This target article puts together pieces of experimental evidence in oculomotor and related research - with special emphasis on the express saccade - in order to enhance our present understanding of the coordination of vision, visual attention, and eye movements necessary for visual perception and cognition.
We hypothethize that an optomotor reflex is responsible for the occurrence of express saccades, one that is controlled by higher brain functions of disengaged visual attention and decision making. We describe a neural network as a basis for more elaborate mathematical models and computer simulations of the optomotor system in primates.
1. Introduction
Visual perception and cognition, movements of the retinal image, and eye movements are so closely related that one wonders how one field can be studied without taking into account the others. The facts are straight forward: (i) During natural viewing conditions a normal adult subject makes 3 - 5 saccades in a second separated by periods of 200 - 300 ms during which the eyes do not make large or fast movements. These periods are usually called "fixations" but this terminology is avoided here because - as will become clear below - periods of no eye movements are not necessarily periods of active and attentive fixation, but can as well be periods where the eyes simply do not move. (ii) If the retinal image as a whole is prevented from moving (by successful voluntary attempts not to move the eyes or by technical means) vision is blurred rapidly and the perception of the retinal image eventually fades away completely within 10 seconds. (iii) The highly inhomogeneous structure of the primate retina with an extremely high density of receptor and ganglion cells in the center, a specialized fovea, and a rapid decline of the cell densities to the periphery makes it almost impossible to have a homogeneous and simultaneous percept of the total visual field without somehow moving the fovea to different positions and without somehow acquiring and integrating information from these successive "looks". The existence of a fovea requires both, eye movements and periods of fixation, i. e. the active suppression of saccadic eye movements. (iv) As a result of a complicated interaction between afferent, central, and efferent neural processes we perceive a complete and stable visual field, which can serve as a frame within which we see motion and within which we move ourselves or parts of our body.
This paper presents a concept which is aimed at an understanding of the coordination of visual processing on the one hand and on the generation or suppression of saccadic eye movements on the other hand. The concept based upon experimental data rather than upon theoretical assumptions includes the aspect of visual attention as an important mechanism that governs both visual perception and saccadic eye movements.
We first consider the result of a simple experiment: the occurrence of express saccades. This experimental finding is controversial because some authors report that they failed to elicit express saccades. Wenban-Smith and Findlay (1991) observed visually guided saccades with latencies of 100 ms but they could not verify these as part of a bimodal distribution. Reuter-Lorenz et al. (1991) also found monomodal distributions, except for one subject who clearly showed a separate express peak. Kalesnykas and Hallett (1987) believe in express saccades but they claim that visually guided saccades have latencies above 120 ms, whereas we maintain that under appropriate conditions the express saccades can have latencies even below 100 ms in agreement with Jttner and Wolf (1992).
Even when the existence of express saccades is accepted there is the question of why such short latency saccades can occur. Is it an effect of warning or alerting which decreases the reaction time (Ross and Ross, 1980) or is it even anticipation which produces an artificial peak in the distribution (Kowler, 1990). There exists also the idea of premotor priming and sensory facilitation that may be regarded as mechanisms that explain the latency reduction. In this article we will put forward the concept of the attentional system controlling the saccade system in such a way that express saccades occur in a certain state of disengaged attention, which enables other preparatory steps - like decision making - to take place before target stimulus onset. This concept is layed out in the form of the three-loop- model.
In its first part the article will take into account mainly work on human subjects and the concept will be developed on this basis. The second part of the article is devoted to the attempt to provide the concept with a neurophysiological and anatomical basis by considering the data from corresponding experiments in monkeys. Both human and monkey data support the assumption that the attentional system controls vision and eye movements and that it has a certain dual functional structure which is by itself under different amounts of voluntary control depending on the amount of practice the subjects have and depending on the state of maturation of the brain. Further evidence supporting the concept will be considered at the end by describing data obtained from dyslexic subjects and from patients with neurological problems.
The article leaves it open to future theoretical work to find a mathematical formulation for the concept. A good possibility, however, using logical AND-gates with certain time constraints as has been proposed earlier (Rogal, Fischer, 1986) to describe the experimental data of experiments on eye - hand - coordination both in man (Fischer, Rogal, 1986) and in monkey (Rogal et al.1985). This principal approach is used to simulate reaction time distributions and to show how the model can work quantitatively.
2. Basic observations Since any rapid shift of gaze by means of a saccadic eye movement (head and body movements are not considered here and the vestibular ocular reflex therefore is neglected) is preceded by a time of no eye movements (microsaccades and slow drifts are not considered here) one can manipulate the visual conditions prior to a saccade to get an idea of what goes on when the eye does not move; i.e. when it is supposed to "see". The most straightforward manipulation is to remove the stimulus that the eye is "looking at", i.e. the fixation point, before the next saccade is generated. Such a situation can then be compared with one in which the fixation point remains visible. To determine whether the visuo-oculomotor system is actually in two different states at the time of the occurrence of a new visual stimulus in these two situations the subject can be required to make a saccade to the new target and the reaction time is used as an index. These two experimental conditions - the gap and the overlap paradigm -
have been used in many of the reaction time studies.
2.1 Methods Since some of the data presented in this article have not been published elsewhere, the experimental details are described below. The temporal and spatial arrangements are illustrated in Fig. 1: during gap trials the fixation point was turned off before the target appeared, during overlap trials it remained on. The visual targets were computer generated white squares (0.2 x 0.2x) presented in random order 4x to the right or left from the small (0.1 x 0.1x) red central fixation point on a green background at a distance of 57 cm from the subjects' eyes. The luminance of all stimuli was well above perceptual threshold. Target onset was synchronized with the frame impulse (frame rate 83 Hz) and, since the stimuli appeared always at the same horizontal level, the physical delay between the desired onset and the real onset of the stimuli was in the order of microseconds. Reaction times were determined offline by an analysis program which detects saccades by a velocity (30x/s) and a duration (15 ms) threshold criterion. The result of the automatic evaluation of a trial was presented on the computer monitor. In case of artifacts or false detections the experimenter could either abort the trial or reanalyse it "manually" under visual control using the cursor. The data were stored and analysed further using a commercial graphics and statistics program (NCSS). Mostly the reaction time data are presented in the form of a histogram using a bin width of 10 ms. Data from different subjects were pooled only, if all these subjects showed similar results. Statistical evaluation of the data is given in case it is necessary to support the interpretation of the results. Mean values are given in certain cases only, because most of the reaction time data are poorly characterized by a single mean value. Rather the distributions and scatter plots are presented, which in many cases are selfexplanatory with respect to the conclusions drawn in this paper. To quantitatively characterize multimodal distributions we have used a 3-Gaussians curve fitting procedure described in detail by Fischer and Weber (1990) and Weber et al. (1991).
Further methodological details are given in the appropriate sections where the data are presented.
Fig. 1 about here
2.2 The Gap-Effect and the Express Saccade Saslow was the first who used gap and overlap trials with human subjects. He reported that under gap conditions saccadic reaction times are considerably shorter (ca. 140 ms) than under overlap conditions (ca. 200 ms) (Saslow, 1967). This phenomenon is now called the gap-effect. (As we will see below, the gap-effect and the occurrence of express saccades are not the same phenomenon).
When the experiment was repeated by Fischer and Ramsperger (1984), it was found that the distribution of saccadic reaction times was bimodal or even trimodal in gap trials: a first peak occurred around 100 ms and a second peak at about 170 ms; so that the mean value was indeed on the order of 140 ms (Fischer, Ramsperger, 1984). The middle and the bottom panel of Fig. 2 show such a distribution. Saccades contributing to the first peak are called express saccades; the others fast regular saccades. The development of the express peak with increasing gap durations from zero to 200 ms is shown by Fig. 2. The top distribution was obtained using the classical condition, i. e. the fixation point was turned off at the same time the stimulus was turned on. Introducing a gap of 100 ms drastically changed the distribution. It now consists of two peaks, one at about 100 ms and another at about 150 ms. Note that all reaction times became shorter: the mean value of 189 ms obtained with gap zero is reduced to 133 ms with gap 100 ms. The introduction of the gap has not only a latency reducing effect but turns the distribution into a bimodal one. We will further discuss this double effect in section 10. Increasing the gap duration further to 200 ms resulted in an increase of the first peak and a decrease of the second peak. In addition there occur reaction times between zero and 85 ms. We will see below that the latter are due to reactions, that the subject made in response to the offset of the fixation point and/or in anticipation of the target. The data presented in Fig. 2 stem from a single subject who had considerable practice in oculomotor investigations.
