Progesterone-associated increase in ERP amplitude correlates with an improvement in performance in a spatial attention paradigm.

Ovarian sex hormones modulate neuronal circuits not directly involved in reproductive functions. In the present study, we investigated whether endogenous fluctuations of estradiol and progesterone during the menstrual cycle are associated with early cortical processing stages in a cued spatial attention paradigm. EEG was monitored while young women responded to acoustically cued visual stimuli. Women with large mean amplitude of the event-related potential (ERP) (80-120 ms following visual stimuli) responded faster to visual stimuli. In luteal women, mean amplitude of the ERP as well as alpha amplitude, an indicator of attentional modulation, correlated positively with progesterone. Further, cerebral asymmetry in ERP amplitude in the alpha frequency band following target presentation was restricted to luteal women. Critically, early follicular women responded slower to right hemifield compared to left hemifield targets. In late follicular or luteal women, we did not detect a right hemifield disadvantage. Progesterone correlated negatively with RTs in luteal women. Therefore, whereas our behavioral data indicate a functional cerebral asymmetry in early follicular women, EEG recording reveal a physiological cerebral hemisphere asymmetry in the alpha frequency band in luteal women. We assume that a progesterone-associated enhancement in synchronization of synaptic activity in the alpha frequency band in luteal women improves early categorization of visual targets in a cued spatial attention paradigm.

Introduction

Electrical cortical activity is segregated in discrete frequency bands (Buzsaki, 2006). Among the five mayor frequency bands, alpha and theta frequencies fluctuate predictably during the menstrual cycle, indicating an association between sex hormone fluctuations and neural activity (Becker et al., 1982, Creutzfeldt et al., 1976, Brötzner et al., 2014). Analysis of EEG data reveal a lower frequency in the alpha band in late follicular phase, when estradiol is elevated but progesterone is low, compared to early follicular phase, when estradiol as well as progesterone is low, or luteal phase, when estradiol as well as progesterone is elevated (Brötzner et al., 2014). Theta oscillations show a higher frequency in the follicular compared to the luteal menstrual cycle phase (Becker et al., 1982). How endogenous changes in sex hormone levels during the menstrual cycle contribute to inter- as well as intra-individual differences in cognitive performance and its underlying neural activity remains a fundamental issue.

Previous studies correlated cognitive performance either with an event-related potential (ERP) or sex hormone level. Following presentation of visual stimuli, the temporal sequence of an ERP consists of C1, P1, and N1. This sequence may represent sensory processing (C1), early categorization (P1), and identification of objects (N1) (Klimesch, 2011). Among the three components, P1, with a post-stimulus latency of approximately 100 ms, may be the earliest equivalent for top-down modulation of sensory input. In goal-directed top-down attention paradigms, expected perceptive contents are categorized as relevant or irrelevant information within a tenth of second (Thorpe et al., 1996, Rousselet et al., 2007; for review see Klimesch, 2011). Furthermore, Hanslmayr and colleagues describe that during a visual discrimination task enhanced early ERP components (P1 and N1 amplitude) are related to good performance (Hanslmayr et al., 2005). Several lines of arguments indicate that at least a fraction of P1 equals synchronized alpha oscillations: (1) P1 latency and period of alpha oscillation are approximately 100 ms, (2) P1 is predicted by phase alignment in alpha (Gruber et al., 2005) and (3) similar time domain of alpha oscillations and attentional blink (Hanslmayr et al., 2011). One influential interpretation of P1 is the inhibition model (Klimesch et al., 2007). According to the inhibition model, phasic synchronization of alpha oscillation is associated with an increase in signal to noise ratio for relevant information, but tonic synchronization with suppression of irrelevant information. Both processes improve working memory and attention performance (Klimesch et al., 2007).

