Abstract

Yoga is a mind–body practice that originated in India thousands of years ago, and which has extended throughout the world in recent years. As it becomes more popular, more studies are being conducted regarding its health benefits in multiple areas, including the human brain, where results have shown that it can reduce stress, modulate neurotransmitters, increase cerebral blood flow, and affect brain structure and function. This review aims to provide a synthesis of the current knowledge on the impact of yoga on human brain structure and function, through the selection and analysis of 23 international peer-reviewed neuroimaging studies with healthy participants. These studies were selected from 216 results on Web of Science, PubMed and PsycInfo after applying the inclusion and exclusion criteria. The final set of studies employed both neuroimaging and neurophysiological techniques, including MRI, fMRI, and EEG. The results show that yoga may exert multiple effects on the brain. However, the heterogeneity of results may be explained by differences in sample characteristics, study designs, and the lack of a consistent definition of yoga and its distinction from meditation. Finally, the limitations of the present review are discussed, along with recommendations for future research aimed at better understanding the neuropsychological health benefits of yoga.

1 Introduction

Yoga is a mind–body practice with historical roots in ancient Indian traditions (Nagendra, 2008; Satchidananda, 2012). While traditionally embedded within a broader philosophical system, contemporary yoga encompasses a range of practices that are increasingly studied for their effects on psychological and biological functioning (Gard et al., 2014a; Schmalzl et al., 2015; van Aalst et al., 2020). In the present review, yoga is considered primarily as a set of practices relevant to brain structure and function, rather than in its historical or philosophical dimensions (Afonso et al., 2021; Bakshi and Srivastava, 2024; Foxen and Kuberry, 2021; Davies et al., 2024).

Contemporary yoga includes diverse styles that differ in their relative emphasis on physical postures (asana), breath regulation (pranayama), meditation (dhyana), and sensory withdrawal (pratyahara) (Schmalzl et al., 2015; Cramer et al., 2016a). Some practices prioritize sustained postures combined with controlled breathing (e.g., Hatha-based approaches), whereas others focus predominantly on meditative components with minimal physical movement. More dynamic styles, such as Vinyasa- or Ashtanga-based practices, integrate continuous movement and breath coordination (Cramer et al., 2016a). These variations are relevant for neuroimaging research, as different components of yoga may differentially engage motor, attentional, interoceptive, and self-regulatory neural systems, thereby contributing to heterogeneity across study findings (Gard et al., 2014a; van Aalst et al., 2020).

Over recent decades, yoga practice has expanded globally and gained widespread popularity, particularly in Europe and North America, largely driven by its use for health-related purposes and as a complementary approach to psychological and medical treatments (Clarke et al., 2015; Cramer et al., 2016b; Zhang et al., 2021). Consequently, increasing research interest has focused on the potential benefits of yoga for physical health, emotional well-being, and cognitive functioning (Kinser et al., 2012; Gothe et al., 2019; Nourollahimoghadam et al., 2021; Bhargav et al., 2024; Gautam et al., 2024; Olex and Olex, 2018; Cramer et al., 2018).

From a neurobiological perspective, yoga has been proposed to influence brain function through multiple interacting pathways, including modulation of the autonomic nervous system, regulation of the hypothalamic–pituitary–adrenal (HPA) axis. It has also been associated with changes in neural networks involved in attention, emotion regulation, and self-referential processing (Taylor et al., 2010; Schmalzl et al., 2015; Aggarwal, 2020). Several theoretical frameworks have been proposed to explain these effects. Taylor et al. (2010) describe an executive homeostatic network (EHN), involving prefrontal, cingulate, and insular regions, which integrates cognitive, emotional, and physiological information to support self-regulation. Similarly, Gard et al. (2014a,b) propose a self-regulatory model of yoga emphasizing the interaction between top-down cognitive control processes and bottom-up physiological regulation across attentional, emotional, and autonomic domains. Importantly, these frameworks provide conceptual models for interpreting potential effects of yoga practice, but they do not constitute direct empirical evidence (see Figures 1, 2) (Davidson et al., 2003; Li et al., 2020; Rathore et al., 2022; Kip and Parr-Brownlie, 2023).

Top-down and bottom-up regulation through yoga practice. Modified from Gard et al. (2014a,b).

The interaction between the HPA pathway, cerebral blood flow and mitochondrial function. PFC, prefrontal cortex; ACC, anterior cingulate cortex; HPC, hippocampus; HRV, heart rate variability; AMY, amygdala, HPA axis, hypothallamic–pituitary–adrenal axis. Modified from Chiarpenello and Brodmann (2024).

