Visual search: Who cares?

This is a visual task
that is important
outside psychology
laboratories (for both
humans and nonhumans).
X
X
X
X X X
XX X
X
X
X X
X
Feature search
X
X O
O
X
O
O
X O XO
X
X
Conjunction search
Treisman & Gelade 1980
Reaction Time (ms)
“Serial” vs “Parallel” Search
Set size
Feature Integration Theory:
(FIT) Treisman (1988, 1993)





Distinction between objects and features
Attention used to bind features together
(“glue”) at the attended location
Code 1 object at a time based on location
Pre-attentional, parallel processing of
features
Serial process of feature integration
FIT: Details


Sensory “features” (color, size,
orientation etc) coded in parallel by
specialized modules
Modules form two kinds of “maps”

Feature maps


color maps, orientation maps, etc.
Master map of locations
Feature Maps

Contain 2 kinds of info

presence of a feature anywhere in the field



there’s something red out there…
implicit spatial info about the feature
Activity in feature maps can tell us
what’s out there, but can’t tell us:


where it is located
what other features the red thing has
Master Map of Locations


codes where features are located, but
not which features are located where
need some way of:



locating features
binding appropriate features together
[Enter Focal Attention…]
Role of Attention in FIT




Attention moves within
the location map
Selects whatever features
are linked to that location
Features of other objects
are excluded
Attended features are
then entered into the
current temporary object
representation
Visual Search Experiments



Record time taken to determine
whether target is present or absent
Vary the number of distracters
FIT predicts that


Feature search should be independent of
the number of distracters
Conjunction search should get slower
w/more distracters
Feature Search: Find red dot
“Pop-Out Effect”
Conjunction: white vertical
1 Distractor
12 Distractors
29 Distractors
Feature Search



Is there a red T in
the display?
Target defined by a
single feature
According to FIT
target should “pop
out”
T
T
T
T
T T
T
TT
T
T
Conjunction Search



Is there a red T in the
display?
Target defined by shape and
color
Target detection involves
binding features, so
demands serial search
w/focal attention
T
X
X
T
T X T T
T
TX T T
X
Typical Findings & interpretation

3000

Feature Target
2500

Conjunction
Target
2000
RT (ms)
Feature targets pop out
Conjunction targets
demand serial search

1500
1000
500
0
1
5
15
Display Size
30
flat display size function
non-zero slope
… not that simple...
X
O
X
X
O
X
O
O
X
O
X
O
X
easy conjunctions - -
depth & shape, and movement & shape
Theeuwes & Kooi (1994)
Is 3D a feature?
Enns & Rensink (1991)

Search is very fast in this situation only when the
objects look 3D - can the direction a whole object
points be a “feature”?
Asymmetries in visual search
Vs
Vs

the presence of a “feature” is easier to find than
the absence of a feature
Kristjansson & Tse (2001)

Faster detection of presence than
absence - but what is the “feature”?
Duncan & Humphreys (1989)


SIMILARITY
visual search tasks are :


easy when distracters are homogeneous and
very different from the target
hard when distracters are heterogeneous and
not very different from the target
There is a continuum of searches
Familiarity and asymmetry
asymmetry for German but not Cyrillic readers
Guided Search
There is a stimulus
Local salience is computed
local differences
create bottom-up
salience
“red”
“steep”
A limited set of coarse, categorical
features are computed
A weighted sum creates an
activation map
The activation map: local salience is weighted
heavily and will attract attention (bottom-up)
Top-down guidance:
Give weight to what you want
Find the green verticals
The activation map guides re-entrant
attentional selection of objects
but you do not “see” the output of
the activation map
Guided Search is a two-stage model
Object
Recognition
?
?
?
?
?
?
First Stage
Bottleneck
?
?
Second Stage
FirstThe
stage
information
guidesSearch
access to the
core
idea of Guided
second stage
?
?
?
?
?
?
First Stage
Bottleneck
?
?
Second Stage
?
?
?
?
?
?
First Stage
Bottleneck
?
?
Second Stage
binding stage
?
?
?
?
?
?
First Stage
Bottleneck
?
?
Second Stage
A vexing problem
Find the 5
Umm…there is no 5
How do you know when to stop?
We know you are not
marking every reject
How do you know when to stop?
The number marked as
rejected is small (~4)
How do you know when to stop?
Palmer (after Broadbent)
cue
250 ms
interval
750 ms
test
100 ms
(Palmer, after Shaw)
Decision Integration Model based on SDT
Processing before decision is assumed to be independent for each
stimulus and may or may not be task-specific
Set size effect can be calculated using the decision integration model
based on SDT (Shaw)
1)
The internal representation of each stimulus is independent of set size
2)
The stimulus representation is noisy; both target and distracters -->
the more distracters in a display, the greater the chance that the
brightness of one will fall in the target range
Decision Integration Model based on SDT (Shaw)
3) The decision is determined by the stimulus representation that
yields the maximum likelihood (max rule) -- stimulus with the
maximum value on any given trial
4) Mean value of distracter’s representation is zero, and its variability is
1

The effect of increasing set size is to shift the distribution of the
maximum stimulus representation generated by the set of distracters
(determined by whichever distracter happens to generate the highest
value).
SDT assumes that the vertical distracters generate a smaller
response from the filters selective to the tilted target
Discriminating target from distractor:

both the mean separation between target and distractors and
the intrinsic variablity of these representations determine how
discriminable the target is from the distractors
 for a given orientation difference between target and distractor,
as distributions variance increases, discriminability decreases
Max rule
Easy search: tilt among vertical
Hard search: tilt (45) among tilted (22)
Response strength
 p (c) depends on the overlap of both distributions
 response to the 45 target is in the same location (~9); response to the tilted
distractor is shifted rightward (~4 to ~7)
The maximum rule
Set Size >1
 for finding a single target, a decision based on choosing the largest response
across the units is close to the best use of the available information, provided that
the responses for each of the units is independent
• noise interval (distracters only)
• signal interval (n-1 distracters & target)
• the observer looks for the largest value of the samples in each presentation and
then chooses the presentation interval that has the larger of the two maximum
values
 the greater the set size, the higher the probability that the maximum
emerges from the noise interval
Easy search
Hard search
Orientation X color conjunction - free viewing
Carrasco, Evert, Chang, & Katz ’95 (fig 1)
Orientation X color conjunction - fixed viewing
Carrasco, Evert, Chang & Katz ’95 (fig 5)
Carrasco, Evert, Chang, & Katz ’95 (fig 2)
Set size X Eccentricity
Carrasco, Evert, Chang, & Katz ’95 (fig 3)
Carrasco, Evert, Chang, & Katz ’95 (fig 4)
Carrasco & Frieder ’97 (fig 1)
RT (msec)
% ERROR
Carrasco & Frieder ’97 (fig 3)
Carrasco & Frieder ’97 (fig 4)
Carrasco & Frieder ’97 (fig 7)
Carrasco & Frieder ’97 (fig 8)
Carrasco & Yeshurun ’98 (fig 8)
Carrasco & Yeshurun ‘98 (fig 9)
Carrasco & Yeshurun ‘98 (fig 11)
Carrasco & Yeshurun ’98 (fig 12)