Estradiol modulated brain connectivity in females during midlife performing an episodic memory task Authors: Abigail A. Testo, Julie A. Dumas Published: 2026 ABSTRACT Midlife in females typically involves a significant neuroendocrine transition that may have effects on brain connectivity that are relevant to both normal and pathological aging. One potential contributing factor to the individual differences in cognition and risk for pathological aging is the brain’s response to decreased levels of estradiol. We assessed the relationship between estradiol level and task-modulated effective connectivity between brain regions during an episodic memory task in females during midlife. We utilized data from the Human Connectome Project – Aging 2.0 release. One hundred and fifty female participants between 40 and 55 years of age were included in our analysis of the relationship between estradiol and connectivity. Pre-, peri-, and postmenopausal participants were included to capture a range of estradiol levels present during midlife. We assessed connectivity during the fMRI FaceName task utilizing Conn Toolbox on a connection level as a pairwise comparison between regions. Five connections associated with estradiol were significant during the recall portion of the FaceName task at a p<.05 FDR corrected level after controlling for age. Significant connections among the superior frontal gyrus left and brain stem and the middle temporal gyrus anterior division right and superior temporal gyrus posterior division right were identified. Negative associations between these regions were also present. There was no significant relationship between estradiol and FaceName task performance. Our results indicated that task-modulated effective connectivity was related to estradiol level during midlife in females between brain regions known to be important for memory functioning during the recall portion of an episodic memory task. These findings may have long term implications for memory function in females during the aging process due to the low levels of estradiol present post-menopause.Supplementary InformationThe online version contains supplementary material available at 10.1007/s11682-026-01167-1. Introduction In human females, midlife typically involves a neuroendocrine aging process known as perimenopause (Brinton et al., 2015). Perimenopause is the transition to menopause, which occurs after one year of amenorrhea and marks reproductive senescence (Brinton et al., 2015). During perimenopause estrogen levels transition from their pre- to post-menopause levels (Hariri & Rehman, 2023; McEwen et al., 1997). These hormonal changes impact both normal and pathological brain aging, such as dementia (Brinton et al., 2015; Newhouse & Dumas, 2015). Prior to menopause, estradiol is the most potent and abundant estrogen in the female body (Hariri & Rehman, 2023) and is associated with positive neurological effects (Ycaza Herrera & Mather, 2015). After menopause, circulating levels of estrogens, including 17β-estradiol (E2) are significantly lower than pre-menopause (Rannevik et al., 1995). The decrease in E2 is largely responsible for many of the symptoms experienced during and after perimenopause (Hariri & Rehman, 2023) including changes in mood, behavior, and cognition (Genazzani et al., 2007). Complaints of declines in cognitive domains such as working memory and attention (Weber et al., 2012) have been reported during perimenopause. These cognitive complaints are both distressing and may reflect true cognitive difficulties (Weber et al., 2012). In most instances, the negative effects of varying hormone levels on cognitive performance are transient and do not persist long into menopause (Greendale et al., 2009). However, perimenopause may present as a time of increased vulnerability for risk factors associated with pathological brain aging (Brinton et al., 2015). It remains unclear what factors contribute to some individuals experiencing a persistent decline in cognition, while others do not. Differences between normal and pathological aging are often reflected in episodic memory performance. Episodic memory is the deliberate recall of autobiographic events and experiences (Tulving, 1972) and relies on brain regions, such as the hippocampus (Rugg & Vilberg, 2013). Tests of encoding and recall have revealed the influence of E2 on brain regions critical for memory functioning. After menopause, E2 depletion decreased hippocampal activation during encoding, but increased activation in the anterior cingulate cortex and dorsolateral prefrontal cortex (DLPFC), indicating a mechanism may exist to compensate for the effects estrogen decline has on the hippocampus (Bayer et al., 2015). In instances of bilateral salpingo-oophorectomy, midlife participants not receiving E2 therapy were found to have lower posterior hippocampal volumes compared to those receiving E2 therapy (Brown et al. 2025a, b). Bilateral salpingo-oophorectomy has also been found to be associated with lower posterior hippocampal activation during the encoding portion of the face-name task in both early midlife participants taking E2 therapy and those who are not compared to age-matched controls (Brown et al. 2025a, b). Additionally, higher levels of E2 after menopause were associated with increased activation of temporal and frontal cortices during encoding, which may enhance memory performance (Schroeder et al., 2025). During recognition, E2 was negatively