An increasing number of neuroimaging studies have investigated the effects of yoga on brain structure and function using techniques such as electroencephalography (EEG), structural magnetic resonance imaging (MRI), and functional MRI (fMRI) (Desai et al., 2015; Fox et al., 2014; Gothe et al., 2018; van Aalst et al., 2020). Structural and functional MRI studies have reported changes in brain regions involved in executive control, emotional regulation, interoceptive awareness, and default mode network (DMN) dynamics (Gard et al., 2014b; Villemure et al., 2015; Gothe et al., 2018; Tomasello et al., 2023; Konecki et al., 2016; Nayan Rishi et al., 2024).

EEG is a widely used technique to study the neural correlates of yoga and meditation, as it provides information about oscillatory brain activity associated with different cognitive and affective states (Bell and Cuevas, 2012; Kora et al., 2021). Previous research has shown that yoga and meditative practices are commonly associated with increases in alpha and theta activity, which have been linked to relaxation, internalized attention, and emotional regulation (Desai et al., 2015; Thomas et al., 2014; Martínez et al., 2020). In more experienced practitioners, changes in higher-frequency bands such as gamma have also been reported, potentially reflecting enhanced moment-to-moment awareness (Thomas et al., 2014; Martínez et al., 2020). These findings support the relevance of EEG measures for examining functional brain changes related to yoga practice.

Despite the growing body of literature on the neural effects of yoga practice, no systematic review to date has provided an integrated synthesis of EEG, MRI, and fMRI findings specifically focused on healthy participants. While previous reviews have addressed aspects of yoga-related neuroimaging, they have either focused on single modalities or included mixed clinical and non-clinical populations, thereby limiting the specificity of conclusions regarding yoga-related neuroplasticity in healthy brains (Desai et al., 2015; van Aalst et al., 2020). The present systematic review aims to address this gap by providing an updated and focused synthesis of neurophysiological and neuroimaging evidence up to 2025, with particular emphasis on functional connectivity and brain network organization in healthy individuals.

2 Methods2.1 Search strategy

The research question (“What structural and functional changes occur in healthy brains with yoga practice?”) was formulated following the Population, Intervention, Comparison, Outcome (PICO) strategy.

The searches for the articles used for this review were performed on Web of Science, PubMed and PsycINFO on February 9, 2025, with the words and Boolean terms: “yoga” AND (“functional connectivity” OR “structural activity” OR “brain structure” OR “brain function”). There were no time range nor specifications set to the search, obtaining 123 results on Web of Science, 59 results on PubMed and 34 results on PsycINFO, for a total of 216 articles.

This systematic literature review was conducted following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines, which establish a series of criteria to follow to assure the quality of the review (Liberati et al., 2009; Moher et al., 2009).

The restriction to English-language publications reflects standard practice in systematic reviews, as English is the predominant language of peer-reviewed journals with high methodological standards. Importantly, the included studies were conducted across diverse geographical and cultural contexts, including non-Western countries such as India—the country of origin of yoga—thereby supporting the cultural breadth of the evidence synthesized in this review.

2.2 Study selection

The articles chosen for this review were screened twice: the first screening was of the title and abstract, and the second screening was of the Methods section. The selection criteria for the screening were as follows:

Inclusion criteria were (a) empirical published studies, (b) published in English, Spanish, or Dutch, (c) included healthy participants, (d) included either an experienced yoga practitioner group and a novice comparison group, and/or a pre-evaluation and post-evaluation of a yoga training group.

Exclusion criteria were (a) systematic reviews, meta-analyses, case studies, book chapters, dissertations, and protocols, (a) studies including participants with cognitive impairment or clinical disorders, (c) studies without a comparison group and without pre–post evaluation, (d) Studies including insufficient yoga exposure (e.g., only basic instructions without formal training), (e) studies reporting only peripheral physiological measures without neuroimaging outcomes.

As shown in Figure 3, of the original 216 results, 149 were excluded. Of the remaining 67, 36 were duplicates, leaving 31 for the second screening. The second screening consisted of reading the full articles, with special attention to the Methods section to identify variables related to the research question: number of participants, their sex, age, and experience with yoga; the procedure followed, the measuring instruments and the structure/function studied. Having done this, one study was excluded as participants were novices and received no formal yoga training (only basic instructions) before the measurements were made; a second study was excluded due to the selection of novice participants with only two yoga sessions accompanied by transcranial direct current stimulation (tDCS), a neuro-stimulation technique that modulates cortical excitability; one study was excluded for only having salivary cortisol level measurements; one study was excluded for being a machine learning accuracy test; one study was excluded for not providing sample characteristics of the healthy participant group, only collective sample characteristics of a mixed group of healthy and unhealthy participants; and three more were excluded for having only one measure (no pre–post) of one group (without a comparison group).

