Neuropsychology Copyright 1996 by the American Psychological Association Inc.
1996, Vol. 10, No. 3, 385-395 0894-4105/96/$3.00
The Right Visual Field Advantage
and the Optimal Viewing Position Effect:
On the Relation Between Foveal and Parafoveal Word Recognition
Marc Brysbaert
University of Leuven, Leuven, Belgium
Franqoise Vitu
Universit6 Ren6 Descartes, Paris
Walter Schroyens
University of Leuven, Leuven, Belgium
Recent developments on the optimal viewing position (OVP) effect suggest that it may be caused
by the same factors that underlie the right visual field advantage in word recognition. This raises
the question of the relationship between foveal and parafoveal word recognition. Three
experiments are reported in which participants identified tachistoscopically presented words that
were presented randomly in foveal and parafoveal vision. The results show that both the OVP
effect and the right visual field advantage for word recognition are part of a larger extended OVP
curve that has the shape of a Gaussian distribution with the mode shifted to the left of the center of
the stimulus word. The shift of the distribution is a function of word length, but not of presentation
duration; it is also slightly moderated by the information value of word beginning and word end.
Tachistoscopic visual half field (VHF) studies are frequently
used to assess the laterality of cognitive functions. They are
based on the fact that stimuli presented in the left half of the
visual field (LVF) are initially projected to the right cerebral
hemisphere, and stimuli shown in the right half (RVF) are sent
to the left cerebral hemisphere. This anatomical feature has
been taken as support for the argument that LVF-RVF
differences are an index of asymmetric functioning of the two
cerebral hemispheres (for reviews see Bradshaw & Nettleton,
1983; Bryden, 1982; Hellige, 1993). Thus, the repeated finding
that words are recognized more easily in the RVF than in the
LVF is considered a consequence of left-hemisphere domi-
nance for language processing. Further evidence for this
position is obtained by finding that individuals with left hand
preference show a reduced RVF superiority for word recogni-
tion relative to persons with right hand preference (Kim, 1994;
but see Brysbaert, 1994c, for a more cautious account).
The interpretation of LVF-RVF differences in word process-
ing as an indication of laterality has not remained unchal-
lenged, however. At least three alternative explanations of the
RVF superiority have been proposed. The first considers the
Marc Brysbaert and Walter Schroyens, Department of Psychology,
University of Leuven, Leuven, Belgium; Franqoise Vitu, Laboratoire
de Psychologie Exp6rimentale, Centre National de Recherche Scienti-
fique Universit6 Ren6 Descartes, Paris, France.
The collaboration between the University of Leuven, Belgium, and
the Universit6 Ren6 Descartes of Paris was made possible by the
Nationaal Fonds voor Wetenschappelijk Onderzoek, a Tournesol
project between France and the Flemish Community of Belgium, and a
European Union Erasmus Studentship. The authors wish to thank
Philip Bryden and two anonymous reviewers for helpful comments on
an earlier draft.
Correspondence concerning this article should be addressed to
Marc Brysbaert, Department of Psychology, University of Leuven,
B-3000 Leuven, Belgium.
RVF superiority for word processing to be a consequence of
the word beginning being more informative than the word end.
This implies that a word presented to the right of the fixation
location has its most informative part nearer to the line of sight
than a word presented left of the fixation location (e.g., Kirsner
& Schwarz, 1986). Given that visual acuity drops steeply away
from the fixation location (Anstis, 1974), such an arrangement
favors the RVF for purely perceptual reasons. Laterality
researchers have argued against this perceptual hypothesis by
showing that the LVF-RVF difference is not influenced by the
information distribution within words (e.g., Bryden, 1986;
Bryden, Mondor, Loken, Ingleton, & Bergstrom, 1990; Brys-
baert & d'Ydewalle, 1990b; Eviatar & Zaidel, 1991). Thus,
Bryden et al. (1990) did not find a difference in VHF
asymmetry for words that could be completed more easily on
the basis of the last letters than on the basis of the first letters.
