Pergamon PII: S0042-6989(97)00287-3

Vision Res., Vol. 38, No. 9, pp. 1295-1302, 1998 © 1998 Elsevier Science Ltd. All rights reserved

Printed in Great Britain 0042-6989/98 $19.00 + 0.00

Adjustment of Fixation Duration in Visual Search IGNACE TH. C. HOOGE,*t CASPER J. ERKELENS*

Received 30 August 1996; in revised form 11 April 1997; in final form 14 July 1997

To investigate whether fixation durations are adjusted to the duration of a foveal analysis task, we designed a search task in which each stimulus element yielded information about the position of the target. We asked subjects to look for the target by making eye movements in the direction indicated by each stimulus element. We explicitly asked the subjects to make the eye movements in the correct direction, but they did not always do this. They made only 65-80% of the eye movements in directions indicated by the stimulus elements. From these results we conclude that fixation durations are not solely determined by the immediate visual stimulus and that subjects encounter difficulties when trying to increase fixation durations to values that would enable them to direct saccades accurately. In a second experiment we shortened the presentation time in order to provide an incentive for the subjects to speed up search. Shortening the presentation time did not affect fixation duration. Therefore, we suggest that fixation duration is controlled by a mechanism that uses estimations of the foveal analysis time of previous fixated stimulus elements. © 1998 Elsevier Science Ltd. All rights reserved.

Vvisual search Saccades Control of fixation duration

INTRODUCTION

Subjects make many eye movements when searching in large displays (Bloomfield, 1972). Periods between saccadic eye movements are called fixations (intersacca- dic intervals). Mean durations of fixation are of the order of 200-500 msec (Enoch, 1959; Ford, White & Lichten- stein, 1959; Gould, 1967, 1973; Luria & Strauss, 1975; Engel, 1977; Widdel & Kaster, 1981; Jacobs, 1986). During fixation, the foveal target is analysed, the peripheral field is sampled and the next eye movement has to be prepared (Viviani, 1990). The preparation of a saccadic eye movement has been estimated to take approx. 135 msec (Becker & Jtirgens, 1979). The analysis of a fixated stimulus takes approx. 100-300msec (Eriksen & Eriksen, 1971; Salthouse, Ellis, Diener & Somberg, 1981) depending on the properties of the stimulus elements and task.

In the search experiments of Engel (1977), Gould (1973), Hooge, Boessenkool and Erkelens (1996) and Hooge and Erkelens (1996), subjects were asked to fixate a target if it was found. They reported that subjects often fixated the target, made an eye movement away from the target and subsequently returned to the target (return saccade). Hooge and Erkelens (1996) also reported missed targets. This means that subjects fixated the target and subsequently continued to search, which implies that fixation duration was too short to recognize

*Helmholtz Instituut, Utrecht University, Princetonplein 5, 3584 CC, Utrecht, The Netherlands.

tTo whom all correspondence should be addressed at the Department of Physiology, Erasmus University, P.O. Box 1738, 3000 DR Rotterdam, The Netherlands [Email: hooge@fysl.fgg.eur.nl].

the target. The occurrence of return saccades and missed targets show that saccade preparation may start before the foveal analysis is completed. Thus, completed foveal analysis is not necessarily the trigger for the subsequent saccade.

The present paper deals with the relationship between visual analysis and saccadic programming, in particular the adjustment of fixation duration to the level of completion of the foveal analysis task. Fixation durations increase with the difficulty of the foveal task (Gould, 1967, 1973; Moffit, 1980; Jacobs, 1986; Jacobs & O'Regan, 1987; Erkelens & Hooge, 1996). However, not much is known about the overlap in time of saccade programming and foveal analysis (Viviani, 1990). Subjects may use different strategies to adjust the overlap between saccade programming and foveal analysis. For example, if subjects allow a large overlap in time, search becomes fast because saccadic dead-time ("idle time": Russo, 1978) is shortened, but mistakes become more likely (Hooge et al., 1996). Can subjects select a specific overlap between foveal analysis and saccade program- ming?

