Neural Correlates of Sexual Orientation in Heterosexual, Bisexual, and Homosexual Women


We used fMRI to investigate neural correlates of responses to erotic pictures and videos in heterosexual (N = 26), bisexual (N = 26), and homosexual (N = 24) women, ages 25–50. We focused on the ventral striatum, an area of the brain associated with desire, extending previous findings from the sexual psychophysiology literature in which homosexual women had greater category specificity (relative to heterosexual and bisexual women) in their responses to male and female erotic stimuli. We found that homosexual women’s subjective and neural responses reflected greater bias towards female stimuli, compared with bisexual and heterosexual women, whose responses did not significantly differ. These patterns were also suggested by whole brain analyses, with homosexual women showing category-specific activations of greater extents in visual and auditory processing areas. Bisexual women tended to show more mixed patterns, with activations more responsive to female stimuli in sensory processing areas, and activations more responsive to male stimuli in areas associated with social cognition.


Studies using physiological measures have found that women tend to have non-specific patterns of genital arousal1,2,3. That is, in contrast to men, women tend to show similar degrees of arousal to erotic stimuli depicting either sex. For example, heterosexual women have generally shown equivalent arousal to both erotic stimuli featuring men and erotic stimuli featuring women. This has been repeatedly demonstrated with vaginal photoplethysmography3,4. This pattern has also been found using less direct measures such as looking time5, pupil dilation6, and fMRI7. Notably, homosexual women’s arousal patterns are more category-specific than heterosexual women’s, although less so than men’s8.

The fact that women’s sexual arousal patterns are less category-specific than men’s has been interpreted as a potential contributor to gender differences in “erotic plasticity”9, which Baumeister has defined as “the extent to which sex drive is shaped by social, cultural, and situational factors.”

Baumeister offered three lines of evidence when he initially proposed that women may have greater erotic plasticity compared with men: (1) women show larger effects of social and cultural factors on sexual attitudes, desire, and behavior; (2) sexual attitude-behavior consistency is lower in women than in men; (3) individual women exhibit more variation in sexual behavior across time than men. Women’s less specific arousal patterns may also contribute to their increased “sexual fluidity”10, which Diamond has defined as an individual’s “capacity for situation-dependent flexibility in sexual responsiveness, which allows individuals to experience changes in same-sex or other-sex desire across both short-term and long-term time periods”11,12.

One might hypothesize that arousal patterns of bisexual women should be similar to the non-specific arousal patterns of heterosexual women; however, studies of women’s arousal patterns have mostly neglected to include bisexual women. Heterosexual women’s arousal does not appear to favor erotic stimuli of either sex, and thus may be considered to reflect a bisexual pattern. (We do not mean to imply that heterosexual women are confused or in denial about their “real preferences”; rather, the findings in need of explanation are why heterosexual women show non-heterosexual arousal patterns in the laboratory). The implication of women’s non-specific arousal patterns for their sexual orientations is difficult to interpret. Most women, like most men, behave and identify heterosexually13,14,15,16,17,18. However, men are more likely than women to identify as completely heterosexual or completely homosexual, and women are more likely than men to identify as bisexual or “mostly heterosexual”19.

If arousal patterns are similar between heterosexual and bisexual women, the question remains what distinguishes the two groups. One possibility, supported by some research, is that bisexual women tend to have greater sexual motivation, which may increase the likelihood of exploring a capacity for attraction to both sexes20,21. Or, bisexual women may be more aware than heterosexual women of their non-specific arousal22, which could partially contribute to bisexual sexual motivation. Alternatively, bisexual women may be more likely than heterosexual women to interpret their non-specific arousal states in sexual or romantic terms.

It is also possible that bisexual women’s arousal patterns differ from those observed in heterosexual women. Consistent with this possibility, recent studies suggest that women with bisexual interests tend to be more aroused by female than by male erotic stimuli23,24,25. Perhaps for some women with female-biased arousal patterns, this bias can motivate non-heterosexual feelings, behavior, and identity.

Interpretations of non-specific arousal patterns in women are further complicated by the fact that female genital arousal exhibits relatively low correlations with subjectively reported sexual arousal, in contrast to the high correlations observed in men26. Discrepancies between existing genital and subjective measures indicate that some women may report substantial subjective arousal without substantial genital arousal, and vice versa. It has also been suggested that non-specific arousal patterns may not indicate affective responses to erotic stimuli, but may instead reflect a kind of protective preparatory response27.

