Multimodal neurophenomenology of advanced concentration absorption meditation: An intensively sampled case study of Jhana

Abstract

Using a combination of fMRI, EEG, and phenomenology ratings, we examined the neurophenomenology of advanced concentrative absorption meditation, namely jhanas (ACAM-J), in a practitioner with over 23,000 h of meditation practice. Our study shows that ACAM-J states induce reliable changes in conscious experience and that these experiences are related to neural activity. Using resting-state fMRI functional connectivity, we found that ACAM-J is associated with decreased within-network modularity, increased global functional connectivity (GFC), and desegregation of the default mode and visual networks. Compared to control tasks, the ACAM-J were also related to widespread decreases in broadband EEG oscillatory power and increases in Lempel-Ziv complexity (LZ, a measure of brain entropy). Some fMRI findings varied by the control task used, while EEG results remained consistent, emphasizing both shared and unique neural features of ACAM-J. These differences in fMRI and EEG-measured neurophysiological properties correlated with specific changes in phenomenology – and especially with ACAM-J-induced states of bliss - enriching our understanding of these advanced meditative states. Our results show that advanced meditation practices markedly dysregulate high-level brain systems via practices of enhanced attention to sensations, corroborating recent neurocognitive theories of meditation as the deconstruction of the brain’s cortical hierarchy. Overall, our results suggest that ACAM-J is associated with the modulation of large-scale brain networks in both fMRI and EEG, with potential implications for understanding the mechanisms of deep concentration practices and their effects on subjective experience.

1. Introduction

Over the past few decades, mindfulness, and more broadly meditation, has received considerable attention in both science and news media. Supported by substantial evidence of positive associations between mindfulness meditation and stress reduction (Kabat-Zinn et al., 2011; Kabat-Zinn and Burney, 1981), mindful practices have increasingly been incorporated into clinical mental health practices (Wielgosz et al., 2019) and as a tool for increasing subjective well-being. Nevertheless, the scientific evidence supporting mindfulness has received criticism for its poor methodology and lack of scientific rigor (Van Dam et al., 2018). Mindfulness meditation, for example, is poorly understood from the perspective of psychology and neuroscience (Tang et al., 2015), despite it being an ancient practice common to most religious, contemplative, and spiritual traditions. This gap in understanding has motivated recent research to examine the specific cognitive mechanisms underlying different forms of meditation, their associated neuroscientific signatures, and what they reveal about the mind and brain (Dahl et al., 2015; Lutz et al., 2008; Slagter et al., 2011).

Only recently have there been attempts to map the writings of ancient texts describing the methods and practices of different kinds of meditation to concepts and processes in cognitive psychology (Dahl et al., 2015; Galante et al., 2023; Laukkonen and Slagter, 2021; Wright et al., 2023). Since the window into the meditative mind is the meditator – researchers have the task of constructing cognitive models of meditation through descriptions of methods used by practitioners to attain these states, their subjective experience during the meditation, as well as objective measures of neural activity co-occurring during these practices and experiences. This rigorous and extensive use of first-person data about the meditator’s experience during meditation to describe, constrain, and interpret third-person or objective neural data – termed neurophenomenology – has been advocated as an effective method to probe not only these non-ordinary states of consciousness but also consciousness and cognition more broadly (Lutz, 2002; Lutz and Thompson, 2003; Timmermann et al., 2023a). The foremost challenge to using first-person, subjectively generated reports to guide our understanding of cognition is that they can be biased or inaccurate – and that there may be no method to objectively verify or corroborate these reports. A potential solution to this dilemma is to recruit phenomenologically trained meditators to report their experience. Adept meditators can be considered phenomenologically trained, as mindfulness meditation enhances introspective awareness. This aligns with the idea that their improved phenomenological abilities are valuable for studying both meditative states and consciousness more broadly (Varela, 1996). These participants are likely to generate more stable and reproducible mental states than untrained subjects, and can also describe these states more accurately than novice participants (Lutz and Thompson, 2003). Indeed, several investigators have previously studied advanced stages or states of meditation as experienced and described by adept meditators (Berkovich-Ohana, 2017; Dor-Ziderman et al., 2013; Laukkonen and Slagter, 2021; Metzinger, 2020; Paoletti et al., 2022). Previous studies also report that the duration of meditation practice can nonlinearly impact large-scale functional network dynamics of the brain (Brefczynski-Lewis et al., 2007) – and this underscores the importance of examining the neurophenomenology of advanced meditation in individuals with an exceptionally long duration of meditation practice, coupled with a mastery of the technique. Specifically for advanced meditative states, studying a single adept meditator enables the recording of high-quality, stable phenomena, strengthened by repeated measurement power—effects that could be blurred in premature group studies. Once specific effects have been identified, these single-case studies can lay the groundwork for future corroborative larger-scale research, as seen in prior work (Dor-Ziderman et al., 2013; Trautwein et al., 2024).

The present study is a neurophenomenological case study examining an advanced meditation practice, the jhanas (“jhānas_” in Pali, the liturgical language of Theravada Buddhism). The jhanas are non-ordinary states of consciousness that can be experienced through deep concentration meditation (Gunaratana, 1992; Ñāṇamoli, 2010; Sparby and Sacchet, 2024), usually as a culmination of long-term, intensive practice. Similar concentration meditation practices and phenomenology have also been mentioned in various non-Buddhist and geographically and temporally different contexts (Wahbeh et al., 2018). For example, similar states are described in the Carmelite tradition (Beauregard and Paquette, 2008); Islamic Sufi meditation (Louchakova-Schwartz, 2011); and Jewish contemplative traditions that involve states of immense concentration (Fisher, 2022) – suggesting that these practices and their experiences may manifest universally as unique signatures of human consciousness. The study of jhanas, therefore, may be informative regarding the nature of consciousness in general. The practice of jhana is classified as being that of Focused Attention (FA) (Dahl et al., 2015) – due to its focus on training the mind to maintain unwavering concentration through object orientation – and we refer to it as advanced concentrative absorption meditation – jhanas (ACAM-J; Yang et al. (2024b)). Phenomenologically, and according to ancient Buddhist texts (Ñāāhenom, 2010), the ACAM-J are eight successive states of increasing concentration. The first four are sometimes referred to as the form ACAM-J while the latter four are referred to as formless ACAM-J. Each of these states presented as having distinct characteristics in terms of experience (Dennison, 2019; Hagerty et al., 2013; Sparby and Sacchet, 2022). The intense concentration that is characteristic of the ACAM-J may also be described as effortless (Tang et al., 2015). During the first ACAM-J, the experience of bliss/happiness (“Piti” in Pali) may stand in the foreground, and then the meditator progresses through states in which contentment and equanimity predominate. As the ACAM-J progress, the practice of concentration provides the background for increasing formlessness or dereification (Dahl et al., 2015), leading to a sense of self-dissolution. To our knowledge, there have been only three prior neuroimaging studies of the ACAM-J (Dennison, 2019; Hagerty et al., 2013; Yang et al., 2024a). The two initial studies, although provide preliminary evidence of ACAM-J being associated with changes in brain activity, lack neurophenomenological rigor. The third study (Yang et al., 2024a) with comprehensive neurophenomenological analysis, identified distinctive patterns of brain activity in cortical and subcortical brain regions associated with ACAM-J, and these activations also correlated with specific jhanic qualities of attention and bliss.

Although generations of practitioners and scholars have mapped the subjective experiences that emerge from practicing ACAM-J, only recent developments in neuro-cognitive theoretical frameworks of meditation and other altered states of consciousness enable scientifically rigorous empirical predictions and testing of falsifiable hypotheses (Cooper et al., 2022; Dahl et al., 2015; Laukkonen and Slagter, 2021; Lutz et al., 2015; Timmermann et al., 2023a; van Elk and Yaden, 2022; Wright et al., 2024). Specifically, these models of consciousness and cognition posit that meditation effectively alters the hierarchical nature of the mind (or brain; Badcock et al., 2019). Specifically, FA meditation may reduce the effect of higher-level abstract processes in the brain’s hierarchy by directing sustained attention on information (e.g. sensation) acquired in the present moment (Laukkonen and Slagter, 2021) and reducing mental processes that employ counterfactual and temporally deep cognition (e. g., rumination, anxiety, regret/relief).