corelated with activation in a prefrontal region (Schroeder et al., 2025). E2 may affect cognition through its ability to influence underlying brain functioning, measured with functional connectivity. Functional connectivity assesses the similarity of signals arising from anatomically distinct brain regions. After menopause, E2 has been found to enhance functional connectivity between the left anterior parahippocampal gyrus and the precuneus and right posterior parahippocampal gyrus (Testo et al., 2024). Studies of connectivity during task-based functional magnetic resonance imaging (fMRI) demonstrated the influence of E2 underlying task performance. E2 concentrations were related to increased hippocampal connectivity and better performance on a memory retrieval task across all menopausal stages (Jacobs et al., 2016). Effective connectivity, defined as the casual influence one brain region exerts on other anatomically distinct regions (Stephan & Friston, 2010), was also influenced by E2. A study of directed effective connectivity found that E2 influenced effective connectivity in networks underlying emotion regulation and reactivity in premenopausal participants in both positive and negative directions compared to placebo, (Derntl et al., 2024). The presence of positive and negative associations between brain regions related to E2 level suggests a complex relationship between E2 and brain connectivity. The purpose of the present study was to assess the level of task-modulated effective connectivity between cortical regions associated with E2 level during an episodic memory task in females during midlife. We investigated how E2 influenced episodic memory task-modulation of functional connectivity patterns during both encoding and recall conditions. We hypothesized that E2 levels would be related to increased connectivity in females between 40 and 55 years of age during the encoding conditions and associated with decreased connectivity during the recall condition based on previous works investigating the relationship between E2 and activation and connectivity. Methods Study design Data utilized in the present study were collected and shared by the Human Connectome Project and were included in the Lifespan Human Connectome Project Aging (HCP-A) 2.0 release. The HCP-A is a large-scale data collection project with the goal of investigating the effects of healthy aging on brain circuitry and is the result of work conducted by co-investigators from Washington University, University of Minnesota, Massachusetts General Hospital, University of California Los Angeles, and Oxford University. Potential participants received a phone screening prior to participation. Individuals diagnosed and treated for major psychiatric disorders were excluded from the HCP-A study, as were individuals diagnosed with severe depression that required treatment for 12 months or longer in the past five years. Individuals diagnosed with neurological disorders such as stroke, brain tumors, and Parkinson’s Disease were also excluded. Potential participants were also required to meet the threshold for their age group on the Montreal Cognitive Assessmentnnn (Nasreddine et al., 2005). The HCP-A 2.0 release contained imaging and behavioral data from 725 participants, 56% of which were female. The present study included data from participants that were classified as female and between 480 months (40 years) and 671 months (55 years + 11 months) of age. The use of hormonal medications did not exclude participants from the analyses. A total of 154 participants included in the HCP-A 2.0 release met this criterion; however, participants with hormonal disorders or missing task imaging data were excluded. One participant was excluded due to polycystic ovarian syndrome diagnosis and a further three were excluded for incomplete imaging data. The final N was 150 females. Hormone data collection HCP-A researchers collected blood samples for hormonal tests following an 8-hour fast. Imaging and biospecimen collection were conducted on the same day. Samples were analyzed by the Washington University’s Core Laboratory. In participants between the ages of 45 and 55 years of age, blood samples were collected between 2 and 6 days following the start of the menstrual cycle to account for hormone level variability (Bookheimer et al., 2019). Participants were classified as pre-, peri-, or post-menopause, as assessed by HCP-A researchers using the Stages of Reproductive Aging Workshop + 10 (Harlow et al., 2012) and menstrual history and menopause status questionnaires. An additional group defined as missed periods by HCP-A researchers included participants who were experiencing irregular periods that could not be conclusively attributed to perimenopause. MRI sequences HCP-A imagining procedures were described in (Harms et al., 2018) and (Bookheimer et al., 2019). Data were collected at four acquisition sites using equivalent Siemens Prisma 3 T MRI scanners. Structural scans included T1-weighted images at 0.8 mm isotropic resolution. Structural scans had a sagittal FOV of 256 × 240 × 166 mm with a matrix size of 320 × 300 × 208 slices. T1 parameters were as follows: TR/TI = 2500/1000, TE = 1.8/3.6/5.4/7.2 ms, and a flip angle of 8 degrees. BOLD scans collected as a part of the HCP-A study were acquired with a 2D multiband gradient-recalled echo echo-planar imaging sequence with a multiband factor of 8, TR/TE of 800/37 ms, a flip angle of 52 degrees, and 2.0 mm