Archives of Sexual Behavior

, Volume 40, Issue 4, pp 755–765

Women’s Sexuality, Well-Being, and the Menstrual Cycle: Methodological Issues and Their Interrelationships

Authors

    • Department of PsychologyUniversity of Hawaii at Hilo
    • Social Sciences DivisionUniversity of Hawaii at Hilo
  • Marites J. Calibuso
    • Department of PsychologyUniversity of Hawaii at Hilo
  • Amanda L. Roedl
    • Department of PsychologyUniversity of Hawaii at Hilo
Original Paper

DOI: 10.1007/s10508-010-9630-3

Cite this article as:
Brown, S.G., Calibuso, M.J. & Roedl, A.L. Arch Sex Behav (2011) 40: 755. doi:10.1007/s10508-010-9630-3

Abstract

Although many studies report that women’s sexual behavior varies across their menstrual cycles, the research findings remain inconsistent. In this study, we addressed two methodological issues in research on the menstrual cycle: how ovulation is measured/inferred and whether data using menstrual cycles or participants’ scores averaged across cycles as units of analysis yield similar results. We also employed an abstinent comparison group in addition to examining how emotional well-being was related to libido and sexual behavior through factor and regression analysis. Data were obtained from 97 participants. There were no significant differences in the results of analyses performed using cycles with known LH surges to determine ovulation versus cycles based on backward counts. However, we concluded that statistical power might be compromised when the known timing of ovulation was less accurate. Likewise, we found few overall differences in the results when we analyzed data using cycles with known LH surges compared to participants’ averaged data across cycles. Women, including those in the abstinent group, reported increased sexual behavior prior to ovulation. Allosexual behavior was positively related to libido, and negatively related to positive and “premenstrual” emotional factors. Autosexual behavior was predicted by libido and an energetic/creative emotional factor. Our findings support hypotheses that women’s sexual behavior is related to both mating and pair-bond formation.

Keywords

Emotional well-beingSexual orientationLH surgeLesbiansMenstrual cycleOvulation

Introduction

Much of the evidence to date supports the idea that women are continuously sexually receptive throughout a menstrual cycle but experience changes in their sexuality just prior to and at the time of ovulation (Gangestad & Thornhill, 2008) and, perhaps, premenstrually (Englander-Golden, Chang, Whitmore, & Dienstbier, 1980). Wilcox, Weinberg, and Baird (1995) found that women were fertile several days prior to and on the day of ovulation and later reported that intercourse frequency peaked on the day preceding and on the day of ovulation (Wilcox et al., 2004). Likewise, Bullivant et al. (2004) found that sexual activity was elevated for the 3 days prior to and 2 days after the luteinizing hormone (LH) surge. Women ovulate 24–36 h after the LH surge. However, other research has reported no significant differences in sexual activity across the menstrual cycle (Brewis & Meyer, 2005), differences in female-initiated but not male-initiated sexual activity at mid-cycle (Adams, Gold, & Burt, 1978), differences in the timing of inter-pair versus extra-pair copulations across the cycle (Bellis & Baker, 1990), and differential sexual attraction to men other than their partners at mid-cycle (Haselton & Gangestad, 2006).

The different findings on women’s sexuality result from a variety of factors, some behavioral and some methodological. Examples of behavioral factors include whether the women studied were partnered or not (Bellis & Baker, 1990) and the degree of sexual attraction they felt towards their partners (Haselton & Gangestad, 2006), whether they were engaging in sex with women or men (Burleson, Trevathan, & Gregory, 2002), the sexual activities that were monitored during the study (initiation of sexual activity: Adams et al., 1978; libido: Bullivant et al., 2004; autosex [sex with self]: Burleson et al., 2002; allosex [sex with a partner]: Bellis & Baker, 1990; sexual fantasies: Bullivant et al., 2004; Gangestad, Thornhill, & Garver, 2002; sexual arousal: Englander-Golden et al., 1980; Slob, Bax, Hop, Rowland, & van der Werff ten Bosch, 1996; see Table 1 for other examples) or whether participants’ sexual behavior was measured objectively and prospectively (Bullivant et al., 2004) or subjectively via questionnaire (Pillsworth, Haselton, & Buss, 2004). Additionally, most of the age ranges were between 18 and 40 (Table 1) but at least one study examined women who were up to 53 years of age (Burleson et al., 2002).
Table 1

Summary of selected articles examining sexual differences across the menstrual cycle

Article

N

Age

Definitions of phases or ovulation

Number of phases/days

Major findings

Adams et al. (1978)

12

27–37

Backward count; Days on standardized 28 day cycle

NA

Female-initiated but not male-initiated sexual activity increased mid-cycle

Bellis and Baker (1990)

