In the Beast Mode: Predator Self-Identifications as a Model of Disagreeable Functioning

Abstract

Some key sources of motivation are likely to be primitive. As a way of assessing such sources of motivation and their links to personality, participants in three studies (total N = 808) were asked to choose animals that they would prefer to be, with all pairs contrasting predator and prey animals. Individuals who select predator animals more often may wish to interact with the environment in a self-serving and callous manner. In support of such thinking, Study 1 linked higher levels of predator self-identification to lower levels of agreeableness and interpersonal warmth. Study 2 extended this model by showing that wishing to be predator animals was linked to self-serving behavior in economic games. Study 3 found inverse relationships between predator preferences and daily agreeableness levels in both between-person and within-person analyses. The findings, in total, highlight a motivation-based orientation to the environment that is disagreeable and self-serving.

When one surveys the history of personality psychology, it is clear that there are at least two grand traditions – the trait tradition and the motivation tradition (Winter et al., 1998). Personality traits are descriptive of ways in which people differ from each other (Fleeson & Jayawickreme, 2015), whereas motivation-based approaches seek to understand why it is that people act as they do (McClelland, 1987). Combining trait and motivational perspectives could be expected to result in a more comprehensive science of personality than either approach considered individually (McClelland, 1951; Winter et al., 1998). There are explanatory limitations to the trait approach, for example, that can be rectified by considering what it is that people are trying to do in their lives (Pervin, 1994).

As described by McClelland et al. (1989) and Winter et al. (1998), among others, many motivation scholars have used projective tests to assess key social motives. These tests, which include the Thematic Apperception Test, versions of the Rorschach test, sentence completion tests, and so forth, aim to assess key social motives implicitly, with the idea that some primitive sources of motivation, which may or may not be linked to explicit views of the self, may be best assessed using materials and tasks that pull for fantasy to some extent (McClelland, 1987; Tuber, 2012). The idea that much of what people do follows from implicit sources of motivation is a powerful one (McClelland et al., 1989), but this tradition of personality assessment is no longer prominent, in part due to the ascendency of the trait perspective (McCrae & Costa, 1999) and in part due to anti-scientific attitudes among some advocates of projective testing (Bornstein, 2001).

As explained by Bornstein (2007), projective tests have primarily been used in clinical settings, in which an analyst seeks to make inferences about a client who might benefit from psychotherapy (also see Tuber, 2012). Projective tests may be presented in an open-ended manner, a conversation between analyst and client is likely to take place, and interpretation of test responses is often reliant on clinical judgment rather objective scoring rules (Bornstein, 2007). Much may be inferred from just a few responses and questions related to reliability and validity are often treated intuitively rather than on the basis of statistical evidence (Bornstein, 2007). Such procedures may be useful in case conceptualization, but they cannot contribute to a nomothetic personality science, which requires standardized testing procedures, objective scoring rules, comparisons across participants, and statistical evidence concerning reliability and validity (Bornstein, 2001, 2007). In response to concerns of this type (Lilienfeld et al., 2000), some traditions of projective testing, such as the Rorschach tradition, have attended closely to questions of reliability and validity (Bornstein, 2012; Meyer & Eblin, 2012; Mihura et al., 2013; Viglione et al., 2012), and a nomothetic approach to personality assessment can be embraced with respect to other projective tests as well (Bornstein, 2007).

In this connection, what we seek to do is to show how one old projective test, which has primarily been used in applied clinical settings (Tuber, 2012), can be modified such that it is suited to nomothetic personality science. In a larger sense, these efforts are also meant to contribute to a resurgence of scientific interest in an important testing tradition in personality psychology (McClelland, 1987). The old projective test in question is the Animal Preference Test (Van Krevelen, 1956). Children, typically, have been asked to imagine that they could be an animal rather than a human being. They are then asked to generate an animal and explain why they generated the animal (or animals) that they did. The test trades on fantasy rather than reality (e.g., one will never become a non-human animal), which could allow individuals to express central, but rarely articulated, needs, such as those related to power or security (Van Krevelen, 1956) and/or autonomy, aggression, or nurturance (Rojas & Tuber, 1991; also see Sevillano & Fiske, 2016).

The Animal Preference Test, as it is typically administered and scored, does not meet scientific assessment standards (Hunsley & Mash, 2007). The use of non-standardized instructions, open-ended response procedures, and clinical interpretation (Van Krevelen, 1956) will almost certainly give rise to a noisy assessment instrument with low levels of reliability and uncertain validity (Lilienfeld et al., 2000). In rectifying such shortcomings, Robinson et al. (2017) created a revised version of the test named the Revised Animal Preference Test (RAPT). Instructions were standardized, clinical inference was eliminated, and participants were always asked which of two animals (e.g., hawk or dove) they would prefer to be. This test presents twelve animal pairs and, in all cases, one of the animals would be motivated to kill and eat the other. People who wish to be predator animals more often are thought to possess motivations of a self-serving and hostile type, potentially aligning the test with motives related to psychopathy.

