, sex, degree of relatedness). These variables were taken directly from participant responses on the survey.

Antisociality

Both childhood and adult antisociality were measured. For childhood antisociality, participants were asked whether, before the age of 15, they engaged in 12 activities indicative of conduct-disordered behavior in childhood (e.g., skipping school often, initiating physical fights often, vandalism). The positive (yes) responses were counted as 1 point and summed to create a total score. Scores could therefore range from 0 to 12, with higher scores indicating more childhood antisociality. These 12 items had a high degree of internal consistency (Cronbach’s α = .83). For adult antisociality, participants were asked whether they had ever been arrested, charged, or convicted for a (1) nonsexual violent offense (e.g., assault), (2) nonsexual nonviolent offense (e.g., theft), and (3) a contact sexual offense (e.g., forcing someone to engage in sexual activity). All three questions had yes/no response options. The original intent was to create a total score for adult antisociality by summing the positive (yes) responses, but internal consistency was very low (α = .41). Therefore, a new variable representing whether the participant had been arrested, charged, or convicted for any type of crime was computed (0 = no, 1 = yes) and used in the analyses.

Physical proximity

Participants were asked a number of questions about their living arrangements with their daughters, including (1) their child’s current age, (2) if they had ever lived in the same residence with their child, (3) how old the child was when they started living with them (from birth to 20 years old or older), and (4) how old the child was when they stopped living with them (still living with child or less than 1 year old to 20 years old or older). Using these questions, and similar to Kresanov et al. (2018), we calculated the proportion of the child’s life (before the age of 18) that the father lived with the child. For example, if the child is 6 years old and the father has lived with the child for all 6 years of their life, their score would be 1. If the father is still currently living with the child, the child is 16 years old, and the father started living with the child when they were 10 (6 years in total), their score would be 0.375 (6/16). If the father has never lived with his child, their score would be 0.

Incest propensity

Based on the premise of egocentric empathy (Fessler & Navarrete, 2004), participants were asked to read four vignettes (adapted from Albrecht et al., 2014) describing third-party sexual activity between a father and his daughter and asked questions that would indicate a propensity for the behavior described. For biological fathers, the vignettes indicated a sexual encounter between a father and his “young daughter,” whereas for sociolegal fathers (step/common-law/adopted), the vignettes indicated a sexual encounter between a father and his “young stepdaughter.” All other aspects of the vignettes were the same regardless of participant characteristics. Two of the vignettes described nonreproductive sexual behavior (genital fondling), whereas two of the vignettes described reproductive sexual behavior (intercourse; Bevc & Silverman, 1993, 2000). Participants were asked “How likely do you think it is that the man in this story will encourage continued sexual contact with his daughter?” and “If you were in a similar situation, how likely would you be to encourage continued sexual contact with your daughter?” These items were rated on a 1 (not at all likely) to 10 (extremely likely) scale. Furthermore, participants were also asked “How arousing do you find this story?” and indicated their response on a 1 (not at all arousing) to 10 (extremely arousing) scale. These 12 items (3 items for each of the four vignettes) were summed to create a total incest propensity score. Scores could range from 12 to 120, and higher scores indicated a higher level of incest propensity. These questions were always presented in the same order as presented above. These 12 items had an acceptable level of consistency (α = .79).

Disgust toward incest

Responding to the same four vignettes as described above, participants were also asked “How disgusting do you find this story?” with responses provided on a 1 (not at all disgusting) to 10 (extremely disgusting) scale and “How morally wrong do you find the man’s behavior in this story?” which was measured on a 1 (not at all wrong) to 10 (extremely wrong) scale. These 8 items (2 items for each of the four vignettes) were summed to create a total disgust toward incest score. Scores could range from 8 to 80, and higher scores indicated a higher level of disgust toward incest. These questions were always presented in the same order as presented above. These 8 items had an acceptable level of consistency (α = .83).