2707

No age data

Forward count on retrospective data; Days on standardized 28 day cycle

3 phases

Intrapair copulations peaked in luteal phase; extrapair copulations peaked pre-ovulatory and were positively correlated with probability of conception

Brewis and Meyer (2005)

>700

18–40

Forward count on retrospective Day 1 of menstruation

4 phases

No evidence that sexual behavior is related to menstrual cycle phase

Bullivant et al. (2004) (2 studies)

19 = 5 cycle

46 ≥ 1 cycle

18–35

Ovukit (LH surge); cervical mucus; BBT and pregnanediol-3-glucuonide

6 phases

Sexual activity, libido and fantasies peaked around ovulation

Burleson et al. (2002) (lesbians, L & heterosexuals, H)

L = 89

H = 139

21–49

19–53

BBT; cervical mucus

6 phases

Allosexual behavior peaked in the follicular phase; autosexual behavior in the follicular and ovulatory phases

Englander-Golden et al. (1980)

26

No data

Backward count to day 13

3 phases;

20 days

Peaks in sexual arousal at ovulation (day-13) and pre-menstrually

Gangestad et al. (2002)

51

18–34

Ovusign to detect LH surge

2 phases

Women initiated and engaged in sex near ovulation; women reported increased attraction and fantasies about men other than their partners and more partner vigilance near ovulation

Garver-Apgar, Gangestad, and Thornhill (2008)

124

M = 21.9

Backward count to day 15 (estimated day of LH surge)

Estimated reproductive hormone levels

Estimated estrogen and progesterone levels correlated with symmetry preference

Graham et al. (2000)

40

19–45

Clearplan easy for LH surges

2 phases

Women reported higher levels of sexual arousal during the periovulatory rather than follicular phase

Harvey (1987)

69

18–34

BBT or backward count to day 14

5 phases

Female-initiated (autosexual and with partner) and male-initiated sexual activity peaked at mid-cycle

Haselton and Gangestad (2006)

38

17–43

Backward count to day 15

2 phases

Women mated to men with low sexual versus investment attractiveness were more attracted to men other than their husbands when fertile

Haselton et al. (2007)

30

18–37

Clearblue to detect LH surge

2 phases

Women judged to dress better when more fertile

Krug, Plihal, Fehm, and Born (2000)

11

21–32

Q-test for LH surge

3 phases

Women reacted maximally to sexual stimuli, judged the stimuli as less offensive and were more high spirited during ovulation

Laeng and Falkenberg (2007)

7

21–24

Backward count to day 15

3 phases

Pupil size change was largest to boyfriend during ovulation

Matteo and Rissman (1984)

7 lesbian couples

21–37

Backward and forward count standardized like Adams et al. (1978)

10 phases

Peaks in orgasm, total encounters, self-initiated encounters at mid-cycle; peaks in sexual fantasies during follicular phase

Miller, Tybur, and Jordan (2007)

11

26.9 ± 5

Forward to days 9–15

3 phases

Women lap dancers earned more tips in their fertile phase

Pillsworth et al. (2004)

173

17–24

Backward count 15 days based on expected day of next menstruation

Regression on probability of conception

Partnered, but not unpartnered, women reported increasing sexual desire as probability of conception increased

Pipitone and Gallup (2007)

51 (both contraception and normal)

17–30

See phases

4 equally spaced samples across cycle

In naturally cycling women increased conception risk associated with greater voice attractiveness

Roberts et al. (2004)

48

19–33

Forward count

2 phases

Women were rated as more attractive when they were in the follicular phase

Roney and Simmons (2008)

72

M = 18 (.1 year)

Reported first day of last menses and forward count

25 days

Correlation between estimated estrogen level in women and testosterone level in preferred male

Rosen and Lopez (2009)

17

18–22

Forward and backward counts

2 phases

Women in the fertile phase made more shadowing errors in the presence of a courtship distracter

Sanders et al. (1983), Bancroft, Sanders, Davidson, and Warner (1983)

19 (from PMS clinic) 36 (not from clinic)

M = 34.3 (5.3)

M = 31.2 (5.8)

Estradiol, LH and progesterone levels

6 phases

Frequency of sexual behavior, female initiated sexual activity, sexual feelings and pleasant thoughts peaked in the mid-follicular phase

Schreiner-Engel, Schaivi, Smith, and White (1981)

30

20–30

Backward count

3 phases

Sexual arousal is greater during the follicular and luteal phases

Schwarz and Hassebrauck (2008)

40

17–38

Backward count 15 days

2 phases

Women were perceived as dressing more provocatively on high fertility days

Sheldon, Cooper, Geary, Hoard, and DeSoto (2006)

56 (30 taking OCs)

M = 21.6 (.9)

Forward count 15 from retrospective 1st day of menstruation

14 days

Day of cycle negatively correlated with sex for intimacy

Singh and Bronstad (2001)

