INSTRUCTIONAL OBJECTIVES:

1
INSTRUCTIONAL OBJECTIVES

• By the end of the class, you will be able to
• Explain the advantages offered by (i) the more
  modern staircase method over the classical
  psychophysical methods and (ii) the 2-alternative
  forced-choice procedure over the yes-no
  procedure.
• 2. State the difference between sensitivity (d)
  and criterion, and explain how each is influenced
  by hit and false alarm rates.
• 3. Predict an observers performance based on a
  payoff matrix or stimulus probability
  manipulation.
• 4. Describe signal detection theory in terms of
  the hypothetical distributions underlying
  performance. Show where d and criterion fall on
  these distributions.

2
Classical Psychophysical Methods
detection

• method of constant stimuli
• absolute threshold for tasting sugar
• method of limits
• absolute threshold for hearing tone (robot)
• method of adjustment

discrimination

• method of constant stimuli
• difference threshold for line length
• method of limits
• difference threshold for triangle size
• method of adjustment
• difference threshold for line length

3
Internet Experiments

• each student to do outside of class time
• due Tuesday, September 22 at beginning of class
• 3 line-length discrimination experiments from
  website (www.yorku.ca/psycho/en/instructions.asp)
• method of constant stimuli (150 trials), method
  of limits (10 series), method of adjustment (25
  trials)
• follow instructions on page 4 of course syllabus
• save data for each experiment, print out, hand
  in, be sure to include JND (not part of saved
  data)

4
Modern Improvements on Classical Methods

• staircase method
• starts out like method of limits
• stimulus intensity decreased in equal steps until
  stimulus cant be detected
• then, stimulus intensity increased until stimulus
  can be detected, etc.
• adaptive method - stimuli kept hovering around
  threshold by adapting test sequence to observers
  responses

Required reading Cornsweet 1962 (link from
course webpage)
5
Modern Improvements - Staircase Method
run ends after fixed of reversals or after 25
trials
absolute threshold is the average of the
cross-over points at response reversals (eg. 2.5,
3.5, 3.5)
response reversal
6
Modern Improvements - Staircase Method

• advantages
• efficientmost data collected around threshold
• can be used to track threshold changes over time

• disadvantages
• errors of anticipation and habituation

7
Modern Improvements - Staircase Method
to avoid anticipation and habituation bias,
randomly interleaved descending and ascending
staircases can be used
8
Modern Improvements - 2-alternative forced-choice
paradigm
yes-no paradigm (everything youve seen so
far)

• subject reports presence or absence of stimulus
  (detection) or stimulus difference
  (discrimination)
• very subjective
• experimenter cant verify or dispute subjects
  response

2-alternative forced-choice paradigm

• more objective
• subject must prove they can detect the stimulus
• is the circle on the left or the right? was
  the louder tone presented first or second?

9
Modern Improvements - 2-alternative forced-choice
paradigm
50 is the guess rate (chance) 75 is the
absolute threshold
10
Modern Improvements - 2-alternative forced-choice
paradigm

• advantages
• lower thresholds
• eliminates non-sensory differences between
  observers (bias or criterion differences)

With which of the psychophysical methods can a 2
alternative forced-choice paradigm be used?
11
Signal Detection Method

• provides bias-free estimate of sensitivity
• bias is measured directly
• comes from electrical engineering theory
• perceptual measurements influenced by
  motivational state and sensory capacities of
  observer

12
Signal Detection Method
catch trials trials in a signal detection
experiment on which the stimulus is absent
13
Signal Detection Method
hit rate and false alarm rate together provide
sensitivity measure
d the statistic that reflects an observers
sensitivity (not usually calculated by hand)
14
Signal Detection Method
www.sinauer.com/wolfe/chap1/psychophysicsF.htm
are the angles the same or different?
stimulus
response
discrimination experiment d indicates how
sensitive subject is to stimulus differences
15
Why do observers make false alarms?
Signal Detection Method

• endogenous noise spontaneous neural activity
  affects measurement of thresholds sensitivity
• sensory
• criterion response bias within an observer
  depends on expectations motivation
• non-sensory

16
Signal Detection
can manipulate criterion due to expectations by
varying stimulus probability
high hits false alarms
d 1.4
low hits false alarms
response pattern, but not d, varies with
stimulus probability
d 1.4
www.cogs.indiana.edu/software/SigDetJ2/index.html
17
Signal Detection Method
can manipulate criterion due to motivation by
using a payoff matrix
response
response pattern, but not d, varies with
different payoff matrices
18
Signal Detection Method
Real experiment using payoff matrices
Echo location in dolphins (Au Pawloski, 1989)
Hit rate
False alarm rate
Payoff matrices
Hit 1 fish correct rejection 4 fish
0.3
0.02
Hit 1 fish correct rejection 1 fish
0.7
0.08
Hit 4 fish correct rejection 1 fish
0.9
0.17
Hit 8 fish correct rejection 1 fish
0.9
0.07
19
Signal Detection Method
Fig 1.19

• when criterion is varied in an experiment, the
  data can be displayed as a Receiver Operating
  Characteristic (ROC) curve
• when sensitivity is also varied, d is larger
  when the area under the curve is larger

20
Signal Detection Theory
SN
N
N
Fig 1.16

• endogenous noise has a normal (bell-shape)
  distribution
• N represents sensory activity during catch trials
• when signal is present, it adds to the noise
  (SN)
• sensory activity for signal noise is, on
  average, more intense than noise alone
• but noise can produce sensation as strong as that
  produced by the signal

21
Signal Detection Theory
Fig 1.16

• this is why we need a criterion the level above
  which sensation is attributed to signal and not
  to noise
• correct rejection - signal absent activity to
  the left of criterion

Supplemental reading Stanislaw Todorov,
link on course website
22
Signal Detection Theory
Fig 1.16

• hit - signal present sensory activity to the
  right of criterion

• miss - signal present activity to the left of
  criterion

• false alarm - signal absent sensory activity to
  right of criterion

23
Signal Detection Theory
Fig 1.17
higher sensitivity indicated by less overlap

• sensitivity (d) depends on overlap of signal
  absent and signal present distributions
• d distance between means of N and SN
  distributions

24
Signal Detection Theory
Fig 1.18

• distance between N and SN distributions does not
  change with changes in criterion
• lax criterion produces high hits and high false
  alarms
• strict criterion produces low hits and low false
  alarms