Vision Research 42 (2002) 1339–1348
www.elsevier.com/locate/visres
Hemianopic visual field defects elicit hemianopic scanning
M.L.M. Tant
a
a,e,*
,
F.W. Cornelissen
b,e
, A.C. Kooijman
b,c,d,e
, W.H. Brouwer
a,e
Department of Psychology, University of Groningen, Section Neuropsychology, University Hospital Groningen, Poortweg 4, 2de verdieping,
P.O. Box 30.001, 9700 RB Groningen, Netherlands
b
Laboratory of Experimental Ophthalmology (LEO), University of Groningen, P.O. Box 30.001, 9700 RB Groningen, Netherlands
c
Department of Ophthalmology, University Hospital Groningen, P.O. Box 30.001, 9700 RB Groningen, Netherlands
d
Visio, National Foundation for the Visually Impaired and Blind, P.O. Box 144, 9752 AC Haren, Netherlands
e
School for Behavioral and Cognitive Neurosciences (BCN), University of Groningen, P.O. Box 145, 9700 AC Groningen, Netherlands
Received 31 July 2001; received in revised form 28 January 2002
Abstract
Previous explanations for the variability in success of compensating for homonymous hemianopia (HH) has been in terms of
extent of the brain injury. In using on-line eye movement registrations, we simulated HH in 16 healthy subjects and compared their
scanning performance on a dot counting task to their own ‘‘normal’’ condition and to real HH patients’ performance.
We evidenced clear parallels between simulated and real HH, suggesting that hemianopic scanning behaviour is primarily visually
elicited, namely by the visual field defect, and not by the additional brain damage. We further observed age-related processes in
compensating for the HH. Ó 2002 Elsevier Science Ltd. All rights reserved.
Keywords: Hemianopia; Brain; Visual; Simulation; Age dependence
1. Introduction
Homonymous hemianopia (HH) is a visual field defect (VFD) in which, for both eyes to the same extent,
half of the visual field is blind. The HH can either be
complete or incomplete, congruent or incongruent, and
with or without macular sparing. This VFD results from
unilateral post-chiasmal brain damage. Whether or not
the HH is complete depends on the relative integrity of
part of the visual stream or pathway. Macular sparing
and congruency of the VFD are more frequently observed with posterior than with anterior lesions. Nearly
80% of patients with unilateral post-chiasmal brain
damage acquire a homonymous VFD (Zihl, 1994).
Common causes are cerebrovascular accident, traumatic
brain injury and tumours (e.g. Kerkhoff, 1999; Zihl,
2000).
*
Corresponding author. Address: Department of Neuropsychology,
University Hospital Groningen, University of Groningen, Poortweg
4-2de verdieping, P.O. Box 30.001, 9700 RB Groningen, Netherlands.
Tel.: +31-50-361-4665; fax: +31-50-361-1706.
E-mail address:
[email protected] (M.L.M. Tant).
Visual field defects often lead to visually related
complaints and dysfunctions. Patients complain for example about having a limited overview, bumping into
obstacles or persons and experience their vision as being
too ‘‘slow’’. These disabilities are related to the degree of
compensation for the visual field loss. For comprehensive reviews, we refer to Kerkhoff (1999) and Zihl
(2000). Oculomotor compensation, this is adaptive visual scanning behaviour, can be assessed by recording
eye movements (e.g. Zangemeister, Meienberg, Stark, &
Hoyt, 1982; Zangemeister & Oechsner, 1996; Zihl, 1995,
1999, 2000).
A paradigm to objectively and quantitatively assess
oculomotor compensational behaviour was introduced
by Zihl (1995, 1999, 2000) and consisted of inspection of
a dot pattern. The stimulus display consists of 20 randomly arranged dots projected onto a screen. Subjects
are asked to fixate the centre of the screen, after which the
dot pattern is presented and eye movements are recorded.
Subjects subsequently scan the pattern and silently count
the number of dots. Upon completion they report the
number of dots. A relatively simple stimulus display was
chosen to restrict visual scanning to the process of visual
sampling without any further identification component
0042-6989/02/$ - see front matter Ó 2002 Elsevier Science Ltd. All rights reserved.
PII: S 0 0 4 2 - 6 9 8 9 ( 0 2 ) 0 0 0 4 4 - 5
1340
M.L.M. Tant et al. / Vision Research 42 (2002) 1339–1348
(Zihl, 1999), or the primary involvement of other complex higher-order visual functions. Using this paradigm,
it was found that in HH typical defective oculomotor
scanning behaviour is characterised by longer scanning
times and scan paths, higher number of fixations and refixations, and, at least in part, longer fixation durations
and shorter saccadic amplitudes (e.g. Zihl, 1995, 1999,
2000). These findings are in concordance with other reports (e.g. Chedru, Leblanc, & Lhermitte, 1974; Ishiai,
Furukawa, & Tsukagoshi, 1987; Kerkhoff, 1999; Meienberg, Zangemeister, Rosenberg, Hoyt, & Stark, 1981;
Neetens, 1994; Zangemeister et al., 1982; Zangemeister
& Oechsner, 1996).