Fig. 2 about here
After training the subjects in the gap experiment for many days, it was found that the first peak increased and the second decreased or eventually disappeared altogether (Fischer, Ramsperger, 1986), thus resulting in a unimodal distribution which consisted almost exclusively of express saccades. Some other research groups (Reuter-Lorenz et al.1991; Wenban-Smith, Findlay, 1991) claimed to have difficulties in replicating the presence of double peaked distributions in the gap paradigm. However, looking at the data of Reuter-Lorenz et al. one finds that some of the distributions are indeed bimodal. For instance, in their Fig. 6 subject REP exhibits clearly separated peaks of express and fast regular saccades in the two gap conditions. Wenban-Smith and Findlay (1991) even question the existence of a separate population of express saccades. However, only three subjects were tested in their study, one of which produced a single peak of express saccades while the others showed "some suggestion of bimodality". In order to clarify this critical point by looking at the interindividual variability in the data of many subjects, we have recently measured the saccadic reaction times of 20 naive adult subjects (Fischer et al.submitted). Analysing the data statistically with the help of a curve fitting procedure it was found that in the gap paradigm the far majority (17 subjects) produced reaction times in the express range and, in addition, 12 of them showed clearly separated peaks of express and fast regular saccades as indicated by a bimodality factor. Only three subjects did not produce any express saccades. Thus, a bimodal distribution of express and fast regular saccades is quite frequently observed, but its presence is not a necessary prerequisite for the existence of an express saccade population. These data show in addition that express sacades can be obtained in naive subjects without any previous training.
The distribution of reaction times from overlap trials exhibit a rather broad peak with values scattered around 200 ms. These saccades are called slow regular saccades. An example of this general result is given in Fig. 3B. In A the distributions for gap trials are shown from the same subject and from the same experimental session. This subject was highly trained and produced a single peak around 100 ms consisting almost exclusively of express saccades (Fig. 3A), a few saccades below 85 ms may be anticipatory and very few fast regular saccades can be seen at about 160 ms.
By contrast, saccadic reaction times are all longer than 150 ms during overlap trials (Fig. 3B). Thus, keeping the fixation point visible changed the distribution of saccadic reaction times completely and the question arises: why?
Fig. 3 about here
Obviously, the difference between gap and overlap trials is not only that in one case the fixation point is absent (gap) while in the other it is present (overlap) at the time the target occurs but also that there is an event (the offset of the fixation point) that precedes the saccade in gap but not in overlap trials. This event - one may argue - elicits a saccade which anticipates the occurrence of the target.
Before we discuss other processes that may give rise to the large difference in reaction times, we will first consider the many pieces of experimental evidence indicating why express saccades cannot be anticipatory or predictive in nature but rather are to be considered as visually guided saccades. (The rest of Fig. 3 will be treated later).
3. Express saccades: anticipatory or visually guided? Express saccades were first identified by Fischer and Boch (1983) in the monkey and subsequently by Fischer and Ramsperger (1984) in man. In both studies, in which gap trials were used, several control experiments were conducted to argue against the notion of anticipation and/or prediction as an explanation of the occurrence of express saccades. Here we list the arguments irrespective of their historical or logical order.
(i) On trials where unexpectedly no target appeared after the gap (catch trials) neither human subjects (Jttner, Wolf, 1992) nor monkeys (Fischer, Boch, 1983) make saccades at the corresponding time of 300 ms or 270 ms after fixation point offset, respectively.
(ii) Express saccades can be obtained in overlap trials both in human subjects (Mayfrank et al.1986) and in monkeys (Boch, Fischer, 1986). This takes some practice or training for monkeys and normal adults, but children and youngsters (age 8 to 17 years) and in particular certain dyslexic children make express saccades in the overlap condition spontaneously and without any practice (Fischer, Weber, 1990).
(iii) If the gap duration is randomly varied between 200 and 220 ms the narrow express peak is displaced by 20 ms (as measured from fixation point offset) in the 220 ms gap trials as compared with the 200 ms gap trials (Fischer, Ramsperger, 1984). If the gap duration is randomly varied between 100 ms and 200 ms or between 200 ms and 300 ms such that the onset time of the target is unpredictable express saccades are still obtained. Examples of the corresponding distributions are shown in Fig. 9.
(iv) The exact reaction times of the express saccades critically depend on the luminance contrast, size, and eccentricity of the target stimulus (Boch et al.1984). This fact alone contradicts the notion of anticipation. It is however not the luminance contrast itself which determines the reaction time, because isoluminant colour contrast target stimuli elicit express saccades of about the same latency (Weber et al.1991).
(v) The number of express saccades depends on the relative luminance of the fixation point (Mayfrank et al.1986).
(vi) The size of the express peak in the monkey can be increased by daily practice. The practice effect, however, occurs only for the saccades of the same size and direction as have been used during the training session, not for other saccades (Fischer et al.1984).
(vii) There are subjects, who produce many express saccades to one side but not to the other side. Fig. 6 presents the data of such a subject.
(viii) The number of express saccades is decreased rather than increased if the subject directs his attention to the target position (see below).
(ix) Anticipatory saccades are obtained and identified by their reaction times below 85 ms (in human) in gap trials with a constant (non randomized) target position. These saccades disappear almost completely when the target position is randomized, while express saccades are still present. This is true regardless of wether only the direction or both the direction and the location are randomized (Weber et al.1992). Anticipatory saccades have a somewhat smaller velocity as compared with the main sequence (Bronstein, Kennard, 1987); whereas express saccades have normal velocities (Fischer, Weber, 1992).
(x) If the target position is randomized between right and left, saccades are sometimes made in the wrong direction. These saccades have latencies below 90 ms and they are anticipatory by definition. This can be seen from Fig. 4, where we plot the reaction time distribution for saccades correct in direction and wrong in direction. These data have been accumulated from a single subject using the methods described above. The minimum reaction time of a visually guided saccade therefore is in the order of 100 ms, which comes very close to the number estimated by other authors (Kalesnykas, Hallett, 1987; Wenban-Smith, Findlay, 1991). The direction errors as well as amplitude errors of these anticipatory saccades can be corrected (after extremely short intervals down to zero!) at express saccade latencies (Fischer, 1986).
(xi) In an "anti"-saccade task express saccades are absent, but anticipatory saccades occur (see Fig. 5). The details of the "anti"-saccade experiment are described elsewhere (Fischer, Weber, 1992).
(xii) Patients with frontal lobe lesions are unable to anticipate but they make high numbers of express saccades (Guitton et al.1985), (Braun et al.1992).
Fig. 4 about here
On the basis of these arguments, it is evident that the notion of the express saccades being anticipatory or predictive in nature must be wrong. It will become clear, however, that anticipatory and express saccades may be both preceded by certain common preparatory processes. The difference is that express saccades are visually guided, (i. e. they need the onset of a visual stimulus as trigger and a target) but anticipatory saccades are not. In addition, anticipatory responses require the presence of the frontal lobe, but express saccades do not (Braun et al.1992).
Several studies of saccade dynamics have shown that saccades elicited in the absence of a visual stimulus and, in particular, predictive or anticipatory saccades have slower velocities than visually guided saccades (for review see Becker, 1989). Consequently, there have been attempts to make use of this velocity difference as a criterion to distinguish between anticipatory and visually guided saccades. Looking at the reaction time of this velocity transition it has been noticed that it occurs surprisingly early, i. e. between 30 and 70 ms (Smit, van Gisbergen, 1989). Nevertheless these authors concluded that visually guided saccades in man cannot have reaction times below 120 ms, thereby confirming the claim of Kalesnykas and Hallett (1987). However, the data of Smit et al. as well as the existence of express saccades with latencies even below 100 ms (given optimal physical and psychological conditions) obviously show that the minimum latency of a visually guided saccade is on the order 100 ms.
4. The Express Saccade: An Optomotor Reflex So far we have argued that the express saccade is visually guided, i. e. it occurs in response to the onset of a visual stimulus. Now we will argue that the express saccade occurs only in response to the onset of a visual stimulus, which is not only the temporal trigger but also the spatial target for the eye movement.
Let us consider a situation which is physically identical with the normal gap task (in which express saccades can be obtained easily and frequently), but the instruction is changed: the subjects are asked to make a saccade in the direction opposite to the side where the stimulus occurs. In this so called anti- task the visual information is no longer in register with the motor information necessary to guide the eye to the desired position. The distribution of reaction times obtained in the normal and in the anti-gap task are shown in Fig. 5. The top distribution shows the usual bimodality with a clear express peak at 100 ms. The second panel shows the reaction times of those saccades that the subject made unvoluntarily to the target rather than to the opposite side (these are "direction errors" in the sense of the anti-task). Interestingly, this distribution shows a clear express peak as well. The bottom panel shows the reaction times of the anti-saccades: an express peak is clearly missing. This experiment shows the impossibility of generating express saccades to a position where no target appears. This point can be made even stronger by a slight change in the experiment. The subject is instructed to make saccades always in the same direction, whereas the stimuli occur randomly at the right or left side. Yet, in this situation express saccades are obtained only on trials that happen to be normal trials even though on anti-trials the subject knows in advance the time, the direction, and the size of the saccade he/she is supposed to make (Fischer, Weber, 1992).
Fig. 5 about here
The conclusion from this experiment is that the express saccade needs the onset of the stimulus at the position to which the saccade will be directed. The express saccade appears as a reflex-like optomotor reaction that brings the fovea to the stimulus or close to it. We will see below that this reflex is usually inhibited and we will consider the different circumstances under which the reflex is disinhibited. For example in neurological patients, in some dyslexics, and in normal adults after practice and/or with certain instructions express saccades occur much more frequently than in normal naive adults. Nevertheless it must be born in mind that there are naive normal adult subjects who make express saccades in the gap task and even in the overlap task (Fischer et al.submitted). The notion that express saccades occur only after training is wrong.