Ovarian steroid hormones modulate neural circuits and cognitive performance not directly related to reproductive behavior. Endogenous fluctuations of the sex hormones, estradiol and progesterone, across the menstrual cycle may correlate with changes in attention and memory performance. Evaluation of ERP provides advantages for analyzing the impact of sex hormones on brain oscillations. First, EEG signals, including ERP, reflect synaptic activity (Buzsaki, 2006). Sex hormones modulate synaptic transmission, where progesterone and its metabolites affect inhibitory, GABAergic synaptic transmission and estradiol affects excitatory, glutamatergic synaptic transmission (Finocchi and Ferrari, 2011). Second, sex hormone level is associated with performance in goal-directed attention (Solís-Ortiz and Corsi-Cabrera, 2008). Third, goal-directed attention is associated with ERP amplitude (Klimesch et al., 2007). Forth, alpha oscillations are functionally and, presumably, physiologically inhibitory (Klimesch, 2011, Klimesch, 2012). Therefore, in the present study, we simultaneously examined performance, ERP, and sex hormone level in young women at three time points during the menstrual cycle using a cued attention paradigm. Our results in a goal-directed attention paradigm demonstrate an association of endogenous progesterone level with response time as well as mean absolute ERP amplitude and alpha ERP amplitude. We discuss our findings in an extended version of the inhibition model of how progesterone modulates synaptic activity underlying alpha oscillations.

Results

Menstrual cycle-dependent changes in RTs: progesterone level correlates negatively with RTs

Dependent t-tests showed that progesterone level is significantly higher during luteal phase compared to early follicular (t(17)=−3.504, p=.003) and late follicular phase (t(17)=−3.044, p=.007).

Table 1 summarize mean and SD for RTs for early follicular, late follicular and luteal phase for the spatial attention test performed during EEG recording (Fig. 1). The main findings were that women responded (1) significantly faster to valid compared to invalid trials during early follicular (F(1,17)=26.231, p<.001, η2=.607), late follicular (F(1,17)=9.058, p=.008, η2=.348) as well as luteal phase (F(1,17)=7.719, p=.013, η2=.312), and (2) consistently – but not statistically significant – slower to right valid and invalid trials compared to left valid and invalid trials, in the early follicular phase (F(1,17)=3.485, p=.079, η2=.170), but not in the late follicular (F(1,17)=.003, p=.959, η2<.001) and luteal phase (F(1,17)=.002, p=.963, η2<.001). Because RTs were slower in right hemifield cued targets compared to left hemifield cued targets in the early follicular, but not in late follicular and luteal phase, we suggest a right hemifield disadvantage in the early follicular phase. RT does not differ within the three cycle phases (p>.05). Further, RTs correlated negatively with accuracy (p<.05). We found no cycle or hormone dependent differences in accuracy. Mean accuracy was between 74 and 100% with a mean of 96.5%. As women started at different stages of the menstrual cycle in our experiments, we tested for the influence of repeated testing by calculating a 3×4 ANOVA with the factors EEG session (first, second, third) and experimental condition (left valid, right valid, left invalid, right invalid). The dependent variable was RT. We found a main effect for the factor EEG session showing the slowest RTs on the first EEG session compared to the second and third EEG session (F(2,34)=12.024, p<.001). To test for EEG session-dependent functional cerebral asymmetry (FCA), we calculated a 2×2 ANOVA with factor validity (valid, invalid) and hemifield (left, right) separately for the first, second and third EEG session. This analysis revealed that functional cerebral asymmetry was neither detectable in the first, second nor third EEG session (p>.2). Thus, irrespective in which menstrual cycle phase women began in our experiments, right hemifield disadvantage was restricted to the early follicular phase.

Table 1

RT (in ms) for early follicular (EFP), late follicular (LFP) and luteal phase (LP).

	Valid trials		Invalid trials
	Left hemifield	Right hemifield	Left hemifield	Right hemifield
EFP	534.56±57.43	543.95±58.38	553.60±59.54	564.29±62.44
LFP	527.29±48.65	527.83±57.20	543.33±69.10	541.89±61.57
LP	531.67±54.79	531.28±52.20	551.14±54.98	552.28±57.17

Fig. 1

Sequence of events in the visuospatial cued attention task. (ISI=inter-stimulus-interval).

A correlative analysis for the behavioral data revealed a significant progesterone, but not estradiol effect. Progesterone levels were negatively correlated with RTs in luteal but not in early follicular and late follicular women (Table 2). In luteal women, progesterone level was negatively correlated with RTs in left valid (r(16)=−.512, p=.030) and right valid trials (r(16)=−.685, p=.002). Estradiol level did not correlate with RTs in valid as well as invalid trials for both hemifields in neither menstrual cycle phase.