In large, it was found that about 40% of the HH
patients spontaneously compensate effectively for their
VFD (Zihl, 1995, 1999, 2000) and that the subjective
visual complaints by HH patients were substantiated by
eye movement recordings during the dot counting task,
confirming its practical relevance. Interestingly, it was
concluded (Zihl, 1999, 2000) that the presence, time
since, and severity of the VFDs could not sufficiently
explain the observed scanning deficit and that additional
factors are crucial for explaining the impaired oculomotor scanning. Zihl suggested that the extent of the
brain injury is a crucial factor and that occipito-parietal
and posterior thalamic brain injury may be responsible
for inefficient compensation.
Nevertheless, there are some peculiar aspects in the
results, which cast doubt on the provided explanation
for the individual differences in the efficiency of compensation for HH. Firstly, it was noted by Zihl (1999)
that the scanning (e.g. in terms of scanning time) was
found to be impaired in this very simple visual sampling
task. Hence, even in a task, in which complex higherorder (that is brain related) functions are not involved, a
disability appears. This calls into question the crucial
importance of the integrity of the brain for the visual
disability. Zihl therefore cautions against the (mis)interpretation of the results in term of ‘‘unspecific’’ cognitive slowing, suggesting in our view, an interpretation
in terms of mere visual slowing. He suggested that the
‘‘slowness of vision’’ may, at least in part, be explained
by the use of hypometric saccades which are provoked
by a homonymous VFD.
Secondly, it was found that the side of the VFD (and
therefore the side of the brain lesion) was not a crucial
factor. Zihl (1999) comments this observation to be
surprising, because of the assumed specialisation of the
right (posterior) hemisphere for visuo-spatial function,
including the spatial guidance of eye movements. If
predominantly higher-order visual (that is brain related)
functions were involved, one would have expected leftsided HH patients (with right-sided brain damage) to
perform worse, due to hemispheric specialisation and
the inherently visuo-spatial nature of the dot counting
task.
Both observations suggest that the deficit in visual
exploration is perhaps not predominantly related to
additional brain damage, but is merely a knock-on effect
of the lower-order dysfunction, in this case the hemianopic visual field loss. In order to preclude the effects
of brain damage, we simulated HH in healthy subjects
and compared the visual exploration to their own
‘normal’ condition. By simulating the hemianopic visual
field loss, we ‘create’ subjects with the lower-order visual
dysfunction, but without higher-order dysfunctions
caused by brain damage. The observed disabilities (if
any) during simulation result from the visual limitation
only and do not require a further explanation in terms of
brain damage. If visual exploration deficits in real HH
are predominantly provoked by the VFD, the visual
exploration behaviour, displayed in simulated and real
HH, should be comparable, including the variability in
performance between individuals. Our primary research
question hence concerns the influence of the pure visual
component on hemianopic visual exploration during a
dot counting task. We also included real HH patients in
this study to compare the patterns of performance with
simulated and real HH.
A secondary question concerns the explanation of
individual differences in the efficiency of compensating
for HH. Apart from differences in extent and site of
brain injury in patients, in healthy subjects there are
large individual differences in higher-order visual and
cognitive abilities, for example differences in visual speed
and other components of intelligence. Some of these
abilities are suggested to be highly dependent on age,
for example perceptual speed and spatial orientation
(Schaie & Willis, 1993) and fluid intelligence (Rybash,
Roodin, & Hoyer, 1995). As was also suggested by
Szlyk and colleagues (Szlyk, Brigell, & Seiple, 1993), it
is quite conceivable that such age-related abilities play
an important role in the efficient compensation of HH.
To investigate the effect of ageing on the efficiency of
compensation, we included both younger and older
adults in the study. It is predicted that the older subjects
will have significantly more problems in coping with
HH.
In summary of the research questions, we expect to
find typical HH scanning performance in healthy subjects with a simulated complete congruent hemianopic
VFD, since we hypothesise that HH scanning is primarily generated by the VFD and not by brain damage.
To fully compare and characterise HH scanning performance, we will perform, in addition to general analysis, also directional, hemispace, and trend analysis
(see further). Secondly, we expect to find the disabilities to be more pronounced in an older age group,
since we assume that individual differences in perceptual
and intellectual abilities, which tend to decrease with
age, are important factors governing the compensation
process.
M.L.M. Tant et al. / Vision Research 42 (2002) 1339–1348
2. Methods
2.1. Subjects
Sixteen healthy subjects participated in this study
(seven males, nine females). Their mean age was 40 years
(range 16–71). Two age groups were included: a younger
group with a mean age of 21 years (range 16–23) and an
older age group with a mean age of 60 years (range 46–
71), each consisting of eight subjects. They showed no
signs of cognitive decline (CST; De Graaf & Deelman,
1991), reported to be right-handed, and had normal or
corrected-to-normal visual acuity. They declared to have
no visually related complaints.