To investigate the relationship between saccade programming and foveal analysis, we designed a new type of search task in which each stimulus element yielded information about the position of the target. We asked subjects to look for the target by making eye movements in the direction indicated by each stimulus element. We analysed saccadic direction and fixation duration. If subjects are able to adjust their fixation durations adequately to the time needed to perform the foveal analysis task, then saccades will be made in the correct direction. Adequate adjustment means that the

1295

1296 I. TH. C. HOOGE and C. J. ERKELENS

A B

C C 0 0 0 0 0 0 0 0 0 C

C C C O O O O O O O O O

C C C C 0 0 0 0 0 0 0 0

C C C 0 0 0 0 0 0 C 0 0

O C O 0 0 0 0 0 0 0 0 0

C C O O O O O O C O O O

FIGURE 1. Stimulus. (A) Shows an example of the stimulus used in the direction-coded condition. Gaps in the "C"s indicate the direction in which the target can be found. (B) Shows an example of a stimulus used in the uncoded condition. Orientation of

each C was chosen randomly from the directions up, down, left and right.

fixation duration is long enough to allow the result of the foveal analysis to be available for saccade programming before saccadic dead-time starts.

EXPERIMENT 1

Methods

Subjects. Three male subjects (the authors, IH and CE, and one naive subject, CG) participated in the experi- ments (aged 30-46 years). None of them showed any visual or oculomotor pathologies other than refraction anomalies. The subjects had normal or corrected to normal vision. All the subjects were experienced in wearing scleral coils for eye movement recording. Subjects CG and CE had no experience in doing this task. Subject IH had some experience because he took part in the pilot experiments.

Apparatus. Subjects sat in front of a large screen at a distance of 1.50 m. The experimental room was com- pletely darkened. Chin and forehead rest were used to restrict head movements. Stimuli were generated by an Apple Macintosh llci personal computer (refresh rate 66.7 Hz, resolution 640 × 480 pixels) and rear-projected on a translucent screen by a Barco Data 800 projection television. Only the green tube was used. The screen measured 1.9 × 2.4 m. Eye movements of the right eye were measured with an induction coil mounted in a scleral annulus in an a.c. magnetic field. This method was first described by Robinson (1963) and refined by Collewijn, van der Mark and Jansen (1975). The dynamic range of the recording system was from d.c. to 100 Hz (3 dB down), with a noise level of less than 10 min arc. Deviation from linearity was less than 1% over a range of ± 2 0 deg. The horizontal and vertical eye positions of the fight eye were measured at a sampling rate of 500 Hz with a National Instruments 12 bits NB-MIO16h analogue to digital converter. Data were stored on disk for further analysis.

Procedure. The stimulus for the search task contained 36 elements placed on an invisible hexagonal grid subtending 35 × 27.5 deg (Fig. 1). The distance between

the centres of adjacent stimulus elements was 6.2 deg. Stimulus elements had diameters of 2.1 deg.

We used two experimental conditions: the direction- coded condition and the uncoded condition. In the direction-coded conditions the stimulus contained one circle (the target) and 35 Cs. The size of the gap in the Cs measured 0.15, 0.30, or 0.60 deg in separate sessions of 75 trials. Orientation of each C (up, down, left or right) was chosen so that the gap faced the target [Fig. I(A)]. Subjects were asked to find the circle by maintaining fixation on it until the end of the trial. We asked subjects to make each eye movement in the direction of the gap in the C and asked the subjects to do their best to follow the indicated directions.

The uncoded condition was the same, except the size of the gap was always 0.30 deg and the orientation of each C (up, down, left or right) was chosen randomly [Fig. I(B)]. Subjects were not told to use the gaps to search, but simply to maintain fixation on the circle until the end of the trial.

Trials started with the presentation of a circle for 1 sec. Subjects were asked to fixate this circle. After 1 sec, another circle was randomly placed on one of the 36 stimulus element positions. Subjects were again asked to make a saccade towards this circle. The saccade towards the second circle was detected on-line. Immediately after the detection of the saccade the circle was replaced by the complete display of 36 stimulus elements. The search display remained on for 7.5 sec.

Data were analysed off-line by a computer program that ran on an Apple Macintosh computer. Saccades were detected by a velocity threshold (100deg/sec). After detection of the saccade the program searched for the onset and offset by using a velocity threshold of 25 deg/ sec. Onsets and offsets were used to compute fixation durations and fixation positions. An amplitude threshold of 2.1 deg was used to remove small corrective saccades.

