A Meta-Analysis of Menstrual Cycle Variation in Olfactory Sensitivity

Lenka Martinec Nováková, Jan Havlíček, S. Craig Roberts Olfactory processing and odor specificity

Anthropological Review • Vol. 77 (3), 331–345 (2014)

ANTHROPOLOGICAL REVIEW Available online at: www.degruyter.com Journal homepage: www.ptantropologiczne.pl

Olfactory processing and odor specificity: a meta-analysis of menstrual cycle variation in olfactory sensitivity

Lenka Martinec Nováková1, 2, Jan Havlíček2, S. Craig Roberts3

1Department of Anthropology, Faculty of Humanities, Charles University, Prague, Czech Republic 2Department of Zoology, Faculty of Science, Charles University, Prague, Czech Republic 3School of Natural Sciences, University of Stirling, Stirling, UK

Abstract: Cycle-correlated variation in olfactory threshold, with women becoming more sensitive to odors mid-cycle, is somewhat supported by the literature but the evidence is not entirely consistent, with several studies finding no, or mixed, effects. It has been argued that cyclic shifts in olfactory threshold might be limited to odors relevant to the mating context. We aimed to test whether the evidence currently available points in the direction of odor-specific or, rather, general changes in olfactory sensitivity and, if the former is the case, to what group of odorants in particular. We carried out a meta-analysis of relevant studies which together used a variety of different odorants, including some found in food, body odor, and some that occur in neither of these. First we tested whether there appears to be an overall effect when all studies are included. Next, we hypothesised that if cyclic changes in olfactory processing are odor-specific and tuned to biologically relevant odors, we should find changes in detection thresholds only for odorants found in body odor, or for those that are perceptually similar to it. In contrast, if threshold patterns are linked to more general fluctuations in odor processing across the cycle, we would not expect changes in relation to any particular odorant group. The results support the view that there is significant cycle-correlated variation. Thresholds were in general significantly lower in the fertile than the non-fertile phases, with effect sizes consistently in this direction. This same conclusion applied to both ‘food’ and ‘musky’ odorants, despite their different evolutionary significance, and to the androgen steroids (androstadienone, androstenone, and androsterone), but could not be applied to phenyl-ethyl alcohol. The results indicate that olfactory sensitivity may be a non-adaptive by-product of the general physiological fluctuations or differences in neural processing experienced across the cycle to a broad spectrum of odorants, rather than being specifically selected for mate choice-related odors.

Key words: menstrual cycle, mate choice, odor threshold, human oestrus, olfaction

Original Article: Received: 16 October 2014; Accepted for publication: 5 November 2014 DOI: 10.2478/anre-2014-0024 © 2014 Polish Anthropological Society 332 Lenka Martinec Nováková, Jan Havlíček, S. Craig Roberts