17

M = 22.4 (4.5)

Forward count from retrospective 1st day of menstruation

2 phases

Women’s scent is more attractive during follicular phase

Slob et al. (1996)

20

21–46

Serum progesterone

2 phases

Differences in sexual arousal dependent of time of test not phase of cycle

Stanislaw and Rice (1988)

1066

19–44

BBT

NA

Day of BBT shift is positively correlated with the day of sexual desire onset

van Goozen, Wiegant, Endert, and Helmond (1997)

19

24–40

Measured estradiol and progesterone in blood via radioimmunoassay

4 phases

Sexual activity (both masturbation and intercourse) increased around ovulatory phase; and remained high; reported self-initiated sexual activity peaks in the ovulatory phase

Wilcox et al. (2004)

38 (IUD) 30 tubal ligation

~= 25–36

Measured estrogen and progesterone metabolites in urine

Days

Intercourse frequency peaks on the day prior to and day of ovulation

Note: Studies were chosen to maximally illustrate methodological differences rather than to provide a comprehensive review of the literature. Ns usually indicate women with natural menstrual cycles unless we indicate otherwise

The most glaring methodological issue in research on sexuality and the menstrual cycle is how the researchers determined or decided the time of ovulation (Table 1). Some researchers employed physiological measurements to determine the LH surge (e.g., Ovukit: Bullivant et al., 2004; Ovusign: Gangestad et al., 2002; Clearplan: Graham, Janssen, & Sanders, 2000), measured cervical mucus (Burleson et al., 2002), or estrogen and progesterone levels (Sanders, Warner, Backstrom, & Bancroft, 1983) or their metabolites (Wilcox et al., 2004), but most studies employed backward or forward counts from Day 1 of menstruation (Table 1). Similar issues surround differences in the number of menstrual phases defined across a cycle (10 phases, Matteo & Rissman, 1984; 6 phases, Bullivant et al., 2004; 2 phases, Rosen & Lopez, 2009; Singh & Bronstad, 2001), the length of each phase (same length, Singh & Bronstad, 2001 or different lengths, Bullivant et al., 2004; Burleson et al., 2002), if data were broken down by cycle (Brown, Morrison, Calibuso, & Christiansen, 2008a) or averaged across women who contributed different numbers of cycles (Bullivant et al., 2004), and the number of cycles women were followed (usually only one but sometimes up to five cycles). Most researchers defined phases based on women’s hormone profiles (broadly proliferative/follicular with high estrogen levels versus secretory/luteal with high estrogen and higher progesterone levels).

Additionally, when examining the research on the relationship between sexual behavior and women’s menstrual cycles, we failed to find a study that incorporated an abstinent comparison group. Some studies indicated that some of their participants were not engaging in intercourse (Pillsworth et al., 2004) but they did not statistically compare the occurrence of libido and autosexual behavior in the abstinent women with sexually active women.

Another issue in women’s sexuality reflects differences in sexual desire or proceptivity (Hill, 1988) versus pair-bond formation (Diamond, 2003; Gonzaga, Turner, Keltner, Campos, & Altemus, 2006). Women experience both sexual desire/libido and feelings of affection toward others (Diamond, 2003). Libido is probably related to sexual mating while affection is related to pair-bond formation (Gonzaga et al., 2006). Diamond (2003) hypothesized that libido and affection are independent emotional/motivational states that can interact with one another but do not necessarily do so.

If libido and affection are independent, then one should be able to separate their effects. Unfortunately, most research into women’s emotional well-being has concentrated on negative (Hourani, Yuan, & Bray, 2004; Walker, 1997) rather than positive emotional states. However, Brown, Morrison, Larkspur, Marsh, and Nicolaisen (2008b) reported that naturally cycling women experienced fewer negative emotions and more energy at mid-cycle than women using hormonal forms of contraception. In fact, they found that the overall emotional well-being of women using hormonal contraception was more negative than the emotional well-being of naturally cycling women.

In this article, we report on data collected on the menstrual cycles and the sexual desires and behavior of naturally cycling women. In our study, we collected data on cycles with confirmed LH surges but also on cycles without confirmed surges. For the latter cycles, we used a backward count method similar to the ones displayed in Table 1. Therefore, we compared data from cycles with known LH surges with those in which ovulation was estimated using a backward count. We defined five menstrual cycle phases with each phase’s length being 3 days to control for phase length and only examined women between 18 and 40 years of age. Most of our participants were followed for three cycles but many were unable to detect LH surges for each cycle. We, therefore, analyzed data on all cycles with LH surges co-varying out the number of cycles a participant contributed and compared those results with analyses that examined averaged scores across participants who reported one or more LH surges.