We define search time as the period between stimulus onset and fixation of the target. Search times measured according to this definition will slightly underestimate real search times, defined by the time needed to find the target, because the target has not been found by the

ADJUSTMENT OF FIXATION DURATION IN VISUAL SEARCH 1297

©

O

O

;>

1'01 CE / 0.8

0.6

0"41 0.2

0.0 . . . .

1.0" CG

0.8-

0.6 ~

0.4 ~

0.2-

0A3

1.0" IH

0.8

0.6

0.4

q I 0.2 - - Uncoded

0.0 0 2500 5000 7500

Search time (ms) FIGURE 2. Cumulative fraction of correct trials vs search time. Thick lines denote the direction-coded condition. Thin lines denote the

uncoded condition•

condition, performance in the coded and the uncoded condition were compared only in the case of stimuli having a gap-size of 0.30 deg.

The steeper slopes of the cumulative curves indicated that search times were shorter in the direction-coded condition (Fig. 2). For subjects CE and CG we found that search was approximately twice as fast in the direction- coded condition. For subject IH search times decreased slightly in the direction-coded condition. The faster speed of directed coded search shows that subjects used direction information to find the target.

In sparse displays, fixation duration depends mainly on the difficulty of the foveal task (Moffit, 1980). A sparse display is defined as a display in which only one stimulus element can generally be analysed during a fixation. Under the direction-coded condition the position of the gap in the C has to be found. Thus, the foveal task is more difficult under the coded condition than under the uncoded condition. Therefore, we expect fixation dura- tions to be longer in the direction-coded than in the uncoded condition. Figure 3 shows that mean fixation duration was longer under the direction-coded condition than the uncoded condition. For subject CE differences between fixation durations of both conditions were small and were not significant. For subjects IH and CG, differences between fixation durations in both conditions were significant (Fig. 3).

Saccade direction. To obtain insight into how subjects scanned the stimuli, we determined the direction of each saccade that had an amplitude larger than 2.1 deg (size of the stimulus elements). Saccade direction is defined as the angle between the horizontal and the line that connects the start and endpoint of each saccade. Figure 4 depicts three representative histograms of saccade directions. Most saccades were made in the directions 60, 120, 180, 240, 300 and 360 deg. Because the stimulus

300-

beginning of target fixation. From the saccade direction "~ and the orientation of the C from which an eye movement ,~ was made, we determine whether this eye movement was =o 2 0 0 made in the correct direction. For example, if the gap in "4 the fixated C was on the right side, subjects had to make an eye movement to the right. An eye movement was judged to be correct if it was made between - 4 5 deg and .~ 100 45 deg of the correct direction.

Results

Coded vs uncoded search. Subjects used the direction information that was available under the direction-coded condition; we expected search performance to be better in the direction-coded condition than in the uncoded condition. To compare performance in the coded and uncoded conditions, we plotted the cumulative number of correct trials against search time (Fig. 2). Since we used only Cs having a gap-size of 0.30 deg under the uncoded

0

/ / / / / I l l I l l l I l l l I I I I I I / I I I I I I I I I I I I I I I I I t 111 1111 I I I I I I I I I I I 1 I I I I I I I I J i l l I I I I l i l l I I I 1 I I I I

I [ ~ c o d e d

--r- I ~ ] uncoded_.r_.

/ / / J

V./x ~ , / / / /

~ , " f i / / / / ¢ t l l J

f f J ~ , / l i t

C E C G IH

Subjec t

¢ z i ~

¢ / / I

z / z ~

¢ z 1 1

/ / / 1 z z z l

f t . - i

f i g ,

c z z l

z z / /

g / z . "

FIGURE 3. Fixation durations. White bars denote fixation durations of the direction-coded condition. Diagonally dashed bars denote fixation durations of the uncoded condition. Error bars denote errors of the

mean.

1298 1. TH. C. HOOGE and C. J. ERKELENS

©

tD

C)

©

tD

O

tD

Z

100"

80"

60"

40-

20"

0 0

100-

80"

60"

40"

20-

0 0

CE 100 IH

80

6 0 IA AI j40 0 . , , , . ! , .~.L, d , : ,~.AL. 0,

60 120 180 240 300 360 0

CG

60 120 180 240 300 360

D i r e c t i o n ( d e g )

60 120 180 240 300 360

90

210 ~ 330

270

FIGURE 4. Histograms of the saccade directions• The figure shows three representative examples of histograms of saccade directions• Gap-size was 0.3 deg. The figure in the bottom right-hand comer depicts the correspondence between directions and

stimulus elements.

elements were placed on a hexagonal grid, these directions correspond to saccades made to adjacent stimulus elements (Fig. 4). We found a few saccades to stimulus elements that were not adjacent (e.g. the 270 deg direction). The peaks in the histogram are narrow. This means that subjects were able to direct their saccades accurately to adjacent stimulus elements.