isotropic voxels covering the whole brain. Face-name encoding task Participants completed the Face-Name Encoding Task (FaceName) (Sperling et al., 2001) in which they were asked to memorize a series of face-name pairs during the encoding portion of the task and later asked to try to silently remember the face-name pairs during the recall portion of the task. The FaceName task utilized in the HCP-A study was described in (Bookheimer et al., 2019) and was completed in a single 276 s run which included encoding, distractor, and recall blocks. Encoding blocks lasted 22 s each, during which participants were shown five face-name pairs for four seconds each and were instructed to memorize the names paired with each face. Participants were asked to press their index finger on a button when the pairs appeared. A Go/No go task distractor block was presented for 20 s following the encoding block. The recall block followed the Go/No go task distractor block and was 22 s long. Participants were shown the same faces with “???” instead of a name. Participants were asked to press their left index finger button when they remembered the name of the presented face, however, an in-scanner memory test was not performed to minimize head motion. Participants were presented with four blocks of five faces each. Recall was tested after the MRI session. Image processing Data in the HCP-A 2.0 release included imaging data that was preprocessed using the HCP minimal pre-processing pipelines (Glasser et al., 2013; Smith et al., 2013) and fMRI denoising (Glasser et al., 2018, 2019). Images were corrected for gradient nonlinearity distortions, as well as echo planar imaging distortions. To account for subject head motion, images were aligned within and across modalities. Data were aligned to Montreal Neurological Institute (MNI) space using FNIRT, an FSL based software used to register brains to a common space (Andersson et al., 2007). Multi-run spatial ICA + FIX was used to remove artifacts such as those from subject head motion, vascular physiology, or the scanner (Glasser et al., 2018, 2019). Connectivity analyses were performed in-house using CONN toolbox(Whitfield-Gabrieli & Nieto-Castanon, 2012) release 22.a (Nieto-Castanon & Whitfield-Gabrieli, 2022) and SPM (Penny et al., 2011) release 12.7771. Anatomical data were segmented into grey matter, white matter, and CSF tissue classes using SPM unified segmentation and normalization algorithm (Ashburner, 2007; Ashburner & Friston, 2005) with the default IXI-549 tissue probability map template. Functional data were denoised using a standard denoising pipeline (Nieto-Castanon, 2020) and highpass filter (Nieto-Castanon, 2022). Software availability Installation instructions for CONN toolbox are available at: https://web.conn-toolbox.org/resources/conn-installation. Statistical analyses Psychophysiological interaction analyses were used to study the changes in functional connectivity across separate encoding and recall conditions. Seed regions included 132 Harvard-Oxford atlas ROIs (Desikan et al., 2006). A generalized psychophysiological interaction model (gPPI) (Friston et al., 1997; McLaren et al., 2012) was defined separately for each pair of seed and target areas defined by the Harvard-Oxford atlas, Functional connectivity changes across conditions were characterized by the multivariate regression coefficient of the psychophysiological interaction terms in each model. ROI-to-ROI analyses included a total of 17,292 connections among the 132 ROIs included in the Harvard-Oxford Atlas and ROIs were sorted using hierarchical clustering. Connection-level inferences were based on parametric univariate statistics and were a were thresholded using a p<.05 pFDR corrected connection threshold (Benjamini & Hochberg, 1995). Age in years was included as a nuisance covariate. After completing the fMRI portion of the study, participants were immediately tested by HCP-A researchers on recall of the face-name pairs presented during the scanning session. HCP-A researchers presented participants with 10 faces shown during the in-canner portion of the task and they were asked to recall the name they saw paired with the face during the MRI task. The number of face-name pairs correctly identified was recorded. To minimize and standardize the retrieval interval the FaceName task was the last scan performed during the scanning (Bookheimer et al., 2019). For the purposes of our investigation recall accuracy refers to the total number of face-name pairs correctly identified after the MRI. This was an exploratory analysis performed to determine if correlations between E2 and performance were present. A Pearson correlation was also performed to assess the relationship between E2 and age in the present sample due to their previously established relationship during midlife. Follow-up and exploratory analyses Follow-up and exploratory imaging analyses were conducted utilizing the same procedures as those utilized in our primary imaging investigation. ROI-to-ROI analyses included a total of 17,292 connections among the 132 ROIs included in the Harvard-Oxford Atlas and ROIs were sorted using hierarchical clustering. Connection-level inferences were based on parametric univariate statistics and were a were thresholded using a p<.05 pFDR corrected connection threshold (Benjamini & Hochberg, 1995). Age in years was included as a nuisance covariate. Results Participants Participants were a mean age 48.01 years (SD = 4.93) years and a mean education of 17.39 (SD = 2.14) years. Of the total 150 participants, 82 participants identified as White, 41 as Black or African American, 8 as Asian, 12 as more than one race, and 7 were classified as unknown race or not reported. A total of 58 participants were classified as premenopausal, 34 as perimenopausal, 52 as postmenopausal, and 6 as experiencing missed periods. Of the 150 participants included in the present study, a total of 25 participants were missing hormone data. In an effort to preserve the sample size and statistical power, we conducted mean imputations for E2 for participants missing hormone data in each menopause group, including the missed periods group using SPSS software. This involved calculating mean E2 for each group and replacing the missing values with the calculated means. No participants that were missing hormone data self-reported taking a hormonal medication or endorsed having altered menstrual cycles as result of birth control or intrauterine device use. The mean E2 level was 83.83pg/ml (SD = 104.78) prior to imputations and 82.78pg/ml (SD = 97.10) following imputations (Table 1; Fig. 1). Table 1DemographicsParticipants (N = 150)Age48.01 years (SD = 4.93)Years of Education17.39 (2.14) yearsEstradiol (pg/mL)83.83pg/ml (SD = 104.78)Estradiol following imputation (pg/mL)82.78pg/ml (SD = 97.10)FaceName Correct Recall (N = 146)7.49(SD) = 2.467Values represent Mean (Standard Deviation) of demographic and blood draw data of participants Table 1DemographicsParticipants (N = 150)Age48.01 years (SD = 4.93)Years of Education17.39 (2.14) yearsEstradiol (pg/mL)83.83pg/ml (SD = 104.78)Estradiol following imputation (pg/mL)82.78pg/ml (SD = 97.10)FaceName Correct Recall (N = 146)7.49(SD) = 2.467Values represent Mean (Standard Deviation) of demographic and blood draw data of participants Demographics Values represent Mean (Standard Deviation) of demographic and blood draw data of participants Fig. 1Bars represent distribution of estradiol levels (pg/mL) of participants following mean imputations Bars represent distribution of estradiol levels (pg/mL) of participants following mean imputations Participants self-reported medication use. Of the 150 participants, 20 endorsed using hormonal medication including hormonal contraceptives or endorsed that their monthly menstrual cycle was altered by birth control or an intrauterine device. Of those 20 participants, 6 were premenopausal, 2 were perimenopausal, and 12 were postmenopausal. For the 146 participants with FaceName retrieval accuracy data available, the mean number of faces recalled was 7.49 (SD = 2.467) out of maximum of 10 faces. Task-modulated effective connectivity Changes in functional association strength associated with E2 were identified. ROIs with altered functional association strengths associated with E2 during recall included the superior frontal gyrus left, brainstem, and lateral occipital cortex, superior division left. A total of five connections were significant (Fig. 2) during the recall portion of the FaceName task, which included connections between the superior frontal gyrus left and brain stem and the middle temporal gyrus anterior division right and the superior temporal gyrus posterior division right. All five connections that were significant indicated negative functional associations among ROIs during recall were associated E2 (Table 2). No connections were significant during the encoding portion of the FaceName Task. See supplementary materials for this analysis without imputed values. Fig. 2Glass brain representation of brain regions of intertest with significant negative correlations. Five connections displayed altered functional association strength as indicated by a negative correlation between selected regions of interest and represented by the blue lines between functionally connected regions. Colored regions represent regions of interest with altered patters of connectivity associated with E2 level Glass brain representation of brain regions of intertest with significant negative correlations. Five connections displayed altered functional association strength as indicated by a negative correlation between selected regions of interest and represented by the blue lines between functionally connected regions. Colored regions represent regions of interest with altered patters of connectivity associated with E2 level Table 2Significant connections in task-modulated effective connectivity analysisConnectionTp uncorrectedp FDRSuperior Frontal Gyrus Left (blue) – Brain Stem (orange)−4.63< 0.00010.0494Middle Temporal Gyrus, anterior division Right (yellow) – Superior Temporal Gyrus, posterior division Right (mauve)−4.60< 0.00010.0494Cerebellum 6 Left (red) – Cerebellum 6 Right (light orange)−4.50< 0.00010.0494Cerebellum 8 Left (dark green) – Lateral Occipital Cortex, superior division Left (light green)−4.50< 0.00010.0494Cerebellum 6 Left (red) – Cerebellum Crus 2 Right (light orange)−4.49< 0.00010.0494Significant connections among regions of interest associated with E2 (p<.05 p-FDR corrected) at a connection level during the recall portion of the task. All significant connections showed a negative correlation between regions of interest. Indicating lower task-modulated effective connectivity associated with E2 Table 2Significant connections in task-modulated effective connectivity analysisConnectionTp uncorrectedp FDRSuperior Frontal Gyrus Left (blue) – Brain Stem (orange)−4.63< 