Twenty-nine patients were included (23 males, 6 females). They showed no evidence of cognitive decline
(CST; De Graaf & Deelman, 1991 and MMSE; Folstein,
Folstein, & McHugh, 1975), aphasia (SAN; Deelman,
Liebrand, Koning-Haanstra, & van der Burg, 1987) or
apraxia (De Renzi, Faglioni, & Sorgato, 1982). Neither
of them showed severe unilateral visual hemi-neglect
(UN) or visual agnosia. The selection procedure for UN
is described elsewhere (Tant, Kuks, Kooijman, Cornelissen, & Brouwer, 2002). All patients had a binocular
optimally corrected acuity of 0.8 or better and contrast
sensitivity within normal ranges. Automated perimetry
was performed using the Humphrey Field Analyzer
(Full Field 246 screening program, age corrected, threezone strategy). Fourteen patients had left-sided HH (10
incomplete, 4 complete). Their mean age was 54 years
(range 29–76), the mean time since lesion was 32 months
(range 6–157). Ten of these patients had macular sparing
(mean 4°, range 2–10°). All were victims of stroke, except two patients, who were surgically operated for tumour. One patient, with left-sided HH, only had (right)
monocular vision. Fifteen patients had right-sided HH
(nine incomplete, six complete). Their mean age was 50
years (range 17–68), the mean time since lesion was 80
months (range 3–390). Eleven of these patients had
macular sparing (mean 4°, range 3–8°). One patient was
surgically operated for hydrocephalus, and one for tumour. Two patients suffered closed head injury. The
remaining patients were victims of stroke.
Keeping the limitations of sensitivity and fixation
control of the perimetric procedure in mind, we observed that the incompletenesses of the HHs consisted
of a small wedge-shaped sparing near the midline in
the upper or lower quadrant. None of patients showed
VFDs which were strikingly incongruent.
2.2. Dot counting task and apparatus
Our dot counting task is based upon the work of Zihl
(1995). We presented in total 29 patterns of dots. The
screen dimensions were 36° and 27° horizontally and
vertically respectively. The dot size was 1°. Dots were
1341
white (luminance 25 cd/m2 ) on a grey background (50%
contrast). The viewing distance was 52 cm. Dot patterns were created by giving individual dots a random
horizontal and vertical offset relative to a rectangular
imaginary 4 5 grid. The random offsets were 2.0° relative to the grid position. As the grid positions were 6.0°
apart, dots never overlapped. Dots were assigned randomly to any of the 20 possible grid positions. We
presented five different patterns consisting of 19, 20 and
21 dots each (15 trials). For the trials with 21 dots, one
extra dot was added to the (generated 20 dot) pattern, so
that it did not overlap any other dot. These patterns
were identical for all subjects. Additionally we presented
patterns consisting of 5–17 dots (two-dot increment),
which each were presented twice (14 trials). Their spatial
distribution was randomly generated on each presentation. The 29 trials were presented in a random order.
Before each trial, a fixation dot (1.5°) was presented in
the centre of the screen. Upon stable fixation, the trial
was initiated. The subject was asked to count the number of dots. When the subject verbally indicated being
finished, the trial was aborted, and the answer registered.
During the experiment the eye movements were recorded using an EyeLink Gaze tracker (SensoMotoric
Instruments GmbH, Teltow, Germany) which registers real-time gaze at 250 Hz. When simulating HH, a
window, with the same properties as the background,
continuously and completely blanked one side of the
screen with reference to the current gaze position. This
could either be left or right of fixation in order to simulate left- or right-sided HH respectively. The length of
the entire system’s delay (from eye movement to screen
update) was 20 ms. Prior to the experiment, the equipment was calibrated using a nine-point grid. The initial
central fixation dot, prior to each trial, was also used
for drift correction which may result from slips of the
Eyelink’s headset. Small head-movements were allowed
(and corrected for) during the experiment. This equipment allows for relatively normal free viewing conditions.
2.3. Procedure
The healthy subjects performed the task on two different occasions. On each occasion, they firstly performed the task in a non-simulation (that is normal) and
subsequently in a simulation condition. During the task
in a simulation condition, the side of the simulated
homonymous hemianopia (sHH) was fixed. On the
second occasion, the side of the sHH was changed for
each subject. During simulation, half of the subjects had
a macular sparing of 2.7° on both occasions. The patients performed the task once. Obviously, no simulation was imposed.
1342
M.L.M. Tant et al. / Vision Research 42 (2002) 1339–1348
2.4. Statistical analysis
The oculomotor parameters are number and duration
of the fixations, number and amplitude of the saccades,
and the length of the scanpath, which is sum of the
saccadic amplitudes. When both number of fixations
and number of saccades are used in the same analysis,
only the number of fixations are reported. As a saccade
typically follows a fixation, both parameters are logically linked and hence the number of one or the other
provides no additional information. We further report
the absolute error in counting the dots (henceforth referred to as error) and the search time. Additionally, we
perform directional and hemispace analysis. Since the
healthy subjects suffered no actual brain damage, directions and side of the hemispace are defined with
respect to the side of the VFD. ‘‘Ipsilateral’’ and ‘‘contralateral’’ hence refer to ‘‘in or towards’’ the affected
and intact visual hemifield respectively. Directional
analysis is performed on the amplitudes and number of
saccades. Hemispace analysis is performed on the fixation parameters. The hemispace is defined in terms of
the centre of the screen, which is also the start of exploration in each trial. To further characterise the
scanning performance, we perform a trend analysis
using the errors, search time, number of fixations, and
length of the scanpath to investigate how the difficulty of
the task (operationally defined by the number of dots in
each pattern) influences performance in each subject
group.