Psychopathy is multi-faceted and is marked by an exploitative, self-centered, socially toxic, and fearless mode of interacting with the physical and social environment (Furnham et al., 2013; Patrick, 2022). Robinson et al. (2024a) pursued a fearlessness perspective concerning this interface, with the idea that higher RAPT scores would be linked to lower levels of negative emotionality and behavioral decisions that courted rather than avoided risk. In Study 1 of the paper, and across four independent samples, it was found that higher RAPT scores were linked to lower levels of neuroticism (average β = −.27), but the RAPT was not a predictor of traits or tendencies related to extraversion (average β = .05). In Study 2 of Robinson et al. (2024a), participants receiving higher RAPT scores were more prone to risky and potentially harmful behaviors. On the basis of results from all three studies, Robinson et al. (2024a) suggested that the motivations captured by the RAPT are linked to lower levels of avoidance motivation, negative affect, and fear, all of which have been emphasized in some theories of primary psychopathy (Crego & Widiger, 2022).

Developmental analyses of psychopathy suggest two primary deficits, one related to reduced threat sensitivity and the other related to reduced affiliative reward (Waller & Wagner, 2019). The latter deficit has recently received attention, for example in the neurobiological analysis of Viding and McCrory (2019) and in results reported by Waller et al. (2021). In a particularly relevant and insightful paper of this type, Foulkes et al. (2014) reviewed findings indicating that people with psychopathic traits have fewer, less intimate, and short-lived personal relationships, in part because they value their own goals over the goals and wishes of others. Similarly, Sherman and Lynam (2017) suggested that psychopathy is likely to be linked to low levels of communal motivation, which would enable self-serving and exploitative behavior.

If the RAPT captures motivational trends related to psychopathy, participants receiving higher RAPT scores are likely to exhibit traits, behaviors, and states that could be characterized as disagreeable (i.e., hostile or cold rather than affiliative and warm: Locke, 2015). In pursuing links of this type, Study 1 pursued the hypothesis that there would be an inverse relationship between RAPT scores and both agreeableness and interpersonal warmth, as they are typically assessed (i.e., in trait-related terms). Given that these relationships were foundational to the studies that followed, we collected data from three independent samples that were analyzed separately, therefore allowing for direct replication (Simons, 2014). In Study 2, we took advantage of the fact that prosocial traits and tendencies, such as agreeableness, can be assessed in economic games (Zhao & Smillie, 2015). On the basis of this literature (Thielmann et al., 2021; Zhao & Smillie, 2015), we hypothesized that higher RAPT scores would be linked to selfish rather than prosocial economic behavior (e.g., in versions of the dictator game: Engel, 2011).

In Study 3, we embedded predator-prey preference items into a daily diary protocol. Doing so will allow us to examine whether higher levels of predator self-identification are linked to lower levels of daily agreeableness in both between-person and within-person multilevel models (Nezlek, 2007). With respect to the latter focus, there is increased recognition that personality tendencies fluctuate within the person (Church et al., 2013) and the motivations captured by the RAPT are also likely to fluctuate across time (McClelland, 1987). On the basis of such precedents, we hypothesized that higher levels of predator self-identification, at a given time, would be linked with lower levels of state agreeableness. In sum, the programmatic studies were designed to provide novel evidence linking the individual differences captured by the RAPT to personality and behavioral tendencies falling within the agreeableness construct realm.

In addition to results involving traits, economic decisions, and daily states, we hypothesized that men would typically score higher on the RAPT than women would. This prediction follows from numerous sources of evidence indicating that men, relative to women, are (on average) more inclined to adopt interpersonal strategies characterized by dominance, competition, and hostility (e.g., Archer, 2009; Gurtman & Lee, 2009; Hsu et al., 2021), which are trends that the RAPT seeks to assess (Robinson et al., 2017). Viewed from this perspective, sex differences would provide a further source of validity evidence for the RAPT. Although controlling for sex would be problematic in distorting the relationships of interest (Hoyle et al, 2023), it is our belief that the RAPT captures similar trends among men and women (Robinson et al., 2024a). To the extent that this is true, sex should not moderate RAPT/outcome relationships, as examined using a multiple regression platform (Aiken & West, 1991).

Study 1

Method

Power Considerations, Participants, and General Procedures

We sought .80 power to detect correlations in the .30 range and G*Power recommended samples sizes of 84. Three samples of participants, who sought course credit for general education classes, signed up for a generically described Personality and Social Cognition study and arrived to the laboratory in groups of 6 or fewer. After signing consent forms, participants were placed in private computer rooms with personal computers. Samples consisted of 107 (48.60% female; 84.11% White; M age = 19.12), 97 (46.39% female; 84.54% White; M age = 19.60), and 128 (37.50% female; 89.84% White; M age = 19.80) participants, who completed the measures of interest, as well as other measures pertinent to other projects, using a MediaLab platform. Data for the project are available at OSF: https://osf.io/7g59u/?view_only=0a18c16008b141fa9198a16fe6b43ae5 (Authors, 2025). Preregistration did not occur with respect to studies or analyses, but replication was sought.

In Studies 1 and 2, the experimenter monitored performance and flagged any participant who completed the entire session, which was meant to take 50 minutes, in less than 20 minutes. Flagged participants were deleted, with additional deletions occurring in the case of insufficient response variation. In Study 3, participants were deleted if they did not provide enough daily reports or if they did not vary their ratings. Such data quality deletions occurred prior to analysis and it is estimated that 5–10 participants per sample were dropped on the basis of standard cleaning procedures for the lab.

