[ Pobierz całość w formacie PDF ]
Spatial Vision
, Vol. 17, No. 4-5, pp. 295 – 325 (2004)
VSP 2004.
Also available online -
Signal detection theory applied to three visual search tasks
— identification, yes/no detection and localization
E. LESLIE CAMERON
1
,
∗
, JOANNA C. TAI
1
, MIGUEL P. ECKSTEIN
2
and MARISA CARRASCO
1
1
Department of Psychology and Center for Neural Science, New York University,
6 Washington Place, 8th Floor, New York, NY 10003, USA
2
Department of Psychology, University of California, Santa Barbara, Santa Barbara,
CA 93106, USA
Received 4 June 2003; revised 10 January 2004; accepted 14 January 2004
Abstract
—Adding distracters to a display impairs performance on visual tasks (i.e. the set-size
effect). While keeping the display characteristics constant, we investigated this effect in three tasks:
2 target identification, yes-no detection with 2 targets, and 8-alternative localization. A Signal
Detection Theory (SDT) model, tailored for each task, accounts for the set-size effects observed in
identification and localization tasks, and slightly under-predicts the set-size effect in a detection task.
Given that sensitivity varies as a function of spatial frequency (SF), we measured performance in each
of these three tasks in neutral and peripheral precue conditions for each of six spatial frequencies
(0.5–12 cpd). For all spatial frequencies tested, performance on the three tasks decreased as set size
increased in the neutral precue condition, and the peripheral precue reduced the effect. Larger set-
size effects were observed at low SFs in the identification and localization tasks. This effect can
be described using the SDT model, but was not predicted by it. For each of these tasks we also
established the extent to which covert attention modulates performance across a range of set sizes.
A peripheral precue substantially diminished the set-size effect and improved performance, even at
set size 1. These results provide support for distracter exclusion, and suggest that signal enhancement
may also be a mechanism by which covert attention can impose its effect.
Keywords
: Visual search; signal detection theory; transient covert attention; yes/no detection;
discrimination; localization.
INTRODUCTION
The visual system is constantly confronted with more information than can be
efficiently processed at one time. To overcome this overload of information, visual
∗
Current address and to whom correspondence should be directed: Leslie Cameron, Department
of Psychology, Carthage College, 2001Alford Park Drive, Kenosha, WI 53140-1994, USA. E-mail:
lcameron@carthage.edu
296
E. L. Cameron
et al.
attention is the process by which one grants priority among sources of visual
information. In the study of visual attention, one of the central phenomena is the
set-
size effect
of visual search. The set-size effect is the decrease in accuracy or increase
in reaction time as a function of the number of distracters. This effect, traditionally
observed in conjunction but not feature searches, has been taken as evidence that
resources are limited and that attention is required and deployed serially for such
tasks (e.g. Treisman, 1993; Treisman and Gelade, 1980; Wolfe, 1994). This
explanation of the set-size effect has been challenged by many (e.g. Carrasco
and Yeshurun, 1998; Eckstein, 1998; McElree and Carrasco, 1999; Nakayama and
Joseph, 1998; Palmer
et al
., 2000; Verghese, 2001). For example, performance may
improve with smaller set sizes because of the removal of misleading information
from irrelevant distracters (distracter exclusion; e.g. Palmer
et al
. 1993).
Signal detection theory applied to visual search
A formal theory to calculate set-size effects and the attentional effects of distracter
exclusion on performance comes from signal detection theory (SDT) models that
take into consideration noise in the visual and decision systems. These models
attribute the set-size effect to the noisy quality of the sensory impressions, which
increases the risk of confusing the target with a distracter as the number of
distracters increases. The set-size effect, then, is the result of an increase in the
probability of the noise from one of the distracters exceeding that of the signal,
causing the observer to choose a distracter instead of the target. These models
account for the set-size effect in some feature and conjunction searches (Eckstein,
1998; Eckstein
et al
., 2000; Foley and Schwartz, 1998; Kinchla, 1974; Palmer,
1994; Palmer
et al
., 1993, 2000; Shaw 1982). Note that neither serial processing
nor limited-capacity attentional resources are required to explain the set-size effect.
On the other hand, when observers are cued to a location or element, these models
assume that the precue allows the observer’s attention to ignore noisy responses
arising from irrelevant distracters (distracter exclusion) and therefore improve
performance.