Introduction – to three – day intervals. In general, he found an increase in sensitivity (decrease In recent years, researchers interested in in odor threshold) in all cycles follow- the evolutionary underpinnings of hu- ing menstruation but, at the same time, man behaviour have recorded in women the individual cycles varied considerably a variety of behavioural and perceptual in the timing and magnitude of this in- changes associated with menstrual cycle crease. For instance, while in some cases phase (review in Gangestad and Thorn- sensitivity peaked shortly after menstru- hill 2008). The cyclic shifts in sensory ation, in others it took longer. Moreover, perception seem to cut across sensory he also noted a second peak in sensitivity modalities (Doty et al. 1982; Dye 1992; during the late luteal phase in three cas- Kuga et al. 1999) and may be linked to es. However, no such changes in sensitiv- general physiological fluctuations or ity to safrole, guaiacol, amyl salicylate, or differences in neural processing experi- pyridine were noted in a smaller number enced across the menstrual cycle. Doty of tested women, except for a mixture of and Cameron (2009) reviewed the ways cholesterol and testosterone. Hence, it through which such a general effect would seem that menstrual cycle-relat- would be possible. They put forward ed changes in olfactory sensitivity would the idea that cyclic changes may reflect pertain exclusively to musky-smelling fluctuations in hormones other than the odorants, suggesting that they might be primary ovarian steroids, suggesting the of special ecological significance to hu- potential involvement of the corticotrop- mans. ic releasing hormone (CRH) – adren- In contrast to the findings ofLe Magnen ocorticotropic hormone (ACTH) – ad- (1952), Mair et al. (1978), who employed renal axis. Alternatively, they propose a signal detection paradigm, reported that these shifts may be controlled by women’s better performance during ovu- central nervous system centres or net- lation compared to menstruation for pen- works such as those that control various tadecalactone (Exaltolide) as well as for other rhythms (e.g. Rusak and Zucker coumarin and cinnamyl butyrate (but not 1979), for instance specific or non-spe- amyl acetate). Other researchers have also cific effects of neurotransmitters or other noted significant cyclic variation in sensi- neuroactive substances that fluctuate on tivity not only to pentadecalactone (Good a monthly basis. et al. 1976; Vierling and Rock 1967), but Of the various sensory modalities, also to other musk-smelling odorants, among the most hotly debated is cyclic such as 5α-androst-16-en-3-one (andros- variation in olfactory thresholds for both tenone; e.g. Renfro and Hoffmann 2013; social (i.e. human body odor-related) Sueda et al. 2003) or musk-ketone (Caru- and non-social odors (review in Doty and so et al. 2001), as well as those that are Cameron 2009). The suggestion that ol- not considered to be of biological rele- factory detection thresholds vary across vance to humans, such as ammonia, anise the menstrual cycle was first reported essence, citral, eugenol (clove odor), fur- over 50 years ago (Le Magnen 1952). In fural, or pyridine (e.g. Caruso et al. 2001; one study, he measured thresholds for Doty et al. 1981). Navarrete-Palacios et al. pentadecalactone (Exaltolide) across ten (2003) also reported menstrual cycle-re- menstrual cycles of five women with two lated fluctuations in amyl acetate thresh- Olfactory processing and odor specificity 333 olds, contrary to the findings of Mair et the present topic, measures of odor de- al. (1978). On the other hand, some re- tection performance, body temperature, searchers have not found any significant heart rate, plasma FSH, LH, progester- changes in sensitivity to androstenone, one, and total estrogens were taken twice citral, pure lemon extract, n-butanol, ni- a day. Most notably, they observed a pos- cotine, pure peppermint extract, pheny- itive correlation between body tempera- lethyl alcohol, or rose water (Hummel et ture and olfactory sensitivity. al. 1991; McNeil et al. 2013; Pause et al. Thus, cycle-correlated variation in 1996; Renfro and Hoffmann 2013). olfactory threshold, with women becom- A major contribution to the debate ing more sensitive to odors mid-cycle, is comes from Doty et al. (1981), who, somewhat supported by the literature but employing a signal detection paradigm, the evidence is not entirely consistent, tested odor detection performance for with several studies finding no, or mixed, furfural every other day across 2 con- effects. A major question that arises is secutive menstrual cycles of both oral whether the evidence currently available contraceptive non-users and users. Fur- points in the direction of odor-specific thermore, concomitant measures of body or, rather, general changes in olfactory temperature, blood pressure, heart rate, sensitivity. Furthermore, if the former nasal airflow, and respiration rate were is the case, what group(s) of odorants also taken as well as levels of estradiol, in particular are affected? For example, estrone, follicle stimulating hormone might increased mid-cycle sensitivity be (FSH), luteinizing hormone (LH), pro- specific to musky-smelling odorants and gesterone, and testosterone. To handle the androstenes, which may possess bi- problems related to averaging data across ological relevance through an influence menstrual cycles of different lengths, the on mate choice? Consistent with this authors used a procedure devised by Doty idea, Lundström et al. (2005) found that (1979). They found three peaks in aver- menstrual cycle stage influenced sensi- age olfactory sensitivity: in the middle of tivity to a biologically-relevant odorant the cycle (around ovulation), in the mid- androsta-4,16,-dien-3-one (androstadi- dle of the luteal phase, and in the second enone), but not to phenylethyl alcohol half of the menstrual phase. Remarkably, (PEA, rose odor). Similarly, Renfro and these fluctuations were observed in both Hoffmann (2013) reported cyclic varia- oral contraceptive non-users and users, tion in sensitivity to androstenone and suggesting that there may not be a caus- 3α-hydroxy-5α-androstan-17-one (anal relationship between levels of gonadal drosterone), but not to pure lemon or hormones and hypophyseal gonadotro- peppermint extract, or rose water. pins and olfactory thresholds but, rather, Here we aimed to further test whe a correlation. ther the menstrual cycle influences odor In a subsequent study, Doty et al. sensitivity and, if so, whether the ex- (1982) aimed to test whether similar tent of such changes might be predict- shifts in olfactory sensitivity could be ed by the biological significance of the found for phenylethyl alcohol in one oral odorants used. To do this, we carried contraceptive user across two menstru- out a meta-analysis of relevant studies al cycles, along with shifts in auditory which together used a variety of different thresholds. Of the variables relevant to odorants, including some found in food, 334 Lenka Martinec Nováková, Jan Havlíček, S. Craig Roberts some found in body odor, and some that Sueda et al. (2003), which was translated occur in neither of these. First we tested from Japanese. Data from a total of thir- whether there appears to be an overall ef- teen independent studies were included in fect when all studies are included. Next, the analysis (274 females). In two studies we hypothesised that if cyclic changes that used both a within-subject and a be- in olfactory processing are odor-specific tween-subject design (Navarrete-Palacios and tuned to biologically relevant odors, et al. 2003; Renfro and Hoffmann 2013), we should find changes in detection we chose to use data from the within-sub- thresholds only for odorants found in ject design because this better controls body odor, or for those that are perceptu- for possible confounding variables related ally similar to it. In contrast, if threshold to interindividual differences in olfactory patterns are linked to more general fluc- processing. tuations in odor processing across the A summary of the studies included in cycle (cf. Becker et al. 1982), we would the meta-analyses, giving details on the not expect changes in relation to any par- odorants used, concentrations, perceived ticular odorant group. quality, volatility (as defined by vapor pressure) sample sizes, the techniques Materials and Methods used to estimate olfactory thresholds, and the frequency and timing of meas- We carried out a literature search (using urements taken, is shown in Table 1. Al- the search engine Google Scholar) with though there is some variability in tech- the two search terms “olfactory thresh- nique, more than half of the studies used old” and “menstrual cycle”, then sifted some form of staircase forced – choice through the results for relevant and suit- technique without feedback, now a stand- able studies. The requirements for in- ard practice in olfactory psychophysical clusion into the meta-analysis were that testing (Keller and Vosshall 2004). a study had measured olfactory thresh- We conducted two initial fixed-effect olds with the menstrual cycle phase as meta-analyses to determine whether: (1) a time variable, and reported (a) an effect there is variation in thresholds across statistic or (b) an effect for which a statis- the cycle (including all studies), and (2) tic could be calculated from the findings differences occurred specifically between (based on this criterion, we have not in- fertile and non-fertile cycle phases. cluded the study by Grillo et al. (2001); Where appropriate, we here condensed however, this study reported lower perio- the effects of several odorants used in vulatory thresholds for a number of odor- single studies by averaging the respec- ants consistent with the pattern of re- tive effect sizes (Schmidt 1990). Next, sults obtained in this meta-analysis, thus we compared whether changes occurred exclusion of this study should not have across the cycle in studies that included: altered our conclusions). For the same (3) odors associated with foods (namely reason, we could only include results for pure anise essence, amyl acetate, citral, androstenone and androsterone from the eugenol, cinnamyl butyrate, coumarin, study by Renfro and Hoffmann (2013). vanillin, and n-butanol), (4a) odors that All the studies have been published in smell musky (namely androstadienone, peer-reviewed scientific journals. All were androstenone, musk ketone, pentadeca- reported in English with the exception of lactone) and (4b) musky odors excluding Olfactory processing and odor specificity 335 - - Schedule 19 subsequent days within 1 month (ex cluding weekends) Data divided for statistical analysis into intervals:7 four-day menstrual, early/late ovulatory, follicular, mid-, and late early, luteal 3 measures: follicular (day 5–8), periovu lar (13–16), luteal (18–23) -