We examined the neurophenomenology of ACAM-J to obtain a deeper, process-based understanding of mental and neural activity during advanced meditation (Sparby and Sacchet, 2022). Using multimodal neuroimaging (EEG, 7T fMRI) and phenomenological assessment of the advanced meditative experience, we tested the hypothesis of disintegration of hierarchical processing in the brain during ACAM-J practice. Specifically, we assessed ultra-high-field 7T fMRI functional connectivity across and within canonical functional brain networks (Yeo et al., 2011), as these networks consistently represent a gradient from unimodal (i.e., sensory) to transmodal (i.e., more abstract functions) organization in the brain (Atasoy et al., 2017; Huntenburg et al., 2018; Margulies et al., 2016; Mesulam, 1998). Activity within these macroscale networks also provides evidence of modularity and hierarchical organization (Bertolero et al., 2015; Betzel and Bassett, 2017; Kok et al., 2016), where higher-order modules within regions form hypotheses related to the causes of current sensory inputs and then send feedback to lower-order regions to “test” these hypotheses. Specifically, the default mode network (DMN) and frontoparietal network (FPN) are thought to be positioned at the top level in the hierarchy of the cognitive landscape (Margulies et al., 2016; Vatansever et al., 2017; Vidaurre et al., 2017), and increased connectivity between these association networks with lower-level functional networks may indicate a breakdown in hierarchical processing. One possibility is that the occurrence of blissful states during ACAM-J is explained by such a high to low-order connectivity – similar to how the hyperconnected brain in psychedelic experiences is associated with increased feelings of immersion and self-dissolution (Tagliazucchi et al., 2014; Timmermann et al., 2023b). In the context of predictive processing, the brain’s electrophysiological (e.g., EEG) activity may also represent a hierarchical organization of the brain’s function (Klimesch, 2012; Snipes et al., 2022; Steinke and Galán, 2011) – lower-frequency oscillatory alpha-to-beta band activity has been linked to top-down predictive signals and higher frequency gamma band activity to bottom-up prediction errors (Arnal and Giraud, 2012; Engel et al., 2001; Engel and Fries, 2010). Our approach enabled the examination of direct changes in neural activity (EEG) in parallel with indirect changes seen through fMRI Blood Oxygen Level Dependent (BOLD) signal, as it relates to subjective experiences during states of ACAM-J.

2. Materials and methods

2.1. Participant

This was a case study with a single adept meditator, aged 52 years old at the time of data collection. This approach of extensively sampling experiences of interest (i.e., ACAM-J) in a single participant has proved to be a powerful tool in neuroscience to derive insights beyond a single or several repeats of a measurement (Levenson et al., 2012; Poldrack et al., 2015). Specifically in the domain of advanced meditation, where considerable meditation expertise and introspection are developed, there is unique value in studying a single practitioner with mastery of the specific meditation practice. The participant was a long-term meditation teacher with over 25 years of meditation experience at the time of data acquisition. The length of training was estimated based on his daily practice and the time spent on meditative retreats. Based on an estimated one to two hours of daily practice since the start of practice and approximately one year of retreat at 14 h per day, we estimate a total practice amount of at least 20,000 h of total meditation practice. Throughout his life span, the participant has practiced within a variety of meditation traditions including Mahasi noting, form and formless ACAM-J - specifically Sutta ACAM-J (Shankman, 2008), Kasina meditation, and Vipassana or Insight meditation (Ingram, 2018), with the ACAM-J being a fundamental and consistent element of the participant’s meditation practice. Before the neuroimaging and electrophysiology data acquisition sessions, the participant was interviewed by a PhD-level clinical psychologist to complete the Mini-International Neuropsychiatric Interview (MINI; Sheehan et al., 1998) and was found to have no diagnosis. The participant also showed no cognitive impairment as assessed by the Mini-Mental State Examination (MMSE; Folstein et al., 1975). The meditator in this study is the same as that in Ganesan et al. (2024) and Yang et al. (2024a) - while the fMRI data are partially shared with these studies, the EEG dataset and analyses presented here are distinct. The Mass General Brigham IRB approved the study, and the participant provided informed consent.

3. Study procedures

3.1. Advanced concentrative absorption meditation – Jhanas (ACAM-J)

During both the MRI and EEG sessions, the participant was asked to meditate in their standard sequence for ACAM-J practice, eyes closed. Our sampling of ACAM-J from the participant was extensive – there were five days of EEG and five more days of fMRI data collection, with each day spanning up to 3 to 4 h of data acquisition. In total, we recorded 27 runs of ACAM-J in the MRI and 29 runs of ACAM-J with EEG. A run started with the participant entering access concentration (AC): the initial focus of attention on the object of meditation (e.g., breath, bodily feelings, visual imagery, or a combination of these) necessary to access the ACAM-J. The participant then progressed sequentially, without pause, through ACAM-J1 to 8, finishing with a post-8th ACAM-J state that we term afterglow (refer to the note in supplementary).

For entering the first four form ACAM-J, the participant indicated that the practice involved focusing on the object of meditation: a combination of the breath, bodily feelings, and width of attention. In contrast, the formless ACAM-J (ACAM-J5–8) transcend sensory concentration and were their own object of meditation (refer to introduction). In general, the participant intended/inclined their mind to begin a progression through the ACAM-J and the mind did as intended in its own time.

For every transition from AC to ACAM-J6, the participant pressed a button to indicate the start of that state. For ACAM-J7 and 8 (i.e., the deep formless ACAM-J), the participant did not press any buttons to indicate the transition into these states. According to the participant, pressing buttons for ACAM-J 7 and 8 would require an exit from these deep meditative states and, therefore, would have interrupted the natural traversing to the subsequent ACAM-J state. Therefore, following a button press indicating a transition to the 6th ACAM-J, the next button press was only when the participant exited ACAM-J8 and entered the post-ACAM-J afterglow. Apart from the frequency of afterglows, the participant invariably progressed through each ACAM-J state in every session. The average duration for a complete run was 14 min. Across all runs (EEG and fMRI), the average duration for the ACAM-J states were as follows: ACAM-J1: 77.15 s; ACAM-J2: 92.45 s; ACAM-J3: 107.64 s; ACAM-J4: 75.28 s; ACAM-J5: 69.14 s; ACAM-J6–8: 180.20 s.

3.1.1. Control conditions

Discussions of research on advanced meditation (Tang et al., 2015) caution that experienced meditators may naturally enter meditative states even during rest. We also discussed the feasibility of using a “pure” resting-state condition with our case study participant, who reaffirmed that maintaining such a state would be effectively unattainable. Thus, using a resting state as a control could complicate the interpretation of results if it were not genuinely “restful” but instead involved some degree of meditative activity. To address this, we developed two control tasks that engaged non-meditative cognition and approximated a resting-state condition by tapping into cognitive processes common during rest. This approach acknowledges that rest is not a blank state of mind, but rather one filled with mind-wandering and fluctuating attention (Smallwood and Schooler, 2015, 2006). Specifically, the first control task was a counting task, in which the participant was asked to count down mentally, without moving the lips, in decrements of 5 from the number 10,000 for 8 min with eyes closed. The second control task involved memory recall, where the participant was instructed to reminisce about events from the past two weeks and silently narrate them in their mind, also for 8 min with eyes closed. The memory recall task engages self-referential thoughts, which are a dominant feature of mind-wandering and typical during rest, where the mind frequently shifts between recalling past and anticipating future events (Seli et al., 2017; Stawarczyk et al., 2011). Similarly, the counting task recruits minimal working memory and reflects wavering attention – a cognitive process common in everyday tasks that require partial but not sustained focus (Smallwood and Schooler, 2015). Thus, the combined cognition in these two tasks should effectively emulate a resting-state condition. Our use of combined controls (counting and memory tasks) provides a more precise comparison than a traditional resting-state control. Resting-state studies may not fully separate meditation effects from general cognitive processes like mind-wandering. By using two control conditions that reflect typical resting-state cognition, we aimed to isolate the neural mechanisms underlying the meditation state.

Our choice of control tasks was also motivated by recent frameworks that explain FA meditation as part of a continuum that deconstructs the brain’s typical hierarchical cognition (Laukkonen and Slagter, 2021). According to this model, FA meditation directs attention to a specific sensory experience (e.g., breath, visual imagery) and reduces temporally deep cognition, such as thinking about the past or future. In the formless ACAM-J, this concentrated experience becomes the object of meditation, while attention to extraneous thoughts and sensory input further decays. Thus, we deliberately chose control tasks that were temporally dense (memory recall) or required attention on a fluctuating process (reverse counting), rather than sustained focus on a single sensation. Our general hypothesis is that the fluctuating attentional process of the counting task serves as a general contrast to the sustained concentration in the ACAM-J, while the memory recall task, as a proxy for self-related thoughts, may uniquely elucidate the self-deconstructive nature of the formless ACAM-J. We collected two runs for each control condition, totaling 16 min of data for each task. The first run of control tasks was collected before any meditation runs, while the second was conducted on the fourth of five data collection days. The order of presentation of the two control conditions within a session was randomized.

