Theories of Visual Search
A standard theory of visual
search tasks assumes that when a person searches for a target in an array of
other items, memory is used in locating the target. The following analysis of
three articles shows that there is both strong support for this highly respected
theory and evidence that this theory may have some flaws in reasoning.
In the article \"Features and Objects in Visual Processing,\" Anne
Treisman states that there are two theoretical levels of visual processing. In
the first level of visual processing, certain components of visual information
are processed instantly and unconsciously. A person does not have to concentrate
on individual parts of the scene. This stage of processing is called the
preattentive stage. During the preattentive stage, the light received by the
visual receptors is translated in to the lines, curves, colors and textures of
the objects. Within the brain, there are two distinct visual areas that
specialize in different areas of processing. The first area processes lines,
curves, color, and texture and other areas deal with movement. After this
initial processing occurs, another area of the brain processes the more complex
and distinct qualities of a scene. Then, all of the components of the objects in
the scene are recombined into whole objects. In order to support the
preattentive stage of visual processing, Treisman devises that parts of objects
that belong to the same object share similar attributes. Attributes such as
color, continuity of lines and curves that define the boundaries between
objects. She uses an experiment to determine which properties of a visual
stimulus make its boundaries stand out from other similar objects. The
properties of an object that make it stand out are used by the visual processing
system in distinguishing the object from ground. In actuality, boundaries are
conspicuous between components that are distinctive in basic properties such as
color, brightness and line orientation but not in the way their properties are
connected or grouped. In an experiment showing evidence of this principle,
subjects are shown a picture in which a region of T’s easily distinguishes
itself from an area of slanted T’s but not from a region of backwards L’s that
are constructed of the same geons as the T’s. This illustrates that line
orientations are important features in the preattentive stage of visual
processing and that the specific configurations or conjunctions of lines are
not. Next, the subjects are shown an array of red O’s and blue V’s and red V’s
and blue O’s. A clear boundary between the O’s And V’s is apparent. This
indicates that basic shape properties, such as line curvature, are important. A
boundary between blue letters and red letters is noticeable. This condition
implies that color is important in creating boundaries. However, a boundary
between red V’s and blue O’s and red O’s and blue V’s is not conspicuous. Early
visual processing involves individual properties but not conjunctions of
properties. These results allude to the fact that analysis of parts and
properties occurs before they are recombined into whole objects. If these
components are noticed before they are joined with objects, they exist on their
own. This fact leads to the possibility of errors in synthesis.
One of
the experiments that Treisman uses to show the possibility of errors in
synthesis involves illusory conjunctions. In an illusory conjunction, properties
of two different objects appear to belong together. In this experiment, subjects
were shown a schematic that contained a dollar sign among a series of S’s and
slanted lines and a series pointed shapes that when conjoined with the lines,
form triangles. Then, the subjects were shown a similar schematic that did not
include a triangle. They reported the presence of dollar signs even though no
signs were present. They also saw dollar signs when they were shown a diagram
that contained only S’s and triangles. The results of this experiment propose
that early visual processing detects features regardless of their location.
Treisman conducted other experiments to prove the existence of early
visual processing. In one of these experiments, she used visual search tasks in
which subjects were asked to locate a target object that was surrounded by other
\"distractor\" objects. Her hypothesis was that if preattentive processing
happens automatically across the complete visual field, a target that is
strikingly different from the objects in its vicinity will stick out. If a
target differs from the distractors in a basic property such as orientation or
curvature, the target is located in about the same time in a series of a few
items as in a series of several items. Thus, the time it takes to find the
target is not influenced by the number of disractors.