4.1 Gain Control of Express Saccades For regular saccades (above 2 degrees in size) the gain, i. e. the ratio between the size of the saccade and the eccentricity of the target, is one or slightly less for larger saccades, which quite often undershoot the targets (Becker, 1972), (Henson, 1978).
The mean value of the gain of express saccades appears to be smaller than that of regular saccades. Fig. 6 shows a scatter plot of saccadic amplitude versus reaction time. The data are obtained with random target position as described above from a single subject. One clearly sees the anticipations (interestingly only to the right side for this subject) characterised by a large scatter of their amplitudes. The express saccades, in particular those to the left side, have smaller amplitudes than the regular saccades. According to our unpublished observation (3 subjects, 4361 saccades) the express gain (target at 4 deg) is 0.91 whereas that of the regular saccades is 0.99. These numbers may depend on the eccentricity of the target, because for very small saccades (below 1 or 2 deg) the proportion of express saccades has been found to decrease (see below) and all saccades begin to show a tendency to overshoot the targets, such that their gain becomes larger than one (Weber et al.1992).
Fig. 6 about here
Under certain circumstances the saccadic gain is, however, changed and the question arises whether or not the express saccades undergo similar gain changes. Since the express saccade obviously uses a rather short pathway (see below), one might expect that its gain is fixed and cannot be modulated by the experimental conditions. Instead, however, the experimental results are quite different:
(i) The global effect, first described for regular human saccades (Findlay, 1982), exists also for express saccades in man. This means, that an express saccade elicited by the simultaneous onset of two stimuli at different but neighbouring positions lands somewhere in the middle rather than at one of the two targets (Weber et al.submitted). (ii) Similarly, express saccades in the monkey are subjected to an adaptive gain change, already described for regular saccades in the monkey (Albano, King, 1989), (Fitzgibbon et al.1985) and in man (Albano, King, 1989), (Deubel et al.1986), (Miller et al.1981). If the first target for the saccade is replaced by a second target at the beginning of the saccade to the first target, then - after some practice time - the saccade to the first target becomes larger or smaller, i. e. the gain increase or decreases, depending on whether the second target occurs farther away or closer than the first target (Latanov et al.submitted). (iii) Finally, we observed that human, but not monkey, express saccades show a considerable range effect, as described for human regular saccades (Kapoula, 1985), (Kapoula, Robinson, 1986): this means that the size of the express saccades to a target at a given position depends on whether the target at this position is presented in random order together with single targets at other positions or whether the target is presented in the same position all the time (Weber et al.submitted). Other intertrial effects will be considered in section 6.3.
These gain changes must occur in the striate cortex or beyond, because electrical stimulation of the superior colliculus of already adapted monkeys results in saccades of the unadapted size (Fitzgibbon et al.1985). If one assumes that prestriate cortical areas do not contribute to the determination of the amplitude of a saccade then it must be concluded that the direct or indirect messages from striate cortex to the oculomotor centers can be modified.
5. The Express Way - Anatomy and Physiology Investigation of the anatomical connections from the retina to the oculomotor centers in the brain stem reveals several possible pathways (for review see Fischer, Boch, 1990):
(i) The shortest connection originates from the large retinal ganglion cells that send their axons directly to the superior colliculus (Perry, Cowey, 1984), which in turn projects to the brain stem (Harting et al.1980). (ii) Another pathway includes the lateral geniculate body and the striate cortex, from where layer V cells send their axons to the superior colliculus (Tootell et al.1988). (iii) A still longer pathway runs through the visual cortex to the frontal eye fields (Barbas, Mesulam, 1980) and from there either directly to the brain stem (Glickstein et al.1985) or indirectly through the superior colliculus (Knzle et al.1976) to the brain stem or even more indirectly through the caudatum and the substantia nigra pars reticulata to the superior colliculus (Astruc, 1971).
The third possibilitiy as a pathway for express saccades must be excluded because express saccades - unlike regular saccades - survive lesions of the frontal eye fields but not lesions of the superior colliculus (Schiller et al.1987). The first possibility, likewise, must be excluded because express saccades cannot be obtained without the striate cortex (Boch, 1989). Higher cortical areas - such as the prelunate cortex (V4) or the parietal cortex (A7) can be involved only indirectly, because the latencies of the visual responses of the cells in these cortical areas are in the same range or even longer than the express saccade reaction times (Tanaka et al.1986) or because their malfunction increases the number of express saccades, as has been shown for example by chemical lesions in area V4 (Weber, Fischer, 1990b). Another intriguing result came from an experiment, in which a monkey was trained to make saccades to a small white stimulus appearing at a particular location in the visual field. In parallel with the training the stimulus preferences of single cortical cells in V4 with receptive fields including the "trained" location was modulated: after several weeks of recording the far majority of these cells responded optimally to the "training" stimulus, while cells with receptive fields elsewhere showed no changes of their stimulus preference. This effect paralleled a spatially selective increase of the proportion of express saccades towards the "trained" stimulus (Weber, Fischer, 1990a). Both long term effects may be regarded as a stimulus selective "sensory priming", which also favours the execution of a very fast response towards an object of special importance in the life of the animal, and therefore may be of behavioural significance.
The indirect control of the parietal cortex over the generation of saccades becomes evident from the fact that electrically elicited eye movements are abolished when the monkey actively fixates a foveal stimulus during the stimulation (Shibutani et al.1984). A very similar observation has been made in the frontal eye fields (Goldberg et al.1986). Nevertheless, the frontal eye fields contribute considerably to the control of saccades, not only directly to the generation of regular saccades, but also indirectly by sending foveal signals for maintaining or releasing fixation (Latto, Coway, 1971).
Finally we consider the temporal aspect: retinal ganglion cells have latencies of their visual responses in the order of 20 ms or more. Another 10 ms are needed for transmission up to the striate cortex. Electrical stimulation in the superior colliculus elicits saccades after about 20 ms from the stimulus. Given a saccadic reaction time of 70 ms there are only 20 ms left for a central computation time, which is needed to evaluate the correct direction and amplitude of the express saccade.
In conclusion, lesion experiments and considerartions of transmission times suggest that the express way includes the lateral geniculate body, the striate cortex, the superior colliculus, and the brain stem. Control over this pathway may be taken - directly or indirectly - by the frontal eye fields and/or by the parietal cortex. The nature of this control mechanism should become clear from experiments in which one attempts to suppress the occurrence of express saccades.
6. Suppression of Express-Saccades Since neither the physical conditions nor anticipation and prediction can account for the occurrence of express saccades, the difference of SRT's in gap and overlap trials calls for another explanation. One possibility is to study the conditions -
besides the overlap paradigm - under which express saccades are suppressed in subjects who can make them in the gap task.
6.1 Engagement of Visual Attention In this section we will discuss the idea that the disappearance of the fixation point in gap trials facilitates the disengagement of visual attention leading to a state in which saccades can be generated after shorter reaction times, whereas during overlap trials visual attention is engaged and saccades are inhibited. The time it takes for visual attention to disengage adds to the reaction time and accounts for the long latencies in overlap trials as compared with the short latencies in gap trials, where the disengagement (and may be other preparatory processes) can take place during the gap.
The first attempt to test the hypothesis was to very briefly (15 ms) blink the fixation point 200 ms before the target occurred. Except for the blink this situation is physically identical with overlap trials. The idea is that the blink can be used as a physical event to initiate the process of the disengagement. The result of this experiment is shown in Fig. 3C. One clearly sees a peak of express saccades and many fast regular saccades. The introduction of the blink (a physical change) has changed the unimodal distribution of slow regular saccades (Fig. 3B) into a bimodal distribution. Fig. 3A shows the result of the usual gap task with a single express peak.
The second attempt to test the hypothesis was to ask the subject to always direct (engage) his attention to one of the two possible positions of the peripheral target during gap trials. Fig. 3D shows the result (same subject). The peak at 100 ms (express saccades) is clearly reduced and a second peak (fast regular saccades) appears. The change in instruction (a mental change) has changed the unimodal distribution of express saccades (Fig. 3A) into a bimodal distribution. Important to note here is that visual attention directed to the target position increased the number of fast regular saccades rather than reducing it. In other words: engaged visual attention produced a "cost" of time in some trials and no "benefit" in others.
If the blink facilitated the disengagement leading to express saccades in overlap trials and if the instruction to direct attention facilitated the engagement in gap trials leading to a reduction of the number of express saccades then the instruction to direct attention to the target position instead of the fixation point should leave saccadic reaction times unchanged in overlap trials because attention is engaged in both situations. The result of this test, shown in Fig. 3E, when compared with the result in Fig. 3B confirms the above prediction.
If, finally, the instruction to pay attention to the target position is given in overlap trials with a blink the number of express saccades should be reduced (Fig. 3F) as compared with the case where attention is directed to the fixation point (Fig. 3C), because the blink has a lower chance of initiating the disengagement. The equivalence of the fixation point and a peripheral stimulus as attention targets for the occurrence of express saccades has been explicitely shown by Braun and Breitmeyer (1988).
