## Introduction

The sociologist Emile Durkheim coined the term “collective effervescence” to describe the feelings of self-transcendence that often arise at mass gatherings such as festivals, pilgrimages, and collective rituals1. Testimonial accounts from attendees of mass gatherings suggest that such self-transcendent experiences may be epistemically and personally transformative, resulting in lasting changes to the self2,3. The philosopher L.A. Paul articulates two key components of transformative experiences: they provide new knowledge that is impossible to attain without having the experience, and they produce changes in personal values and priorities that cannot be anticipated4. Here, we investigate the psychological qualities of transformative experiences at secular mass gatherings and test the possibility that such experiences are associated with enduring changes in moral orientation.

Support for our hypothesis comes from research showing that collective gatherings such as rituals5, ceremonies6, raves7, and sporting events8,9 are associated with feelings of self-expansion and “group identity fusion”10,11,12,13 a psychological state characterized by intense feelings of merging or oneness between the self and the group14,15. This phenomenon is well-documented across cultures16 and is associated with endorsement of group values17,18, enhanced group loyalty6,19, increased generosity20,21,22,23,24,25,26, and cooperation27,28,29,30. Such prosocial changes are thought to emerge from psychological processes whose original adaptive function lay in promoting the survival of the group at the expense of the self31. For example, anthropological accounts suggest that humans throughout history have often engaged in ritualistic behaviors (e.g., ceremonial dancing) that amplify identity fusion before performing acts of extreme self-sacrifice32.

Past work on mass gatherings and identity fusion has tended to focus on how such experiences can amplify prosocial feelings and behavior directed toward the members of one’s own social group. Here, by contrast, we investigated the potential for transformative experiences at mass gatherings to expand the boundaries of one’s moral circle beyond the group to include all of humanity33,34,35. Recent research suggests that experiences of self-transcendent emotions are associated with increased identification with all humanity—i.e., increased concern for all other human beings—and motivations to help distant others36,37. Other work has shown that such experiences are associated with heightened feelings of social connectedness38. Building on this and other recent theoretical work4, we conceptualized transformative experiences at mass gatherings as precipitating events that may lead to changes in values and behavior characterized by an increased sense of connection to other human beings and an expanded moral circle. Thus, we tested whether self-reported transformative experiences at secular mass gatherings were accompanied by increased feelings of universal connectedness, and investigated whether such experiences would be associated with enduring changes in generosity and moral expansion.

For practical and ethical reasons, it is difficult to generate in the lab the kinds of intensely transformative experiences that people report having at mass gatherings. Previous research on such experiences has relied on retrospective surveys that prompt participants to report on experiences they had in the past7,25, yet such approaches are subject to the vagaries of personal recollection. We sought to build on this work by studying the psychological qualities of transformative experiences as they occurred. To do this, we adopted a lab-in-the-field approach in which we collected data from participants as they attended one of six secular multi-day mass gatherings across five field sites in the US and the UK (see Fig. 1 and Table 1). We define secular mass gatherings as events with a total attendance greater than five hundred39 and no explicitly religious component. We focused on secular events in order to ensure that any effects observed were not the result of explicit reference to the divine, which past research has shown to foster self-transcendent experiences40 but is not the focus of this investigation. We also focused our research efforts on non-political events in order to avoid the potential confounding effects of specific ideological messages on prosocial attitudes and behavior.

Mass gatherings are highly immersive social experiences that may strip away belief systems and aspects of the self-concept like layers of an onion. Such experiences may thus not have their impact immediately upon arrival but instead take many days to impart their full effect41. For this reason, a central question in our research was how transformative experiences and their putative prosocial correlates unfold over time both during the events themselves and in the weeks and months after attendance. We thus focused our research on events with a total duration of three or more days and employed a data collection strategy that sampled participants at different timepoints within each event (see Fig. 1 and SOM 1.2). This allowed us to use time-at-event as a predictor in our models while controlling for key characteristics in our population sample. For example, while it is possible that people who attend mass gatherings are on average more likely to report transformative experiences or display prosocial attitudes and behavior, we were able to study the temporal dynamics of transformative experience and its prosocial correlates with our study population while holding these factors constant by having all participants complete our study at a relatively random timepoint during each event (see “Methods” for more information on our sampling strategy). Further, by sampling participants in the weeks and months following mass gathering attendance, we were additionally able to examine whether correlations between self-reported transformative experience and prosocial attitudes and behavior evolve over time.