Table 2

Correlation between progesterone-concentration and RT for valid trials.

	RT left hemifield	RT right hemifield
EFP progesterone	−.207	−.098
LFP progesterone	−.025	.225
LP progesterone	−.512⁎	−.685⁎⁎

p<.05.

p<.01.

Mean absolute ERP amplitude is associated with RTs as well as progesterone level

The main behavioral findings were that women (1) responded significantly faster to valid compared to invalid trials, (2), revealed significant correlations between progesterone and RTs in luteal women and (3) showed a right hemifield disadvantage in the early follicular phase. A RT advantage for valid compared to invalid trials in a cued attention paradigm was also reported by Freunberger and colleagues as well as by Sauseng and colleagues, who showed a larger P1 and alpha power for task-irrelevant trials on ipsilateral sites (Freunberger et al., 2008) and 10 Hz frequency specific effects in visual short-term memory using TMS (Sauseng et al., 2011). Because we found progesterone-associated differences in RTs in luteal women, we focused on progesterone-associated differences in the EEG signature during responses in attention tasks. In general, the EEG signature of a cued attention task contained a negativity provoked by an auditory cue and an ERP induced by a visual target (Fig. 2). Our analysis included only the first 120 ms following presentation of the visual task. This temporal segment is sufficient for an early categorization of the target (Klimesch et al., 2007). To analyze EEG correlates of cued attention task, we segregated the EEG signal in two segments. Defining task-onset as 0 ms, the first post-task segment reached from 0 to 80 ms, and the second post-task segment from 80 to 120 ms, which included the P1. EEG signals were the mean of the posterior electrodes P3 (left parietal cortex), Pz, and P4 (right parietal cortex).

Fig. 2

Association between RTs and ERP amplitudes in left and right hemifield trials. Representative standard ERPs following left valid (A) or right valid (B) hemifield presentation from the luteal woman with the fastest and slowest RTs. Average ERPs from women in the luteal phase having either RTs above or below the median of RTs following left valid (C) or right valid (D) hemifield presentation. Note discontinuity in the initial phase of N1. Comparison of ERPs at the individual level revealed larger P1 and N1 in the woman with fastest response time (A, B). Average alpha filtered ERPs from women in the luteal phase having either RTs above or below the median of RTs following left valid (E) or right valid (F) hemifield presentation. Overall larger amplitude is associated with fast RT. ERPs are averaged over P3, Pz, P4 electrodes. (RT=response time).

Representative standard ERPs following left valid or right valid hemifield presentation from the luteal woman with the fastest and slowest RTs, respectively, is shown in Fig. 2A, B. The main difference was a larger P1–N1 complex in the woman with the fastest compared to the woman with a slowest RT. Fig. 2C, D visualize average ERPs from women having either RTs above or below the median of RTs. Surprisingly, in left valid hemifield trials average ERPs from luteal women with fast RTs revealed a smaller P1 than expected from ERP signature from a single woman. This discrepancy regarding P1 and N1 amplitudes between ERPs recorded from an individual and ERPs averaged from several women is most likely due to different temporal onsets of short-lived P1 and, accordingly, P1 overlaps with long-lived N1 so that the initial part of N1 is contaminated with the P1 signal. Table 3 summarizes correlations between mean absolute ERP amplitude and RT in early follicular, late follicular and luteal women. Critically, we found significant correlations between RTs and mean amplitude only in luteal women, but not in early follicular women, where we observed a right hemifield disadvantage. Significant correlations between RT and ERP amplitude were identified for left valid as well as right valid hemifield presentations.

Table 3

Correlations between RT and ERP amplitude.

	ERP mean amplitude (0–80 ms)		ERP mean amplitude (80–120 ms)		Alpha P1–N1 amplitude
	Left hemifield	Right hemifield	Left hemifield	Right hemifield	Left hemifield	Right hemifield
EFP_RT	−.246	−.247	−.425	−.251	−.583⁎	−.394
LFP_RT	−.199	.076	−.213	.099	−.411	−.266
LP _RT	−.501⁎	−.596⁎⁎	−.642⁎⁎	−.621⁎⁎	−.636⁎⁎	−.341

p<.05.

p<.01.