To analyse the data by the healthy subjects, we
performed a MANOVA on a doubly repeated measures design. When significant multivariate effects were
observed, the univariate effects were inspected. When
necessary, we additionally performed simple-main-effects analysis, to untangle the interaction components.
For this last type of analysis, only P-values will be reported. The design is graphically depicted in Fig. 1. We
included the factors of age (young/old), macular sparing
(yes/no), and sequence (A/B) as between-subjects factors. The sequence factor represents whether subjects
were first imposed with a left- (A) or right-sided sHH
(B). Within-subject factors were measurement (first or
second occasion) and mode (normal/simulated HH).
The measurement-, mode- and age- factors are interpretable as main-effects factors. Since the sparing factor
has no relevance in the normal mode and can only exert
effects during simulation, a main effect of sparing will
arise as a mode sparing interaction. Effects of the side
of the sHH are inferred from simple-main-effect analysis
on the measurement mode sequence interactions.
Namely, a significant measurement sequence interaction, within the simulation mode, reveals differential
performances by left- and right-sided sHH subjects.
When observed, further analysis and inspection of the
means will then reveal the nature of the effects of the side
of sHH and the consistency of the difference between
left- and right-sided sHH in both sequences. Absence of
this two-way interaction indicates no difference in leftversus right-sided sHH. Although statistically interactional, the effects of both sparing and side of the sHH
will be reported as main effects.
The patient data conform to a repeated measures
design with side of the HH as a between-subjects variable. We analysed the same parameters as for the healthy subjects.
3. Results
3.1. General analysis: multivariate
MANOVA failed to reveal significant multivariate
main effects of measurement and age, suggesting that
nor repetition of the experiment or age did have any
overall influence on the data. There was a significant
mode-effect (F ð7; 2Þ ¼ 147, P < 0:007), suggesting an
overall effect of the simulation. We observed no effect of
sparing, indicating that macular sparing did not lead to
better performance in sHH. We found no measurement
mode interaction, confirming the absence of learning
effects in both modes. Age, however, did interact significantly with mode (F ð7; 2Þ ¼ 62, P < 0:016). The age
effect will be explored further. We observed a significant
measurement mode sequence interaction (F ð7; 2Þ ¼
27, P < 0:036), suggesting a possible influence of the side
of the sHH. Simple-main-effect analyses will be performed to reveal the nature of these effects and interactions.
3.2. General analysis: univariate
Fig. 1. Graphical depiction of the healthy subjects design. Betweensubject factors are sequence, age, and sparing. Within-subject factors
are measurement and mode.
In the simulation mode (compared to the same subjects in the normal mode), subjects took more time and
made more errors in counting the dots. They fixated
more and the mean fixation duration was longer. Also
the scanpath was prolonged (Fig. 2). All parameters
showed significant differences, except the saccadic amplitudes (F-range: 11–184, P-range: 0.011–0.0001).
M.L.M. Tant et al. / Vision Research 42 (2002) 1339–1348
1343
Fig. 2. Simulated HH (black) provokes longer search times, more errors, more fixations, longer fixation durations, and longer scanpaths than in
normal conditions for the same subjects (white). Patient data (grey bars) for comparison. Error bars are 1 S.E.
In comparison to right-sided sHH, subjects with leftsided sHH made more errors (F ð1; 8Þ ¼ 13, P < 0:007)
and presented a longer search time (F ð1; 8Þ ¼ 11, P <
0:01) (Table 1). Simple-main-effects analysis revealed
that none of the parameters produced significant differences in normal modes. Significantly worse performance by left-sided HH was also observed in the patient
group, but only for the errors (F ð1; 27Þ ¼ 9, P < 0:005)
(Table 1).
The effect of age was apparent in the search time
(F ð1; 8Þ ¼ 19, P < 0:002) and number of fixations
(F ð1; 8Þ ¼ 19, P < 0:006). The increase for both parameters in the simulation mode was greater for the
older age group (Fig. 3). Exploratory, we plotted the
search time per dot (this is relative search time) in
function of the trial order, and observed that, although
Table 1
Effects of the side of the VFD in simulated (sHH) and real hemianopia
(HH)
Side of the VFD
Errors
Search time (s)
sHH
HH
sHH
HH
Left
Right
0.95 (0.11)
0.68 (0.10)
13.2 (0.83)
10.1 (0.67)
0.34 (0.11)
0.25 (0.10)
10.2 (0.83)
8.8 (0.65)
Left-sided sHH subjects take more time and make more errors than
right-sided sHH. Left-sided HH patients make more errors than rightsided HH. The difference between left- and right-sided HH was not
statistically significant for the search time. Standard errors between
brackets.
Fig. 3. Subjects from the older age group had longer search times and
made more fixations (and saccades) in the simulated conditions.
N: Normal condition, sHH: simulated condition. Error bars are 1 S.E.
1344
M.L.M. Tant et al. / Vision Research 42 (2002) 1339–1348
hemifield, for both left- and right-sided sHH. This pattern of results was paralleled in the patient group.