3.1.2. Phenomenology

In the present study, we employed a neurophenomenological approach. That is, we used systematic first-person descriptions of the experience to provide a more precise explanation of objective or third-person neuroimaging data (Lutz and Thompson, 2003). Specifically, our phenomenologically trained subject (i.e., with the ability to generate, identify, and rate stable and reproducible ACAM-J states and their experiences) systematically evaluated the mental and physiological processes relevant to the experience of ACAM-J as they manifested during his meditation. Prior to any neuroimaging scans, the case study participant first listed these phenomenological items to characterize the experience during meditation. The items were then reviewed to define appropriate rating scales and terminology. Of note, these phenomenological items were consistent with the typical experience of ACAM-J in advanced practitioners, as codified in Buddhist texts (Khema, 2022) and by long-term practitioners (Brasington, 2015). Immediately after each run (EEG or fMRI), phenomenological ratings were obtained from the participant using Likert-type rating scales (see Supplementary). We carefully framed questions by consulting the participant to ensure clarity and minimize recall bias. The participant also noted specific characteristics of that run in an open response format and this response was recorded either as audio and transcribed (in fMRI) or as typed text (in the EEG acquisition room). We also implemented several measures to address potential issues with subjective reporting. The subject was given the option to skip specific items if he had no awareness of the state or could not remember the given state’s characteristics. Moreover, a “back” option was provided if they wished to re-rate certain items. It is worth noting that these nonconscious or no-memory responses were minimal, accounting for less than 1% of the total number of responses, suggesting that the subject was able to provide meaningful and accurate phenomenological ratings throughout the study. Some of these phenomenology items (see supplementary for a list and specific questions) were also rated following the control conditions. These ratings allowed us to compare the phenomenology between control and ACAM-J states. We included phenomenology items congruent with the five dominant factors (Gunaratana, 1992, 1988) commonly and distinctly associated with each ACAM-J. The “stability” and “width” of attention were rated across all ACAM-J. Additionally, the participant rated the intensely pleasant bodily feeling of “bliss or joy” during ACAM-J2. For the third ACAM-J, the participant rated “cool bliss” – which is a more subtle (mental) pleasant experience achieved through additional concentration. When the practice of the third ACAM-J matures, in the fourth ACAM-J, feelings of bodily pleasure go away, and the experience is neither pleasant nor unpleasant (i.e., neutral). The participant rated this as “equanimity”, also termed “jhanic neutral qualities”. From the fifth ACAM-J onwards, the states receive the designation “immaterial” or “formless” because they are achieved by surmounting all perceptions of material form and are characterized by “grades of formlessness”. The participant rated the “grades of formlessness” for these formless ACAM-J (i.e., ACAM-J5–8). We also asked the participant how often (0 – Never to 10 - Always) and to what degree (0 – Not at all to 10 - Extremely) did ACAM-J phenomenology arose during the control tasks.

3.1.3. Differences between ACAM-J and control phenomenology

Since we collected two runs of data for each of the control conditions, and in each modality (i.e., fMRI and EEG), we only had 4 data points for control-related phenomenology. Due to the disparity and small number of data points for the control rating compared to ACAM-J (total 56, 27 fMRI, and 29 EEG runs), we did not perform any formal statistical test to establish differences between control and ACAM-J ratings, and present to the reader only mean differences in ratings.

3.1.4. fMRI acquisition

The fMRI data were acquired across five consecutive days. Each day, the participant completed 3–4 h of MRI data collection depending on levels of fatigue and comfort. Data were acquired using an ultra-high-field 7T MR scanner (Siemens MAGNETOM Terra) using a 32-channel head coil. Functional imaging was performed using a single-shot two-dimensional echo planar imaging sequence with T2*-weighted BOLD-sensitive MRI (repetition time (TR) = 2.9 s, echo time (TE) = 30 ms, flip angle (FA) = 75°, field of view (FOV) = [189×255], matrix = [172×232], GRAPPA factor = 3, voxel size = 1.1 × 1.1 × 1.1 mm3, 126 slices, interslice distance = 0 mm, multi-band factor = 3). Slice acquisitions were acquired for the whole brain, with interleaved slices, sagittal orientation, and anterior-to-posterior phase encoding. We also acquired opposite phase-encoded (i.e., posterior-to-anterior) slices with the same parameters to perform distortion correction. Whole-brain T1-weighted structural images were also acquired with parameters: TR = 2.53 s, TE = 1.65 ms, inversion time = 1.1 s, flip angle = 7°, 0.8mm isotropic resolution, FOV = 240×240, GRAPPA factor = 2, bandwidth =1200 Hz/Px. To account for physiology-related signal fluctuations, during fMRI acquisition, we also recorded the timing of cardiac and respiratory cycles from the participant throughout each scanning session by the use of a piezoelectric finger pulse sensor (ADInstruments, Colorado Springs, CO, USA) and a piezoelectric respiratory bellow (UFI, Morro Bay, CA, USA) positioned around the chest, respectively. The recording was performed using a PowerLab data acquisition system (PowerLab 4/SP, ADInstruments, New Zealand) and LabChart (LabChart, AD Instruments, New Zealand), with a sampling rate of 1000 Hz. The same acquisition system also recorded the MR scanner trigger pulses synchronized with the acquisition of each image volume.

4. fMRI data analysis

4.1. fMRI preprocessing

Preprocessing steps were conducted at the level of the session and for all runs within that session. A session started each time the participant’s position was localized in the scanner and lasted till the next localization was conducted. For each session, one reverse phase-encode EPI was collected for distortion correction across all runs for that session. Preprocessing steps consisted of (1) de-spiking (3dDespike, AFNI); (2) RETROspective Image CORrection (RETROICOR; Glover et al., 2000) using (ricor, AFNI) to regress out effects of physiological (cardiac and respiratory) noise. We used a total of 10 regressors: four cardiac phase regressors (2nd order), four respiration phase regressors (2nd order), respiration volume per time convolved with the respiration response function (Birn et al., 2008), and heart rate convolved with the cardiac response function (Chang et al., 2009); (3) Slice time correction (3dTshift, AFNI); (4) Distortion correction using opposite phase-encoded EPI (3dNwarpApply, AFNI); (5) motion correction (3dvolreg, AFNI) was done by registering each volume to a volume with the minimum outlier voxels within the brain-mask (i.e., a low-motion volume, determined based on the data); (6) Registering the anatomical dataset (T1) to a standard (MNI152_2009) template (@Sswarper, AFNI). This step includes multiple substeps: (6a) Bias-field correction of T1 image (3dUnifize, AFNI); (6b) Skull-strip T1 (3dSkullStrip, AFNI); (6c) Nonlinearly warp the T1 to the standard dataset (3dQwarp, AFNI). Registration from EPI to the standard template was done by concatenating all the transformation matrixes (distortion correction, motion correction, registration to anatomical, anatomical to standard) into one single registration. Additional preprocessing steps included: (7) Scrubbing: we scrubbed any volume with motion > 0.3 mm and had more than 5% outlier voxels (AFNI). Scrubbed volumes were removed from the data (no more than 10% of total volumes were scrubbed for any session); (8) Regress: We regressed out eroded CSF mask timecourse and motion parameters (3 translations, 3 rotations) per run and conducted band-pass filtering (0.01 – 0.1Hz) by putting them in one single regression (3dDeconvolve, AFNI); (9) spatial smoothing (FWHM) of 6 mm (3dBlurInMask, AFNI). Next, the data from the entire ACAM-J run was segmented to separate data corresponding to each state. We excluded ACAM-J segments that were less than ~30 s (10 TRs) in all analyses, which resulted in the exclusion of two data segments. For each of the control datasets, we cut the 8-minute run into consecutive ~1 min (20 TRs) segments, close to the average duration for each ACAM-J state, yielding a total of 16 segments for each control condition from the two runs of control data.

4.1.1. Within-network modularity (fMRI)

To examine possible changes in the hierarchical (modular) organization within specific resting-state functional networks (RSNs; Yeo et al., 2011), we quantified and compared modularity between ACAM-J and control conditions. Broadly, modularity indicates the degree of segregation of a network into modules (or clusters). Networks with high modularity have dense connections between nodes within the same module but sparse connections between nodes across different modules. To analyze the modularity of the canonical RSNs, we first calculated pairwise functional connectivity between areas across the 100 regions defined by the Schaeffer atlas (Schaefer et al., 2018). Specifically, Pearson correlation analyses were performed between each pair of regions, resulting in a connectivity matrix comprising 100×100 connections (or ‘edges’). This connectivity matrix was computed for each ACAM-J state and each run, as well as for each control condition and each segment. The matrices were then segregated by network, isolating edges corresponding to the seven RSNs defined by Yeo et al. (2011), which include the visual network (VN), somatomotor network (SMN), dorsal attention network (DAN), ventral attention or salience network (SN), affective network (AFN), frontoparietal control network (FPN), and the default mode network (DMN).

We calculated the modularity of these networks using the Louvain algorithm, implemented via the Brain Connectivity Toolbox (Version 1.1.1.0; (Rubinov and Sporns, 2010)) in MATLAB R2022b. This algorithm computes the modularity score by assessing how much more densely connected the nodes within each network are relative to a random network. The modularity values were compared between the control conditions (counting, memory, and combined) and individual ACAM-J states for each RSN using nonparametric permutation tests with 10,000 permutations. These tests were FDR-corrected separately for each control condition and network across the six pairwise comparisons between each ACAM-J state and specific control. Using the same analysis, we also separately compared the control conditions with the form and formless ACAM-J, grouping the first four form ACAM-J (ACAM-J1–4) and the formless ACAM-J (ACAM-J5–8) into distinct categories. This additional analysis aimed to capture potential effects that may not be evident when comparing with individual ACAM-J states. For the form and formless ACAM-J comparisons, FDR corrections were applied separately, as these analyses combined multiple ACAM-J states and served as complementary to the primary comparisons. This correction approach was consistently applied across all metrics analyzed in this study. To account for the difference in sample sizes between ACAM-J and control conditions (e.g., 27 ACAM-J runs and 16 control runs), a random subset of the ACAM-J modularity values, equal to the sample size of the control condition, was sampled without replacement for each permutation. The p-value for each test was computed as the proportion of times the absolute difference in means between the shuffled samples exceeded the observed difference in the original data.

