Comparing psychedelic and 
meditation experience reports with 
natural language processing
Konsta Kallio-Mannila1, Rosa Salmela2 & Jussi Jylkkä3
Psychedelics and meditation are known for their potential to induce personally meaningful and even 
transformative experiences. However, it is unclear how similar these experiences are, or how they 
differ from each other. This explorative study used natural language processing (NLP) methods to 
compare reports of personally meaningful subjective experiences facilitated by either psychedelic 
substances or meditation. Participants (N = 197) wrote open-ended narrative reports about their 
most meaningful experience facilitated either with psychedelics (n = 134) or meditation (n = 63). These 
reports were analysed with text similarity analyses, topic modeling and sentiment analysis. The 
semantic and lexical contents of the reports were highly similar and both groups expressed positive 
emotions on average. However, psychedelic experience reports were more emotionally charged, 
showing higher levels of positive and negative sentiments compared to more neutral meditation 
experiences. These results suggest that the two types of subjective experiences might be quite similar 
in general, but emotional intensity could be a distinguishing factor between them. Challenges with 
the NLP methods and the dataset limit the conclusions that can be drawn from the study. However, it 
offers new hypotheses and suggestions for future research on transformative experiences.
Keywords Psychedelics, Hallucinogens, Meditation, Transformative experience, Natural language 
processing, Text similarity analysis, TF-IDF, Topic modeling, LDA, Sentiment analysis
Humans have long sought out experiences that are deeply meaningful, impactful, and potentially 
transformative1. These powerful experiences can take many forms—from spiritual and religious encounters to 
intense psychological insights. What they share is their ability to profoundly affect an individual, often leading 
to lasting changes in beliefs, values, identity, and behaviour2. Transformative experiences have been defined 
as "brief experiences, perceived as extraordinary and unique, entailing durable and/or irreversible outcomes, 
which contribute to changing individuals’ self-conception, worldviews, and view of others, as well as their own 
personality and identity" (p. 14)1. They typically involve an epistemic expansion—gaining new knowledge or 
insight about oneself and the world—as well as heightened emotional complexity and intensity1. Among many 
practices that can lead to transformative experiences, two have gained particular scientific and cultural interest 
in recent years: psychedelic use and meditation3,4.
Psychedelic and meditation experiences have been argued to share phenomenological similarities5–7 but 
have rarely been directly compared to each other. In this study, we used natural language processing (NLP) to 
compare open reports of personally meaningful experiences, facilitated by either psychedelics or meditation. 
The aim was to examine similarities and differences between the two types of reports, and also to examine the 
usefulness and value of NLP in this context.
Psychedelics have been used for centuries for their mystical, spiritual and therapeutic properties, and 
scientific interest in these substances has grown dramatically during the recent years. Many of these substances 
have deep indigenous and ceremonial roots—for instance, psilocybin-containing mushrooms, ayahuasca, and 
mescaline-bearing cacti such as peyote have long been used in ceremonial and healing contexts by various 
Indigenous cultures of the Americas. The term “psychedelic” literally means "mind-revealing" and refers to a 
class of substances that can induce dramatic changes in perception, self-awareness, and inner experience8,9. 
While “classical” psychedelics (e.g., psilocybin, LSD, DMT, ayahuasca) are defined based on their common 
mechanism of action as agonists of serotonin 2A (5-HT2A) receptors10, recent research has begun to consider 
a broader range of substances that can induce psychedelic-like experiences through diverse mechanisms. 
Ketamine acts as an NMDA receptor antagonist, producing detachment and perceptual alterations. It may 
1Department of Psychology, University of Turku, Turku, Finland. 2Turku Research Institute for Learning Analytics, 
University of Turku, Turku, Finland. 3Department of Psychology, Faculty of Arts, Psychology and Theology, Åbo 
Akademi University, Turku, Finland. email: konsta.kalliomannila@gmail.com
OPEN
Scientific Reports |        (2025) 15:43903 1| https://doi.org/10.1038/s41598-025-27724-0
www.nature.com/scientificreports
share a common therapeutic mechanism with classical psychedelics through glutamate-driven neuroplasticity9. 
MDMA, primarily affecting serotonin release and reuptake, produces empathogenic effects and mild perceptual 
changes11. Psychoactive variants of the cannabis plant contain tetrahydrocannabinol (THC) that acts as an 
agonist on the endocannabinoid receptor CB1 and can produce psychedelic-type effects, including mystical-type 
experiences12. Philosopher Aidan Lyon has argued that whether a substance counts as a psychedelic depends 
primarily on its psychoactive effects, and thus a broad range of substances can be considered as psychedelic13. 
This hypothesis is supported by qualitative data that various psychoactive substances have psychedelic-type 
effects14. Therefore, we have adapted a wide definition of psychedelics that encompasses substances capable of 
inducing significant alterations in consciousness, regardless of their primary mechanism of action.
Current research on psychedelics has mainly focused on treating mental health disorders3,9,14–16, since 
this is an area where these substances hold great potential. However, the therapeutic mechanism and the 
phenomenological content of the psychedelic experience are still areas where our knowledge is limited. It has 
been argued that the subjective effects are necessary for the therapeutic benefits of psychedelics17–19, and this 
encourages us to study the phenomenological content of the experiences further.
Meditation, on the other hand, has been practiced for millennia and is also known to induce altered states 
of consciousness, which can be personally meaningful and even transformative1. It can be defined as a set of 
practices which aim to control one’s cognitive processes, especially those related to the self and attention20, and 
it can alter the sense of self, time and space even in laboratory settings21,22 or produce mystical-type experiences 
and insights23. Meditation programs have been shown to have small to moderate well-being effects on multiple 
dimensions of psychological stress, such as anxiety and depression4.
These meditation practices consist of many techniques, including silent sitting or lying down, breathwork, 
voicework, such as chanting or mantra repetition and bodywork such as body scan practices. Laukkonen and 
Slagter24 categorize meditation practices into three fundamental types: focused attention (FA), open monitoring 
(OM), and non-dual awareness (ND). FA meditation involves sustaining attention on a specific object, such 
as the breath, to reduce mind-wandering. OM meditation expands awareness to observe all present-moment 
experiences without judgment. Mindfulness meditation, which has gained significant popularity in recent 
years, often combines elements of both FA and OM, emphasizing non-judgmental awareness of one’s thoughts, 
feelings, and sensations25. ND meditation differs from the other practices by aiming to dissolve the subject–
object distinction entirely, potentially revealing a state of “pure awareness” without conceptual content24.
Laukkonen and Slagter24 propose that these practices gradually decrease the “temporal depth” of mental 
processing, bringing attention closer to immediate sensory experience and away from abstract thinking about past 
and future, potentially leading to alterations in one’s sense of self and reality. Their theory is based on predictive 
coding framework, according to which the brain continually generates and updates models of the world to 
minimize prediction errors between expectations and sensory input26. Meditation is proposed to progressively 
reduce reliance on such predictive models, allowing perception to be less shaped by prior expectations and more 
directly grounded in present experience. Neuroimaging studies further indicate that meditation modulates large-
scale brain networks (e.g., the default mode, salience, and executive control networks) and produces structural 
and functional changes in regions involved in attention, emotion regulation, and self-awareness27.
The states induced by psychedelics and meditation have been compared before, but the studies have 
concentrated more on the changes in brain activity28,29 and well-being effects30 than the phenomenology of 
the experiences. Also, some studies have examined the well-being effects5 and phenomenological effects31 of 
combining psychedelic use with meditation. However, as far as we are aware, prior to the collection of the present 
dataset, so far no one has empirically compared the phenomenology of psychedelic experiences to meditation. 