The Revised Animal Preference Test

Participants completed the Revised Animal Preference Test (RAPT). This test, like older versions of the animal preference test (e.g., Van Krevelen, 1956), asks participants to indicate animals that they would like to be, if they were an animal rather than a human being. Stimuli for the RAPT, however, consist of pairs of animals (e.g., dove-hawk, lion-zebra) and participants choose their preferred animal. Each participant received a total RAPT score that was equivalent to the percentage of 12 trials on which a predator was chosen.

The average participant, in Study 1, displayed some attraction to predator animals, with means of .74 (95% CI = .69 to .79), .69 (CI = .64 to .74), and .77 (CI = .73 to .81) in Studies 1a through 1c, respectively. Of more importance, standard deviations were substantial (SDs = .25, .27, .23), indicating the presence of pronounced individual differences that were reliable across choices, with Cronbach alphas of .83, .83, and .74 in Studies 1a through 1c, respectively.

Personality Variations in Agreeableness

Personality-related variations in agreeableness were assessed using a 10-item IPIP scale that maximizes correlations with factor markers for this trait (Goldberg, 1999). This scale, which has been used in many studies (e.g., Robinson et al., 2024b), also correlates highly with alternative assessments of agreeableness based on the NEO and BFI (John & Srivastava, 1999) and it emphasizes the close theoretical link between agreeableness and interpersonal warmth (Fetterman et al., 2017). Participants indicated whether 10 statements (e.g., “I have a soft heart”, “I feel little concern for others”, which is reverse-scored) describe them (1 = very inaccurate; 5 = very accurate) and agreeableness levels were computed by averaging across items (Study 1a: M = 3.82; SD = .77; α = .84; Study 1b: M = 4.18; SD = .55; α = .83; Study 1c: M = 4.06; SD = .61; α = .85).

Interpersonal Warmth

Variations in interpersonal warmth were assessed by administering markers from the Interpersonal Adjective Scale-Revised (IAS-R: Wiggins et al., 1988). This is an extensively validated scale (Gurtman, 2009; Jones & Paulhus, 2001) that serves as the basis for validating other circumplex scales (e.g., Markey & Markey, 2009) and organizes over 100 measures with interpersonal content (Wiggins & Broughton, 1991). The IAS-R uses adjectives, which are favored in some assessment traditions (Goldberg, 1992), and participants were asked whether 16 adjectives targeting the warmth dimension (e.g., “charitable”, “sympathetic”, “cruel”, which is reverse-scored) describe them in general (1 = extremely inaccurate; 6 = extremely accurate). After reverse-scoring cold-hostile items, we averaged across markers (Study 1a: M = 4.92; SD = .79; α = .93; Study 1b: M = 4.96; SD = .76; α = .92; Study 1c: M = 4.95; SD = .69; α = .93).

Results

The predator self-identification continuum was z-scored (which will not affect ts, ps, or βs, but which will result in more interpretable bs) and entered as a predictor of agreeableness and interpersonal warmth in a series of simple regressions. The results of these simple regressions, which are displayed in Table 1, reveal that higher levels of predator self-identification were linked to lower levels of agreeableness as well as to lower levels of interpersonal warmth. Because these relationships were evident in all sub-studies, we conclude that the motivations assessed by the RAPT are inimical to agreeableness and warmth.

Table 1.

Simple Regressions Linking Higher Levels of Predator Self-Identification to Lower Levels of Agreeableness and Interpersonal Warmth, Study 1

Study and Outcome	b [95% CI]	t	p	β
Study 1a
Agreeableness	−.187 [−.331, −.044]	−2.59	.011	−.24
Interpersonal Warmth	−.264 [−.408, −.120]	−3.63	<.001	−.33
Study 1b
Agreeableness	−.131 [−.240, −.022]	−2.38	.019	−.24
Interpersonal Warmth	−.303 [−.444, −.162]	−4.27	<.001	−.40
Study 1c
Agreeableness	−.160 [−.265, −.056]	−3.05	.003	−.26
Interpersonal Warmth	−.232 [−.348, −.117]	−3.99	<.001	−.34

The predatory strivings that characterize the RAPT should, it was hypothesized, vary by sex. In support of this hypothesis, ANOVAs indicated the presence of significant sex differences in average RAPT scores in Studies 1a (male M = .86; female M = .62), F(1, 105) = 31.40, p < .001, η_p² = .23, 1b (male M = .84; female M = .51), F(1, 95) = 59.57, p < .001, η_p² = .39, and 1c (male M = .88; female M = .59), F(1, 126) = 70.45, p < .001, η_p² = .36. Nonetheless, in a series of multiple regressions, which centered sex (male = −1; female = +1) in accordance with recommendations in the literature (e.g., Aiken & West, 1991), sex by RAPT interactions were absent with respect to variations in agreeableness and interpersonal warmth, all |ts| < 1.50, all ps > .150. Hence, the RAPT assesses similar tendencies among males and females.

Discussion

The RAPT is a projective test, but it is a different sort of projective test than other projective tests (Carter et al., 2013). As a result of the focused nature of the test, its structure, and objective scoring rules, the individual differences captured by the RAPT were reliable and valid, such that higher levels of predator self-identification were linked to lower levels of agreeableness and warmth. The desires captured by the RAPT can therefore link up with personality traits (seemingly unlike TAT-based motives: Winter et al., 1998), though both agreeableness and warmth are defined in largely motivational terms (Graziano & Tobin, 2002). In Study 2, we sought to capitalize on this motivational basis in examining relations between RAPT scores and prosocial behavior.