Rather than concentrating on sexually active women, we included an abstinent comparison group to compare with our allosexually active heterosexual and lesbian/bisexual women. If sexual behavior does peak during ovulation, we should be able to detect peaks in libido and autosexual behavior in the abstinent women that might be obscured in allosexually active women because of their partner’s desires. Additionally, the relationship between emotional well-being and type of sexual activity was examined.

In summary, we examined two methodological issues concerning research on the menstrual cycle: (1) Do results differ when ovulation and menstrual phase are determined using LH surge and first day of menstruation data versus first day of menstruation and a backward count method? and (2) Do results differ when cycles were used as units of data analysis versus when participants’ averaged scores across cycles were used in the analyses? We also examined specific hypotheses about sexual behavior and the menstrual cycle. First, we tested whether abstinent women exhibited the same patterns of sexuality in terms of libido and autosexual behavior as sexually active heterosexuals and lesbians. Second, we examined how sexual behavior was related to emotional behavior through factor analysis and multiple regression.

Method

Participants

The protocols for this study and the consent forms were approved by the institutional review board at the University of Hawaii. The data were collected from January 2005 to December 2007. Participant recruitment occurred at the University of Hawaii at Hilo through classroom visits and campus fliers. Potential participants attended an information session about eligibility. Between 70 and 80% of the participants met the eligibility criteria (see below) and participation rate was 95% of those who met the eligibility criteria. Eighty-nine percent of the participants (N = 136) completed at least one menstrual cycle. At enrollment, participants were told that they would be asked to contribute data (see below) for three menstrual cycles and would be paid $150 for each completed cycle. For this article, we present data on 97 participants who did not use hormonal forms of contraception.

Participants met the following eligibility criteria: (1) aged 18-40 (M = 25.7 years; SD = 5.6); there were no significant group differences in age; (2) reported regular 25–35 day menstrual cycles; (3) not using an intrauterine device; (4) no hysterectomies; and (5) no autoimmune disease or use of medication affecting menstrual cycles, immune system function or reproductive hormones.

Procedure

Upon enrollment, participants met with a research associate to review study protocols and provide informed consent. Height and weight measurements were taken to determine body mass index (BMI = wt in kg/ht in m2; range, 17.9–44.5; group means were not significantly different). Bioelectrical impedance analysis (RLJ Systems BIA-101A) was used to determine percent body fat (range, 14–45; group means were not significantly different). We measured body mass and percent body fat as potential control variables because too little or too much body mass or fat can influence reproductive hormones (Ferin, Jewelewicz, & Warren, 1993; Siiteri, 1987). Participants received training on how to collect data on their LH surges. The LH surge was determined using Assure Ovulation Predictor kits (the 9-day version, Contraception Technology, San Diego, CA). The length of a participant’s reported cycle determined when she began using the Ovulation Predictor Kit (e.g., if her cycle was 24 days, she began using the kit on Day 4 after menstruation while if her cycle was 30 days, she began using the kit on Day 10 after menstruation). Participants were instructed to test their urine for LH at the same time each day and to read the results within 3–5 min of the test. Participants continued testing their urine until they obtained a positive result or they menstruated. Ovulation occurs between 24 and 36 h after the LH surge. Participants began collecting data on the first day of the next menstrual bleed following the meeting.

Participants completed daily questionnaires each evening of their menstrual cycle without recording retrospective data. Data were collected on libido using a three-point scale (1 = less than usual, 2 = about the same as usual, 3 = more than usual), so each participant could act as her own control, and sexual behavior including autosex (sex with self) and allosex (sex with another individual). For the sexual data each participant reported daily on whether she had sex with herself (0 = no autosex; 1 = autosex) and/or another person (0 = no allosex; 1 = allosex). The daily questionnaire also included 24 questions to assess participants’ emotional well-being. Participants recorded the degree that they experienced each emotion using a three-point scale similar to the one reported above for libido. Sixteen negative emotions were derived from the DSM-IV for Premenstrual Dysphoric Disorder and eight more (Table 2) were included to measure positive emotions and energy levels (Brown et al., 2008b).
Table 2