Conversely, if a
target is distinguished only by a conjunction of properties or if it is
characterized by its specific combination of components, the time it takes to
find the target or rule out the presence of the target increases with the number
of distractors. Subjects involved in these experiments are forced to focus on
each item in the schematic in order to figure out how the properties are
conjoined. In a trial in which a target is not present, it takes subjects longer
to notice the lack of the target, because they have to analyze all of the
distractors. In subsequent experiments concerning visual search tasks, subjects
were presented a situation in which they had to locate a target distinguished by
its lack of a feature that is present in the other distractors(i.e. a diagram of
O’s and Q’s). In this case the discriminating factor between the O’s and Q’s is
that one object is a circle and the other is a circle intersected by a line
segment. Selected results from the experiment indicated a significant difference
in the search time based on whether the target was a Q or an O. When the target
was the contained the line, the search time was not related to the number of
distractors. The target was stood out from the distractors. In contrast, when
the target lacked the line, the search time increased linearly with the number
of distractors. The search time increased in this condition because the items
had to be scanned serially. The results of this experiment defend the theory
that a combined neural signal early in visual processing conveys the presence
but not the absence of a distinctive feature. Treisman theorizes that this
evidence illustrates possible feature maps. These experiments dealt with simple
properties of line orientation and curvature. A tilted line was more
distinguishable from vertical line distractors. A curved line target stood out
from straight line distractors. The converse targets did not stand out.
In the second section of the article, Treisman discusses how focused
attention is necessary for combining features in a given location of a scene and
for making structured representations of objects and their spatial relationships
with other objects. A piece of evidence that shows that conjunctions need
attention comes from experiments in which subjects identified a target in a
display and commented on its location. In one type of display, a basic feature
distinguishes the target from the distractors. In the remaining displays, the
target was different from the distractors in the way its features were attached.
Treisman and her colleagues hypothesized that subjects would be able to identify
a target based on an individual feature. However, they would incorrectly state
the location of the feature.
In the last part of the Treisman study, an
experiment was carried out to explore the role of prior knowledge in the
conjoining of properties. The study focused on illusory conjunctions. Results of
the study indicated that prior knowledge and expectations help a person to use
attention effectively to combine features, but prior knowledge does not
influence illusory exchanges of features to make abnormal objects appear normal.
In short, illusory conjunctions seem to arise at a stage of visual processing
that happens before conjunctions that do not make sense are corrected in the
late stages of visual processing.
At the end of this article, Anne
Treisman devises a model of visual processing. The visual system commences by
coding a certain number of simple and useful properties in a series of feature
maps. These maps preserve the spatial relationships of the properties. However,
spatial relationships are not readily available to the late states of visual
processing since this part of processing occurs automatically. Focused attention
is important in the late stages of visual processing. In this stage the location
of the objects is coded. Attention uses this information , simultaneously
selecting , via links to the individual feature maps, all the features that are
currently present in a specific location. These are stored in a temporary file.
The model now assumes that the integrated information about the properties and
structural relations in each object file is compared with objects that are
recognized due to prior knowledge. Conscious awareness depends on
representations that integrate information about particular objects from the
analyses of sensory features and from prior knowledge and constantly refresh the
information. When a significant discontinuity in space or time occurs, the
original file on an object is eliminated if it no longer is relevant to the
visual experience. The object then disappears and the visual system commences
processing of new visual stimuli.
An article that adds a new perspective
to how visual systems operate is titled \"Visual Search has no memory.\" The
research in this article was conducted by Todd. S. Horowitz and Jeremy Wolfe. A
series of experiments were run in which participants were asked to pick out the
letter T among a screen of L’s. In one version of the experiment, the
researchers relocated the letters several times a second, which was too rapid
for the visual system to follow. Remarkably, the subjects looking at scrambled
screen found the constantly \"drifting\" letter as quickly as those looking at
stationary screens. It seems that the brain does not waste any memory space
storing information in short-term visual tasks, responding instead to whatever
happens to be seen from moment to moment. Other aspects of memory compensate and
keep a constant image intact. In other words, people exhibit a kind of visual
amnesia when doing these kinds of searches. The visual system remains locked
into an ongoing present state of awareness, noting what it sees at any given
moment but retaining no record of what it looked at before or what it saw there.
The results of the experiment show that the visual system does not accumulate
information about object identity over time. These new findings somewhat
discredit a standard theory that states that observers use memory for locations
when searching for objects.
