visual identification and spatial location in alzheimer’s disease
TRANSCRIPT
Brain and Cognition 52 (2003) 155–166
www.elsevier.com/locate/b&c
Visual identification and spatial location in Alzheimer�s disease
Annamarie Stehli Nguyen,* Charles Chubb, and F. Jacob Huff
Departments of Cognitive Sciences (AS and CC) and Neurology (FJH), Institute for Brain Aging and Dementia, University of California,
Irvine, 19722 MacArthur Blvd., Suite 150, Irvine, CA 92612, USA
Accepted 7 October 2002
Abstract
Twelve patients with Alzheimer�s disease (AD) and 15 healthy elderly control subjects were shown sets of luminance-defined
letters, texture-defined letters, luminance-defined squares, and texture-defined squares. They were asked to name the letters or point
to the target square on each page. The stimuli were graded into four levels of difficulty based on the amount of contrast between the
figure and the background. Performance was measured in terms of the maximum level of difficulty at which the participant correctly
identified or located the three figures. Contrary to expectations, no significant difference was found between the performance of AD
patients and control subjects on texture discrimination tasks vs. luminance discrimination tasks. However, results indicate that AD
patients are impaired in performing a task requiring them to locate a texture-defined target of known shape in a noisy background
field. By contrast, AD patients show no significant deficit in a task requiring them to locate a texture-defined shape in a known
location. This argues that the observed deficit in the location task is not due to a failure in the system that discriminates target
texture from background texture (since both location and identification tasks require the same textural discriminations), but rather
to an impairment of the system responsible for ‘‘finding things’’ (i.e., locating known targets at unknown locations). This obser-
vation suggests that AD patients may suffer selective damage to the dorsal ‘‘Where’’ pathway, which is responsible for localizing
objects in space.
� 2003 Elsevier Science (USA). All rights reserved.
Keywords: Vision; Alzheimer�s disease; Spatial location
1. Introduction
Visual impairments continue to be widely overlooked
as major symptoms of Alzheimer�s disease (AD) despitethe fact that in recent years both behavioral and neu-
ropathological studies have demonstrated that visual
processes are not spared in patients with the disease
(Cronin-Golomb, 1995). Mendez, Mendez, Martin,
Smyth, and Whitehouse (1990) asserts that two major
reasons for such a dismissal are: (1) inconsistenciesamong AD vision studies and (2) that patients with AD
typically have normal visual acuity and thus do not
complain of visual problems explicitly. Cronin-Golomb
argues, however, that visual problems can in fact be
found in the majority of AD patients. Therefore, she
maintains that assessments of their visual ability should
be considered useful in terms of providing new diag-
* Corresponding author. Fax: +949-824-1811.
E-mail address: [email protected] (A. Stehli Nguyen).
0278-2626/03/$ - see front matter � 2003 Elsevier Science (USA). All rights
doi:10.1016/S0278-2626(03)00031-9
nostic information and as a source of insight into the
dynamics of the disease�s more renowned cognitive andfunctional symptoms.
Perhaps the strongest evidence that visual dysfunc-
tion is an important issue for Alzheimer�s research are
case studies such as the one presented by Levine, Lee,
and Fisher (1993), which suggests the existence of an
actual variant of the disease characterized chiefly by
visual impairments. The patient exhibited extensive vi-
sual dysfunction including losses of peripheral vision,spatial skills, and visuomotor ability. These underlying
deficits were manifested in many forms including im-
paired eye–hand coordination, disorientation, mis-
matching of clothes, inappropriate object selection, and
extreme driving problems (Levine et al., 1993). Consis-
tent with Cronin-Golomb�s (1995) assertion, the patienthad normal visual acuity for the first several years,
demonstrated little awareness as to the extent of his vi-sual problems, and tended to dismiss his problems with
‘‘every-day tasks’’ as being a result of fatigue, other
reserved.
156 A. Stehli Nguyen et al. / Brain and Cognition 52 (2003) 155–166
people�s mistakes, or bad environmental conditions(Levine et al., 1993).
It was later clinically established that for the most
part the patient�s trouble with reading, calculating
numbers, disorientation, and several other functions,
was a result of the visual impairment as opposed to
being memory or language based. Although memory
and language disabilities became evident as the disease
progressed, the degeneration of this patient�s visualabilities continued to be the dominant form of impair-
ment (Levine et al., 1993). The results of the post-mor-
tem examination of this patient confirmed that he had
been correctly diagnosed with Alzheimer�s Dementia.
Interestingly, extensive atrophy and the highest density
of neurofibrillary tangles and senile plaques were found
in the occipitoparietal region which is considered to be
very important for higher-order visual processing (Le-vine, et al.). Conversely, the hippocampus and amyg-
dala, which are typically the main sites of atrophy (and
known to be crucial areas for memory function), were
only mildly deteriorated (Levine et al.). This pattern of
degeneration strongly confirmed the clinical impression
that much of the patient�s exhibited cognitive and be-
havioral deficiencies could be attributed to his visual
impairments.Although the discussed case study represents an ex-
treme case of visual dysfunction as a result of AD,
current research has shown that similar deficits can be
found in most ‘‘typical’’ Alzheimer�s patients as well.