Factor analysis on 24 emotions

Emotion

Factor

Premenstrual

Positive affect

Energetic/creative

Absentminded

.84

.18

−.06

Affection

−.15

.69

−.13

Alert

−.08

.06

.79

Anger

.84

−.06

.15

Anxious

.80

.10

.14

Burst of energy

.01

.38

.73

Creative

.41

.37

.52

Depressed

.89

−.16

.05

Difficulty concentrating

.82

.10

−.22

Excited

.23

.73

.26

Happy

−.28

.76

.21

Helpless

.89

−.13

.02

Irritable

.73

−.14

.22

Lethargic

.75

.05

−.36

Lonely

.84

−.19

.03

Mood Swing

.77

−.16

.29

Orderliness

.36

.10

.53

Out of control

.90

.14

.08

Well-being

−.10

.70

.32

Overwhelmed

.74

.00

.11

Restless

.66

−.05

.27

Sadness

.88

−.20

.02

Tension

.82

−.01

.12

Withdrawn

.78

−.23

.08

Note: Factors with loadings >.50 are in bold

Of the 97 participants who provided data for the study, 8 participants never detected an LH surge and 89 detected at least one LH surge. Of the 89 participants who had detectible LH surges, 9 contributed data to both the abstinent and heterosexual groups and 2 contributed data to both the heterosexual and lesbian groups (the latter participants had one cycle in which they exclusively engaged in sex with male(s) and another cycle in which they exclusively engaged in sex with female(s)) so we had N = 100 (Table 3). If a participant did not detect an LH surge for a particular cycle, we used a backward count method to estimate the date of ovulation (Table 4). We could, therefore, compare the two different methodologies LH surge (N = 185) versus backward count (N = 71) of estimating the time of ovulation and determining menstrual cycle phase. For participants with detectable LH surges, the cycle then was divided into 5 phases based on the day of the LH surge and time of menstrual bleeding. The menstrual phase consisted of the first day of menstruation (Day 1) and the subsequent 2 days, the follicular phase of days 5, 6, and 7 after Day 1, the ovulatory phase of the day of the LH surge and the subsequent 2 days, the luteal phase of days 7, 8, and 9 after the LH surge and the premenstrual phase of the 3 days prior to the next menstrual flow. Data were averaged across the 3 days defining a phase for all analyses.
Table 3

Number of participants (N), number of cycles with documented LH surges (cycles), and the number of cycles each participant contributed with an LH surge for the analyses

Abstinent

Heterosexual

Lesbian

N

Cycles

N

Cycles

N

Cycles

33

58

45

85

22

42

14

1 cycle

17

1 cycle

8

1 cycle

13

2 cycles

16

2 cycles

8

2 cycles

6

3 cycles

12

3 cycles

6

3 cycles

Note: Analyses were based on either averaged values in each phase across the participants’ cycles or used cycles with LH surges, covarying the number of cycles each participant contributed. The number of participants contributing 1, 2 or 3 cycles with LH surges did not differ across the groups, χ2(4) < 1

Table 4

Methods used to determine cycle phase based on whether the LH surge was detected or not (backward count)

 

Phase of menstrual cycle

Menstrual

Follicular

Ovulatory

Luteal

Premenstrual

LH surge detected

Days 1, 2, 3 after day 1 of menstruation

Days 5, 6, 7 after day 1 of menstruation

Day of LH surge and next 2 days

Days 7,8,9 after LH surge

3 days prior to next menstrual bleed

Backward count

Days 1, 2, 3 after day 1 of menstruation

Days 5, 6, 7 after day 1 of menstruation

−16, −15, −14 days prior to menstrual bleed

−9, −8, −7 days prior to menstrual bleed

3 days prior to next menstrual bleed

Participants who detected at least one LH surge during the study were divided into three groups: women abstaining from sex for a particular cycle (n = 33); women engaging in sex with only male partner(s) for a particular cycle (n = 45); and women engaging in sex with only female partner(s) (n = 17) or both male and female partners (n = 5). Because all of the latter participants identified themselves as lesbians, we placed them in the lesbian group (n = 22). We analyzed data for these groups in two ways: (1) we used data from all cycles that had an LH surge and (2) we averaged a participant’s data across the cycles she contributed. Therefore, in the first analysis N was based on number of cycles with LH surges and in the second analysis N was based on number of participants contributing at least one cycle with an LH surge (Table 3).

We subsequently examined the data across three independent variables: detectible LH surge or not (2 levels); participants’ sexual behavior (abstinent, heterosexual, lesbian) and menstrual cycle phase (5 levels). Additionally, we tested whether results differed if we examined data from all cycles with LH surges or whether we averaged data across the participants’ cycles.

Experimental Design and Statistical Analyses

Our first analyses examined two methodological issues: (1) Did results differ if ovulation and menstrual cycle phases were constructed using the first day of menstruation and a known LH surge (LH surge) versus using the first day of menstruation and a backward count method (backward count)? (2) Did results differ if cycles with LH surges were used as units of analysis versus using participants’ averaged data across phases for cycles with LH surges? All analyses were conducted using SPSS 15.

We did not examine an effect using a mixed univariate analysis of variance (ANOVA) unless the multivariate effect was significant. To examine whether analyses using LH surge versus backward count methods (Table 4) differed, we used repeated measures ANOVAs. We examined LH surge versus backward count (Method) and the participants’ groups (abstinent, heterosexual or lesbian) across the dependent variables of libido, autosex, and allosex in 2 (Method) × 3 (Group) × 5 (Phases) ANOVAs. These analyses were based on a total of 185 cycles with an LH surge and 71 cycles using a backward count method. We followed up the ANOVAs with pair-wise comparisons and trend analyses.