In conclusion, visual attention acts on the preparation of visually guided saccades in a way suggesting that directed (engaged) visual attention increases the overall reaction time, whereas disengaged attention reduces it. The change in reaction time, however, is not a shift of a (unimodal) distribution but rather a modulation of the size of (at least) three distinct peaks in the distribution. In particular, engaged visual attention tends to eliminate the express peak as well as the peak of fast regular saccades. In the light of the reflex hypothesis one can say that engaged attention inhibits the saccade system and especially the reflex. Attentive fixation in this context appears as just a special state of engaged attention, in which attention is directed to a foveal stimulus thereby inhibiting the saccade system.
The idea of the attentional system acting in three different steps - disengage - move - engage - , derived here from eye movement data alone, was first proposed by Posner (1984), who studied manual reaction times of patients with parietal lobe lesions.
Posner's concept is that a "benefit" in reaction time is obtained when attention is enaged at the position where the stimulus is going to be presented, and a "cost" when the stimulus appears somewhere else. We have found that the subjects can maintain the state of disengaged attention for some time in the order of some hundred milliseconds and that the most dramatic changes in reaction time are observed when the target stimulus is presented during the time of disengaged attention (Mayfrank et al.1986). Recently, Makeben and Nakayama (paper submitted) found in normal human subjects that shifts of visual attention as measured by a vernier detection task is speeded considerably by the introduction of a temporal gap between fixation mark offset and cue onset. They concluded that their results support directly the three step concept of attention, especially that of the disengagement. There exists of course a large body of work on attention in the psychology literature, but here we have considered only those aspects that are closely related to the saccade system.
Other authors have also used the gap task and they found express saccades in only some of their subjects (Reuter-Lorenz et al.1991). However their methods were not identical to the simple task we used, because they randomly inserted catch trials and they always used a warning tone even in overlap trials. These authors nevertheless obtained a gap-effect. But its size was unusually small in the order of 30 ms, which we attribute to a latency decrease resulting from the introduction of the tone (see also chapter 6.1.1). Moreover they found no gap-effect at all with anti-saccades, which is in clear contrast to our results (Fischer, Weber, 1992); neither did they find a decrease of manual latencies which is also at variance with earlier studies in man (Ross, Ross, 1981; Fischer, Rogal, 1986) and in monkeys (Rogal et al.1985). Despite this difference in the experimental results Reuter-Lorenz et al. attributed the gap-effect to a facilitation of premotor programming in the superior colliculus. This idea is not necessarily in contrast with the concept of attentional disengagement, because they leave open what the physiological/psychological meaning of "facilitation" is, from where the corresponding signal arises, and along which pathways it arives at the superior colliculus. For example one could regard the facilitation as a neural sign of attention becoming disengaged.
Further investigations of the gap-phenomenon come from Ross and Ross (1980, 1981). These investigators obtained a considerable gap-effect (about 75 ms with gap durations of 100 and 300 ms) with saccadic (Ross, Ross, 1980) as well as with manual choice reaction times (Ross, Ross, 1981). Unfortunately all reaction times below 130 ms were excluded from their analysis, so that eventually occurring express saccades would have been eliminated. These authors consider the extinction of the fixation stimulus prior to target onset a warning event, which is assumed to exert a facilitating effect on the latency of saccadic or manual responses to a peripheral target. The resulting reduction of latencies is assumed to reflect a "general preparatory or alerting process that affects the preparatory steps related to the programming of the saccade to the target stimulus that follows". Although this claim is not at variance with our concept of saccade preparation, we regret to see that the Ross and Ross papers lack any discussion about the neurophysiological nature of such facilitatory processes.
A general shortcoming of the facilitation concept, proposed also by Reulen (1984), is that it cannot explain the occurrence of more than one mode in the latency distribution, neither can it account for the occurrence of express saccades in overlap trials, where there is not external stimulus triggering the facilitation process. We will take up this points again in section 10.
6.1.1 Non-Visual Stimuli If the offset of the fixation point initiates the disengagement of visual attention thus triggering one process in the preparation of the next saccade it should be possible to disengage one's attention also on the command of a non visual stimulus. Kimmig (1986) has used a short accoustic signal preceding target onset by 200 ms in the overlap task. Whether express saccades were obtained or not, depended on wether the subject paid attention to the tone or to the fixation point. Fig. 7 shows this effect. In A the result of the control experiment with overlap trials is shown. In B and C the physical conditions were the same, but the subject was told to pay attention to the fixation point (B), whereas in C the tone had to be attended. As a result an express peak was absent in B but present in C. The distribution in B, nevertheless, differs from that in A: the tone, even when not attended, reduced the reaction times. What exactly the tone does for the peparation of the saccade in this case is not clear. It is evident, however, that whatever it is, it cannot be prevented voluntarily. Note also, that the tone does not by itself trigger an express saccade - this is done by the onset of the target. The tone rather provides a signal that can be used to disengage attention, which in turn will enable express saccades to occur in response to the target. The same of course applies to the offset of the fixation point or any other external (sensory) or internal event that gives rise to a disengagement of attention.
Ross and Ross (1981) also found a considerable reduction of saccadic latencies with the onset or offset of an auditory signal 100 or 300 ms prior to the onset of the saccade target, comparable to their gap-effect obtained with visual events (even though these authors considered only saccades with latencies above 130 ms). They emphasize the generality of what they call the facilitating effect of a nonspecific warning event preceding target onset. We feel, however, saying that the tone is used by the subject as a warning signal is not an explanation but rather another verbal description of the latency decreasing effect of the tone. Reuter-Lorenz et al. (1991), used an overlap condition with a warning tone occurring either 200 or 300 ms before target onset, which resulted in lateny distributions with mean values of about 180 ms (mostly fast regular saccades). This is very fast when compared with other studies, where mean values in the order of 200 to 250 ms are reported (Saslow, 1967; Mayfrank et al.1986; Kalesnykas, Hallett, 1987; and others), and is likely to result from the introduction of the tone. This assumption might explain the small latency differences (gap-effect) between gap and overlap distributions in that study, because a kind of gap-effect is already present with the overlap-tone condition.
Fig. 7 about here
Nevertheless, all these data show that the signal for the disengagement may be generated through different sensory systems, the visual system (e. g. the offset of the fixation point) being only one possibility. The fact that during overlap trials without any physical event preceding target onset express saccades can be obtained (see section 7) shows in addition that the disengagement signal can also be generated internally.
6.2 Small Saccades Fig. 8 shows a scatter plot of saccadic reaction times versus saccade size of a single human subject obtained from a gap experiment, in which the targets appeared randomly at different postions between 0.5 and 10 degrees from the fixation point. The horizontal band at 100 ms represents the express saccades. Saccades above 5 degrees are almost exclusively of the express type. Between 2 and 4 degrees one clearly sees regular saccades in addition to the express saccades. Below 2 degrees express saccades are virtually absent and, below 1 degree, the regular saccades have increasingly long latencies (Weber et al.1992). An increase of saccadic latency for small saccades (below 0.5 deg) has been reported earlier (Wyman, Steinman, 1973b), but these authors did not distinguish between different groups of saccades according to their latencies.
Fig. 8 about here
The experiment with small saccades was repeated in a monkey. The animal was trained in a fixation and in a saccade gap task and produced a reasonable number of express saccades to a target positioned at 4 deg in the left visual field. If the target eccentricity was stepwise decreased below 1 deg express saccades disappeared but regular saccades were still present in quite the same way as in man.
This experiment shows that the fovea is surrounded by a region into which one can make saccades but these cannot be of the express type. Even extensive training of the subjects (man and monkey) did not lead to small express saccades. Thus it looks as though an express way exists only for larger saccades. Alternatively, the offset of the fixation point and the following disengagement could disinhibit only that part of saccade system, which generates saccades larger than 1 or 2 degrees.
In addition it was found that anticipatory saccades were largely reduced in number as compared to the case in which the target appeared always at larger distances, say 4 deg to the right. This observation shows that anticipatory saccades and express saccades have something in common: they both need the disengagement of visual attention as one prerequisite for their occurrence. The express saccade, however, in contrast to the anticipatory saccade needs the physical onset of the target at a sufficiently large distance from the fovea.
The reduction of the number of anticipations for small eccentricities raises another question. Is it possible to make small saccades - voluntarily or unvoluntarily - in the absence of a target? The answer comes from an experiment in which the target again appeared always in the same position at 1 degree but on some trials - unpredictibly for the subject - no target appeared (catch trials). The task was always to make a saccade to the 1 degree position. The result was that on both types of trials many saccades were triggered by the offset of the fixation point and express saccades were absent. The important observation was, that the size of the saccades obtained on the catch trials was clearly greater than 1 degree and clearly larger than the size of the saccades obtained on the target trials. This observation confirms previous work (Haddad, Steinman, 1973) reporting that subjects could make voluntary saccades as small as the miniature saccades during fixation only if the targets were visible. This explains the reduction of the number of small anticipations, because the anticipations must be regarded as saccades that are initiated before the information of the target position was properly received.
In a further experiment, we applied the overlap paradigm with small (1x) and with larger (4x) saccades. Comparison of SRTs in gap and overlap conditions with the same target eccentricities revealed that a clear gap effect was also present for the small saccades: while slow regular saccades were obtained mostly with the overlap condition, the gap favours the occurrence of fast regular saccades. Therefore it can be concluded that the signal of a disengagement of attention (facilitated by fixation point offset) is provided for small saccades as well. We assume that the "express way" does not exist or is interrupted for saccades in the close vicinity of the fovea. Most likely this implies that the intratectal anatomical connections from the superficial to the deep layers of the superior colliculus (Moschovakis et al.1988); (Paige, Sargent, 1991) spare the center of the visual field representation.
