Outline: Psychophysics and Signal Detection Theory
Introduction
Stimulus
Sensation, Transduction
Perception
Psychophysics
Allied fields: Neuroscience, Computer Vision
Marr and levels of explanation
Computational
Algorithmic
Implementation
Psychophysics
Define
Limits on psychophysical/behavioral method
Threshold/Limen
Definition
Methods for measuring it:
Ascending/Descending Limits
Adjustment
Constant stimuli/Signal Detection/Yes-no procedure
2-interval forced choice
Adaptive procedures
Psychometric function
Definition
Ogive
Random threshold? No!
Detection versus discrimination
Just noticeable difference
Weber's Law: ΔI/I = k, where I is physical stimulus strength, ΔI is the just noticeable
difference, and k is a
constant.
Weber's Law fails at low magnitudes
Magnitude of Sensation
Fechner: number of jnd's, so that (with
Weber's Law) S = c log I
Plateau/Stevens: magnitude estimation/production, S = k In
Invariances [if we cover this]
Ricco's Area (10')
Bloch's Law (e.g. 100 msec)
Auditory equivalent
Signal Detection Theory
"Threshold" doesn't exist or is
arbitrary
Unify detection and discrimination
Dice example or tumor detection example
Detectability
Outcomes: hit, miss, false alarm, correct reject
Criterion, hence no fixed threshold
Payoff matrix
A Priori probabilities
Receiver Operating Characteristic (ROC curve):
Probability of a Hit on signal trials
as a function of
the probability of a False Alarm on noise-only trials
Analogy to human observers in Yes-No tasks
Yes-No versus 2IFC
But: Hecht, Schlaer & Pirenne: detection of a single quantum
per receptor