This study aims to fill this gap by exploring how reports of personally meaningful experiences from psychedelic 
use compare with those from meditation practice. The objective is to uncover what these experiences share in 
common and how they differ, focusing on narrative reports from individuals who have used psychedelics or 
practiced meditation in natural settings.
The present dataset included a fully open question where the participant was asked to describe all aspects 
of the experience, as well as three thematic questions, including one about insight experiences. The latter were 
previously qualitatively analysed by Jylkkä et al.32. In that study, the insights were classified into three main 
types: mystical-type, psychological, and philosophical-existential (including value insights). Mystical-type 
insights were slightly more common in meditation reports and value insights were more frequent in psychedelic 
reports; otherwise, the reports were very similar. Here, we take a different approach and apply NLP methods 
to the complete narrative reports to provide a more comprehensive comparison of the experiences. The main 
difference to the previous qualitative analysis is that the NLP approach can be considered as less driven by the 
researcher’s subjective conceptions and interpretations of the data. The nature of this study is explorative and 
therefore no hypotheses were set beforehand.
NLP refers to a set of methods for analysing textual information with computers. NLP methods are often 
very well fitted for exploratory text analysis, and they offer speed, efficiency, scalability and consistency over 
traditional qualitative text analysis by humans33. NLP can be viewed as bridging the gap between quantitative 
and qualitative approaches: like qualitative methods, it can capture nuanced meanings and themes in text, while 
offering the systematic, reproducible analysis typical of quantitative methods. Therefore, NLP methods were 
chosen for comparing the narrative reports about psychedelic and meditation experiences. More specifically, the 
participants’ open narrative reports were analysed using text similarity analyses, topic modeling and sentiment 
analysis. These methods, when applied together, should provide a comprehensive way to examine the content, 
common themes, and emotional aspects of the participants’ written experiences.
Scientific Reports |        (2025) 15:43903 2| https://doi.org/10.1038/s41598-025-27724-0
www.nature.com/scientificreports/
Methods
Procedure
Two online surveys were distributed internationally with the aim to recruit people with a personally meaningful 
experience facilitated by psychedelic substances or meditation practice. The survey language was English. The 
two surveys were identical in every other part, except for one asked about a psychedelic experience and the other 
about meditation. The study recruited participants through a combination of targeted outreach to psychedelic and 
meditation societies, as well as social media-based snowball sampling. To qualify for participation, individuals 
needed to be at least 18  years old and have experienced a personally highly meaningful event facilitated by 
either a psychoactive substance or meditation. Psychoactive substances were broadly defined as any compounds 
capable of inducing altered states of consciousness, including but not limited to LSD, psilocybin, cannabis, 
MDMA, ketamine, and ibogaine. This was done to focus on experiences that the participants considered as 
personally meaningful or transformative, irrespective of which type of psychoactive substance or meditation 
they were facilitated by.
Materials
The structure of the surveys was as follows: first the participants answered background questions about their 
demographics and history of psychedelic use and meditation practice. To probe their current psychological state, 
the participants were asked to fill in standardised questionnaires about psychological well-being, peace of mind, 
body appreciation, psychological flexibility and values (these data are reported elsewhere, here the focus is on 
comparing the experience reports). Then they were asked about one experience with psychedelics or meditation 
that they considered to be the most meaningful to them. This focus was deliberate, as it aligns directly with the 
objective to analyse meaningful personal experiences. The section included an open narrative report where 
the participants were instructed to describe their experience in detail. Moreover, the survey included thematic 
open questions about insights32, body-related experiences, and experiences related to sense of connection or 
alienation (reported elsewhere).
Finally, the participants were asked to evaluate their experience with the Mystical Experience Questionnaire 
(MEQ30), a validated survey designed to measure the intensity and qualities of mystical experiences, often 
associated with psychedelic use or deep meditative states34. The MEQ30 questionnaire results are reported in 
order to provide a psychometrically valid reference point for the present NLP analyses. The MEQ30 assesses 
various aspects of mystical-type experiences, including feelings of unity, transcendence of time and space, 
ineffability, and positive mood, providing a way to operationalise deeply subjective spiritual or mystical states34.
Several other studies were conducted on this dataset, and therefore multiple questionnaires mentioned 
here are excluded from the current study. We are focusing only on the participants’ open narrative reports and 
reporting the MEQ scores as a psychometrically valid reference point. The instruction for the narrative report 
was as follows:
Please write down everything you experienced during the session/experience (what happened, where, who 
was present, thoughts, feelings, images, scenarios) as accurately and in as much detail as possible. Remember 
that every detail is important. Do not attempt to make the experience more structured, organized, logical, or 
complete than how you remember it. Do not change the description with omissions, additions, conclusions, or 
embellishments. In case you would like to comment on the experience (e.g., explain how the experience relates 
to your everyday life, etc.), please write those comments in parentheses so that the comments are clearly 
separated from the description of the actual experience you had.
This study was approved by the Research Ethics Committee of Åbo Akademi University (#15092022). Informed 
consent was obtained from all participants prior to their involvement in the study. The eligibility criteria were 
being over 18 years of age and having a previous personally meaningful experience, facilitated by psychedelics 
or meditation. The survey was anonymous, and participation was voluntary. No compensation was given for 
participants. The study adhered to the EU General Data Protection Regulation (GDPR) in its handling and 
processing of research data, and the Declaration of Helsinki regarding research ethics.
Analyses
The open narrative reports were analysed within a Jupyter Notebook using the Python programming language 
along with many open-source packages. All the analyses presented here were performed within the same 
notebook which is available online (see Data and code availability). The study was preregistered at the Open 
Science Framework (https://osf.io/5pa8f) where it was mentioned that NLP methods will be used to analyse the 
data, however, the details of the analyses were not decided beforehand.
This study involved text similarity analyses, topic modeling, and sentiment analysis to analyse the similarity of 
reported personally meaningful psychedelic and meditation experiences. These narrative reports were analysed 
separately for the psychedelic and meditation groups. In the NLP context, individual open-ended responses are 
typically referred to as "documents," while a collection of such responses is termed a "corpus." This paper adheres 
to this standard NLP terminology.
Mystical experience questionnaire
MEQ30 total scores between the groups were compared to provide a psychometrically valid reference point to 
the present NLP analyses. Since the scores were not normally distributed, a Mann–Whitney U-test was used for 
comparison.
Scientific Reports |        (2025) 15:43903 3| https://doi.org/10.1038/s41598-025-27724-0
www.nature.com/scientificreports/
Preprocessing
Before the actual analyses, the narrative reports were first spell checked automatically using Google Sheets’ 
spell checker functionality and misspelled words were corrected manually by the authors. The reports were also 
pre-processed using Python’s NLTK library35, which contains widely used tools for natural language processing. 
Pre-processing included stop word removal and lemmatization of the words inside the documents. Stop 
words are words which are considered not to contain any value for the actual analysis (e.g. “the”, “of ”, “from”). 
Lemmatization is a process where each word is reduced to its root form. For example, “singing” becomes “sing” 
in the process. This helps to consolidate different forms of the same word, which should improve the accuracy 
of the analysis.
Text similarity analyses
Several similarity analyses were conducted to recognise the most common words and themes from the narrative 
reports. First, we simply checked the relative word frequencies in both corpora to see which are the most 
common words that the participants in the psychedelic and meditation groups have used. This provides a baseline 
understanding of the most common concepts in each group’s experiences, allowing for direct comparison 
between the groups. We assessed the lexical similarity of the two corpora by calculating the lexical overlap of the 
unique words and by comparing the relative frequencies of the individual words using the log-likelihood test.
