Abstract
BACKGROUND: A fundamental unsolved problem in psychophysical detection experiments is in discriminating guesses from the correct responses. This paper proposes a coherent solution to this problem by presenting a novel classification method that compares biomechanical and psychological responses. METHODS: Subjects (13) stood on a platform that was translated anteriorly 16 mm to find psychophysical detection thresholds through a Adaptive 2-Alternative-Forced-Choice (2AFC) task repeated over 30 separate sequential trials. Anterior-posterior center-of-pressure (APCoP) changes (i.e., the biomechanical response R(B)) were analyzed to determine whether sufficient biomechanical information was available to support a subject's psychophysical selection (R(Ψ)) of interval 1 or 2 as the stimulus interval. A time-series-bitmap approach was used to identify anomalies in interval 1 (a₁) and interval 2 (a₂) that were present in the resultant APCoP signal. If a₁ > a₂ then R(B) = Interval 1. If a₁ < a₂, then R(B)= Interval 2. If a₂-a₁ < 0.1, R(B) was set to 0 (no significant difference present in the anomaly scores of interval 1 and 2). RESULTS: By considering both biomechanical (R(B)) and psychophysical (R(Ψ)) responses, each trial run could be classified as a: 1) HIT (and True Negative), if R(B) and R(Ψ) both matched the stimulus interval (SI); 2) MISS, if R(B) matched SI but the subject's reported response did not; 3) PSUEDO HIT, if the subject signalled the correct SI, but R(B) was linked to the non-SI; 4) FALSE POSITIVE, if R(B) = R(Ψ), and both associated to non-SI; and 5) GUESS, if R(B) = 0, if insufficient APCoP differences existed to distinguish SI. Ensemble averaging the data for each of the above categories amplified the anomalous behavior of the APCoP response. CONCLUSIONS: The major contributions of this novel classification scheme were to define and verify by logistic models a 'GUESS' category in these psychophysical threshold detection experiments, and to add an additional descriptor, "PSEUDO HIT". This improved classification methodology potentially could be applied to psychophysical detection experiments of other sensory modalities.