Multivariate analysis failed to reveal any effect of the
side of the HH, but presented a significant directionality
effect (F ð2; 26Þ ¼ 11, P < 0:000). Inspection of the univariate analysis showed the saccadic amplitudes to
be smaller in ipsilateral than in contralateral direction
(F ð1; 27Þ ¼ 19, P < 0:000) (Table 2).
3.4. Hemispace analysis
Fig. 4. The age effects in sHH are especially evident in the beginning of
the experiment. N: Normal condition, sHH: simulated condition,
Y: Younger age group, O: Older age group, HH: patient group.
always present, the age effects are specially evident in the
beginning of the task (Fig. 4). We also observed slight to
moderate age effects in the patient population, as evidenced by the Pearson’s correlation of age with search
time (rð29Þ ¼ 0:38, P < 0:05) and number of fixations
(rð29Þ ¼ 0:42, P < 0:05).
3.3. Directional analysis
Multivariate directional analysis on the healthy subjects data, including the number and amplitude of the
saccades, was significant (F ð2; 13Þ ¼ 4, P < 0:05) for the
mode direction interaction. The saccadic amplitudes
in either direction did not differ in the normal mode, but
there was a significant directional effect in the simulation mode (F ð1; 14Þ ¼ 5, P < 0:05). Namely, ipsilateral
saccadic amplitudes were smaller than contralateral
amplitudes (Table 2). There was no effect of the side of
the sHH, indicating that the amplitudes of saccades into
the blind hemifield are smaller than into the seeing
Table 2
Directional and hemispace analysis
3.5. Trend analysis
Ipsilateral
Saccadic amplitude
(degree)
N
sHH
HH
Number of fixations
N
sHH
HH
A multivariate hemispace analysis was performed on
the number and durations of the fixations by the healthy
subjects. A multivariate mode field interaction was
found (F ð2; 13Þ ¼ 11, P < 0:002).
Univariate analysis showed a significant effect of
fixation duration (F ð1; 14Þ ¼ 9, P < 0:010). This effect
was however not apparent in the simulation condition.
In the normal mode, the durations proved to be longer
when they occurred on the right side of the screen than
on the left side (352 and 411 ms for left and right
hemispace respectively, P < 0:004). Tentatively, in the
simulation mode, inspection of the means would suggest
ipsilateral fixation durations (577 ms) to be longer than
contralateral ones (533 ms), but this difference proved
not to be significant.
There was also a significant effect of hemispace on the
number of fixations (F ð1; 14Þ ¼ 8, P < 0:012). In the
normal mode, there were as many fixations in either left
or right hemispace, but clearly more fixations in the
ipsilateral hemispace in the simulation mode (P < 0:009)
(Table 2). There was no interaction with the side of the
sHH, indicating that, in both left- and right-sided sHH,
subjects fixated more on the same side of the screen as
their VFD.
This pattern of results was paralleled in the patient
group. We observed a multivariate effect of hemispace
ðF ð2; 26Þ ¼ 55, P < 0:000Þ and no effect of side of the
HH. Both left- and right-sided HH patients fixated more
in the ipsilateral hemispace (F ð1; 27Þ ¼ 93, P < 0:000)
(Table 2).
8.4 (0.59)
9.08 (0.96)
7.00 (0.18)
7.3 (0.39)
12.5 (1.2)
15.7 (0.94)
Contralateral
8.7 (0.55)
10.81 (1.2)
8.2 (0.31)
7.2 (0.41)
8.9 (1.1)
9.1 (0.48)
Saccadic amplitudes are smaller in ipsilateral than in contralateral
direction in simulated (sHH) and real hemianopia (HH). In the normal
condition (N), there were no differences in amplitudes between saccades to the left or to the right (in table termed ipsilateral and contralateral respectively). In both HH groups, there are more fixations in
ipsilateral than in contralateral hemispace. We observed no such difference in the normal condition. Standard errors between brackets.
To assess the relative difficulty level of the patterns,
induced by the number of constituent dots, we expressed
the errors, search time, number of fixations, and length
of the scanpath as relative measures in dividing them by
the number of dots in the patterns. These parameters
hence indicate the performance per dot. We then performed a trend analysis by way of polynomial contrasts,
separately for the normal and simulation mode. If the
dot counting task functionally is performed in the same
manner in both groups (N and sHH), the same trends
should appear. If different or additional trends appear,
the number of dots assert a different influence on the
M.L.M. Tant et al. / Vision Research 42 (2002) 1339–1348
1345
Table 3
Trend analysis assessing the influence of the number of dots for the normal (N) and simulated condition (sHH) and for the patients (HH)