For example, in summarizing her work, Cronin-Golomb
(1995) reports finding selective losses of visual capacities
such as contrast sensitivity, color perception, and ste-
reoacuity in AD patients which can be differentiatedfrom losses associated with normal aging or such other
age-related conditions as Parkinson�s disease. Similarly,Mendez et al. (1990) found that all of their AD subjects
demonstrated impairments in basic figure-ground tasks
and that high percentages of the patients also performed
significantly worse than normals on other complex vi-
sual tasks such as recognition of common objects, color
recognition, spatial localization, and visual synthesis. Inaddition, Katz and Rimmer (1989) reported that brain
studies of Alzheimer�s patients have revealed high
numbers of neurofibrillary tangles and neuritic plaques
in the visual association and higher-order visual asso-
ciation areas of cortex even in patients whose primary
deficits are not visual.
Besides seeking evidence supporting the claim that
most Alzheimer�s patients suffer from visual impairment,Cronin-Golomb (1995) has been interested in evaluating
the extent to which the difficulties exhibited by AD pa-
tients on cognitive tasks and in daily life, can be at-
tributed to such visual disturbances. In a direct
approach to this issue, an analysis was run in which
performance on the visual tasks (e.g., color discrimina-
tion, contrast sensitivity, and motion detection), were
independent variables with the dependent variables be-ing the performance on the cognitive tasks (e.g., Boston
Naming Test and Luria Mental rotation) (Cronin-Go-
lomb). It was found that some of the vision tasks ac-
counted for up to 50% of the variance in cognitive test
performance with some vision-cognitive test combina-
tions demonstrating the predictive ability of the vision
test to be better than that of dementia severity (Cronin-
Golomb). Although vision performance did not out-shine dementia severity as a predictive source in all
cases, these findings support the usefulness of visual
tasks as diagnostic tools and the idea that at least some
of the cognitive deficits exhibited by Alzheimer�s pa-
tients reflect visual-perceptual problems.
Because of its importance for success in everyday
tasks, spatial vision has been the focus of much Alz-
heimer�s research. One specific capacity that has receiveda great deal of attention from researchers is spatial
contrast sensitivity, also known as contrast sensitivity. It
is thought that measurement of contrast sensitivity
provides a more complete evaluation of visual dys-
function than do tests of visual acuity. One reason this
has been proposed is that adequate spatial vision re-
quires the ability to process both coarse and fine spatial
patterns. Standard visual acuity tests (e.g., Snellen acu-ity) fail to isolate these dependencies of visual perfor-
mance on spatial scale. In addition, it has been shown
that losses in the ability to perceive spatial contrast can
exist despite the maintenance of normal visual acuity,
suggesting that contrast sensitivity may be a more sen-
sitive measure of visual function (Katz & Rimmer, 1989;
Nissen et al., 1985).
The traditional way of assessing contrast sensitivity isthrough the use of test stimuli comprising sinusoidal
gratings of light and dark bars that vary in contrast
(Nissen et al., 1985). Spatial frequency is thus defined as
the number of pairs or cycles of light and dark bars per
degree of visual angle, with high frequencies corre-
sponding to fine patterns and low frequencies corre-
sponding to coarse patterns. In Nissen et al.�s task,
subjects were presented with stimuli of five differentspatial frequencies and were asked for each one whether
or not they saw stripes. The assessment of contrast
sensitivity was calculated by considering the mean con-
trast of stimuli for which the subjects answered yes and
that for which they answered no. The finding was that
the Alzheimer�s patients exhibited lower sensitivity to allfrequencies.
Cronin-Golomb et al. (1991) used both a computer-ized forced-choice method and a chart format to assess
contrast sensitivity. Results for both methods showed
Alzheimer�s patients to be less sensitive to contrast than
elderly controls at all frequencies with more patients
demonstrating impairment at low frequencies as op-
posed to higher frequencies. This was true for patients at
all severity levels (Cronin-Golomb et al.).
A. Stehli Nguyen et al. / Brain and Cognition 52 (2003) 155–166 157
Although contradictory results in terms of contrastsensitivity have been presented by other authors, both
Cronin-Golomb (1995) and Katz and Rimmer (1989)
attribute such inconsistencies to crucial methodological
differences such as subject selection and task type. Katz
and Rimmer suggested that the exhibition of decreased
contrast sensitivity by Alzheimer�s patients seems likelybecause of its consistency with findings of reduced ax-
onal diameter in large-diameter ganglion cells (magno-cellular cells) that have proven to be essential for
transmitting contrast sensitivity information.
Another capacity of spatial vision that is thought to
be crucial for successful object recognition and orien-
tational stability is that of figure-ground or spatial dis-
crimination (Carlson, 1994). As previously mentioned,
Mendez et al. (1990), found figure-ground tasks to be
especially troublesome for Alzheimer�s subjects. Suchtasks involve the ability to differentiate the object from
the background (in terms of their retinal images) based
on some perceptual characteristic (Regan & Hong,
1994). As Regan and Hong asserted, the stimuli most
often used in these tasks are those for which the contrast
between figure and ground is defined by their respective
luminance levels. However, spatial discrimination can
also be assessed by manipulating several other percep-tual characteristics. An example is the discrimination of
numbers based on color properties that is required for
success on the standard Ishihara Color Plates primarily
used to diagnose color blindness (in McCleary, Shankle,
Mulnard, & Dick, 1996).
McCleary et al. (1996) used Ishihara plates to eval-
uate figure-ground processing in dementia patients. It
was found that although prior research has shown thatAD does not disrupt red–green vision, the non-color
blind Alzheimer�s subjects made significantly more er-
rors on red–green plates than normals. This suggested
that the processing deficits were due to form discrimi-
nation problems and not problems perceiving color per
se (McCleary et al.). In addition to correlating highly
with dementia severity, performance on the number
naming task was significantly worse for Alzheimer�spatients than for subjects with other forms of dementia.
