Introduction

Yoga is prevalent among adults in Western societies, where it is often perceived as a complementary form of exercise. According to a nationally representative survey of US adults in 2017, 14.5% had engaged in yoga in the past 12 months (Wang et al., 2019). There is growing consensus among researchers regarding the need to explore the effects of different styles of yoga (Cowen & Adams, 2005; Larkey et al., 2009; Riley & Park, 2015; Ross & Thomas, 2010). Some factors that differentiate these yoga styles are speed of movement, difficulty of postures, focus on flow and transitions between the movements, and length of time that poses are held (Morone & Greco, 2007). There is preliminary evidence suggesting that the physical and physiological benefits of yoga differ by style. A randomized controlled trial (RCT) compared Hatha to Ashtanga yoga among young adults, in which the Ashtanga yoga intervention included sun salutations, holding postures that build strength, breathing exercises and relaxation. Both groups showed improvements in strength and flexibility, yet only the Ashtanga group improved in perceived stress and diastolic blood pressure (Cowen & Adams, 2005). Ashtanga yoga is among the styles that focus on flow-based, moderate to high intensity postures and sequences. The description of the intervention indicates that the postures may have been fairly difficult; however, the intensity and perceived exertion of the intervention was unspecified.

Sun salutation is another form of flow-based moderate intensity yoga sequence. It involves a sequence of 10–12 postures, completed in quick succession and accompanied by breathing that is synchronized with each movement. The sun salutation sequence comes from Hatha yoga, which is suitable for beginners, and its variations are also found in Ashtanga and Vinyasa yoga (which focuses on flow-based movements). Only two studies have tested the potential cognitive benefits of sun salutation, in young adults (Goud Kondam, 2017; Phansikar & Mullen, 2022). However, in the study by Goud Kondam (2017), sun salutation was one part of the intervention that included other yoga postures. While the intensity of the intervention was not described, participants were instructed to perform the sun salutations slowly. The sun salutation lasted for 30 min and the remaining yoga intervention, which is not described, was done for an additional 30 min. In our previous study, we found preliminary evidence for feasibility and favorable effects on psychosocial functioning after engagement in a flow-based yoga intervention guided by a video (Phansikar & Mullen, 2022). Specifically, low-active adults (18–45 years of age) with symptoms of stress, were randomly assigned to 12 rounds of sun salutation and breathing exercises instructed via iPad, treadmill walking at moderate intensity (55–70% heart rate maximum), or an attentional video control condition. The sun salutation intervention consisted of the 12 rounds of the sun salutations sequence performed in Hatha yoga, and concluded with guided breathing and relaxation. The intensity of the intervention was light to moderate (mean heart rate = 100 beats per minute). The sun salutation group showed significant improvements in stress, state anxiety and self-efficacy for doing yoga postures, as well as perceived focus of attention, when compared to the other trial arms. In addition, the majority of the participants found the video to be clear and understandable, and an easy modality to learn the poses, but suggested a larger screen display. There was no statistically significant effect of the intervention on cognitive functioning, but the mean scores were trending in the expected direction. It is likely that the null effects on cognitive functioning may have been due to the short duration or lower intensity of the intervention. From the studies mentioned above, the intervention in our current study is most similar to our acute study. The difference is that the current study involves sun salutations A and B which include more difficult postures than Hatha yoga sun salutation; is longer in duration (50 min vs 30 min); and is designed to be of moderate intensity. The common aspect of our intervention with the prior literature, is the inclusion of guided breathing and relaxation, whereas the primary difference is that our intervention is designed to be of moderate intensity.

While Hatha yoga remains widely studied among healthy populations, the longitudinal cognitive and psychosocial benefits of such moderate-intensity forms of yoga have not been tested. Moreover, the feasibility of delivering a remote moderate intensity yoga intervention focusing on metrics of safety, enjoyment and adherence have also not been thoroughly explored. Utilization of digital health technology may be advantageous for increasing accessibility to yoga via at-home, instruction-on-demand. Studies have used synchronous (e.g., real-time videos or videoconferencing) and asynchronous (e.g., pre-recorded videos) technology to facilitate home-based yoga interventions. A side effect of the Covid-19 pandemic has led to an increase in remote practice of yoga (Gothe & Erlenbach, 2021; Jasti et al., 2020). Even though studies have been done with synchronous and asynchronous delivery methods, these delivery methods are still not a widely used format for conducting yoga interventions.