A second alternative explanation attributes the RVF superi-
ority for word recognition to the use of languages read from
left to right. The reading habits associated with these lan-
guages make it more easy to process a word right of fixation
than a word left of fixation; and indeed the first experiments
with languages read from right to left revealed a reversed VHF
asymmetry (e.g., Mishkin & Forgays, 1952; Orbach, 1952).
However, subsequent and better controlled studies using
native speakers reported RVF advantages both for languages
read leftward and for languages read vertically (for reviews,
see Faust, Kravetz, & Babkoff, 1993; Malamed & Zaidel,
1993).
A final alternative interpretation of the RVF advantage for
word recognition assigns it to (attentional) scanning factors
(e.g., Efron, 1990; Heron, 1957). It is assumed that for verbal
stimuli the RVF is scanned before the LVF, which because of
stimulus decay results in a RVF advantage. The problem with
the scanning hypothesis, however, is that so far no one has
explained why in word processing the RVF is scanned before
385
386 BRYSBAERT, VITU, AND SCHROYENS
the LVF and why this sequence changes for different kinds of
stimuli (e.g., why the RVF is scanned first for digit words and
second for bar graphs representing digits; Boles, 1986). In
addition, Kim and Levine (1991) showed that although atten-
tional factors accounted for about half of the interindividual
variance in VHF asymmetries, there still was a significant
additional effect of cerebral asymmetry (see also Brysbaert,
1994a). On the other hand, Mondor and Bryden (1992)
showed that a significant 55-ms RVF advantage in a lexical
decision task could be reduced to a nonsignificant 28-ms RVF
advantage if a valid attentional cue was given 50 ms prior to the
stimulus onset (but see Hardyck, Chiarello, Dronkers, &
Simpson, 1985).
Other studies have tried to validate the laterality account by
looking at VHF asymmetries for different kinds of words.
Experiments with split-brain patients (i.e., individuals in whom
the commissures have been sectioned to treat epilepsy) had
shown that the isolated right hemisphere contains a limited
capacity for word comprehension (e.g., Zaidel, 1983). Thus, it
was expected that the RVF superiority could differ as a
function of word frequency, imageability, concreteness, emo-
tionality, syntactic role, or word length. On the whole, the
results have been negative for word frequency, imageability,
concreteness, and emotionality. Low-frequency words do not
lead to a larger RVF superiority than high-frequency words
(e.g., Brysbaert & d'Ydewalle, 1990b; Koenig, Wetzel, &
Caramazza, 1992; McMullen & Bryden, 1987; Mohr, Pulver-
mtiller, & Zaidel, 1994). Highly imageable words induce the
same RVF advantage as words with low imageability (Boles,
1983; Hernandez, Nieto, & Barroso, 1992; McMullen &
Bryden, 1987). Abstract words do not differ from concrete
words (Boles, 1983; Eviatar, Menn, & Zaidel, 1990), and
emotional words give rise to the same VHF asymmetry as
neutral words (Eviatar & Zaidel, 1991).
On the other hand, there is some evidence that the RVF
advantage is larger for function words than for content words
(Bradley & Garrett, 1983; Chiarello & Nuding, 1987; Mohr et
al., 1994), and virtually all findings point to an increasing RVF
advantage with increasing word length (Bruyer & Janlin, 1989;
Brysbaert & d'Ydewalle, 1990b; Bub & Lewine, 1988; Ellis,
Young, & Anderson, 1988; Eviatar & Zaidel, 1991; Young &
Ellis, 1985; but see Bruyer & Ducarme, 1990). The latter is due
to the fact that the LVF score drops more steeply as a function
of word length than the RVF score. Some authors (e.g., Young
& Ellis, 1985) have interpreted this finding as evidence for two
different modes of word processing in the left and the right
cerebral hemisphere, whereas others (e.g., Schwartz, Montag-
ner,& Kirsner, 1987) saw it as evidence for the importance of
visual acuity in the emergence of VHF asymmetries (see the
perceptual hypothesis above).