To compare the semantic similarity of the reports, we used semantic textual similarity analysis, which 
quantifies how similar two pieces of text are in meaning rather than just word overlap. We employed one of the 
leading embedding models, OpenAI’s text-ada-00236, for transforming the documents into vector embeddings in 
a high-dimensional space. These embeddings represent the semantic content of the reports, capturing contextual 
and relational meanings beyond simple word matching. The embeddings were aggregated for calculating the 
cosine similarity index37 across and within the groups.
Term frequency-inverse document frequency (TF-IDF) analysis was performed in order to find the 
words which distinguish the two corpora. TF-IDF is a numerical statistic that weighs a word’s relevance in a 
document against its commonness across all documents, highlighting words that are uniquely significant to each 
document38. It consists of two components: term frequency (TF) and inverse document frequency (IDF). TF 
measures how often a word appears in a document while IDF shows how unique a word is across all documents 
in a corpus. IDF is calculated as the logarithm of the ratio of the total number of documents in the corpus to 
the number of documents containing the word. TF-IDF score simply multiplies these two components and the 
formula for word w would look like this:
 
tf − idf(w) = incidence ofw in a document
total words in a document
× log 10
(
total number of documents
number of documentswithw
)
In order to find distinguishing words from both corpora, the psychedelic and meditation narrative reports 
were combined into one corpus with two documents. The first document contained all the psychedelic reports 
concatenated into one long string, and the second document contained all the meditation reports. In this way, 
if a word was included in both psychedelic and meditation reports, the IDF component became zero, and thus 
the whole TF-IDF score was zero for this particular word. Therefore, TF-IDF analysis ends up emphasising the 
words which appear many times in either psychedelic or meditation reports, but not in both.
Topic modelling
Latent Dirichlet Allocation (LDA) is an unsupervised learning algorithm that is used to identify topics present 
in a set of documents. In the context of text modeling, the idea is to summarise longer texts by representing 
them as a set of underlying latent topics. LDA is based on probabilistic graphical models and assumes that 
documents consist of mixtures of topics and that each topic consists of a mixture of words. This algorithm helps 
in discovering what topics are present in a corpus and the degree to which each document exhibits these topics39.
The selection of the parameter k (number of topics) is a challenging and somewhat arbitrary task with 
multiple approaches that often provide conflicting results40. Therefore, some kind of trial-and-error process 
is usually needed for finding the best k for any practical use case. We determined appropriate k-value through 
experimentation and following Wheeler and colleagues’41 simulation study guidelines on statistical performance 
of LDA based on sample size and response length. Their approach suggested five to six topics for psychedelic 
documents and two for meditation documents. We chose five topics for both groups to ensure analytical 
consistency and facilitate direct comparisons between psychedelic and meditation experiences.
Sentiment analysis
The documents from the two different groups were also analysed in terms of their expressed sentiments. This 
was done in Python with the popular Valence Aware Dictionary and Sentiment Reasoner (VADER) model for 
sentiment analysis42. VADER is a lexicon and rule-based approach, which allows for accurate sentiment scoring 
without extensive training datasets. We chose VADER for sentiment analysis because it has been shown to 
outperform other lexicon-based approaches by producing consistent results across different datasets43. VADER 
recognises three classes of sentiments—positive, neutral, and negative—and provides scores for them ranging 
from 0 to 1. These three sentiments are also combined into a compound score which ranges from -1 (extremely 
negative) to 1 (extremely positive). The sentiment scores were calculated for each document in the corpus, and 
the aggregate statistics were compared between the psychedelic and meditation groups. The sentiment scores 
were not normally distributed, so we used the Mann–Whitney U test to compare the distribution of sentiments 
across the groups. Effect sizes were calculated using the rank-biserial correlation. False discovery rate (FDR) was 
controlled due to multiple comparisons, following the procedure recommended by Benjamini and Hochberg44.
Scientific Reports |        (2025) 15:43903 4| https://doi.org/10.1038/s41598-025-27724-0
www.nature.com/scientificreports/
We do not share the original narrative reports to protect participant anonymity. Moreover, we do not present 
example snippets from the original reports because, unlike qualitative content analysis which examines selected 
text segments, NLP methods analyse patterns across entire corpora. Presenting isolated quotes would therefore 
be arbitrary and could give readers a misleading impression of the reports, as such excerpts would not represent 
the overall patterns that NLP methods detect. However, example snippets from the insight reports can be found 
in Jylkkä et al.32.
Results
213 participants from various countries answered the structured online survey about psychedelic or meditation 
experiences. 10 participants answered the survey, but didn’t write the open reports, so they were excluded from 
the analyses. Also, six responses were excluded, because three participants responded multiple times to the 
survey. Therefore, 197 respondents’ answers were analysed in the study—134 in the psychedelics group and 63 
in the meditation group. Participants’ characteristics and their prior experience of psychedelics or meditation 
are presented in Tables 1, 2, 3 and 4.
Participants in the meditation group were typically older (M = 51.4  years, SD = 14) compared to the 
psychedelic group (M = 39.27  years, SD = 13.04). In terms of gender distribution, 63.43% of the psychedelic 
group participants were male, while in the meditation group the gender distribution was more balanced (50.79% 
male). The experiences occurred in different settings, with psychedelic experiences often taking place in familiar 
settings and with other people present, while meditation experiences predominantly occurred in ceremonial 
spaces/retreats (50.79%) and alone (49.21%).
Psychedelic experiences were most commonly facilitated by classical psychedelics (psilocybin, LSD, 
ayahuasca, DMT, 5-MEO-DMT and mescaline), with various other substances used less frequently. However, 
also MDMA and cannabis were quite common in the psychedelic group. For meditation, silent sitting or lying 
down was the most common technique for inducing the meaningful experience.
The MEQ30-total scores were analysed as a psychometrically valid34 reference point to the NLP analyses. 
The score was found to be relatively high on average (maximum score is 150): M = 97.39, SD = 32.94 for the 
psychedelic group and M = 97.18, SD = 34.2 for the meditation group. The results indicated no significant 
difference between the groups regarding their questionnaire responses (U = 3900.5, p = 1.0).
Document length was measured for both groups, and these results are presented in the histograms below 
(Fig.  1). The average lengths of answers in words were 400.69 (SD = 444.78) and 270.49 (SD = 271.78) for 
psychedelics and meditation groups respectively. As we can see from the histogram, the word distribution was 
heavily skewed to the right—most of the participants wrote relatively short reports, but some wrote also multiple 
times more than the average.
Text similarity analyses
The raw frequency analyses showed which words were most common in the psychedelic and meditation 
documents. Top 20 most common words are shown for both groups in Figs.  2 and 3 below. The numbers 
presented are relative to the number of total words in both datasets in order to make the figures easier to compare 
for the reader.