The purpose of the present study was to investigate the feasibility of a remote, moderate-intensity sun salutation yoga intervention, delivered via synchronous live sessions and asynchronous [self-guided] videos, and to examine potential psychosocial and cognitive effects. Feasibility was assessed through tracking of adverse events, participant reported enjoyment, and attendance records associated with both delivery formats.

Methods

Sample

Participants (n = 86) were recruited from a mid-Western University town, through email list-servs distributed to University staff and employees, and a database of community members maintained by the lab. Sample size was calculated apriori using G-power analysis (Faul et al., 2007), accounting for type of analysis, type I (alpha level of 0.05) and type II error (80% power), and medium effect size (0.50). The required sample size was 68. We also accounted for 20% attrition for a final apriori sample size of 88 participants. Participants were full-time working adults (at least 39 h per week), between the ages of 18 and 64, and low active. The criteria of low-active and inexperience with yoga was met when participants self-reported engaging in physical activity (e.g., going to the gym or walking), or similar activities (e.g., mindfulness and tai-chi), for two or fewer times per week for 30 min over the past three months. Also, participants had to not be engaged in yoga regularly in the past 10 years, able to exercise at a moderate intensity, have access to internet and a laptop/computer, and be free from exclusionary conditions listed on the Physical Activity Readiness Questionnaire (Chisholm et al., 1975). Additionally, participants had to self-report having at least three of six stress symptoms from the Generalized Anxiety Disorder checklist in the Diagnostic and Statistical Manual V (American Psychiatric Association, 2013), during the screening phone call. Participants were excluded if they had any self-reported neurological or psychiatric disorders, diagnosed hypertension, were unwilling to wear a wrist-worn accelerometer—Fitbit Charge 3 (Fitbit Inc, USA), or were unwilling to be randomized into one of the two groups. The study was approved by the Institutional Review Board (IRB) of the University and pre-registered at ClinicalTrials.gov. All participants signed an IRB approved informed consent document prior to participation in the study. The study CONSORT diagram is presented in Fig. 1.

Fig. 1
figure 1

CONSORT

Full size image

Procedure

At baseline, participants completed online questionnaires (~ 30 min) assessing demographic factors, physical activity, and psychosocial functioning. This was followed by a 30-min appointment with a staff member via Google Meet, to complete the cognitive assessments. Following this, participants picked up a Fitbit device to be worn throughout the study, from the laboratory. The Fitbit was partially configured prior to pick-up (profile setup involving age, gender, and self-reported height and weight). The initial 7-day wear period, prior to the intervention, served as participants’ baseline physical activity. Participants who completed baseline assessments were randomly assigned to a yoga group or a waitlist control group. Randomization was done using block-order stratification through Microsoft Excel, with an allocation ratio of 1:1. The block size was 10, and stratification included age and sex. Group allocation was revealed to participants only after completion of all baseline assessments, via individual emails. The person conducting randomization was not involved in conducting assessments or the intervention. Two follow-up assessment were administered, one at mid-point of the intervention, i.e. end of week four (a 10-min survey on stress, affect, self-efficacy, and self-regulation), and one at the conclusion of the 8-week intervention (full baseline battery was repeated). All aspects of the study, except for Fitbit pick-up and drop-off, were conducted remotely.

Yoga intervention

Participants were asked to engage in 50-min yoga sessions three times/week over an eight-week period. Supervision was titrated from remote individual and group instructional sessions to self-guided video sessions. All synchronous sessions were conducted remotely on Zoom (max capacity of 10) with a certified and trained yoga teacher and a trained staff member for supervision. The trained staff member observed all participants throughout the session. If one or more participants performed an incorrect posture, corrective cues would be provided (e.g., remember to tuck in your elbows). If any errors persisted, participants would be requested to stay online to meet with the instructor briefly at the conclusion of the group session to receive further guidance to improve technique and reduce chances of injury. Class recordings were provided for self-guided sessions. The yoga intervention primarily consisted of the sun salutation sequence from Hatha yoga and sun salutations A and B often performed in Ashtanga and Vinyasa yoga. Details about the intervention content and delivery method for each week is shown in Table 1. The poses and movements were done for 35–40 min, followed by 5–7 min of breathing and 5 min of guided relaxation. Participants logged the start and end of their yoga sessions on the Fitbit device, and filled out a one-minute post-session log recording their heart rate, exertion, stress, and affect.