Neuroimaging assessments may shed light on the neural systems that are involved in responding to a given paradigm. Functional magnetic resonance imaging (fMRI) is a neuroimaging approach that allows for the indirect assessment of brain activity by tracking ratios of oxygenated and deoxygenated blood a proxy for neural firing. When used in the context of presenting erotic stimuli, this non-invasive neural measure could provide a converging line of evidence for interpreting the genital and subjective arousal findings described above. In this study, we used fMRI to specifically focused on the “reward system” in order to address the question: to what extent is there an affective significance to findings from the literature on women’s sexual orientation and genital arousal?

The part of the “reward system” that we focused on is the ventral striatum, a dopamine-sensitive area of the brain that is a reliable measure of reward-related processing–and in particular, wanting and “incentive motivation”28,29–including with respect to sexual orientation30. Most neuroimaging studies of sexual response have focused on men31,32, but the ventral striatum has also been found to reliably activate in studies of women’s responses to erotic stimuli33,34,35. However, until now, no studies have measured neural responses to erotic stimuli in bisexual women.

The present investigation primarily focused on two hypotheses: (1) Homosexual women may show greater category-specificity than non-homosexual women in brain activity, as suggested by the genital arousal literature; (2) Bisexual women may show larger biases towards female stimuli, compared with heterosexual women. We tested these hypotheses with respect to subjective and neural responses to erotic pictures and erotic videos. We used two different kinds of erotic stimuli because of their potentially non-overlapping strengths and weaknesses. Erotic pictures may be particularly well-suited for assessing the initial appraisal of sexual stimuli, but their brevity may not reflect the kinds of experiences that drive sexuality in the real world. Erotic videos may allow for the measurement of more intense states, but their extended duration may also provide opportunities for self-regulatory efforts to modify erotic responses.



Participants were 26 heterosexual women, 26 bisexual women, and 24 homosexual women, recruited from a variety of publicly-posted and online advertisements seeking (paid) volunteers for a neuroimaging study of sexual orientation and arousal. Bisexual women were required to have had at least two previous sexual partners and one romantic partner (of three months or greater duration) of each sex. Homosexual and heterosexual participants all met these criteria with respect to their respective preferred sexes.

After responding to advertisements, participants were screened for inclusion using online questionnaires. Participants provided information about sexual orientation, sexual interests, and personality, in addition to answering screening questions relevant to medical eligibility for fMRI research. Participants were required to be right handed, non-claustrophobic, free from ferromagnetic implants, and not currently taking psychiatric medications. Participants were informed of the risks and nature of the study and agreed to participate in all portions of the research. This study was approved by the Institutional Review Board of Northwestern University and carried out in accordance with its guidelines. Informed consent was obtained from each participant for every portion of the study in which they participated.

Participants’ sexual orientation was assessed using self-reported identities (i.e. “Homosexual”/“Gay”, “Bisexual”/“Bi”, “Heterosexual”/“Straight”), as well as with a modified Kinsey score, which asked participants about their sexual fantasies throughout adulthood as well as in the past year. The scale ranged from 0 to 6, with 0 corresponding to an exclusively heterosexual orientation and 6 corresponding to an exclusively homosexual orientation. Responses to questions about adulthood and about the past year were averaged to create a Kinsey score for each participant. The average Kinsey score was 0.8 for heterosexual women (SD = 0.7, range = 0–2), 2.63 for bisexual women (SD = 0.7, range = 2–4.5), and 5.2 for homosexual women (SD = 0.68, range = 4–6).

Participants’ ages ranged from 21 to 46 years old. Mean ages were 29.7 for heterosexual women (SD = 5.86, range = 25–46), 30.27 for bisexual women (SD = 6.41, range = 21–48), and 29 for homosexual women (SD = 3.12, range = 25–38). The sample of 76 participants was racially and ethnically diverse, with 23 non-Caucasian participants including two Latina participants, ten African-American participants, four Asian-American participants, and seven participants who identified otherwise or who identified as multiethnic/multiracial. Groups did not significantly differ either with respect to age (F(2,73) = 0.348, p = 0.708) or ethnicity (c2(2, N = 76) = 2.94, p = 0.23). We also confirmed that ethnicity did not significantly impact responses to the erotic stimuli.

