All varieties of dreaming remain a mystery. Lucid dreams in particular, or those characterized by awareness of the dream, are notoriously difficult to study. Their scarce prevalence and resistance to deliberate induction make it difficult to obtain a sizeable corpus of lucid dream reports. The consequent lack of clarity around lucid dream phenomenology has left the many purported applications of lucidity under-realized. Here, a large corpus of 55k dream reports from 5k contributors is curated, described, and validated for future research. Ten years of publicly available dream reports were scraped from an online forum where users share anonymous dream journals. Importantly, users optionally categorize their dream as lucid, non-lucid, or a nightmare, offering a user-provided labeling system that includes 10k lucid and 25k non-lucid, and 2k nightmare labels. After characterizing the corpus with descriptive statistics and visualizations, construct validation shows that language patterns in lucid-labeled reports are consistent with known characteristics of lucid dreams. While the entire corpus has broad value for dream science, the labeled subset is particularly powerful for new discoveries in lucid dream studies.
Figures & tables
Figure 1: \par Example DreamViews dream journal entry. The categories labeling system, seen at the bottom, was used to extract labels for the corpus. This entry is from the author’s journal.
Table 1: \par Description of data files.
Figure 2: \par Total corpus size. Left axis values correspond to bars and right axis values correspond to cumulative distribution lines. Note that the final cumulative distributions endpoints provide a rough visualization of total corpus counts. The top panel shows the number of unique users that posted each month, where repeat/gold users are those that also posted in a prior month. The bottom panel shows the amount of unique posts each month, where colors indicate the lucidity label of posts. Note that the popular lucid dreaming film Inception saw releases in July (USA theater) and December (USA home video) of 2010.
Figure 3: \par Number of posts per user.
Figure 6: \par Number of words per dream report. n : sample size, xˉ : mean, σxˉ : standard deviation, x~ : median
Table 2: \par Top category and tag labels.
Figure 7: \par Reported user locations.
Figure 8: \par Reported user age and gender.
Figure 9: \par User-generated label frequencies, per post and per user. Within each section, the top number represents the number of posts and the bottom number represents the number of unique users that contribute to the post count. Non-lucid labels occur the most frequently, lucid labels second, and nightmare labels least. Venn sets are sized proportionally to user-normalized post frequencies.
Figure 10: \par The frequency of users contributing lucid and non-lucid posts. The center panel highlights the number of users at each lucid and non-lucid post frequency. The dashed lines close a box around all post frequencies that include ≥\par\par lucid and ≥\par\par non-lucid, and the inner text provides the total number of users this is true for. The dotted lines close a box around all post frequencies that includes ≥\par\par lucid and ≥\par\par non-lucid, motivated by the finding that \par dream reports provides a stable estimate of dream content within individuals [ schredl1998\par\par ] . Distributions in the marginal panels indicate the raw and cumulative number of users who contribute to either lucid or non-lucid post frequencies.
Figure 11: \par Lucidity JSD word shift. Words and phrases that most distinguish user-labeled lucid and non-lucid dream reports are consistent with known lucidity characteristics. These results offer content validity for user-generated lucidity labels.
Figure 12: \par Nightmare fear word shift. Words and phrases that most distinguish user-labeled nightmares from all other dream reports are consistent with known nightmare characteristics. These results offer content validity for user-generated nightmare labels.
Figure 15: \par LIWC validation results. Dream reports labeled as lucid by the user contain a higher percentage of insight and agency dictionary words than those labeled as non-lucid dream reports. These results offer convergent validity for the user-labeled dream reports.