Archives of Sexual Behavior

, Volume 41, Issue 1, pp 63–71

Considering the Role of Conditioning in Sexual Orientation


    • Department of PsychologyKnox College
Original Paper

DOI: 10.1007/s10508-012-9915-9

Cite this article as:
Hoffmann, H. Arch Sex Behav (2012) 41: 63. doi:10.1007/s10508-012-9915-9


The effects of learning on sexual orientation are rarely addressed in the literature. At the very least, such processes provide a means of elaborating upon orientation predispositions. Some aspects of our mates may inherently elicit a sexual response, but other attributes do so only after sexual experience with them. Animal research shows robust, direct effects of conditioning processes on partner preference with a few studies showing plasticity in preference for sex of partner. Descriptive research in humans suggests effects of experience on partner preference and, although experimental demonstrations of human sexual conditioning are neither numerous nor robust, sexual arousal is conditionable in women and men. With modern developments in learning theory (e.g., expectancy learning and evaluative conditioning), it seems appropriate to renew the investigation of contributions and limitations of conditioning processes to explaining how cues acquire erotic value and to attempt some integration between the sexual conditioning literature and research on sexual orientation or more generally sexual partner preference.


Sexual orientationPartner preferenceExpectancy learningEvaluative conditioning


Learning processes allow for behavioral flexibility in response to unpredictability in the environment and such processes have been shown, mainly in animals but in some instances in humans, to affect many aspects of sexual behavior (Akins, 2004; Domjan & Akins, 2011; Pfaus, Kippin, & Centeno, 2001; Woodson, 2002), including partner preference (e.g., Coria-Avila et al., 2008a; Coria-Avila, Ouimet, Pacheco, Manzo, & Pfaus, 2005; Coria-Avila et al., 2008b; Coria-Avila, Triana Del Rio, Montero-Dominguez, Hernandez, & Manzo, 2009; Ismail, Gelez, Lachapelle, & Pfaus, 2009; Nash & Domjan, 1991; Pfaus, et al., 2001; Woodson, Balleine, & Gorski, 2002; Young & Wang, 2004). Yet, while a few older theories of the development of sexual orientation, such as Bem’s (1996) exotic becomes erotic theory (EBE) and Storms’ (1981) “timing of puberty” model as well as some recent observations regarding erotic plasticity (Baumeister, 2000; Diamond, 2000, 2006, 2009; Diamond & Fagundes, 2008), implicate experience as influential in the development of sexual orientation, little clear or convincing evidence supports such a role (Rahman, 2005). However, Freund and Blanchard (1993) stated that it is unlikely that we are born with preformed erotic images of women, men or children but, rather, we are born with differential sensitivities to particular classes of stimuli (e.g., hairiness, smoothness) and these sensitivities interact with experience to produce desired objects. At the very least, learning processes provide a means of elaborating upon orientation predispositions. Some aspects of our mates may inherently elicit a sexual response but other attributes may do so only after sexual experience with them. The present paper will provide a contemporary perspective on (classical) conditioning as well as review research showing the effects of experience, in particular conditioning, on partner preference in animals and humans.

Conditioning: Procedures and Mechanisms

The effect of early life experiences on sexual partner preference has sometimes been referred to as sexual imprinting. Whether imprinting is a distinct type of learning has yet to be established and such a process may share mechanistic commonalities with classical and operant conditioning (e.g., Hollis, ten Cate, & Bateson, 1991; Sullivan, 2001), the procedures most often used in sexual conditioning studies.

Classical or Pavlovian conditioning consists of pairing an initially innocuous cue (the conditioned stimulus; CS) and a biologically significant one (the unconditioned stimulus; US) and results in learning that these events are related which, in turn, can yield a conditioned emotional or motivational state and/or behavioral response (the conditioned response; CR). Operant conditioning involves learning that one’s own actions can change environmental events. More specifically, if a behavior is followed by rewarding or punishing consequences, it can result in changes in the frequency of that behavior. Classical conditioning appears most directly related to how cues or “objects” can acquire sexually arousing properties.