Typically, performance on visual search tasks has been assessed by using either
yes-no detection tasks with reaction time as the main or only dependent variable
(e.g. Egeth
et al
., 1984; Treisman, 1993; Treisman and Gormican, 1988; Wolfe,
1994) or discrimination tasks with accuracy as the dependent variable (Baldassi and
Burr, 2000; Morgan
et al
., 1998; Solomon and Morgan, 2001). This study examines
human performance on three different tasks (yes/no detection with 2-targets,
2-target identification (Baldassi and Verghese, 2002; Carrasco
et al
., 2000, 2003)
and 8-alternative localization) with accuracy as the primary dependent variable, and
applies an SDT model to search performance on the three different tasks.
Analysis of three visual search tasks
297
The role of sensory factors in visual search
Performance on visual search tasks can be affected by sensory factors. For example,
the set-size effect becomes more pronounced as eccentricity increases (Carrasco
et al
., 1995) and this effect is neutralized for features and substantially diminished
for conjunctions when stimulus visibility is equated across eccentricity (i.e. with
a cortical magnification factor; Carrasco and Frieder, 1997). The finding that
search performance for orientation, spatial frequency (SF) and color is closely
related to discrimination thresholds for the respective dimensions suggests that early
visual processes determine search performance (Verghese and Nakayama, 1994).
Likewise, stimulus content and spatial resolution predict search time in multiple
fixation searches for both features and conjunctions (Geisler and Chou, 1995).
In most visual search studies whose results have been interpreted in terms of
a serial deployment of attention, the display is presented until observers respond
(e.g. Treisman, 1993; Treisman and Gelade, 1980; Wolfe, 1994). Therefore, it is
impossible to discriminate between the effects of eye movements and those of covert
attention. To assess the effects of covert attention, it is necessary to present short
duration displays so as to prevent the possibility of eye movements while the display
is present (e.g. Carrasco
et al
., 1995, 2004; McElree and Carrasco, 1999). Here,
we used brief displays and placed the stimuli at a constant eccentricity, to equate
retinal and field eccentricity and to try to equate for discriminability (
cf
. Carrasco
and McElree, 2001; Eckstein, 1998), and with an inter-stimulus distance designed
to prevent masking and crowding effects (Bouma, 1970; Toet and Levi, 1992).
The role of attention in visual search
Search studies have shown that although performance in some feature searches is
impaired in the presence of distracters, directing attention to the target location
reduces this effect (e.g. Baldassi and Burr, 2000; Carrasco and McElree, 2001;
Carrasco and Yeshurun, 1998; Foley and Schwartz, 1998; Morgan
et al
., 1998;
Palmer, 1994). A ‘peripheral’ (exogenous, transient) precue, which appears
adjacent to the location of an upcoming target, has been shown to draw attention
effectively and to result in enhanced performance in detection, discrimination
and visual search tasks (e.g. Baldassi and Burr, 2000; Carrasco and McElree,
2001; Carrasco and Yeshurun, 1998; Carrasco
et al
., 2004; Kahneman
et al
., 1983;
Morgan
et al
., 1998; Nakayama and Mackeben, 1989).
Previous studies have shown that transient covert attention improves performance
on tasks that rely on the detection or discrimination of high spatial frequencies (e.g.
Balz and Hock, 1997; Nakayama and Mackeben, 1989; Yeshurun and Carrasco,
1999). However, relatively less is known about how transient covert attention
improves performance on tasks that examine a range of spatial frequencies (SFs).
Previous studies from our laboratory (Cameron
et al
., 2002; Carrasco
et al
., 2000,
2001) have shown that attention improves performance across the entire contrast
sensitivity function when a single target is presented without distracters. However,
298
E. L. Cameron
et al.
in the few visual search experiments in which SF has been manipulated, only one
or two spatial frequencies were tested (e.g. Baldassi and Burr, 2000; Carrasco
et al
., 1998; Foley and Schwartz, 1998; Morgan
et al
., 1998). Given that contrast
sensitivity varies across the range of SF that we perceive (Campbell and Robson,
1968) and that a search asymmetry has revealed that a low SF target is detected
among high SF distracters faster and more accurately than the opposite condition
(Carrasco
et al.
, 1998), here we investigated whether SF affects either the extent of
the set-size effect, the effect of peripheral precueing or their interaction.
Whereas most studies in the visual search literature have inferred the role of
attention (e.g. Treisman and Gelade, 1980; Treisman and Gormican, 1988; Wolfe,
1994, 2000; Wolfe
et al
., 1989), in this study we examined its role directly
by peripherally precueing the target location (
cf.