The observation, that RTs correlated significantly with mean absolute amplitude of ERP in luteal, but not in early or late follicular women, indicate an impact of ovarian steroid hormones on this association. Accordingly, we next analyzed the association between progesterone and estradiol, respectively, and mean absolute amplitude of ERP. Interestingly, we found significant associations between progesterone and mean absolute post-stimulus amplitude in luteal, but not in early or late follicular women (Table 4, Fig. 3A). We did not identify a significant association between estradiol and mean absolute amplitude of ERP.

Table 4

Correlations between progesterone-concentration and ERP amplitude.

	ERP mean amplitude (0–80 ms)		ERP mean amplitude (80–120 ms)		Alpha P1–N1 amplitude
	Left hemifield	Right hemifield	Left hemifield	Right hemifield	Left hemifield	Right hemifield
EFP progesterone	.390	.446	.458	.455	.257	.143
LFP progesterone	.140	.019	.111	.294	.302	.282
LP progesterone	.485⁎	.627⁎⁎	.479⁎	.602⁎⁎	.562⁎	.520⁎

p<.05.

p<.01.

Fig. 3

Associations between progesterone level and ERP amplitudes. Average ERPs (A) and alpha filtered ERPs (B) from women in the luteal phase having either progesterone level above or below the median of progesterone concentration following left valid or right valid hemifield presentation. Noticeably high progesterone concentration is associated with high ERP amplitude. ERPs are averaged over P3, Pz, P4 electrodes. (P=progesterone).

Alpha oscillation, RTs and progesterone

Since the second post-stimulus segment between 80 and 120 ms equals the period of an alpha oscillation (~100 ms), we correlated post-stimulus alpha P1–N1 amplitude difference with RTs and progesterone, respectively. Alpha P1–N1 amplitude difference revealed significant correlations with RTs in left valid trials in early follicular and luteal women (Table 3, Fig. 2E, F). Similar to the standard ERP, progesterone correlated with post-stimulus alpha P1–N1 amplitude difference in left and right valid hemifield trials only in luteal, but not early or late follicular women (Table 4, Fig. 3B).

Visual hemifield presentation provoke an asymmetry in the EEG signal in luteal women

Our behavioral experiments revealed a right hemifield disadvantage, meaning that right valid trials provoked slower RTs than left valid trials in early follicular women. Traditionally, this is interpreted as a functional cerebral asymmetry. Therefore, we compared the EEG signal in the left parietal (electrode P3) and right parietal cortex (electrode P4) following valid hemifield presentations. In the alpha filtered ERP, we detected: (1) cerebral asymmetries following valid task presentations in ipsi- but not contralateral brain areas, and (2) significantly smaller alpha P1–N1 amplitude differences in P3 in left valid trials compared to alpha P1–N1 amplitude differences in P4 following right valid trials (t(17)=−2.212, p=.041, d=.5) (Fig.4). Thus, in contrast to behavioral data, hemispheric asymmetry was largest in the luteal phase. In early and late follicular women, we did not detect significant cerebral hemisphere asymmetries.

Fig. 4

Hemispheric asymmetry in the EEG signal following ipsi- or contralateral presentation of visual stimuli. Alpha P1–N1 amplitude differences are significantly smaller in P3 after left hemifield stimulation compared to P4 following right hemifield stimulation in luteal women.

Discussion

Our findings provide additional evidence that fluctuations in ovarian sex hormones are involved in fluctuations in cognitive performance and further indicate that progesterone is a modulator in neuronal circuits related to attention. Using a cued spatial attention paradigm, we observed (1) significant correlations between progesterone and RTs as well as mean absolute ERP amplitude in luteal, but not in follicular women; (2) a significant correlation between progesterone and alpha P1–N1 amplitude difference in luteal women, (3) a functional cerebral asymmetry (right hemifield disadvantage) in early follicular women, and (4) a physiological hemispheric asymmetry in the alpha frequency band in the luteal women. This may indicate that an increase in progesterone enhances synchronization in the alpha frequency band and, accordingly, improves attention performance in women.