To examine how sexual behavior and emotions varied across the menstrual cycle, we analyzed data from cycles with LH surges of participants in the abstinent, heterosexual, and lesbian groups and across averaged data per participant (second methodological question). In analyses using LH surges, we covaried out the number of cycles a participant contributed.

We also conducted a factor analysis using a principal components analysis with a varimax rotation across the 24 emotional variables (Table 2). We used participants’ averaged scores for each variable from cycles with LH surges. Using the three factors from the factor analysis, we computed a factor score for each menstrual cycle phase and used multiple regression analysis to examine whether the participants’ three emotional factor scores were related to their libido (using the abstinent, heterosexual, and lesbian groups) and whether libido and the emotional factor scores were related to allosexual (using the heterosexual and lesbian groups) and autosexual behavior scores. We conducted the multiple regression in two ways, using all cycles with an LH surge and cycles averaged across participants. For the multiple regressions, we collapsed the data across the cycle phases by averaging across the phases so that the relationship would be linear rather than curvilinear.

Results

Methodological Issue I: Did Results Differ When the LH Surge Was Used to Determine Menstrual Phase Versus a Backward Count Method?

To test the above research question, we conducted multivariate and univariate 2 (Method) × 3 (Group) × 5 (Phases) ANOVAs across the libido, autosex, and allosex dependent variables. There were no significant main effects for Method (LH surge vs. a backward count method for determining the menstrual cycle phase) whether we used a multivariate analysis F(3, 252) < 1 or the univariate analyses of the sexual dependent variables: libido F(1, 250) < 1, autosex F(1, 250) = 1.4, and allosex F(1, 172) < 1. Additionally, none of the two-way (Method × Groups; Method × Phases) or three-way (Method × Groups × Phases) interactions were significant.

The phase effect was significant across all three dependent variables: libido F(4, 1000) = 16.67, p < .0001; autosex F(4, 1000) = 5.05, p = .001; allosex F(4, 688) = 5.68, p < .0001. We discuss the phase effects in more detail below using data with phases based on the LH surge and cycles as the unit of analysis.

Methodological Issue II: When Multiple Menstrual Cycles Were Examined Across Participants, Did Results Differ When the Data Were Organized by Participants and Averaged Across Cycles Versus When Data Were Organized and Analyzed by Cycles?

To examine the above methodological issue, we conducted two separate series of analyses, the first using the participants’ averaged data and the other using cycle data. These results are shown in Table 5. The methods yielded almost identical means but the SDs, as a whole, were slightly higher when all cycles with LH surges were used in the analyses rather than averaged values across participants. We found one difference between the two methods of examining the data. When data were averaged across participants, there were no significant effects among the groups (abstinent, heterosexual, lesbian). However, when we examined the data using cycles as units of analysis, we found that lesbians (M = .18) engaged in significantly more autosexual behavior than heterosexual women (M = .08) but neither differed from the abstinent group (M = .15), F(1, 179) = 5.12; p = .007. Both methods yielded significant phase effects. Below, we report the F ratios based on the analyses using all cycles with LH surges.
Table 5

Methodological issue II, sexual behavior and cycle phases

 

N

Menstrual

Follicular

Ovulatory

Luteal

Premenstrual

M

SD

M

SD

M

SD

M

SD

M

SD

Libido

Cycle

185

1.78FOLP

.47

2.11MOLP

.48

2.20MFLP

.55

1.94 MFO

.44

2.00MFO

.48

Libido

Ave Ss

100

1.78

.40

2.10

.42

2.20

.44

1.94

.37

2.20

.42

Autosex

Cycle

185

.09FOL

.21

.14 M

.26

.16MLP

.29

.12 M

.25

.11O

.23

Autosex

Ave Ss

100

.10

.19

.15

.24

.16

.26

.12

.22

.10

.19

Allosex

Cycle

185

.16FOLP

.26

.32 M

.33

.30 M

.30

.24 M

.28

.30 M

.31

Allosex

Ave Ss

100

.17

.25

.30

.29

.29

.27

.24

.23

.28

.27

Note: Methodological issue II examined whether results differ when cycles versus participants were used as units of analysis. Comparison of means and SDs using data based on all cycles with a confirmed LH surge (Cycle) or based on averages across participants (Ave Ss). Data for autosexual and allosexual behavior were obtained by scoring whether or not a participant engaged in masturbation or in sex with another person (1 = engaged; 0 = did not engage) and then averaging the values across the 3 days of a phase. M, F, O, L, and P denote that the referent phase was significantly different from the menstrual, follicular, ovulatory, luteal, or premenstrual phases, respectively

Hypothesis I:

Did abstinent women exhibit the same patterns of sexuality in terms of libido and autosexual behavior as sexually active heterosexuals and lesbians?