McCleary et. al. interpreted these findings as indicating
that brain structures involved in the recognition of
common but visually complex objects (like the numbers
in the task) are specifically affected by AD. These suc-
cessful results combined with the simplicity of the task
suggests that such stimuli may provide a useful diag-
nostic tool for AD.As a kind of extension of McCleary et al.�s (1996)
work, the present research introduced a task much like
the number-naming Ishihara task but using different
figure-ground stimuli. Specifically, the present task was
designed to assess the ability of AD patients and
age-matched normal controls to recognize both lumi-
nance-defined and texture-defined letters. Regan and his
colleagues have conducted several experiments involvingtexture-defined letters proving such stimuli to be ade-
quate for figure discrimination tasks (e.g., Regan &
Hong, 1994; Regan & Simpson, 1995).
Many current models of visual processing draw a
crucial distinction between first-order vs. second-order
stimuli (e.g., Arsenault, Wilkinson, & Kingdom, 1999;
Bergen & Landy, 1991; Graham, 1994; Lu & Sperling,
1996; Sutter & Graham, 1995). Sinusoidal gratings (suchas are used in testing contrast sensitivity) are examples
of first-order stimuli. The detection and/or classification
of such stimuli requires the visual system to perform an
accounting of the relative average light levels in different
regions of the visual field. To detect a sinusoidal grating,
for example, the observer must be able to sense that
more light emanates from the bright bars than from the
dark bars. Such computations are basic to all of spatialvision. Indeed V1 simple cells, the first functional level in
the cortical hierarchy of visual processing, are devoted
specifically to the task of sensing such patterns of lu-
minance in the visual field.
However, the visual system is able to draw many
distinctions (to detect boundaries, gradients and other
sorts of qualitative modulations of the stimulus field)
that are based not on differences in average luminance,but on other second-order properties. Consider, for ex-
ample, Fig. 1c. This figure shows one of the second-order
stimuli used in the experiment reported here. Prior to any
conscious effort on the part of the observer, the visual
system extracts the shape of the texture-defined square
from the field of background texture. Note, however,
that there is no difference in average luminance between
the square and the background. To detect the square, thevisual system must detect that the pattern of luminance
within the square is systematically different from that of
the background. This task requires that (i) first-order
sensors be used to discriminate individual black bars
from the background (based on luminance differences),
and then (ii) higher level neurons process the output
from these first-order sensors in order to detect the
presence and discriminate the shape of the texture-de-fined square. Although it is by no means clear that there
exists a unique cortical region responsible for all second-
order processing, V4 has been implicated in some sorts
of second-order judgments (e.g., Carlson, 1994).
Because texture-based discrimination logically re-
quires processing beyond that required for luminance-
based discrimination (Huxlin &Merigan, 1998; Merigan,
1996), we conjectured texture-based discriminationmight be more vulnerable to effects of AD than analo-
gous discriminations featuring luminance-defined stim-
uli. If this were true, then texture-defined stimuli (and
other second-order stimuli), might provide a diagnostic
tool enabling earlier detection of AD than tests using
first-order stimuli. In addition, because each of the
processes is proposed to take place in different areas of
Fig. 1. (a) Hundred percentage contrast version of luminance-defined (LD) letter F; (b) 80% contrast version of luminance-defined (LD) letter F; (c)
100% contrast version of texture-defined orientation (TDo) letter F; (d) 80% contrast version of texture-defined orientation (TDo) letter F; (e) 100%
contrast version of texture-defined granularity (TDg) letter F; (f) 80% contrast version of texture-defined granularity (TDg) letter F.
158 A. Stehli Nguyen et al. / Brain and Cognition 52 (2003) 155–166
visual cortex, patterns of performance by Alzheimer�spatients could provide important insight as to the cor-
tical progression of the disease.
2. Materials
2.1. General features of stimuli
All stimuli were printed onto standard 8.5 in.� 11 in.
(21.59 cm� 27.94 cm) pages. The entire stimulus field
was a square with one side equaling 7 in. (17.78 cm) inlength. Stimuli were composed of textures generated by
painting different sorts of micropatterns into a regular
grid of small square regions, called texels (for texture
elements). For each stimulus (in all conditions), the
density of the texels remained the same. There were 48
texels per side of the square stimulus field giving a
density of about 49 texels per square inch (about 9 per
square centimeter). The stimuli were presented to theparticipants at arms length (about 30 in.; 76.20 cm). At
this viewing distance the entire display subtended 13.30�
A. Stehli Nguyen et al. / Brain and Cognition 52 (2003) 155–166 159
of visual angle in width, with each texel subtending16.2min.
2.2. Stimulus conditions
Six different sorts of stimuli were used in this exper-
iment. Stimuli differed with respect to (i) task type (two
types) and (ii) discrimination type (three types). The two
types of tasks employed were an identification task anda location task. In the identification task, on each ex-
perimental trial the participant attempted to identify a
letter occurring in the middle of the display. In the lo-
cation task, the participant attempted to find and point
to a small, square target region at an unknown location
in the stimulus field. Thus, in the identification task,
there was no uncertainty about stimulus location (it was
always in the center of the display), whereas in the lo-cation task, there was no uncertainty about target
identity (it was always a square of the same fixed size).