Carrasco and McElree, 2001;
Carrasco and Yeshurun, 1998; Carrasco
et al
., 2004). This is the first study to
examine the effect of a peripheral precue in three different visual search tasks:
2-target yes/no detection, 2-target identification and 8-alternative localization.
MODELING
Signal detection models for search tasks
Signal detection based models are often used to calculate the performance degrada-
tion with increasing set size or performance enhancement with decreasing set size,
expected from the effect of independent noise associated with the observers’ inter-
nal response to each relevant item in the visual search task. The models can be
used as a benchmark to test whether set-size effects can be accounted for without
resorting to limited attentional resources. These types of models have been success-
fully used to predict human performance for detection and localization of targets in
a variety of conditions (Burgess and Ghandeharian, 1984; Eckstein and Whiting,
1996; Foley and Schwartz, 1998; Green and Swets, 1966; Palmer
et al
., 1993; Shaw,
1980; Solomon
et al
., 1997; Sperling and Dosher, 1986; Swensson and Judy, 1981).
For complex tasks and those involving memory, human performance often degrades
beyond what is expected from the independent noise prediction (Palmer
et al
., 1993;
Shaw, 1980).
Single filter models for single target tasks
In most simple search tasks a single target may appear among distracters (all
distracters being identical). In the typical yes/no task, a single array of elements
is presented to the observer on each trial and the target has a 50% probability of
being present. The observer’s task is to decide whether the signal was present. In
the 2-interval forced choice task (2 IFC), two arrays of elements are presented to
the observer sequentially through time. One of the two intervals consists of a target
and distracters while the other interval consists of only distracters. The observer’s
task is to decide which interval contained the target.
Analysis of three visual search tasks
299
For these tasks, the standard SDT model assumes that the observer monitors a
single filter or detector tuned to the target (or tuned to discriminate maximally
between the target and distracter). The filter elicits a higher response to the target
than the distracter; however, its response is stochastic due to internal noise (e.g.
neural noise and/or decision noise).
In a yes/no task the model responds ‘target present’ if the internal response
exceeds a decision threshold or criterion. The proportion of trials in which the
model correctly responds ‘target present’ in a target present trial (hit rate,
h)
is
calculated by computing the probability of the target or any of the distracters in the
display exceeding the criterion. Similarly, the false alarm rate (proportion of trials
in which the model incorrectly responds ‘target present’ in a target absent trial) is
calculated by computing the probability of any one of the distracters exceeding the
criterion (see Table 1, yes/no detection of orientation with one oriented target among
vertical distracters).
In a 2-interval forced choice task, the model chooses the interval associated with
the highest (maximum) response. Performance (proportion of correct selection of
the target present interval) is calculated by computing the probability of the filter’s
response to the target taking the maximum value (see Table 1, 2-interval forced
choice detection).
Two filter models for two-target tasks
In this study, as in some previous work (Baldassi and Burr, 2000; Baldassi
and Verghese, 2002; Carrasco and McElree, 2001, Carrasco
et al
., 2003, 2004;
Morgan
et al
., 1998), the search task involves one of two possible targets appearing
among a single type of distracter. The observer’s task is to decide which of two
targets has appeared (identification task), whether either of the two targets has
appeared (detection task), or to identify the spatial location of either of the targets
(localization task).
Given that two types of targets can appear, it is reasonable to assume that the
observer must monitor two filters, one tuned to each of the possible targets (or
alternatively each filter tuned to discriminate between each of the targets and the
distracters; Carrasco
et al
., 2000). The model performance is then determined by
the response of two filters to each element in the display rather than a single filter’s
response. In the following sections we develop two-filter SDT models for each of
the tasks: (a) 2-target identification; (b) 2-target yes/no detection; (c) 8-alternative
localization. The mathematical expressions for each of the tasks are summarized
in Table 1. In addition, for comparison, Table 1 also includes the traditional SDT
expressions corresponding to tasks in which there is only one type of target and
the observer is assumed to monitor the response of a single perceptual filter to the
elements in the display. Related treatments of the 2-target identification task taking
into account the two possible targets have been previously developed in Carrasco
et al
. (2000) and Baldassi and Verghese (2002) (see Note 1). In the neutral condition
where the precue did not give any information about target location, we set
n
to the
[ Pobierz całość w formacie PDF ]
zanotowane.pl doc.pisz.pl pdf.pisz.pl adbuxwork.keep.pl
Spatial Vision
, Vol. 17, No. 4-5, pp. 295 – 325 (2004)
VSP 2004.