Procedure

Ethical approval for this study was granted by the Royal Ottawa Health Care Group’s Research Ethics Board (Reference # 2015016) and the University of Ottawa Ethics Board (Reference # H-07-17-39). Participants for this study were recruited by Qualtrics (https://www.qualtrics.com/uk/online-sample/)—a survey platform and project management company. Qualtrics has developed an online community of individuals (panels) interested in participating in survey research. Individuals interested in being a panel member sign-up online, fill out a comprehensive demographic questionnaire, and are sent invitation e-mails when a new survey is available that they are eligible to participate in. Panel members are paid an agreed amount by Qualtrics for participating in a survey, and Qualtrics was paid by us for each survey participant. Prior research has demonstrated that Qualtrics panel members have similar demographic, geographic, and political characteristics to the U.S. population (Heen, Lieberman, & Miethe, 2014) and that in general, research panels can be considered a close approximation of the general population (Goodman, Cryder, & Cheema, 2013; Paolacci, Chandler, & Ipeirotis, 2010).

Qualtrics recruited participants for this study from their online research panel. Invitations to participate were sent to potential participants who met the basic inclusion criteria of the study (male, over 18 years old, had a child). Interested participants were directed to the survey, where they provided informed consent by reading a detailed consent form, followed by the statement “Click ‘I Agree’ to indicate that you understand the information above and would like to participate in this study or ‘I Disagree’ if you do not want to do the survey.” Data collection for this survey was conducted in two waves. Due to the small number of sociolegal fathers recruited in the first wave of data collection (7%), the second wave restricted the sample to sociolegal fathers. Furthermore, 21 participants from Wave 1 of the survey were subsequently excluded because despite instructions, they answered the questions in the survey about their son, not their daughter. In Wave 2 of the survey, participants who did not follow instructions (i.e., indicated they were answering the questions about a sociolegal son as opposed to a sociolegal daughter) were redirected out of the survey. Data for Wave 1 were collected between November 16 and November 25, 2017. Data for Wave 2 were collected between February 28 and March 19, 2018. In Wave 1 of the survey, participants who had multiple female children were asked to answer the questions concerning proximity about their oldest biological daughter. If they didn’t have any biological daughters, then they provided responses about their oldest sociolegal daughter. In Wave 2 of the survey, participants were asked to provide responses about their oldest sociolegal daughter.

Using Qualtrics to recruit participants meant that identifying information about participants was not recorded in the data set, but the Qualtrics server that stored the electronic data did record participant IP addresses. Therefore, because of the potential for a participant’s responses to be connected to their identities, questions concerning undetected criminal behavior, where a specific victim could be identified (e.g., a child), were not asked.

Results

See the Supplemental Materials for details regarding planned statistical analyses and preparing the data for analyses (pattern of missing data, ceiling and floor effects, statistical assumptions).

Degree of Genetic Relatedness

As expected, sociolegal fathers had less physical proximity with their daughters than biological fathers, d = −1.73, 95% confidence interval (CI) [−1.94, −1.52]. However, biological and sociolegal fathers did not differ in rates of disgust toward incest (d = −0.13, 95% CI [−0.29, 0.03]) or incest propensity (d = 0.03, 95% CI [−0.13, 0.18]), controlling for the participant and their daughters’ current age. Lastly, while biological and sociolegal fathers did not differ in rates of childhood antisociality, sociolegal fathers were more likely to have a criminal history than biological fathers (d = 0.30, 95% CI [0.10, 0.52]). See Table 2 for more details regarding these analyses.

Table 2.

Differences Between Biological and Sociolegal Fathers.

	Mean Rank/M (SD)		N	H	p	d [95% CI]
	Biological Fathers	Sociolegal Fathers
Physical proximity	435.55/0.89 (0.27)	208.19/0.45 (0.32)	620	265.54	<.001	−1.73 [−1.94, −1.52]
Childhood antisociality	298.98/1.72 (2.49)	317.00/1.92 (2.50)	617	1.75	.187	0.11 [−0.05, 0.26]
			N	F	p	d [95% CI]
Disgust toward incest	326.24/72.05 (11.85)	300.18/69.82 (13.40)	619	2.51	.114	−0.13 [−0.29, 0.03]
Incest propensity	302.52/31.43 (12.15)	308.44/31.98 (12.01)	606	0.10	.751	0.03 [−0.13, 0.18]
	% Yes		N	χ2	p	d [95% CI]
Adult criminal history	23.1%	33.1%	628	7.61	.006	0.30 [0.10, 0.52]

Note. H = Kruskal–Wallis ranked analysis of variance (ANOVA). F = Quade’s ranked analysis of covariance (ANCOVA)—controlling for participant and child’s age. M (SD) reported for descriptive purposes only; Kruskal Wallis ranked ANOVA and Quade’s ranked ANCOVA calculated from ranked data. χ² = Pearson χ² test. Bolded values indicate statistical significance, p < .05. Cohen’s d is positive when sociolegal fathers had a higher score than biological fathers, and negative when biological fathers had a higher score than sociolegal fathers.