. Done technique Threshold (Whissell-Buechy 3AFC and Amoore 1973) twice at each dilution step. Threshold: highest con secutive dilution correctly chosen. No feedback. Rhinomanometry and Threshold: olfactometry. smallest volume of air at which odor is perceived. - - N 22 (total) 9 complete data (3 + 6 hor monal contra ceptive non- + users)† 60 - (Be (volatile) –5 3 –2 –7 10 10 10 10 * * * * Vapor pressure Vapor [mm Hg] at 25°C 2.61 doukian 2014) 7.51 9.13 5.84 20.8 (volatile) - (Bedoukian Perceived quality Perceived Extra- ordinarily persis tent, musk-like odor Strong, lemon-like odor warm, light, Fine, odor, sweet musky closest to natural musks 1986:301–303) Characteristic, penetrating odor - Concentration(s) Binary dilution steps, each step being half the concentration of the step before. Start ing from step 0 for the saturated solution (1.1 ppm, w/v), steps 4–16 were tested. Liquid diluted to 22% in distilled water Undiluted Undiluted Undiluted Undiluted Pure liquid substance

(subset) Odorant(s) Pentadecalactone Pentadecalactone (Exaltolide) (musky) Ammonia Pure anise essence (food) Citral (food) Pure clove essence (food) Musk-ketone (musky) Pyridine Study Amoore et Caruso et (1) al. (1975) (2) al. (2001) Table 1. Summary of studies included in the meta-analysis Table perceived used, concentration(s) meta-analysis, the of purposes the for included been have they which in subset the used, odorants of details gives Table quality irrespective of concentration, volatility as defined by vapor pressure (if known or applicable), sample sizes (N), olfactory threshold techniques employed and frequency and timing of measurements taken. The quality descriptions have been taken from Burdock (2009) and vapor pressure data from the publicly accessible PubChem Project Database (NCBI 2014), unless stated otherwise. Higher values of vapor pressure indicate a tendency of 2014). Morris and (Vo volatile considered are 25°C at Hg mm 0.1 than greater pressure a vapor with odorants and readily more vaporize to a substance †Number of participants and analyses actually used in the meta-analysis. 336 Lenka Martinec Nováková, Jan Havlíček, S. Craig Roberts - - - - Schedule Measures taken re Measures taken peatedly every other day across 2 consecu tive menstrual cycles. 2 measures: mid-cy cle, menses Cycle split into 5 phases (two preovu two ovulatory, latory, postovulatory) 2 measures: fertile, non-fertile - - - -

technique Threshold , 3AFC, concen Signal detection paradigm, 350 trials (175 blank). On a given trial diluent (blank) presented first, followed ei ther by diluent or odorant. confidence in whether Rate an odorant presented. given. Feedback Signal detection para digm:100 trials (50 blank); report if stimulus is odorant or blank. Feedback given. staircase Ascending and Beauchamp (Wysocki 1984) trations increase until 3 correctly discerned in succession, lowest is noted. Threshold: mean Repeated. of 2 lowest concentrations. No feedback. staircase, 3AFC,Ascending concentrations increase until 2 correctly discerned in suc cession, triggering a reversal. 7 reversals. Threshold: mean of concentrations at the last 4 reversals. No feedback. (Hummel et al. 1997) 9 3 N 14 22 –2 –2 10 * 10 * Vapor pressure Vapor [mm Hg] at 25°C 2.21 (volatile) 3.8 8.68 See above - - - - (Jacob et (Bremner et al. Perceived quality Perceived Quality rarely rec ognised at concen trations employed here (below recog nition threshold) (Arctander 1969) See above urinous; Sweaty, sweet, floral; odor less 2003; Gower et al. and 1998; Wysocki Beauchamp 1984) Characteristic rose- odor like Putrid, sweaty, urinous al. 2006) See above and –6 - 10 * v/v. –5 *10 Concentration(s) Custom-prepared for each individual and detectable 55% – 75% of the time during pilot trials. Range between 7.23 7.35 Dilutions prepared successively by adding 40 ml of propylene glycol (double-distilled water for nicotine) to 40 ml of the preceding solution. Maximum concentrations: 1.44 mmol/l (andros tenone), 4130 mmol/l (PEA), 151 mmol/l (nicotine) Geometric series dilutions in propylene glycol ranging from 16) (dilution μM 0.091 to 3000 μM (dilution 1) Geometric series dilutions in propylene glycol ranging from 16.3 μM (dilution 16) to 0.54 M (dilution 1) -