Fixation durations. What is the relationship between the fixation duration and the difficulty of the foveal task? In an earlier experiment we found that fixation durations increased with decreasing gap-size (Hooge & Erkelens, 1996) and therefore, we expected this to occur under the direction-coded condition. Figure 5 shows that as expected, mean fixation duration in the direction-coded condition increased with decreasing gap-size.

Fixation durations and saccade directions. If subjects adjusted their fixation durations to the time needed for both foveal analysis and programming of the correct direction, then eye movements should have corresponded to the direction of the gap. Figure 6 shows the fractions of saccades made in the correct direction, indicated by the gap. Fractions of correctly directed eye movements ranged from 0.65 to 0.80. These fractions are far above chance level, which is close to 0.25. Thus, subjects did not restrict all saccadic eye movements in directions indicated by foveally presented information.

Recent experiments have shown that foveal analysis and eye movement programming may overlap in time (Hooge & Erkelens, 1996; Hooge et al., 1996). If the

result of the foveal analysis is not available soon enough, we assume that it is not used for programming the direction of the next saccade. To check the relationship between fixation duration and saccade direction, we plotted the mean durations of fixations preceding incorrectly and correctly directed eye movements (Fig. 7). For all subjects, durations of fixations preceding incorrect eye movements (diagonally dashed bars) were

300-

280- -

= 260- o

"~ 240-

o 220-

200-

----O-- C E

C G

180 . . . . . . . . . . 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7

Gap-s ize (deg)

FIGURE 5. Fixation duration vs gap-size• White bars denote the mean fixation durations of the direction-coded conditions. Error bars are the

standard errors of the mean.

ADJUSTMENT OF FIXATION DURATION IN VISUAL SEARCH 1299

.,..~

1.0

0 . 8 "

0 . 6

0 . 4

0.2" - - - C h a n c e - + IH

- CG w

----<)--- CE

. 0 i i i i i

0 .00 0 .15 0 .30 0.45 0 .60 0.75

Gap-size (deg)

FIGURE 6. Fractions of correctly directed eye movements. Fraction of correctly directed eye movements is defined as the sum of the correctly directed eye movements divided by the total number of eye

movements.

shorter than durations of fixations before correct eye movements (white bars). Differences were largest for subject IH (50-100 msec). For subject CG, the duration of fixation preceding incorrectly directed saccades increased with decreasing gap-size and were 25 -70 msec shorter than durations of fixations preceding correctly directed saccades. The small differences between the durations of fixations preceding incorrectly and correctly directed saccades for subject CE were not significant.

EXPERIMENT 2

To test whether subjects change their scanning rate if they are pressed for time, we shortened the presentation time. If they were to scan at a higher fixation rate, we would expect the fraction of eye movements made in the correct direction to be smaller, because there was less time available to analyse the stimulus elements. Experi- ment 2 was the same as the direction-coded condition of Experiment 1, except presentation time was reduced to 1.5, 2.25 or 3 sec in separate sessions. We chose the presentation time to be 3 sec or less because the longest time subjects needed to find the target was approximately 4 sec (Fig. 2). Subjects IH and CE took part in this experiment.

Results

Did subjects search faster if the presentation time was shortened? Fig. 8 depicts the cumulative fraction of targets found against search time. For both subjects, shortening the presentation time did not affect search performance. The three cumulative curves almost coin- cided. Fixation durations were independent of presenta- tion time (Fig. 9) as was the fraction of eye movements made in the correct directions (Fig. 10). The results of Experiment 2 show that reduced presentation time did not reduce fixation duration.

DISCUSSION

Fixation durations and saccade directions

Results of this experiment show that subjects set their fixation duration in such a way that in most cases they used the result of the foveal analysis for programming saccades. The majority of the saccades were made in the directions of the gaps. However, the subjects also made saccades in other directions. These saccades were directed mainly to adjacent stimulus elements (Fig. 4). For two of the three subjects, fixation durations preceding correctly directed eye movements were longer than fixation durations preceding incorrectly directed eye movements. It is difficult to draw conclusions about individual analysis times from the relationship between fixation duration and saccade direction, because we do not know the distribution of saccade programming times.