We designed a variety of experimental procedures and methods appropriate for the mass gathering context, including measures of generosity and moral expansion that did not rely on the use of money, which was prohibited at some of our field sites (see SOM 4 for full descriptions of measures and verbatim participant instructions). We also collected detailed measures of participants’ use of psychoactive substances, as past work has implicated psychedelic substance use in transformative experience and prosocial behavior7,42,43. In related work, we have examined the emotional consequences of psychedelic substance use at mass gatherings44, finding that their use is associated with increased positive mood—an effect mediated by transformative experience and universal connectedness. Here, we sought to examine the nature and prosocial correlates of transformative experiences that do not necessarily arise from psychedelic substance use. To this end, we controlled for substance use in all analyses of onsite data, and compared the psychological qualities of transformative experience that arose in the presence versus absence of psychedelic substances (see SOM 1.2 for further details).

Finally, we investigated whether the “set and setting” of transformative experiences is associated with the nature of prosocial change45,46. While these concepts where originally conceived to understand the psychological consequences of psychedelic substance use47, we surmised they may have similarly meaningful consequences for mass gathering participation alone. Related to “set” (i.e., mindset), it is known that desires and expectations can have powerful effects on subsequent experiences, for example in the placebo effect48 and self-fulfilling prophecies49. Conversely, the absence of a desire or expectation for change may preclude actual change, a phenomenon known in clinical psychology as “client resistance”50. Accordingly, we sought to characterize the relationships between expecting and desiring transformative experiences, on the one hand, and actual self-reported transformative experiences on the other, predicting that anticipation would positively predict transformative experience.

Setting may also be important: transformative experiences may be particularly associated with moral expansion when they occur in the context of explicitly communal event norms51,52. This question remains unexplored because, with some exceptions18,27, past research on transformative experiences and mass gatherings has focused largely on singular events. Here, we took a comparative approach, exploiting natural variation across field sites in the salience of communal norms. Specifically, some of our sites promoted communal norms by operating a gift economy that explicitly prohibited the use of money, while other sites promoted individualistic norms by operating a market economy where money could be exchanged for goods and services53. If setting matters, prosocial behavior may be more strongly associated with transformative experiences that occur within gifts as opposed to market economies.

Overall, we set out to answer several questions about transformative experiences at secular multi-day mass gatherings, building upon and extending past work on mass gatherings, identity fusion, and prosocial behavior. First, we sought to describe the psychological qualities of transformative experiences as they unfolded over time and document how they relate to expectations and desires for transformation. Second, we examined the prosocial correlates of transformative experiences, investigating their potential to enhance generosity and expand the moral circle by engendering feelings of universal connectedness. Finally, we measured the persistence and evolution of transformative experiences and their prosocial correlates in the weeks and months following mass gathering attendance.

## Results

We collected data at six events across five field sites that varied on several key attributes, including the total number of attendees, location, and the presence of a gift versus market economy (Table 1). Over a data collection period of 5 years, we were able to obtain data from 1215 onsite participants. Gift economy field sites included Burning Man (an event in central Nevada; n = 473) and Burning Nest (a “Regional Burn” event in the UK officially associated with the Burning Man organization; n = 100). Market-economy sites included Lightning in a Bottle (a festival in California; n = 391), Latitude (a festival in the UK; n = 168), and Dirty Bird Campout (a festival in California; n = 83). See Table 1 for more information on data collection, SOM 3.1 for detailed descriptions of each event, and SOM 3.2 for detailed participant demographics. We collected self-reported frequency of gift-giving and money handling at each event, and confirmed that people gave more gifts (t(1036) = 12.60, P < 0.001), received more gifts (t(1082) = 16.30, P < 0.001), and handled less money (t(576) = 11.80, P < 0.001), at gift economies than market economies.