In sum, laterality research has shown that parafoveally
presented words are more easily recognized if they are
presented to the right of the fixation location than if they are
presented to the left of the fixation location. This is a quite
robust phenomenon that generalizes across different kinds of
words with different information values of beginning and end,
and across languages with different reading directions. The
predominant interpretation of the RVF word advantage is that
it is due to structural characteristics of the visual pathways and
the cerebral cortex, although possible additional effects of
reading habits and attention allocation have been acknowl-
edged (e.g., Bryden, 1986; Bryden & Mondor, 1991; Hellige,
1986).
The Optimal Viewing Position Effect
in Foveal Word Recognition
In VHF experiments words are usually presented in parafo-
veal vision, with the nearest letter at a distance of minimally
one degree of visual angle from the fixation location. Under
normal reading conditions, one degree of visual angle agrees
with three to four letter spaces (Rayner & Pollatsek, 1989, p.
119). The precaution of parafoveal presentation is taken
because laterality researchers fear that foveal presentation of a
word may lead to bilateral projection, either because of small
eye movements or because the fovea itself is bilaterally
represented in the visual cortex (see e.g., McKeever, 1986;
Young, 1982). Recent findings on the optimal viewing position
(OVP) effect in foveal word recognition, however, suggest that
this concern may be incorrect.
In the early 1980s O'Regan and colleagues discovered that
the efficiency of foveal word recognition depends on the letter
fixated within the word (Nazir, O'Regan, & Jacobs, 1991;
O'Regan, 1981; O'Regan & Jacobs, 1992; O'Regan, L6vy-
Schoen, Pynte, & Brugaill~re, 1984). They found that word
processing was optimal when observers were allowed to fixate
between the first and the middle letter of the words. This was
true for word lengths ranging from 5 to 11 letters. Response
latencies increased and accuracy decreased if observers were
forced to look at the extremes of words, and more so when they
were forced to look at the end than when they were forced to
look at the beginning.
O'Regan and colleagues ascribed this pattern of results to a
combination of three factors. First, due to the decrease of
visual acuity outside the center of fixation (e.g., Anstis, 1974),
recognition is more difficult for words presented away from the
fixation location. This is even true for distances of less than one
degree; that is, for stimuli displayed well within the foveal area.
If the decrease of visual acuity were the only significant factor,
the OVP would lie in the middle of a word and processing time
would be a perfect U-shaped curve of the letter initially fixated
(unless one accepts that the decrease of acuity is smaller in the
RVF than in the LVF; Nazir et al., 1991). The faster process-
ing of a word after fixation on the first half than after fixation
on the last half was further explained by two additional factors,
namely the fact that in the language studied (French) words
are read from left to right, and the fact that most words can
more easily be guessed from their first letters than from their
last letters. These two linguistic factors lead to the so-called
word-beginning superiority effect.
The analogies between the explanation of the OVP effect
and the alternative explanations of the VHF asymmetry are
obvious. However, subsequent research has not confirmed
O'Regan's interpretation of the OVP effect entirely. First,
there is little evidence that words in a language read from left
to right are also processed in a left-to-right manner. This is
quite convincingly shown in a study by Radeau, Morais,
Mousty, Saerens, and Bertelson (1992), who looked at the
FOVEAL AND PARAFOVEAL WORD RECOGNITION 387
same language as O'Regan (i.e., French). Radeau et al.
examined the time course of lexical access in written-word
recognition by focusing on the uniqueness point (UP) effect. In
earlier studies (Radeau & Morais, 1990; Radeau, Mousty, &
Bertelson, 1989), they had shown that in spoken-word recogni-
tion, the recognition time of a word depended strongly on the
position of its UP; that is, the point at which the word can be
distinguished from other words that begin with the same
letters. Words with an early UP (e.g., aptitude) were recog-
nized more rapidly than words with a late UP (e.g., machin-
erie). In their study on printed word recognition, they repli-
cated the UP effect with an incremental presentation of the
word letters, but not with the normal presentation of the entire
word. The incremental presentation situation consisted of
progressively displaying the letters of a word on the screen,
starting from the left, with each new letter coming into position
at a speed designed to match the duration of the spoken word
delivery. For example, the word
spaghetti
would involve the
successive displays s,
sp, spa, spag, spagh, spaghe, spaghet,
spaghett,
and
spaghetti,
at approximately the same rate as the
word is commonly pronounced. The finding that the UP effect
is present in spoken word recognition and in the incremental
presentation situation, but not in the normal simultaneous
presentation situation, strongly suggests that printed French
words are not processed in left-to-right serial order, but that all
letters are processed in parallel. Further evidence against the
reading habit hypothesis comes from a study on the OVP effect
in Arabic words, which are read from right to left (Farid &
Grainger, in press). These words do not give rise to an OVP
shift to the right half of the words, but a much smaller shift
toward a symmetric function.