Progesterone-associated improvement in performance correlates with progesterone-associated enhancement in mean absolute ERP and alpha amplitude

Analysis of top-down modulation of visual cortical neurons at the single-unit level in rhesus macaques (Macaca mulatta) indicates involvement of two physiologically distinct neuronal populations in attentional processing (Mitchell et al., 2007, Chen et al., 2008). Whereas one population belongs to pyramidal neurons, the second population includes GABAergic neurons characterized by a spontaneous resting activity of 9.4 Hz (Mitchell et al., 2007), which is within the alpha frequency band. Thus, interpretation of EEG signals recorded during cued attention tasks should include activity of excitatory, pyramidal neurons and inhibitory, GABAergic neurons. The present EEG study focused approximately on the first tenth of a second following target presentation. This temporal domain is sufficient for an early categorization process of a target (Klimesch et al., 2007). In a top-down attention paradigm, like the cued attention paradigm used in the present study, expectancy is a selection mechanism among sensory inputs in cortical areas. In EEG recordings, a method of extracellular recording, enhancement in excitability is reflected in an increase in negativity. In the present study, we identified in valid trials significant correlations between mean absolute ERP amplitude and RTs within the first tenth of second following target presentation. The first segment (0–80 ms) may represent an increase in excitability due to a top-down control of sensory input. Enhancement of excitability decreases the threshold for relevant or expected sensory input. The second segment (80–120 ms) includes the P1 component of the ERP. P1 as well as the P1–N1 complex may represent a synchronized synaptic input in the alpha frequency band (~10 Hz). Critically, alpha oscillation is functionally and, presumably, physiologically inhibitory (Klimesch, 2012). Thus, non-synchronized neuronal activity within the first 80 ms may also induce competition among local neuronal networks, which culminates in synchronization of inhibitory networks specific for the target. Amplitude of P1–N1 difference or alpha amplitude may be an indicator of the magnitude of synchronization. Increase in alpha activity either increase signal to noise ratio and allows processing of relevant information (contralateral hemisphere) or suppression of irrelevant information (ipsilateral hemisphere) (Klimesch et al., 2007, Klimesch, 2012).

At the single cell level, estradiol increases, but progesterone decreases neuronal excitability (Majewska et al., 1986, Wong and Moss, 1992, Spencer et al., 2008, Finocchi and Ferrari, 2011). As progesterone and its metabolites affects inhibitory, GABAergic synapses, fluctuations of endogenous progesterone during menstrual cycle might affect synchronization of inhibitory networks. In the present study, we find that women with fast RTs show higher progesterone level compared to women with slow RTs. Further, progesterone correlates positively with alpha P1–N1 amplitude difference. Thus, assuming that alpha oscillations are inhibitory at the physiological level, an increase in progesterone may enhance inhibition via fine tuning rhythmic synchronization of neural networks leading to improvements in cognitive processing. Critically, the inhibition model of alpha oscillations predicts that an increase in functional inhibition causes an increase in alpha amplitude. An increase in alpha amplitude, specifically in P1–N1 difference, may increase signal to noise ratio as well as tonic inhibition of networks processing irrelevant information (Klimesch et al., 2007). Both mechanisms improve cognitive processing.

We summarize our results in a progesterone-dependent alpha-inhibition model. This model combines the “inhibition model” (Klimesch, 2011) with the physiological consequences of progesterone on neuronal excitability as well as on alpha oscillations. The progesterone-dependent alpha-inhibition model predicts in our cued spatial attention paradigm that an increase in progesterone is associated with (1) tonic mutual inhibition of ipsilateral cerebral hemispheres illustrated by a larger alpha P1–N1 amplitude difference in the ipsilateral hemisphere and (2) increase in signal to noise ratio in the contralateral hemisphere via enhancing GABAergic synaptic transmission visualized by larger alpha P1–N1 amplitude difference in women high in performance compared to women low in performance.