(subset) Odorant(s) Furfural Pentadecalactone (Exaltolide) (musky) Androstenone (musky) Nicotine Phenylethyl Alcohol (PEA) Androstadienone (musky) Phenylethyl Alco hol (PEA) Study Doty et al. Good et al. Hummel et Lundström (3) (1981) (4) (1976) (5) al. (1991) (6) et al. (2005) Table 1. cont. Table Olfactory processing and odor specificity 337 - - - - - Schedule 2 measures: ovula tion, menses Exaltolide, indi For viduals tested on up to 3 successive days around ovulation un til a clear increase in sensitivity observed. 3 measures: early follicular – menstru ation (days 1–5), late follicular – ovulation (days 11–14), mid-lu teal (days 21–26); tailored to individual cycle length 4 measures: menses, ovulatory, follicular, luteal 3 measures: follicular, luteal ovulatory, - -

technique Threshold Signal detection paradigm: 60 trials (30 blank), report if stimulus was odorant or given. blank. Feedback staircase, 3AFC, Ascending concentrations increase until 2 correctly discerned in succession, triggering a reversal. 7 reversals. Threshold: mean of concentrations at the last 4 reversals. No feedback. (Hummel et al. 1997) Staircase, 4AFC, 2 odor No feedback. ants, 2 blanks. Staircase, 3AFC. Threshold: lowest of 2 consecutive concentrations for which 4 successive correct iden tifications occurred. No feedback. - 5 N 12 17 15 (all with Exal tolide; 6 with AA, 6 with coumarin and CB) - (Be –5 –4 10 –3 * 10 * Vapor pressure Vapor 10 * [mm Hg] at 25°C 3.5 (volatile) 1 9.8 2.61 doukian 2014) 7.0 (volatile) See above See above - Perceived quality Perceived Fruity, banana, Fruity, sweet, fragrant, powerful odor slightly Fruity, floral odor Sweet, fresh, hay- odor similar to like vanilla seeds See above or ba Fusel- sweet nana-like odor See above See above - (Hummel et Concentration(s) Dilution in dibutyl phthalate, 0.038 mM Dilution in dibutyl phthalate, 0.78 mM Dilution in dibutyl phthalate, 0.96 mM Dilution in dibutyl phthalate, 1.6 mM 16 geometric series dilutions al. 1997) Series of 8 half log concentrations, ranging from −log 9.5 to −log 6.0 Series of 14 half deci mal log steps (diluted in diethyl phthalate), maximum: 1:2 (v/v) dilution, minimum: 1:6 300 000 (v/v).

(subset) Odorant(s) Amyl acetate (AA) (food) Cinnamyl butyrate (CB) (food) Coumarin (food) Pentadecalactone (Exaltolide) (musky) N-butanol (food) Amyl acetate (food) Citral (food) - Study Pause et Pause Mair et al. McNeil et Navarre (7) (1978) (8) al. (2013) (9) et te-Palacios al. (2003) (10) al. (1996) Table 1. cont. Table 338 Lenka Martinec Nováková, Jan Havlíček, S. Craig Roberts - - Schedule 2 measures: perio luteal (ap vulatory, proximately one week after first visit) 4 measures: menses, ovulatory, follicular, luteal 2–4 times, arbitrary days in cycle - - -

technique Threshold Staircase, 3 AFC. Con centrations increase until correctly identified, then the higher one presented.next If again correct, reversed. If again correct at reversal, as threshold. If taken incorrect at reversal, increase until correctly identified. Threshold: average Reversed. consecutive correct identifi cations and reversal. Staircase, 3AFC (3-way forced choice; no feedback) with 5 ascending 9 flasks concentrations (5–1), 4 after smelling each blanks; report if stimulus was odorant or blank. Thresh old: individual scored 1–5 according to concentration detected. N 16 19 73 - - (Ragha (Be –8 –4 –5 10 10 10 * * * Vapor pressure Vapor [mm Hg] at 25°C See above 2.79 va 2011) See above 1.18 2.61 doukian 2014) - (De Ro Perceived quality Perceived See above (grapey), Fruity camphoraceous or aromatic, musky urine-like, or odorless vira 2008:542) See above See above Characteristic, vanilla-like creamy, odor See above - -

–11 (1), , 10 –3 –9 M to 10 mM μ , 10 –7 0.1 , 10 Concentration(s) –5 7 geometric series di lutions in safflower oil (ratio 1:10), ranging from Stock solution of 1 mg of Exaltolide in 10 ml absolute ethanol dilut ed with distilled water to yield 5 dilutions containing 10 10 mg of Exaltolide per ml of water