Each stimulus (irrespective of task type) comprised
two regions filled with different textures, a target region
and a background region. The type of discrimination
required by a given stimulus was determined by the
textures used to define the target and background re-
gions. In the luminance-defined (LD) conditions, thetexture filling the target region had higher mean lumi-
nance than the texture filling the background. In the
texture-defined (orientation) (TDo) conditions, the pre-
dominant orientation of the texture filling the target
region was up-and-to-the-right, while the orientation of
the background texture was down-and-to-the-right. In
the texture-defined (granularity) (TDg) conditions, the
texture filling the target region was finer grained, onaverage, than the background texture. The TDo and
TDg conditions require a purely second-order discrim-
ination because the space-average luminance is con-
stant across both subfields (i.e., all texels have the same
space-average luminance). Conversely, the LD stimuli
require a purely first-order discrimination because tex-
ture element form is constant across both subfields
(see Figs. 1 and 2).For each of our six experimental conditions, there
were four levels of difficulty. Task difficulty was in-
creased by increasing the similarity of the textures used
to define the target vs. background regions. Every tex-
ture used in this study was a mixture of at most two
types of micropatterns. The two types of micropatterns
used to generate the textures in LD stimuli differed in
luminance (one being gray, the other black) but wereotherwise identical (both were small bars, diagonally
oriented up-and-to-the-right). In the easiest level of the
LD condition, all of the micropatterns filling the target
region were gray, whereas all of the micropatterns in the
background were black. In this case, the difference be-
tween target and background is maximally salient; ac-
cordingly we refer to this as the 100% contrast level. The
task is made more difficult by intermixing the two typesof micropatterns, both in the target region and the
background. Thus, for example, by allowing 10% of the
micropatterns in the target region to be black and 90%
to be gray, while 10% of the background micropatterns
are gray and 90% are black, we obtain a contrast level
equal to 90%� 10% ¼ 80%.
More generally, by allowing X% if the micropatterns
in the target region to be black and ð100� X Þ% to begray, while X% of the background micropatterns are
gray and ð100� X Þ% are black, we obtain a contrast
level equal to ð100� 2X Þ%. Thus, for X% ¼ 50%, con-
trast level drops to 0%, at which point the textures in
target and background become identical, and the task
becomes impossible. The four levels of contrast used in
the current study, for both identification and location
tasks, were 100, 80, 60, and 40%.In the TDo conditions, for both the identification and
location tasks, all textures were composed of two types
of micropatterns: one type was a small black bar ori-
ented up-and-to-the-right (55 counterclockwise from
horizontal); the other was a small black bar oriented up-
and-to-the-left (125 counterclockwise from horizontal).
Each bar subtended 14.32min of visual angle in length
at the standard viewing distance, and 2.39min in width.Crucially, both types of micropattern had the same total
luminance. Levels of difficulty were introduced in the
same fashion into the TDo conditions as the LD
conditions. In the 100% contrast condition, all micro-
patterns in the target region were oriented up-and-
to-the-right, whereas all background micropatterns
were oriented up-and-to-the-left. More generally, in the
ð100� 2X Þ% contrast condition, ð100� X Þ% of micro-patterns in the target region were oriented up-and-to-
the-right, while X% were oppositely oriented; at the
same time, ð100� X Þ% of the background micropat-
terns were oriented up-and-to-the-left, while X% were
oppositely oriented. The four levels of contrast, for both
identification and location tasks, were 100, 80, 60, and
40%.
Similarly, in the TDg conditions, for both the iden-tification and location tasks, all textures were composed
of two types of micropatterns: one type comprised a
single black (square) granule; the other type comprised
four smaller square granules, separated in space. Each
larger granule subtended 7:16� 7:16 min2 of arc. Eachsmaller granule subtended 3:58� 3:58 min2. The total
area subsumed by the four smaller granules was equal to
that subsumed by the single larger granule. Thus, thetotal luminance of the four-granule micropattern was
equal to that of the single-granule micropattern. Levels
of difficulty were introduced in the same fashion into the
TDg conditions as the LD and TDo conditions. In the
100% contrast condition, exclusively four-granule
micropatterns were painted into the target region,
whereas only single-granule micropatterns were assigned
Fig. 2. (a) Hundred percentage contrast version of luminance-defined (LD) patch; (b) 80% contrast version of luminance-defined (LD) patch; (c)
100% contrast version of texture-defined orientation (TDo) patch; (d) 80% contrast version of texture-defined orientation (TDo) patch; (e) 100%
contrast version of texture-defined granularity (TDg) patch; and (f) 80% contrast version of texture-defined granularity (TDg) patch.
160 A. Stehli Nguyen et al. / Brain and Cognition 52 (2003) 155–166
to the background. As for the TDo and LD stimuli, the
four levels of contrast used for the TDg stimuli, for both
identification and location tasks, were 100%, 80%, 60%,
and 40%.
2.3. Identification task stimuli
In the identification test, the target region took the
shape of a block, capital letter, and was centered in the
stimulus field. Only 11 letters were used; however, par-
ticipants were not informed of this restriction. The 11
letters used were D, E, F, H, J, L, M, U, W, X, and Y.
The letters shared an equal height of 4.875 in.
(12.3825 cm) and were capitalized. Nine of the 11 letters
(as well as their order) were selected at random for each
contrast level-discrimination type combination.
2.4. Location task stimuli
In the location task, the target patch was a 1 in.
(2.54 cm) square (subtending 1.9� in width at the stan-
dard viewing distance) located at a randomly selected
A. Stehli Nguyen et al. / Brain and Cognition 52 (2003) 155–166 161
location within the background field. For each dis-crimination type, and each contrast level, nine stimuli
were generated.
2.5. Letter naming stimuli
For a control task in which subjects were asked to
name normally printed letters, the stimuli consisted of a
series of letters using a font size of 18. Ten of the letterswere printed in black on a white background, and the
other 10 were printed in white on a black background.