Trends
Linear
Quadratic
Error
N
sHH
HH
F ð1; 15Þ ¼ 13, P < 0:002
F ð1; 15Þ ¼ 19, P < 0:001
F ð1; 27Þ ¼ 38, P < 0:000
ns
F ð1; 15Þ ¼ 5, P < 0:046
#
Search time
N
sHH
HH
ns
F ð1; 15Þ ¼ 15, P < 0:001
F ð1; 27Þ ¼ 5, P < 0:026
ns
ns
F ð1; 27Þ ¼ 15, P < 0:001
Number of fixations
N
sHH
HH
F ð1; 15Þ ¼ 24, P < 0:000
F ð1; 15Þ ¼ 42, P < 0:000
ns
ns
F ð1; 15Þ ¼ 8, P < 0:012
F ð1; 27Þ ¼ 7, P < 0:015 #
Length of scanpath
N
sHH
HH
F ð1; 15Þ ¼ 11, P < 0:004
F ð1; 15Þ ¼ 50, P < 0:001
F ð1; 27Þ ¼ 56, P < 0:000
ns
ns
F ð1; 27Þ ¼ 29, P < 0:000 #
# indicates presence of higher-order trends. ns: not statistically significant.
performance, suggesting functionally different subcomponents or processes. The results are summarised in
Table 3. For all but one parameter, at least one additional trend was present in the simulation mode, compared to the normal mode. For the length of the
scanpath, only linear trends were present in both modes,
but in simulation mode being far more distinct (as evidenced by the F-value being almost five times higher).
This overall pattern was observed to be similar for the
patient data, except that occasionally also additional
higher-order trends were present (Table 3). For illustrative purposes, we plot the search time per dot in
function of the number of dots (Fig. 5). It can be observed that in sHH, there is relatively more time consumption for the patterns with less dots, as evidenced
by the linear trends. We similarly observed relatively
Fig. 5. Trends for relative search time in function of number of dots.
In the normal condition (N) no trend is present. In the simulated
condition (sHH), a linear trend was found. For the patients (HH), both
a linear and quadratic trend was present. Different trends suggest
different functional components.
increasing number of fixations and the length of the
scanpaths with decreasing number of dots. The reverse
pattern was found for the errors (not in the figures).
4. Discussion
We did observe hemianopic scanning behaviour in
healthy subjects without brain damage with an imposed
HH. This suggests that hemianopic scanning behaviour
is largely visually elicited, namely by the VFD. The
parallels between simulated and real HH are evidenced
by several findings. Firstly, we found elevated search
times, errors, number and duration of fixations and
length of scanpath (Fig. 1) in sHH compared to the
normal condition. We did not observe a main effect of
saccadic amplitude. These findings are in perfect concordance with previous findings reported by Zihl (1995,
1999, 2000) for real HH patients and confirmed by our
own patient data (Fig. 1). In sHH, we also found in
general longer fixation durations, which was not observed in our patient data (Fig. 1). Zihl (1999) reported
the mean fixation duration to be longer in some (‘‘impaired’’) and shorter in other (‘‘unimpaired’’) HH patients. We did not create these subgroups, and hence, in
Zihl’s view, are likely to have a ‘‘pooled’’ patient population in this respect. This could account for the total
null-effect of fixation duration in our patient group
(compared to the normal condition, Fig. 2). The finding
that we observe fixation duration increase, fortified by
the elevation of the other parameters, suggests that,
in many respects, our sHH subjects resemble the ‘‘impaired’’ HH patients. The observation that, for most
parameters, the performance in sHH is more deviant
(from the normal condition) than in HH, is agreement
with this suggestion. Alternatively, the HH patients did
have (more) time to adapt to their VFD, while for the
1346
M.L.M. Tant et al. / Vision Research 42 (2002) 1339–1348
sHH subjects, the acquisition of the VFD was very recent.
The second parallel between real and simulated HH
concerns the side of the VFD, which was previously
reported not to affect any oculomotor parameter in real
HH patients (Zihl, 1999). This was confirmed by our
patient data. We did however find left-sided HH patients to count less accurately than right-sided HH patients (Table 1). Zihl did not observe this difference. In
his paradigm, only one pattern (one trial) was presented
and subjects made no errors. Since we presented 29
trials, we were more likely to observe errors, and hence
our data are not optimally comparable in this respect.
The absence of effects of the side of the VFD in oculomotor parameters and the worse error performance in
left-sided HH was paralleled in sHH (Table 1). We did
additionally find a longer search time in left-sided sHH.
Hence, the results in sHH parallel the results in HH, in
that the side of the VFD does not differentially influence oculomotor performance. Left-sided sHH subjects,
however, tend to make more errors and need more time
than subjects with right-sided sHH. This was also the
case with respect to the errors for the HH patient group.
Thirdly, directional and hemispace analyses further
confirm the same pattern of results in sHH and HH
(Table 2) and are in concordance with previously reported findings. Differential hemifield distribution of the
fixations has previously been reported (e.g. Chedru et al.,
1974; Ishiai et al., 1987; Kerkhoff, 1999; Meienberg et al.,
1981; Zangemeister & Oechsner, 1996; Zihl, 1995, 1999,
2000). Also in our data, both sHH and HH fixated more
in the ipsilateral hemispace. Saccadic dysmetria and
more specifically ipsilateral hypometric saccades are
considered typical for hemianopic scanning (e.g. Chedru
et al., 1974; Ishiai et al., 1987; Meienberg et al., 1981;
Neetens, 1994; Zangemeister et al., 1982; Zangemeister
& Oechsner, 1996; Zihl, 2000). Our data, both for sHH
and HH, confirm saccadic amplitudes in ipsilateral direction to be smaller than in contralateral direction. As
in previous studies (e.g. Zihl, 1995) no effects of the side
of the VFD were found.