4.1.2. Between-network segregation (fMRI)

We examined segregation between the 7 RSNs by calculating a resting-state functional connectivity (RSFC) matrix. This matrix represented pairwise functional connectivity between each of the seven RSNs of interest, with greater connectivity between networks indicating lesser segregation. FMRI time series for each of the seven networks were extracted using the network components as a region of interest (ROIs). For each pair of RSNs, these time series were then entered into a regression, resulting in beta values representing the strength of functional connectivity between RSNs. The regression was conducted both ways – that is, with each RSN as a dependent variable in one model and as an independent variable in the second model. The beta values were then averaged to generate 7 × 7 matrices for control and ACAM-J states. We then examined the difference in these beta values between the control (counting, memory, combined) and ACAM-J states (individual ACAM-J plus groups of form and formless ACAM-J) using nonparametric permutation tests with 10,000 permutations (of independent samples) on each of the quadrants of the matrices. FDR correction was applied separately per control condition and network, across six comparisons each time.. Hedges’ g was used to quantify the effect size of the difference in pairwise network connectivity between control and ACAM-J.

4.1.3. Global functional connectivity (fMRI)

Global functional connectivity (GFC) was obtained by taking the average normalized Fisher Z score of Pearson correlation coefficient value from each brain area to all other brain areas. Brain areas in the cortex were defined according to the 100 regions resolution of the Schaeffer cortical atlas (Schaefer et al., 2018). Global connectivity at the network level was obtained by averaging the GFC values for all Schaeffer parcellations corresponding to each of the seven networks by Yeo et al. (2011). We then examined the difference in GFC values between the control (counting, memory, combined) and ACAM-J states (individual ACAM-J plus groups of form and formless ACAM-J) using nonparametric permutation tests with 10,000 permutations (independent samples) for each control-ACAM-J pair. FDR correction was applied separately per control condition and network, across six comparisons each time.

4.2. EEG data acquisition

The meditator was seated approximately 50 cm before a computer monitor inside an acoustically and electrically shielded booth for EEG data acquisition. EEG cap placement adhered to the standard 10–20 EEG positioning method. Continuous EEG activity was recorded from a customized 96-channel actiCAP system using an actiCHamp amplifier (Brain Products GmbH, Gilching, Germany). Impedances were kept below 5 kΩ. The ground (GND) channel was embedded in the cap and was located anterior and to the right of Channel 10, which roughly corresponds to electrode Fz. Channel 1 (Cz) served as the online reference channel during data acquisition. All signals were digitized at 500 Hz using BrainVision Recorder software (Brain Products).

4.3. EEG data analysis

4.3.1. EEG preprocessing

Offline analyses were performed using FieldTrip (Oostenveld et al., 2011) and EEGLab (Delorme and Makeig, 2004), both MATLAB-based toolboxes. The data was first downsampled to 250 Hz, demeaned, and band-pass filtered with cut-offs of 1 and 45 Hz. Data points with gross muscle and other artifacts were first manually removed following a visual inspection. These artifacts primarily occurred at the very start and end of sessions for short periods and were otherwise minimal. During the same visual inspection, channels that were noisy throughout the recording were identified. These channels were repaired using interpolation, that is: replacing the bad channel with the weighted average of their neighbouring channels. Neighbouring channels were defined using a ‘triangulation’ method as implemented in FieldTrip. A mean of 0.96 channels (SD =1.46) was interpolated for consistent noise or visible bridging with other channels. The EEG time course was then detrended to remove any linear trends, followed by referencing (to the average of all channels). Independent Component Analyses (ICA) were subsequently performed to remove residual noise. We used an automatic component rejection algorithm (IClabel; Pion-Tonachini et al., 2019) to discard non-brain components. IClabel provides separate probability estimates for each ICA component, including the probability that the component is related to brain activity, muscle activity, eye movements, heart activity, or channel noise. Using these probability outputs, we removed ICA components that exhibited less than 5% brain contributions and non-brain combined probability of greater than 80%. The mean number of rejected components was 4.4 (SD = 4.91). Rejected components were primarily related to artifacts from (a) Eye: 22%; (b) Heart: 12%; (c) Muscle: 58%; and Channel noise: 8%. ACAM-J and control runs were preprocessed using identical steps. For each ACAM-J run, the data from the entire run was cut into separate data segments corresponding to each ACAM-J state. For all analyses, we excluded ACAM-J segments that were less than ~30 s, resulting in the exclusion of a total of 4 segments of data. Consistent with the MRI analyses, data from each control run (of 8 mins) were cut into consecutive one-minute segments, yielding a total of 16 segments for each control condition for each of the two runs of control data.

4.3.2. Spectral analysis (EEG)

Frequency analysis was conducted using a fast Fourier transform using a Hanning window between 1 and 30 Hz at 0.5 Hz frequency intervals, and power at 30–45 Hz was obtained using Slepian multitapers with a smoothing of +/− 3 Hz. The spectrum was divided into the following canonical frequency bands: delta (1–4 Hz), theta (4–8 Hz), alpha (8–13 Hz), beta (13–30 Hz), and gamma (30–45 Hz). A cluster-based permutation statistical method (Maris and Oostenveld, 2007) was adopted to assess the significance of the difference in power between meditation and control (counting, memory, combined) conditions for the different frequency bands. This statistical method controls for the type I error rate arising from multiple comparisons using a nonparametric Monte Carlo randomization. First, channel-wise comparisons in power values were made using an independent samples t-test. A cluster was then defined across adjacent electrodes and frequencies for which t-statistics of the same sign were above the 97.5th quantile (i.e., cluster forming alpha = 0.05, two-sided) of the t-distribution, and the cluster-level test statistics are defined as the sum of t-values within that cluster. If this produces more than one cluster, the maximum is taken as the cluster-level statistic. A null distribution was then obtained by calculating the cluster-level statistic in several shuffled versions of the data (we used 7500 permutations). Cluster-corrected p-values were defined as the proportion of random partitions whose cluster-level test statistic exceeded the one obtained in the original (non-shuffled) data. The significance level for the cluster permutation test was set to 0.025 (corresponding to a false alarm rate of 0.05 in a two-sided test). Spectral analyses were performed separately for each frequency band by each ACAM-J state. As our analyses intended to explore all possible changes in oscillatory power during ACAM-J, we did not further correct the significance level (e.g., significance alpha = 0.025/ divided by the number of tests) for tests across the different frequency bands.

4.3.3. Signal diversity (EEG)

To compare our results with recent findings that indicate that psychedelic states are related to increased EEG signal diversity (Schartner et al., 2017; Timmermann et al., 2023b, 2019), we performed signal diversity analysis using the Lempel-Ziv 1976 algorithm (LZ76). The EEG signal at each electrode was binarized using its mean for each two-second epoch. The LZ76 algorithm then generated a dictionary of unique sub-sequences that provides a raw value that quantifies the temporal diversity for each signal (denoted here as LZs). FieldTrip cluster-based permutation (7500 permutations) testing of independent sample t-statistics was used to determine channels that exhibited significant differences between ACAM-J states and control conditions (using the same procedure as the spectral analyses for multiple-comparison correction).

4.4. Inter-run variability of phenomenology and neural activity

To examine the stability of both the phenomenological ratings and neural metrics, we calculated several reliability/variability measures. For each phenomenology rating, we computed the intra-individual coefficient of variation (i.e., SD/Mean of values). For fMRI measures, specifically GFC and Modularity, we examined the inter-run reliability of these values using intra-class correlation (ICC; 103) across runs for each network. This index indicates the extent to which measures of functional connectivity (GFC, Modularity) within each ACAM-J are likely to be similar and between different ACAM-J are likely to be different from each other. Specifically, ICC estimates and their 95% confidence intervals were calculated using custom R code (R Core Team and others, 2013) from (Ten Hove et al., 2022) based on a mean-rating (k = 27), inter-run consistency, and 2-way random-effects model. For reliability of activity within networks refer to (Ganesan et al., 2024). For EEG, we similarly looked at the ICC of power values averaged across all electrodes for each significant clusters from the ACAM-J vs. counting control comparison.