3. Hypotheses and method of analysis
In each of the identification and location tasks, all
participants and controls were shown a series ofstimuli of increasing difficulty for each of the LD,
TDo, and TDg discrimination types. In the identifica-
tion task, for each page presented, the participants
were asked to name the letter they saw. In the location
task, participants were asked to point to the small
square. Performance was measured in terms of the
maximum level of difficulty (which corresponds to the
minimum level of contrast) at which the participantcould still correctly respond. The main analysis was a
comparison of performance between the Alzheimer�spatients vs. normals for each type of discrimination in
each task.
It was expected that for all three discrimination types,
the Alzheimer�s patients would reach threshold at a
lower level of difficulty than normals. This prediction
was based on the findings of the studies previously dis-cussed which have shown Alzheimer�s patients to (1)
perform worse than controls on other figure-ground
tasks (Mendez et al., 1990; McCleary et al., 1996) and
(2) demonstrate a decreased ability to perceive spatial
contrast (Cronin-Golomb et al., 1991; Nissen et al.,
1985). It was also anticipated that the difference between
the Alzheimer�s patients� performance and the controls�might be greater for the TDo and TDg conditionscompared to the LD conditions. This expectation
stemmed from the authors� conjecture that cognitive
functions involved in processing second-order stimuli
may be affected by AD earlier than those responsible for
processing first-order stimuli. Such a result would sug-
gest that second-order stimuli could be useful in devel-
oping a perceptual diagnostic tool. In addition, because
processing of first-order and second-order stimuli areproposed to take place in different areas of visual cortex,
this finding would be highly relevant to Alzheimer�sbrain studies.
A control task was also given. This was a simple letter
naming task designed to assess the ability of the par-
ticipants to recognize normally printed, high contrast
letters and name them correctly.
Thus, this study was designed to contribute to severaltheoretical issues involving the role of visual perception
studies to the research on Alzheimer�s disease. Particu-larly, this study investigates the usefulness of second-or-
der visual stimuli as diagnostic tools for AD. In addition,
it is thought that the results should be able to provide
important information as to the progression of the
disease through visual cortex, and the degree to which
cognitive symptoms of AD are a reflection of perceptiondeficits.
3.1. Subjects
A total of 16 patients and 15 elderly controls were
tested for this experiment. The patients were obtained
from the UC Irvine Alzheimer�s Dementia Research
Center. All patients that come to this center receivestandard physical, neurological, neuropsychological,
and neuroimaging tests as part of the center�s standardassessment battery. Most of the patients participating in
this task did so during one of their first visits to the
center. Confirmation of the diagnosis of possible or
probable AD was made following the completion of the
comprehensive evaluation. Data for three patients were
not analyzed because these patients had not received aclinical diagnosis of probable or possible AD at the time
of data analysis. Four patients received mixed diagno-
ses, meaning that both AD and some other dysfunction
were listed as the primary diagnoses. Data analysis in-
cluded these four subjects. All diagnoses were made
according to NINCDS criteria.
Participants in all stages of dementia severity, as
measured by Mini Mental State Exam (MMSE) scores(Folstein, Folstein, & McHugh, 1975), were invited to
participate. Data for one patient were excluded due to
failure to comprehend basic instructions and task com-
ponents. The control participants were obtained from
the subject pool of elderly controls that is also main-
tained by the UC Irvine AD Research Center.
The mean age for the patients was 79 years ðn ¼ 12Þand the controls had a mean age of 75 s ðn ¼ 15Þ. Bothgroups had a mean education level of 15 years. The
MMSE scores in the patient group ranged from 6 to 28
with a mean of 19.9. Control subjects� scores ranged
from 27 to 30, with a mean of 28.7. Exclusion criteria for
both groups included: corrected Snellen visual acuity of
less than 20/50, the presence of an uncorrected astig-
matism, history of psychiatric illness, and a failure to
comprehend English.All participants signed a consent form pertaining to
this study that was separate from consent given for the
standard tasks at the UC Irvine AD Research Center. It
was made clear that participation in this task was vol-
untary, and that a choice not to participate would not
affect their eligibility for, or the quality of the services
they receive at the center.
162 A. Stehli Nguyen et al. / Brain and Cognition 52 (2003) 155–166
3.2. Procedure
Testing the Alzheimer�s patients took place at the UCIrvine AD Research Center. The control patients had a
choice of being tested at the research center or in their
homes. For the patients, testing occurred between sec-
tions of their standard evaluation. For the control par-
ticipants, a separate appointment was made for the task.
All participants were asked to complete each of the taskcomponents unless they appeared to be abnormally
frustrated or agitated.
Participants were first given the preliminary letter
naming task in which the series of 20 normally printed
letters (10 black on white and 10 white on black) were
shown one at a time after instructing the participant to
name the letter they saw on each page. The number
correct out of 10 for each type was recorded.After the naming task, the identification task was ad-
ministered. First the participant was presented with the
series of LD letters. These stimuli had four levels of deg-
radation and were presented using a staircase approach.
Starting with the highest level of contrast, letters within
the level were presented, and the participant was asked to
name the letter they saw on each page. If they correctly
named three letters (in any order) within the level, theymoved on to the next level. This was continued for each
level. When they missed three in a row within a level, they
were finished. The participant�s response to each stimuluswas marked as either correct or incorrect. The final score
assigned to a participant was the greatest level of difficulty
at which the participant was able to respond correctly to
three stimuli. The five possible scores were 0, 1, 2, 3, and 4.