Also available online -
Signal detection theory applied to three visual search tasks
— identification, yes/no detection and localization
E. LESLIE CAMERON
1
,
∗
, JOANNA C. TAI
1
, MIGUEL P. ECKSTEIN
2
and MARISA CARRASCO
1
1
Department of Psychology and Center for Neural Science, New York University,
6 Washington Place, 8th Floor, New York, NY 10003, USA
2
Department of Psychology, University of California, Santa Barbara, Santa Barbara,
CA 93106, USA
Received 4 June 2003; revised 10 January 2004; accepted 14 January 2004
Abstract
—Adding distracters to a display impairs performance on visual tasks (i.e. the set-size
effect). While keeping the display characteristics constant, we investigated this effect in three tasks:
2 target identification, yes-no detection with 2 targets, and 8-alternative localization. A Signal
Detection Theory (SDT) model, tailored for each task, accounts for the set-size effects observed in
identification and localization tasks, and slightly under-predicts the set-size effect in a detection task.
Given that sensitivity varies as a function of spatial frequency (SF), we measured performance in each
of these three tasks in neutral and peripheral precue conditions for each of six spatial frequencies
(0.5–12 cpd). For all spatial frequencies tested, performance on the three tasks decreased as set size
increased in the neutral precue condition, and the peripheral precue reduced the effect. Larger set-
size effects were observed at low SFs in the identification and localization tasks. This effect can
be described using the SDT model, but was not predicted by it. For each of these tasks we also
established the extent to which covert attention modulates performance across a range of set sizes.
A peripheral precue substantially diminished the set-size effect and improved performance, even at
set size 1. These results provide support for distracter exclusion, and suggest that signal enhancement
may also be a mechanism by which covert attention can impose its effect.
Keywords
: Visual search; signal detection theory; transient covert attention; yes/no detection;
discrimination; localization.
INTRODUCTION
The visual system is constantly confronted with more information than can be
efficiently processed at one time. To overcome this overload of information, visual
∗
Current address and to whom correspondence should be directed: Leslie Cameron, Department
of Psychology, Carthage College, 2001Alford Park Drive, Kenosha, WI 53140-1994, USA. E-mail:
lcameron@carthage.edu
296
E. L. Cameron
et al.
attention is the process by which one grants priority among sources of visual
information. In the study of visual attention, one of the central phenomena is the
set-
size effect
of visual search. The set-size effect is the decrease in accuracy or increase
in reaction time as a function of the number of distracters. This effect, traditionally
observed in conjunction but not feature searches, has been taken as evidence that
resources are limited and that attention is required and deployed serially for such
tasks (e.g. Treisman, 1993; Treisman and Gelade, 1980; Wolfe, 1994). This
explanation of the set-size effect has been challenged by many (e.g. Carrasco
and Yeshurun, 1998; Eckstein, 1998; McElree and Carrasco, 1999; Nakayama and
Joseph, 1998; Palmer
et al
., 2000; Verghese, 2001). For example, performance may
improve with smaller set sizes because of the removal of misleading information
from irrelevant distracters (distracter exclusion; e.g. Palmer
et al
. 1993).
Signal detection theory applied to visual search
A formal theory to calculate set-size effects and the attentional effects of distracter
exclusion on performance comes from signal detection theory (SDT) models that
take into consideration noise in the visual and decision systems. These models
attribute the set-size effect to the noisy quality of the sensory impressions, which
increases the risk of confusing the target with a distracter as the number of
distracters increases. The set-size effect, then, is the result of an increase in the
probability of the noise from one of the distracters exceeding that of the signal,
causing the observer to choose a distracter instead of the target. These models
account for the set-size effect in some feature and conjunction searches (Eckstein,
1998; Eckstein
et al
., 2000; Foley and Schwartz, 1998; Kinchla, 1974; Palmer,
1994; Palmer
et al
., 1993, 2000; Shaw 1982). Note that neither serial processing
nor limited-capacity attentional resources are required to explain the set-size effect.
On the other hand, when observers are cued to a location or element, these models
assume that the precue allows the observer’s attention to ignore noisy responses
arising from irrelevant distracters (distracter exclusion) and therefore improve
performance.