In any case the earlier notion of an oculomotor dead zone (Rashbass, 1961) seems to be wrong for regular saccades (Wyman, Steinman, 1973a), but true for express saccades.
6.3 Intertrial Effects We have seen in section 4.1 that the gain of saccades can be modulated by the context within which the saccade is being made. We now consider the possibility that the probability of the occurrence of an express saccade on a given trial may depend also on what happens on other trials.
Jttner and Wolf (1992) did the following experiment: they randomly mixed target trials with catch trials (trials in which no target occurred) in an otherwise normal gap condition. The result was that the higher the proportion of catch trials in a block the lower the proportion of express saccades. In particular, they found that the probability for an express saccade was most strongly reduced on the trials following a catch trial. The authors believe that this effect has to do with the process of the decision to make or not to make a saccade, because as a rule on catch trials no saccades were made by the subjects and consequently the decision to make a saccade was not made (Jttner, Wolf, 1992).
We modified this experiment by randomly mixing long gap trials (instead of catch trials, which can be considered as trials with an infinitely long gap) with short gap trials. Under these conditions the subject makes a saccade on every trial. Fig. 9 illustrates the result: in A and B the gap duration was 100 ms, but in A the gap 100 trials were randomly mixed with gap 0 trials, whereas in B they were mixed with gap 200 trials. Similarly, in C and D the gap duration was 200 ms, but in C the gap 200 trials were mixed with gap 100 trials (same experimental session as in B) and in D they were mixed with gap 300 trials. Clearly, the physical conditions of a given trial do not determine alone, what kind of saccade is going to occur. The last experiment shows that it is not just a question of the decision to make no saccade on one trial and to make one on the next trial.
Fig. 9 about here
The results of this section show that there are internal states of readiness that determine wether express saccades are generated or not. The effect of intermingling catch trials or long gap trials seems to be that the subject in order not to make mistakes does not use the offset of the fixation point as frequently to start the disengagement but rather waits more often until the target appears, before he/she starts any preparatory processes for the next saccade. One can say, the subjects maintains fixation until the target appears.
7. Effects of practice and age on saccadic reaction times
First of all, we recall that the data discussed so far are obtained from highly trained subjects. Naive and normal subjects above age 20 may produce everything between zero and 50 % express saccades during gap trials and usually no express saccades during overlap trials. Practice using gap trials increases the number of express saccades both in monkey (Fischer et al.1984) and in man (Fischer, Ramsperger, 1986).
The important point here is that with practice express saccades can be also obtained from adult subjects in overlap trials, i.e. when there is no physical event preceding the onset of the target. Again this has been shown for monkey (Boch, Fischer, 1986) as well as for human subjects (Fischer, 1987).
The fact that instructions concerning the direction of visual attention quite drastically change the sizes of the different peaks in the SRT-distribution offers an explanation of what is changed by practice: the ability to disengage one's attention without the help of the onset of the new target. The disengagement is relatively easy and largely facilitated in gap trials, but relatively difficult in overlap trials. The disengagement is a state, in which one cannot stay for long periods of time, because with increasing gap durations from zero the number of express saccades reaches a maximum around values of 200 - 300 ms and then decreases continously to zero with longer gap durations up to 800 ms (Mayfrank et al.1986). The disengagement, therefore appears as a state which is reached more or less automatically before any saccade during natural inspection of a large visual field. To reach this state voluntarily at a given time and to hold it over a certain time interval may indeed need practice for a normal and naive adult observer, but may be the "normal" state for children or subjects with certain problems leaving them in a state of almost permanent disengagement (see section 8 and 9).
The small number of express saccades during gap trials and their virtual absence during overlap trials in normal adults changes abruptly if one looks at younger subjects. Whereas there are only minor differences between age 50 and age 25, the number of express saccades in the gap task is much higher for youngsters of age 15 - 16 and children of age 8 - 10. Fig. 10 shows the distribution of naive subjects of three different age groups (22 - 28, 15 - 17, and 9 - 10 years). The figure not only shows that even at age 16 the control of eye movements is not fully developed, but also that the number of express saccades is highest in early life (it remains open, what happens at pre- school age).
Fig. 10 about here
In conclusion, maturation beyond age 16 - 18 years decreases the chances of producing express saccades, but by practice an adult subject may regain this ability. It looks as though with increasingly difficult tasks in life - such as learning to read -
more control over the saccadic systems is necessary. This control - at least in part - is provided by the attentional system and has the consequence that reflex-like optomotor reactions , i. e. express saccades, are rarely seen in the gap trials and almost completely absent in overlap trials.
8. Dyslexia and saccades
We now consider the possibility that for one or the other reason a child does not develop the attentional control over the saccade system, either because the attentional system is not properly developed or because the control action is not properly passed onto the saccade system. In any case one expects unusual patterns of saccadic eye movements and changes of saccadic reaction times in non-cognitive tasks and - of course - during reading and durind any other task requiring proper coordination of vision and eye movements.
Such reports were published by Pavlidis (1981, 1985), who claimed that dyslexic children often overshoot the targets when they are asked to scan them by saccadic eye movements. The result - according to Pavlidis - was an increased number of regressions in the eye movements records (Pavlidis, 1981, 1985). Olson (1983) - in his attempt to replicate the results of Pavlidis (1981) - failed completely and concluded that dyslexic children have normal eye movements (Olson et al.1983).
On the basis of the express saccade and its possible relationship to visual attention we looked at saccadic latencies of dyslexic children using gap and overlap trials (Fischer, Weber, 1990). We found that dyslexics clearly differ from controls in their latency distribution. In particular we found a group of dyslexics (group II) who produce almost exclusively express saccades regardless of whether gap or overlap conditions were used (Fischer, Weber, 1990). This most interesting observation was later supported and extended by Biscaldi and Fischer (1992), who found another 4 children among 12 other dyslexics who showed this striking pattern of saccadic latency distribution. Fig. 11 shows the distributions of saccadic reaction times of control children and of two groups of dyslexic children all tested in the overlap paradigm. First of all, one sees that children make more short but also more long latency saccades. They seem to be unable to set properly the time they make the eye movement. Quite clearly, there is a preponderance of express saccdes in the group II dyslexics, which differentiates them not only from adults but also from the age matched controls and from other dyslexics.
Fig. 11 about here
Biscaldi and Fischer (1992) also used the light test of Pavlidis. In their data the number of regressions was normal for all the dyslexics - in disagreement with Pavlidis - but other parameters of sequences of the saccades , for example their number and size, were abnormal. In disagreement with Olson it is concluded that the eye movement patterns, in particular the reaction times, of many if not all dyslexics at age 9 to 11 years are different from normal readers. The analysis of the eye movements in non- cognitive tasks may be even used as a diagnostic tools to tell dyslexics apart from normals or from children with other problems that might have contributed to their deficit in reading.
The main conclusion of this section is that understanding the role of attention controlling the saccadic system, gives insight into why reading is impaired when this control mechanism does not work properly. The problem for such subjects is twofold: on the one hand they have difficulties in their attentional system, on the other hand and as a consequence of the attentional problem they have difficulties in producing proper saccades at proper times. More specifically, too short latencies and in particular express saccades under overlap conditions are obtained if the subject fails to engage attention consistently. These subjects are ready to move their eyes whenever a new target appears. One can even say that their attention is very likely to be distracted. The opposite happens with subjects who cannot consistently disengage their attention. They stay in the engaged state with the consequence of prolonged reaction times. Finally, one could even consider the possibilty that a subject has difficulties in switching from one state into the other: once engaged they cannot disengage in time, once disengaged they cannot engage in time. In this case one predicts clearly bimodal distributions with an "too early" peak and a "too late" peak as compared with control subjects. These aspects of dyslexia and of the concept of engaged/disengaged attention are rather speculative at the moment and need more theoretical and experimental evaluation.
The attentional problem may not neccessarily show up in a neuropsychological test, because it may be that though a subject may have difficulties in switching to the engaged state, once there he/she can perform perfectly well in any task which requires engaged visual attention but no saccades.
It would be very interesting to study the development of the saccadic and the attentional system from birth to see at what time express saccades and slow regular saccades can be obtained. Certainly, infants below two months of age can make target directed saccades, but their latencies are broadly scattered and rather long and their amplitudes are much too small (Aslin and Salapatek, 1975). These saccades may be the result of an still immature reflex, the use of which is necessary for the development of the saccade system, which only later will be controlled by cortical structures.
9. Clinical Observations
In this section we consider data from neurological patients contributing to our present understanding of saccade generation and visual attention.
