Structuring first person reports: a coding instrument for phenomenological dimensions (pilot)
Evan Richard Nolan
OSF Preprints October 12, 2026 preprint DOI: 10.17605/osf.io/dxvn9 (opens in new tab)
Study at a glance
AI-extracted from the abstract| Characteristics | Practical guide or manual |
|---|---|
| Sample size | 60 |
| Population | 30 dream reports from DreamBank and 30 recalled waking memories from Hippocorpus |
| Key points | The authors propose a five-dimension coding instrument for first-person experience reports and an analysis plan written before any human scores existed. The plan focuses on whether the instrument measures what it claims, checking for confounds from report length and source corpus before any comparison between dream and waking states. No results are reported; human coding, inter-rater reliability, and LLM coder validation are in progress. |
Abstract
A pilot coding instrument for turning free text first person reports of experience into structured, comparable data. The manual defines five dimensions, each with definitions, boundary cases and worked examples: vividness, world anchoring (internally generated versus externally anchored), agency, reality status at the time, and interpretive load (how much of a report asserts meaning or metaphysical claims rather than describing what was experienced). The pilot applies the instrument blind to 30 dream reports (DreamBank) and 30 recalled waking memories (Hippocorpus), each coded in full and as its first 120 words. The analysis plan was written before any human scores existed and focuses on whether the instrument measures what it claims, checking for confounds from report length and source corpus, before any comparison between states. Status (October 2026): instrument and analysis plan complete; human coding, inter rater reliability and an LLM coder validated against human codes are in progress. No results are reported yet. Report texts are not shared here, as redistribution terms for the source corpora are not fully specified. Code was developed with AI assistance (Claude Code).