We were able to replicate all aspects known to be
typical for hemianopic scanning behaviour in simulated
HH. These healthy subjects did not suffer brain damage,
but were imposed with a simulated homonymous hemianopic VFD. It follows that the typical HH scanning
behaviour is largely due to the VFD (visually elicited)
and not to concomitant brain damage. To further explore underlying components in the scanning behaviour,
we performed the trend analysis in function of the
number of dots (Table 3). We assume that functionally
different components will result in different trends. In
the normal condition, we observed linear relationships
between the number of dots and the errors, number of
fixations and length of the scanpath per dot. The search
time per dot did not seem to be influenced by the
number of dots. These same trends appeared in sHH,
suggesting the same underlying mechanisms. However,
in nearly all parameters, also other (higher-order) trends
were observed, suggesting additional components. It is
reasonable to assume that these additional trends are
brought about by the simulated VFD, since this was the
only difference with the normal condition. These additional trends are suggested to be visually elicited. We
already argued that the scanning behaviour, displayed in
sHH, is in many respects identical to real HH. This was
partly confirmed by the trend analysis, in that most
trends present in sHH, also appeared in HH (Table 3).
Upon visual inspection of the trends, a paradox appeared. Fig. 5 shows that, relatively, the search time
increases with decreasing number of dots. A similar
pattern was observed for the number of fixations and the
length of the scanpath. We suggest that this pattern can
be explained as the time cost and effort for sHH to check
the whole visual field. When healthy subjects, in the
normal condition, fixate the centre of the screen, they
can parafoveally perceive the full screen and spatially
represent the dot pattern, as to effectively and economically organise their scanning pattern. They will cluster
neighbouring dots and devote more attention and time
to densely crowded parts of the screen, and no attention
to empty parts. This immediately available spatial representation is not available in sHH. In addition and as
a result, even (eventually apparent) empty parts of the
screen require visual exploration. It follows that, as the
result of the VFD, apparently easier configurations lead
to more dysfunction in sHH. In such a configuration, a
priori clustering and identification of unimportant parts
of the screen, leads to gain in effectivity in the normal
condition, contrary to sHH. We suggest that this is
partly at the basis of the visual slowness reported in HH:
a visual slowness brought about by the absence of an
immediately available spatial representation and the
need to standardly fully explore all parts of the screen,
also when this is, as ultimately appears, not necessary.
However, the additional trends in HH, compared to
sHH (e.g. quadratic in Fig. 5), suggest that still additional components are into play in real HH scanning.
Although different subjects comprise the sHH and HH
groups, and hence are not ideally comparable, this
suggests that also brain damage functionally influences
the scanning behaviour. More dots most likely summon
more visuo-spatial, memory and organisational functions. Brain damage is likely to affect (some of) these
functions, which are likely to interact reciprocally with
adequate visual exploration and proper cerebral representation of space, hence resulting in the appearance of
additional trends. Alternatively, the appearance of the
additional trends could be statistically induced by generally better performance by the HH subjects (compared
to sHH). As a result, HH subjects sooner perform
at their maximal effectivity, inducing flattening of the
M.L.M. Tant et al. / Vision Research 42 (2002) 1339–1348
(relative) performance curve, which will appear as additional trends. With this statistical alternative in mind,
we would like to indicate that the additional trends can
at least be suggestive for the additional impact of brain
damage on the HH scanning behaviour, but also that
psychometrically fully comparable data is needed to
support our suggestion.
In summary, we can conclude that HH scanning behaviour is largely visually elicited, namely by the VFD.
We further suggest that subtle interplay of brain-related
functions and the VFD complete real HH scanning.
Our interest in the effects of age were aroused by
Szlyk et al. (1993) who suggested that age-related losses,
when compounded by CVA-associated impairments, significantly influenced visuo-spatial driving related skills.
Such an age-related loss could be fluid intelligence, defined as the ability to new-problem solving. Our healthy
subjects were exposed to a new experience (sHH) for
which adaptive behaviour was required. We found this
compensation indeed to be worse for the search time
in the older age group. This age effect remained when
the log-transformed values were used, as suggested by
Cornelissen and Kooijman (2000).
Clear differences were also observed for the number of
fixations, in that the increase in the sHH condition was
far greater in the older age group (Fig. 3). These findings
were paralleled in the patient group. Hence, becoming
(simulated) hemianopic seems more disabling for older
subjects. It would follow that on second simulation, these
effects would weaken, since it then is no longer a new
situation. The absence of a learning effect seems to
contradict this, but since on both occasions the side of
the sHH was changed, it cannot be considered a valid test
for our hypothesis. We therefore explored the compensation effects within the simulation conditions by trial
order. The rationale is that the sHH is very new at the
first trial, but less with increasing trials. If the older age
group is less capable of new-problem solving, it should
be most prominent during the first trials. This is exactly
what we observed (Fig. 4). This pattern, although still
very prominent, was slightly reduced on second occasion
for the older age group (not in figure). For the younger
age group, patterns on both occasions were identical (not
in figure). We therefore conclude with Szlyk and colleagues that age-related processes are related to hemianopic compensation, but we add this to be the case even
if the disabilities are merely visually elicited and hence
are not specific for brain damaged subjects.