Functional cerebral asymmetry and physiological lateralization of hemispheres

Cerebral hemispheres are mutual inhibitory (Innocenti, 2009, Bocci et al., 2014). Accordingly, in top down controlled attention tasks, neural equivalent of expectancy of a target may include a cue-induced increase in excitability in the contralateral, but an increase in inhibition in ipsilateral hemisphere. When RTs following left as well as right hemifield presentation are similar, we assume equal mutual hemispheric inhibition at the functional level. Functional equality in RTs is present in luteal women, when progesterone is elevated. A decline in progesterone in early follicular women correlates with functional inequality visualized by larger latency in RTs in right compared to left hemifield presentation. Thus, at the behavioral level right hemisphere is dominant in attention tasks. Dominance of right hemisphere has been identified in several attention tasks (Petersen and Posner, 2012, Somers and Sheremata, 2013). Mutual inter-hemispheric inhibition at the physiological level is visualized in differences in ERP or alpha-amplitude. Ipsilateral alpha amplitude is larger in right than left visual field presentation. This asymmetry in amplitude is statistically significant in luteal women. Thus, suppression of the dominant right hemisphere requires synchronization of a larger inhibitory neuronal network than suppression of the subdominant left hemisphere. One interpretation of larger right hemisphere synchronization is that subdominant areas in the left hemisphere may trigger synchronization of a larger inhibitory network in the dominant, right hemisphere when progesterone is elevated. An alternative interpretation is that the dominant right hemisphere suppresses the subdominant left hemisphere more efficiently and, thus, decreases interferences in information processing. In both cases, progesterone enhances synchronization in alpha frequency band and therefore leads to suppression of irrelevant information in the ipsilateral hemisphere and minimizes interferences between cerebral hemispheres.

Thus, our findings may contribute to elucidate an interesting paradox regarding the impact of sex hormones on functional cerebral asymmetry and physiological hemisphere laterality. On the one hand, the progesterone-mediated interhemispheric decoupling model by Hausmann and Güntürkün predicts that an increase in progesterone decrease hemisphere asymmetry (Hausmann and Güntürkün, 2000). This model states that hemispheres are coupled when the dominant hemisphere suppresses homotopic areas of the subdominant hemisphere. Glutamatergic neurons, projecting form the dominant to the subdominant hemisphere, synapse on pyramidal neurons, which activate GABAergic neurons. An increase in progesterone decouples cerebral hemispheres and, thus, decreases functional cerebral asymmetry. Accordingly, functional cerebral asymmetry is only detectable in menstrual cycle phases with low progesterone level (Hausmann and Güntürkün, 2000). On the other hand, the Hampson model predicts that an elevation of ovarian sex hormones facilitates left hemisphere processing. Accordingly, hemispheric lateralization is associated with an increase in sex hormones (Hampson, 1990).

In conclusion, we suggest that functional cerebral asymmetry at the behavioral level in early follicular women is related to dominance of the task specific hemisphere. An increase in cerebral asymmetry at the EEG level in luteal women is associated with an increase in alpha power, which may represent suppression of the dominant by the subdominant hemisphere and may improve performance in relation to progesterone level.

Any interpretation of a biological significance of improved performance in luteal compared to early follicular women is at the moment speculative. However, if our laboratory findings are translatable to daily routines, two applications are imaginable. Rating of attractiveness of women by men as well as preference for masculinity by women depends on the phase of the menstrual cycle (Little et al., 2007, Little and Jones, 2012, Puts et al., 2013). Women are perceived as more attractive in fertile compared to non-fertile menstrual cycle phases (Roberts et al., 2004, Puts et al., 2013). Interestingly, rating by men of female facial and vocal attractiveness correlated negatively with progesterone (Puts et al., 2013). Further, women high in progesterone prefer men perceived as supportive (Jones et al., 2005). Thus, a positive association between progesterone and performance in cued attention in the present study as well as in selection of likely supportive mates (Jones et al., 2005) may indicate a progesterone-dependent modulation of top-down processes of expected features in women. In addition, luteal phase may represent the earliest stage where a woman has conceived a child. Both conditions require a higher demand on attention to scan and respond to expected or unexpected social stimuli or to avoid potential precarious situations.

A small sample size of 18 women participating in our study is a source of concern. However, each of the 18 women was repeatedly tested at three distinct menstrual cycle phases. Accordingly, we collected a total of 54 EEG recordings and equivalent behavioral data. Furthermore, statistical analysis of specific hypotheses regarding the association of progesterone with RTs or mean amplitudes revealed significant correlations.

In conclusion, we suggest that improved performance in luteal women is associated with progesterone-dependent increase in alpha oscillations, which is related to tonic suppression of irrelevant information, but phasic increase in signal to noise ratio of relevant information.