Table 1 Weekly content of the yoga sessions
Full size table

Waitlist control group. Participants in the waitlist control group were asked to carry on with their lives as usual, and refrain from starting any new physical activity practice. Participants also wore a Fitbit throughout the duration of the study, enabling us to track their step count. They were given access to the intervention videos at the end of the study.

Measures

Demographics

Participants completed a baseline survey about their demographic information.

Cardiorespiratory fitness

Cardiorespiratory fitness was estimated using Jurca et al.’s equation (Jurca et al., 2005) with a higher score indicating higher estimated cardiorespiratory fitness: (Gender (0 for women, 1 for men) × 2.77) − (Age in years × 0.10) − (Body mass index × 0.17) − (Resting heart rate × 0.03) + score from the physical activity question + 18.07. Participants were asked to report their usual level of aerobic physical activity from one (no activity) to five (highly active).

Feasibility

Assessment of feasibility included attendance and adherence, dropouts from the study, adverse events, program evaluation and enjoyment. Attendance for the supervised sessions was recorded by the yoga instructor. Adherence to self-guided sessions was assessed via Fitbit-logged yoga sessions, and post-session logs. Program enjoyment and evaluation questions were asked only to the yoga group. Enjoyment was assessed using the revised Physical Activity Enjoyment Scale (Massey et al., 2021) on a 7-point bipolar Likert rating scale, with a minimum score of 0 and maximum score of 49. Program evaluation included questions about various aspects of the video instruction to understand participants’ perception about learning yoga through videos, rated on a 7-point Likert rating scale (strongly agree-strongly disagree). For example, “The video was clear and understandable.”

Perceived stress and anxiety

The Perceived Stress Scale (PSS; Cohen et al., 1983) is a 10-item scale that measures the extent to which situations are perceived as stressful. Two composite scores, assessing stress reduction (lower scores indicating stress reduction) and counter-stress/co** (higher score indicating better co**) were calculated by summing the relevant items, based on prior research showing a two-factor solution (Golden-Kreutz et al., 2004). The Brief Job Stress Questionnaire (BJSQ; Kawada & Otsuka, 2011), is a 15- item scale measuring job demands, job control, and support. Total score was calculated by summing all items of each subscale after required reverse coding. Higher scores indicated a lower level of job stress. The Visual Analogue Scale for stress is a one-item measure asking participants to rate their perceived stress in the moment, on an unmarked ruler with endpoints labelled ‘none’ (0) and ‘as bad as it could be’ (100), with higher scores indicating a higher level of perceived stress. The Hospital Anxiety and Depression Scale (HADS) (Zigmond & Snaith, 1983) is a 14-item scale measuring depression and anxiety. Total score was calculated by summing responses on the seven items measuring anxiety. A total score of 0–7 indicated ‘normal’, 8–10 indicated ‘borderline abnormal’, and 11–21 indicated ‘abnormal’. State anxiety was measured using the state anxiety subscale (20 items) of The State Trait Anxiety Inventory (STAI; (Spielberger, 2010). A total score was calculated by summing the items (after required reverse-coding). Higher score indicated a higher level of anxiety.

Executive functioning

Participants completed tasks assessing executive functioning while on a Google Meet video call, in the presence of a study staff member. For the two tasks completed on Psytoolkit website (Stroop task and Task switching paradigm), participants were required to share their screen with the staff member while doing the task, for supervision. For the other two tasks which entailed a verbal response from the participant, the staff member shared their screen to present the stimuli. The staff member recorded the participants’ verbal responses on a scoring sheet.

Administration of cognitive tasks virtually is increasing after the COVID-19 pandemic. There is evidence for the reliability and validity of this delivery method, for some tasks. For example, Kim et al (2019) found no significant differences in performance metrics (response choice and time) in a psycholinguistic task assessed via Psytoolkit or Eprime (the original, validated software). In another study, few measures from the Weschler Adult Intelligence Scale that are administered verbally were validated for virtual administration, with the participant undergoing these assessments from home (Fox-Fuller et al., 2021).

Stroop task Stroop task measures response inhibition or response interference control. It was administered via Psytoolkit (Stoet, 2010, 2017). The task included a series of color words (e.g. blue, red) that were either congruent or incongruent with the color in which the word was printed. Participants were asked to respond to the print color, using the keys ‘b’, ‘n’, and ‘m’, corresponding to blue, red, and green colors respectively. Reaction time was recorded. A cost score was calculated, with shorter cost scores indicating better performance.