Study 2

Prosocial motivations of an agreeableness type are thought to give rise to prosocial behavior, which can be assessed, experimentally and objectively, through the use of economic games (Zhao & Smillie, 2015). Decomposed games, in which interdependence between parties is absent, are particularly useful (Murphy & Ackermann, 2014), and we concentrate on such games in the present study. In the dictator game, for example, participants are free to keep all of the money that has been endowed to them or they could donate some of this money to another person (Engel, 2011). We have suggested that individuals receiving higher RAPT scores possess lower levels of prosocial motivation, which would be evident in terms of lesser donations to the hypothetical other (Balliet et al., 2009). Across two sub-studies, prosocial behavior was assessed in a dictator game, a public goods game, and in terms of point allocations made on trials of a social value orientation measure.

Method

Power Considerations, Participants, and General Procedures

Based on the meta-analysis of Thielmann et al. (2020), we sought .80 power to detect zero-order correlations in the .25 range. G*Power (correlation: bivariate normal model) recommended sample sizes of 123 and we sought to exceed this figure to some extent. Two samples of participants, who sought course credit for general education classes, signed up for a generically described Personality and Social Cognition study and arrived to the laboratory in groups of 6 or fewer. After providing consent, participants were placed in private rooms with personal computers. Sample sizes were 138 (49.28% female; 91.30% White; M age = 19.36) in Study 2a and 216 (60.19% female; 86.57% White; M age = 18.87) in Study 2b and all measures were collected using a MediaLab interface.

The Revised Animal Preference Test

Predatory strivings were assessed with the 12-pair version of the RAPT, as in Study 1. The average participant displayed some differential desire to be predator animals, with means of .74 (95% CI = .70 to .78) and .67 (95% CI = .63 to .70) in Studies 2a and 2b. Standard deviations were large, however (SDs = .25 and .28), and individual differences were reliable across animal pairs (αs = .83 and .85).

Study 2a Outcomes

We sought to assess prosocial motivations and behaviors through the use of economic games (Zhao & Smillie, 2015). In Study 2a, we presented a version of the Dictator Game, which assesses cooperative or altruistic tendencies in the absence of opportunities for the other to retaliate (Hilbig et al., 2013). Participants were told that they would decide how to split a given amount of money between themselves and “another person” whose identity was not specified. The six trials asked participant how much of $10, $20, $50, $100, $500, and $1000 they would give to the other person, with the remainder to be kept for the self. To weight the trials equally, we z-scored each trial and then averaged across trials (M = .00; SD = .84; α = .94).

In Study 2a, we also presented a “public goods” game (Thielmann et al., 2020). Participants were asked to imagine that they listen to a particular public radio station every day. They were also told that the station can only operate through the donations of its listeners. Although donating money is entirely voluntary, a lack of donations could lead to the cessation of services. Participants were then asked how much money they would donate to the station per year, with 6 options ranging from $0 (coded as 1) to $100 (coded as 6). The average participant was willing to donate about $15 per year, but individual differences, which were of key interest, were also apparent (M = 3.52; SD = 1.46).

Study 2b Outcomes

For the sake of replication, a different version of the Dictator Game (Engel, 2011) was used in Study 2b. The instructions for the game were identical to those of the dictator game in Study 2a, but there were 5 rather than 6 trials and all trials involved the fixed amount of $10. Because the endowment was always the same, we simply averaged across trials, with higher numbers reflecting higher levels of prosocial behavior (M = $4.16; SD = $2.06; α = .95).

Prosocial behavior can also be assessed as a component of Social Value Orientation (SVO), which is a decomposed game designed to assess social motives in the absence of interdependence (Balliet et al., 2009). In Study 2b, we administered the 9-item Triple Dominance Scale version of this test, which is the most commonly used measure of this type (Murphy & Ackermann, 2014). Participants were asked to imagine that they were randomly paired with another person, referred to as “Other”. On each of 9 trials, they could choose an option (A, B, or C) that results in points being allocated to self and other, with more points being more desirable. Each trial presented one option that was prosocial because it maximized points for the other (e.g., 500 points for the self and 500 points for the other). Another option reflects a proself orientation as it maximizes points for the self (e.g., 560 points for the self and 300 points for the other). The third option, which is not commonly endorsed (Au & Kwong, 2004), is competitive because it maximizes the discrepancy between self and other (e.g., 490 points for the self and 90 points for the other).

We scored the SVO measure by computing the percentage of trials, for a given participant, that reflected prosocial (M = 59.83%; SD = 41.50%; α = .96), proself (M = 23.25%; SD = 32.12%; α = .94), and competitive (M = 16.92%; SD = 33.29%; α = .97) orientations. Although the last score was reliable, many participants never chose the competitive option (Au & Kwong, 2004). For this reason, our hypotheses primarily focus on prosocial and proself percentages.

Results

We hypothesized that higher levels of predator preference would be linked to lesser prosocial behavior. This hypothesis was examined by z-scoring the predator self-identification measure, for a given study, and entering it as a predictor of each of the outcomes for that study, in simple regressions. As displayed in Table 2, the results of these analyses support the point that predatory strivings are inversely linked to prosocial behavior. This was true concerning the dictator game in both studies, the public goods game in Study 2a, and the SVO measure that was administered in Study 2b. In the latter case, participants who preferred to be predator animals made point allocations that favored the self and were less prosocial.