In conclusion, HH scanning behaviour, as assessed by
eye movement recordings during a dot counting task,
can largely be accounted for by the VFD. It follows that
most typical HH oculomotor dysfunctions, as for example ipsilateral hypometric saccades, do not result
from the brain damage but are visually elicited. Agerelated processes, in this case worse compensation to
these visually elicited disabilities, were apparent. The
1347
implication of this study is that at least some typical HH
disabilities and complaints, as for example slowness of
vision and prolongation of scanpaths, can no longer be
merely associated to brain damage, as they also do appear in subjects with sHH. A further implication would
be that these visually elicited impairments can be most
pronounced during (seemingly) the simpler situations.
This can have also ramifications both for rehabilitation
and diagnosis. Firstly, these results suggest that, at least
for some HH patients, more emphasis can be devoted to
visual than to cognitive components in rehabilitation.
Secondly, diagnosing higher-order visuo-spatial impairment can only occur in the light of concomitant lowerorder visual impairment.
Acknowledgements
We would like to thank E.M. Havik for collecting the
healthy subjects data. F.W. Cornelissen was supported
by Visio.
References
Chedru, F., Leblanc, M., & Lhermitte, F. (1974). Visual searching in
normal and brain-damaged subjects. Cortex, 9, 94–111.
Cornelissen, F. W., & Kooijman, A. C. (2000). Does age change the
distribution of visual attention? A comment on McCalley, Bouwhuis, and Juola (1995). Journals of Gerontology Series B Psychological Sciences and Social Sciences, 55, 187–190.
De Graaf, A., & Deelman, B. G. (1991). Cognitive screening test
[cognitive screening test]. Lisse: Swets and Zeitlinger.
De Renzi, E., Faglioni, P., & Sorgato, P. (1982). Modality-specific and
supramodal mechanisms of apraxia. Brain, 105, 301–312.
Deelman, B. G., Liebrand, W. B. G., Koning-Haanstra, M., & van der
Burg, W. (1987). SAN test. Lisse: Swets and Zeitlinger.
Folstein, M. F., Folstein, S. E., & McHugh, P. R. (1975). ‘‘Minimental state’’. A practical method for grading the cognitive state
of patients for the clinician. Journal of Psychiatric Research, 12,
189–198.
Ishiai, S., Furukawa, T., & Tsukagoshi, H. (1987). Eye-fixation
patterns in homonymous hemianopia and unilateral spatial neglect.
Neuropsychologia, 25, 675–679.
Kerkhoff, G. (1999). Restorative and compensatory therapy approaches in cerebral blindness—a review. Restorative Neurology
and Neuroscience, 15, 255–271.
Meienberg, O., Zangemeister, W. H., Rosenberg, M., Hoyt, W. F., &
Stark, L. (1981). Saccadic eye movement strategies in patients with
homonymous hemianopia. Annals of Neurology, 9, 537–544.
Neetens, A. (1994). Revalidation of homonymous hemianopic patients. In A. C. Kooijman, P. L. Looijestijn, J. A. Welling, & G. J.
van der Wildt (Eds.), Low vision. Research and new developments
in rehabilitation (pp. 296–300). Amsterdam: IOS Press.
Rybash, J. M., Roodin, P. A., & Hoyer, W. J. (1995). Adult
development and aging (3rd ed.). Boston, MA: McGraw-Hill.
Schaie, K. W., & Willis, S. L. (1993). Age difference patterns of
psychometric intelligence in adulthood: Generalizability within and
across ability domains. Psychology and Aging, 8, 44–55.
Szlyk, J. P., Brigell, M., & Seiple, W. (1993). Effects of age and
hemianopic visual field loss on driving. Optometry and Visual
Science, 70, 1031–1037.
1348
M.L.M. Tant et al. / Vision Research 42 (2002) 1339–1348
Tant, M. L. M., Kuks, J. B. M., Kooijman, A. C., Cornelissen, F. W.,
& Brouwer, W. H. (2002). Grey Scales uncover similar attentional
effects in homonymous hemianopia and visual hemi-neglect.
Neuropsychologia, 40, 1474–1481.
Zangemeister, W. H., Meienberg, O., Stark, L., & Hoyt, W. F. (1982).
Eye-head coordination in homonymous hemianopia. Journal of
Neurology, 226, 243–254.
Zangemeister, W. H., & Oechsner, U. (1996). Evidence for scanpaths
in hemianopic patients shown through string editing methods. In
W. H. Zangemeister, H. S. Stiehl, & C. Freksa (Eds.), Visual
attention and cognition (pp. 197–220). Amsterdam: Elsevier Science.
Zihl, J. (1994). Rehabilitation of visual impairments in patients with
brain damage. In A. C. Kooijman, P. L. Looijestijn, J. A. Welling,
& G. J. van der Wildt (Eds.), Low vision. Research and new
developments in rehabilitation (pp. 287–295). Amsterdam: IOS Press.
Zihl, J. (1995). Visual scanning behavior in patients with homonymous
hemianopia. Neuropsychologia, 33, 287–303.
Zihl, J. (1999). Oculomotor scanning performance in subjects with
homonymous visual field disorders. Visual Impairment Research, 1,
23–31.
Zihl, J. (2000). Rehabilitation of visual disorders after brain injury.
Hove, East Sussex: Psychology Press Ltd.