Psychedelics Meditation
M SD Max Min M SD Max Min
Demographics
 Age 39.27 13.04 79 18 51.4 14 80 27
 Income level [1–5]a 3.26 1.07 5 1 2.95 1.25 5 1
 Education level [1–7]b 5.98 1.68 7 1 6.46 1.24 7 2
Previous experience
 Frequency [0–6]c 4.31 1.45 6 0 4.87 0.89 6 2
 Meditation history [0–6]d 3.58 2.51 6 0 5.7 0.98 6 1
Table 1. Psychedelic and meditation group characteristics. aIncome level was measured on an ordinal scale 
from 1 to 5, estimating difference to the average in one’s home country (1 = Much below average, 2 = Below 
average, 3 = Average, 4 = Above average, 5 = Much above average). bEducation level was measured on an ordinal 
scale from 1 to 7 (1 = Primary education; 2 = Lower Secondary education, 3 = Higher Secondary education, 
4 = Vocational education, 5 = University: Bachelor’s degree, 6 = University: Master’s degree, 7 = University: 
Doctoral degree). cFrequency of previous psychedelic experiences was measured on an ordinal scale from 0 to 
6 (0 = Never, 1 = Once, 2 = Twice, 3 = 3–5 times, 4 = 6–10 times, 5 = 10–50 times, 6 = Over 50 times). Frequency 
of meditation practice was measured on an ordinal scale from 0 to 6 with the question ”Do you practice 
meditation?” (0 = No, 1 = I have tried once or twice but do not practice regularly, 2 = Few times per year, 3 = Few 
times per month, 4 = Every week, 5 = Daily, 6 = Several times a day). dHistory of meditation practice was 
measured on an ordinal scale from 0 to 6 with the question “For how long have you practiced meditation?” 
(0 = I do not meditate, 1 = Less than a month, 2 = 1–6 months, 3 = 7–12 months, 4 = Between 1 and 2 years, 
5 = Between 2–5 years, 6 = Over 5 years). This question was also asked from the psychedelics group.
 
Scientific Reports |        (2025) 15:43903 5| https://doi.org/10.1038/s41598-025-27724-0
www.nature.com/scientificreports/
These figures indicate that there is overlap in the most common words. Expanding the analysis to all words 
used, 27.78% (1405 out of 5058 unique words) were shared between the two corpora. When analysing each 
unique word separately with the log-likelihood test, only 440 words  (8.70%) showed statistically significant 
(p < .05) differences in usage between the two groups. This low percentage of significant differences, even without 
correcting for multiple comparisons, suggests a high degree of similarity in word usage between the psychedelic 
and meditation groups.
The analysis of the document vector embeddings revealed a high degree of similarity between the psychedelic 
and meditation groups. The cosine similarity between the average embeddings for each group was 0.972 
(theoretical range from -1 to 1), indicating that the average embeddings of the two groups are very similar in 
the high-dimensional space. Cosine similarity was also used to calculate within-group similarities to assess the 
consistency of the meaning of documents within each group. The average cosine similarity was 0.839 within the 
psychedelics group and 0.838 within the meditation group.
To analyze which words are different between the groups, we used TF-IDF analysis. The words with the 
greatest TF-IDF scores are presented in the Figs. 4 and 5 for psychedelic and meditation groups respectively.
Experience facilitated by* Count Percentage
Psilocybin 64 47.76
LSD 47 35.07
Cannabis 30 22.39
MDMA 24 17.91
Ayahuasca 12 8.96
5-MEO-DMT 10 7.46
DMT 10 7.46
Ketamine 6 4.48
Mescaline 4 2.99
Other 4 2.99
Salvia 1 0.75
Table 3. Psychedelic group: substances used to facilitate the Experience. *Please note that some participants 
used multiple substances to facilitate the Experience, and the percentages do not therefore add up to 100%.
 
Psychedelics Meditation
N % N %
Gender
 Male 85 63.43 32 50.79
 Female 48 35.82 28 44.44
 Other 1 0.75 3 4.76
Region
 Nordic countries 63 47 20 31.70
 Other Europe 39 29.10 17 27
 North America 21 15.70 17 27
 Other (Asia, Africa, South America & Oceania) 9 6.70 9 14.30
Setting of the Experience
 Home 54 40.30 22 34.92
 Ceremonial space/retreat 22 16.42 32 50.79
 Friend’s home 21 15.67 0 0
 Nature 18 13.43 3 4.76
 Public space 9 6.72 3 4.76
 Other private space 7 5.22 2 3.17
 Therapeutic space 3 2.24 1 1.59
Who was present
 One other person 47 35.07 1 1.59
 Several people, all familiar 35 26.12 6 9.52
 Alone 28 20.90 31 49.21
 Several people, some unfamiliar 24 17.91 25 39.68
Table 2. Psychedelics and meditation groups’ gender and region distributions, and settings of the Experience.
 
Scientific Reports |        (2025) 15:43903 6| https://doi.org/10.1038/s41598-025-27724-0
www.nature.com/scientificreports/
The TF-IDF analysis highlights unique words in each group. In the psychedelic group, terms related to 
substance use and effects were emphasised, such as “trip”, “lsd”, “mushroom”, “hallucination” and “visuals”. Words 
indicating dosage (“dose”, “gram”) were present as well. For the meditation group, words specific to meditation 
practices and traditions were highlighted, including “nyams”, “dzogchen” and “ngondro”. Words related to 
spiritual concepts (“divinity”, “prayer”) and practice settings (“garden”, “roof ”) were also notable.
Topic modeling (LDA)
As described in the Methods section, the number of topics parameter was set to k = 5 for both groups. The 
resulting five topics for both groups along with the 10 most important words within each topic are presented 
below in Figs.  6 and 7. The most important words per topic were chosen according to descending β, the 
probability of a word contributing to the given topic. Topics are sorted by their average contributions (γ). γ is 
a metric which describes how much each topic contributes to each document in the corpus, and by averaging 
these values, we can see how much these topics contribute on the level of the whole corpus.
The psychedelic group’s topics focused on personal experiences and interactions, with Topic 4 contributing 
most significantly. The meditation group’s topics, particularly Topic 5, centred on practice-related terminology 
and personal experiences.
Fig. 1. Histogram of the document length in words for psychedelic and meditation groups.
 
Experience facilitated by* Count Percentage
Silent sitting or lying down 34 53.97
Other technique 17 26.98
Mindfulness 15 23.81
Breathwork 5 7.94
Voicework 4 6.35
Bodywork 3 4.76
Table 4. Meditation group: techniques used to facilitate the Experience. *Please note that some participants 
used multiple meditation techniques to facilitate the Experience, and the percentages do not therefore add up 
to 100%.
 
Scientific Reports |        (2025) 15:43903 7| https://doi.org/10.1038/s41598-025-27724-0
www.nature.com/scientificreports/
Sentiment analysis
To analyse the sentiment of each document in both groups, the VADER package in Python was used. VADER 
provides four outputs: positive, neutral, negative, and compound scores. The positive, negative, and neutral 
scores represent the proportion of the text that falls into those categories. By contrast, the compound score is 
a normalized and weighted composite score that calculates the sum of all the lexicon ratings, which have been 
normalized between -1 (extreme negative) and + 1 (extreme positive). This score can be interpreted as a measure 
of the overall sentiment of the text. Box plots of compound sentiment scores for both groups are presented in 
Fig. 8 and the complete results from the sentiment analysis are shown in Table 5 below.
The psychedelic group showed slightly higher presence of negative sentiments (M = 0.07, SD = 0.05) than 
the meditation group (M = 0.06, SD = 0.07), and the difference remained significant after correcting for multiple 
comparisons (U = 5265, p < .01, pFDR < .05, r = .20). The positive sentiment scores were also slightly higher for 
the psychedelics group (M = 0.15, SD = 0.08) compared to the meditation group (M = 0.13, SD = 0.08), and also 
this result remained significant after correction (U = 5097.5, p < .05, pFDR < .05, r = .17). On the other hand, quite 
logically, the meditation group showed a higher tendency toward neutral sentiments (M = 0.81, SD = 0.09) 
Fig. 3. Most common words in the meditation group and their relative occurrences.
