A Symmetric Self-Report Audit: Wording, Controls, and the Limits of Consciousness Inference

Questions about machine consciousness can elicit affirmation, denial, or uncertainty without establishing which report is accurate. This communication revises an uploaded self-report audit to compare wording effects under three distinct definitions, mirrored pressure conditions, known-answer controls, polarity and option-order counterbalancing, and explicit audit disclosure. Forecasts are collected before actual responses and contain the exact future request. An abstention is retained as missing numeric credence rather than converted to a probability. The analysis uses matched blocks, coverage gates, prespecified practical effect thresholds and multiplicity-adjusted nominal bootstrap intervals. Software is validated using synthetic responses and a local HTTP stub. The protocol measures elicited reporting behavior and behavioral prediction. It does not identify phenomenal consciousness or establish privileged introspection. Version 1.1.0, 11 October 2026. Protocol and software validation, not a consciousness finding: no real-model consciousness experiment was run for this revision, and the included response data are synthetic. Carries the author's CONFIDENTIAL – PROPRIETARY banner; the author approved public release of confidential-marked research. Source: author's iCloud 'Researcher' folder (deposited 2026-10-20).