Task switching paradigm This is a measure of cognitive flexibility, i.e., a participant’s ability to adapt quickly to changing rules and the cost associated with it. Participants had to categorize the stimuli presented according to shape (circle or rectangle) or color (blue or yellow), with a switch in the rules occurring every two trials. Participants were cued to the rule before the stimulus was presented. Participants responded with ‘b’ key corresponding to circle or yellow, and ‘n’ key corresponding to rectangle or blue. Reaction time and accuracy were recorded, with lower reaction time and higher accuracy indicating better performance.

Digit span forward and backward This is a measure of the storage and manipulation capacity of working memory (Wechsler, 2008). Participants were shown a series of digits on a computer screen. The minimum length of digits presented was 3 (e.g., 638) and the maximum length was 9 (e.g., 628,295,702). In the forward task, participants were asked to recall the digits exactly as they were presented, while in the backward task, they had to recall them in the reverse order. Accuracy was recorded, with higher accuracy indicating better performance.

Digit symbol substitution test (DSST) DSST is a task measuring processing speed. Participants were shown a code-key in which every digit matches a particular symbol (Hodes et al., 2013). Then, they were shown a series of symbols on the screen and participants had to say the corresponding digit aloud, as quickly and accurately as possible. Participants were given 120 s to complete as many digit-symbol pairings as possible. Accuracy was recorded, with higher scores indicative of better performance.

Statistical analysis

Data were analyzed in SPSS v.27. Data were checked for outliers (± 2.5 standard deviations from the mean) and no outliers were present. Feasibility data were summarized using descriptive and univariate statistics. All data was analyzed with intention to treat. Three participants did not complete post-testing assessments, and data for them was imputed using multiple imputation. A 2 (Group) × 2 (Time) repeated measures analysis of covariance (ANCOVA) was conducted to identify differences in executive functioning between the two groups from baseline to the 8-week follow-up. A p-value of 0.05 or less determined statistically significant differences. Similarly, 2 × 2 ANCOVAs were conducted to assess difference in stress and anxiety outcomes, separately. Partial eta squared values were used to determine the magnitude of each effect size (0.01 = small, 0.06 = medium, 0.14 = large) (Olejnik & Algina, 2000). Covariates used in the analysis were cardiorespiratory fitness, Fitbit steps at baseline, past yoga and exercise experience (number of months), and education (only for cognitive variables).

Results

Participants (n = 86) were mostly female (81.4%), non-Hispanic White (61.62%) with an advanced college degree (51.2%). The mean age was 41 years (SD = 10.60). Baseline characteristics are shown in Table 2, with stratification by trial arm.

Table 2 Baseline characteristics of the sample
Full size table

Feasibility

The overall attendance across the intervention was 82%. This implies that on average, participants attended 2.46 sessions per week. Attendance (see Table 3) was much higher for synchronous supervised sessions led by the instructor (97.15%) as compared to self-guided sessions (63.50%). Attendance for each week is reported in Table 3, along with mean heart rates. Overall, all the participants enjoyed the study (70.7% reported yes, 29.3% reported sometimes, and 0% reported no). The mean score on the Physical Activity Enjoyment Scale was 31.13. Most participants (56.5%) sometimes found the study to be a burden on their life, 39% did not find the study to be a burden at all, and 4.9% found it to be a burden. All of the participants would recommend the program to family or friends (75.6% reported yes, 24.4% reported sometimes). A majority of the participants (78%) liked the video instruction for yoga, 17% liked it sometimes, and 4.9% did not like the instruction. On average, participants found the yoga videos to be clear and understandable [M = 6.44 (0.95)], easy to follow [M = 6.34 (0.1.06)], easy to learn the poses through the video [M = 6.20 (0.1.26)], and would use videos to continue engaging in yoga at home [M = 6.24 (1.39)].

Table 3 Average attendance and heart rates for yoga sessions per week
Full size table

Two participants officially withdrew from the study, due to family emergency and low adherence to the study. One person was lost to follow-up due to a family emergency. There was one adverse event in the yoga group potentially related to the intervention. The participant reported feeling pain in their upper back, and although the source was unknown, it corresponded with a yoga session at which point the pain worsened and caused discomfort in their daily life. The participant did not engage in yoga after the adverse event but remained in the study and completed follow-up assessments. Another participant from the yoga group was injured but it was unrelated to the study (fall down the stairs).