 
Fig. 2. Most common words in the psychedelic group and their relative occurrences.
 
Scientific Reports |        (2025) 15:43903 8| https://doi.org/10.1038/s41598-025-27724-0
www.nature.com/scientificreports/
compared to the psychedelics group (M = 0.78, SD = 0.09); difference remaining significant after correcting 
for multiple comparisons (U = 3270.5, p < .05, pFDR < .05, r = − .18). There was no significant difference in the 
compound sentiment scores between the groups (U = 49,170, p = .062, pFDR = .062, r = .13).
Discussion
Previous research indicates neurocognitive similarities between psychedelic and meditation experiences5, and 
both practices have been argued to be capable of facilitating personally meaningful or even transformative 
experiences1. However, research is limited comparing the phenomenology of these two types of experiences. In 
this study, we analysed and compared narrative reports of personally meaningful psychedelic and meditation 
experiences using NLP methods, in order to examine their similarities and differences.
The results from the text similarity analyses revealed commonalities between psychedelic and meditation 
experience reports. Most notably, the vector embeddings showed very high semantic similarity (cosine 
similarity = 0.972) between the average representations of each group. In practical terms, this means that when 
considering the overall semantic content and themes across all documents, the “average” experience reports 
across the groups were very similar in meaning and the concepts they discussed. The within-group document-
level similarity was also on a high, albeit lower level: 0.839 and 0.837 for psychedelic and meditation reports 
respectively. This lower within-group similarity likely reflects greater individual variation in how experiences 
are described within each group, while the higher between-group similarity captures shared broad themes that 
emerge when averaging across reports, as the aggregation process tends to smooth out individual differences and 
highlight common semantic elements. The semantic textual similarity analyses were corroborated by the lexical 
analysis, where only 8.87% of words showed significant differences in usage between groups. Also, the fact that 
the TF-IDF analysis—designed to highlight differences between corpora—only revealed superficial vocabulary 
distinctions related to induction methods, not experiential differences, further supports the similarity of the 
actual experiences.
These findings provide empirical support for Chirico’s and colleagues’1 theoretical framework of 
transformative experiences. While they identify different methods of inducing transformative experiences—
including but not limited to meditation and psychedelics—they suggest these experiences share fundamental 
phenomenological features such as epistemic expansion and emotional complexity. Our results align with this 
theoretical framework, demonstrating that despite different induction methods, the subjective experiences show 
remarkable semantic and lexical overlap. The higher semantic variation found in individual reports compared 
Fig. 4. Psychedelic group top 20 words with the greatest TF-IDF scores.
 
Scientific Reports |        (2025) 15:43903 9| https://doi.org/10.1038/s41598-025-27724-0
www.nature.com/scientificreports/
Fig. 6. Psychedelic group topics ordered by average topic contribution along with their most important words.
 
Fig. 5. Meditation group top 20 words with the greatest TF-IDF scores.
 
Scientific Reports |        (2025) 15:43903 10| https://doi.org/10.1038/s41598-025-27724-0
www.nature.com/scientificreports/
to group averages also supports the observation of Chirico et al.1 that while transformative experiences have 
universal features, they retain distinct individual characteristics.
While Latent Dirichlet Allocation (LDA) has been shown to perform better than other topic models in human 
evaluation45 and theoretically offers advantages in summarizing large text datasets, its application to psychedelic 
and meditation experience reports proved challenging. Our analysis revealed that the highest-contributing 
topics across both psychedelic and meditation reports contained largely the same important words in different 
Fig. 8. Box plots of compound sentiment scores for both groups.
 
Fig. 7. Meditation group topics ordered by descending average topic contribution along with their most 
important words.
 
Scientific Reports |        (2025) 15:43903 11| https://doi.org/10.1038/s41598-025-27724-0
www.nature.com/scientificreports/
orders, primarily emphasizing emotional, temporal and social themes that were already captured by word 
frequency analyses. This redundancy with simpler analyses, combined with the lack of distinct interpretable 
themes, raises questions about LDA’s added value in this context. This limitation can be also observed from study 
by Qiu and Minda46, which compared experience reports about different psychedelic substances. This suggests 
that LDA may be less effective for analysing these types of open-ended experiential narratives compared to more 
structured texts, such as scientific publications47.
According to the present sentiment analysis, both psychedelic and meditation experiences were highly 
positive on average, though psychedelic experiences showed greater emotional variance. The key difference 
was that psychedelic reports contained both higher positive and negative sentiment scores, whereas meditation 
reports were more emotionally neutral. Though effect sizes were small, the results are consistent with previous 
research showing increased emotional lability with psychedelics, particularly LSD48, while meditation research 
typically reports increased positive affect without corresponding emotional variability49,50. This emotional 
intensity difference may be explained by increased cognitive entropy caused by the psychedelic substances48 and 
could be important given that emotional intensity during psychedelic experiences predicts positive wellbeing 
outcomes51. Even though these findings are based on indirect sentiment analysis with a modest sample size, 
they suggest that emotional lability might be at least one key difference between the psychedelic and meditation 
experiences.
The current results—high similarity in the general content of the narrative reports, but differences in the 
emotionality—are quite aligned with two predictive coding theories about psychedelics52 and meditation24. 
Both theories focus on the relaxation of high-level priors and top-down predictions which in turn leads to 
increased bottom-up signalling in the brain during the psychedelic and meditation experiences. If these models 
are accurate, it could show as similar narrative reports between the two groups, which was the result here. 
Also, the differences in the sentiment analysis can be explained by the differences in these models. The REBUS 
model by Carhart-Harris and Friston52 suggests that the relaxation of top-down constraints in brain processing 
allows for more varied and intense emotional experiences. On the other hand, Laukkonen and Slagter24 suggest 
that meditation gradually brings practitioners closer to immediate sensory experience and away from abstract 
thinking. This gradual process might result in a more balanced emotional state, consistent with the less extreme 
sentiment scores observed in the meditation group.
Chirico et al.1 identify mystical experience as one of the subtypes of transformative experience. According to 
them, these experiences can be induced both with psychedelics and meditation. This is interesting in relation to 
our results, since both groups received high and almost identical MEQ30-total scores on average: M = 97.39 in 
the psychedelic group and M = 97.18 in the meditation group. The similarity in these scores between the groups 
aligns with the text similarity analyses. However, the MEQ30 questionnaire and the whole mystical experience 
construct have been criticised for inviting study participants to interpret their experience through this mystical 
lens53. The MEQ30 contains some metaphysically loaded wordings, which might bias the respondents to use 
similar terms while describing their experience. Our study tried to avoid this by administering MEQ30 as 
the very last questionnaire of the whole survey, so that the participants would not use the terminology of the 
questionnaire in the open reports. Nevertheless, both groups still received high and similar scores from the 
questionnaire.
The convergence we observe—both in MEQ30 scores and semantic similarity of narrative reports—
demonstrates that within our sample of predominantly Western participants, psychedelic and meditation 
experiences were described in remarkably similar ways. However, this cannot be directly interpreted as evidence 
for universal, context-independent mystical states. The high semantic similarity we found could indicate 
genuinely overlapping phenomenology between psychedelic and meditation experiences, but it could also partly 
be explained by the fact that participants were describing their most meaningful experience with psychedelics 
or meditation. This may have led them to write about similar themes of transformation regardless of the specific 
qualities of the experiences themselves. In addition, since the most participants are from Western countries and 
all of them have answered the survey in English, they might be inclined to use similar cultural terminologies to 
explain their experiences.