A score of 0 indicates that the participant was unable torespond correctly to three stimuli even at the easiest level
(in actuality, this score was never given). A score of 1 in-
dicates that the participant was able to respond correctly
to three stimuli at the easiest level, but missed three in a
row at level 2. To achieve a score of 4, a participant had to
respond correctly to three stimuli at the hardest level. This
procedurewas repeated exactly using theTDo stimuli and
then the TDg stimuli.The final task administered was the location task. For
this task the participants were asked to point to the
square or patch in the stimulus field that was different
from the background. The assessment procedure and
scoring system were the same as those used for the
identification task. Also, participants completed the
discrimination types in the same order as in the identi-
fication task: LD first, then TDo, and TDg last.For each subject, the entire procedure took about 15–
30 min.
4. Results
Data were analyzed for 12 patients and 15 controls.
The dependent variable in each of our six experimental
conditions was the maximum level of difficulty at whichthe participant responded correctly to three stimuli. The
five possible scores were 0, 1, 2, 3, and 4.
It became clear during the experiment that the cali-
bration of the TDg stimuli created a ceiling effect for the
scores in both the identification and location tasks (all
patients performed very well on these tasks). Therefore,
data analysis was only carried out for the LD identifi-
cation, LD location, TDo identification, and TDo lo-cation tasks. Results for the LD and TDo location tasks
are given in Fig. 3; results for the LD and TDo letter
identification tasks are given in Fig. 4.
We chose to treat the dependent variable, achieve-
ment level, as ordinal in our analysis rather than as
metric. Because the processes mediating performance
are not known, this conservative option seemed prudent.
Accordingly, we used the sample median as our teststatistic rather than the mean. Thus, a bootstrap pro-
cedure was used to evaluate the difference between the
median score for patients and the median score for
controls in the various conditions. In accordance with
standard bootstrapping procedure, we simulated our
experiment 1000 times, each time randomly resampling
the empirically obtained data to generate the simulated
data set. For each simulated data set, the median for thepatients was calculated and compared to the calculated
median of the controls. This produced a proportion of
the 1000 comparisons in which the patient median was
greater than or equal to the control median. This pro-
portion was taken to estimate the p-value of the null
hypothesis that the patient population median is no
lower than the control population median. This proce-
dure was carried out for: (a) LD identification, (b) LDlocation, (c) TDo identification, and (d) TDo location.
Table 1 gives the estimated p values for each of these
four conditions.
Patients were found to be significantly impaired
compared to controls in the LD location task. Specifi-
cally, in all but six out of 1000 simulated data sets the
patient median was strictly less than that of the controls.
The primary data for the location task using LD stimuliin AD patients and control subjects are illustrated in
Figs. 3a and b.
5. Discussion
The motivation of this resea rch was to evaluate the
ability of Alzheimer�s patients, compared to controls, toperform figure-ground tasks featuring first-order and
second-order stimuli. We know at the outset the patients
were of mixed etiologies and varying severity which may
have contributed to the results obtained. Contrary to
our hypothesis the patients did not perform significantly
worse on all of the tasks presented. It had also been
predicted that the difference in performance between
Fig. 4. Histograms for (a) controls� performance with luminance-defined (LD) letter naming; (b) patients� performance with luminance-defined (LD)letter naming; (c) controls� performance with texture-defined orientation (TDo) letter naming; and (d) patients� performance with texture-defined
orientation (TDo) letter naming.
Fig. 3. Histograms for (a) controls� performance with luminance-defined (LD) patch location; (b) patients� performance with luminance-defined (LD)patch location; (c) controls� performance with texture-defined orientation (TDo) patch location; and (d) patients� performance with texture-defined
orientation (TDo) patch location.
A. Stehli Nguyen et al. / Brain and Cognition 52 (2003) 155–166 163
Table 1
Probability that patients� median score is less than controls� medianscore as estimated by bootstrap procedure
Stimulus type Task type
Letter naming Patch location
Luminance .020 .994��
Texture(o) .356 .755
Note. Data were analyzed for 12 patients and 15 controls.** p < :01.
164 A. Stehli Nguyen et al. / Brain and Cognition 52 (2003) 155–166
normals and patients would be especially prominent in
the tasks involving the second-order stimuli (i.e., the
TDo stimuli). The obtained results did not support this
hypothesis, and instead suggested that the AD patients
were not suffering from any selective primary visual
deficits involving the processing of first-order or second-
order stimuli. This finding is similar to the results found
by McCleary et al. (1996) in evaluating performance onthe Ishihara tasks. Because it had been established that
the patients were not color blind, it was stated that the
errors made by patients must be attributed to other
types of deficits. That is to say that the processes nec-
essary for color vision were not selectively impaired by
the disease. Similarly in the present experiment it seems
that other features of the stimuli and task had more
impact on performance than did luminance or textureper se.
Surprisingly, the data suggest that AD patients may
be selectively impaired on the location task, as com-
pared with the identification task, where little evidence is
seen of any impairment. Although results for the TDo
location task fail to reach statistical significance, they
tend strongly in the same direction as the LD location
task. Moreover, it should be noted that (i) the (strictlyordinal) statistical test used here is much weaker than
corresponding (metric) tests comparing means, and (ii)
the scale of stimulus difficulty we used includes only four
levels of contrast. This implies that there are very few
possible values for the median to take, further restricting
the power of our statistical test.
In light of these observations, the significance of the
results of the LD location task is remarkable. However,a glance at the raw data (see Figs. 3a and b) shows why
the difference between patient and control performance
achieves statistical significance: only two out of 12 of the
AD patients succeeded at the most difficult level; how-
ever, 12 out of 15 controls succeeded at this level.