A basic observation was reported by Guitton and his group: patients with unilateral lesions of the frontal lobe were unable to suppress saccades to a suddenly appearing visual stimulus at one side when they were requested to make saccades to the opposite side (anti-saccade task). When a few hundred milliseconds later another stimulus occurred they made saccades towards this new target after reaction times of about 100 ms, i. e. they made express saccades (Guitton et al.1985). The inability to suppress initial glances at potentially distracring stimuli after frontal lobe lesions was also reorted earlier (Milner, 1982). One could argue that the inhibitory action through the frontal eye fields onto the saccade generating system was no longer effective thus allowing for express saccades. This seems indeed to be the case whenever frontal patients are tested in a gap condition. In the overlap condition, however, the saccadic reaction times are in the normal range. The suppressive effect of the presence of a fixation point on the occurrence of express saccades is intact in these patients (Braun et al.1992), indicating that the occurrence of express saccades with frontal lesions is highly dependent on the state of fixation. The source of the inhibition may be the parietal cortex, because electrically elicited saccades are abolished in monkeys who are actively fixating a foveal target (Shibutani et al.1984). Braun et al. (1992) found that with lesions of the dorsolateral parietal cortex express saccades were significantly reduced, but this effect was the result of a general large scatter of the latency distributions of these patients. Parietal lesions thus appear to cause a destruction of stimulus triggered saccade timing affecting the whole spectrum of saccadic reaction times, rather than any special latency population.
The mechanisms of visual attention have long been associated with the parietal cortex. For instance, from Posner et al. (1984) we know that patients with lesions in the parietal cortex are unable to use a foveal cue to direct their attention to the periphery. It was concluded that visual attention must be disengaged from whereever it had been engaged before it can be moved to another part of the visual field. Our present notion of the mechanisms of visual attention as derived from eye movement data alone very much support this idea. Further evidence in favour of the "disengage-move-engage" hypothesis comes from measurements of visual attention without eye movements (Mackeben et al.person. communication). They used vernier acuity measurements in the parafoveal region and introduced a gap between fixation point offset and the onset of a spatial cue which indicated where the vernier target would be presented. Performance was best with gap durations of 100 - 300 ms. The authors conclude that subjects can accelerate their "attentional deployment" if they are given enough time to disengage their attention by means of fixation point removal before cue presentation.
10. Theoretical Considerations
As a possible basis for mathematical models or computer simulations, Fig. 12 depicts schematically the most important anatomical projections that connect the retina with the eye movement generating structures (EM) in the region of the brain stem.
The pathway mediating the express saccades is drawn by the single heavy lines. This pathway, obviously, is not always functionally available, because otherwise any sudden appearance of a visual stimulus would lead to a saccadic eye movement, which immediately would create an impossible chaotic situation. (The direct projection from the retina to the superior colliculus and from striate cortex to the brain stem alone (thin lines) cannot mediate saccadic eye movements because without striate cortex or without the superior colliculus express saccades are abolished. The significance of these connections remain therefore unclear in this context.) It is assumed that a cortical control mechanism, presumably through the frontal eye fields, acts on this first, most primitive reflex loop as to prevent the generation of reflex-like saccades. It is only when this inhibitory action is stopped that a saccade can be initiated directly through this pathway by the onset of a stimulus. Such a saccade occurs as an express saccade. The disinhibition may be direct or indirect through the nucleus caudatus and the substantia nigra pars reticulata (Hikosaka, Wurtz, 1983). Similarly, the second loop is only available, if another inhibitory action, presumably through the parietal cortex, is taken away. This control - we propose on the basis of the data described above - is taken by visual attention being engaged or disengaged. In the disengaged state the inhibitory action is stopped, saccades are permitted, in the engaged state the generation of saccades is inhibited either directly by the projection from the parietal cortex to the superior colliculus or indirectly through the frontal eye fields. The inhibition of saccades may also or in addition occur at the brain stem level. For example the continuous firing of the pause or omni-pause cells may drive this inhibitory mechanism.
Fig. 12 about here
The neurophysiological aspects of saccade generation are summarized by Fischer and Boch (1990), where the different types of presaccadic modulation of neural activity in the superior colliculus, the frontal eye fields, the parietal cortex, and the visual association cortex are described. It is essential to notice that in each of these structures different aspects of the saccade are important: the voluntary aspect in the frontal eye fields (Bruce, Goldberg, 1984; Pierrot-Deseilligny et al.1991), the attentional aspect in the parietal cortex (Mountcastle et al.1987; Robinson et al.1978), and in the visual cortex (Haenny, Schiller, 1988), the saccade as a pending motor action in the superior colliculus (Goldberg, Wurtz, 1972).
In this paper we have concentrated on the significance of visual attention for the generation of saccades. Little has been said about the decision process preceding each single saccade. However, a complete discussion of saccade generation must include this aspect. Without it one would not understand the existence of the three different modes in the distribution of the saccadic reaction times. The model outlined and described below, therefore, will take into account decision making and computation of the metrics of the saccade as well was the processes of visual attention.
10.1 A computer simulated model The three loop model, proposed by Fischer (1987) is outlined in the top of Fig. 13. Besides the afferent visual inputs ("vis", the signals from the fixation point and from the stimulus) and the efferent output to the oculomotor nuclei ("mot"), the model includes three central states denoted by "att" (for attention), "dec" (for decision to make a saccade), and "com" (for computation of saccade size and direction). A central state, as indicated in the middle, has three input lines (a = afferent, r = random, and c = central) and one output line. By r we denote a random process which is governed by other brain functions that may act on the optomotor system, e. g. general alertness or effects of the instruction for the subject. The output line becomes active according to the logical rule that c and at least one of the other input lines (a or r) are active. Once this condition is fullfilled the output becomes active after a transition time variing randomly (mean T and standard deviation s). T and s are model parameters set for each of the three central states, for the afferent, and for the efferent process. In principal, each of these processes has the same effect as the facilitation generator proposed by Reulen (1984). However, instead of using it only for the time after fixation point offset, we use this principle for all states.
Fig. 13 about here
The model also includes the possibility of any state to become inactive at random with a given probability. Before the final output line of the model can become active all three input lines to the motor state "mot" must still be active. This is realized by the links between "att" and "mot" and between "dec" and "mot" together with a threefold AND operation at the mot-level.
The heavy lines in the three (otherwise identical) graphs on top of the figure show the model pathways for the three types of saccades according to their latencies. The histograms at the bottom of Fig. 13 display the result of a computer simulation of 200 gap trials and 200 overlap trials. One clearly sees the three peaks in the overlap case and the two peaks in the gap case. Note that anticipations occur in gap trials much more often than in overlap trials. The model also produces a few cases, representing socalled "misses" with rather long reaction times (above 300 ms). The effect of the gap - in the light of the model - is to initiate "att", i. e. the disengagement of attention, which then after a delay can initiate also "dec". If after another delay "dec" also initiates "com" an anticipatory saccade is the result. If the target occurs and leads to the initiation of "com" an express saccade is the result. If "att" did not initiate "dec" the occurrence of the target will do this and the result is a fast regular saccade. Slow regular saccades are obtained very rarely in the gap condition as the chances of a deactivation of "att" within the gap are rather small. The opposite is true in the overlap condition: during attentive fixation "att" remains usually inactive until the target occurs. This then leads to a preponderance of slow regular saccades. For the simulation shown in the figure we have on purpose set the parameters as to produce a few express saccades in the overlap condition. A smaller chance of random switches of "att" and "dec" would have decreased or even completely eliminated the express and the fast regular peak leaving a monomodal distribution as one obtains it usually in the overlap task (see for example in Fig. 3B).
10.2 Other theoretical concepts
There exists a large body of work concerned with models of the saccade system. A good review is given by Becker (1989). However most of this work is devoted to an understanding of how the brain stem structures determine the size and direction of a saccade from what is called the "retinal error" signal. This process would be included in process C of the present model and its details are not considered here. The present approach is concerned with the total loop from the retina to the brain stem including the influences from higher brain functions. To explain the effect of the gap on the saccadic reaction times Reulen (1984) has proposed a facilitation model. This model explains the latency reducing effect of the introduction of a gap but it has several disadvantages: first, one does not know what is ment by "facilitation" in terms of a physiological or psychological process; second, it cannot explain the occurrence of short lateny saccades in overlap trials as they are observed in children and in specially trained adults; third, the model fails to produce bimodal distributions. The latter point is very important when considering the double effect of the gap condition. In accordance with the experimental findings the three-loop-model produces a reduction of the duration of the latencies and the appearance of a bimodality or even a trimodality. The latency distributions become unimodal when the gap durations are too short and yet the latencies are reduced as compared to the no-gap or overlap case. None of the models we are aware of can produce this basic feature of the saccade system. The latency reduction and the bimodality are both inherent features of the model and do not need two different explanations.
Since the gap task is most favourable for generating express saccades and since in a condition where the gap duration as well as the target position are kept constant the offset of the fixation point may be considered as a cue which can be used by the subject to activate what has been called "premotor priming" (for review see Oakley and Eason, 1990). This concept, however, has the difficulty to explain the occurrence of express saccades during overlap trials, where there is no cue preceding the target onset. Another difficulty arises from the fact that express saccades can be obtained with unpredictable gap durations and/or with unpredictable target positions: how could the priming process possibly facilitate a target directed movement when the size and direction of the movement are not yet known? Also, we have shown that the express saccade have smaller amplitudes, undershoot the target more often and by a larger amount as compared with regular saccades (see for example Fig. 6) A paper concerning this issue has been submitted. This observation would imply that the premotor priming makes the movement worse in accuracy rather than leading to an improvement. Finally, the priming concept predicts shorter latency for valid as compared to neutral or invalid cues. Saccadic latencies, eventhough shorter for valid than for invalid cues, are shortest in the gap task with no spatial cues (Biscaldi et al.1989). This means that the knowledge of the future target location if anythimg increases the reaction time rather than decreasing it. The disengagement- engagement concept offers an explanation of this experimental finding by assuming that the cue - valid or not - may catch ones attention thereby leading to an increase in latency.