Stress and anxiety

There was a significant group × time interaction effect on the stress reduction composite of the PSS F(1, 80) = 4.038, p = 0.048, ηp2 = 0.048 in favor of the yoga group, but not on the other counter-stress composite F(1, 80) = 1.327, p = 0.253, ηp2 = 0.016 (see Table 4 for mean scores). There was no significant group × time interaction on job demands F(1, 80) = 1.534, p = 0.219, ηp2 = 0.019; job control F(1, 80) = 0.125, p = 0.724, ηp2 = 0.002; or support F(1, 80) = 1.651, p = . 203, ηp2 = 0.020. There was a significant group × time interaction effect on the stress visual analog scale with a large effect size F(1, 80) = 21.458, p < 0.001, ηp2 = 0.211, with the yoga group reducing in their perceived stress. Additionally, there was a significant difference from baseline to week four on the stress reduction composite of PSS F(1, 80) = 4.635, p = 0.034, ηp2 = 0.05 and the stress valence scale F(1, 80) = 18.571, p < 0.001, ηp2 = 0.188, favoring the yoga group. There was a significant group × time interaction effect from baseline to 8-week follow-up on the state anxiety subscale of STAI with a medium effect size F(1, 80) = 5.056, p = 0.027, ηp2 = 0.059. The yoga group had significantly lower state anxiety post-intervention as compared to the control group. There was a significant group × time interaction effect on HADS-A with a large effect size F(1, 80) = 10.304, p = 0.002, ηp2 = 0.114. The yoga group had significantly lower trait anxiety post-intervention and had a score in the normal range (< 7) as compared to the control group.

Table 4 Cognitive and psychosocial measures by group and time
Full size table

Executive functioning

The group × time interaction effect from baseline to 8-week follow-up was statistically significant on digit span forward F(1, 79) = 4.238, p = 0.043, partial ƞ2 = 0.051 and digit span backward F(1, 79) = 5.679, p = 0.020, partial ƞ2 = 0.067. The yoga group had higher accuracy as compared to the control group (see Table 4 for mean scores). There was no significant group × time interaction effect on digit symbol substitution task F(1, 79) = 0.398, p = 0.530, partial ƞ2 = 0.005, Stroop cost score F(1, 79) = 0.458, p = 0.501, partial ƞ2 = 0.006. and task-switch cost score F(1, 79) = 0.040, p = 0.842, partial ƞ2 = 0.001.

Discussion

This study tested the feasibility of an 8-week fully remote (live-streamed and recorded videos) moderate intensity yoga intervention for reducing stress and anxiety and improving cognitive functioning, among working adults facing symptoms of stress, during a pandemic. Overall, our data suggest that the intervention was relatively safe. Adherence was high and a majority of participants enjoyed the study and would recommend it to others. Participants liked the supervised video instruction as well as the recorded videos. The yoga intervention resulted in a decrease in perceived stress and anxiety, and an increase in working memory capacity.

Feasibility

Even though the intervention was mostly safe, the one adverse event (e.g., back pain) early on, potentially related to the intervention, underscores that for low-active adults, yoga may still pose a risk to unconditioned physiological systems, joints and musculature. These incidents are still possible, even when postures and transitions are introduced via slow progressions led by a qualified instructor. Instructors should also include sufficient warm-ups and provide warnings of possible exacerbation of pre-existing conditions at the onset of any new yoga program. Adherence was high and a majority of participants enjoyed the study and would recommend it to others. Participants liked the supervised video instruction as well as the recorded videos.

Not surprisingly, there were differences in the attendance rate between the supervised and self-guided sessions. Our yoga intervention varied in program length, frequency and session duration compared to other studies in the literature and involved a hybrid approach towards synchronous and asynchronous delivery. These differences may account for some of the discrepancies from published attendance rates of similar studies, however there is little attendance data on synchronous vs. asynchronous delivery. In a study comparing synchronous vs. asynchronous delivery of a mindfulness intervention, adherence rates were similar across the groups (M = 63%), but completion rate of the intervention was lower for the asynchronous group (41% vs. 19%) (Allexandre et al., 2016). The degree of adherence to these programs ultimately will depend on the priorities, needs, motivation, and self-regulatory resources available to the participants.