There are several possible mechanisms that may ac-
count for this deficit. Mendez et al. (1990) reported that
AD patients have problems with figure-ground tasks.However, the fact that a significant difference between
patients and controls was evident only for the location
task and not for the identification task suggests that
figure-ground discrimination per se cannot account for
the results. Because the identification task posed no
extraordinary challenge to the AD patients, we infer
that the perceptual apparatus is intact that draws dis-tinctions between the textures composing the back-
ground and the target for all types of discriminations
required in this study.
This suggests that it is the search for the target that is
somehow impaired in the AD patients. The location task
requires the observer to distribute his/her attention
broadly over the stimulus field in order to single out that
region that spontaneously emerges as the most likelytarget patch candidate. It may be that AD patients have
difficulty distributing their attention sufficiently broadly
over space, and are thus forced to adopt a sub-optimal,
serial search strategy using a restricted attentional win-
dow. In this case, one might expect AD patients to ex-
hibit a different, perhaps more variable, pattern of eye
movement fixations in performing the location task than
do normal controls. Alternatively, it may be that ADpatients are able to distribute their attention in the re-
quired manner, but that the mechanism that triggers the
emergence of the target is somehow compromised, so
that a higher signal-to-noise ratio is required to detect
the target. In general, the impairment suggested by these
results is consistent with the finding of impaired spatial
attention in AD in several previous studies using
other location–detection and visual alerting tasks(Buck, Black, Behrmann, Caldwell, & Bronskill, 1997;
Greenwood, Parasuraman, & Alexander, 1997; Nebes &
Brady, 1989).
The current results are highly suggestive, given the
emerging evidence that visual processing splits into
distinct streams, one concerned with object identifica-
tion (the ‘‘What’’ stream), and the other concerned with
localizing objects in space (the ‘‘Where’’ stream) (Carl-son, 1994; Lennie, 1980; Livingstone & Hubel, 1994;
Merigan, Freeman, & Meyers, 1997; Ungerleider &
Mishkin, 1982). In the ventral ‘‘What’’ stream, infor-
mation is channeled through VI into inferior temporal
cortex. In the dorsal ‘‘Where’’ stream, information is
channeled through VI into parietal cortex. The identi-
fication task involves ‘‘identity uncertainty’’: the patient
must decide which letter of the alphabet is on the page.In controls, to successfully complete the location task,
one does not have to identify or classify the figure. In-
stead, the uncertainty in the task is locational in that one
must find where the figure is in the stimulus field. Thus,
the results suggest that the ability to process locational
information may be selectively impaired by AD. Thus
the results of the current experiment may suggest that
AD patients suffer from a selective deficit involving the‘‘Where’’ pathway. Although visual cortex (Broadman
areas 17 and 18) contains relatively few neurofibrillary
tangles in AD (Braak, Braak, & Kalus, 1989; Arnold,
Hyman, Flory, Damasio, & Van Hoesen, 1991),
documentation of DNA damage in visual cortex neu-
rons suggests that they may degenerate by tangle-inde-
pendent mechanisms and that oxidative damage may
A. Stehli Nguyen et al. / Brain and Cognition 52 (2003) 155–166 165
contribute to such mechanisms (Su, Deng, & Cotman,1997). The parietal cortex contains relatively more
neurofibrillary tangles which correlates with the degree
of impairment in spatial attention (Buck et al., 1997).
AD patients also exhibit increased losses of neurons
from the magnocellular layer of the lateral geniculate
nucleus (LGN) (Katz & Rimmer, 1989). As explained by
Carlson (1994), while the ‘‘What’’ system receives input
from both the parvocellular and magnocellular layers,the ‘‘Where’’ system receives most of its input from the
magnocellular layer. The magnocellular layer also me-
diates contrast sensitivity over the lower spatial fre-
quencies (Katz & Rimmer, 1989), which has been shown
to be impaired in many AD patients (e.g., Cronin-Go-
lomb et al., 1991; Nissen et al., 1985). These observa-
tions suggest that the results of the current study may
reflect selective damage to the magnocellular layer of theLGN in AD patients, rather than damage to parietal
cortex. In this case we should expect AD patients to be
selectively impaired at motion judgments, which are
thought to be mediated exclusively by the magnocellular
pathway.
Some authors feel that the ‘‘Where’’ pathway in vi-
sual cortex plays a large role in guiding people�s actionsin relation to objects in the visual field (Goodale andMilner in Carlson, 1994). The hypothesis that the
‘‘Where’’ pathway in visual cortex suffers selective
damage in AD receives support from case studies of
patients with behavioral symptoms involving incorrect
object selection and disorientation (Levine et al., 1993;
Katz & Rimmer, 1989). Thus, further work to investi-
gate the relationship between Alzheimer�s disease and
damage to the ‘‘Where’’ pathway may provide insight asto how this neurological damage is reflected in behavior.
Acknowledgments
The authors thank Julene Johnson, Ph.D., and other
clinical staff of the Institute for Brain Aging and De-
mentia for assessment and referral of subjects. This
study was supported by Alzheimer Disease Research
Center grants from the State of California and the
National Institute on Aging, Carl W. Cotman, Ph.D.,
Principal Investigator.
References
Arnold, S. E., Hyman, B. T., Flory, J., Damasio, A. R., & Van
Hoesen, G. W. (1991). The topographical and neuroanatomical
distribution of neurofibrillary tangles and neuritic plaques in the
cerebral cortex of patients with Alzheimer�s disease. Cerebral
Cortex, 1, 103–116.
Arsenault, A. S., Wilkinson, F., & Kingdom, F. A. (1999). Modulation
frequency and orientation tuning of second-order texture mecha-
nisms. Journal of the Optical Society of America A, 16(3), 427–437.