Most theories of visual search assume that
when observers search a complex display for a target, the display items are
scanned systematically and the same locations are not analyzed repeatedly. Arni
Kristjannson discusses the flaws in the research findings of Wolfe and Horowitz
in the \"In Search of Remembrance: Evidence for Memory in Visual Search.\"
Kristjannson conducts an experiments similar to those of Wolfe and Horowitz. In
the first experiment, there is a static condition and a relocation condition. In
the relocation condition, the target was moved to a position previously occupied
by a distractor several times per second. In the static condition the display
items remained in the same location in the series of frames shown. In both
conditions were randomly rotated in each successive frame. The purpose of using
these two conditions is to analyze whether location-based memory is used during
visual search since the only difference between the two conditions was that the
location of the target fluctuated with each successive frame in the relocation
condition. Results showed that when the target was reallocated, it took longer
for subjects to locate the target. These results strongly support the role of
memory in visual search tasks. Kristajannson points out some major divisions in
the results of this experiment and those of Horowitz and Wolfe. First, locations
were constantly reused in the present experiment. In Horowitz and Wolfe’s
experiment, relocated items appeared randomly anywhere within the display, often
appearing in locations that were previously lacked and item. Also, there were no
appearances of items in locations that were previously blank in the present
experiment, whereas in Wolfe’s experiment, items were continuously popping up in
blank locations. Lastly, the results of Horowitz and Wolfe’s experiment may not
even be related to the models of search involved in the present visual search
task.
There is a series of potential flaws in the experiments used in each
of the three articles to support theories of visual search. First, the subjects
may be overcome with boredom from the repetitiveness of the experiment and may
give an answer in order to complete the experiment as fast as possible. The
subjects may also become confused with the constant rotation of the targets in
some experiments and they may make up an answer. The methods of these experiment
do not take into the account the fact that some individuals perform better in
visual search tasks than others. Lastly, the physical, mental, and emotional
state of the subjects during the time of the experiment may influence the
results. For example, a person who is mentally stimulated during the experiment
may respond to stimuli at a faster rate that someone who is physically
exhausted. Also, individuals who perform well on visual search tasks are more
likely to volunteer to participate in visual task experiments. It is not known
whether or not these external factors have an effect on the results of these
experiments. However, the visual system seems to operate consistently regardless
of the intrusive factors.
Theories
of Visual Search
A standard theory of visual search tasks assumes that
when a person searches for a target in an array of other items, memory is used
in locating the target. The following analysis of three articles shows that
there is both strong support for this highly respected theory and evidence that
this theory may have some flaws in reasoning.
In the article \"Features
and Objects in Visual Processing,\" Anne Treisman states that there are two
theoretical levels of visual processing. In the first level of visual
processing, certain components of visual information are processed instantly and
unconsciously. A person does not have to concentrate on individual parts of the
scene. This stage of processing is called the preattentive stage. During the
preattentive stage, the light received by the visual receptors is translated in
to the lines, curves, colors and textures of the objects. Within the brain,
there are two distinct visual areas that specialize in different areas of
processing. The first area processes lines, curves, color, and texture and other
areas deal with movement. After this initial processing occurs, another area of
the brain processes the more complex and distinct qualities of a scene. Then,
all of the components of the objects in the scene are recombined into whole
objects. In order to support the preattentive stage of visual processing,
Treisman devises that parts of objects that belong to the same object share
similar attributes. Attributes such as color, continuity of lines and curves
that define the boundaries between objects. She uses an experiment to determine
which properties of a visual stimulus make its boundaries stand out from other
similar objects. The properties of an object that make it stand out are used by
the visual processing system in distinguishing the object from ground. In
actuality, boundaries are conspicuous between components that are distinctive in
basic properties such as color, brightness and line orientation but not in the
way their properties are connected or grouped. In an experiment showing evidence
of this principle, subjects are shown a picture in which a region of T’s easily
distinguishes itself from an area of slanted T’s but not from a region of
backwards L’s that are constructed of the same geons as the T’s. This
illustrates that line orientations are important features in the preattentive
stage of visual processing and that the specific configurations or conjunctions
of lines are not. Next, the subjects are shown an array of red O’s and blue V’s
and red V’s and blue O’s. A clear boundary between the O’s And V’s is apparent.