Holmes and O'Regan (1987) tried to manipulate the infor-
mation distribution within a word and looked at the influence
on the OVP effect. They selected in the French dictionary two
groups of 10 words: One group of words had the property of
being uniquely defined by their first six letters (e.g.,
perquisi-
tion, attroupement);
the other group was uniquely determined
by the last six letters (e.g.,
circonspecte, interrogatif).
Both
groups of words were matched pairwise for length (10-12
letters) and for frequency. The task was semantic judgment:
The test word appeared as the first word of a short phrase that
could make sense or not. Oculomotor behavior was measured
as a function of the imposed position where the eyes started
fixating in the word. The main finding was that although the
information distribution had some effect on the OVP pattern,
it did not reverse the phenomenon: No word-end superiority
effect was present for the words that were uniquely determined
by their last letters. This seems to indicate that the word-
beginning superiority effect is not completely due to the fact
that words on the average can be guessed better after knowl-
edge of the first letters than after knowledge of the final letters.
Similar results were reported by Farid and Grainger (in press)
for prefixed and suffixed French words, together with the
intriguing finding that the information distribution within a
word had more effect in Arabic than in French.
The insensitivity of the word-beginning superiority effect to
manipulations of reading habits and the informativeness of
word beginnings and ends made Brysbaert (1994b) wonder
whether the asymmetry of the OVP effect, just like the RVF
advantage, could be (partially) explained by hemispheric
specialization (see also Brysbaert & d'Ydewalle, 1988). Indi-
viduals were diagnosed as left-hemisphere or right-hemisphere
dominant for visual word processing on the basis of three VHF
tasks: word naming, object naming, and clock-face reading.
The first task was included for its obvious similarity with the
OVP manipulation. The second task was added to ensure that
the RVF superiority for word naming was not due to reading
habits (i.e., observers had to name line drawings of five
common objects that were symmetric around the vertical axis).
The clock-face reading task, finally, was intended to draw upon
processes of the nondominant cerebral hemisphere and was
expected to yield VHF superiorities opposite to those of the
first two tests. It was included to guarantee that the VHF
superiorities in the first tests were not due to attentional
imbalances.
On the basis of the three tests, 9 participants were diagnosed
as left-hemisphere dominant and 9 other participants as
right-hemisphere dominant. These individuals then named
foveally presented words after fixation on different letters. The
predictions of the hemispheric specialization hypothesis were
confirmed: Observers with left-hemisphere dominance prof-
ited more from fixations on the beginning of a word than
observers with right-hemisphere dominance, whereas the re-
verse was true for fixations on the end of a word. Thus,
left-hemisphere dominant participants had a significantly larger
word-beginning superiority effect than right-hemisphere domi-
nant participants. This was already shown to be true for
naming three-letter words that subtended a visual angle of less
than one degree.
The findings of Brysbaert (1994b) suggest that there is no
real difference between foveal and parafoveal word processing
and that cerebral asymmetry is an important factor in the
explanation of both the RVF advantage in parafoveal word
recognition and the OVP effect in foveal word recognition.
This adds further support to the growing body of doubts about
the bilateral cerebral representation of foveal vision in humans
(reviewed by Brysbaert, 1994b; see also Sugishita, Hamilton,
Sakuma, & Hemmi, 1994, for additional recent evidence).
The experiments described below were devised to further
investigate the relationship between the RVF advantage and
the OVP effect. They are based on the OVP paradigm.
