Signal detection theory cannot distinguish

Transcription

Signal detection theory cannot distinguish
Signal detection theory cannot distinguish perceptual and response-based biases:
Evidence from the Muller-Lyer illusion and application for action-specific effects
2, John Wixted3
Jessica K. Witt12, J. Eric T Taylor
y 2, Mila Sugovic
g
1Colorado State University, 2Purdue University, 3UC-San Diego
Action-Specific Perception
SDT & Muller-Lyer Illusion
Alternative Techniques
Witt & Sugovic (in preparation)
Performance influences p
perceptual
p
jjudgments:
g
Athletes who perform better than others report
that the target looks bigger, closer, or slower.
Sample
S
l display
di l off the
th experiment.
i
t The
Th top
t left
l ft line
li was the
th standard
t d d
line and was the same length on each trial. The bottom right line
was the comparison line. On each trial, this line was set to 1 of 6
possible lengths and had the tails oriented out (as depicted here) or
the tails oriented in. Ps indicated which line was longer.
1) Fish moves across screen @ 1-6 speeds
2) Press trigger to release net – attempt to intersect fish
3) Speed bisection task: Judge whether fish moved more like the slow
speed or more like the fast speed
Witt et al.(2011)
Witt et al.(2008)
Witt & Sugovic (2010)
Witt & Proffitt (2005)
Judgment-Based Measure
Direction of the effect
is specific to the task:
Parkour experts judge
walls (aka obstacles)
to be lower.
Taylor et al.(2011)
Heavy objects seem
lighter when anticipating
help from another person
Canal-Bruland
& van den
Kamp (2009)
Action-Based Measure
Proportion of trials on which participants
responded that the comparison line was
longer as a function of actual line length
and tail orientation. The gray dashed
arrows indicate the PSE for each tail
orientation
i t ti condition.
diti
Table 1. Means (and Standard Deviations) from SDT analyses with the Muller‐Lyer illusion. Targets within reach as
a result of using a tool
look closer to actors
and to observers
Witt & Dorsch (2009)
Bloesch et al. (2012)
Davoli et al., (2012)
Linkenauger et al. (submitted)
Witt et al.(2005)
Doerrfeld et al.
al (2011)
Question: Does Perception Change?
Classifications:
Hits = correctly classifying long lines as long
False Alarms = incorrectly classifying short
lines as long
• 3 Sources of Support:
• Prior demonstrations: Mack et al., 2011; Wixted & Stretch, 2000
• Empirical: Muller-Lyer illusion influences c, not d’
• Theoretical: Bias in perception and bias in decision both create
effects in c, not d’
References
2.73 (0.82)
1.50 (.62)
Tails-Out
2.50 (1.06)
-1.41 (.71)
ns
n.s.
***
Eq ations Used
Equations
Used:
d' = [ z(H) – z(FA) ] / √2 (1)
c = -0.5 * [ z(H) + z(FA) ] (2)
d’
Judgment-Based Measure
Changes that Influence d’
b
c
Means are changed
Baseline
• Both of the changes (depicted in a and c)
would result in changes in d’
• Neither change is predicted to occur for
the Muller-Lyer illusion (or for actionspecific effects)
• So no predicted effect in d’
d for either.
either
Spread is changed
Balcetis, E. & Dunning, D. (2010). Wishful seeing: Desirable objects are seen as closer. Psychological Science, 21, 147-152.
Bloesch,E.K., Davoli,C.C., Roth,N., Brockmole,J.R., & Abrams,R.A. (2012). Psychonomic Bulletin & Review.
Changes that Influence c
Cañal-Bruland, R. & van der Kamp, J. (2009). Psychonomic Bulletin & Review, 16, 1100-1105.
Davoli, C. C., Brockmole, J. R., & Witt, J. K. (2012). Journal of Experimental Psychology: Human Perception and Performance, 38.
Doerrfeld, A., Sebanz, N.,& Shiffrar, M. (2011). Expecting to Lift a box together makes the load look lighter. Psychological Research.