This indicates that basic shape properties, such as line curvature, are
important. A boundary between blue letters and red letters is noticeable. This
condition implies that color is important in creating boundaries. However, a
boundary between red V’s and blue O’s and red O’s and blue V’s is not
conspicuous. Early visual processing involves individual properties but not
conjunctions of properties. These results allude to the fact that analysis of
parts and properties occurs before they are recombined into whole objects. If
these components are noticed before they are joined with objects, they exist on
their own. This fact leads to the possibility of errors in synthesis.
One of the experiments that Treisman uses to show the possibility of
errors in synthesis involves illusory conjunctions. In an illusory conjunction,
properties of two different objects appear to belong together. In this
experiment, subjects were shown a schematic that contained a dollar sign among a
series of S’s and slanted lines and a series pointed shapes that when conjoined
with the lines, form triangles. Then, the subjects were shown a similar
schematic that did not include a triangle. They reported the presence of dollar
signs even though no signs were present. They also saw dollar signs when they
were shown a diagram that contained only S’s and triangles. The results of this
experiment propose that early visual processing detects features regardless of
their location.
Treisman conducted other experiments to prove the
existence of early visual processing. In one of these experiments, she used
visual search tasks in which subjects were asked to locate a target object that
was surrounded by other \"distractor\" objects. Her hypothesis was that if
preattentive processing happens automatically across the complete visual field,
a target that is strikingly different from the objects in its vicinity will
stick out. If a target differs from the distractors in a basic property such as
orientation or curvature, the target is located in about the same time in a
series of a few items as in a series of several items. Thus, the time it takes
to find the target is not influenced by the number of disractors.
Conversely, if a target is distinguished only by a conjunction of properties
or if it is characterized by its specific combination of components, the time it
takes to find the target or rule out the presence of the target increases with
the number of distractors. Subjects involved in these experiments are forced to
focus on each item in the schematic in order to figure out how the properties
are conjoined. In a trial in which a target is not present, it takes subjects
longer to notice the lack of the target, because they have to analyze all of the
distractors. In subsequent experiments concerning visual search tasks, subjects
were presented a situation in which they had to locate a target distinguished by
its lack of a feature that is present in the other distractors(i.e. a diagram of
O’s and Q’s). In this case the discriminating factor between the O’s and Q’s is
that one object is a circle and the other is a circle intersected by a line
segment. Selected results from the experiment indicated a significant difference
in the search time based on whether the target was a Q or an O. When the target
was the contained the line, the search time was not related to the number of
distractors. The target was stood out from the distractors. In contrast, when
the target lacked the line, the search time increased linearly with the number
of distractors. The search time increased in this condition because the items
had to be scanned serially. The results of this experiment defend the theory
that a combined neural signal early in visual processing conveys the presence
but not the absence of a distinctive feature. Treisman theorizes that this
evidence illustrates possible feature maps. These experiments dealt with simple
properties of line orientation and curvature. A tilted line was more
distinguishable from vertical line distractors. A curved line target stood out
from straight line distractors. The converse targets did not stand out.
In the second section of the article, Treisman discusses how focused
attention is necessary for combining features in a given location of a scene and
for making structured representations of objects and their spatial relationships
with other objects. A piece of evidence that shows that conjunctions need
attention comes from experiments in which subjects identified a target in a
display and commented on its location. In one type of display, a basic feature
distinguishes the target from the distractors. In the remaining displays, the
target was different from the distractors in the way its features were attached.
Treisman and her colleagues hypothesized that subjects would be able to identify
a target based on an individual feature. However, they would incorrectly state
the location of the feature.
In the last part of the Treisman study, an
experiment was carried out to explore the role of prior knowledge in the
conjoining of properties. The study focused on illusory conjunctions. Results of
the study indicated that prior knowledge and expectations help a person to use
attention effectively to combine features, but prior knowledge does not
influence illusory exchanges of features to make abnormal objects appear normal.
In short, illusory conjunctions seem to arise at a stage of visual processing
that happens before conjunctions that do not make sense are corrected in the
late stages of visual processing.