Stimuli and Procedure

Subjects experienced two experimental paradigms in the scanner: first erotic pictures were shown (over a period of ~21 minutes), and then erotic videos were shown (over a period of ~19 minutes) after a brief rest period. Picture stimuli were shown before video stimuli for all participants in an attempt to promote stimulus engagement. That is, it was assumed that potentially less intense stimuli might be better presented earlier in the experimental session while attentional resources are highest. Further, there was concern that first showing more intense stimuli would reduce engagement with subsequent stimuli. As such, pictures and videos stimuli were not counterbalanced with respect to each other.

Participants watched stimuli while laying down with a combination of earplugs (to minimize scanner noise) and over-ear headphones (for video sound and communication with experimenters). Images were displayed via projector onto a wall, which was made viewable to participants via an angled mirror placed above the eyes.

Erotic pictures paradigm

The present study employed a subset of the picture stimuli used in Safron et al.36 and Sylva et al.7. Pictures depicted a nude man, a nude woman, or a same-sex couple (i.e., either two men or two women) engaged in explicit sexual contact. Erotic stimuli featuring both individual nudes and same-sex pairs engaging in explicit sexual interaction is common in research on sexual arousal and sexual orientation2,4,37, which is not the case when stimuli featuring male-female couples is presented. However, erotic stimuli featuring explicit sexual activity in same-sex couples tends to be substantially more arousing compared with pictures of single nudes4. Such stimuli are similar to pictures of nude individuals, in the sense that only men or women, but not both, are depicted in a given picture. Thus, sexual arousal induced by them is relatively unambiguous in terms of the gender to which participants are responding.

In each of two 10.5-minute runs (ordering counterbalanced), participants viewed 40 erotic pictures featuring male models and 40 erotic pictures featuring female models. Each picture was shown for 3.5 seconds, followed by a variable-duration fixation cross presented for either 1.5, 6.5, or 11.5 seconds. Variable-duration baselines were utilized for superior deconvolution of the BOLD signal in a rapid event-related design for fMRI (in which evoked signals are never allowed to return to baseline levels). During the presentation of each picture, participants used buttons held in their right hands to rate that image on a scale of −2 to +2 (respectively: “strongly disliked,” “disliked,” “liked,” “strongly liked”), with no option of 0 for neutral ratings. Neutral options for ratings were not provided for the sake of consistency with previous research using the same stimuli. Note: Subjective ratings of pictures were lost for some participants due to a button-box equipment error.

Erotic videos paradigm

Following picture assessment, participants were shown six video clips depicting individual masturbating men and six video clips depicting individual masturbating women. Depicted individuals appeared sexually aroused but did not reach orgasm. To estimate baseline responses, six natural landscape videos were shown.

In each of two 9.25-minute runs (ordering counterbalanced), videos were presented for 15 seconds each, followed by a 15-second distraction task requiring participants to indicate via button-press when a number in a series decreased by an interval other than seven. This task was intended to facilitate a return to emotional and physiological baseline. 15-second stimulus presentations were chosen as a desirable stimulation period in an fMRI block design, which can potentially be more sensitive than event-related designs38.

After leaving the scanner, participants viewed the videos once more and provided ratings of each clip. Videos were rated using a 5-point scale for degree of sexual appeal, ranging from “not at all” (0) to “very much” (4), with a midpoint of “somewhat” (2).’

fMRI signal extraction methods

Image acquisition

A Siemens Trio 3 T magnet and 12-channel RF head coil were used to collect T2*-weighted gradient-recalled EPI images from the whole brain (32 3-mm slices with a 0.99-mm interslice gap; TR = 2500 ms; TE = 20 ms; flip angle = 80°; FOV = 200 × 220 mm, 120 × 128 acquisition matrix). Slices were taken along the plane connecting the anterior and posterior commissures, with a 1.72 mm × 1.72 mm × 3.99 mm resolution, with more refined axial dimensions intended to produce less distortion and signal dropout in sub-cortical areas, although possibly at the expense of signal-to-noise ratio. During each picture run, 250 whole-brain volumes were collected, and during each video run, 220 whole-brain volumes were collected, with the first four volumes discarded to account for initial magnetization effects. For anatomical localization, a structural MRI scan consisting of T1-weighted images was conducted after the testing runs (160 1-mm axial slices; TR = 2.1 ms; TE = 4.38 ms; flip angle = 15°; FOV = 220 mm; 256 × 192 matrix).

Image pre-processing

Image pre-processing and analysis was performed using SPM 12b (Wellcome Trust Centre for Neuroimaging, London, UK), and implemented in Matlab v 8.1.604 (The MathWorks Inc., MA, USA).