4.5. Neurophenomenology

Across the various neural metrics described above, we assessed how each is associated with the phenomenology of the ACAM-J. Given a large number of possible permutations between brain indexes and phenomenology, we only report correlation analyses between the primary characteristic experience of each ACAM-J, ACAM-J-wide phenomenological ratings, and neural data that showed consistent differences in the comparison between ACAM-J and control in the primary neural analyses. Specifically, associations between network modularity, GFC, EEG power, and LZ values were examined against phenomenological ratings of ACAM-J2 “bliss/joy”, ACAM-J3 “cool bliss”, ACAM-J4 “equanimity” (i.e., “jhanic neutral qualities”), ACAM-J5–8 “grades of formlessness”, “ACAM-J-wide stability” and “width of attention”.

For fMRI, we used nonparametric Spearman correlations (due to the non-normal distribution of the rating) between network modularity, GFC, and phenomenological ratings to test relations between functional connectivity changes and meditation experience. Due to the exploratory nature of these correlations, we report and interpret uncorrected results in the discussion but specify correlations that also survive FDR correction across the set of phenomenological items (Fig. 5). Correlational analysis results for all ratings and networks are provided in supplementary Fig. S1–2. For EEG, we used a cluster-based permutation method to identify spatial clusters that correlated with phenomenology. For both EEG spectral power as well as LZ values, Spearman’s correlation coefficients were calculated between the power for each frequency band (and LZs), for each channel, across runs. This correlation value was then used to calculate a cluster statistic, similar to the spectral analysis (i.e., cluster forming alpha = 0.05, two-sided). A null distribution of the cluster-statistic was then obtained by calculating the cluster-level statistic in 7500 permuted versions of the data, where any association between EEG power (or LZs) and phenomenological ratings was broken by randomly permuting the values of the ratings. Cluster-corrected p-values were defined as the proportion of random partitions whose cluster-level test statistic exceeded the one obtained in the original (non-shuffled) data. The significance level for the cluster permutation test was set to 0.025 (corresponding to a false positive rate of 0.05 in a two-sided test).

Fig. 5. fMRI Neurophenomenology of ACAM-J:

DMN GFC was positively, and modularity negatively correlated with the characteristics of ACAM-J experiences of “bliss/joy” in ACAM-J2. VN GFC and modularity were also related to the “stability of attention” and “bliss/joy” in the second ACAM-J. Correlations with p < 0.05 are plotted in red, and those that also survive FDR correction are denoted with an * before the p-value (*p).

4.6. Correlation between fMRI GFC and EEG spectral power

Although this study did not record EEG and fMRI simultaneously and the fMRI and EEG runs were not of equal numbers nor paired, therefore precluding pairwise correlations, we nevertheless conducted exploratory correlational analyses between EEG spectral power at delta, alpha, beta, and gamma band activity, as well as global signal diversity (LZ), with the fMRI measure of global functional connectivity (GFC) for networks that displayed the most pronounced effects in our primary ACAM-J vs. control analysis. Specifically, we calculated the mean GFC for each ACAM-J state and control task for the DMN, FPN, and VN networks, and the mean spectral power within each band (or LZ) for the same. GFC and spectral power were then correlated across ACAM-J and controls (Spearman’s correlation) for each EEG electrode and rho values were displayed on a topographic map. We did not perform any type of multiple-comparison correction (e.g., cluster-based permutation) or significance test for these analyses since they were exploratory in nature and these correlations were made between on only eight value-pairs (i. e., the six different ACAM-J states examined in this study plus two controls).

5. Results

5.1. Phenomenology of ACAM-J

Immediately after each run (EEG and fMRI), phenomenological ratings were obtained from the participant in-scanner using Likert-type rating scales (see Methods for additional detail on how ratings were collected and Supplementary for the questionnaire used). Across all runs (a total of 56 ACAM-J and 4 control runs across fMRI and EEG), the difference in mean of phenomenological ratings revealed that the ACAM-J states induced reliable changes in conscious experience (Fig. 1). We note that since we specifically mapped phenomenological items to describe the ACAM-J-specific experience, it is anticipated that these ratings are primarily relevant to specific ACAM-J states and will be relatively stable for control runs. Refer to Supplementary Table S1 for descriptive Statistics (Mean, SD, variability) of key phenomenological ratings in each ACAM-J and Fig. S3 for correlations between all ratings. The reliability of phenomenological change was generally high (coefficient of variability = 0.10 to 0.60). Stability and width of attention, as well as bliss/joy in the second ACAM-J, showed relatively higher variability (0.3 – 0.6) compared to other measures – suggesting that the phenomenology of these specific measures fluctuates across runs.

Fig. 1. Phenomenology of ACAM-J.

Across all runs (fMRI and EEG), phenomenological ratings revealed that the ACAM-J states induced reliable changes in conscious experience compared to controls. Control ratings were similar across tasks. Refer to Supplementary Table S1 for the coefficient of variability for key ratings. For the second ACAM-J, the pleasant state of “bliss/Joy” was heightened, which evolved into the state of increased “cool bliss’ in ACAM-J3. In ACAM-J4, jhanic neutral qualities or “equanimity” was experienced. Additionally, from the third ACAM-J onwards, the meditator reported increased “width of attention.” From ACAM-J5 onwards, “physical sensations” were also reduced compared to control conditions. Consistent with the typical experience of the formless ACAM-J, ACAM-J5–8 had increased “grades of formlessness”. There was also a reduced experience of “sounds,” and “narrative thoughts” across all ACAM-J.

Across runs, phenomenological ratings for the two control tasks were similar. Compared to controls, there was reduced experience of “sounds” and “narrative thoughts” across all ACAM-J. Additionally, from the third ACAM-J onwards, the meditator reported increased “width of attention.” For the second ACAM-J, the pleasant state of “bliss/joy” was heightened and evolved into the state of increased “cool bliss’ in ACAM-J3. In ACAM-J4, the participant experienced jhanic neutral qualities or “equanimity.” Consistent with the typical experience of the formless ACAM-J, ACAM-J5–8 exhibited increased “grades of formlessness.” From ACAM-J5 onwards, “physical sensations” were also reduced compared to control conditions. Of note, these results are expected given that we developed the items to specifically map phenomenology as described by our meditator. Ratings also showed that during the first run of the control task, there was some ACAM-J phenomenology that crept into the control tasks (Supplementary Table S3).

5.2. Resting-State functional connectivity (RSFC)

Here we illustrate results comparing the ACAM-J to the combined controls (i.e., counting + memory) because this approach better emulates a contrast with a resting-state condition (see section on Control conditions). Figures comparing the counting and memory control tasks are provided in the supplementary (Supplementary Fig. S1 and S2, respectively). Refer to the discussion for an analysis of possible reasons explaining the heterogeneity of results when comparing the ACAM-J to the two different control tasks.

5.2.1. Modularity

We assessed fMRI-measured modularity as a measure of hierarchical integrity of canonical resting-state networks (Visual=VN; somatomotor=SMN; dorsal attentional=DAN; ventral attentional/salience=SN; affective=AFN; frontoparietal=FPN; default-mode=DMN) during ACAM-J. Assessment of the difference in modularity between the ACAM-J and both the control tasks combined (counting and memory), the most consistent decreases in modularity were confirmed to be found across the wholebrain (ACAM-J6–8; formless ACAM-J), DMN (all except ACAM-J4 & 5; across form and formless ACAM-J), FPN (ACAM-J2,5–8; across form and formless ACAM-J), VN (ACAM-J3,5–8; in formless ACAM-J), and SN (ACAM-J2,5–8; in formless ACAM-J) networks (Fig. 2A). Compared to the counting control task, ACAM-J significantly decreased the within-network modularity of the DMN (all ACAM-J except ACAM-J4–5; across form and formless ACAM-J), FPN (all ACAM-J), VN (ACAM-J3, 6–8; in formless ACAM-J), and SN (ACAM-J2,6–8) (FDR-corrected; Figure S1A). Decreased in modularity were also found compared to the memory control task, only in a limited number of ACAM-J in the VN (ACAM-J3,5–8; formless ACAM-J), (Supplemental Figure S2A).

Fig. 2. Reduced RSN modularity and increased GFC during ACAM-J compared to combined (both) control conditions.

Networks analysed (VN= Visual; SMN = somatomotor; DAN = dorsal attentional; SN = ventral attentional/salience; AFN = affective; FPN = frontoparietal; DMN = default-mode). (A) Analysis of modular integrity (i.e., modularity) for ACAM-J (blue boxes) versus both controls combined (black box) showed significant reductions in modularity in 4 out of 7 networks (DMN, FPN, VN, SN; FDR-correction, *p < 0.05, **p < 0.01, ***p < 0.001), across both form and formless ACAM-J. (B) Increases in global functional connectivity (GFC) in 6 of 7 networks (all except AFN; FDR-correction, *p < 0.05, **p < 0.01, ***p < 0.001), as well as the whole-brain (Total GFC), are found across a range of ACAM-J (but primarily in the formless ACAM-J). See Supplementary Materials for results comparing the counting control and memory control separately (Fig. S1B and S2B).