Millière and colleagues7 have theorized that both psychedelics and meditation can disrupt the brain processes 
related to self-consciousness, potentially leading to a variety of experiences that are often described with the 
term self-loss. The changes in self-consciousness in psychedelic and meditative states are also emphasized in 
the predictive coding models about these states24,52. Therefore, it’s interesting to notice that in the present NLP 
analyses, these changes in self-consciousness are not emphasized at all in the present results. However, this 
might be due to the limitation of the NLP methods. Experiences of self-loss could be described using diverse 
vocabulary and metaphors that may not be captured by the present NLP analyses.
The present results can be compared to a previous study on the same dataset that utilised qualitative content 
analysis to identify insights in the psychedelic and meditation reports. Jylkkä et al.32 identified three main insight 
Sentiment type Psychedelics mean (SD) Meditation mean (SD) Mean difference U-statistic Initial p value Adjusted p value Effect size (r)
Compound 0.63 (0.61) 0.56 (0.59) 0.07 4917 .062 .062 .13
Positive 0.15 (0.08) 0.13 (0.08) 0.02 5097.5 .019* .025* .17
Negative 0.07 (0.05) 0.06 (0.07) 0.01 5265 .005* .021* .20
Neutral 0.78 (0.09) 0.81 (0.09) − 0.03 3270.5 .011* .022* -.18
Table 5. Sentiment scores between groups.
 
Scientific Reports |        (2025) 15:43903 12| https://doi.org/10.1038/s41598-025-27724-0
www.nature.com/scientificreports/
themes across the reports: mystical-type, psychological, and compassion-related. The insight themes occurred 
equally frequently in the two types of reports, with only two exceptions: mystical-type insights were more 
frequent in the meditation reports, and value insights were more common in the psychedelic group. Moreover, 
no differences in the insights were found depending on which type of drug was used to facilitate the experience 
(classic psychedelic, non-classic psychedelic, or combination of the two). The authors propose that the higher 
frequency of value insights in the psychedelic reports could be due to higher emotional salience of psychedelic 
experiences, given the close link between values and affect54. The results of our sentiment analysis support 
this hypothesis. Moreover, Jylkkä et al.32 speculated that the higher occurrence of mystical-type insight in the 
meditation reports could be an artefact, given that the meditators often framed their reports using meditation-
specific technical terms, which often have metaphysical meanings. This hypothesis is supported by the present 
TF-IDF analysis where the main differentiating factors were mainly trivial psychedelic- or meditation related 
words. Finally, regarding the lack of phenomena related to self-loss in the current analyses, it is noteworthy 
that these aspects were discovered in the qualitative analysis by Jylkkä et al.32, suggesting that failing to find 
this phenomenon in the current NLP analysis reflects a limitation of this method. However, in sum, the results 
from the previous qualitative analysis are in line with the present NLP results, supporting the hypothesis that 
personally meaningful psychedelic and meditation experiences are quite similar.
Limitations
The nature of the collected dataset and the study design pose some challenges for drawing conclusions from the 
analyses. The psychedelic and meditation groups were quite different in terms of their size and demographic 
factors (see Tables 1, 2) and the small sample size precluded statistical control for these potential confounders 
in the NLP analyses. Moreover, the reports were collected retrospectively with independent samples design. 
This means that many other potential confounding factors could not be controlled for and recall bias might be 
common among the participants. They might not accurately remember their experiences at the time of reporting 
and could interpret the experiences in the light of their current knowledge and beliefs, further distorting their 
reports. Furthermore, the sample size was relatively small for NLP analyses, especially in the meditation group, 
which could amplify the impact of outliers and reduce the statistical robustness of the findings. Finally, the 
number of self-report questionnaires about the content of the experience was limited, since only MEQ30 about 
mystical experiences34 was administered. For instance, including the Awe Experience Scale55 or the PIQ scale 
on insights56 could have offered a more nuanced and multidimensional assessment of participants’ experiences 
to support the NLP analyses.
In addition to the matching problems, lack of standardisation within the groups poses challenges for 
interpreting the data. Psychedelic group included experiences induced with a wide variety of substances and 
the meditators had used several different techniques. Even though some researchers have argued that personally 
meaningful experiences can be qualitatively similar across multiple substances with different mechanisms of 
action1,9,13,14, it would be optimal to have data where the groups were more homogeneous. For example, one 
potentially useful approach could be to compare experiences facilitated by classical psychedelics with sitting 
meditation. In the present study this could not be done since the samples would have become too small for 
robust analyses. Therefore, the present study does not allow inferences to be made about specific substances or 
meditation techniques used by the participants.
Despite these challenges, the systematic survey used in this study provides a significant improvement over 
less structured data sources like those freely accessible on the Internet (e.g., Erowid; https://www.erowid.org/) 
which have been previously used for gathering information about psychedelic experiences46,57. Gathering data 
directly from internet forums has many disadvantages, such as lack of demographic data and lack of instructions 
for writing the report. Compared to analysing reports from Internet forums, the current survey ensures that all 
participants are responding to the same prompts, reducing variability in the answers.
The present study could be criticised for neglecting negative or challenging experiences that can be facilitated 
by both psychedelics58 and meditation59. The study explicitly sampled “personally meaningful experiences”, 
which could have led to a positivity bias, given that most such experiences are positive. It is noteworthy that in 
the sentiment analysis some of the reports showed very high negative sentiment scores, which could indicate 
negative experiences. However, the sentiment analysis ignores the context of the sentiment words, thus we 
cannot draw conclusions about the presence of negative experiences.
A final limitation worth noting pertains to the NLP methods used. While vector embeddings can capture 
semantic relationships between words and phrases, the black-box nature of these models makes it challenging 
to fully understand how they arrive at these results37. On the other hand, more understandable traditional 
NLP approaches that rely on exact word matching can be criticized for missing semantic meanings of words. 
However, qualitative analysis using human raters has its own limitations, particularly researcher subjectivity 
in interpretation60. Therefore, combining computational and human-based approaches could provide 
complementary insights, leveraging the systematic nature of NLP while maintaining the nuanced understanding 
that comes from human interpretation.
Conclusions
This study employed NLP methods to compare reports of personally meaningful experiences facilitated by 
psychedelics and meditation. Analyses of word frequencies, distinctive terms (TF-IDF) and semantic similarities 
revealed strong similarity between the two, with substantial overlap in common words and themes. Additionally, 
both groups expressed predominantly positive sentiments, though psychedelic reports were more emotionally 
charged, displaying higher levels of both positive and negative sentiment, while meditation reports tended to 
be more neutral. These findings align with a previous qualitative content analysis of the same dataset32. While 
sentiment analysis and vector embeddings provided useful insights, the NLP approach as a whole did not provide 
Scientific Reports |        (2025) 15:43903 13| https://doi.org/10.1038/s41598-025-27724-0
www.nature.com/scientificreports/
very deep understanding of the reports, since distinct themes could not be recognized from the LDA topic 
models. Despite the limitations, this study serves as foundation for future research, which should focus on larger, 
matched samples with standardized substances and meditation techniques. Future studies can use NLP as a 
complementary tool alongside structured questionnaires and qualitative analysis to enhance our understanding 
of these complex experiences.
Data and code availability
Data is available per request to the authors. The GitHub repository containing the Jupyter Notebook used for 
performing the analyses can be found from:  h t t p s : / / g i t h u b . c o m / k o n s t a k m / p s y c h e d e l i c s _ v s _ m e d i t a t i o n _ n l p.