350"

300"

250"

200"

150"

100"

50"

0

, ~ 350"

300- O

" ~ 250-.

200" • 150-

O 100- . t m l , l

¢~ 50-

"~"~ 0 350-

300-

250-

200i

150- J

100"

50-

0

CE

"--t- - - / -'¢ -¢ I f / / , c / J

, ' / i I , c / , c / /

t ' f , c ¢ , / /

t ' / J , c / / / / / / / / , c / / , c / / ¢ , / /

CG - ' l "

/ / / , C / /

f / J f l J t ' / / t ' / / f / /

t ' / . / i ' l l , c f / d ' / ' , ~

i / / f f / f / / f . / ' j

f J /

IH

, s l , / / A f f ~

/ / A / / A / / A

/ / A / / A / / A

0.15

" T -

/ / A

, t , C A , C / A / . c A / , C A , C / A t ' / A / / A ¢ ' / A

/ / A / . c A / f A

t ' / A

"--I'-

¢ / z , t / /

, c / / / / , c / / / , c / / , t / /

, c . ¢ ' / f / / / J , c / . t . ¢ / / / , t , - j , t , c / , ¢ / /

0.3

l c- ct t Incorrect [

- - - t -

I

D 0.6

Gap-size (deg) FIGURE 7. Fixation duration vs gap-size. White bars denote durations of fixations preceding saccades in the correct direction. Diagonally dashed bars denote durations of fixations preceding saccades in the

incorrect direction•

1300 I. TH. C. HOOGE and C. J. ERKELENS

1.0 q.., ©

= 0 . 8 o

~ o.6

• 0.4

E 0.2 L)

0.0 0

CE I / - .~"-¢'-- |

1000 2000

IH I _.:....,---~ --1

/::& ! I ! I

! - - 2 2 5 0 ms

f : - -1500ms

~ J I I

3000 0 1000 2000 3000

Search time (ms)

FIGURE 8. Cumulative fraction of correct trials vs search time. Thick line denotes a presentation time of 1.5 sec. Intermediate line denotes a presentation time of 2.25 sec. Thin line denotes a presentation time of 3.0 sec.

The results suggest that subjects adopted a strategy of setting fixation duration to a value too short to allow complete foveal analysis of the directional cue. Thus, eye movements were in error.

Evidence for indirect control

From a recent experiment (Hooge & Erkelens, 1996) we concluded that the control of fixation duration is indirect, because subjects made many return saccades (see Introduction). The occurrence of return saccades implies that co-operation between the visual and the motor system is not the same as it is in a process- monitoring model (Rayner, 1978). In a process-monitor- ing model, eye movement programming always starts after the visual analysis is complete. However, that experiment (Hooge & Erkelens, 1996) did not rule out the possibility that direct control (process-monitoring) of saccades may occur if a particular task requires it. Therefore, we explicitly asked subjects to make eye movements in the correct direction. To perform this task

260"

240

= o

~ 22O

o

200 . , . . .~

0 CE m I H

180 , , , 750 1500 2250 3000

Presenta t ion t ime (ms)

FIGURE 9. Fixation duration vs presentation time.

i

3750

correctly, fixation duration has to be long enough to allow the motor system to use the result of the foveal analysis. We found that subjects did not always allow foveal analysis to be complete. A process-monitoring model (Rayner, 1978) predicts eye movements only in the correct directions, because in the model eye movement programming starts after the foveal analysis has been completed. On the basis of our new experiment we reject the process-monitoring hypothesis.

Findlay (1995) and McPeek and Nakayama (1995) also found evidence that saccade programming starts before visual analysis is complete. Findlay (1995) used stimuli consisting of seven green circles and one red circle placed in a circular arrangement. Subjects were asked to make an eye movement to the red target. In 25% of trials the stimulus contained two red targets instead of one. The two red targets were adjacent to each other or separated by one green distractor. When stimuli contained two targets, subjects made many eye movements to inter- mediate positions. Latencies of eye movements that ended between the two red targets were short (185 msec instead of 300 msec for correctly directed eye move- ments). McPeek and Nakayama (1995) used a pop-out visual search task consisting of three red and green stimulus elements (one red or one green target among two green or two red distractors). Subjects had to respond by making a saccadic eye movement from a fixation cross to the odd coloured target. They found that subjects sometimes make eye movements to a distractor. Incorrect saccades were observed much more frequently when the target colour of the current trial differed from the colour of the target in the previous trial. The latencies of these incorrect saccades were often short and were usually followed by corrective saccades with extremely short latencies (10-100msec). The results of McPeek and Nakayama (1995) and Findlay (1995) suggest that the timing of saccades is not triggered by the visual processes that occur during the preceding fixation.