We supplemented our onsite data with online surveys administered 0–2 weeks before (“pre-event”, n = 600), 0–4 weeks after (“immediate follow-up”, n = 1866), and 6 months after (“6-month follow-up”, n = 710) attendance, which allowed us to track relationships between transformative experience and moral orientation over time (Fig. 1). Links to an online survey were distributed to e-mail addresses collected from onsite participants (“targeted”) as well as newsletters sent out to all mass gathering participants (“untargeted”). In addition, we were able to collect a small number of additional variables in a supplementary sample collected through the Burning Man Census, an annual post-event online survey distributed to Burning Man attendees (n = 6649). Finally, we collected comparison data (n = 98) from a “virtual” mass gathering that took place among Burning Man attendees in the context of the COVID-19 pandemic (see SOM 1.6 for details).

Two within-subjects, repeated-measures (longitudinal) samples were obtained by combining datasets from multiple timepoints using an anonymous identifier code unique to each participant: one that compared responses from the “pre-event” and “immediate follow-up” samples (“pre-post”; n = 184), and one that compared responses from the “onsite” and “immediate follow-up” samples (“onsite-post”; n = 148). “Cross-sectional” analyses refer to those occurring within a single time period (e.g., onsite, immediate follow-up, etc.); “longitudinal” refer to those that examine relationships between variables collected over time within subjects (Fig. 1). Longitudinal analyses assessed test-retest reliability on all primary measures (transformative experience, universal connectedness, moral expansion, and generosity), all rs ≥ 0.40, all Ps < 0.0001, except that onsite generosity did not significantly correlate with immediate follow-up generosity, r = 0.15, P = 0.23, possibly because of reduced within-subjects sample size for this measure (n = 62; see SOM 1.1 and Supplementary Table S1 for details).

### Transformative experience

First, we assessed changes in self-reported transformative experience over time onsite. Regression analysis showed that, as predicted, rates of transformative experience increased significantly over time, B = 0.23, SE = 0.03, t(1178) = 7.61, P < 0.001; Fig. 2). In order to estimate the average expected rate of transformative experience following mass gathering attendance, we calculated mean self-reported transformation among all onsite participants. Results showed that, overall, 63.2% of participants reported being at least “somewhat” transformed, and 19.5% said they were “absolutely” transformed (M = 4.08, SD = 2.06). Overall, then, these data show the prevalence of self-reported transformative experience at these events and suggest that rates of transformative experience increase over time.

Next, we examined demographic, affective, and behavioral predictors of transformative experience. In our main models, we did not observe associations between transformative experience and gender, age, or income (all Ps > 0.3). Transformative experience was negatively associated with educational attainment, B = −0.12, SE = 0.05, t(1177) = −2.52, P = 0.012 and the consumption of alcohol, B = 0.29, SE = 0.13, t(1177) = −2.24, P < 0.025. Transformative experience was positively associated with mood, B = 0.33, SE = 0.06, t(1178) = 5.19, P < 0.001, and the use of psychedelic substances, B = 0.37, SE = 0.13, t(1177) = 2.75, P = 0.006. We also built exploratory models to examine the contribution of additional behavioral variables to transformative experiences. These analyses suggested that increased reports of transformative experiences over time could be partially attributed to the formation of new social relationships, gift exchange, and dancing (see SOM 1.13 for details).

To probe the qualities of participants’ transformative experiences, we asked them a number of questions (see Fig. 3 and “Methods” for a complete list). The most frequently reported two qualities of participants’ experiences were feeling socially connected to something larger than oneself (M = 4.92, SD = 1.88) and perceiving something new about others (M = 4.55, SD = 1.83). The least frequently reported qualities were feeling as though one’s self had dissolved (M = 2.81, SD = 1.90) and feeling like a different person than they were before (M = 3.06, SD = 2.00). Figure 3 displays the relative prevalence and 95% confidence intervals of the base rate of each quality. Time-based analysis suggested that the prevalence of each of these qualities increased significantly over time spent at the event, all Ps < 0.05, with the exception of feeling spiritually connected to something larger than oneself and expressing one’s true self (see SOM 1.8). Overall, this analysis suggests that the most prevalent attributes of transformative experiences were socially oriented (e.g., toward others and the community). In contrast, psychedelic substance use most strongly predicted changes to perceptions of reality and oneself, suggesting that transformative experiences elicited by psychedelics may differ in certain key respects from those arising from mass gathering participation alone (see SOM 1.2 for details).