Table 2.

Simple Regressions Linking Higher Levels of Predator Self-Identification to Lower Levels of Prosocial Behavior, Study 2

Study and Outcome	b [95% CI]	t	p	β
Study 2a
Dictator Giving	−.155 [−.295, −.014]	−2.18	.031	−.18
Public Good Donation	−.367 [−.606, −.129]	−3.04	.003	−.25
Study 2b
Dictator Giving	−.345 [−.618, −.072]	−2.49	.014	−.17
SVO Prosocial %	−.076 [−.131, −.021]	−2.73	.007	−.18
SVO Proself %	+.048 [+.005, +.091]	+2.22	.028	+.15
SVO Competitive %	+.028 [−.017, +.073]	+1.24	.215	+.08

Note: SVO = Social Value Orientation

As in Study 1, there were sex differences in average RAPT scores in Studies 2a (male M = .86; female M = .62), F(1, 136) = 39.68, p < .001, η_p² = .36, and 2b (male M = .83; female M = .56), F(1, 214) = 66.04, p < .001, η_p² = .24. In multiple regressions, however, sex by RAPT interactions were absent in predicting the Study 2 outcomes, all |ts| < 1.50, all ps > .150. The RAPT, therefore, captures similar motivational trends among males and females.

Discussion

People who wish to be predator animals may wish to be able to do what they want, in a self-favoring way, much as the classic psychopath does (Crego & Widiger, 2022). Evidence in favor of links of this type was found in Study 2, in which it was shown that participants who wished to be predator animals to a greater extent displayed less altruism in a dictator game, donated less to a public good, and made point allocations consistent with a “pro-self” orientation when distributing resources between self and other. Correlations among these decomposed game measures tend to be modest (Balliet et al., 2009), but the individual differences captured by the RAPT predicted behavior in all cases. We conclude that individuals who wish to be predator animals also favor the self when resources can be hoarded rather than shared, at least when others are not particularly familiar (Arbuckle & Cunningham, 2012).

Study 3

We conceptualize predator self-identifications in terms of an orientation to the social environment that is both self-favoring and exploitative. This orientation, like other motivational orientations (McClelland, 1987), is likely to vary by time and context. To examine questions concerning the malleability of the orientation in question, predator preference items were inserted into a daily diary protocol. Doing so has not been done before and it will allow us to examine relationships between predator preferences and (daily) agreeableness levels in both between-person and within-person terms (Nezlek, 2007). In both cases, we hypothesized that higher levels of predator preference would be linked to lower levels of agreeableness.

Method

Power Considerations, Participants, and General Procedures

Daily diary designs are powerful, though exact power calculations are error-prone (Nezlek, 2012). In planning sample size, we therefore followed rules of thumb based on simulation data. Maas and Hox (2005) found that estimates of regression coefficients, variance components, and standard errors are reliable and unbiased when there are 50 level 2 units (in our case, participants) and we sought to exceed this figure. Scherbaum and Ferreter (2009) recommend obtaining 900 total observations and we sought to exceed this figure as well.

A “Daily Diary Study” was posted to SONA and undergraduate students seeking credit for their psychology classes signed up for it. Electronic consent was obtained and demographic information was provided, following which the diary portion of the study began. We asked participants to try to meet the goal of providing paired (morning and evening) reports for a total of 8 days and we gave them up to 19 days to meet this goal, following which data collection ceased. For each of these 19 days, emails with participant number information and a link to a Qualtrics survey were sent at 12 p.m. and 6 p.m. In each case, participants needed to complete the survey within 2 hours or it was considered missing. Email lists were continuously revised by removing participants who met the goal of 8 days of paired reports, though we ultimately needed to relax criteria to bolster sample size, with all such decisions being made prior to analyses.

Such considerations led us to create two datasets, the first of which we will refer to as “unpaired” and the second of which we will refer to as “paired”. In both cases, we counted the number of completed reports per participant (or, in the case of the paired dataset, the number of morning-evening pairs for the same day) and then sorted by these compliance numbers, prior to analyses. The goal was to retain 110–120 participants per dataset, but to delete participants whose compliance levels were very poor (e.g., only 2 reports completed). In balancing these two criteria, we ultimately retained unpaired data for 122 participants (66.39% female; 93.44% White; M age = 18.66) who completed at least 8 total reports. Within this dataset, the average participant completed 16.77 reports, resulting in 2046 rows. A secondary goal was to examine temporal trends within days, which resulted in a secondary paired dataset. Within this dataset, we retained 111 participants who had at least 5 sets of paired reports – i.e., a morning report and an evening report for the same day. This dataset consisted of 838 rows/days.

Survey and Variables

The study was conceptualized as a “daily metaphor study” and the survey therefore included a number of metaphor-inspired items (e.g., whether open or closed is preferred on a particular day: Robinson et al., 2024b). Pertinent to the present study, we included two animal preference items and an agreeableness item. The animal preference pairs were hawk-dove and fox-rabbit. With respect to each pair, participants were asked to indicate their current degree of preference to be one animal (e.g., a fox) versus the other (e.g., a rabbit). In both cases, they needed to move a slider from center position to the point on the slider that indicated their current preference. Each slider recorded 201 positions, from a decided preference to be one animal (e.g., −100 = definitely a dove) to a decided preference to be the other (e.g., +100 = definitely a hawk), with position feedback (e.g., +36) displayed to guide responding. After moving the slider to the desired position, participants made a mouse click and the preference response was recorded.