PRISMA flow diagram of the study selection process.

No restrictions were applied regarding participants’ age range, the minimum duration or the intensity of yoga practice, or the year of publication, in order to provide a comprehensive synthesis of the available neuroimaging evidence of yoga.

During this process, some study validity criteria were taken into account, such as the attrition bias in the studies, which was not higher than 10% in any case. Another aspect was how participants were selected and assigned to the experimental or control group: all studies selected participants based on availability and, in some cases, a certain amount of experience with yoga, whilst the control groups (if there were any) were matched in age, sex, education, and other variables, with no experience. And, finally, in cases of doubt as to maintaining or discarding certain studies, experts in the field were consulted.

Figure 3 summarises the screening and selection procedure, whilst Table 1 presents the final 23 articles’ titles, first authors, year of publication, journals, countries, number of citations according to Google Scholar in March 2025, and a code for each publication. These 23 articles were first categorised based on if they were functional studies, structural studies, or both; following which they were analysed to determine their methodological aspects, such as their sample characteristics, types of yoga, imaging techniques, and study design (use of a pre-post evaluation, use of a comparison group, yoga training duration, etc.). This methodological information, along with a summary of each study’s results, is reflected in the Appendix.

Articles included in the systematic review.

The methodological quality and risk of bias of the included studies were formally assessed using validated appraisal tools appropriate to each study design. Of the 23 studies included in this review, one randomized controlled trial was evaluated using the Cochrane Risk of Bias 2 (RoB 2) tool (Sterne et al., 2019), while the remaining 22 non-randomized studies were assessed using the ROBINS-I tool (Sterne et al., 2016). The results of this assessment are summarized in Table 2.

CodeStudy designQuality assessment toolMain sources of biasOverall risk of bias1Cross-sectional (experts vs. controls)ROBINS-IConfounding, participant self-selectionModerate risk2Non-randomized intervention studyROBINS-IConfounding, lack of randomization, self-reported outcomesModerate risk3Pre–post design without randomizationROBINS-IConfounding, participant selectionModerate risk4Cross-sectional (experts vs. controls)ROBINS-IConfounding due to prior experienceModerate risk5Pre–post design without randomized controlROBINS-IConfounding, absence of active controlModerate risk6Cross-sectional (experts vs. controls)ROBINS-IConfounding (age, cognitive activity)Moderate risk7Cross-sectional (experts vs. controls)ROBINS-IConfounding, participant selectionModerate risk8Cross-sectional studyROBINS-IConfounding, self-selectionModerate risk9Population-based observational studyROBINS-IResidual confoundingModerate risk10Cross-sectional (experts vs. controls)ROBINS-ILifestyle-related confoundingModerate risk11Cross-sectional + meditation state comparisonROBINS-IConfounding, participant selectionModerate risk12Cross-sectional (experts, novices, controls)ROBINS-IConfounding by level of experienceModerate risk13Cross-sectional studyROBINS-IConfounding, participant selectionModerate risk14Randomized controlled trialRoB 2Lack of blinding, self-reported outcomesSome concerns15Cross-sectional (experts vs. controls)ROBINS-IConfounding, participant selectionModerate risk16Cross-sectional (older adults)ROBINS-IHealthy aging-related confoundingModerate risk17Cross-sectional (experts, novices, controls)ROBINS-IHigh confoundingModerate risk18Non-randomized pilot studyROBINS-ISelection bias, lack of preregistrationModerate risk19Cross-sectional (experts vs. controls)ROBINS-IPsychosocial confoundingModerate risk20Cross-sectional (experts vs. novices)ROBINS-IConfounding, participant selectionModerate risk21Non-randomized longitudinal interventionROBINS-IConfounding, mixed outcome measuresModerate risk22Cross-sectional studyROBINS-IConfounding, participant selectionModerate risk23Intensive non-randomized interventionROBINS-IConfounding, participant selectionModerate risk

Methodological quality and risk of bias assessment of the included studies.