At the end of this article, Anne
Treisman devises a model of visual processing. The visual system commences by
coding a certain number of simple and useful properties in a series of feature
maps. These maps preserve the spatial relationships of the properties. However,
spatial relationships are not readily available to the late states of visual
processing since this part of processing occurs automatically. Focused attention
is important in the late stages of visual processing. In this stage the location
of the objects is coded. Attention uses this information , simultaneously
selecting , via links to the individual feature maps, all the features that are
currently present in a specific location. These are stored in a temporary file.
The model now assumes that the integrated information about the properties and
structural relations in each object file is compared with objects that are
recognized due to prior knowledge. Conscious awareness depends on
representations that integrate information about particular objects from the
analyses of sensory features and from prior knowledge and constantly refresh the
information. When a significant discontinuity in space or time occurs, the
original file on an object is eliminated if it no longer is relevant to the
visual experience. The object then disappears and the visual system commences
processing of new visual stimuli.
An article that adds a new perspective
to how visual systems operate is titled \"Visual Search has no memory.\" The
research in this article was conducted by Todd. S. Horowitz and Jeremy Wolfe. A
series of experiments were run in which participants were asked to pick out the
letter T among a screen of L’s. In one version of the experiment, the
researchers relocated the letters several times a second, which was too rapid
for the visual system to follow. Remarkably, the subjects looking at scrambled
screen found the constantly \"drifting\" letter as quickly as those looking at
stationary screens. It seems that the brain does not waste any memory space
storing information in short-term visual tasks, responding instead to whatever
happens to be seen from moment to moment. Other aspects of memory compensate and
keep a constant image intact. In other words, people exhibit a kind of visual
amnesia when doing these kinds of searches. The visual system remains locked
into an ongoing present state of awareness, noting what it sees at any given
moment but retaining no record of what it looked at before or what it saw there.
The results of the experiment show that the visual system does not accumulate
information about object identity over time. These new findings somewhat
discredit a standard theory that states that observers use memory for locations
when searching for objects.
Most theories of visual search assume that
when observers search a complex display for a target, the display items are
scanned systematically and the same locations are not analyzed repeatedly. Arni
Kristjannson discusses the flaws in the research findings of Wolfe and Horowitz
in the \"In Search of Remembrance: Evidence for Memory in Visual Search.\"
Kristjannson conducts an experiments similar to those of Wolfe and Horowitz. In
the first experiment, there is a static condition and a relocation condition. In
the relocation condition, the target was moved to a position previously occupied
by a distractor several times per second. In the static condition the display
items remained in the same location in the series of frames shown. In both
conditions were randomly rotated in each successive frame. The purpose of using
these two conditions is to analyze whether location-based memory is used during
visual search since the only difference between the two conditions was that the
location of the target fluctuated with each successive frame in the relocation
condition. Results showed that when the target was reallocated, it took longer
for subjects to locate the target. These results strongly support the role of
memory in visual search tasks. Kristajannson points out some major divisions in
the results of this experiment and those of Horowitz and Wolfe. First, locations
were constantly reused in the present experiment. In Horowitz and Wolfe’s
experiment, relocated items appeared randomly anywhere within the display, often
appearing in locations that were previously lacked and item. Also, there were no
appearances of items in locations that were previously blank in the present
experiment, whereas in Wolfe’s experiment, items were continuously popping up in
blank locations. Lastly, the results of Horowitz and Wolfe’s experiment may not
even be related to the models of search involved in the present visual search
task.
There is a series of potential flaws in the experiments used in each
of the three articles to support theories of visual search. First, the subjects
may be overcome with boredom from the repetitiveness of the experiment and may
give an answer in order to complete the experiment as fast as possible. The
subjects may also become confused with the constant rotation of the targets in
some experiments and they may make up an answer. The methods of these experiment
do not take into the account the fact that some individuals perform better in
visual search tasks than others. Lastly, the physical, mental, and emotional
state of the subjects during the time of the experiment may influence the
results. For example, a person who is mentally stimulated during the experiment
may respond to stimuli at a faster rate that someone who is physically
exhausted. Also, individuals who perform well on visual search tasks are more
likely to volunteer to participate in visual task experiments. It is not known
whether or not these external factors have an effect on the results of these
experiments. However, the visual system seems to operate consistently regardless
of the intrusive factors.
|
|