Typically, performance on visual search tasks has been assessed by using either
yes-no detection tasks with reaction time as the main or only dependent variable
(e.g. Egeth
et al
., 1984; Treisman, 1993; Treisman and Gormican, 1988; Wolfe,
1994) or discrimination tasks with accuracy as the dependent variable (Baldassi and
Burr, 2000; Morgan
et al
., 1998; Solomon and Morgan, 2001). This study examines
human performance on three different tasks (yes/no detection with 2-targets,
2-target identification (Baldassi and Verghese, 2002; Carrasco
et al
., 2000, 2003)
and 8-alternative localization) with accuracy as the primary dependent variable, and
applies an SDT model to search performance on the three different tasks.
Analysis of three visual search tasks
297
The role of sensory factors in visual search
Performance on visual search tasks can be affected by sensory factors. For example,
the set-size effect becomes more pronounced as eccentricity increases (Carrasco
et al
., 1995) and this effect is neutralized for features and substantially diminished
for conjunctions when stimulus visibility is equated across eccentricity (i.e. with
a cortical magnification factor; Carrasco and Frieder, 1997). The finding that
search performance for orientation, spatial frequency (SF) and color is closely
related to discrimination thresholds for the respective dimensions suggests that early
visual processes determine search performance (Verghese and Nakayama, 1994).
Likewise, stimulus content and spatial resolution predict search time in multiple
fixation searches for both features and conjunctions (Geisler and Chou, 1995).
In most visual search studies whose results have been interpreted in terms of
a serial deployment of attention, the display is presented until observers respond
(e.g. Treisman, 1993; Treisman and Gelade, 1980; Wolfe, 1994). Therefore, it is
impossible to discriminate between the effects of eye movements and those of covert
attention. To assess the effects of covert attention, it is necessary to present short
duration displays so as to prevent the possibility of eye movements while the display
is present (e.g. Carrasco
et al
., 1995, 2004; McElree and Carrasco, 1999). Here,
we used brief displays and placed the stimuli at a constant eccentricity, to equate
retinal and field eccentricity and to try to equate for discriminability (
cf
. Carrasco
and McElree, 2001; Eckstein, 1998), and with an inter-stimulus distance designed
to prevent masking and crowding effects (Bouma, 1970; Toet and Levi, 1992).
The role of attention in visual search
Search studies have shown that although performance in some feature searches is
impaired in the presence of distracters, directing attention to the target location
reduces this effect (e.g. Baldassi and Burr, 2000; Carrasco and McElree, 2001;
Carrasco and Yeshurun, 1998; Foley and Schwartz, 1998; Morgan
et al
., 1998;
Palmer, 1994). A ‘peripheral’ (exogenous, transient) precue, which appears
adjacent to the location of an upcoming target, has been shown to draw attention
effectively and to result in enhanced performance in detection, discrimination
and visual search tasks (e.g. Baldassi and Burr, 2000; Carrasco and McElree,
2001; Carrasco and Yeshurun, 1998; Carrasco
et al
., 2004; Kahneman
et al
., 1983;
Morgan
et al
., 1998; Nakayama and Mackeben, 1989).
Previous studies have shown that transient covert attention improves performance
on tasks that rely on the detection or discrimination of high spatial frequencies (e.g.
Balz and Hock, 1997; Nakayama and Mackeben, 1989; Yeshurun and Carrasco,
1999). However, relatively less is known about how transient covert attention
improves performance on tasks that examine a range of spatial frequencies (SFs).
Previous studies from our laboratory (Cameron
et al
., 2002; Carrasco
et al
., 2000,
2001) have shown that attention improves performance across the entire contrast
sensitivity function when a single target is presented without distracters. However,
298
E. L. Cameron
et al.
in the few visual search experiments in which SF has been manipulated, only one
or two spatial frequencies were tested (e.g. Baldassi and Burr, 2000; Carrasco
et al
., 1998; Foley and Schwartz, 1998; Morgan
et al
., 1998). Given that contrast
sensitivity varies across the range of SF that we perceive (Campbell and Robson,
1968) and that a search asymmetry has revealed that a low SF target is detected
among high SF distracters faster and more accurately than the opposite condition
(Carrasco
et al.
, 1998), here we investigated whether SF affects either the extent of
the set-size effect, the effect of peripheral precueing or their interaction.
Whereas most studies in the visual search literature have inferred the role of
attention (e.g. Treisman and Gelade, 1980; Treisman and Gormican, 1988; Wolfe,
1994, 2000; Wolfe
et al
., 1989), in this study we examined its role directly
by peripherally precueing the target location (
cf.