Functional (EPI) volumes were first corrected for slice timing. Each participant’s volumes were then registered to the mean slice, after which the registered volumes were resliced, used to create a mean resliced image, and then co-registered to the structural (T1) image. All EPI images, including the mean resliced image, as well as the structural (T1) scans were then spatially normalized to Montreal Neurological Institute (MNI) space, and re-sampled to 3 × 3 × 3 mm (27 mm3) resolution. Normalized functional images were then smoothed to an 8 mm full-width-at-half-maximum Gaussian kernel.

Signal to noise ratio and head coverage exclusions

To exclude participants with poor signal due to either head motion or scanner conditions, average signal-to-noise ratio (SNR) over time was calculated for each subject (after preprocessing, using a mask that included only voxels with appreciable EPI signal). The SNR ratio for each voxel (mean divided by standard deviation) was averaged across all voxels in the brain39. Participants whose picture data SNR was less than one standard deviation below the mean were excluded from picture analyses. Similarly, participants whose video data SNR was less than one standard deviation below the mean were excluded from video analyses.

Based on these criteria, fourteen participants (five heterosexual, five bisexual, and four homosexual) were excluded from fMRI and subjective picture analyses, and sixteen participants (six heterosexual, six bisexual, and four homosexual) were excluded from fMRI and subjective video analyses. After exclusions were performed for SNR, we included a total of twenty-one heterosexual women, twenty-one bisexual women, and twenty homosexual women in fMRI picture analyses. Video analyses after SNR exclusion included eighteen heterosexual women, eighteen bisexual women, and twenty homosexual women. To check the validity of our SNR criterion, head motion plots were visually inspected for all participants (Parrish, et al.39). Excluded participants had highly variable head positions as compared to included participants. An additional validity-check was performed using evoked responses to erotic pictures minus a fixation-cross baseline. Excluded participants had substantially reduced activity in visual cortices as compared to included participants.

An additional thirty-two participants (twelve heterosexual, twelve bisexual, and eight homosexual) were excluded from subjective picture rating analyses due to insufficient subjective data resulting from a button-box equipment error. Five participants (three bisexual and two homosexual) were excluded from subjective video analyses for the same reason. Thus, after exclusions were performed for insufficient subjective data, we included a total of nine heterosexual women, nine bisexual women, and twelve homosexual women in subjective picture analyses, and twenty heterosexual women, seventeen bisexual women, and eighteen homosexual women in subjective video analyses.

For whole-brain analyses, mean functional scans were individually examined to identify participants with substantial cutoffs in head coverage. As a result, one heterosexual female who had substantial frontal lobe cutoff was excluded from whole-brain analyses in addition to those participants excluded for SNR.

First-level analyses

For both the video and picture assessments, a standard general linear model (GLM)40 was used in identifying hemodynamic changes for each participant, and a high-pass filter (cutoff 128 s) was used to remove low-frequency temporal noise.

Estimated average activity was calculated for each participant’s separate responses to male pictures, female pictures, male videos, and female videos (contrasted with fixation cross for pictures and neutral nature scenes for videos). These estimates were used for region of interest analyses. For whole-brain analyses, estimated average activity was also calculated for each participant’s response to male compared with female pictures and videos. For both the picture and video assessments, each participant’s responses to each stimulus contrast of interest were concatenated within stimulus type, using data from both the 1st and 2nd runs.

Ventral striatum region of interest analyses. An a priori region of interest (ROI) analysis was performed on the ventral striatum—centered on the nucleus accumbens—as this was the area most likely to indicate desire. The ventral striatum and hypothalamus are the only two areas that have been shown to be specifically associated with sexual (as opposed to general) arousal41,42. We focused on the ventral striatum because it likely has higher validity for reflecting sexual incentive value compared with the hypothalamus, which contains a variety of nuclei with heterogeneous functions (including sexual arousal) that would be difficult to disambiguate with the limited spatial resolution of 3 T fMRI.

The ventral striatum ROI mask used in the present study was drawn on an MNI template brain using the WFU PickAtlas toolbox for SPM 843. It was anatomically defined as a dilated intersection of the ventral anterior caudate and putamen. The resulting ventral striatum ROI is shown in Fig. 1.

Figure 1
Figure 1

Mask used as the ventral striatum (VS) ROI, drawn using an average brain in the WFU PickAtlas toolbox for SPM 8. MNI coordinates displayed: x = 0, y = 17, z = −8.

see full details

Leave a Reply

Your email address will not be published. Required fields are marked *