(subset) Odorant(s) Androstenone Androsterone Androstenone (musky) Phenylethyl Alcohol (food) Vanillin Pentadecalactone (Exaltolide) (musky) Study Renfro Renfro Sueda et Vierling (11) and Hoffmann (2013) (12) al. (2003) (13) and Rock (1967) Table 1. cont. Table Olfactory processing and odor specificity 339 the androstenes, i.e. musky odors that heterogeneity by calculating values of Q, are not found in human body odor. Fi- which follows a chi-square distribution nally, we assessed the specific effects on with k-1 degrees of freedom, where k is thresholds for (5) PEA, and (6) the an- the number of studies (Wolf 1986), under drostenes, since these were each used in the null hypothesis that all studies share several studies. a common effect size (Hedges and Olkin We calculated effect sizes (r) for in- 1985). Rosenthal’s failsafe N, the number dividual studies using transformations of filed, new or unretrieved studies show- from Rosenthal and DiMatteo (2001). ing null effects which would be needed In cases where p values were reported to produce an overall effect (Rosenthal in the absence of a test statistic, each p 1991), was calculated as N= (ΣZ)² /2.706 value was converted to its associated – k (where k=number of studies). one-tailed standard normal deviate Z, and then to an effect size calculated us- Results ing the formula r=√Z2/N (Rosenthal and DiMatteo 2001). Where a result was re- Results of the individual analyses are ported as not statistically significant, but shown in Table 2. Analysis 1 shows that it was not possible to calculate an effect detection thresholds are subject to varia- size, we set an effect size at zero, which tion in menstrual cycle phase, when com- is considered a conservative procedure. bining all studies and across a wide range For analyses 1–4, where single studies of odorants used in testing. Analysis 2 had tested more than one odorant, ef- confirms that this variation holds when fect sizes were averaged within-study. restricting studies to those that explic- Effect sizes for each study were weighted itly compare fertile and non-fertile cycle according to N-3 (Rosenthal and DiMat- phases. Sensitivity to odorants is higher teo 2001). We estimated between-study during the fertile phase than non-fertile

0.1

0.2

fect Size 0.3

0.4

0.5

Standard Error of Ef 0.6

0.7

–1 –0.5 0 0.5 1 1.5 Effect Size Fig. 1. Funnel plot for Analysis 1. Straight lines (inverted funnel) define a region within which 95% of points might lie in the absence of both heterogeneity and publication bias 340 Lenka Martinec Nováková, Jan Havlíček, S. Craig Roberts

Table 2. Summary of results of the individual meta-analyses Mean ES ± SE denotes mean effect size estimate ± standard error, 95% CI is the 95% confidence interval for the mean ES, Z and p are the associated mean Fisher Z and its p-value, Q and critical Q refer to estimated between-study heterogeneity and its referential critical value, which should not be exceeded if variability across effect sizes is not to exceed what would be expected based on sampling error Mean Meta-analysis Studies N 95% CI Z p Q Critical Q ES±SE 1) Threshold change 1–13 13 0.316±0.07 0.188/0.444 4.84 <0.001 7.95 26.21 2) Fertile/non-fertile 2, 4–12 10 0.301±0.08 0.142/0.459 3.72 <0.001 6.45 21.67

3) Food 2, 7–10, 12 6 0.290±0.10 0.098/0.482 2.96 <0.01 4.28 15.09

4a) Musky including 1, 2, 4–7, 9 0.511±0.07 0.370/0.652 7.12 <0.001 18.64 20.09 androstenes 11–13 4b) Musky excluding 1, 2, 4, 7, 13 5 0.582±0.09 0.414±0.750 6.79 <0.001 14.69 13.28 androstenes 5) PEA 5, 6, 12 3 0.041±0.15–0.248/0.330 0.28 <0.390 .11 5.99