It is possible that people participate in mass gatherings with strong expectations and desires to be transformed, and these expectations and desires may create self-fulfilling prophecies. To test this question, we tested in a cross-sectional analysis whether expectations and desires for transformation predicted reported transformative experience. Consistent with this notion, both expecting (B = 0.16, SE = 0.04, t(1174) = 4.40, P < 0.001) and desiring (B = 0.23, SE = 0.03, t(1176) = 6.83, P< 0.001) a transformative experience positively predicted participants’ actually having one. It should be noted that, because these measures were collected at the same time as reports on transformative experience, it is possible that transformative experiences onsite subsequently increased reports of expectations and desires for one. Some evidence that speaks against this possibility is that self-reported expectations (M = 3.19, SD = 1.91) and desires (M = 3.71, SD = 2.09) collected onsite were, if anything, lower than those collected in the pre-attendance sample (M = 4.06, SD = 1.98, and M = 5.38, SD = 1.86; t(548) = 11.8, P < 0.001 and t(629) = 14.4, P < 0.001, respectively). We cannot, however, rule out the possibility that transformative experience impacted reports of anticipation.

Notwithstanding the strong relationship between the likelihood of expecting or desiring a transformative experience and the likelihood of having one, there was some evidence the transformative experiences we observed onsite over time were not merely due to the self-fulfilling effects of anticipation. First, we found that, of the 49.6% of participants who did not even somewhat expect to be transformed, nearly half (46.7%) reported being at least somewhat transformed. Similarly, of the 41.5% of people who did not even somewhat desire to be transformed, 42.2% reported being at least somewhat transformed. These results show that considerable proportions of people who neither expected nor desired to be transformed nevertheless reported a transformative experience, thereby supporting the possibility that anticipation is not a necessary condition for transformation. In addition, expectations and desires did not correlate with any of our measures of prosocial attitudes behavior (all Ps > 0.4), suggesting that changes in such variables were not the result of anticipation (for additional details regarding the effects of anticipation, see SOM 1.5). Finally, data collected at a “virtual” mass gathering held during the COVID-19 pandemic showed that reports of transformative experience were significantly greater onsite than online, despite greater desires for transformation in the latter (see SOM 1.6). Overall, these results suggest that while anticipating a transformative experience is positively associated with having one, anticipation is neither a necessary nor sufficient condition for having a transformative experience.

Next, we tested whether reports of transformative experience varied across cultural settings. To do this, we included the cultural context variable (gift versus market economy) and its interaction with time (days onsite) as predictors of self-reported transformative experience, additionally controlling for event location (US versus UK). We then tested the significance of the main effect of setting and its interaction with time on a transformative experience. Results showed no significant effects of cultural context, either as a main effect, B = −0.10, SE = 0.11, t(1176) =  −0.86, P = 0.391, or as an interaction with time, B = 0.05, SE = 0.06, t(1176) = 0.76, P = 0.448. We also examined whether any of the qualities of transformation were more prevalent at gift economies than market economies by examining the main effect of setting on each quality. Results showed that participants of gift economies felt more socially connected to something larger than themselves, B = 0.38, SE = 0.15, t(762) = 2.48, P = 0.013. No other effects of cultural setting on transformation qualities were significant. This suggests that the majority of self-reported aspects of mass gathering attendance are consistent regardless of whether or not the event operates a gift economy.

### Prosocial correlates of transformative experience

Next we sought to test the relationship between the self-reported transformative experience and prosocial orientation. Our analysis focused on two primary dependent variables: generosity and moral expansion. We tested whether these prosocial measures were correlated with self-reported transformative experience and whether they increased over time. In addition, because we hypothesized that changes in prosocial behavior would occur as a result of increased feelings of connectedness to other human beings, we also examined the relationship between these measures and the degree of overlap people reported feeling between themselves and all human beings (“universal connectedness”; see “Methods”).