Several types of learning processes most likely simultaneously affect any given learning situation. For example, McGuire, Carlisle, and Young (1965) and Junginger (1997) described the development of sexual fetishes with a two-factor model. An accidental pairing of some object with arousal or orgasm gives the stimulus erotic value. This object can then elicit approach and/or other appetitive volitional behaviors involving the object that are positively reinforced by sexual reward. In addition, a few studies implicate observational learning in sexual conditioning (Galef, 2008; Köksal & Domjan, 1998; White, 2004), yet it is unclear if/how vicarious learning is mechanistically distinct from classical or operant processes. For example, Mineka and Cook (1993) have shown that observational fear learning in chimps is “reducible” to classical conditioning.

Although traditionally viewed as a lower level reflexive process with limited applicability to human behavior, contemporary learning theorists believe that classical conditioning most likely serves a number of important functions in complex organisms. It can prepare organisms for interaction with biologically significant cues or events by generating expectancies (signal or expectancy learning: Domjan, 2005; Hollis, 1997; Rescorla, 1988; Timberlake, 2001) but may also alter preferences for (De Houwer, Baeyens, & Field, 2005) and attitudes about (Baccus, Baldwin, & Packer, 2004; Dijksterhuis, 2004; Gawronski & Bodenhausen, 2006; Karpinski & Hilton, 2001; Livingston & Drwecki, 2007) stimuli associated with such cues or events (evaluative conditioning).

Evaluative conditioning (EC) is a more recently recognized form of classical conditioning that is believed to involve an associative transfer of affective value, or valence, as a result of exposure to CS-US pairings (De Houwer et al., 2001). In contrast to signal learning, EC has been almost exclusively researched in humans (largely because the subjective experience of liking or disliking is more difficult to measure in animals). The prototypical EC paradigm involves pairing of a neutral CS (e.g., a neutrally rated face) with pictures of liked or disliked stimuli, and results in a change in liking for the CS. Evaluative conditioning appears to be robust in some instances (yet fails to appear in others; see De Houwer et al., 2005; Hofman, De Houwer, Perugini, Baeyens, & Crombez, 2010; Rozin & Barnes, 1998) and has been demonstrated in a variety of paradigms employing diverse cues, including CSs from a range of sensory modalities and more biologically relevant USs as well as with direct and indirect measures (De Houwer et al., 2001; Hofman et al., 2010).

Signal and evaluative conditioning can be dissociated within the same paradigm (e.g., Hermans, Vansteenwege, Crombez, Baeyens, & Eelen, 2002), suggesting they are distinct, although not necessarily independent processes. Further, EC appears less sensitive to CS–US contingency and modulation (Baeyens, Crombez, De Houwer, & Eelen, 1996; Olson & Fazio, 2001; but see Hardwick & Lipp, 2000; Lipp & Purkis, 2005) and more resistant to extinction (Baeyens et al., 1988; Hermans, Crombez, Vansteenwegen, Baeyens, & Eelen, 2002; but see Lipp & Purkis, 2005, 2006). In addition, there is evidence that EC can be acquired without awareness of the CS–US contingency (e.g., De Houwer, Hendrickx, & Baeyens, 1997; Dickinson & Brown, 2007; Walther & Nagengast, 2006, but see Hofmann et al., 2010) and that its expression may be more automatic (Neumann, Forster, & Strack, 2003; Öhman & Mineka, 2001; Yin & Knowlton, 2006).

Yet, even if we accept that classical conditioning can alter responses to and/or preference for stimuli in humans, we are still unsure of how such conditioning effects occur. It is important to distinguish between the phenomenon of classical conditioning and the mechanism behind it. Moreover, whether there is a singular Pavlovian conditioning process that occurs in all organisms or whether there are different types of classical conditioning, perhaps some that are unique to humans, is still an open question.