Overall, most non-randomized studies were judged to present a moderate risk of bias, primarily due to confounding factors and participant self-selection into yoga practice. Importantly, bias related to outcome measurement was generally low, as most studies relied on objective neuroimaging techniques such as MRI, fMRI, or EEG. The single randomized controlled trial showed some concerns of bias, mainly related to the lack of participant blinding and the use of self-reported psychological outcomes.

3 Results

Given the heterogeneity of the included studies, the interpretation of the reported neural effects required consideration of participant characteristics and practice-related variables. As detailed in the Appendix, findings varied according to participants’ level of experience, duration and intensity of yoga practice, age, and the specific type of yoga examined. Studies involving long-term or highly experienced practitioners—often with several years or decades of regular practice—more consistently reported structural and functional brain differences, including changes in default mode network (DMN) organization, increased grey matter volume in regions such as the insula and hippocampus, and greater network integration.

In contrast, intervention and pre–post studies involving novice participants or short-term training programs, ranging from a few days to several weeks, more frequently reported state-dependent or training-related effects. These included reductions in anxiety and negative affect, decreased amygdala reactivity, and transient modulation of DMN activity. Participant age also appeared to influence the pattern of findings, with studies in older adults emphasizing network efficiency, resilience, and preservation of cognitive function. By comparison, studies with younger or mixed-age samples more often reported changes related to emotional regulation, attentional control, and self-referential processing. Finally, the type of yoga practiced differed substantially across studies, with more meditative traditions being more strongly associated with DMN modulation, while more physically oriented practices were more often linked to hippocampal and sensorimotor-related changes.

Of the 23 articles, three studied structural changes through MRI, 15 studied functional connectivity changes (six EEG, nine fMRI), and five studied both types of changes with MRI and fMRI. Regarding the sample characteristics, most studies had predominantly female samples, with only six studies that had more men than women. The average ages ranged between 16.4 and 66.5 years, and experience with yoga varied between none prior to the study training and a maximum experience of 30 years. Sahaja Yoga Meditation (SYM) was the most studied type of yoga (five studies), followed by investigations that studied multiple types of yoga (four studies).

Regarding functional neuroimaging paradigms, the included studies were differentiated based on whether they employed resting-state conditions or other acquisition paradigms. Resting-state paradigms were used in studies examining spontaneous brain activity or functional connectivity in the absence of an explicit task (Barrós-Loscertales et al., 2021; Fingelkurts et al., 2016b; Fingelkurts et al., 2016a; Gard et al., 2014a,b; Gard et al., 2015; Martínez et al., 2020; Santaella et al., 2019; Thomas et al., 2014; Villemure et al., 2015). In contrast, the remaining studies employed task-based paradigms, meditation- or practice-induced states, baseline-plus-meditation designs, or structural neuroimaging approaches, and therefore do not constitute resting-state paradigms (Dodich et al., 2019; Fialoke et al., 2024; Gothe et al., 2018; Gotink et al., 2018; Hernández et al., 2016; Hernández et al., 2018; Malipeddi et al., 2024; Novaes et al., 2020; Pérez-Diaz et al., 2024; Shrivastava et al., 2023; Simon et al., 2017; Singleton et al., 2021; Tymofiyeva et al., 2021; Vishnubhotla et al., 2021).

3.1 Structural changes

In total, eight studies analysed structural changes in the human brain with yoga practice, which can be divided into WM changes and GM changes, summarised in Figure 4.

Structural changes in the human brain with yoga practice. L, left; R, right; WM, white matter. Only statistically significant results are shown.

3.1.1 White matter changes

Only one study (Pérez-Diaz et al., 2024) focused on the white matter (WM) changes, noting that in five of the seven WM paths analysed, people with experience in yoga had a higher number of tracts, such as between both amygdalae and the left anterior cingulate cortex (ACC). In the remaining two paths, yoga practitioners had a lower number of tracts than inexperienced participants, including in the left anterior insula (AI).

3.1.2 Grey matter changes

Of the seven studies that analysed GM changes, two found that the total intracranial volume did not change between yoga practitioners and non-practitioners (Gotink et al., 2018; Hernández et al., 2016), despite there being some GM volume changes in certain areas, such as a decrease in right amygdala GM volume (Gotink et al., 2018), and an increase in GM volume stretching from the right AI to the right ventromedial orbitofrontal cortex (vmOFC), which correlated positively with depth of mental silence and daily frequency of thoughtless awareness (Hernández et al., 2016).