When restricting the sample to participants who scored in the top 10th percentile on incest propensity, as a group of fathers who are most likely to commit incest (scores above 43, N = 65), sociolegal fathers had higher scores on childhood antisociality compared to biological fathers, H = 4.65, p = .031, d = 0.56, 95% CI [0.05, 1.07]. Furthermore, sociolegal fathers were more likely to have a criminal history compared to biological fathers (30.6% vs. 6.9%, respectively), χ²(1, N = 65) = 5.62, p = .018, d = 0.97, 95% CI [0.08, 1.86].

Physical Proximity, Disgust Toward Incest, and Incest Propensity

After examination of the statistical assumptions associated with linear regression for each path of the mediation model, multivariate normality and homoscedasticity of residuals was found to be violated. Thus, a Box–Cox transformation (Box & Cox, 1964) was performed on all dependent variables in the regression model (disgust toward incest, λ = 5.00; incest propensity, λ = 0.20). Although improved, some paths still exhibited evidence of multivariate non-normality of residuals. However, while multivariate normality of residuals is important for computing valid prediction intervals, when regression is used to simply examine the associations between variables, it is robust to violations of multivariate normality of residuals with large (N > 500) samples (Lumley, Diehr, Emerson, & Chen, 2002). The participant’s age, their child’s age, as well as childhood and adult antisociality indicators were included in all models as covariates. To increase interpretation, all variables were standardized (z-scores) prior to modeling. Three multivariate outliers were identified and removed from the model (see the Supplemental Material for more information).

Contrary to expectations, physical proximity was not associated with incest propensity, b = −.01, SE = .04, 95% CI [−.09, .08)]. Physical proximity was also not associated with disgust toward incest, b = −.02, SE = .05, 95% CI [−.11, .07]. However, lower levels of disgust toward incest were associated with an increased propensity for incest among fathers, b = −.26, SE = .04, 95% CI [−.34, −.17]. The indirect effect was tested using a bootstrap estimation approach with 5,000 samples. These results indicated that disgust toward incest did not mediate the relationship between physical proximity and incest propensity, b = −.005, SE = .01, 95% CI [−.02, .03].

To account for the possibility that it is only physical proximity during the theorized critical window that influences incest avoidance, we also calculated the proportion of the child’s life, before the age of 6, the father had lived with his child. Using this new variable, physical proximity was still not associated with incest propensity, b = −.03, SE = .04, 95% CI [−.11, .06], or with disgust toward incest, b = .02, SE = .04, 95% CI [−.07, .10]. The indirect effect was tested using a bootstrap estimation approach with 5,000 samples. The results indicated that disgust toward incest did not mediate the relationship between physical proximity and incest propensity, b = −.004, SE = .01, 95% CI [−.03, .02]. Based on the similar results found for both operationalizations of physical proximity, all further analyses were conducted using the original physical proximity variable that accounts for the proportion of the first 18 years of the child’s life a father lived with his daughter.

The relationship between physical proximity, disgust toward incest, and incest propensity was further examined among biological and sociolegal fathers separately, using moderated mediation. Disgust toward incest did not mediate the relationship between physical proximity and incest propensity for biological (b = .03, SE = .02, 95% CI [−.001, .07]) or sociolegal (b = .02, SE = .02, 95% CI [−.02, .07]) fathers. There was not a meaningful difference in the magnitude of the difference between these indirect effects, index ² = .003, SE = .03, 95% CI [−.06, .05].

