Erowid's Experience Report Vaults offer a unique resource for understanding the effects of psychoactive drugs, many of which are poorly described in the scientific literature. Erowid's Vaults contain thousands of reports describing people's experiences. These are a potential gold mine of wisdom and insight. Of course, mining this gold currently takes serious effort. One must read through dozens or more reports to get a sense of the community's experience with a substance, and if one selects a non-representative subset of reports to read, any conclusions one draws may be incorrect. We hope to implement an objective technique for summarizing and visualizing the meaning of these reports.

We will use Latent Semantic Analysis (LS Analysis), a statistical technique that analyzes patterns of word usage in order to interpret text.1 LS Analysis is a well-established technique similar to what is used by search engines to figure out which documents best match search terms. LS Analysis takes a collection of text documents and analyzes the frequency with which individual words occur in each document and across the entire collection: the word "insight" might occur three times in one report, zero times in another, and hundreds of times across the collection. From these word counts, a mathematical matrix is created within which relationships between words, reports, and words and reports can be characterized. Because we know what drug is being described in each report, we are able to look for relationships between individual words or combinations of words, and specific drugs.

From a mathematical point of view, this matrix can be considered a multi-dimensional space. Each unique word in the collection acts as a dimension to define the space and a report's position along each word-dimension is determined by the number of times the word occurs in the report. This allows us to "map" the location of each report or drug within the space. Because spaces with more than three dimensions are inherently difficult to visualize, we will use statistical tools to find maps with fewer dimensions that preserve the relative locations of reports.

We think LS Analysis can help characterize the effects of various substances: which are more visual; which are entactogenic; which are more likely to lead to a bad trip. With this method, drug experiences could even be compared to dreams or "spontaneous" religious experiences. All this should not only be interesting to Erowid visitors, but may also suggest hypotheses for further scientific research.

We are currently testing the efficacy of LS Analysis on a limited subset of the Experience Vaults. Our initial focus is on the more commonly used and better-understood drugs like MDMA and LSD as well as some lesser-known psychedelics for which there are 100 or more reports. We have written code to count words and form matrices and are now beginning to explore the resulting spaces, learning which subspaces are most interesting.

Since Moreau de Tours's work in the mid–nineteenth century2, researchers have suggested that altered states might have certain common elements or dimensions, regardless of how the states are induced. For example, Dittrich reported3 evidence that many altered states include changes in sense of self, feelings of altered reality, and sensory changes. It will be fascinating if we can find evidence of these common underlying dimensions to altered states in the Erowid experience reports.

Latent Semantic Analysis provides an objective way to quantify the meanings of experiences, yet making these meanings clear will be challenging. Work in the field of information visualization suggests a solution: take advantage of the human visual system and its ability to rapidly understand landscapes or other scenes. For example, results of our analyses could be displayed as three-dimensional maps. Each drug could be represented as an island with position, topography, or other properties indicating how much the drug is associated with certain changes in mental state. Longitude might indicate how much the drug changes "ego boundaries", from egotistic west to ego-less east. Latitude might display "sensory changes", from the spare north to the rich sensory confusion of the tropical south. On each island, altitude of mountains could suggest how readily the drug leads to euphoric bliss; towns could indicate prosocial effects. Once learned, such a map could facilitate rapid comprehension of how drugs are similar and different.

Our project brings to mind Terence McKenna's concept of visible language. He described DMT and other tryptamine psychedelics as sometimes allowing people to experience language as visible "high-speed sculpture". One day the results of our analyses could be represented in an interactive virtual reality. As we work on our comparatively drab analysis, we occasionally imagine future Erowid visitors jacking in and navigating the visible language matrix of this community's wisdom.

This project is a collaboration between Matthew Baggott (a neuroscience graduate student at UC Berkeley), David Presti (a lecturer in Molecular and Cellular Biology at UC Berkeley), Juan Carlos Lopez (a researcher at the Addiction Pharmacology Research Lab in San Francisco), and Jeremy Coyle (a cognitive science undergraduate student at UC Berkeley).