The mechanism(s) underlying conditioning in humans has been conceptualized in a number of ways. Initial explanations derived from animal research proposed the formation of a direct stimulus–response (S–R) association. Although nonhuman evidence suggests the learning of a direct link (Carew, Hawkins, & Kandel, 1983; Colwill & Rescorla, 1985; Rozin & Zellner, 1985), the modern interpretation of Pavlovian conditioning (even within much of the animal literature) is that the CS comes to predict the US (Rescorla, 1988). Most agree that conditioning forms expectancies. The debate is on the level of cognition at which this occurs. Some have proposed that associative learning, and in particular classical conditioning, in humans depends on higher-level cognitive processes that give rise to propositional learning (Brewer, 1974; Mitchell, De Houwer, & Lovidbond, 2009). Others support a dual process model indicating that learning can result from controlled reasoning as well as more passive automatic processes (Baeyens et al., 2001; Kirsch, Lynn, Vigorito, & Miller, 2004; Öhman & Mineka 2001; Razran, 1971). However, this distinction is perhaps too simplistic.

A continuum or an intermediate stage between automatic and controlled processes (e.g., unconscious expectancies/goal pursuit, preconscious declarative knowledge) appears to exist (Custers & Aarts, 2010; Kihlstrom, 1987; Kirsch et al., 2004; Rose, Haider, & Buchel, 2005) and learning could potentially result from/include any number of these processes that, in turn, may be coordinated at some level. For example, Baeyens et al. (2001) proposed that signal learning is governed by an expectancy system that requires more cognitive resources to process or translate complex information resulting in anticipation of an object or event. On the other hand, they suggested that EC was mediated by a more “primitive” referential system that employs rudimentary learning or performance rules (e.g., Hebbian synapses) resulting in changes in affective value that can influence the direction of behavior (approach/avoid) and modulate (facilitate/suppress) responses generated by the expectancy system. A meta-analysis of EC studies found that contingency awareness moderated the effects of EC, yet, interestingly, learning was stronger in aware participants (Hofmann et al., 2010). Hence, there may be different types of EC mediated by different mechanisms, including conscious propositional knowledge (De Houwer, 2007) or even other (e.g., non-associative: Davey, 1994) processes.

In summary, classical conditioning has been shown to be more complex than originally expected. It is applicable to lower and upper level behaviors and it can contribute to the generation, either directly or indirectly, of a range of CRs, including physiological, subjective (cognitive and emotional), and behavioral ones. Perhaps both signal and evaluative conditioning are involved in the development of human partner preferences. Although certain characteristics of our mates may be inherently arousing, learning processes may enhance and expand on such predispositions, e.g., “neutral” stimuli associated with our mates may acquire erotic value and/or they may simply come to predict or signal sexual interaction.

Evidence for the Effects of Sexual Conditioning on Partner Preference


The empirical animal literature contains numerous convincing demonstrations that conditioning impacts many aspects of appetitive and consummatory sexual behavior in a wide range of species, including insects, fish, birds, rodents and primates. The effects observed include conditioned approach, conditioned courtship, conditioned copulatory behaviors, conditioned place preference (which can model learning the location of mates), as well as demonstrations that organisms will engage in various instrumental responses to obtain sexual reward (Domjan & Akins, 2011; Domjan & Holloway, 1998; Pfaus et al., 2001). Sexual learning has also been evidenced using physiological and other non-behavioral measures. For example, cues paired with access to a receptive female can increase serum luteinizing hormone (LH) and testosterone levels in male rats (Graham & Desjardins, 1980), sperm volume and concentration in quail (Domjan, Blesbois, & Williams, 1998), and the number of offspring in quail (Adkins-Regan & MacKillp, 2003; Mahometa & Domjan, 2005; Matthews, Domjan, Ramsey, & Crews, 2007) as well as in blue gourami fish (Hollis, Pharr, Dumas, Britton, & Field, 1997).