Experimental procedures

Participants

22 women gave informed consent to participate in the present EEG study. Individuals had no history of neurological or psychiatric diseases and were not taking medications. Two women were excluded because they had no menstruation since one year and two because they did not follow task instruction and moved their eyes away from the fixation cross. The remaining 18 women (age: 24.06±4.66, 2 left handed) had a regular menstrual cycle (mean cycle length: 29.44±1.9 days). Eleven women were students from the University of Salzburg (Department of Biology, Department of Psychology), three women were students from a vocational secondary school in Salzburg and four women were employees in Salzburg. All subjects got 30€ remuneration for participation. Our study was approved by the ethics committee of the University of Salzburg.

Naturally cycling women were tested three times, once during their early follicular phase (low estrogen and progesterone), once during ovulation (estrogen peak), and once during their mid-luteal phase (high estrogen and progesterone). Early follicular phase ranged from onset of menstruation plus five days. Late follicular phase (ovulation) was estimated using a commercial ovulation test (Pregnafix®Ovulationstest) as well as by verbal reports. Ovulation was approximated as fourteen days before onset of menstruation. Mid-luteal phase spanned from day three post ovulation to five days before the onset of menstruation. Nine naturally cycling women had their first EEG session during early follicular phase, five women during ovulation and four women during mid-luteal phase. With four exceptions, the three EEG sessions were a maximum of one cycle apart.

Visuospatial cued attention task

A fixation cross was presented 5.5° visual angle above the center of the screen and visual targets (“p” or “q”) were viewed on a computer screen with a visual angle of 1.5° (Sauseng et al., 2011). Targets were 12.7° to the left or right of the center, which was labeled with cross. Distance between participant and screen was 80 cm. Participants had normal or corrected to normal vision. Each trial consisted of an acoustic cue and a visual target (Fig. 1). A 500 Hz tone required focusing of attention to the left hemifield (without moving eyes away from the fixation cross), and a 1000 Hz tone, which directed attention to the right hemifield. Following a jittered interval of 600 to 800 ms after the acoustic cue, a visual target was presented at the screen for 83 ms. The target (“p” or “q”) was presented either on the left or on the right hemifield. In valid trials, target was presented at the hemifield indicated by the acoustic cue, in invalid trials, the target was presented at the opposite hemifield indicated by the tone. The paradigm consisted of 400 trials, of which in half attention had to be directed to the left and in half to the right hemifield. In 75% of the trials, target location was congruent with the cued visual hemifield (valid trials). The inter-trial interval lasted between 2000 and 3000 ms. Participants were asked to respond as fast as well as accurate as possible by pressing the left mouse button with their index finger of the right hand for “p” and the right mouse button with their middle finger of their right hand for “q”. Before women performed the experiment, they practiced one block with 50 trials. Stimuli were presented using Presentation Software (version .71, 2009, Neurobehavioral Systems Inc., Albany, CA, USA).

Salivary sex hormone analysis

To determine sex hormone levels, each participant provided a saliva sample before an EEG-session. Samples were taken by direct expectoration into sterile tubes. Saliva samples were then stored in a freezer at −20 °C. Before sex hormone quantification, particles in saliva samples were removed by centrifugation (2355 g for 15 min). Salivary estradiol and progesterone concentration were measured using Demetitec Salivary Estradiol ELISA kids. The mean and standard deviation of estradiol levels were 3.94±1.82 pg/ml in early follicular phase, 4.88±2.75 pg/ml in late follicular phase and 5.20±4.22 pg/ml in luteal phase. The mean and standard deviation of progesterone levels were 62.34±57.74 pg/ml in early follicular phase, 65.23±31.31 pg/ml in late follicular phase and 133.27±102.95 pg/ml in luteal phase.

EEG-recordings

Acquisition of EEG data

32 Ag–AgCl electrodes were used to record EEG signals. Electrode position was according to the 10–20 – system (Jaspers, 1958). Electrodes were referenced to a nose electrode. Signals were amplified with a BrainAmp amplifier (Brain Products, Inc., Gilching, Germany) using a sampling rate at 1000 Hz. To eliminate 50 Hz oscillation, a notch filter at 50 Hz was applied and recording bandwidth was set from .016 to 100 Hz. Eye movements were controlled by two electrodes set at vertical and horizontal positions near the right eye. Impedance was kept below 8 kΩ.