Bergen, J. R., & Landy, M. S. (1991). The computational modeling of
visual texture segregation. In M. S. Landy, & J. A. Movshon
(Eds.), Computational models of visual processing (1, pp. 253–271).
Cambridge, MA: MIT Press.
Braak, H., Braak, E., & Kalus, P. (1989). Alzheimer�s disease: Arealand laminar pathology in the occipital isocortex. Acta Neuropa-
thologica, 77, 494–506.
Buck, B. H., Black, S. E., Behrmann, M., Caldwell, & Bronskill, M. J.
(1997). Spatial and object-based attentional deficits in Alzheimer�sdisease: Relationship to HMPAO_SPECT measures of parietal
perfusion. Brain, 120, 1229–1244.
Carlson, N. R. (1994). Vision. In Physiology of behavior (fifth ed., pp.
141–179). Boston: Allyn and Bacon.
Cronin-Golomb, A. (1995). Vision in Alzheimer�s disease. The Geron-tologist, 35, 370–376.
Cronin-Golomb, A., Corkin, S., Rizzo, J. F., Cohen, J., Growdon, J.
H., & Banks, K. B. (1991). Visual dysfunction in Alzheimer�sdisease: Relation to normal aging. Annals of Neurology, 29, 41–52.
Folstein, M. F., Folstein, S. F., & McHugh, P. R. (1975). ‘‘Mini-
mental state’’: A practical method for grading the mental state
of patients for the clinician. Journal of Psychiatric Research, 12,
189–198.
Graham, N. (1994). Non-linearities in texture segregation. Higher-
order processing in the visual system. In J. A. Good, & G. R. Bock
(Eds.), Ciba Foundation symposium (184, pp. 309–329). Chichester,
UK: Wiley.
Greenwood, P. M., Parasuraman, R., & Alexander, G. E. (1997).
Controlling the focus of spatial attention during vision search:
Effects of advanced aging and Alzheimer�s disease. Neuropsychol-ogy, 11, 3–12.
Huxlin, K. R., & Merigan, W. H. (1998). Deficits in complex visual
perception following unilateral temporal lobectomy. Journal of
Cognitive Neuroscience, 10, 395–407.
Katz, B., & Rimmer, S. (1989). Ophthalmologic manifestations of
Alzheimer�s disease. Survey of Ophthalmology, 34, 31–43.Lennie, P. (1980). Parallel visual pathways: A review. Vision Research,
20, 561–594.
Levine, D. N., Lee, J. M., & Fisher, C. M. (1993). The visual variant of
Alzheimer�s disease: A clinicopathologic case study. Neurology, 43,
305–313.
Livingstone, M., & Hubel, D. (1994). Segregation of form, color,
movement, and depth: Anatomy physiology and perception. In H.
Gutfreund, & G. Toulouse (Eds.), Biology and computation: A
physicist�s choice (pp. 740–749). Singapore: World Scientific.
Lu, Z.-L., & Sperling, G. (1996). Second-order illusions: Mach bands,
Chevreul, and Craik-O�Brien-Cornsweet. Vision Research, 36(4),559–572.
McCleary, R., Shankle, W. R., Mulnard, R. A., & Dick, M. B. (1996).
Ishihara test performance and dementia. Journal of the Neurolog-
ical Sciences, 142, 93–98.
Mendez, M. F., Mendez, M. A., Martin, R. N., Smyth, K. A., &
Whitehouse, P. J. (1990). Complex visual disturbances in Alzhei-
mer�s disease. Neurology, 40, 439–443.Merigan, W. H. (1996). Basic visual capacities and shape discrimina-
tion after lesions of extrastriate area V4 in macaques. Visual
Neuroscience, 13, 51–60.
Merigan, W., Freeman, A., & Meyers, S. P. (1997). Parallel processing
streams in human visual cortex. Neuroreport: An International
Journal for the Rapid Communication of Research in Neuroscience,
80, 3985–3991.
Nebes, R. D., & Brady, C. B. (1989). Focused and divided attention in
Alzheimer�s disease. Cortex, 25, 305–315.Nissen, M. J., Corkin, S., Buonanno, F. S., Growdon, J. H., Wray, S.
H., & Bauer, J. (1985). Spatial vision in Alzheimer�s disease.
Archives of Neurology, 42, 667–671.
Regan, D., & Hong, X. H. (1994). Recognition and detection of
texture-defined letters. Vision Research, 34, 2403–2407.
166 A. Stehli Nguyen et al. / Brain and Cognition 52 (2003) 155–166
Regan, D., & Simpson, T. (1995). Multiple Sclerosis can cause visual
processing deficits specific to texture-defined form. Neurology, 45,
809–815.
Su, J. H., Deng, G., & Cotman, C. W. (1997). Neuronal DNA
damaging precedes tangle formation and is associated with up-
regulation of nitrotyrosine in Alzheimer�s disease brain. Brain
Research, 774, 193–199.
Sutter, A., & Graham, N. (1995). Investigating simple and complex
mechanisms in texture segregation using the speed–accuracy
tradeoff method. Vision Research, 35(20).
Ungerleider, L. G., & Mishkin, M. (1982). Two cortical visual systems.
In D. J. Ingle, M. A. Goodale, & R. J. W.Mansfield (Eds.),Analysis
of visual behavior (pp. 549–586). Cambridge, MA: MIT Press.
Further reading
Nothdurft, H. C., & Parlitz, D. (1993). Absence of express saccades to
texture or motion defined targets. Vision Research, 33, 1367–1383.