Studies examining the effects of learning on partner preference are most relevant to the question of the development of sexual orientation. Partner preference refers to what type of partner is approached or towards whom sexual behaviors are aimed. Maternally-associated odors influence adult mate preference in male rats (Fillion & Blass, 1986). Moreover, studies show that cross-fostered birds (Immelmann, Pröve, Lassek, & Bischof, 1991; Spence & Smith, 2007; ten Cate, Verzijden, & Etman, 2006) as well as mammals (e.g., Kendrick, Hinton, Atkins, Haupt, & Skinner 1998) prefer characteristics of foster over biological parents in mate selection. Such effects have been observed in both males and females although there appear to be sex differences: Kendrick et al. found that preferences for characteristics of foster parents were weaker and reversible in females and Manrique and Gutierrez (2006) found that rearing conditions seem to have a greater impact on partner preference in female quail.

Experience during adolescence or adulthood can also influence partner preference. Pfaus et al. (2001) proposed a flexible critical period in which salient features of the first sexual partner come to be preferred in subsequent partners. Sexual conditioning and conditioned partner preference in particular has been most commonly explored in Japanese quail and rats.

Galef (2008) reported that female quail select a male that has been observed mating with another female and this preference appears to be lasting. On the other hand, males briefly (for 24–48 h) avoid a female that they have seen mating with another male. Domjan and colleagues found that males show preferences for partners bearing cues that were previous associated with copulatory access or visual exposure to a female (for review, see Domjan & Akins, 2011; Domjan & Holloway, 1998). Further, if species typical CSs (e.g., head and neck cues) were used, sexual conditioning was generally more robust, that is, occurring in a single trial (Hilliard, Nguyen, & Domjan, 1997), occurring with a significant delay between CS and US (Akins, 2000) and such learning appeared more resistant to extinction (Krause, Cusato, & Domjan, 2003; Domjan & Hall, 1986), blocking (Köksal, Domjan, & Weisman, 1994), and CS pre-exposure effects (Cusato & Domjan, 1998).

Even though prepared cues may engage learning mechanisms somewhat differently than arbitrary ones, evidence for plasticity involving what could be considered innately predisposed stimuli exists. For example, Nash and Domjan (1991) found a conditioned cross-strain preference—that male brown quail preferred proximity to blonde versus brown female quail if they had copulated with blonde but not brown mates. Moreover, in their second experiment, male quail acquired a same-sex preference after copulating with a sexually receptive non-reproductive male and this preference was just as readily learned as the cross-strain preference. Reversal of this response, subsequently conditioning a preference for a female after acquisition of the preference for a male, was readily learned. However, for those originally conditioned with the female as the CS, reversal learning did not occur. Differences in the quality of sexual reward may have mediated the asymmetry in reversal learning, but it most likely reflects a heterosexual bias in partner preference. Nonetheless, these finding show that, under some circumstances, such a predisposition can be overridden.

Rats are a non-monogamous species; hence, the effects of learning on partner preference would appear to override an innate preference for novelty (Pfaus et al., 2001). Male rats show a conditioned ejaculatory preference (preferring to ejaculate although not necessarily mount or intromit) with females bearing cues (e.g., scented with an odor) associated with prior copulation (Kippin & Pfaus, 2001). Male rats also learn to avoid certain partners that bear an odor paired with internal malaise or with thwarted copulation. Female rats find the ability to pace bouts of copulation rewarding (Paredes & Alonso, 1997) and they show increased solicitations and higher magnitude lordosis for males bearing cues paired with the ability to pace mating (Coria-Avila et al., 2005). Female rats also show cross-strain conditioned partner preferences (Coria-Avila et al., 2006) and Woodson et al. (2002) showed that sexual experience interacted with steroid exposure to shape the sex of partner preference in female rats. Recently, Coria-Avila et al. (2009) have documented the development of male–male partner preferences in sexually naïve male rats if they experienced cohabitation with that male paired with injection of quinpirole, a dopamine (D2) agonist and female–female partner preferences when the partner female was paired with injection of quinpirole and oxytocin (Cibrian-Llanderal et al., submitted).

