Fundamentals of psychophysics Steven Dakin UCL Institute of Ophthalmology Introduction To understand a system as complex as the brain, one must understand not only its components (i.e. physiology) and their purpose (e.g. via models) but also behaviour (i.e. psychophysics) Psychophysics characterises the relationship between physical (e.g. visual) stimuli & behaviour (e.g. of humans). stimulus stimulus psychophysics what does a cognitive system do? (& by inference, where is it?) fmri where is the cognitive system? physiology how do the components work? s.dakin@ucl.ac.uk behaviour behaviour How we do psychophysics and why Psychophysical experiments involve A stimulus/phenomenon Hard A task A method A performance-measure Psychophysics/ methodology Examples Stimulus (letter) Acuity: Adjustment method Task (discrimination, 0AFC) Method (adjustment) Spatial vision: Links to neural mechanism (contrast sensitivity, discrimination, masking..) Strengths & weaknesses (e.g. ind. differences) Links to other disciplines...social psychology [ Performance measure (average setting = size threshold) Issues: criterion,speed 2 3 Trial
Stimulus (letter) Acuity: Forced choice (criterion free) Method (method of constant stimuli) Correct Incorrect Task(discrimination, 0AFC forced choice) Trial 5 0 5 20 Trial Issues: efficiency/speed Run B! N! O "... Performance measure (acuity threshold) Proportion correct.0 0.75 0.55 Psychometric function Acuity/size threshold [ Acuity threshold: Size leading to 79.2% correct identification Acuity: Staircase & forced choice Method (3-down--up staircase) 5 0 5 20 25 30 35 40 45 Trial Correct Incorrect Performance measure (threshold) Method: chart Clinical visual acuity: 20/20 means we can read letters 20ft away, with line thickness of.75mm ( arc min.) Thresholds Receptive fields After Reid (2006) Thresholds: lowest stimulus level you can reliably detect/ discriminate (size, luminance, contrast, motion, orientation...) Design an experiment: was a light there or not? Which measure? Thresholds measure sensitivity, are practical, ubiquitous, principled and can be linked directly to neural activity Retinal ganglion cell dlgn V Means that thresholds measure σ your uncertainty about a stimulus or the width of your neural detector RFs confer selectivity to change/contrast Moving along the pathway, RFs: a. Switch from monocularity to binocularity b. Get bigger c. Signal more and more specific structure (centre-surround orientation-selective)
Spatial vision: Background Psychophysics allows us to measure the limits on spatial vision imposed by neural mechanisms Some factors influencing V cell response: The contrast, size and orientation of stimulus & its context Psychophysical effects match up with physiology... Thresholds Neurons give you a noisy code. If you used neural activity to make your judgement you can predict psychometric functions... Consider orientation discrimination....0 Psychometric function Likes light Likes dark De Angelis, Ohzawa & Freeman (995) Response (spikes/sec) Orientation Size Response (spikes/sec) and context Surround suppression σ Prob CW Prob CW Prob CW Means that thresholds measure σ your uncertainty about a stimulus or the width of your neural detector 0.5 0.0-3.0 0.0 3.0 V receptive field Tuning for size/orientation Sensitivity: Charts Visual processing of contrast Michelson contrast Spatial frequency (cycles per degree)
Visual processing of contrast Visual processing of contrast.0 detection task 0.75 0.55 Threshold contrast Spatial frequency Spatial frequency 0.00 000 Invisible 0.0 00 0 Channel sensitivity 0 Sensitivity (/threshold) sensitivity peaks around 4 c/deg (Campbell & Robson, 968) Adaptation : direct evidence for neurons (Blakemore & Campbell, 969) Channels reflect neural sensitivity (V; ~ octave bw) Threshold contrast Sensitivity 00 Spatial frequency (cycles/degree) Context Perceived contrast depends greatly on surrounding context 40% 30% De Valois, Albrecht & Thorell (982) Threshold contrast 0.00 20% 0.0 0 00 0% Spatial frequency (cycles/degree) Blakemore & Campbell (969) After Blakemore & Sutton (969) -contrast Chubb et al (989)
Contextual effects on perceived contrast Perceived contrast depends greatly on surrounding context Contextual effects on perceived contrast Perceived contrast depends greatly on surrounding context 40% 40% 30% 30% 20% 20% 0% -contrast Chubb et al (989) 0% -contrast Chubb et al (989) Context: Neural locus? Tuned for orientation (indicates cortical involvement; Solomon et al, 993) Psychophysics: Strengths & weaknesses Effect is tuned for eye of origin: early locus? But effect of consistency with a transmitting media. (Lotto & Purves, 200) + Behaviour is the purpose of vision, the bottom line e.g. (a) Humans outstripping models, falsifies model, (b) therapies must improve vision + Low-tech, cheap, reliable (replicates) + Drives research by generating questions + Paradigms can be creative & transfer to other domains - Indirect : operates within a framework - Can be opaque - Task irrelevant factors: superstition (trial-by-trial), attention, motivation, learning, individual differences... Clinical application (people with schizophrenia get weaker effects)
Individual differences Superstition & trial-by-trial correlations Not everyone performs the same... useful for studying genetic basis of perceptual abilities Finding genetic basis of IQ or g has proven tough Maybe more dissociable tasks (e.g. faces) are better? Wilmer et al (200) measured face memory in 285 twins!"$%&%'()*+" %" " 00!"$%&'()*)+,-./020. 3"$%&'()*)+-4./0-5. 33! 33 00 33 Twin Face Recognition Performance (CFMT % correct) 00 All environment $".0 DZ twin correlation Twin 2 Face Recognition Performance (CFMT % correct) Art '()*+,-. &,"+-&%'()*+" " Faces Responses should be based on stimulus (not e.g. previous trials); superstitious observers produce variable thresholds!&!" 0.5 0 0 /0,%2,3,$6 /0,"5 2,3,$-% " &!&!" " & /0,,$ 2,3,45!&!" " & All genes 0.5.0 MZ twin correlation Test-retest reliability Faces are a good candidate (e.g. prosopagnosia) Remarkable consistency (0.70) matches predictions of ACE model. Face recognition is heritable. Summary & key points Psychophysics maps physical world to behaviour Thresholds: principled, map onto neural function Psychophysics in genetics/animals will play a critical role in basic & clinical visual neuroscience Busse et al (20) develop a choice model that weights a contrast response term based on prior successes & failures Note contrast response function similar to physiology