The fox-rabbit slider response was multiplied by −1 such that higher numbers indicated a preference to be the predator animal (fox). At the level of report, wishing to be a fox was correlated with wishing to be a hawk (α = .72) and we therefore averaged across hawk-dove and fox-rabbit responses. We further examined the reliability of the Study 3 measure by revisiting Study 2b data, which had a large sample size. A score based on hawk-dove and fox-rabbit preferences correlated, in Study 2b, with the total RAPT score at .78, suggesting that a two-item measure composed of hawk-dove and fox-rabbit pairs can result in a reliable quantification of predator preference. Returning to the current study, the average predator preference score was −2.31, but the standard deviation across reports was 53.90, indicating a considerable degree of variability.

Report-specific tendencies toward agreeableness were assessed in a manner consistent with previous research (Fetterman et al., 2017; Fetterman et al., 2018; Fleeson, 2001). Specifically, participants indicated whether (1 = strongly disagree; 5 = strongly agree) they were “agreeable” during a particular morning or afternoon. Agreeableness levels tended to be moderate (M = 3.60) and varied across reports (SD = .94).

Results

Intraclass Correlation Analyses

We computed intraclass correlation coefficients (ICCs) for each of the measures, with unpaired reports nested within participants. The ICC estimate for the predator preference measure was .45, meaning that approximately half of the variance for this measure varied between persons and approximately half of the variance for this measure varied within persons (West et al., 2011). The ICC for agreeableness was .28, indicating a greater degree of within-person malleability, relative to predator preferences.

Did Fluctuations in Predator Preference Predict Fluctuations in Agreeableness?

Using the “unpaired” dataset, we examined whether within-person variations in predator preference predicted within-person variations in agreeableness. For this multilevel analysis (MLM), we person z-scored predator preferences, which will remove between-person variance (Enders & Tofighi, 2007). We then performed a “level 1” (within-person) MLM using the PROC MIXED procedures of SAS, with the intercept allowed to vary at random (Singer, 1998). Variations in agreeableness retained their original units, except that we ran a second model with a z-scored outcome, resulting in a standardized b that can be considered a measure of effect size (Lorah, 2018). As hypothesized, results indicated that fluctuations in predator preference were linked to fluctuations in agreeableness, b = −.081 [−.128, −.033], t = −3.35, p < .001, standardized b = −.086.

Did Average Levels of Predator Preference Predict Average Levels of Agreeableness?

We created a “level 2” (between-person) predator preference score by averaging across all predator preference responses for a given participant (M = −1.84; SD = 36.62). This average score was then z-scored and added to the unpaired dataset. We then performed a level 2 MLM (Singer, 1998). At the between-person level, higher average levels of predator preference were linked to lower average levels of agreeableness, b = −.108 [−.200, −.015], t = −2.30, p = .023, standardized b = −.115, replicating Study 1.

Analyses Involving Sex

Average levels of predator preference (as revealed by an ANOVA) were higher among males (M = 21.21) than females (M = −14.39), F(1, 120) = 30.13, p < .001, η_p² = .20, replicating prior studies. In MLMs, however, sex (−1 = male; +1 = female) did not moderate within-subject, b = −.013 [−.064, .037], t = −0.52, p = .604, standardized b = −.014, or between-subject, b = .003 [−.102, .108], t = 0.06, p = .952, standardized b = .003, relationships between predator preference and agreeableness.

Analyses Involving the Paired Dataset

We used the paired dataset to examine within-day relationships involving predator preference and agreeableness. After person-centering the predator preference measure (Enders & Tofighi, 2007), predator preferences earlier in the day did not predict predator preferences later in the day, b = 2.084 [−1.510, 5.678], t = 1.14, p = .255, standardized b = .039, highlighting just how state-dependent predator preferences can be. Predator preferences earlier in the day, however, predicted levels of agreeableness later in the day, b = −.097 [−.158, −.035], t = −3.09, p = .002, standardized b = −.105. By contrast, agreeableness levels earlier in the day did not predict predator preferences later in the day, b = −1.635 [−4.700, 1.430], t = −1.05, p = .295, standardized b = −.031. These results provide support for the contention that predator preferences assess components of motivation that affect interpersonal functioning.

As a more conservative test of directionality, we reran the directional models after controlling for morning levels of the outcome. In the first such model, early-day predator preferences predicted changes in agreeableness (from earlier in the day to later in the day) in a disagreeable direction, b = −.077 [−.140, −.014], t = −2.39, p = .017, standardized b = −.083. By contrast, early-day agreeableness levels did not predict changes in predator preference (from earlier in the day to later in the day), b = −1.484 [−4.650, 1.682], t = −0.92, p = .358, standardized b = −.028. Within-person relationships between predator preference and agreeableness therefore appear to favor predator preference as a predictor of agreeableness rather than vice versa.