Other GM volume changes are varied among the studies. For example, two studies found an increase in volume for the insula, one in left AI clusters (Hernández et al., 2016) and the other in the left mid-insula (Villemure et al., 2015) both correlating positively with amount of experience. This latter study also found increases in the left frontal operculum, the right middle temporal gyrus (both of which correlated positively with experience), the right primary visual cortex, the posterior cingulate cortex (PCC), and the right superior parietal lobule (all three of which had a positive correlation with weekly practice).

There were contradictory results regarding the GM volume of the left hippocampus in yoga practitioners, with two studies finding an increase in volume that correlated with amount of practice (Gothe et al., 2018; Villemure et al., 2015), whilst one other found a decrease in volume with no significant correlations (Gotink et al., 2018).

One study analysed cortical thickness in yoga practitioners (Singleton et al., 2021), finding increases in the left isthmus cingulate, the precuneus and the PCC, all of which presented a positive correlation with total weighted scores for ego development in the Maturity Assessment Profile (MAP). Another study (Dodich et al., 2019) found an increase in GM density in the right inferior frontal gyrus, correlating with increased self-reported well-being.

3.2 Functional connectivity changes

A total of 20 studies analysed functional connectivity (FC) changes in the human brain with yoga practice, through EEG measurements and fMRI measurements.

3.2.1 Electroencephalogram results

As shown in Figure 5, six studies used EEG as their imaging technique. Two studies (Fingelkurts et al., 2016b; Fingelkurts et al., 2016a) analysed changes in operational synchrony (OS), with one finding a decrease in global Default Mode Network (DMN) OS (Fingelkurts et al., 2016a), and both finding an increase in OS in the frontal DMN operational module and a decrease in the bilateral posterior DMN operational module. These changes in DMN activity correlated with self-reported calmness, happiness, self-agency, simpler subjective experiences and slower thought speed.

EEG results of functional connectivity changes in the human brain with yoga practice. DMN, default mode network; PFC, prefrontal cortex; R, right; S1, primary somatosensory cortex. Only statistically significant results are shown.

Another two investigations studied changes in frequency band power (Malipeddi et al., 2024; Shrivastava et al., 2023). Shrivastava et al. (2023) reported an increase in alpha, delta and theta band power in the prefrontal cortex (PFC) in all states. Malipeddi et al. (2024) found larger increases in the yoga group compared to the control group in different bands for different areas () depending on the state the participants were in. Specifically, in the meditation state, experienced practitioners had no change in band power, whilst novices and non-practitioners had reduced band power. These band power results correlated positively with self-reported well-being, meditation depth, and non-attachment in practitioners, whilst controls reported higher drowsiness, stress, mental distress and mind-wandering.

Of the remaining two EEG studies (Martínez et al., 2020; Thomas et al., 2014), one focused on sample entropy (SE) changes (Martínez et al., 2020), finding increases in global alpha, frontal low gamma and centro-frontal high gamma bands in a meditation group, and an increase in high gamma in a Hatha Yoga group. The second study (Thomas et al., 2014) focused on current source density (CSD), finding increases in beta and gamma bands in experienced yoga practitioners compared to novice practitioners; and increases in alpha1, alpha2, delta and theta bands in novice yoga practitioners, compared to experienced practitioners. Novice practitioners reported higher meditation depth than experienced practitioners, with the biggest difference during the mantra condition, which is when alpha1 activity was highest in novices.

3.2.2 Functional magnetic resonance imaging results

As summarised in Figure 6, there were 14 fMRI studies included in this review. Three studies (Fialoke et al., 2024; Simon et al., 2017; Vishnubhotla et al., 2021) found a decrease in DMN structural activity, specifically in the PCC (Fialoke et al., 2024; Simon et al., 2017), the pregenual ACC, the precuneus, the medial prefrontal cortex (mPFC) and both inferior parietal lobes. One study (Fialoke et al., 2024) found no DMN deactivation despite an increase in auditory and motor region activity. Another study (Tymofiyeva et al., 2021) found a decrease in right amygdala activation. All other structural activation and FC results are varied, with no coincidences nor contradictions between studies.

fMRI results of functional connectivity changes in the human brain with yoga practice. R, right; L, left; RS, resting state; MS, meditation state; FC, functional connectivity. Only statistically significant results are shown.

Regarding decreases in connectivity, one study (Vishnubhotla et al., 2021) found a reduction in FC pathways within the DMN, the dorsal attention network (DAN) and the salience network (SN); and other studies found FC decreases from the right angular gyrus to the left occipital cortex (Barrós-Loscertales et al., 2021), the left insula to the mid cingulate cortex (Barrós-Loscertales et al., 2021), and the rostral ACC to the th