Although there does not appear to be any experimental work on partner preference in primates, behavioral observations suggest an effect of experience on mate choice in male rhesus monkeys. They mount males and females equally as juveniles but mount females almost exclusively as adults. Having ejaculated with a female better predicted this transition to female mounting partners than did increased testosterone (Wallen, 2001).

In summary, work with nonhumans, in particular quail and rat models, show clear, robust evidence that learning affects sexual partner preference. Such evidence includes demonstrations that conditioning can impact preferences for characteristics typically considered to be innately predisposed.


Money (1986) suggested that the development of a lovemap reflected sexual imprinting in humans. He believed humans are born prepared but sexuoerotically unfinished and hence susceptible to early input.

No studies have directly examined sexual conditioning in prepubescents or even adolescents. However, descriptive research on human partner preference suggests an effect of early experience on later sexual preferences. For example, homogamy—matching for both physical and personality characteristics between opposite sex parent and current sexual partner (Gyuris, Járai, & Bereczkei, 2010; Little, Penton-Voak, Burt, & Perrett, 2002; Wiszewska, Pawlowski, & Boothroyd, 2007)—has been observed with adoptive parents (Bereczkei, Gyuris, & Weisfeld 2004) and such effects were moderated by the quality of the relationship with the opposite sex parent (Bereczkei, Gyuris, Koves, & Bernath 2002; Gyuris et al., 2010; Wiszewska et al., 2007). Rantala, Pölkki, and Rantala (2010) showed that the hairiness of a woman’s father correlated with that of her current partner as well as her general preference for male body hair. Santtila et al. (2009) found that males with early sexual experience desired and had younger partners in adulthood. Also, Perrett et al. (2002) found that adults born to older parents were more attracted to partners with older faces than those born to younger parents. Finally, Bickham et al. (2007) compared heterosexual women (who were behaviorally Kinsey 0’s or 1’s) and found that experimenting sexually with females and masturbating using images of females prior to the age of 18 predicted sex contact with females, masturbating to images of females, fantasizing about a female sex partner when having sex with male, and voyeurism directed at women in adulthood. They proposed that because Kinsey 1’s have a primarily heterosexual orientation and because analogous early female–male behaviors were not significant predictors of adult female–female behavior (i.e., such behavior did not differ between the Kinsey 0’s and 1’s) argue against the effect being due to a same-sex predisposition.

Fetishes and other paraphilias emerge before adulthood (see Money, 1984; Seto, Abramowitz, & Barbaree, 2008) and some data support effects of experience in contributing to atypical patterns of arousal. For example, Weinberg, Williams, and Calhan (1994) found that, among gay men with a foot fetish, many recalled early positive, sexual experiences with other men’s feet and some reported conditioning-like, including one trial learning, effects.

In a meta-analysis of data from sex offenders, Jespersen et al. (2009) documented an association between reported sexual abuse history and sexual offending. Specifically, sex offenders differed in sexual (but not physical or emotional) abuse history from non-offenders, that those who offended against children were more likely to report sexual abuse than those who offended against adults, and that those who were pedophilic were more likely than non-pedophilic offenders to have a sexual abuse history. Seto and Lalumière (2010) showed similar effects for adolescent sex offenders. Burton (2003) reported that adolescent sex offenders who were sexually abused tended to perpetrate the same acts they had experienced themselves. Adolescent offenders who reported being sexually abused also showed greater arousal to children and to coercive sex than non-abused offenders (Becker, Hunter, Stein, & Kaplan, 1989; Becker, Kaplan, & Tenke, 1992; Hunter, Goodwin, & Becker, 1994). Although the abuse–abuser correlation does not fit with a straightforward conditioning interpretation (e.g., one would think victims would be more attracted to those significantly older than themselves), it demonstrates an effect of early experience on sexual behavior and potentially on partner choice.