Discussion

By embedding predator preference items into a daily diary protocol, and by obtaining both morning and afternoon reports, we were able to provide additional insight into the relationship between predator preferences and agreeableness. In a replication of Study 1, higher average levels of predator preference were linked to lower average levels of (daily) agreeableness. As a complement to this “trait” perspective on the relationship, we found that shifts in predator preference, within-person, were associated with shifts in daily agreeableness levels. These results suggest that the predator preference orientation is, to some extent, state-related in nature. Further, this orientation appeared to predict subsequent levels of agreeableness, but not vice versa, providing some support for the idea that an attraction to predator animals shifts social functioning in a disagreeable direction.

General Discussion

Personality science will be better off if it can supplement a trait-related understanding of the person with a motivation-related understanding (Winter et al., 1998). If projective tests, which are classically used to assess key forms of social motivation (McClelland, 1987), continue to fade in their usage in personality and clinical science (Bornstein, 2001), something important will have been lost. In the present research, we have shown that revising a classic projective measure such that it is more structured and less reliant on clinical interpretation (Lilienfeld et al., 2000) can result in a test that captures important strivings that are likely to contribute to individual differences in personality and social behavior.

The Revised Animal Preference Test (RAPT) is psychometrically sound and it appears to capture motivational trends related to psychopathy, which is marked by a self-serving, callous, and exploitative interpersonal style (Foulkes et al., 2014; Patrick, 2022). Robinson et al. (2024a) found that individuals receiving higher RAPT scores, relative to lower scores, were comparatively fearless, which is a classic, though debated, component of psychopathy (Patrick, 2022). Fearlessness, though, need not be disagreeable (Crego & Widiger, 2022), and modern conceptions of psychopathy emphasize the centrality of disagreeable functioning, whether from a trait perspective (Miller & Lynam, 2015) or by emphasizing deficiencies in affiliative reward (Waller et al., 2021) or communal motivation (Sherman & Lynam, 2017). In aligning the RAPT with these new sources of theorizing (e.g., Viding & McCrory, 2019; Waller & Wagner, 2019), we conducted a series of studies designed to show that wishing to be predator animals means, in part, wishing to be disagreeable.

Before reviewing key findings, it is useful to revisit the topic of sex differences. Men, relative to women, are thought to possess interpersonal strivings that, on average, favor dominance, competitiveness, and hostility to a greater extent (Archer, 2009; Buss, 1990; Gurtman & Lee, 2009). Given that predator animals are perceived to be (and arguably are: Lima, 2002) both dominant and hostile (Sevillano & Fiske, 2016), one should expect men to be drawn to predator animals to a greater extent. This expectation was confirmed and it is reasonable to suggest that the relevant sex differences provide a source of validity evidence for the RAPT. Even so, sex differences are, by themselves, not particularly explanatory (Hyde, 2005), and the motivational trends captured by the RAPT are likely to mediate a number of sex differences in interpersonal motivation and behavior (Robinson et al., 2017). Of more pertinence to the present studies, it was shown that the RAPT functions similarly among men and women (i.e., there were no sex by RAPT interactions), attesting to validity of the test among both sexes.

Turning to key findings, the results of Study 1 were robust in suggesting that the motivations captured by the RAPT are inimical to high levels of agreeableness or interpersonal warmth. The results of Study 2 were important in showing that they are also linked to selfish and less prosocial behavior in economic games. The results of Study 3 provided many insights. The desire for the self to be a predator animal fluctuates across time and these fluctuations are systematically linked to within-person variations in daily agreeableness. Such results link predator preferences to an emerging literature on personality states (Church et al., 2013), with the idea being that one’s personality changes across time and context, with such changes being linked to motivations and goals that also fluctuate (McCabe & Fleeson, 2016). The motivations captured by RAPT-like items appear to be consequential in predicting these shifts, at least with respect to variations along the disagreeable-to-agreeable dimension.

Winter et al. (1998) suggested that traits and motives are essentially independent (i.e., uncorrelated). If this is true, there would be little commerce between two important research traditions and the motivation tradition, at least as reflected in projective tests, would become an insular one (Bornstein, 2001, 2007). At least with respect to the RAPT, such a lack of integration need not occur because it is apparent that the individual differences captured by the RAPT can be linked to personality traits, whether neuroticism (Robinson et al., 2024a) or agreeableness (present studies). That predator preferences are linked to lower levels of both neuroticism and agreeableness, however, suggests that the individual differences captured by the RAPT are configural in nature, just as variations in psychopathy appear to involve multiple traits and facets interacting in a combinatory manner (Miller & Lynam, 2015). Motivation-based perspectives, we think, could result in new ways of understanding the taxonomic basis of individual differences.

Questions, Limitations, and Future Directions

There is a puzzle to the RAPT that should be mentioned. On the RAPT, the average person displays a decided preference to be predatory over non-predatory animals, with means consistently departing from .5. Yet, traits related to primary psychopathy and callous-unemotional functioning tend not to be normative (Viding & McCrory, 2019). This puzzle can be resolved by pointing to the make-believe nature of the RAPT, which allows test-takers to imagine themselves as fierce and fearless beings without ascribing callous or hostile traits to themselves. One implication that might follow is that the average person possesses some primitive desires that are self-centered as well as disagreeable. These desires, though they may be restrained in accordance with social norms, can be identified using a fantasy-based test. This analysis suggests that the RAPT should have particular value in predicting behaviors that follow from primitive sources of motivation, perhaps due to their spontaneous nature (McClelland et al., 1989).