We want to emphasize the fact that the model has a rather simple structure consisting essentially of three identical central elements. Yet it produces rather complex sets of data which very closely resemble what one sees in real life experiments on saccadic reaction times. It remains open for future mathematical work to elaborate the model (Aiple, Huber, in prep.) and to embed it into other brain functions, like for example visual perception and/or visual-vestibular interaction and/or locomotion.
Acknowledgements: This work was supported by the Deutsche Forschungsgemeinschaft, SFB 325, Teilprojekt C5 and C7. The help of Dr. Franz Aiple and cand. phys. Wolfgang Huber in computer programming is greatfully acknowledged.REFERENCES
Albano JE, King WM (1989) Rapid adaptation of saccadic amplitude in humans and monkeys. Investigations in Ophthalmology and Visual Science 30: 1883-1893
Aslin R, Salapatek P (1975) Saccadic localisation of visual targets by the very young human infant. Perception & Psychophysics 17: 293-302
Astruc J (1971) Corticofugal connections of area 8 (frontal eye field) in macaca mulatta. Brain Research 33: 241-256
Barbas H, Mesulam M-M (1980) Differential afferent input to subdivisions within the frontal eye fields (area 8) of macaque. Neuroscience Abstracts 6 Nr. 114.4: 316
Becker W (1972) The control of eye movements in the saccadic system. In: Dichgans J, Bizzi E (eds) Cerebral Control of Eye Movements and Motion Perception (Bibl. Ophthalmol., Vol 82). Karger, Basel, pp 233-243
Becker W (1989) Metrics. In: Reviews of Oculomotor Research 3. Wurtz R, Goldberg M (eds) The Neurobiology of Saccadic Eye Movements. Elsevier Amsterdam, pp 13-67
Biscaldi M, Boch R, Fischer B (1989) Effects of directed attention on saccadic and manual reaction times. Perception 18: 521
Biscaldi M, Fischer B (1992) Saccadic eye movements of dyslexic children in non-cognitive tasks. In: Wright S, Groner R (eds) Studies in visual information processing. Elsevier North Holland
Boch R (1989) Saccadic reaction times after chemical lesions in striate and prestriate cortex of the rhesus monkey. Investigations in Ophthalmology and Visual Science 30: 184
Boch R, Fischer B (1986) Further observations on the occurrence of express-saccades in the monkey. Experimental Brain Research 63: 487-494
Boch R, Fischer B, Ramsperger E (1984) Express-saccades of the monkey: reaction times versus intensity, size, duration, and eccentricity of their targets. Experimental Brain Research 55: 223-231
Braun D, Breitmeyer BG (1988) Relationship between directed visual attention and saccadic reaction times. Experimental Brain Research 73: 546-552
Braun D, Weber H, Mergner T, Schulte-Mnting J (1992) Saccadic reaction times in patients with frontal and parietal lesions. Brain in print.
Bronstein AM, Kennard C (1987) Predictive eye saccades are different from visually triggered saccades. Vision Research 27: 517-520
Bruce C, Goldberg M (1984) Physiology of the frontal eye fields. Trends in Neuroscience 7: 436-441
Deubel H, Wolf W, Hauske G (1986) Adaptive gain control of saccadic eye movements. Human Neurobiology 5: 245-253
Findlay JM (1982) Global visual processing for saccadic eye movements. Vision Research 22: 1033-1045
Fischer B (1986) Express saccades in man and monkey. Progress in Brain Research 64: 155-160
Fischer B (1987) The preparation of visually guided saccades. Reviews in Physiology, Biochemistry and Pharmacology 106: 1-35
Fischer B, Boch R (1983) Saccadic eye movements after extremely short reaction times in the monkey. Brain Research 260: 21-26
Fischer B, Boch R (1990) Cerebral Cortex. In: Carpenter R (ed) Vision and Visual Dysfunction: Vol. 9: Eye movements. The Macmillan Press Ltd.
Fischer B, Boch R, Ramsperger E (1984) Express-saccades of the monkey: effect of daily training on probability of occurrence and reaction time. Experimental Brain Research 55: 232-242
Fischer B, Ramsperger E (1984) Human express saccades: extremely short reaction times of goal directed eye movements. Experimental Brain Research 57: 191-195
Fischer B, Ramsperger E (1986) Human express saccades: effects of randomization and daily practice. Experimental Brain Research 64: 569-578
Fischer B, Rogal L (1986) Eye-hand-coordination in man: a reaction time study. Biological Cybernetics 55: 253
Fischer B, Weber H (1990) Saccadic reaction times of dyslexic and age-matched normal subjects. Perception 19: 805-818
Fischer B, Weber H (1992) Characteristics of "anti" saccades in man. Experimental Brain Research 89: 415-424
Fischer B, Weber H, Biscaldi M, Aiple F, Otto P, Stuhr V. Separate populations of visually guided saccades in humans: reaction times and amplitudes. Submitted to Experimental Brain Research
Fitzgibbon E, Goldberg M, Segraves M (1985) Short term saccadic adaptation in the monkey. In: Keller E, Zee D (eds) Adaptive processes in visual and oculomotor systems. Pergamon Press, pp 329-333
Glickstein M, May JG, Mercier BE (1985) Corticopontine projection in the macaque: the distribution of labelled cortical cells after large injections of horseradish peroxidase in the pontine nuclei. Journal of comparative Neurology 235: 343-359
Goldberg ME, Bushnell MC, Bruce CJ (1986) The effect of attentive fixation on eye movements evoked by electrical stimulation of the frontal eye fields. Experimental Brain Research 61: 579-584
Goldberg M, Wurtz R (1972) Activity of the superior colliculus in the behaving monkey. II Effect of attention on neuronal responses. Journal of Neurophysiology 35: 560-574
Guitton D, Buchtel HA, Douglas RM (1985) Frontal lobe lesions in man cause difficulties in suppressing reflexive glances and in generating goal-directed saccades. Experimental Brain Research 58: 455-472
Haddad GM, Steinman RM (1973) The smallest voluntary saccade: Implications for fixation. Vision Research 13: 1075-1086
Haenny P, Schiller P (1988) State dependent activity in monkey visual cortex. I. Single cell activity in V1 and V4 on visual tasks. Experimental Brain Research 69: 225-244
Harting JK, Huerta MF, Frankfurter AJ, Strominger NL, Royce GJ (1980) Ascending pathways from the monkey superior colliculus: An autoradiographic analysis. Journal of comparative Neurology 192: 853-882
Henson DB (1978) Corrective saccades: effects of altering visual feedback. Vision Research 18: 63-67
Hikosaka O, Wurtz RH (1983) Effects on eye movements of a GABA agonist and antagonist injected into monkey superior colliculus. Brain Research 272: 368-372
Jttner M, Wolf W (1992) Occurrence of human express saccades depends on stimulus uncertainty and stimulus sequence. Experimental Brain Research 89: 678-681
Kalesnykas RP, Hallett PE (1987) The differentiation of visually guided and anticipatory saccades in gap and overlap paradigms. Experimental Brain Research 68: 115-121
Kapoula Z (1985) Evidence for a range effect in the saccadic system. Vision Research 25: 1155-1157
Kapoula Z, Robinson DA (1986) Saccadic undershoot is not inevitable: saccades can be accurate. Vision Research 26: 735-743
Kimmig H (1986) Express Sakkaden beim Menschen: die Rolle der Aufmerksamkeit in der Vorbereitungsphase zielgerichteter Blicksprnge. Dissertation, Fachbereich Medizin Universitt Freiburg
Kowler E (1990) The role of visual and cognitive processes in the control of eye movement. In reviews of oculomotor research 4: Kowler E (ed) Eye movements and their role in visual and cognitive processes. Elsevier Science Publishers, p 1ff.
