Deadpan Prevalence Validation
Tags: voice-analysis • deadpan • prevalence • validation
Deadpan Prevalence Validation
Summary
The deadpan hypothesis moved from candidate retrieval to representative prevalence measurement.
A deterministic 140-post sample was drawn from the corrected 1,342-post research view. Every post was annotated and human-reviewed.
Result
Sixteen sampled posts contained a qualifying absurd-humorous premise. Eleven used deadpan delivery and five used overt or explanatory delivery.
The conditional deadpan rate is 68.75%, with a 95% Wilson interval of 44.40% to 85.84%.
The predeclared frequency threshold required the lower bound to be at least 20%. The result passes.
Interpretation
“Frequently” is supported only within the correct denominator: Facebook posts containing a discernible absurd or impossible humorous premise.
Deadpan posts account for 7.86% of the complete sample. The result does not mean that most Facebook writing is deadpan.
Lesson
Candidate retrieval establishes existence and recurrence. It cannot establish prevalence.
A defensible frequency claim requires:
- A representative sample.
- A denominator defined before annotation.
- A threshold defined before seeing results.
- Human-reviewed labels.
- Uncertainty reported with the point estimate.
Next Step
Validate and commit the deadpan prevalence increment. Later prevalence studies should use independent annotation or a larger sample before raising confidence beyond medium.