The present studies focused squarely on the RAPT because the goal was to better understand the nature of this test and its correlates. But we also recognize that there are other implicit and projective measures, including some that have been used to predict aggressive behavior. These include scoring systems for the Rorschach (Baity & Hilsenroth, 2002) and the Thematic Apperception Test (Kalliopuska, 1992; Matranga, 1976) as well as versions of the Implicit Association Test (Banse et al., 2015) and conditional reasoning tests (James & LeBreton, 2010). At the risk of oversimplifying, we believe that the RAPT captures something different than these other tests and scoring systems. The RAPT does not assess implicit associations (Banse et al., 2015) or motivated reasoning (James & LeBreton, 2010) and it is not focused on perceptions of ambiguous stimuli (Baity & Hilsenroth, 2002) or story-telling concerning ambiguous stimuli (Matranga, 1976). What the RAPT does assess, likely more so than these other measures, concerns desires for the self to be like a predator – that is, capable of imposing one’s will, often through dominant and hostile behaviors (Lima, 2002). Stated in other terms, the RAPT centers on preferences, the self, and object relations in ways that seem more direct than other tests.

Although we have relied on theorizing concerning psychopathy (e.g., Levenson, 1992; Patrick, 2022; Viding & McCrory, 2019) in making the present predictions, it should be recognized that psychopathy appears to be multi-faceted (Crego & Widiger, 2022; Patrick, 2022), with evidence supporting the idea of primary and secondary subtypes (Docherty et al., 2016; Drislane et al., 2014; Lee & Salekin, 2010). Primary psychopathy is primarily defined in terms of personality features such as boldness, callousness, and manipulativeness (Drislane et al., 2014; Frick & Ray, 2015), whereas secondary psychopathy seems to be marked by vulnerability and emotional impulsivity (Docherty et al., 2016; Skeem et al., 2007). The motivations captured by the RAPT appear to be more strongly linked to primary than to secondary psychopathy (Penzel et al., 2021), as captured by descriptions of the “classic” psychopath (Crego & Widiger, 2022; Hare, 1996; Levenson, 1992; Meloy, 2006).

We focused on links between predator self-identification and agreeableness among young adults, who are in the process of establishing careers and social networks (Arnett, 2000). The RAPT should assess similar motivations among children and adolescents given that the animal preference test is often used to identify problematic strivings among such age groups (Rojas & Tuber, 1991). The animal preference test has been similarly used among older clients (Tuber, 2012), as are related tests that ask family members to choose animals to represent each other (Arad, 2004). Motives related to agency and communion, which likely govern animal choices (Rojas & Tuber, 1991) continue to matter throughout the lifespan (Diehl et al., 2004), and the RAPT should therefore have value among older adults. It should be noted, though, that implicit motives tend to decline in strength with increasing age (Denzinger et al., 2016) and similar declines could occur with respect to average RAPT scores. Even so, the RAPT assesses similar individual differences among groups with lower average scores, as evident in our analysis of differences between and within sex. Nonetheless, we would like to see research using the RAPT with diverse samples (e.g., juvenile delinquents, mid-career managers).

In Study 2, higher scores on the RAPT were linked to lesser prosociality in dictator, public goods, and social value orientation games, but games such as the trust game or commons dilemmas could also be used (Thielmann et al., 2020). Also, games with real rather than hypothetical consequences (Thielmann et al., 2015) could be used, though performance in simplified games may be easier to interpret (Balliet et al., 2009). Asking people about their relationship goals and relationships would be interesting. If we are correct, individuals receiving higher RAPT scores may value relationships to a lesser extent and they may have fewer meaningful friendships (Foulkes et al., 2014). In other words, predatory strivings may give rise to social networks that are smaller and less intimate.

In the current studies, we primarily adopted a deficit-based frame, with higher levels of predator self-identification being linked to lower levels of agreeableness and warmth. In future research, it would be valuable to pursue a presence-based frame by focusing on tendencies and behaviors that should be stronger at higher levels of predator self-identification. Predators are aggressive (Lima, 2002) and self-identified predators are likely to be prone to aggression, particularly of a proactive (instrumental) type (Meloy, 2006). Higher levels of predator self-identification may also be linked to a broader class of antisocial behaviors, including deviance, that are likely to follow from ways of interacting with the world that are both hostile and exploitative (Burt & Donnellan, 2009). There would seem to be a close affinity between the motivations captured by the RAPT and descriptions of a “dark core” that is likely to underlie socially toxic behaviors (and what is common to the dark triad traits of narcissism, Machiavellianism, and psychopathy: Furnham et al., 2013), with this dark core consisting of motivations that are self-centered, exploitative, and self-justifying (Moshagen et al., 2018). Links of this type would be worth establishing as a basis for making further predictions concerning the individual differences captured by the RAPT.

Conclusions

In terms of primitive fantasy, many individuals harbor desires to be like predator animals, who are fierce, autonomous, and hostile (Rojas & Tuber, 1991). As hypothesized, such individuals are less agreeable and more self-centered in their decision-making and behavior. The present research therefore illustrates the value of modifying projective tests such that their capacity to speak to personality process is more evident.

Data availability:

Data are available: https://osf.io/7g59u/?view_only=0a18c16008b141fa9198a16fe6b43ae5