Received: 6 March 2025; Accepted: 5 November 2025
References
 1. Chirico, A. et al. Defining transformative experiences: A conceptual analysis. Front. Psychol. 13, 790300.  h t t p s : / / d o i . o r g / 1 0 . 3 3 8 9 / f 
p s y g . 2 0 2 2 . 7 9 0 3 0 0 (2022).
 2. Yaden, D. B., Haidt, J., Hood, R. W., Vago, D. R. & Newberg, A. B. The varieties of self-transcendent experience. Rev. Gen. Psychol. 
21, 143–160 (2017).
 3. Galvão-Coelho, N. L. et al. Classic serotonergic psychedelics for mood and depressive symptoms: A meta-analysis of mood 
disorder patients and healthy participants. Psychopharmacology 238, 341–354 (2021).
 4. Goyal, M. et al. Meditation programs for psychological stress and well-being. JAMA Intern. Med. 174, 357.  h t t p s : / / d o i . o r g / 1 0 . 1 0 0 1 
/ j a m a i n t e r n m e d . 2 0 1 3 . 1 3 0 1 8 (2014).
 5. Holas, P. & Kamińska, J. Mindfulness meditation and psychedelics: potential synergies and commonalities. Pharmacol. Rep. 75, 
1398–1409. https://doi.org/10.1007/s43440-023-00551-8 (2023).
 6. Letheby, C. Psychedelics and meditation: A neurophilosophical perspective. In Routledge Handbook on the Philosophy of Meditation 
(ed. Repetti, R.) 209–224 (Routledge, 2022).
 7. Millière, R., Carhart-Harris, R. L., Roseman, L., Trautwein, F. M. & Berkovich-Ohana, A. Psychedelics, meditation, and self-
consciousness. Front. Psychol. 9, 1475. https://doi.org/10.3389/fpsyg.2018.01475 (2018).
 8. Carhart-Harris, R. L. & Goodwin, G. M. The therapeutic potential of psychedelic drugs: Past, present, and future. 
Neuropsychopharmacology 42, 2105–2113 (2017).
 9. Vollenweider, F. X. & Kometer, M. The neurobiology of psychedelic drugs: Implications for the treatment of mood disorders. Nat. 
Rev. Neurosci. 11, 642–651 (2010).
 10. Nichols, D. E. Psychedelics. Pharmacol. Rev. 68, 264–355 (2016).
 11. Nichols, D. E. Differences between the mechanism of action of MDMA, MBDB, and the classic hallucinogens. Identification of a 
new therapeutic class: Entactogens. J. Psychoact. Drugs 18, 305–313 (1986).
 12. Earleywine, M., Ueno, L. F., Mian, M. N. & Altman, B. R. Cannabis-induced oceanic boundlessness. J. Psychopharmacol. (Oxford, 
England) 35, 841–847 (2021).
 13. Lyon, A. Psychedelic Experience: Revealing the Mind (Oxford University Press, 2024).
 14. Breeksema, J. J., Niemeijer, A. R., Krediet, E., Vermetten, E. & Schoevers, R. A. Psychedelic treatments for psychiatric disorders: A 
systematic review and thematic synthesis of patient experiences in qualitative studies. CNS Drugs 34, 925–946 (2020).
 15. Luoma, J. B., Chwyl, C., Bathje, G. J., Davis, A. K. & Lancelotta, R. A meta-analysis of placebo-controlled trials of psychedelic-
assisted therapy. J. Psychoact. Drugs 52, 289–299 (2020).
 16. Romeo, B., Karila, L., Martelli, C. & Benyamina, A. Efficacy of psychedelic treatments on depressive symptoms: A meta-analysis. J. 
Psychopharmacol. 34, 1079–1085 (2020).
 17. Ko, K., Knight, G., Rucker, J. J. & Cleare, A. J. Psychedelics, mystical experience, and therapeutic efficacy: A systematic review. 
Front. Psychiatry 13, 917199. https://doi.org/10.3389/fpsyt.2022.917199 (2022).
 18. McCulloch, D.E.-W. et al. Psilocybin-induced mystical-type experiences are related to persisting positive effects: A quantitative 
and qualitative report. Front. Pharmacol. 13, 841648. https://doi.org/10.3389/fphar.2022.841648 (2022).
 19. Yaden, D. B. & Griffiths, R. R. The subjective effects of psychedelics are necessary for their enduring therapeutic effects. ACS 
Pharmacol. Transl. Sci. 4, 568–572 (2021).
 20. Dahl, C. J., Lutz, A. & Davidson, R. J. Reconstructing and deconstructing the self: cognitive mechanisms in meditation practice. 
Trends Cogn. Sci. 19, 515–523 (2015).
 21. Berkovich-Ohana, A., Dor-Ziderman, Y., Glicksohn, J. & Goldstein, A. Alterations in the sense of time, space, and body in the 
mindfulness-trained brain: A neurophenomenologically-guided MEG study. Front. Psychol. 4, 912.  h t t p s : / / d o i . o r g / 1 0 . 3 3 8 9 / f p s y g . 
2 0 1 3 . 0 0 9 1 2 (2013).
 22. Wahbeh, H., Sagher, A., Back, W., Pundhir, P. & Travis, F. A Systematic review of transcendent states across meditation and 
contemplative traditions. EXPLORE 14, 19–35 (2018).
 23. Zanesco, A. P., King, B. G., Conklin, Q. A. & Saron, C. D. The occurrence of psychologically profound, meaningful, and mystical 
experiences during a month-long meditation retreat. Mindfulness 14, 606–621 (2023).
 24. Laukkonen, R. E. & Slagter, H. A. From many to (n)one: Meditation and the plasticity of the predictive mind. Neurosci. Biobehav. 
Rev. 128, 199–217 (2021).
 25. Hölzel, B. K. et al. How does mindfulness meditation work? Proposing mechanisms of action from a conceptual and neural 
perspective. Perspect. Psychol. Sci. 6, 537–559 (2011).
 26. Friston, K. The free-energy principle: A unified brain theory?. Nat. Rev. Neurosci. 11, 127–138 (2010).
 27. Tang, Y. Y., Hölzel, B. K. & Posner, M. I. The neuroscience of mindfulness meditation. Nat. Rev. Neurosci. 16, 213–225 (2015).
 28. Moujaes, F. et al. Comparing neural correlates of consciousness: From psychedelics to hypnosis and meditation. Biol. Psychiatry 
Cogn. Neurosci. Neuroimaging 9, 533–543 (2023).
 29. Timmermann, C. et al. A neurophenomenological approach to non-ordinary states of consciousness: Hypnosis, meditation, and 
psychedelics. Trends Cogn. Sci. 27, 139–159 (2023).
 30. Heuschkel, K. & Kuypers, K. P. C. Depression, mindfulness, and psilocybin: Possible complementary effects of mindfulness 
meditation and psilocybin in the treatment of depression. A review. Front. Psychiatry 11, 224.  h t t p s : / / d o i . o r g / 1 0 . 3 3 8 9 / f p s y t . 2 0 2 0 . 0 
0 2 2 4 (2020).
 31. Azmoodeh, K., Thomas, E. & Kamboj, S. K. Meditation trips: A thematic analysis of the combined naturalistic use of psychedelics 
with meditation practices. Exp. Clin. Psychopharmacol. 31, 756–767 (2023).
 32. Jylkkä, J. et al. Meditation and psychedelics facilitate similar types of mystical, psychological, and philosophical-existential insights 
predictive of wellbeing: A qualitative-quantitative approach. Conscious. Cogn. 133, 103901 (2025).