6) Androstenes 5, 6, 11, 12 4 0.369±0.13 0.112/0.627 2.81 <.01 1.98 11.35 phases. We calculated Rosenthal’s fail- what might be expected based on sam- safe N for these two analyses as 111.85 pling error. An exception was analysis 4b and 54.59, and by applying Rosenthal’s (musky odors without the androstenes), criterion of 5k+10, we have toleranc- where heterogeneity in effect sizes across es of 75 and 60. This suggests that the studies suggested the presence of addi- findings of the first analysis are robust tional moderator variables. to possible file-drawer bias, while the second may not be. Fig. 1 shows a fun- Discussion nel plot for Analysis 1, which, however, suggests publication bias towards stud- We employed a meta-analytic approach ies with smaller standard errors (great- to summarise the research literature er sample sizes) and particularly those addressing the effects of the menstru- showing greater effect sizes. al cycle on human olfactory threshold. Analyses 3 and 4 concerned thresh- The results, taking into account the olds for odors found in foodstuffs and thirteen studies investigating olfactory musky odors, respectively. In both cases, threshold across the menstrual cycle, thresholds are found to be lower dur- support the view that there is significant ing the fertile than non-fertile phases. cycle-correlated variation. Thresholds A similar result was found for studies are in general significantly lower in the using androstenes (analysis 6), but there fertile than the non-fertile phases, with appeared to be no evidence for cycle-cor- effect sizes consistently in this direction related variation in sensitivity to PEA (even for the one involving PEA). This (analysis 5). same conclusion applies to both ‘food’ Homogeneity analyses showed that and ‘musky’ odorants, despite their dif- values of Q generally fell below the crit- ferent evolutionary significance, and to ical value, indicating that variability ob- the androgen steroids (androstadienone, served across effect sizes did not exceed androstenone, and androsterone) which Olfactory processing and odor specificity 341 are putative human semiochemicals in- doubtedly lies in the above-mentioned fluencing behavior (Saxton et al. 2008), diversity of olfactory measurement tech- but could not be applied to PEA, at least niques employed (Doty et al. 1986; Doty based on existing evidence. et al. 1995). The manner in which the Although, on the whole, variabili- individual menstrual cycle phases are ty observed across effect sizes did not defined and data combined across- cy exceed what might be expected based cle phases also plays a major role (Doty on sampling error for most analyses, 1979). Namely, in few studies the cir- an analysis of musky odors that are not culating hormone levels have actually found in human body odor (4b) did re- been measured, which is more accurate veal heterogeneity in effect sizes across in comparison to, for example, count- studies that suggests the presence of ad- ing methods. Further, in some studies ditional moderator variables. Since it did the olfactory measures might have been not pertain to musky odorants including taken too sparsely to detect any patterns the androstenes (4a) or the androstenes of change. Differences in sensitivity in alone (6), this heterogeneity can be as- either nostril may also account for some cribed to various discrepancies in meth- of the variability observed (Purdon et al. odology employed to test pentadecalac- 2001). Also, Sueda et al. (2003) discuss tone (Exaltolide) and/or musk-ketone the exclusion of individuals who have thresholds. Potential sources include, a particularly poor sense of smell. They for instance, the use of different types of initially found no significant differences psychophysical paradigms (staircase vs. in olfactory threshold for androstenone, signal detection procedures). Most im- but were able to detect cyclic changes portantly, considerable variation stems only when they took into account indi- from the use of single ascending or de- vidual variation in absolute threshold scending series (e.g. Amoore et al. 1975; levels. It is conceivable that had other Vierling and Rock 1967), which is highly studies applied a similar criterion to re- unreliable as the thresholds tend to fluc- strict analyses to normosmic subjects tuate rather wildly on repeated measures within a sample, more pronounced or (e.g. Stevens et al. 1988). Further, large consistent effects may have been ob- step sizes in odorant concentration could served. A further explanation for the have obscured small, but consistent, apparent lack of congruency in results changes. Manipulations that are clearly relates to the absence of cyclic variation inconsistent with best practice include in the studies using the odorant PEA. threshold testing on up to 3 successive These null findings cannot be simply ex- days around ovulation until a clear in- plained by methodological inadequacies, crease in sensitivity could be observed since in two of the studies using PEA the (Mair et al. 1978) and pooling the data of researchers did find significant effects for normally cycling women and those who other odorants (androstadienone: Lund- used hormonal contraceptives (Amoore ström et al. 2005; androstenone: Sueda et et al. 1975). al. 2003). It could be then suggested that If cycle-related change in odor pro- PEA is among the more volatile of the cessing appears domain-general, why odorants that have been tested to date. then is there so much variability in find- This, however, is clearly not the case (see ings across studies? Some variation un- Table 1) and the odorant cannot even be 342 Lenka Martinec Nováková, Jan Havlíček, S. Craig Roberts simply characterized as volatile based on Doty and Cameron (2009) (for details the criterion outlined by Vo and Morris see Introduction). Another possibility is (2014). Mair et al. (1978) suggested that that, as already noted above, the cyclic the patterns of threshold changes across changes in hydration of nasal mucus may different odorants might be due to differ- influence chemicals depending on their ent volatility. Volatility is caused by the volatility (Mair et al. 1978). However, evaporation or rapid sublimation of an it should be noted that the individual odorant and is proportional to the vapor mechanisms may not be mutually exclu- pressure of the substance and inverse- sive and may in fact work in concert. ly proportional to its molecular weight. Researchers in this area have not ne- Mair et al. (1978) noted that involatile glected the possibility that changes in esters (e.g. pentadecalactone, coumarin, odor perception of biologically relevant and cinnamyl butyrate) might be strong- odors across the cycle could potentially ly retarded by nasal mucus (which alters mediate changes in preferences. For in- across the cycle), while the volatile ester stance, Gangestad and Thornhill (1998) amyl acetate might escape this effect and explicitly discussed this possibility in thus exhibit relatively stable threshold their article reporting changes in wom- levels. However, recent work by Vo and en’s preferences for the body odor of Morris (2014) makes such a clear-cut symmetrical men. They assessed wom- distinction problematic. While odor- en’s ratings of intensity (not to be equat- ants with a vapor pressure greater than ed with threshold measurements) of 0.1 mm Hg at 25°C are volatile, those men’s body odor as well as of its pleas- with smaller values may fall under all antness and sexiness. Mean intensity rat- three categories: non-volatile, volatile, ings did tend to be greater among fertile or semi-volatile, making the suggestion than non-fertile women or women us- raised by Mair et al. (1978) more difficult ing hormonal birth control (though this to test than previously assumed. contrast was only marginally significant The lack of odor specificity in relation if a one-tailed test is applied), a pattern to their biological relevance might seem expected if fertile women are more sen- surprising in light of studies that have sitive to odors and consistent with this demonstrated cyclic changes in prefer- meta-analysis. However, intensity scores ences for odors potentially involved in tended to be negatively associated with mate choice. The results indicate that ol- pleasantness scores, the correlation be- factory sensitivity may be linked to gen- tween women’s mean intensity ratings eral physiological fluctuations or differ- and their preference for symmetry was ences in neural processing experienced near-zero and controlling for mean inten- across the cycle to a broad spectrum of sity ratings left the association between odorants in the same way as they appear fertility risk and preference for the scent to be for visual, gustatory, and other stim- of symmetry completely unchanged. uli (Becker et al. 1982; Dye 1992; Kuga et Havlicek et al. (2005) found a preference al. 1999). Our results are therefore con- for odors of dominant men when rated sistent with the suggestion that changes by women in their fertile phase, but not in sensitivity may be linked to a general by women in the non-fertile phase. Simi- effect in odor processing, whose mecha- larly to the above-reviewed study, the as- nism has been proposed in a review by sociation was not mediated by intensity Olfactory processing and odor specificity 343 ratings. To sum up, women’s olfactory of data, in the writing of the report, or preferences for cues of male quality (e.g. in the decision to submit the article for symmetry) shown in the fertile period publication. The authors declare that the cannot simply be attributed to cyclical research was conducted in the absence of changes in perceived intensity. Converse- any commercial or financial relationships ly, Hummel et al. (1991) reported men- that could be construed as a potential strual cycle-related variation in hedonic conflict of interest. ratings for the odorous steroid androstenone but not for other odors, while Authors’ Contributions thresholds did not vary. The results of the present meta-anal- Conceived and designed the meta-analy- ysis of thirteen studies that investigated sis: SCR, JH, and LMN. Conducted the olfactory thresholds across the menstru- search for references: SCR and LMN. al cycle point in the direction of signif- Analyzed the data: SCR and LMN. Wrote icant cycle-related, general shifts in ol- the manuscript: SCR, JH, and LMN. factory sensitivity to a number of odors, including those with different evolution- Conflict of interest ary significance. In the future, further in- tegration of studies on cyclic variation in The Authors declare that there is no olfactory sensitivity as a continuing pro- Conflict of interest. cess is encouraged, expanding research questions to, for instance, the occurrence Corresponding author and relative magnitude of similar fluctuations in hormonal contraceptive users. Lenka Martinec Nováková, Department This might contribute to a more profound of Anthropology, Faculty of Humanities understanding of the underlying mecha- Charles University, U Křĭže 8, 158 00 nisms involved in the cyclic changes in Prague 5, Czech Republic odor thresholds and shed more light on e-mail address: whether such changes can be viewed as [email protected]. a general phenomenon. Finally, a similar meta-analytic approach, employed here References for cyclic changes in olfactory threshold, should also be applied to studies on cy- Amoore JE, Popplewell JR, Whissell-Buechy clic fluctuations in olfactory preferences. D. 1975. Sensitivity of women to musk odor: no menstrual variation. J Chem Ecol 1(3):291–97. Acknowledgements Arctander S. 1969. Perfume and Flavor Chem- icals: (aroma Chemicals). Montclair, NJ: JH and LMN are supported by the Czech TheAuthor. Science Foundation (GA14-02290S) Becker D, Creutzfeldt OD, Schwibbe M, and Charles University Research Cen- Wuttke W. 1982. Changes in physiologi- tre (UNCE 204004). SCR was support- cal, EEG and psychological parameters in ed by a British Academy Mid-Career women during the spontaneous menstru- Fellowship. The funding sources had al-cycle and following oral-contraceptives. no involvement in study design, in the Psychoneuroendocrinology 7(1):75–90. collection, analysis, and interpretation 344 Lenka Martinec Nováková, Jan Havlíček, S. Craig Roberts

