The continuity hypothesis of dreams - a widely studied model of dreaming - suggests that the content of dreams is largely continuous with the waking experiences of the dreamer. Given the unprecedented nature of the experiences during the pandemic of COVID-19, we studied the continuity hypothesis in the context of such a pandemic. To that end, we implemented a state-of-the-art deep-learning algorithm that can accurately extract mentions of virtually any medical condition from text and applied it to two sets of data collected during the COVID-19 pandemic: 2,888 dream reports (dreaming life experiences), and 57M tweets mentioning the pandemic (waking life experiences). We found that the health expressions that were shared by both sets were common COVID-19 symptoms (e.g., coronavirus, anxiety, coughing, and stress), suggesting that dreams reflected people’s real-world experiences. On the other hand, we found that the health expressions that distinguished the two sets reflected differences in thought processes: health expressions in waking life reflected a linear and logical thought process and, as such, described realistic symptoms or related disorders (e.g., body aches, nasal pain, SARS, H1N1); by contrast, those in dreaming life reflected a thought process likely based on the activation of the visual and emotional areas of the brain and, as such, described either conditions not necessarily associated with the pandemic’s virus (e.g., maggots, deformities, snakebites), or conditions of surreal nature (e.g., teeth suddenly falling out, body crumbling into sand). Our results confirm that, in addition to the sources of health data being researched lately (e.g., psychological conditions inferred from social media posts, physiological readings from commercial wearables), dream reports, if interpreted correctly, represent an understudied yet valuable source of people’s health experiences in the real world.

In this study, we compared textual expressions of health experiences in waking discussions and in dreams. To do so, we resorted to a state-of-the-art deep-learning method (https://dl.acm.org/doi/10.1145/3368555.3384467) capable of extracting mentions of any medical condition from text, and applied it to two main data sources:

1) A collection of 57M tweets posted in relation to COVID-19 (https://publichealth.jmir.org/2020/2/e19273/)

2) A set of written dream reports that describe the dreams of 2,888 people during the pandemic (https://psycnet.apa.org/record/2020-71980-003)

This dataset contains two main sources of data:

A) The list of medical entities extracted from Twitter and dream reports, together with their frequency

B) A co-occurrence graph of the 1,732 unique medical conditions we found in dream reports. In this graph, nodes represent all the medical conditions extracted from the dream reports. Two nodes are connected by an edge if they both appeared in the same dream report. Edges are weighted by the number of dreams in which the two conditions co-occurred.

See README.txt for details on the different fields