Using menstrual cycle phase data based on a detected LH surge and cycles as the unit of analysis, we examined whether the groups differed in their sexual behavior across phases. First, the MANOVA across the three dependent variables was significant, F(6, 362) = 18.68, p < .0001. The above significance was predictable given the fact that the abstinent group by definition differed from the heterosexual and lesbian groups in their allosexual behavior. Therefore, we conducted another MANOVA analyzing data for libido and autosexual behavior, which also reached statistical significance, F(4, 362) = 2.84, p = .02. Using univariate analyses and all three groups, the phase effects for libido, F(4, 728) = 22.88, p < .001, and autosexual behavior, F(4, 728) = 5.43, p < .001, were significant. Women’s libido peaked during the ovulatory phase. There was a significant increase in libido from the menstrual to the follicular phase and, again, from the follicular to the ovulatory phase (Table 5). Libido significantly decreased from the ovulatory to the luteal phase, but the luteal and premenstrual phases did not differ from each other (Table 5); this was a significant cubic trend, F(1, 182) = 31.46; p < .001. Women’s autosexual behavior also peaked in the ovulatory phase. There was a significant increase in autosexual behavior from the menstrual to the follicular and ovulatory phases (Table 5). There was a significant decrease in autosexual behavior from the ovulatory phase to the luteal and premenstrual phases; this quadratic relationship was significant, F(1, 182) = 18.84; p < .001. The overall phase effect for allosexual behavior was significant, F(4, 500) = 5.5; p < .001. Sexually active heterosexuals and lesbians engaged in significantly more allosex during their follicular, ovulatory, luteal, and premenstrual phases than in their menstrual phase. Although the pair-wise comparisons did not reflect the subsequent decrease and increase in allosexual behavior, there was a significant cubic trend, F(1, 125) = 11.85; p < .001.

Hypothesis II:

How were the emotional factors related to sexual behavior and libido?

The factor analysis resulted in three factors with eigenvalues >1 and which accounted for 67% of the total variance in emotions. We used variables with loadings greater than .5 in our interpretations of the factors. Factor 1 loaded heavily on emotions involved in premenstrual syndrome, Factor 2 on positive affect, and Factor 3 on energetic/creative (Table 2).

The regression analysis predicting libido from the emotional factors was significant (Table 6 presents the analyses using cycles with LH surges and data averaged across participants). Positive affect was significantly positively related to libido in both analyses. The regression predicting autosexual behavior, using the abstinent and lesbian groups and all cycles with LH surges (the heterosexual group was not included because, using cycles, the heterosexuals displayed significantly less autosexual behavior than the lesbians), was also significant (Table 6) with two significant positive predictors, the energetic/creative factor and libido. When the regression analysis was conducted using the participants’ averaged data and including the heterosexuals, then the multiple R was still significant. The energetic/creative factor and libido maintained their significant positive relationships with autosexual behavior, and the positive affect factor also reached significance albeit in a negative direction (Table 6). The regression predicting allosex using the heterosexual and lesbian groups was also significant (Table 6) for both analyses. The significant predictors were the premenstrual and positive affect factors, which had negative relationships, and libido, which had a positive relationship with allosexual behavior.
Table 6

Multiple correlations between emotional factors and sexuality

Model:

Predicted

Predictors

R

Adjust R2

β weight

Libido

 

.35 (.34)

.12a (.11a)

 
 

Premenstrual factor

.06 (.04)

Positive affect factor

.39a (.37a)

Energetic/creative factor

−.04 (−.02)

Autosex

 

.26 (.22)

.06a (.04a)

 
 

Premenstrual factor

.08 (.07)

Positive affect factor

−.10 (−.16c)

Energetic/creative factor

.18a (.16b)

Libido

.19a (.16a)

Allosex

 

.29 (.29)

.08a (.07a)

 
 

Premenstrual factor

−.16a (−.15c)

Positive affect factor

−.12c (−.18c)

Energetic/creative factor

.02 (.05)

Libido

.28a (.29a)