Bedoukian. Safety Data Sheet [Internet]. son, editor. Cognition and the menstrual Danbury, CT: Bedoukian Research. cycle. New York: Springer Verlag. 67–97. Bedoukian PZ. 1986. Perfumery and flavoring Gangestad SW, Thornhill R. 1998. Menstrual synthetics. Wheaton, IL: Allured Publish- cycle variation in women’s preferences for ing Corp. the scent of symmetrical men. Proc R Soc Bremner EA, Mainland JD, Khan RM, Sobel B 265(1399):927–33. N. 2003. The prevalence of androstenone Gangestad SW, Thornhill R. 2008. Human anosmia. Chem Senses 28(5):423–32. oestrus. Proc R Soc B 275(1638):991– Burdock GA. 2009. Fenaroli’s Handbook of 1000. Flavor Ingredients, Sixth Edition. Boca Good PR, Geary N, Engen T. 1976. The ef- Raton: Taylor & Francis. fect of estrogen on odor detection. Chem Caruso S, Grillo C, Agnello C, Maiolino L, Senses 2(1):45–50. Intelisano G, Serra A. 2001. A prospec- Gower DB, Nixon A, Mallet AI. 1998. The tive study evidencing rhinomanometric significance of odorous steroids in- axil and olfactometric outcomes in women lary odour. In: S Van Toller and Dodd GH, taking oral contraceptives. Hum Reprod editors. Perfumery. London: Chapman & 16(11):2288–94. Hall. 47–75. De Rovira D. 2008. Dictionary of Flavors. Grillo C, La Mantia I, Triolo C, Scollo A, La Ames, IA: John Wiley & Sons. Boria A, Intelisano G, Caruso S. 2001. Doty RL. 1979. Procedure for combining Rhinomanometric and olfactometric var- menstrual cycle data. J Clin Endocr Metab iations throughout the menstrual cycle. 48(6):912–18. Ann Oto Rhinol Laryn 110(8):785–89. Doty RL, Cameron EL. 2009. Sex differences Havlícek J, Roberts SC, Flegr J. 2005. Wom- and reproductive hormone influences on en’s preference for dominant male odour: human odor perception. Physiol Behav effects of menstrual cycle and relationship 97(2):213–28. status. Biology Letters 1(3):256–59. Doty RL, Gregor TP, Settle RG. 1986. Influ- Hedges LV, Olkin I. 1985. Statistical methods ence of Intertrial Interval and Sniff-Bottle for meta-analysis New York: Academic Volume on Phenyl Ethyl-Alcohol Odor Press. Detection Thresholds. Chem Senses Hummel T, Gollisch R, Wildt G, Kobal G. 11(2):259–64. 1991. Changes in olfactory perception Doty RL, Hall JW, Flickinger GL, Sondheimer during the menstrual cycle. Experientia SJ. 1982. Cyclic changes in olfactory and 47(7):712–15. auditory sensitivity during the menstrual Hummel T, Sekinger B, Wolf SR, Pauli E, Kob- cycle: no attenuation by oral contraceptive al G. 1997. ‘Sniffin’ Sticks’: Olfactory per- medication. In: W Breipohl, editor. Olfac- formance assessed by the combined test- tion and endocrine regulation. London: ing of odor identification, odor discrim- IRL Press. 35–42. ination and olfactory threshold. Chem Doty RL, McKeown DA, Lee WW, Shaman P. Senses 22(1):39–52. 1995. A study of the test-retest reliabil- Jacob TJC, Wang LW, Jaffer S, McPhee S. ity of ten olfactory tests. Chem Senses 2006. Changes in the odor quality of an- 20(6):645–56. drostadienone during exposure-induced Doty RL, Snyder PJ, Huggins GR, Lowry LD. sensitization. Chem Senses 31(1):3–8. 1981. Endocrine, cardiovascular, and psy- Keller A, Vosshall LB. 2004. Human ol- chological correlated of olfactory sensitiv- factory psychophysics. Curr Biol ity changes during the human menstrual 14(20):R875-R878. cycle. J Comp Physiol Psych 95(1):45–60. Kuga M, Ikeda M, Suzuki K. 1999. Gustatory Dye L. 1992. Visual information processing changes associated with the menstrual cy- and the menstrual cycle. In: JTE Richard- cle. Physiol Behav 66(2):317–22. Olfactory processing and odor specificity 345