Knzle H, Akert K, Wurtz RH (1976) Projection of area 8 (frontal eye field) to superior colliculus in the monkey. An autoradiographic study. Brain Research 117: 487-492
Latanov A, Weber H, Fischer B. The double step effect on express saccades in man and monkey. Submitted to Experimental Brain Research
Latto R, Coway A (1971) Fixation changes after frontal eye-field lesions in monkeys. Brain Research 30: 25-36
Mayfrank L, Mobashery M, Kimmig H, Fischer B (1986) The role of fixation and visual attention in the occurrence of express saccades in man. European Archive of Psychiatry and Neurological Sciences 235: 269-275
Miller JM, Anstis T, Templeton WB (1981) Saccadic plasticity: parametric adaptive control by retinal feedback. Journal of Experimental Psychology [Human Perception] 7: 356-366
Milner B (1982) Some cognitive effects of frontal-lobe lesions in man. Philosophical Transactions of the Royal Society of London [Biology] 298: 211-226
Moschovakis AK, Karabelas AB, Highstein SM (1988) Structure-function relationships in the primate superior colliculus. II. Morphological identity of presaccadic neurons. Journal of Neurophysiology 60: 263-302
Mountcastle V, Motter B, Steinmetz M, Sestokas A (1987) Common and differential effects of attentive fixation on the excitability of parietal and prestriate (V4) cortical visual neuronsin the macaque monkey. Journal of Neuroscience 7: 2239- 2255
Oakley M, Eason R (1990) The conjoint influence of spatial selective attention and motor set on very short latency VERs. Neuropsychologia 28: 487-497
Olson RK, Kliegl R, Davidson BJ (1983) Dyslexic and normal readers' eye movements. Journal of Experimental Psychology [Human Perception] 9: 816-825
Paige GD, Sargent EW (1991) Visually-induced adaptive plasticity in the human vestibulo-ocular reflex. Experimental Brain Research 84: 25-34
Pavlidis GT (1981) Do eye movements hold the key to dyslexia? Neuropsychologia 19: 57-64
Pavlidis GT (1985) Eye movement differences between dyslexics, normal, and retarded readers while sequentially fixating digits. American Journal of Optometry and Physiological Optics 62: 820-832
Perry VH, Cowey A (1984) Retinal ganglion cells that project to the superior colliculus and pretectum in the macaque monkey. Neuroscience 12: 1125-1137
Pierrot-Deseilligny C, Rivaud S, Gaymard B, Agid Y (1991) Cortical control of memory-guided saccades in man. Experimental Brain Research 83: 607-617
Posner MI, Walker JA, Friedrich FJ, Rafal RD (1984) Effects of parietal injury on covert orienting of attention. Journal of Neuroscience 4: 1863-1874
Rashbass C (1961) The relationship between saccadic and smooth tracking eye movements. Journal of Physiology 159: 326-338
Reulen JPH (1984) Latency of visually evoked saccadic eye movements. Biological Cybernetics 50: 251-262
Reuter-Lorenz P A, Hughes H C, Fendrich R (1991) The reduction of saccadic latency by prior offset of the fixation point: An analysis of the gap effect. Perception & Psychophysics 49, 167- 175
Robinson D, Goldberg M, Stanton G (1978) Parietal association cortex in the primate: sensory mechanisms and behavioral modulations. Journal of Neurophysiology 41: 910-932
Rogal L, Fischer B (1986) Eye-hand-coordination: a model for computing reaction times in a visually guided reach task. Biological Cybernetics 55: 263-273
Rogal L, Reible G, Fischer B (1985) Reaction times of the eye and the hand of the monkey in a visual reach task. Neuroscience Letters 58: 127-132
Ross L, Ross S (1980) Saccade latency and warning signals: stimulus onset, offset, and change as warning events. Perception & Psychophysics 27: 251-257
Ross M, Ross L (1981) Saccade latency and warning signals: effects of auditory and visual stimulus onset and offset. Perception & Psychophysics 29: 429-437
Saslow MG (1967) Latency for saccadic eye movement. Journal of the Optic Society of America 57: 1030-1033
Schiller PH, Sandell JH, Maunsell JH (1987) The effect of frontal eye field and superior colliculus lesions on saccadic latencies in the rhesus monkey. Journal of Neurophysiology 57: 1033-1049
Shibutani H, Sakata H, Hyvarinen J (1984) Saccade and blinking evoked by microstimulation of the posterior parietal association cortex of the monkey. Experimental Brain Research 55: 1-8
Smit AC, van Gisbergen JA (1989) A short-latency transition in saccade dynamics during square-wave tracking and its significance for the differentiation of visually-guided and predictive saccades. Experimental Brain Research 76: 64-74
Tanaka M, Weber H, Creutzfeldt OD (1986) Visual properties and spatial distribution of neurones in the visual association area on the prelunate gyrus of the awake monkey. Experimental Brain Research 65: 11-37
Tootell RB, Hamilton SL, Switkes E (1988) Functional anatomy of macaque striate cortex. IV. Contrast and magno-parvo streams. Journal of Neuroscience 8: 1594-1609
Weber H, Aiple F, Fischer B, Latanov A (1992) Dead zone for express saccades. Experimental Brain Research 89: 214-222
Weber H, Fischer B (1990a) Cortical mechanisms for vision revealed by saccadic eye movements in trained rhesus monkeys. In: Iwai E (ed) Vision, Memory and the Temporal Lobe. Elsevier Science Publishing, pp 323-330
Weber H, Fischer B (1990b) Effect of a local ibotenic acid lesion in the visual association area on the prelunate gyrus (area V4) on saccadic reaction times in trained rhesus monkeys. Experimental Brain Research 81: 134-139
Weber H, Fischer B, Bach M, Aiple F (1991) Occurrence of express saccades under isoluminance and low contrast luminance conditions. Visual Neuroscience 7: 505-510
Weber H, Latanov A, Fischer B. Context dependent gain changes of express and regular saccades in man and monkey. Submitted to Experimental Brain Research
Wenban-Smith M, Findlay J (1991) Express saccades: is there a separate population in humans? Experimental Brain Research 87: 218-222
Wyman D, Steinman RM (1973a) Small step tracking: Implications for the oculomotor "dead zone". Vision Research 13: 2165-2172
Wyman D, Steinman RM (1973b) Latency characteristics of small saccades. Vision Research 13: 2173-2175
Fig. 1: Spatial (left) and temporal (right) arrangements for a typical Gap (upper) and Overlap (lower) paradigm used in studies in man. Unless stated otherwise the target stimuli were always presented randomly 4 deg to the left and right of the fixation point.
Fig 2.: Effect of gap duration on the distributions of saccadic reaction times in man. The reaction times of right (R) and left (L) directed saccades are accumulated. The gap durations, numbers (N) of saccades, the mean values, and the standard deviations are given in each panel. Anticipatory saccades with latencies below 80 msec are excluded from the calculations.
Fig. 3: Effect of central (left part) and peripheral (right part) attention on the distribution of saccadic reaction times in gap (A and D), overlap (B and E), and overlap with blink trials (C and F). Data from a single subject with target location randomized between right and left. Further description in the text.
Fig. 4: The distribution of saccadic reaction times of 2114 saccades (randomly to the right or left) from a single subject. The reaction times of the 75 direction errors the subject made unvoluntarily are depicted below.
Fig. 5: Distributions of saccadic reaction times in a normal gap task (upper), and in an anti gap task (lower). Note the absence of an express peak in the lower graph. The middle panel shows the reaction times of the saccades the subjects made unvoluntarily to the target rather than in the opposite direction as required in the anti task. Note the presence of an express peak. The data are taken from Fischer and Weber, 1992.
Fig. 6: Scatter plot of saccadic reaction time versus saccade size for right (positive) and left (negative) directed saccades of a subject, who produced an asymmetric pattern of reaction times. The data points below 80 msec are from anticipatory saccades to the right as well as from direction errors, i. e saccades to the right when the target occurred at the left. Note the complete absence of anticipations to the left side. The graph also shows the difference in amplitude of express versus regular saccades.
Fig. 7: Saccadic reaction times obtained in the overlap task (A) and with an additional tone (B and C). In B the subjected attended to the fixation point ignoring the tone, in C the subject attended the tone ignoring the fixation point. Note that the distribution turns into a bimodal one when the tone is attended.
Fig. 8: Scatter plot of saccade size versus reaction time. Express saccades can be seen between 2 and 10 degrees. Fast regular saccades occur between 0.5 and 4 degrees. Below 0.5 degrees reaction times increase drastically. The data are taken from Weber et al.1992.
Fig. 9: Saccadic reaction times from trials with randomly mixed gap durations in different combinations. Part A depicts the data from gap 100ms trials when mixed with gap zero trials. Part B shows the results from 100 ms trials when mixed with gap 200 ms trials. Part C: same but for the gap 200 ms trials when mixed with gap 100 ms trials. Part D: data from gap 200 ms trials mixed with gap 300 ms trials.
Fig. 10: Effect of age on the distribution of saccadic reaction time. Note the strong asymmetry between right and left directed saccades for the adult group (upper panel). In all cases the target was randomly presented to the right or left.
Fig. 11: Saccadic reaction times of normal and two groups of dyslexic readers obtained in overlap trials. Note the big population of express saccades in the group II dyslexics. The data are taken from Biscaldi et al., submitted paper.
Fig. 12: Schematic drawing of the anatomical projections connecting the retina of the eye with the efferent eye movement generating structures (EM) in the region of the brain stem. Three loops can be seen as indicated by the single, the double, and triple heavy lines. V1 = striate visual cortex (the lateral geniculate nucleus is ommitted), Pre = prestriate visual cortex including areas V2 to V5 and area MST, Par = parietal cortex, FEF = frontal eye field, SC = superior colliculus, Nc = nucleus caudatus, Sn = substantia nigra pars reticulata. Figure taken from Weber et al.1992.
Fig. 13: Outline of the three loop model as described in the text. The lower part shows the result of two simulations of the model: the gap condition is depicted above and the overlap condition below. Note that the difference between the two is not a simple decrease of reaction time but rather a difference in the size of the peaks in the distributions.