Note: Non-italicized results reflect findings calculated using all cycles with an LH surge; italicized results in parentheses reflect findings calculated using participants’ averaged data

ap < .001, p < .01, p < .05

Discussion

Although much research has been conducted on how women’s sexual behavior varies across the menstrual cycle, the research findings are not always consistent (Table 1). Part of the reason for the inconsistency is that researchers employ different methods in their work. Our research examined two methodological issues involved in research on the menstrual cycle: whether ovulation was determined using the LH surge or through a backward count method and whether analyses were performed on “cycles” or “participants.” We found no significant differences in the results of analyses performed using cycles with known LH surges versus cycles based on backward counts. Our overall conclusion is that the two methods of determining menstrual cycle phase yielded similar results. However, we also conclude that statistical power might be compromised when researchers use backward counts because the overall differences between the means were lower in these analyses than in those using LH surges. Therefore, studies with null research results, like those reported in Bellis and Meyers (2005), who not only used a backward count method but also retrospective data (participants were required to remember the date they started to menstruate), should be examined with caution. Furthermore, past studies that based their menstrual cycle phases on the LH surge consistently reported differences in women’s sexual desires and behavior prior to ovulation (Table 1). For example, Bullivant et al. (2004) reported increased sexual activity, libido, and sexual fantasies around ovulation, Graham et al. (2000) reported higher levels of sexual arousal in the periovulatory period, Haselton, Mortezaie, Pillsworth, Bleske-Rechek, and Frederick (2007) reported that women dressed better during their fertile phase, and our study found that women’s libido and sexual behavior increased prior to ovulation.

Similarly, we only found two significant differences when we analyzed data using cycles with LH surges versus data averaged across the participants’ scores. In the ANOVAs, we found a difference in autosexual behavior between the groups using cycles as data points but not when we averaged the participants’ data. We also found that positive affect was significantly related to autosex using averaged data but it failed to reach significance when we used the cycle data. All other findings were extremely similar across the two ways of handling the data.

Our results provide strong support for the hypothesis that, although women are receptive to sexual activity across their menstrual cycles, their sexual behavior and libido cluster at mid-cycle. First, we found that abstinent women reported increased libido and autosexual behavior at mid-cycle. The sexual behavior of abstinent women is less likely to be influenced by the sexual desires of others and therefore more likely to reflect the natural fluctuations in sexuality of women. Additionally, like previous researchers (Bullivant et al., 2004; Wilcox et al., 2004), we found that women, regardless of their sexual orientation, engaged in significantly more allosexual activity during their follicular and ovulatory phases with a second smaller peak pre-menstrually than at other times during the cycle. Therefore, it can be concluded that, although women exhibit continuous receptivity throughout their cycles, they have peaks in their sexual behavior before and during ovulation and again pre-menstrually.

We also found that emotions were related to women’s sexuality. Libido was related to positive affect. However, libido and positive affect varied in their relationship to allosexual behavior. Libido, after the effects of positive affect were partialled out, was positively related to allosexual behavior. However, positive affect, after the effects of libido were partialled out, was negatively related to allosexual behavior. These findings support theories such as Diamond’s (2003) that at least two systems are involved in women’s sexual behavior, one involving sexual arousal and related to mating behavior, and the other involving affection and related to attachment behavior. The fact that our results predict that sometimes libido will be positively related to sexual behavior while, at the same time, attraction is negatively related to sexual behavior helps explain the emotions and motivations behind findings like those of Haselton and Gangestad (2006). They found that women who were pair-bonded with men with low sexual attractiveness, relative to investment attractiveness, were more likely to report being attracted to other men when they were ovulatory (libido and affection acting independently) but were not attracted to other men outside of ovulation (libido and affection acting together). The peak in sexual behavior prior to menstruation is probably also related to pair-bonding.

Indeed, the physiological and brain scan literature supports the idea of two separate systems, one involving libido and the other involving pair-bond formation. Research indicates that reproductive hormones, like androgen and estrogen, are associated with libido (Fisher, Aron, Mashek, Li, & Brown, 2002), and fMRI studies find that the ventral striatum and hypothalamus are areas associated with intense sexual arousal (Walter et al., 2008). In contrast, pair-bonding most likely involves oxytocin and vasopressin (Fisher, Aron, & Brown, 2006) and the activation of the basal ganglia (Aron et al., 2005; Phan, Wager, Taylor, & Liberzon, 2002) and the cingulate cortex (Bartels & Zeki, 2004).

Autosexual behavior was positively related to both libido and the energetic/creative emotional factor and negatively related to positive affect. Both men and women reported engaging in more autosex after the sexual revolution in the 1960s (Dekker & Schmidt, 2002; Kontula & Haavio-Mannila, 2003). More research on autosexual behavior and its relationship to creativity and positive affect among individuals of varying sexual orientations needs to be conducted.

Acknowledgements

We thank Marci Arizumi, Louona Larkspur, Ariel Marsh, Lynn Morrison, Nicola Nicolaisen, Tiffany Freitas, Michelle McNamee, Sarah Asakawa, Aizelle Boado, and Lisa Oliver in helping to recruit the participants, and in data collection, organization, and analysis. This research was supported by NIH grant #S06-GM0873-33.

Copyright information

© Springer Science+Business Media, LLC 2010