Observers are asked to look at a fixation location, and words
are presented in such a way that the center of the word is
shifted to one or the other side. The major difference from the
classic OVP paradigm is that words can be shifted so much that
the participants no longer fixate on a letter of the word, but
either on some letter spaces in front of the word or on those
behind the word. We will call the new paradigm the extended
OVP paradigm (EOVP). It combines the classic OVP para-
digm and the usual VHF paradigm (with parafoveal word
presentation). By presenting words randomly in foveal and
parafoveal vision, a direct comparison between both presenta-
tion conditions can be made. In line with several previous OVP
experiments (e.g., Nazir et al., 1991; Nazir, Heller, & Suss-
mann, 1992), a perceptual identification task was used, with
identification accuracy as the dependent variable.
In the first experiment, the EOVP effect was established for
five-letter words over a relatively broad range of presentation
388
BRYSBAERT, VITU, AND SCHROYENS
1,0
~'~ 0.9
0.8
0.7
0.6
0.5
O
~ 0.4
= 0.3
Q
0.2
~ 0.1
0.0
Figure 1.
position relative
Experiment 1.
o----o 14 ms
H 28 ms
42 ms
H 56 ms
~ 70 ms
*-----* 84 ms
~ v . v :
-6 -4 -2 0 2 4 6
fixation position relative to word center
Word recognition probability as a function of fixation
to the word center and presentation duration:
times. In Experiment 2, the range of durations was limited and
the effect of word length was investigated. Finally, in Experi-
ment 3 we looked at the effect of different information values
for word beginning and word end.
Experiment 1
In Experiment 1, we examined the EOVP effect for French
words of five letters. This was done by shifting the word
relative to the fixation position, so that the observer either
looked at the center of the word, at the first or the last letter of
the word, or two and four letter positions in front of and
behind the word. In addition, presentation time was varied
between 14 and 84 ms. The dependent variable was the
percentage of words correctly identified.
Method
Participants.
Participants were 35 students from the Universit6
Ren6 Descartes and the Universit6 Catholique (Paris, France). All
were native French speakers, had normal or corrected-to-normal
vision, and were unaware of the research hypothesis. Twenty-seven
participants were female. Participation in the experiment was on a
voluntary basis.
Stimuli.
Stimuli were 280 French five-letter words with mean
frequency of 78 per million
(Trdsor de la languefranqaise,
1971). They
were divided in 35 lists of eight words matched for frequency.
Thirty-five stimulus lists were needed because of the orthogonal
variation of word position (seven levels) and presentation duration
(five levels). Word position was manipulated so that observers looked
either four letter positions in front of the word, two letter positions in
front of the word, on the first letter of the word, on the middle letter of
the word, on the last letter of the word, two letter positions behind the
word, or four letter positions behind the word. Presentation times
varied between 14 and 84 ms (i.e., multiples of the 70-Hz refresh rate
of the CRT monitor). Presentation times of 28, 42, 56, and 70 ms were
used for all word positions. The presentation time of 14 ms was not
used for the word positions in the extreme parafoveal conditions (i.e.,
four letter positions in front of and behind the word) because it was
expected that words would not be recognizable for these particular
pairs of duration and location. Instead, they were replaced by a 84-ms
presentation condition, making a total of 35 different location-
duration combinations. The lists of words were distributed over the
conditions according to a Latin square so that each list was seen in
each condition once (hence 35 participants). After the assignment of
presentation conditions to the individual words, the list was permuted
at random for each participant (Brysbaert, 1991). There were no
orthographic, semantic, or syntactic constraints for the inclusion of
words in the lists (e.g., verb forms could be included, together with
adjectives and nouns). Words were presented in MS-DOS text mode
using the default font and 80 x 25 character spacing.
Procedure.