Studies of clinical conditioning, that is, using behavioral therapy approaches (e.g., masturbatory conditioning and aversion therapy) to alter sexual arousal patterns (which in some ways parallel partner preference research in mature nonhumans), have sometimes been used to support the effect of “later” learning processes on sexual arousal patterns. Such an approach has most often been aimed at changing “deviant” arousal patterns and at times has included sexual orientation. However, the effectiveness of these procedures is difficult to assess because much of the work lacks consistent/appropriate control groups, measures and/or follow-up. At best, some change in arousal patterns has been documented, although it is unclear how enduring such effects are. Further, “success” varies by individual and may depend on motivation for change, personality variables, degree of previous desired, “normative” interest, and sex (Akins, 2004; Bancroft, 1974; Beckstead, 2012). For example, Bancroft (1974) reported that although relatively little research has been done with women, they appeared more likely than men to show change due to clinical conditioning. Such an observation is consistent with the idea that sexual orientation appears more plastic in women (Baumeister, 2000; Diamond, 2000, 2006, 2009; Diamond & Fagundes, 2008) and that, as mentioned earlier, partner preference may be more malleable in females in general.

Ideally, we seek experimental evidence for the influence of conditioning on the development of arousal patterns in humans. There is still relatively little empirical evidence of sexual conditioning using humans and no studies clearly show that conditioning affects partner preference. Most conditioned arousal studies have tested adult male subjects, using visual stimuli (e.g., photographs of prepared or arbitrary cues as CSs; photographs or film depicting sexual images as USs) and have measured learning via changes in genital responding (Lalumière & Quinsey, 1998; Plaud & Martini, 1999). For example, Plaud and Martini found conditioned increases in penile circumference to a slide of a penny jar after it was paired with slides of nude or partially nude females. More recent research has demonstrated conditioned sexual arousal in women as well as learning with more diverse CSs (e.g., olfactory cues), USs (e.g., vibrogenital stimulation), and measures of conditioned responding e.g., subjective arousal, general affective ratings, skin conductance, and neural activity (Both, Laan et al., 2008; Both, Spiering et al., 2008 Hoffmann, 2007, 2011; Hoffmann & Janssen, 2006; Klucken et al., 2009). In addition, evidence of preparedness (better learning with sexually relevant vs. irrelevant cues: Hoffmann, Janssen, & Turner, 2004) and learning without awareness (Both, Laan et al., 2008; Hoffmann et al., 2004) have also been shown. Nonetheless, conditioned sexual responses in humans are often weak and short-lived, which sharply contrasts with the effects found in nonhumans. Methodological differences could account for this discrepancy. Most animal work is done with sexually naïve subjects (although see Snowdon, Tannebaum, Schultz-Darken, Ziegler, & Ferris, 2011) using powerful USs (e.g., copulation) whereas human studies use sexually experienced (young) adults and erotic films as the US. When masturbation was the US, CRs were stronger and more enduring (Kantorowitz, 1978). Sexual interaction (oral or penetrative sex) was used as the US in one unpublished human study and although learning occurred it was not necessarily stronger than that observed when erotic film served as the US. However, this comparison was confounded by differences in retention interval (Hoffmann, Peterson, & Garner, 2009).

In summary, descriptive and some experimental research support a role for experience, and in particular conditioning, in the development of sexual arousal patterns in humans but more work is needed to confirm and clarify such effects.


Innate predispositions in the form of attraction to particular stimulus characteristics (features of the opposite sex in most instances) most likely guide sexual partner preference. However, what specifically we are predisposed to be attracted to, remains unclear. Further, the diversity of erotic taste shows a powerful role for experience in the development of sexual arousal patterns in humans. Nonetheless, we are still a long way from demonstrating if and how experience can affect partner preference or sexual orientation in men and women. Further, sex differences in such effects (e.g., in terms of degree or timing of influence) most likely exist. Yet, consideration of how learning processes may interact with predispositions to develop a fuller, more complete account of the origins of sexual orientation seems in order.


A version of this article was presented at the University of Lethbridge Workshop, The Puzzle of Sexual Orientation: What It Is and How Does It Work?, Lethbridge, Alberta, Canada, June 2010.

Copyright information

© Springer Science+Business Media, LLC 2012