0.00010.0494Middle Temporal Gyrus, anterior division Right (yellow) – Superior Temporal Gyrus, posterior division Right (mauve)−4.60< 0.00010.0494Cerebellum 6 Left (red) – Cerebellum 6 Right (light orange)−4.50< 0.00010.0494Cerebellum 8 Left (dark green) – Lateral Occipital Cortex, superior division Left (light green)−4.50< 0.00010.0494Cerebellum 6 Left (red) – Cerebellum Crus 2 Right (light orange)−4.49< 0.00010.0494Significant connections among regions of interest associated with E2 (p<.05 p-FDR corrected) at a connection level during the recall portion of the task. All significant connections showed a negative correlation between regions of interest. Indicating lower task-modulated effective connectivity associated with E2 Significant connections in task-modulated effective connectivity analysis Significant connections among regions of interest associated with E2 (p<.05 p-FDR corrected) at a connection level during the recall portion of the task. All significant connections showed a negative correlation between regions of interest. Indicating lower task-modulated effective connectivity associated with E2 Estradiol and facename task performance The results of the Pearson correlation analysis showed no relationship between E2 and FaceName task performance (r(144) = −0.004, p =.965). E2 level was correlated with age in years (r(148) = −0.346, p <.001). FaceName performance was correlated with years of education (r(144) = 0.185, p =.025). Exploratory imaging analyses investigating the relationship between out of scanner recall performance and connectivity were also conducted. No significant associations were identified during encoding or recall. Discussion Task-modulated effective connectivity during an episodic memory task was related to E2 level during midlife in females. E2 level was associated with negative functional associations between brain regions during the recall portion of the FaceName task, highlighting the potential role of E2 underlying episodic memory functioning and episodic memory network within the brain. This was consistent with previous studies that found that E2 modulated effective connectivity (Derntl et al., 2024). We controlled for age in our imaging analysis, strengthening the argument that the observed differences in functional association strength between ROIs during recall were the result of E2 level rather than an effect of chronological age. One potential interpretation of our finding is that at higher levels of E2 resulted in more efficient completion of the recall task. Previous work has established the positive effects estrogens have on cognition (Ycaza Herrera & Mather, 2015). It may be the case that with higher levels of E2 present, strong functional associations between brain regions associated with episodic memory were not required to accomplish the task. Episodic memory function is supported by brain regions within the temporal and frontal cortices humans (Dickerson & Eichenbaum, 2010). Our results indicated negative functional associations among ROIs in these cortices associated with E2 without subsequent associations with task performance, suggesting a potential increase in efficiency. In our analysis ROIs with altered functional association strengths as a result of task modulation and E2 level included the superior frontal gyrus (SFG), middle temporal gyrus (MTG), and superior temporal gyrus (STG). Alterations to underlying functioning during an episodic memory task to the SFG and regions within the temporal lobe are consistent with their established functions. The ventrolateral temporal, medial parietal, lateral parietal, medial frontal, and lateral frontal regions have all been found to support episodic memory function in humans (Dickerson & Eichenbaum, 2010). Associations between episodic memory and the temporal and frontal cortices have also been previously found to be related to E2 (Schroeder et al., 2025). E2 concentrations have also previously been associated hippocampal connectivity (Jacobs et al., 2016). It is unclear why the hippocampus was not among the ROIs identified as having altered functional association strengths associated with E2 level in our study. It may have been the case that the hippocampus was consistently recruited during the recall portion of the task and that the range of E2 levels present in the current task were not sufficient to influence its recruitment in this instance. Task-modulated effective connectivity during recall was also altered between multiple subregions of the cerebellum. Previous studies of menopause-related changes to cerebellar connectivity identified lower cerebello-striatal and cerebello-cortical connectivity at rest in post- compared to premenopausal females (Ballard et al., 2022). The same study also found that postmenopausal females had lower connectivity within the cerebellum compared to their premenopausal counterparts (Ballard et al., 2022). Taken together, the results of prior studies, and those described in the current study emphasized the importance menopause as a brain aging event that influenced cerebellar connectivity. Differences in functional connectivity associated with E2 level may be related to underlying structural changes associated with brain aging. Reduced grey matter volume in brain regions involved in cognition in perimenopausal women compared to age-matched men have previously been identified (Mosconi et al., 2021). However, previous studies have found that grey matter volume