Voluntary control of fixation duration

Why did subjects make saccades in incorrect direc-

ADJUSTMENT OF FIXATION DURATION IN VISUAL SEARCH 1301

©

o o

o

o

1° t 0.8

0 . 6

0.4"

0 . 2

O---- CE IH

_..------O- O O---"

. . . . Chance- - -

° 0 i i i i

750 1500 2250 3000 3750

Presentation t ime (ms)

FIGURE 10. Fractions of correctly directed saccades vs presentation time.

tions? A strategy to direct all saccades correctly is to scan at such a rate that fixation durations are long enough to ensure that saccade programming and foveal analysis do not overlap in time. In other words: subjects could have carried out the search task much more slowly. Presenta- tion time was long enough to allow the subjects to scan at a lower fixation rate (Fig. 2). However, subjects did not extend their fixation durations to values that would have enabled them to programme their eye movements in the correct directions. Perhaps subjects cannot voluntarily extend their fixation durations during a cognitive task such as visual search. We also found that subjects did not reduce fixation duration when presentation time was shortened. Our findings corroborate results obtained by Widdel and Kaster (1981). From their experiments they concluded that fixation duration is not a variable that can be easily influenced by practice.

Control of timing of saccades We conclude that the control of fixation duration is

indirect. This means that the result of the foveal analysis does not act as a trigger for eye movement programming. We also found that subjects were not able to slow down search. Thus, subjects were not able to control the durations of their fixations voluntarily. An important question is then: how are fixation durations controlled? We suggest that the timing of saccades during search is controlled by a mechanism that estimates how much time is needed for the foveal discrimination task. The estimation of the duration of foveal analysis occurs during subsequent fixations. The duration of the next fixation is calculated and pre-programmed from the estimated value of the foveal analysis time.

Hooge and Erkelens (1996) found evidence in favour of this theory. They used a search task in which subjects had to find a circle in a field of six "C"s. Gap-size was varied or constant between trials. When gap-size was constant, fixation durations were longer for small gap- sizes. Gap-size was less influential when it was varied

from trial to trial. This implies that history plays a role in the adjustment of fixation durations.

A model of indirect control of fixation durations, which uses estimations of visual processing times gathered during previous fixations, predicts that if the difficulty of the visual task remains the same during subsequent fixations, the fixation duration will adapt to a constant value. McPeek and Nakayama (1995) found evidence for this idea. They used a pop-out visual search task consisting of three stimulus elements. Subjects had to respond to the odd coloured target (one red or one green target among two green or two red distractors) by making a saccadic eye movement. The experiment was carried out under two conditions. In the blocked condition, target colour was similar in all trials. In the mixed condition the target could be green or red among red and green distractors. Under the blocked condition latencies were shorter than under the mixed condition. Under the mixed condition, as a function of run length of the same target colour, mean saccadic latencies decreased from 225 to 202 msec for one subject, from 212 to 195 msec for a second subject and from 231 to 184 msec for a third subject. McPeek and Nakayama (1995) showed that the adjustment of fixation durations is made according to the difficulty of the visual analysis on previous trials.

The data for the present experiment do not provide a complete explanation of the observed fixation durations and errors made by the subjects. The main reason is that we do not have enough knowledge about the underlying mechanisms, such as the analysis process and saccade programming.

The main conclusions are: (1) single fixation durations do not reflect the amount of time needed for complete processing of foveal information; (2) subjects do not extend their fixation durations; and (3) reducing the presentation time also does not affect fixation durations. We suggest that fixation durations are controlled indirectly by a mechanism that uses estimations of the foveal analysis time of previous fixated stimulus elements.

REFERENCES

Becker, W. & Jtirgens, R. (1979). An analysis of the saccadic system by means of double step stimuli. Vision Research, 19, 967-983.