Lastly, sensitivity analyses were conducted to examine differences in the magnitude of the indirect effect, when predicting nonreproductive versus reproductive incest. After an examination of statistical assumptions, these variables were also transformed with a Box–Cox transformation (Box & Cox, 1964; disgust toward nonreproductive incest, λ = 5.00; disgust toward reproductive incest, λ = 5.00; propensity for nonreproductive incest, λ = −0.60; propensity for reproductive incest, λ = 0.90) and then standardized (z-scores). Three multivariate outliers were identified and removed in the nonreproductive incest model, while four multivariate outliers were identified and removed from the reproductive incest model. Furthermore, a single residual outlier was identified and removed in the reproductive incest model (see Supplemental Material). Disgust toward nonreproductive incest did not mediate the relationship between physical proximity and nonreproductive incest propensity, b = .002, SE = .004, 95% CI [−.003, .01], 84% CI [−.004, .01]. This model accounted for approximately 2% of the variance in nonreproductive incest propensity, r² = .02. Furthermore, disgust toward reproductive incest did not mediate the relationship between physical proximity and reproductive incest propensity, b = −.01, SE = .01, 95% CI [−.03, .01], 84% CI [−.02, .004]. This model accounted for 5% of the variance in reproductive incest propensity, r² = .05. Examining the 84% CIs for these two indirect effects reveals that they do overlap; hence, there was not a statistically significant difference (p > .05) in the magnitude of the indirect effect between predicting nonreproductive versus reproductive incest.

Discussion

The results obtained in this study do not support the Westermarck hypothesis as a mechanism that facilitates incest avoidance in our sample of 632 fathers. Physical proximity was not found to be associated with incest propensity nor did disgust toward incest mediate the relationship between physical proximity and incest propensity. These null results were consistent across biological and sociolegal fathers, as well as across propensity for reproductive versus nonreproductive incestuous behavior. These results are contrary to Parker and Parker (1986) but consistent with Williams and Finkelhor (1995) and Kresanov et al. (2018), suggesting that the accumulating evidence supports the applicability of the Westermarck hypothesis for avoidance of sibling incest but not father–daughter incest. We extended the study by Kresanov et al. by differentiating between different types of sexual behavior. Whether sexual behavior was potentially reproductive or not did not influence the lack of association between physical proximity and incest propensity.

We replicated prior research showing that lower levels of disgust toward incest were associated with higher levels of incest propensity (Kresanov et al., 2018). These results suggest that while physical proximity may not drive incest avoidance for fathers, disgust toward incest may still be the proximate mechanism that facilitates incest avoidance. Other kinship cues more closely related to paternity uncertainty, such as physical resemblance (phenotypic similarity) and perceived partner fidelity, may instead play a more critical role than physical proximity (e.g., Anderson, Kaplan, & Lancaster, 2007; Apicella & Marlowe, 2004; Burch & Gallup, 2000; Fox & Bruce, 2001). In a direct test of these kinship cues on incest avoidance, Billingsley et al. (2018) found that increased certainty of partner fidelity was associated with decreased levels of sexual arousal toward daughters as well as increased disgust toward sexual activity with daughters. With regard to physical resemblance, however, Billingsley et al. did not find a meaningful association between physical resemblance and sexual arousal or disgust toward sexual activity with a daughter.

Alternative Explanations

Post hoc power analysis assuming a small effect (2% variance explained; Cohen, 1988) showed we had an ample sample size to detect a mediation effect, if one was present (Fritz & MacKinnon, 2007). Therefore, it is necessary to begin to postulate alternative explanations for these null results. One alternative explanation is that the theorized incest avoidance mechanism only exists in the daughter, not the father (Rice & Harris, 2002). The costs associated with inbreeding depression are much greater for daughters than their fathers (Haig, 1999) due to differences in minimal parental investment, with an obligate 9 months of pregnancy and infant care for the daughter if she was potentially fertile (Trivers, 1972). Additionally, if there is a critical window in early childhood for the development of incest avoidance (e.g., under the age of 6; Shepher, 1971), then a father would be well outside this critical window by the time he had a child, but his daughter would go through this window. Therefore, incest avoidance mechanisms may have specifically evolved in human females to ensure that they avoid incest, but not in fathers. Due to the power dynamics associated with father–daughter relationships, daughters may be unable to resist sexual advances by fathers, despite the avoidance they experience. Indeed, among gorillas, adolescent females leave their harems early in life, perhaps as a mechanism designed to ensure inbreeding does not occur with fathers or brothers (Ghiglieri, 1999). Furthermore, among spotted hyenas, it is the males that migrate to other packs to search for mates, likely due to female mate choices that avoid inbreeding (Höner et al., 2007). Another possibility was raised by Turner and Maryanski (2005), who suggested that because the nuclear family is a relatively new system in our evolutionary history, natural selection has not had time to select for strong kinship cues used by fathers to inform incest avoidance. Therefore, fathers rely more heavily on cultural incest prohibitions (i.e., the incest taboo) to regulate sexual activity away from their daughters.