changes in brain regions that support cognitive functioning stabilize or recover after menopause, with recovery of grey matter volume post menopause being related cognitive preservation (Mosconi et al., 2021). It may be the case that our findings are indicative of functional changes occurring across menopause stages in conjunction with underlying structural changes in brain regions that support episodic memory function. In the present study, E2 did not have effects on recall performance out of the scanner as it did on connectivity observed during the MRI. The relatively good task performance of most participants included in the present study may have obscured performance-based associations with E2. Roughly 30% of participants with performance data available had perfect performance on the task. It may be the case that differences in the strengths of functional associations during the recall condition as a result of E2 levels would have been reflected in task performance in individuals who performed poorly with the FaceName task, rather than in those who performed well. Previous studies have investigated connectivity and episodic memory performance. One study found higher associations between anterior cingulate cortex and hippocampus correlated with performance on a test of episodic recall, Buschke Selective Reminding Test in postmenopausal participants (Seitz et al., 2019). If a more normal distribution of performance scores was present in the current study, we may have observed similar relationships between estradiol and recall performance. E2 was not associated with alterations in functional association strength between brain regions during the encoding portion of the FaceName task. This suggests that the relationship between E2 level and functional association strengths during an episodic memory task differs between encoding and recall. Future studies should investigate if the differences in connectivity during encoding and recall persist later in life once E2 has stabilized at a lower level. Additionally, future studies should investigate the relationship between other hormones which fluctuate during midlife, including progesterone and follicle stimulating hormone, and connectivity during an episodic memory task. Progesterone levels are significantly lower after menopause (Henderson, 2018); however, even at lower levels progesterone has been found to be positively correlated with verbal memory and global cognition in postmenopausal women within six years of menopause (Henderson et al., 2013). Follicle stimulating hormone levels also differ based on menopause stage, during perimenopause follicle stimulating hormone levels increase and remain elevated after menopause (Harlow et al., 2012). Due to the fluctuating levels of multiple hormones during midlife it is possible that interactions between hormones such as E2, progesterone, and follicle stimulating hormone contributed to our results. Future studies should investigate the relationship between progesterone and follicle stimulating hormone levels during midlife and connectivity during an episodic memory task. The present study was limited by the between-subjects nature of its design. Individual differences in functional association strength between brain regions during an episodic memory task may exist between participants. A within subjects’ design would have allowed us to observe the effects of different levels of E2 on individual participants and compare them over time. The HCP-A is a longitudinal study and future data releases will include data from the 150 participants included in the present study; however, follow-up data were not available at the time of this analysis. Additionally, HCP-A is a multisite study, and it may be the case that the site data was collected from played a role in out finding. Site was not included as a confounding variable at the time of this investigation due to the standardization protocols across HCP-A data collection sites. Similarly vascular confounders such as cerebrovascular reactivity and cerebral blood flow were not assessed. Lastly, we chose to include both participants taking exogenous hormone and those who were not. It is conceivable that the relationship between connectivity during the FaceName task and exogenous and endogenous hormone differs and warrants further investigation. Furthermore, the inclusion of participants taking exogenous hormones may have skewed the data utilized when performing the mean imputations. Conclusions Our results indicated that task-modulated effective connectivity between brain regions was related to E2 level during midlife in females. Patterns of functional connectivity were modulated by the recall blocks of the FaceName task, a test of episodic memory, associated with E2 levels. Our findings regarding SFG, MTG, and STG connectivity were in line with previous work which demonstrated the importance of these regions for memory functioning (du Boisgueheneuc et al., 2006; Taing et al., 2021). Changes in connectivity among these regions may have long term effects on the aging process that extend beyond midlife. Our findings regarding cerebellar connectivity highlighted the importance of E2 in cerebellar functioning during midlife. Future studies should investigate if differences in task-modulated effective connectivity between encoding and recall persist following midlife. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 (DOCX 279 KB) Supplementary Material 1 (DOCX 279 KB)