In theory, performing sessions on one’s own would require a higher level of self-efficacy and self-regulation. Loss of instruction, feedback and encouragement provided by the instructor may have resulted in lower self-efficacy. For our sample of full-time working adults, unique challenges in terms of balancing work, personal life, childcare responsibilities, and being time-pressed for engaging in physical activity may have resulted in lower attendance for self-guided sessions (Hoare et al., 2017). Indeed, while participants enjoyed the intervention, they found the overall study to be a burden on their life, sometimes. Future studies can investigate how many and which combination of behavioral change strategies to use, to maximize engagement in self-guided yoga interventions. Individualized or targeted interventions can be developed, using behavior change techniques according to participant characteristics such as demographics and baseline fitness levels and delivery method of the intervention (Carraça et al., 2021).

Stress and anxiety

We found that participants in the yoga group (vs. control) had a statistically significant reduction in perceived stress and general and state anxiety. Previous studies have shown similar results. An acute sun salutation intervention of 30 min resulted in a decrease in state anxiety (Phansikar & Mullen, 2022). In studies testing Dru yoga or “power yoga” interventions (which focus on movement sequences different than the present study) participants decreased stress and anxiety as compared to a control group (Hartfiel et al., 2012; Maddux et al., 2018). Additionally, there is growing support for stress and anxiety reduction following low-intensity Hatha yoga interventions, with a moderate to large effect size (Zoogman et al., 2019). Our results corroborate these findings and provide evidence that moderate intensity yoga interventions have similar effects on stress and anxiety reduction, with a moderate to large effect size. These results are of substantial value given that there was an approximate increase of 25% in the prevalence of mental health conditions due to the pandemic (Taquet et al., 2021), and a reduction in human mobility was an independent predictor of an increase in these mental conditions. However, we did not see any significant changes in factors related to job stress. This may be because the questionnaire is not validated among a sample of US adults and the questions pertain to aspects of the work and work environment, such as having control over projects and timelines, and support from peers, rather than one’s response to the stressors.

The stress and anxiety reduction from yoga practice may occur from several mechanisms. One such mechanism is via greater interoceptive awareness, i.e. focus on internal states, that is facilitated by focusing on breathing patterns during movement. In our study, participants reported having an internal focus of attention throughout the intervention (Ms ± SD = 7.65 ± 2.66) on a one-item scale assessing focus of attention from 0 (external) to 10 (internal). Interoceptive awareness affects brain areas that appear to regulate emotional experience and result in greater attention to the present moment (Farb et al., 2013). An eight week Kripalu yoga intervention increased interoceptive awareness which correlated with stress reduction, among adults between the ages of 23–67 (Park et al., 2021). In line with this, Gard et al (2014) proposed a theoretical framework of how yoga helps develop emotional self-regulatory pathways through both top-down and bottom-up processing, and an interaction of these processes contributes to better emotional self-regulation and stress reduction in daily life. Other potential mechanisms include an increase in mindfulness (Park et al., 2021), adaptations to the musculoskeletal system (Francis & Beemer, 2019), positive psychological well-being (Park et al., 2021), and physiological mechanisms including reduction in cortisol (Pascoe & Bauer, 2015).

Executive functioning

We found that participants in the yoga group had a statistically significant improvement in working memory forward and backward performance, as compared to the control group. Whereas the improvement in Stroop task and task switching were not statistically significant, reductions in reaction time were in the expected direction. Previous studies have shown that yoga is associated with an improvement in cognitive functioning (Gothe & McAuley, 2015; Luu & Hall, 2016). However, both meta-analyses pointed out that there is insufficient evidence because of small sample sizes and poor quality of studies. The impact of yoga on cognitive functioning has been more widely studied among clinical populations and clinical or healthy older adults with a mean age of > 60 years, and not among healthy young to middle- aged adults (Chobe et al., 2020; Hoy et al., 2021). The null effects observed in our 8-week remote yoga study, with a mean age of 41 years, may be attributable to the relatively younger age of our sample. Additionally, most research has been done on low-intensity Hatha yoga postures which has shown positive cognitive effects (Gothe & McAuley, 2015; Luu & Hall, 2016). The effects of moderate-intensity yoga may differ, but limited research is available for comparison, and no study has compared cognitive effects of varying intensities of yoga on the same sample.

We found a positive effect on working memory, but no effect on other domains. In a review of aerobic, resistance exercise, and mind–body exercises and their effect on cognitive functioning, tai-chi improved working memory, but not other types of functioning (Northey et al., 2018). It may be that working memory is most responsive to mind–body or cognitively engaging interventions. Indeed, our intervention required participants to remember sequences of postures that changed every week and to remember various dynamic modifications and additions to each posture in the sequence, which may have impacted their working memory more than the other constructs.