5.2.2. Global functional connectivity

To determine the effects of ACAM-J on brain connectivity, we computed global functional connectivity (GFC), which indexes the average FC (i.e., correlation) for a given region with all other regions in the brain (see Materials and Methods for details). Significant increases in GFC were identified for the wholebrain (ACAM-J5; and across formless ACAM-J), DMN (ACAM-J3,5–8; across formless ACAM-J), VN (ACAM-J3,5–8; across formless ACAM-J), FPN (ACAM-J5; across formless ACAM-J) and DAN (ACAM-J3,5–8; across formless ACAM-J) (Fig. 2B) when comparing the combined control tasks and ACAM-J. Significant increases in GFC were identified again in the DMN (ACAM-J2–3,5–8; across formless ACAM-J), FPN (across all ACAM-J), VN (ACAM-J3,5–8; across formless ACAM-J), and SN (ACAM-J5; across formless ACAM-J) (FDR-corrected across all ACAM-J-control pairwise comparisons for a specific network; Supplemental Figure S1B) when comparing the counting control task and ACAM-J. Compared to the memory control task, this increase in GFC was only observed in the VN and DMN across formless ACAM-J (Supplemental Figure S2B). Notably, the most consistent ACAM-J-wide increases compared to control conditions were in the DMN, and VN and in the formless ACAM-J (i.e., ACAM-J5–8).

5.2.3. Between-network segregation

To complement these analyses, we calculated functional connectivity between canonical functional networks to examine between-network segregation during ACAM-J. Comparing the ACAM-J with combined controls showed that the most consistent increases in between-network connectivity during ACAM-J were between the DMN and other networks, especially during formless ACAM-J (Fig. 3). Compared to the counting control task, the ACAM-J also revealed decreased between-network segregation (i.e., increased between-network connectivity) between the FPN, DMN, and other networks (FDR-corrected across all pairwise tests; Supplemental Figure S1C). Compared to the memory control condition, ACAM-J were only related to increased connectivity of DMN with the VN and SMN, mostly in the formless ACAM-J (i.e., ACAM-J5–8) (Supplemental Figure S2C). Compared to the memory condition, the FPN network was less connected with other networks during ACAM-J6–8.

Fig. 3. Between-network Functional Connectivity.

Decreased between-network segregation was especially pronounced between the DMN (and to some extent FPN) network, and other networks during the ACAM-J compared to the combined (Both) control conditions (*p < 0.05, FDR-corrected across all pairwise tests). Hedges’ g was used to quantify the effect size of the difference in pairwise network connectivity between control and ACAM-J. Comparison to the memory control task only showed increases in DMN connectivity in a limited number of ACAM-J (ACAM-J5–8), while comparing the ACAM-J with the counting control showed similar results to both controls combined (Supplementary Fig. S1C and Fig. S2C).

5.2.4. Inter-run reliability of RSFC measures

The results of the ICC (inter-run reliability) analysis for each network’s RSFC metrics are presented in Supplementary Table S2A. Moderate (0.5 < ICC < 0.75) to good (0.75 < ICC < 0.90) reliability was observed for networks that exhibited significant differences (from control) in these metrics, such as the DMN and VN. FPN modularity, while significantly different from control, was observed to have a low (< 0.10) average ICC value, indicating reduced within-jhana reliability across runs. Our previous separate analyses showed that activity in regions of the FPN has high reliability for several jhanas, and including phenomenology items as explanatory variables increased reliability for these metrics (Ganesan et al., 2024).

5.3. EEG spectral analysis and signal diversity

We assessed EEG power in different frequency bands to establish the impact of ACAM-J on the synchronized neural activity of the brain. Compared to the combined controls analyses revealed decreased oscillatory power during ACAM-J across multiple frequency bands (Fig. 4A), with the change in power most pronounced in the alpha, theta, and delta (1–13 Hz) bands during ACAM-J2–4 and 6–8.). There was also a more spatially limited decrease in posterior beta power. Plotting the mean of the power spectrum across all channels for each ACAM-J (Fig. 4B), it is evident that whole-brain power was decreased most prominently in the alpha and delta bands. Similar power differences were found when comparing the ACAM-J to the counting and memory control conditions separately (Supplementary Figure S1D and S2D). The ICC (inter-run reliability) of EEG power for significant clusters is tabulated in Supplementary Table S2B. All clusters showed good reliability across runs. Additionally, widespread (somewhat lateralized to the right hemisphere) increases in signal diversity, as determined via Lempel-Ziv complexity (LZ), was observed during the ACAM-J compared to the combined control (Fig. 4A, right panel). Comparing the counting and memory controls to the ACAM-J also revealed similar increases in LZ across ACAM-J (Supplementary Figure S1D and S2D, right panel).

Fig. 4.

Effect of ACAM-J on EEG spectral power and signal diversity. Filled circles correspond to clusters p < 0.01 and asterisks for clusters p < 0.05 (A) There were widespread broadband decreases in power, especially in the alpha band during ACAM-J2–4, and 6–8. Signal diversity (estimated via Lempel-Ziv, [LZ]) was also increased across ACAM-J compared to the combined (both) control conditions; see supplementary for comparison to memory and counting task separately. (B) Whole-brain spectral power exhibited consistent differences, especially in the delta-to-alpha bands during ACAM-J3 and the formless ACAM-J (6–8).

5.4. Parallel changes in fMRI and EEG

Our exploratory analysis looking at correlations between EEG power, LZ, and fMRI GFC (Figure S6) across all ACAM-J and controls mirrors the modality-independent changes in both EEG and fMRI measured activity (Figs. 2 & 4). The topoplots in Figure S6 are intended to depict the general trend across the brain on how EEG spectral power/signal diversity (LZ) is associated with the GFC of large-scale functional networks. These trends advance our understanding of the ACAM-J-specific findings in a meaningful way, by demonstrating that the directionality of change within EEG and fMRI metrics significantly interrelates between modalities and is consistent with independent ACAM-J vs. control comparisons. For example, GFC increases in the DMN, VN, and FPN showed a pattern of being negatively associated with widespread delta-to-alpha power (as seen by negative rho across the brain), a finding that parallels the ACAM-J vs. control results showing a reduction in power in these bands. Similarly, increases in LZ across the brain were positively associated with increases in GFC across networks.

5.5. Neurophenomenology

We then examined how changes in fMRI and EEG measures relate to phenomenology during ACAM-J. The correlation between run-by-run phenomenology ratings and neural metrics revealed that GFC for DMN was positively, and modularity negatively correlated with the characteristic ACAM-J experience of bliss/joy in ACAM-J2. GFC in the VN was also positively correlated with ACAM-J-wide stability of attention, bliss/joy in ACAM-J2, and equanimity in ACAM-J4. Modularity in the VN was negatively correlated with the “stability of attention”, “bliss/joy” in ACAM-J2, and “cool bliss” during ACAM-J3 (Fig. 5). An opposite pattern was found during the formless ACAM-J where DMN GFC showed a trend (p = 0.06) of being negatively, and modularity positively correlated with grades of formlessness. For correlations between within-network modularity, GFC, and all phenomenology ratings, refer to supplemental Figures S4–5.

Paralleling the fMRI results, “bliss/joy” in ACAM-J2 was also associated with widespread decreases in EEG power in the beta and gamma bands (Fig. 6, third column), as well as decreased LZ. The “width of attention” across ACAM-J was positively correlated with widespread increases in LZ. Regarding relations among other phenomenological ratings and EEG metrics, “stability of attention” across ACAM-J was positively correlated with widespread increases in power in the alpha and theta bands (Fig. 6, first column), whereas the “width of attention” was negatively associated with widespread alpha power (Fig. 6, second column). Formlessness (i.e., “grades of formlessness) in the formless ACAM-J5–8 was positively related to alpha power across a large brain topography (Fig. 6, columns 6–7).

Fig. 6. EEG Neurophenomenology of ACAM-J:

Filled circles correspond to clusters p < 0.01 and asterisks for clusters p < 0.05. Stability of attention across ACAM-J was positively correlated with widespread synchronization in the alpha and theta bands (first column), whereas the width of attention was negatively associated with widespread alpha power and positively associated with LZ (second column). Bliss/joy in ACAM-J2 was also associated with widespread beta and gamma desynchronization (third column) and decreased LZ. Formlessness in the ACAM-J5–8 was positively related to alpha power across a large brain topography (columns 6–7).

6. Discussion

The current study provides novel insights into the neuroscience of the advanced meditative states known as jhanas, which are ecstatic non-ordinary states of consciousness characterized by profound concentration. It presents a unique neurophenomenological investigation of advanced concentration absorption meditation (ACAM-J) using extensively sampled multimodal neural data and detailed analyses – providing an important development in the understanding of the neural correlates of advanced meditative states (Galante et al., 2023; Sparby and Sacchet, 2022; Wright et al., 2023). The study significantly complements recent findings (Yang et al., 2024a) of reliable changes in brain activity patterns associated with ACAM-J and their relation to phenomenology. Our results revealed alterations in the brain’s functional connectivity, including network modular disintegration and desegregation, and changes in EEG activity, including decreased broadband spectral power (especially marked in delta-to-alpha bands) and increased spontaneous signal diversity (or entropy). Relations were observed between specific phenomenological measures, particularly between changes in DMN connectivity and decreased oscillatory power, with the pleasant experience of bliss/joy in ACAM-J2. More broadly, these results reinforce the view that fine-grained first-person (phenomenological) data can differentiate and help to understand better the neural dynamics of non-ordinary states of consciousness (Timmermann et al., 2023a).