Stimuli were presented on a CRT monitor connected to
an IBM compatible microcomputer. A character space subtended one
third of a centimeter horizontally, so that there were three character
spaces per degree of visual angle at a viewing distance of 57 cm. At the
beginning of a trial two vertically aligned lines appeared on the center
of the screen with a gap between them. Observers were asked to fixate
the gap. Five hundred milliseconds later, a word was shown on the text
line that coincided with the gap between the two vertical lines. Words
could be presented on any of the seven positions described above and
for all possible durations. Immediately after the stimulus time had
elapsed, the word was replaced by a mask, which consisted of the
ASCII Code 178 repeated five times and aligned horizontally. The
fixation lines remained visible throughout the total presentation time
of target and mask. The mask stayed on the screen for 800 ms, after
which the screen was blanked and a prompt appeared at the bottom of
the screen. Participants had to type in their response. They were
encouraged to guess if they were not sure about the correct answer.
The 280 test trials were preceded by 20 practice trials. Each participant
was seen individually in a quiet room. The experiment lasted about
1
hr.
Results
Figure 1 displays the percentage of correct word identifica-
tions as a function of presentation location and stimulus
duration (see also the left part of Table 1 for the exact values).
Table 1
Recognition Rates for the Duration~Position Conditions of Experiment 1, Lambda Indices, and Estimates of the Unknown Parameters
in the Nonlinear Regression Analysis
A(-2/+2) A(-4/+4) A(-6/+6) Prob OVP
SD
Duration -6 -4 -2 0 2 4 6 Est. 5% Est. 5% Est. 5% Est. 5% Est. 5% Est. 5%
14 ms -- .00 .04 .07 .02 .00 -- .90 1.08 -- -- .07 .01 -.37 .13 2.06 .17
28 ms .06 .22 .49 .57 .32 .08 .03 .71 .35 1.18 .52 .85 .87 .58 .03 -.63 .14 3.40 .19
42 ms .19 .44 .75 .81 .55 .25 .10 .92 .37 .82 .36 .72 .50 .82 .03 -.67 .15 4.32 .22
56 ms .31 .51 .79 .89 .65 .34 .13 .72 .38 .71 .35 1.13 .44 .87 .07 -.64 .32 4.87 .50
70 ms .41 .64 .86 .89 .71 .41 .21 .88 .43 .92 .35 .94 .38 .90 .03 -.85 .18 5.63 .31
84 ms .42 -- -- -- .27 -- -- .69 .36
Note.
Prob = probability of word recognition at the optimal viewing position (OVP); Est. = estimated; 5% = 5% confidence interval. Dashes
indicate that these conditions were not used.
FOVEAL AND PARAFOVEAL WORD RECOGNITION 389
Analyses of variance (ANOVAs) are not needed to see that
both variables had a profound effect: Words had higher
chances of being recognized if they were centered around the
line of sight and if they were presented for a longer period of
time. Furthermore, the functions relating recognition rate to
stimulus position are not symmetric: Chances of recognizing
the word are higher for fixations to the left of the word center
(i.e., on the first half of the word and in front of the word) than
for fixations to the right of the word center (i.e., on the last half
of the word and behind the word). This becomes especially
clear when the lambda index is calculated for the different
eccentricities, The lambda index is a common laterality index
of accuracy (Brysbaert & d'Ydewalle, 1990a; Sprott & Bryden,
1983) and is obtained by applying the following equation to the
values of Table 1:
A = ln(P
.....
t(RVF)/Pwrong(RVF)) -- ln(P
...... t(LVF)/Pwrong(LVF))
Three lambda indices can be calculated for the presentation
times ranging from 28 to 70 ms: one for the difference at
positions 2 and -2, one for positions 4 and -4, and one for
positions 6 and -6. Two more indices can be calculated: one
for positions 2 and -2 in the 14-ms presentation condition,
and another for positions 6 and -6 in the 84-ms presentation
condition. All values are tabulated in the middle part of Table
1, together with the 5% confidence intervals. I As can be seen,
all lambda indices fall within the same range. On the average,
lambda was .83 for positions -2/+2, .91 for positions -4/+4,
and .87 for positions -6/+6.