Data-analysis

EEG data were analyzed using BrainVisionAnalyzer 2.0 (Brain Products, Inc., Gilching, Germany). Raw EEG data were re-referenced to earlobe-electrodes and filtered with an IIR bandpass filter between .5 and 40 Hz. EEG data were corrected for EOG artifacts using ocular correction based on Gratton and Coles (Gratton et al., 1983). Remaining artifacts e.g., due to eye movements, blinks, muscle activity, etc., were excluded by manual visual inspection. Because of inter-individual variety in the dominant alpha frequency, IAF was estimated (Klimesch, 1997). To calculate the IAF in resting conditions with eyes closed, five minutes were segmented into consecutive 2000 ms and analyzed using a Fast-Fourier-Transformation (FFT). After averaging we detected visually the highest peak of the P3 and P4 electrode within a frequency window from 8 to 12 Hz. For alpha ERP-analyses, frequency bands were adjusted to mean IAF individual alpha frequency. Accordingly to the mean IAF (9.8 Hz), the non-segmented data were bandpass filtered in a frequency range between 7.8 and 11.8 Hz for the alpha band. For the alpha filtered and non-filtered data 800 ms epochs were extracted from the data beginning 300 ms preceding visual target presentation and ending 500 ms after target onset. Trials with response time below or above 300–900 ms were excluded. Only trials with correct responses were included and further analyzed. ERPs for two experimental conditions (left and right valid hemifield presentation) were obtained by averaging over trials. ERPs for invalid experimental conditions were not analyzed because we did not find an effect of progesterone on RT in invalid trials.

Alpha P1–N1 amplitude difference

For alpha filtered ERPs, individual early ERP-components were semi-automatically detected. P1 was detected as the largest positive amplitude between 50 to 150 ms and N1 as the largest negative amplitude between 150 and 250 ms. P1–N1 amplitude difference was calculated subtracting N1 amplitude from P1 amplitude.

Mean absolute ERP amplitude

Mean absolute ERP amplitude was calculated as the absolute value of the amplitude of a defined time window of the ERP. Defining task-onset as 0 ms, the first post-task segment reached from 0 to 80 ms, and the second post-task segment from 80 to 120 ms, which included the P1.

P3, Pz, and P4 were our electrodes of interest because they showed the largest effects in our paradigm. Figures were created using BrainVisionAnalyzer 2.0 (Brain Products, Inc., Gilching, Germany).

Behavioral data

Response time was determined on an individual subject level. We extracted the median of response time collectively for each experimental condition (left hemifield presentation valid, right hemifield presentation valid, left hemifield presentation invalid, right hemifield presentation invalid) and separately for correct and incorrect trials.

Statistical analyses of behavioral measurements.

PASW Statistics 18 (SPSS) was used for statistical analysis. To specify the contribution of sex hormones on RTs, we correlated sex hormone levels for each menstrual cycle phase with RT in four experimental conditions: left valid and invalid trials as well as right valid and invalid trials. Sex hormone levels were associated with RTs using the Pearson correlation coefficient (2-tailed). We also correlated in each menstrual cycle phase and for each experimental condition accuracy and RTs using the Pearson correlation coefficient (2-tailed). To calculate the validity effect and the right hemifield disadvantage, RTs for correct responses for each cycle phase were subjected to a 2×2 ANOVA (Greenhouse-Geisser) with factors validity (valid, invalid) and visual hemifield (left, right). Cycle Phase differences in RT were calculated using a 3×4 ANOVA (Greenhouse-Geisser) with factors cycle phase (EFP, LFP, LP) and experimental condition (left valid, right valid, left invalid, right invalid). The dependent variable was RT.

Statistical analyses of general ERP effects

For statistical analysis we averaged the mean absolute ERP amplitude and the P1–N1 amplitude difference for P3, Pz and P4 electrode. Sex hormone levels and RT were associated with mean absolute ERP amplitude (separately for 0–80 ms and 80–120 ms) and alpha P1–N1 amplitude difference using the Pearson correlation coefficient (2-tailed). Calculations were done only for valid trial conditions (left and right hemifield) and separately for all cycle phases. Hemisphere lateralization in early ERP amplitudes in each menstrual cycle phase was evaluated using dependent t-tests. In each test, we compared left with right ipsilateral alpha P1–N1 amplitude difference. The same analyses were done for left and right contralateral amplitudes.