It is important to note that in most yoga and cognition studies, assessments of cognitive functioning were conducted in the lab. Our study is one of the few to conduct cognitive assessments at home, in a person’s naturalistic environment. This may have affected participants’ performance, especially on reaction time tasks. A study showed that reaction times were slower on a web-based version of the Stroop task as compared to the standardized version (Backx et al., 2020), and that the particular software and hardware of the computer on which the task in administered may affect reaction times (Calcagnotto et al., 2021). Another study by showed no difference in reaction time delivered via Psytoolkit at home vs. an E-Prime task delivered in the laboratory, however, they used a psycholinguistic choice reaction time task (Kim et al., 2019). Overall, insufficient research is available to determine whether a more naturalistic environment influences reaction time in higher-order tasks. Future studies should seek to refine protocols for conducting remote online assessments of cognitive functioning. Additionally, researchers should closely examine the impact of mind–body interventions on working memory vs. other domains of cognitive functioning and test the unique features of these physical activity interventions which may impact working memory.

The findings of this study should be viewed with caution due to several limitations. One of the major limitations is the lack of an active control group, limiting the comparison of our intervention with other activities. In addition, the study sample was homogenous as participants were mostly female, non-Hispanic White, and employees at the local university. Our recruitment efforts were restricted to email list-serv distributed to university faculty and staff and a database of contacts, from constraints imposed by the pandemic. Another limitation is the possibility of demand characteristics and social desirability bias, as it was not possible to mask study assignment for the participants. While we did not reveal the specific purpose of the study, but they may have been able to guess it from exposure to our assessments. Another limitation to consider is the use of post-session logs for reporting heart rate and as one of the methods for tracking adherence. We could not ensure that participants logged their sessions (e.g., heart rate data) in a consistent or timely manner. Additionally, it is important to keep in mind that cognitive tasks were administered virtually and some factors differed from the standard administration of the tasks. For example, digit span is usually administered verbally/auditorily by reading out the stimulus set, whereas it was presented visually in our study. In spite of this, the nature of the task still relies on the same cognitive domain broadly (working memory). Future research will need to disentangle if there are important implications given the task differences during virtual administration, as well as investigate other confounding factors. For example, we do not know whether the screen size of the participants’ device may have played a role while reading on-screen stimuli. Lastly, we did not assess fluctuations in social or environmental factors as a function of study participation as it was beyond the scope of the study. At post-intervention, there was no significant difference in perceived loneliness due to the pandemic between the yoga and control group [Ms (SD) = 4.72 (1.20) vs 4.78 (1.55)], making it unlikely that isolation affected the results of our study. Future studies may investigate how a virtual yoga intervention may affect social factors.

Our study has several strengths. This study was pre-registered (via Clinicaltrials.gov) single-blind randomized controlled trial, which is among the most rigorous scientific methods. The interventionist was a certified yoga instructor, who trained in the specific yoga protocol used in this study, for over six months. The remote delivery of the intervention and the assessments was a strength of this study. Especially during the times of a pandemic, providing an intervention remotely to participants meant no exposure to other people and ensuring the safety of one’s health. Additionally, they did not have to face barriers such as access to transportation, time spent traveling, and pandemic-related regulations. The intervention was delivered over Zoom, a mobile and web-based application, which most participants had already been using prior to the intervention. The intervention facilitated autonomy and encouraged self-regulation, by transitioning from individualized sessions to completely self-guided sessions. This improves the study’s ecological validity, as it was not a tightly controlled lab experiment and allowed participants to engage in physical activity on their own time within their home or other natural environment. Other strengths of this study include the use of a Fitbit as an objective method to track attendance, heart rate after yoga sessions, and engagement in habitual physical activity.

It is important to note that this study was conducted during a pandemic. All of our participants were facing an unprecedented situation. They may have responded differently to our outcome variables, and to the remote intervention, because of the pandemic.

Conclusion

This study showed that it is feasible to deliver a moderate-intensity yoga intervention using synchronous and asynchronous technology, at home, for working adults with symptoms of stress, during a pandemic. The intervention resulted in a reduction in stress and anxiety, and an improvement in working memory but did not significantly affect other domains of cognitive functioning. Findings from this study may be useful for individuals looking to engage in moderate intensity flow-based yoga and improve short-term memory, without having to compromise on contact safety during the pandemic.