 33. Wang, Y., Tian, J., Yazar, Y., Ones, D. S. & Landers, R. N. Using natural language processing and machine learning to replace human 
content coders. Psychol. Methods https://doi.org/10.1037/met0000518 (2022).
Scientific Reports |        (2025) 15:43903 14| https://doi.org/10.1038/s41598-025-27724-0
www.nature.com/scientificreports/
 34. Barrett, F. S., Johnson, M. W. & Griffiths, R. R. Validation of the revised Mystical Experience Questionnaire in experimental 
sessions with psilocybin. J. Psychopharmacol. 29, 1182–1190 (2015).
 35. Loper, E., & Bird, S. NLTK: The natural language toolkit. Preprint at https://arxiv.org/abs/cs/0205028 (2002).
 36. Greene, R., Sanders, T., Weng, L., & Neelakantan, A. New and improved embedding model. OpenAI Blog.  h t t p s : / / o p e n a i . c o m / b l o 
g / n e w - a n d - i m p r o v e d - e m b e d d i n g - m o d e l (2022).
 37. McGillivray, B., How to Use Word Embeddings for Natural Language Processing [How-to Guide]. Sage Research Methods: Doing 
Research Online. https://doi.org/10.4135/9781529609578 (2022).
 38. Rajaraman, A. & Ullman, J. D. Data mining. In Mining of Massive Datasets (eds Rajaraman, A. & Ullman, J. D.) 1–17 (Cambridge 
University Press, Cambridge, 2011).
 39. Blei, D., Ng, A. & Jordan, M. I. Latent Dirichlet allocation. J. Mach. Learn. Res. 3, 993–1022 (2003).
 40. Rüdiger, M., Antons, D., Joshi, A. M. & Salge, T.-O. Topic modeling revisited: New evidence on algorithm performance and quality 
metrics. PLoS ONE 17, e0266325. https://doi.org/10.1371/journal.pone.0266325 (2022).
 41. Wheeler, J. M., Cohen, A. S., Xiong, J., Lee, J., & Choi, H. J. Sample size for latent Dirichlet allocation of constructed-response 
items. In Quantitative Psychology: The 85th Annual Meeting of the Psychometric Society. IMPS 2020 (eds. Wiberg, M., Molenaar, 
D., Gonzalez, J., Böckenholt, U. & Kim, J.S.) Springer Proceedings in Mathematics & Statistics, vol. 353, 263–273 (Springer, 2021). 
https://doi.org/10.1007/978-3-030-74772-5_24
 42. Hutto, C., & Gilbert, E. VADER: A parsimonious rule-based model for sentiment analysis of social media text. In Proceedings of the 
International AAAI Conference on Web and Social Media, 8, 216–225 (2014).
 43. Ribeiro, F. N., Araújo, M., Gonçalves, P., André Gonçalves, M. & Benevenuto, F. SentiBench—A benchmark comparison of state-
of-the-practice sentiment analysis methods. EPJ Data Sci. 5, 23 (2016).
 44. Benjamini, Y. & Hochberg, Y. Controlling the false discovery rate: a practical and powerful approach to multiple testing. J. R. Stat. 
Soc. Ser. B (Methodol.) 57, 289–300 (1995).
 45. Chang, J., Gerrish, S., Wang, C., Boyd-Graber, J., & Blei, D. Reading tea leaves: How humans interpret topic models. In Advances 
in Neural Information Processing Systems, 22. (2009).
 46. Qiu, T. T., & Minda, J. P. Recreational psychedelic users frequently encounter complete mystical experiences: Trip content and 
implications for wellbeing. Preprint at https://doi.org/10.31234/osf.io/xrbzs (2021).
 47. Yu, D. & Xiang, B. Discovering topics and trends in the field of Artificial Intelligence: Using LDA topic modeling. Expert Syst. Appl. 
225, 120114 (2023).
 48. Carhart-Harris, R. L. et al. The paradoxical psychological effects of lysergic acid diethylamide (LSD). Psychol. Med. 46, 1379–1390 
(2016).
 49. Harte, J. L., Eifert, G. H. & Smith, R. The effects of running and meditation on beta-endorphin, corticotropin-releasing hormone 
and cortisol in plasma, and on mood. Biol. Psychol. 40, 251–265 (1995).
 50. Johnson, S., Gur, R. M., David, Z. & Currier, E. One-session mindfulness meditation: A randomized controlled study of effects on 
cognition and mood. Mindfulness 6, 88–98 (2015).
 51. Roseman, L. et al. Emotional breakthrough and psychedelics: Validation of the Emotional Breakthrough Inventory. J. 
Psychopharmacol. 33, 1076–1087 (2019).
 52. Carhart-Harris, R. L. & Friston, K. J. REBUS and the anarchic brain: Toward a unified model of the brain action of psychedelics. 
Pharmacol. Rev. 71, 316–344 (2019).
 53. Sanders, J. W. & Zijlmans, J. Moving past mysticism in psychedelic science. ACS Pharmacol. Transl. Sci. 4, 1253–1255 (2021).
 54. Kähönen, J. Psychedelic unselfing: Self-transcendence and change of values in psychedelic experiences. Front. Psychol. 14, 1104627 
(2023).
 55. Yaden, D. B. et al. The development of the Awe Experience Scale (AWE-S): A multifactorial measure for a complex emotion. J. 
Posit. Psychol. 14, 474–488 (2019).
 56. Davis, A. K. et al. Development of the Psychological Insight Questionnaire among a sample of people who have consumed 
psilocybin or LSD. J. Psychopharmacol. (Oxford, England) 35, 437–446 (2021).
 57. Tagliazucchi, E. Natural language as a window into the subjective effects and neurochemistry of psychedelic drugs. Preprint at 
https://doi.org/10.20944/preprints202111.0201.v1 (2021).
 58. Carbonaro, T. M. et al. Survey study of challenging experiences after ingesting psilocybin mushrooms: Acute and enduring positive 
and negative consequences. J. Psychopharmacol. 30, 1268–1278 (2016).
 59. Farias, M., Maraldi, E., Wallenkampf, K. C. & Lucchetti, G. Adverse events in meditation practices and meditation-based therapies: 
a systematic review. Acta Psychiatr. Scand. 142, 374–393 (2020).
 60. Braun, V. & Clarke, V. Using thematic analysis in psychology. Qual. Res. Psychol. 3, 77–101 (2006).
Author contributions
JJ conceptualized the study design and supervised the project. KK conducted the analyses and wrote the first 
draft of the manuscript. RS contributed to the linguistic and natural language processing components of the 
manuscript and took part in revising the manuscript. The final manuscript was written by KK and revised by 
RS and JJ.
Declarations
Competing interests
The authors declare no competing interests.
Additional information
Correspondence and requests for materials should be addressed to K.K.-M.
Reprints and permissions information is available at www.nature.com/reprints.
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and 
institutional affiliations.
Scientific Reports |        (2025) 15:43903 15| https://doi.org/10.1038/s41598-025-27724-0
www.nature.com/scientificreports/
Open Access  This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 
4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in 
any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide 
a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have 
permission under this licence to share adapted material derived from this article or parts of it. The images or 
other third party material in this article are included in the article’s Creative Commons licence, unless indicated 
otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence 
and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to 
obtain permission directly from the copyright holder. To view a copy of this licence, visit  h t t p : / / c r e a t i v e c o m m o 
n s . o r g / l i c e n s e s / b y - n c - n d / 4 . 0 / . 
© The Author(s) 2025 
Scientific Reports |        (2025) 15:43903 16| https://doi.org/10.1038/s41598-025-27724-0
www.nature.com/scientificreports/