6.1. fMRI: functional connectivity and neurophenomenology

Regarding fMRI functional connectivity, our results included increased GFC and decreased modularity across transmodal cortices, specifically the DMN and FPN, as well as the sensory VN, during the ACAM-J compared to the combined control tasks. DMN and VN also showed good reliability across runs, while the reliability of FPN modularity was low (< 0.10) – which may reflect the dynamic nature of FPN modularity as it varies across ACAM-J runs, to mediate context-dependent cognitive processes during meditation (e.g., differences in the initial object of meditation). Our previous separate analyses also showed that activity (as opposed to modularity) in regions of the FPN has high reliability for several jhanas, and including phenomenology items as explanatory variables increased reliability for these metrics (refer to Ganesan et al., 2024). Increased DMN connectivity during ACAM-J is consistent with previous studies on meditation. Several studies have reported increased functional connectivity between the DMN and FPN during focused meditation (Brewer et al., 2011; Jang et al., 2011), as well as general mindfulness and mindfulness meditation (Bremer et al., 2022); reviewed in (Sezer et al., 2022) – findings that have recently been corroborated by meta-analysis of brain activity and connectivity (Ganesan et al., 2022). The DMN, a ‘task-negative’ network, is typically associated with a self-referential or internally focused information processing (Raichle et al., 2001), and is generally anti-correlated with the ‘task-positive’ FPN network, which supports cognitive control and allocation of resources to externalized information processing (Fox et al., 2005). Greater connectivity between these networks during focused mindfulness meditation suggests a breakdown of this internal/external dichotomy. Prior studies have also reported weaker functional connectivity between DMN regions involved in self-referential processing and emotional appraisal in experienced meditators (Taylor et al., 2013), which is consistent with our results of decreased brain modularity during ACAM-J. These similarities across studies suggest that increased DMN connectivity and DMN-FPN interactions may be common neural correlates of meditation including advanced meditative states.

When examining the findings for each control task separately, many fMRI functional connectivity (FC) differences were present with the counting task but not the memory task. This suggests that these specific results may be driven by cognitive processes uniquely engaged by the counting task. We hypothesized that the counting task, involving fluctuating attention, would serve as a general contrast to the sustained focus required in ACAM-J, while the memory task, which emphasizes self-referential thought and mind-wandering, would help clarify the self-deconstructive nature of the formless ACAM-J. Robust FC differences between the DMN and FPN were observed when comparing ACAM-J to the counting task, but these differences were absent in the memory task, where increased global functional connectivity (GFC) was only seen in the formless ACAM-J for the DMN and VN. This suggests some shared functional connectivity patterns between ACAM-J and the memory control task, though this does not imply they share the same cognitive or phenomenological processes. Phenomenological ratings indicated that some ACAM-J-related experiences emerged during the first fMRI control run (see Supplementary Table S3). To address this potential confound, supplementary analyses using only the “purest” control runs showed that FC differences were stronger for the counting control (increased GFC and decreased modularity), while the memory task results remained unchanged. These findings suggest that the phenomenological differences between ACAM-J and the counting task are more pronounced than those between ACAM-J and the memory task. Future studies employing diverse control conditions with distinct cognitive and neural properties could help better isolate the unique characteristics of ACAM-J. Nonetheless, the core findings of increased connectivity and decreased modularity were still present when comparing ACAM-J to both control tasks combined, reinforcing the validity of our conclusions. Additionally, while the fMRI results varied between control tasks, EEG findings were more consistent. which suggests that while some aspects of ACAM-J overlap with resting-state cognition, distinct features—particularly in brain entropy and attentional dynamics—differentiate ACAM-J.

DMN functional connectivity was found to also relate to the ACAM-J-specific experiences of bliss/joy during the second ACAM-J and formlessness in the formless ACAM-J. These associations indicate possible affective processes underlying these states. The DMN and FPN are at the top of the hierarchy of cognition (Margulies et al., 2016), and – according to the predictive processing hypothesis - are dedicated to interpreting incoming external information using existing internal information or priors acquired through experience and learning. A possible explanation for the experience of joy in the second ACAM-J being associated with increased functional connectivity and decreased modularity of the DMN is that this phenomenon may be experienced by curtailing the influence of prior beliefs in the pleasant object of meditation (e.g., the breath, visual mental imagery), and thereby amplifying the inherently pleasant and novel quality of the sensation. The object of meditation might be internally generated visual imagery – as evidenced by increased VN GFC during ACAM-J which was also positively related to bliss/joy – and may reflect a sort of “seeing with eyes-shut” experience similar to those reported during psychedelic experiences (Carhart-Harris et al., 2016). Additionally, since the influence of prior beliefs and hierarchical cognition is positively related to activity and modularity in the DMN (Fischmeister et al., 2017; Smallwood et al., 2021) and other transmodal networks such as the FPN (Hogeveen et al., 2022), ACAM-J may generate joyful states by disintegrating the modularity of these top-down networks and facilitating increased information exchange between these networks and lower-level sensory networks such as the VN (Lutz et al., 2019). The only other fMRI study of ACAM-J (with a sample of only 2 runs) also reported increased activity in the reward region of the brain (Hagerty et al., 2013) and linked their findings to pleasurable experiences during ACAM-J.

Specifically, during the formless ACAM-J (i.e., ACAM-J5–8) the object of focus shifts from being concrete (e.g., the breath, visual imagery) to being very subtle and related to the fundamental structure of experience itself, such as infinite yet void space and consciousness, which may deconstruct the concept of self. In some advanced practitioners, such deconstruction of consciousness may lead to the experience of cessations (Chowdhury et al., 2023; Laukkonen et al., 2023). Our current findings of increased DMN global connectivity and decreases in modularity during the formless ACAM-J are consistent with this disintegration of subject-object dichotomy – conceptualized and represented by a higher degree of functional connectivity between networks commonly associated with internal and external attention (Josipovic, 2019, 2014). The formless ACAM-J (ACAM-J5–8) also showed relatively greater decreases in modularity and increases in GFC, particularly in the DMN, suggesting a more pronounced shift in the brain’s FC dynamics as the participant progressed into these deeper meditative states. Notably, although DMN global connectivity showed an increase and modularity decreased in the formless ACAM-J, “grades of formlessness” showed a pattern (p = 0.06) of being negatively correlated with DMN global connectivity (and positively with DMN modularity). This finding may highlight a key difference in the experience of formlessness (in formless ACAM-J) versus non-dual awareness as practiced in other meditation types (Josipovic, 2019, 2014, 2010). Although the background voidness in the formless ACAM-J may be enhanced by increased connectivity between DMN and other networks, the focused attention needed to explicitly notice the underlying “grades of formlessness” may require functional segregation of the DMN (Schooler et al., 2011). Future research should examine these hypotheses with independent measures of non-dual and meta-awareness.

6.2. EEG: spectral power, signal diversity, and neurophenomenology

Our EEG results included decreased oscillatory power during ACAM-J compared to both the control tasks, with widespread power reduction in the alpha, delta, and theta frequency bands and posterior beta during ACAM-J3, 4, and 6–8. Additionally, we found widespread increases in signal diversity during the ACAM-J compared to both control tasks. Increased alpha power is related to top-down processing and high-level psychological functioning (Bastos et al., 2015, 2012; Klimesch, 2012). Specifically, alpha oscillations may aid in predicting and thereby facilitating the processing of lower-level sensory inputs (Mayer et al., 2015), effectively increasing the brain’s efficiency in interacting with the environment through inhibitory mechanisms. Alpha-suppression, therefore, may be related to the disintegration of this hierarchical cognition and related inhibition. More generally, the decrease in low-frequency multiband power during the ACAM-J may reflect the attenuation of top-down predictive/inhibitory processes (Arnal and Giraud, 2012; Engel et al., 2001). For example, theta oscillations are related to the inhibition of task-irrelevant brain areas (Snipes et al., 2022), both during cognitive demands (Cavanagh and Frank, 2014; Ratcliffe et al., 2022) and temporally local sleep (Siclari and Tononi, 2017). The decrease in theta oscillations observed during ACAM-J, therefore, may indicate the curtailing of this inhibition, leading to an unconstrained mode of cognition.