To measure generosity at our field sites, we used a modified dictator game54. In the classic version of the game, participants are given a monetary endowment then given the opportunity to donate some portion of it to an anonymous stranger. In order to make the game suitable for gift economies (where the use of money is discouraged), we modified the game such that participants were endowed with ten tickets that could each be redeemed for a prize from a “mystery box” containing items of value for eventgoers. Without knowing the exact items in the box, participants decided how many tickets they wanted to give to an anonymous stranger by placing them in an envelope which was then handed to the experimenter. We observed an average donation of 62% of the tickets, a level that is notably higher than donations typically observed in dictator games, which average around 28%55 (see Fig. 4A). However, levels of generosity did not significantly change over time onsite, B = 0.05, SE = 0.07, t(769) = 0.74, P = 0.448, and there was no significant relationship between transformative experience and generosity onsite, B = 0.11, SE = 0.06, t(762) = 1.63, P = 0.103 (see Fig. 4B). Furthermore, there was no significant association between universal connectedness and generosity, B = 0.07, SE = 0.07, t(762) = 0.97, P = 0.334. Thus, while overall rates of generosity were substantially higher at multiday mass gatherings than typically observed in laboratory studies of dictator games, there was no evidence that generosity was positively associated with transformative experience or with time spent attending mass gatherings.

To measure moral expansion, we used a hypothetical social discounting measure where participants were asked to indicate how much free time they would be willing to spend doing a favor for people at different social distances. We indexed moral expansion by plotting the amount donated for each distance and calculating the area under the curve for each participant56,57. Supplementary analyses showed responses on this measure correlated with the classic, hypothetical money-based measure of social discounting, as well as with an incentivized measure of charitable giving across various social distances (see SOM 1.3).

Moral expansion significantly increased with time onsite, B = 0.07, SE = 0.02, t(862) = 2.94, P = 0.003, indicating that participants were willing to spend more time helping more socially distant strangers with every passing day spent at mass gatherings (see Fig. 5).

In addition, moral expansion was positively associated with universal connectedness, B = 0.09, SE = 0.03, t(960) = 3.43, P = 0.001, which itself positively associated with transformative experience, B = 0.08, SE = 0.03, t(1126) = 3.11, P = 0.002. These observations raised the possibility that the effect of time on moral expansion was mediated by its effect on the transformative experience and universal connectedness. To test this hypothesis, we constructed a mediation model testing the significance of the indirect effect from time to moral expansion through transformative experience and universal connectedness. Results showed that a significant indirect effect, B = 0.002, SE = 0.001, P = 0.028, CI95[0.000, 0.004]). The direct effect of time on moral expansion remained significant when including this effect in the model, B = 0.071, SE = 0.23, P = 0.002, CI95[.026, 0.117], consistent with partial mediation. A statistical model testing the relationship between time and moral expansion through universal connectedness and transformative experience (as opposed to the reverse, as tested above) showed no significant effects of time on universal connectedness and no significant effects of transformative experience on moral expansion, lending credence to the originally hypothesized variable ordering (see SOM 1.11). Overall, these results suggest that moral expansion increases over time in part as a result of transformative experience and universal connectedness (see Fig. 5).

In supplementary analyses, we explored the possibility that the relationship between time and moral expansion was mediated by group identity fusion—a variable reflecting people’s sense of overlap with other event attendees as opposed to other humans in general. First, we substituted group identity fusion for universal connectedness in the model predicting moral expansion from time onsite and transformative experience (see Fig. 4B). Group identity fusion did not significantly predict moral expansion (B = 0.06, SE = 0.04, t(809) = 1.50, P = 0.135 versus B = 0.09, SE = 0.03, t(960) = 3.43, P = 0.001 for universal connectedness), and the indirect effect of time onsite to moral expansion through group identify fusion was not significant, B = 0.003, SE = 0.002, P = 0.155, CI95[−0.001, 0.006]. Moreover, a model that allowed both variables to compete for variance by including both as covariates showed that universal connectedness predicted moral expansion even when controlling for group identity fusion (B = 0.09, SE = 0.03, t(798) = 3.13, P = 0.002). These results suggest that changes in moral expansion are due more to generalized increases in universal connectedness than to increased connection other eventgoers (see SOM 1.12). It is important to note that universal connectedness and the group identity fusion scales differed in their extremity, with the latter indicating total immersion of the self in the group, while the former indicated only a high degree of overlap, so means of these measures cannot be compared directly. Nevertheless, these results are consistent with the idea that prosocial change at multiday mass gatherings may be the result of a form of universal identity fusion characterized by connectedness to all humanity36,58,59.