The lack of a reliable effect due to presentation time and
stimulus position can further be demonstrated by calculating a
4 x 3 ANOVA with repeated measures on the lambda indices
of the individual participants for the presentation times of 28,
42, 56, and 70 ms. This analysis failed to return significant
effects due to duration (F < 1,
MSE
= 1.55) and position
(F < 1, MSE
= 1.84). It should be noted, however, that the
ANOVA was handicapped because there were only eight
observations per cell per participant. This quite often required
a correction factor of +0.2 if no word had been recognized and
-.2 if all words had been recognized (these correction factors
were chosen because they returned a value of ln(.2/7.8) =
-3.66 when no word had been recognized and a value of
1n(7.8/.2) = +3.66 when all words had been identified).
The reason the lambda indices are virtually the same for all
eccentricities is that the identification probability as a function
of stimulus location is quite well captured by a Gaussian
distribution shifted to the left of the word center, as is shown in
Figure 2. The curves of Figure 2 were obtained by nonlinear
regression analyses (with the quasi-Newton and simplex algo-
rithms; Statsoft, 1991), using the following equation:
Pc ....
t
= prob * exp {-sqr[(eyepos -
OVP)/SD]},
in which
ecorrect
stands for the percentage of correct identifica-
tion (i.e., the values on the ordinate of Figures 1 and 2), eyepos
for the deviation between the line of sight and the center of the
word (i.e., the values on the abscissa of Figures 1 and 2), prob
for the probability of word recognition at the OVP, and
SD
for
the standard deviation of the Gaussian distribution. The right
O----O 14 ms
28 ms
1.0 ~ 42 ms
;~ 0.9 ,O' ",',II H L,l,, ab--! 56 ms
= 0s '"',','~t: ......... a:', % .--.70ms
0.7 ," ,' ,' ',, ',:q
Z o6
,, ,,,,,, ......... ° ,#,,>,
0.5
,' ,II,,' al' ",, ', ',",
:~O 0.4 9,' ,,,',el ,,,,' ',,,, ',, ',,,~,
,,",,,',,' ,' ', ",,-,',,,
021
.... '"' '""
0.t
, :::,'" ......
o
o
"",, '"":~I ":: ....
0.0 '....,,,,,!.. -. .......... ." . .......
o
.......... = 'T!!,,,,tl.'.'.'.'.!~d:::
-6 -4 -2 0 2 4 6
fixation position relative to word center
Figure 2.
Word recognition probability as a function of fixation
position and presentation duration (Experiment 1), together with the
best fitting Gaussian curves (see Table 1).
part of Table 1 gives the estimates and the .05 confidence
intervals for the three unknown variables: prob, OVP, and
SD
(these values were not estimated for the 84-ms presentation
duration because there were only two data points there; see
Figure 1).
Stimulus duration had a clear effect on prob (i.e., the height
of the curve), and on
SD
(the width of the curve), but not on
OVP (the shift of the curve to the left), at least not for the
three stimulus durations that are least susceptible to floor or
ceiling effects (i.e., presentation times of 28, 42, and 56 ms).
The fact that all recognition rates fall on normal distributions
that are shifted with the same amount to the left means that
the laterality scores for the different positions and durations
must be the same if they are properly corrected for the overall
recognition rate.
Discussion
Given the similarities between the OVP effect and the RVF
advantage for words, we hypothesized that both phenomena
might be due to the same origins (see also Brysbaert, 1994b).
Figures 1 and 2 show that this is indeed the case: Both the OVP
effect and the VHF asymmetries at different eccentricities
appear to be part of a more general EOVP curve that can be
described as a Gaussian curve slightly (i.e., about two thirds of
a letter) shifted to the left of the word center. There are no
apparent discontinuities between foveal and parafoveal word
recognition. In the next experiments we examined the effects
of word length and information distribution within words on
the EOVP curve.
1 The standard deviation of the lambda index is obtained with the
equation
SD
= sqrt(1/N+Rw +
1/N--RVF +
1/N+LvF + I/N--LVF)
in which N + and N- are the number of correct and incorrect
identifications (Brysbaert & d'Ydewalle, 1990b; Sprott & Bryden,
1983). Because the equation assumes independence of observations,
the calculated values of the present studies are likely to be slight
underestimates (because not all observations came from different
subjects). That is why the estimates are always given in parallel with
analyses of variance based on repeated measures.