Lessard, D.A., Linkenauger, S.A., & Proffitt, D.R. (2009). Look before you leap: Jumping ability influences distance perception. Perception
d
Linkenauger, S.A., Bulthoff, H.H., & Mohler, B.M. (submitted). Go-go-gadget arms! The influence of virtual arm reach on perceived…
No perceptual bias
Response bias for “Long”
Li k
Linkenauger,
S
S.A.,
A Witt
Witt, JJ.K.,
K &P
Proffitt,
ffitt D.R.
D R (2011).
(2011) Journal
J
l off Experimental
E
i
t l Psychology:
P
h l
Human
H
Perception
P
ti and
dP
Performance,
f
37
37.
Mack,M.L., Richler,J.J., Gauthier,I., & Palmeri,T.J. (2011). Indecision on decisional separability. Psychonomic Bulletin & Review, 18.
Taylor, J.E.T., Witt, J.K., & Sugovic, M. (2011). When walls are no longer barriers: Perception of wall height in parkour. Perception,40.
e
Baseline
Stefanucci, J. K., & Proffitt, D. R. (2009). Journal of Experimental Psychology: Human Perception and Performance, 35, 424-438.
• Anticipated results based on a responsebased bias (a) and a perceptual bias (c)
• Both types of biases result in the same
relative distributions of hits and false alarms,
so no change to d’
• Both types of biases lead to an increase in
hits and false alarms, leading to effects in c
Witt, J. K., & Dorsch, T. (2009). Perception, 38, 1328-1340.
Witt, J. K., Linkenauger, S. A., Bakdash, J. Z., & Proffitt, D. R. (2008). Psychonomic Bulletin and Review, 15, 581-585.
Witt, J. K., & Proffitt, D.R. (2005). Psychological Science, 16, 937-938.
Witt, J. K., Schuck, D. M., & Taylor, J. E. T. (2011). Action-specific effects underwater. Perception, 40, 530-537.
Witt, J. K., & Sugovic, M. (2010). Performance and ease influence perceived speed. Perception, 39, 1341-1353
Wixted,J.T., & Stretch,V. (2000). The case against a criterion-shift account of false memory. Psychological Review, 107, 368-376.
f
F(2, 36.05) = 5.01, p < .05, ηp2 = .22
Action-Based Measure
Experiment 2: Judge speed of fish
on a scale of 1-7:
Note. n.s. = not significant. *** = p < .001.
• Hypothetical distributions of perception of short lines (blue curves) and long lines (pink curves)
• Black line represents the decision criterion location for classifying a line as “long” rather than “short”
• Hits: Correctly classifying long lines as long
• False Alarms: Misclassifying short lines as long
• Area under pink curve to right of black line
• Area under blue curve to right of black line
a
F(2, 36) = 16.73, p < .001, ηp2= .48
c
Overview of SDT Measures
• What is being affected: Perception or Responses?
• Can we use Signal Detection Theory (SDT) to separate out processes
related to perception and processes related to the response?
• Seems like SDT is designed to do just that
• Answer: NO
• SDT cannot separate perceptual biases from response biases
Tail
Orientation
Tails-In
Perceptual bias to
see as longer
No response bias
Thus, c cannot be used to discriminate
between decision-based and
perceptual biases
F(2, 20) = 7.64, p < .01, ηp2 = .43
F(2, 22.03) = 5.10, p < .05, ηp2 = .32
Discussion
• Signal Detection Theory cannot be used to differentiate between
perceptual biases and response-based biases
• Such as the Muller-Lyer illusion and action-specific effects
• A technique that is valid must be able to prove that the Muller-Lyer illusion
is perceptual (or researcher must claim the illusion is not perceptual)
SDT researchers must rely on other techniques to evaluate
• Without SDT,
whether action-specific effects are perceptual
• Action-based measures (see also Balcetis & Dunning, 2010; Witt et al. 2010)
• Indirect measures (see Stefanucci & Proffitt, 2009; Witt, 2011)
• Functional Specificity (Lessard et al., 2009; Linkenauger et al., 2011; Witt et al., 2004)
• Rule out methodological artifacts/demands (Witt & Sugovic, in press, submitted)
Perceive environment in terms of ability
y to act on it
For More Information
Contact Jessi Witt (jkwitt@purdue.edu), Purdue University
Starting at Colorado State University in Fall 2012
Web: www.psych.purdue.edu/~jkwitt