We examined whether the effects of time on generosity and moral expansion varied for events with market versus gift economies by looking at interactions between time and setting on universal connectedness, moral expansion, and generosity. We found no main effects of cultural setting on these measures, nor any interactions between cultural setting and time, all Ps > 0.05. Thus we find no evidence that the effect of time on transformative experience and prosocial orientation is moderated by local event norms.

### Persistence of transformative experience and its prosocial correlates

Next, we examined the relationships between transformative experience, generosity, and moral expansion in the weeks and months following event attendance. Because we considered participants following the event to have received a “full dose” of event attendance (descriptive analysis showed that 99.3% of the sample indicated they stayed for more than 1 day), we dropped “days onsite” from these analyses and focused on the relationships between transformative experience, generosity, and moral expansion, with universal connectedness as a potential mediator. Because the majority of the follow-up samples consisted of Burning Man attendees, it was not possible to perform cross-event comparisons on these data.

Immediately following attendance (n = 8515), 19.5% of participants indicated they had “absolutely” had a transformative experience and 71% said they had at least “somewhat” had a transformative experience. Similar figures were evident six months after attendance (n = 696), when 27% of participants reported “absolutely” having a transformative experience and 72.7% reported at least “somewhat” having a transformative experience (Fig. 2). This suggests that the subjective sense of transformation following participation in mass gatherings persists over time.

### Analytic approach

All onsite data were analyzed using multilevel linear regression models using the lme4 and lmerTest packages in R. Component path analysis (that is, direct relationships between the predictor, the mediator, and the dependent variables) included field site as a random intercept and controlled for demographics (i.e., gender, age, education, religiosity, and income), and “incidental variables” (i.e., mood, expectations and desires of having a transformative experience, and the use of psychoactive substances [binary-coded as −0.5 or 0.5 and including euphorics, hallucinogens, stimulants, alcohol, narcotics, and cannabis]). Mediation analyses (i.e., tests of indirect relationships between two variables through one or two intermediary variables) were modeled using linear regression via the lavaan package in R, controlling for all demographic and incidental variables. All mediation analyses report standardized betas and confidence intervals. Models testing the moderating effects of set (expectations and desires) and setting (gift vs. market economies) controlled for all incidental variables as well as event size and location. Follow-up studies and within-subjects effects were modeled using linear regression adjusting for demographics and incidental variables. See SOM 1.5 for more details regarding effects of expectations and desires.

All data were cleaned and analyzed using R data analysis software with packages broom 0.7.11, cowplot 1.1.1, tidyverse 1.3.1, forcats 0.5.1, lavaan, 0.6-9, lm.beta 1.5-1, lme4 1.1-27.1, psych 2.1.9, and sjPlot 2.8.10. Analyses examining behavioral and attitudinal effects across events were performed with a mixed model regression using package lmerTest 3.1-3 in R with event set as random intercept and all other predictors as fixed factors. Mediation analyses were conducted using the lavaan package in R (version .5-23.1097). Online surveys were collected in Qualtrics 2015–2020. In all models, parameter estimates were obtained through “normal” maximum likelihood estimation (using the biased sample covariance matrix), with standard errors based on the observed information matrices. Missing values were estimated using a full information maximum likelihood procedure (FIML).

All significant effects are reported in complete regression tables in SOM 5. Small differences between the component path and mediation analyses are due to the fact that the former modeled event-level random effects while the mediation analyses pursued a simple linear approach. Differences in degrees of freedom across statistical tests are the result of missing data or of the fact that certain variables were not assessed at some of the earlier data collection efforts; all instances in which this is the case are noted in the Methods. All other variables and analyses are described in the SOM.

### Reporting summary

Further information on research design is available in the Nature Research Reporting Summary linked to this article.