A surprising finding of this study was that sociolegal fathers did not exhibit a higher propensity for incest with their daughters than biological fathers (d = 0.03), while controlling for both the participant and their child’s age. Even if physical proximity is not a valid kinship cue that informs incest avoidance for fathers, we would still expect sociolegal fathers to have a higher propensity for incest, based on other kinship cues (e.g., physical resemblance). The reason for this finding is unclear, although a similar result was found by Kresanov et al. (2018). It is possible that the effects of socially desirable responding are more prominent for sociolegal than for biological fathers for this topic—possibly because of increased suspicion of sociolegal fathers with regard to child welfare (e.g., societal awareness of the “Cinderella effect”; Daly & Wilson, 1998). Despite attempts to reduce these effects by utilizing third-party descriptions and indirect questions, this could still be influencing the relationships examined.

In support of our hypothesis and consistent with results from a previous meta-analysis (Pullman et al., 2017), among participants with scores within the top 10th percentile on incest propensity, sociolegal fathers reported more childhood and adult antisociality than biological fathers. These results again suggest that contemporary theories that highlight antisociality as a key facilitating factor for child sexual abuse may be more applicable to sociolegal than to biological fathers. Furthermore, these results also highlight the need to control for the influence of antisociality when examining the relationship between kinship cues and incest avoidance among fathers, as antisociality is likely to explain more of the variance in incest propensity for sociolegal than for biological fathers.

Limitations

All of the variables examined in this study were reliant on self-report. Although we used vignettes and indirect questions to attempt to reduce ceiling and floor effects that can be caused by reliance on self-report with such sensitive and taboo topics, it is possible that socially desirable responding influenced our results. Indeed, because Qualtrics logs IP addresses, these panel participants know that their responses, while confidential, are not entirely anonymous. Using methods not reliant on self-report would add incremental validity to any study examining incest avoidance mechanisms. For example, De Smet, Van Speybroeck, and Verplaetse (2014) used facial electromyography to measure disgust toward descriptions of sibling incest. Studies of this nature, while an improvement, can also suffer from limitations related to the laboratory environment (direct contact between researcher and participant). This method could be improved further if it was possible to implement an implicit disgust and/or arousal toward incest analog that can be administered in an anonymous online survey.

Another limitation associated with this study is the vignettes used to measure disgust and arousal toward incest specified a sexual encounter with a “young” biological or step-daughter. This wording was chosen to facilitate participants “being in the shoes” of the actor in the vignette, to provide an estimate of the degree of disgust and arousal they feel toward the idea of sex with their own child (i.e., egocentric empathy; Fessler & Navarrete, 2004). However, this wording means that it is not possible to disentangle the amount of disgust/arousal experienced from the idea of sex with a biological or sociolegal daughter from the amount experienced toward the idea of sex with a young girl in general (e.g., contrasting sex with a 15-year-old vs. 5-year-old daughter). Future research using vignettes to measure incest propensity should employ a factorial design (i.e., vignettes involving sexual activity with a related child, sexual activity with a related adult, sexual activity with an unrelated child, and sexual activity with an unrelated adult) to disentangle these findings.

Finally, as with all cross-sectional studies, the direction of the relationships reported here is not known. While some evidence of covariation (e.g., between disgust toward incest and incest propensity) was found, conclusions about cause and effect cannot be made.

Conclusions

This study builds upon the scarce literature about incest avoidance in fathers, suggesting that physical proximity is not a viable kinship cue used by fathers to regulate incest avoidance toward their daughters. Nonetheless, the results do indicate that disgust toward incest may be a proximate mechanism that facilitates incest avoidance among fathers, using kinship cues other than physical proximity.