Le Magnen J. 1952. Les phénoménes olfac- Rosenthal R, DiMatteo MR. 2001. Meta-anal- to-sexuels chez l’homme. Archives des ysis: Recent developments in quantitative Sciences Physiologiques 6:125–60. methods for literature reviews. Annu Rev Lundström JN, McClintock MK, Olsson MJ. Psychol 52:59–82. 2005. Effects of reproductive state on ol- Rusak B, Zucker I. 1979. Neural regula- factory sensitivity suggest odor specificity. tion of circadian rhythms. Physiol Rev Biol Psychol 71(3):244–47. 59(3):449–526. Mair RG, Bouffard JA, Engen T, Morton Saxton TK, Lyndon A, Little AC, Roberts SC. TH. 1978. Olfactory sensitivity during 2008. Evidence that androstadienone, the menstrual cycle. Sensory Processes a putative human chemosignal, modu- 2(2):90–98. lates women’s attributions of men’s at- McNeil J, Cameron JD, Finlayson G, Blundell tractiveness. Horm Behav 54(5):597–601. JE, Doucet E. 2013. Greater overall ol- Schmidt H. 1990. Methods of meta-analysis. factory performance, explicit wanting for London: Sage. high fat foods and lipid intake during the Stevens JC, Cain WS, Burke RJ. 1988. Varia- mid-luteal phase of the menstrual cycle. bility of olfactory thresholds. Chemi Sens- Physiol Behav 112:84–89. es 13(4):643–53. Navarrete-Palacios E, Hudson R, Reyes-Guer- Sueda K, Arao Y, Nishida N. 2003. Changes rero G, Guevara-Guzmán R. 2003. Lower in olfactory perception for androstenone olfactory threshold during the ovulatory during the menstrual cycle. Aroma Re- phase of the menstrual cycle. Biol Psychol search 4:51–55. 63(3):269–79. Vierling J, Rock J. 1967. Variations in olfacto- NCBI. 2014. PubChem. 2004 ed. Bethesda, ry sensitivity to exaltolide during the men- MD: National Center for Biotechnology strual cycle. J Appl Physiol 22(2):311–15. Information. Vo UUT, Morris MP. 2014. Nonvolatile, semi- Pause BM, Sojka B, Krauel K, Fehm Wolfsdorf volatile, or volatile: Redefining volatile for G, Ferstl R. 1996. Olfactory information volatile organic compounds. J Air Waste processing during the course of the men- Manage 64(6):661–69. strual cycle. Biol Psychol 44(1):31–54. Whissell-Buechy D, Amoore JE. 1973. Purdon SE, Klein S, Flor-Henry P. 2001. Men- Odour-blindness to musk: simple reces- strual effects on asymmetrical olfactory sive inheritance. Nature 242(5395):271– acuity. J Int Neuropsych Soc 7(6):703-09. 73. Raghava GPS. 2011. DrugPedia: A Wikipedia Wolf FM. 1986. Meta-analysis quantitative for Drug Discovery. Chandigarh, India. methods for research synthesis London: Renfro KJ, Hoffmann H. 2013. The relation- Sage. ship between oral contraceptive use and Wysocki CJ, Beauchamp GK. 1984. Ability to sensitivity to olfactory stimuli. Horm Be- smell androstenone is genetically deter- hav 63(3):491–96. mined PNAS Biol Sci 81(15):4899–902. Rosenthal R. 1991. Meta-analytical procedures for social research. London: Sage.