Carrasco and McElree, 2001;
Carrasco and Yeshurun, 1998; Carrasco
et al
., 2004). This is the first study to
examine the effect of a peripheral precue in three different visual search tasks:
2-target yes/no detection, 2-target identification and 8-alternative localization.
MODELING
Signal detection models for search tasks
Signal detection based models are often used to calculate the performance degrada-
tion with increasing set size or performance enhancement with decreasing set size,
expected from the effect of independent noise associated with the observers’ inter-
nal response to each relevant item in the visual search task. The models can be
used as a benchmark to test whether set-size effects can be accounted for without
resorting to limited attentional resources. These types of models have been success-
fully used to predict human performance for detection and localization of targets in
a variety of conditions (Burgess and Ghandeharian, 1984; Eckstein and Whiting,
1996; Foley and Schwartz, 1998; Green and Swets, 1966; Palmer
et al
., 1993; Shaw,
1980; Solomon
et al
., 1997; Sperling and Dosher, 1986; Swensson and Judy, 1981).
For complex tasks and those involving memory, human performance often degrades
beyond what is expected from the independent noise prediction (Palmer
et al
., 1993;
Shaw, 1980).
Single filter models for single target tasks
In most simple search tasks a single target may appear among distracters (all
distracters being identical). In the typical yes/no task, a single array of elements
is presented to the observer on each trial and the target has a 50% probability of
being present. The observer’s task is to decide whether the signal was present. In
the 2-interval forced choice task (2 IFC), two arrays of elements are presented to
the observer sequentially through time. One of the two intervals consists of a target
and distracters while the other interval consists of only distracters. The observer’s
task is to decide which interval contained the target.
Analysis of three visual search tasks
299
For these tasks, the standard SDT model assumes that the observer monitors a
single filter or detector tuned to the target (or tuned to discriminate maximally
between the target and distracter). The filter elicits a higher response to the target
than the distracter; however, its response is stochastic due to internal noise (e.g.
neural noise and/or decision noise).
In a yes/no task the model responds ‘target present’ if the internal response
exceeds a decision threshold or criterion. The proportion of trials in which the
model correctly responds ‘target present’ in a target present trial (hit rate,
h)
is
calculated by computing the probability of the target or any of the distracters in the
display exceeding the criterion. Similarly, the false alarm rate (proportion of trials
in which the model incorrectly responds ‘target present’ in a target absent trial) is
calculated by computing the probability of any one of the distracters exceeding the
criterion (see Table 1, yes/no detection of orientation with one oriented target among
vertical distracters).
In a 2-interval forced choice task, the model chooses the interval associated with
the highest (maximum) response. Performance (proportion of correct selection of
the target present interval) is calculated by computing the probability of the filter’s
response to the target taking the maximum value (see Table 1, 2-interval forced
choice detection).
Two filter models for two-target tasks
In this study, as in some previous work (Baldassi and Burr, 2000; Baldassi
and Verghese, 2002; Carrasco and McElree, 2001, Carrasco
et al
., 2003, 2004;
Morgan
et al
., 1998), the search task involves one of two possible targets appearing
among a single type of distracter. The observer’s task is to decide which of two
targets has appeared (identification task), whether either of the two targets has
appeared (detection task), or to identify the spatial location of either of the targets
(localization task).
Given that two types of targets can appear, it is reasonable to assume that the
observer must monitor two filters, one tuned to each of the possible targets (or
alternatively each filter tuned to discriminate between each of the targets and the
distracters; Carrasco
et al
., 2000). The model performance is then determined by
the response of two filters to each element in the display rather than a single filter’s
response. In the following sections we develop two-filter SDT models for each of
the tasks: (a) 2-target identification; (b) 2-target yes/no detection; (c) 8-alternative
localization. The mathematical expressions for each of the tasks are summarized
in Table 1. In addition, for comparison, Table 1 also includes the traditional SDT
expressions corresponding to tasks in which there is only one type of target and
the observer is assumed to monitor the response of a single perceptual filter to the
elements in the display. Related treatments of the 2-target identification task taking
into account the two possible targets have been previously developed in Carrasco
et al
. (2000) and Baldassi and Verghese (2002) (see Note 1). In the neutral condition
where the precue did not give any information about target location, we set
n
to the
[ Pobierz całość w formacie PDF ]