It is worth noting that a number of previous studies have reported increased alpha power during mindfulness meditation (Lee et al., 2018; Lomas et al., 2015), which is thought to be related to enhanced present-moment awareness, whereas we found the ACAM-J to be associated with decreased widespread alpha oscillations. Interestingly, the ACAM-J-wide stability of attention was indeed positively associated with alpha power across the brain, whereas the width of attention was negatively associated with alpha power. Together, these findings suggest that stability and width (of attention) are both phenomenologically and neurally distinct phenomena, with stability requiring attentional control, marked by increased alpha and theta power (Josipovic, 2014; Tang et al., 2015), and width of attention possibly relating to how widespread this attention is during the ACAM-J state. As alpha power is commonly associated with an attentional suppression mechanism (Foxe and Snyder, 2011), the overall reduction of alpha power during the ACAM-J may be explained by the distinct nature of these advanced meditative states, reflecting a deeper yet wider state of concentration compared to other meditation practices. Our finding that only the width (and not stability) of attention showed a positive association with EEG signal diversity (i.e., LZ, a metric that overall also increased during ACAM-J) supports the interpretation that ACAM-J may induce the transition to a more entropic and differentiated pattern of neural activity (Carhart-Harris, 2018) through increased width of attention.

In the second ACAM-J, the subjective experience of bliss/joy was associated with decreased beta and gamma oscillations, which have been associated with top-down predictions and prediction errors, respectively (Engel and Fries, 2010; Strube et al., 2021) – and a modest decrease in LZ, which is a measure of signal entropy. According to a hierarchical schema of cognition, beta synchronization may be specifically related to the activation of predictive priors, whereas gamma band activity may represent the subsequent errors in the prediction (van Pelt et al., 2016). The reduction of both beta and gamma power related to bliss/joy suggests that this phenomenon is associated with the reduced engagement of general predictive processing in the brain. This suggests that during ACAM-J2, the brain experiences a state of relaxation and “letting go”, which may be related to the pleasant subjective experience of bliss/joy. More generally, high-frequency beta waves have been associated with an awakened, aroused state involved in conscious thought and logical thinking (Abhang et al., 2016), and gamma waves with working memory and movement (Abhang et al., 2016; Whitham et al., 2007) – the attenuation of these power bands therefore indicates the decay of conscious thought and arousal.

Interestingly, we found that alpha power in the formless ACAM-J (5–8) was positively related to the experience of formlessness and stability of attention, a finding that parallels the fMRI connectivity findings of a negative correlation between formlessness and DMN GFC. There is a whole brain drop in alpha power during formless ACAM-J and this decrease is associated with an increased width of attention. Conversely, stability of attention is positively related to alpha and theta power. These findings corroborate that (a) stability and width of attention are phenomenologically and neurally distinct; and (b) the deliberate experiencing of formlessness in formless ACAM-J may require engaging a process similar to increased stability of attention (increased alpha, increased VN GFC) that opposes the generally high width of attention (decrease alpha, decreased DMN GFC) during those states.

6.3. Relation to other theories of consciousness

It is worth considering how our findings align with other theories of consciousness, offering a distinct perspective on ACAM-J. For instance, Global Neuronal Workspace Theory (Mashour et al., 2020) posits that specialized brain regions facilitate conscious access by broadcasting information across neural networks. Our results, showing decreased modularity and increased connectivity between networks such as the FPN, VN, and DMN, suggest that ACAM-J states may involve the “broadcasting” of pleasant sensations from lower-order sensory regions to other areas of the brain, thus sustaining altered states of consciousness through focused attention. In contrast, Higher-Order Thought (HOT) Theory (Brown et al., 2019) proposes that conscious awareness emerges from higher-order thoughts about one’s mental states. Interpreted through this framework, it is the awareness of interoceptive or sensory phenomena, such as breath or bodily sensations, that gives rise to phenomenology—such as experiences of bliss or joy. Our findings, which demonstrate associations between DMN-VN connectivity and phenomenology during meditation, indicate that ACAM-J practice likely engages higher-order cognitive processes. These processes may transform basic sensations into more complex and enriched subjective experiences.

6.4. Comparison to psychedelics

Our findings can be related to studies showing similar alterations in functional connectivity (DMN, FPN, VN) during psychedelic states (Carhart-Harris et al., 2016; Tagliazucchi et al., 2016; Timmermann et al., 2019). These states result from a breakdown in hierarchical brain processing, leading to a globally integrated state of consciousness and ego-dissolution or self-transcendence (Tagliazucchi et al., 2016), similar to non-dual awareness in advanced meditation (Sacchet et al., 2024). However, our results suggest a distinct neurophenomenology for ACAM-J. Psychedelic intensity is linked to increased global connectivity in the DMN, FPN, and SN, greater signal diversity or entropy. (Timmermann et al., 2023b, Timmermann et al., 2019), and temporally dynamic experiences (e.g., bliss, imagery, ego-dissolution) associated with distinct brain states of network integration and segregation (Carhart-Harris et al., 2016; Luppi et al., 2021; Tagliazucchi et al., 2016). In contrast, ACAM-J involve focused attention, reflected in correlations between VN connectivity and attention metrics (e.g., stability, width of attention), with differences in connectivity and modularity primarily observed in the DMN and VN during formless ACAM-J (3 and 5–8). In EEG studies, psychedelics show widespread alpha suppression and increased delta power during immersive experiences ( Timmermann et al., 2019), or reductions across all frequency bands (Muthukumaraswamy et al., 2013; Riba et al., 2004; Tagliazucchi et al., 2016), often linked to specific phenomenology. ACAM-J, however, shows a more complex relationship between EEG metrics and phenomenology, where certain experiences (e.g., stability, width of attention, formlessness) exhibit inverse associations with general EEG trends (Chowdhury et al., 2023).

These neural differences likely reflect distinctions in cognition and experience between psychedelics and ACAM-J. ACAM-J, with its deliberate focus and sustained attention, contrasts with the unpredictable, passive nature of psychedelic experiences. Psychedelics are usually characterized by vivid, pervasive visual imagery (Studerus et al., 2010), while in ACAM-J, such imagery, if present, tends to be more sporadic and spontaneous (Lindahl et al., 2014). ACAM-J’s focus on sustained attention may involve distinct network integration mechanisms compared to the effects of psychedelics. Further research could clarify these neurophenomenological differences.

6.5. Limitations

The current study used a single case design in which we extensively sampled phenomenological and neural data during ACAM-J as practiced by a single adept meditator. This approach of extensively sampling events of interest (i.e., ACAM-J) in a single participant has proved to be a powerful tool in neuroscience to derive insights beyond a single or several repeats of a measurement (Levenson et al., 2012; Poldrack et al., 2015). Despite the case study design being a major strength of the current study, such a protocol simultaneously reduces the generalizability of the current results. Replication of the current results across participants (i.e., a group study of advanced ACAM-J practitioners) represents an important avenue for future research in this area. For instance, one explanation for why we did not find significant relationships between some of our phenomenological ratings and neural metrics is that these ratings showed little variability and therefore precluded the likelihood of detecting linear associations with neural measures. The phenomenology items were also specifically developed with the input of the participant and therefore may be argued as being idiosyncratic. However, these items were generally congruent with the typical experience of ACAM-J as stated in ancient texts as well as other adept practitioners (Brasington, 2015; Khema, 2022). The current study’s strength lies in monitoring a single adept meditator, enabling high-quality and stable data collection. However, there exist variations in strategies used by other meditators to attain ACAM-J, which may yield distinct neural and phenomenological signatures. For example, a recent study (Volodina et al., 2021) identified distinct subgroups among meditators with specific strategies reflected in neural signatures, highlighting the value of group studies for exploring such variations. Future studies with multiple participants of varying proficiency in levels of ACAM-J practice may introduce greater variability in the intensity of the phenomenological experiences and allow for the establishment of the broader validity of the current findings. Relatedly, future studies may explore whether variation in ACAM-J expertise, practice duration, and phenomenology also explain any variation in objective measures of concentration or cognitive control outside of meditation across practitioners. As each ACAM-J state builds on the previous one, it is not possible to randomly assign the order of ACAM-J across runs, and a possibility is that time order effects (e.g., early vs. late ACAM-J) drive some of our findings – this also cautioned against conducting direct statistical comparisons between form and formless ACAM-J. Similarly, although the randomized order of control tasks within sessions mitigates possible confounding effects, their timing across days—before meditation on the first and last days—may introduce sequence and nonspecific effects (e.g., time-of-day, fatigue). Future studies may devise shorter control tasks that can be paired and counterbalanced with each meditation run to control these effects. Nevertheless, differences in modularity and global connectivity spanned across both early and later ACAM-J, and EEG power and LZ changes did not show any apparent time-dependent trends, reducing the likelihood of our results being an artifact of order effects.

7. Conclusion

The current study provides important insights into the neural correlates of advanced meditation states, specifically ecstatic advanced concentrative absorption meditation known as jhanas (ACAM-J), and highlights the potential for multimodal neuroimaging and neurophenomenological techniques for investigating altered states of consciousness. Further research with group cohorts is needed to replicate and extend these findings, and to elucidate the underlying mechanisms of altered states of consciousness in greater detail – in particular advanced states of meditation (Wright et al., 2023) and related meditative development (Galante et al., 2023). This research promises to provide a scientific understanding of advanced meditation that will facilitate the sharing of these experiences in both clinical and non-clinical contexts (Sacchet et al., 2024).

Data/code availability statement

The data that support the findings of this study are available on request from the corresponding author with a formal data-sharing agreement. The data are not publicly available due to privacy or ethical restrictions.