Bloomfield, J. R. (1972). Visual search in complex fields: size differences between target disc and surrounding discs. Human Factors, 142, 139-148.

Collewijn, H., van der Mark, F. & Jansen, T. C. (1975). Precise recording of human eye movement. Vision Research, 15, 447-450.

Engel, F. L. (1977). Visual conspicuity: visual search and fixation tendencies of the eye. Vision Research, 17, 95-108.

Enoch, J. M. (1959). Effect of the size of a complex display upon visual search. Journal of the Optical Society of America, 493, 208-286.

Eriksen, C. W. & Eriksen, B. A. (1971). Visual perception processes rates and backward and forward masking. Journal of Experimental Psychology, 89, 306-313.

Erkelens, C. J. & Hooge, I. Th. C. (1996). The role of peripheral vision in visual search. Journal of Videology, 1, 1-8.

Findlay, J. M. (1995). Eye movements and peripheral vision. Optometry and Vision Science, 72, 461-466.

Ford, A., White, C. T. & Lichtenstein, M. (1959). Analysis of eye

1302 I. TH. C. HOOGE and C. J. ERKELENS

movements during free search. Journal of the Optical Society of America, 49, 287-292.

Gould, J. D. (1967). Pattern recognition and eye movement parameters. Perception and Psychophysics, 29, 399-407.

Gould, J. D. (1973). Eye movements during visual search and memory search. Journal of Experimental Psychology, 98, 184-195.

Hooge, I. Th. C. & Erkelens, C. J. (1996). Control of fixation during a simple search task. Perception and Psychophysics, 58, 969-976.

Hooge, 1. Th. C., Boessenkool, J. J. & Erkelens, C. J. (1996). Stimulus analysis times measured from saccadic responses. In A. M. L., Kappers, C. J., Overbeeke, G. J. F. Smets & P. J. Stappers (Eds), Studies in ecological psychology: Proceedings of the Fourth European Workshop on Ecological Perception (pp. 37-40).

Jacobs, A. M. (1986). Eye-movement control in visual search: how direct is visual span control? Perception and Psychophysics, 391, 47-58.

Jacobs, A. M. & O'Regan, J. K. (1987). Spatial and/or temporal adjustments of scanning behaviour to visibility changes. Acta Psychologica, 65, 133-146.

Luria, S. M. & Strauss, M. S. (1975). Eye movements during search for coded and uncoded targets. Perception and Psychophysics, 17, 303- 308.

McPeek, R. M. & Nakayama, K. (1995). Repetition of target color affects saccadic latency and accuracy. Eighth European Conference on Eye Movements, Abstract Programme, 65.

Moffit, K. (1980). Evaluation of the fixation duration in visual search. Perception and Psychophysics, 274, 370-372.

Rayner, K. (1978). Eye movements in reading and information processing. Psychological Bulletin, 85, 618-660.

Robinson, D. A. (1963). A method of measuring eye movement using a scleral search coil in a magnetic field. IEEE Transactions in Biomedical Electronics, 10, 137-145.

Russo, J. (1978). Adaptation of cognitive processes to the eye movements system. In J. W. Senders & R. A. Monty (Eds), Eve movements and the higher psychological .functions (pp. 89-112). Hillsdale, NJ: Lawrence Erlbaum.

Salthouse, T. A., Ellis, C. L., Diener, D. C. & Somberg, B. L. (1981). Stimulus processing during eye fixation. Journal of Experimental Psychology: Human Perception and Per]brmance, 73, 611-623.

Viviani, P. (1990). Eye movements in visual search: cognitive, perceptual and motor control aspects. In E. Kowler (Ed.), Eve movements and their role in visual and cognitive processes. Reviews ofoculomotor control research (Vol. 4, pp. 353-393). Amsterdam: Elsevier.

Widdel, H. & Kaster, J. (1981). Eye movement measurements in the assessment and training of visual performance. In J. Moral & K. F. Kraiss (Eds), Manned systems design, methods, equipment and applications (pp. 251-270). New York: Plenum Press.

Acknowledgements--We thank Pieter Schiphorst for technical assis- tance, Chris van Groeningen for being a subject, Eileen Kowler and two anonymous referees for helpful comments and Sheila McNab for improving style and language.

Section 5

Clinical Vision Sciences