Fixational eye movements across vertebrates: Comparative dynamics, physiology, and perception

Abstract

During visual fixation, human eyes are never still. Instead, they constantly produce involuntary “fixational eye movements.” Fixational eye movements overcome neural adaptation and prevent visual fading: thus they are an important tool to understand how the brain makes the environment visible. The last decade has seen a growing interest in the analysis of fixational eye movements in humans and primates, as well as in their perceptual and physiological consequences. However, no comprehensive comparison of fixational eye movements across species has been offered. Here we review five decades of fixational eye movement studies in non-human vertebrates, and we discuss the existing evidence concerning their physiological and perceptual effects. We also provide a table that summarizes the physical parameters of the different types of fixational eye movements described in non-human vertebrates.

Introduction

When we fixate our gaze on an object of interest, our eyes are never still. Instead we constantly produce small involuntary eye movements, generally called “fixational eye movements.” If these eye movements are eliminated, our perception of stationary objects fades, due to neural adaptation (Ditchburn & Ginsborg, 1952; Martinez-Conde, Macknik, Troncoso, & Dyar, 2006; Riggs & Ratliff, 1952; Troncoso, Macknik, & Martinez-Conde, 2008). When our eyes are free to move across the image once again, visual perception reappears (Yarbus, 1967). Due to their role in counteracting visual adaptation, fixational eye movements are an important tool to understand how the brain makes the environment visible, both in normal and pathological vision (Martinez-Conde, 2006). Further, because we fixate our gaze most of the time during visual exploration (Martinez-Conde, 2006; Otero-Millan, Troncoso, Macknik, Serrano-Pedraza, & Martinez-Conde, 2008), fixational eye movements are often responsible for driving our visual experience. 

Fixational eye movements can help us understand the underpinnings of visual awareness (Martinez-Conde & Macknik, 2007b) in a number of ways:

1.  
   Fixational eye movements drive the visibility (and counteract the fading of) stationary objects during fixation. Thus fixational eye movements can help constrain the spatiotemporal characteristics of visible stimuli. Moreover, the neural responses triggered by fixational eye movements along the visual pathway must encompass the neural code for visibility (Martinez-Conde, 2006; Martinez-Conde, Macknik, & Hubel, 2000, 2002).
2.  
   Fixational microsaccades in human subjects may drive perceptual alternations for a variety of multi-stable stimuli (Martinez-Conde, 2006; Martinez-Conde & Macknik, 2007b; Martinez-Conde et al., 2006; Troncoso, Macknik, & Martinez-Conde, 2008; Troncoso, Macknik, Otero-Millan, & Martinez-Conde, 2008; van Dam & van Ee, 2006).
3.  
   Fixational microsaccades may indicate attentional or cognitive engagement (Engbert & Kliegl, 2003b; Galfano, Betta, & Turatto, 2004; Hafed & Clark, 2002; Martinez-Conde & Macknik, 2007b; Otero-Millan et al., 2008).
4.  
   Human subjects are unaware of their fixational eye movements (Martinez-Conde, Macknik, & Hubel, 2004). That is, despite the continuous motion caused by fixational eye movements, the world remains perceptually stable when we fixate. This perceptual stability can be foiled: the “visual jitter” illusion shows that in the absence of fixational eye movement compensation the world is seen as unstable and jittery (Murakami & Cavanagh, 1998). Thus the neuronal circuits, populations, etc., that constitute the neural correlates of visual awareness must sustain perceptual stability during fixation.

Research in fixational eye movements is one of the fastest moving fields of vision research today. The last decade has seen a proliferation of fixational eye movement studies, including a few reviews of the published literature (Engbert, 2006; Martinez-Conde, 2006; Martinez-Conde et al., 2004). However, the differences and similarities in fixational eye movements across species, as well as their potential significance, have been neglected so far. Despite wide-reaching interest in this topic (Martinez-Conde & Macknik, 2007a), a comprehensive comparative review has not been published. Here we review the comparative dynamics, the physiology, and the perceptual consequences of fixational eye movements across all species studied to date. To the best of our knowledge, our review includes all Medline-indexed studies (as well as several book chapters and other non-indexed publications) of fixational eye movements in non-human vertebrates. These span 48 years of research: from the first studies in the cat, published in 1960 (Hebbard & Marg, 1960; Pritchard & Heron, 1960), to the most recent studies in the salamander and archer fish, published in 2007 and 2008 (Baccus, Olveczky, Manu, & Meister, 2008; Olveczky, Baccus, & Meister, 2007; Segev, Schneidman, Goodhouse, & Berry, 2007). As a result of our analysis we have generated a table (Table 1) that summarizes the physical parameters of the different types of fixational eye movement across non-human vertebrates. We will relate, whenever possible, fixational eye movements in non-human vertebrates to human fixational eye movements. 

Table 1

 

Characteristics of fixational eye movements in non-human vertebrates, according to different studies.

	Type of eye movement	Amplitude	Frequency/ intersaccadic interval	Duration	Max speed	Mean speed	Conjugate	Study
Mammals
Primate	Microsaccade	8.4–16.2 min (means for 2 monkeys)	2.3–2.5 Hz (mean frequencies for 2 monkeys)	at least 8 msec	–	–	–	Horwitz and Albright ( 2003)
	Microsaccade	~40 min (mean)	–	–	–	~30 deg/sec	–	Snodderly et al. ( 2001)
	Microsaccade	~20 min (mean)	~3–5 Hz	29 msec (mean)	–	~30 deg/sec	–	Martinez-Conde et al. ( 2000)
	Microsaccade	48 min (mean)	0.3–1.4 Hz (means for 2 monkeys)	25 ms (mean)	9–110 deg/sec (median: 40 deg/sec)	~30 deg/sec	–	Bair and O'Keefe ( 1998)
	Microsaccade	10.1 min (median)	0.597 sec (median)	20 msec (mean)	–	–	–	Leopold and Logothetis ( 1998)
	Microsaccade	9.9–40.3 min (medians for 4 monkeys)	0.8–7.4 sec (medians for 4 monkeys)	–	–	–	–	Skavenski et al. ( 1975)
	Microsaccade	40 min (mean; minimum amplitude: 23 min)	–	–	–	–	–	Steinman et al. ( 1973)
	Drift	–	–	–	–	6 min/sec (mean for 1 monkey)	–	Bair and O'Keefe ( 1998)
	Drift	–	–	–	–	0.42–11.91 min/sec (means for 4 monkeys) a	–	Skavenski et al. ( 1975)
Cat	Microsaccade	“somewhat smaller” than in humans	“far fewer” than in humans	–	–	–	No (mostly monocular, and some binocular)	Hebbard and Marg ( 1960)
	Microsaccade	35 min b	–	–	–	–	–	Pritchard and Heron ( 1960)
	Drift	–	–	–	–	14.7 min/sec (mean for 3 cats)	–	Winterson and Robinson ( 1975)
	Drift	–	–	–	–	–	–	Hebbard and Marg ( 1960)
	Drift	“Usually ~25 min, but can exceed 2 deg” b	–	–	–	30 min/sec	–	Pritchard and Heron ( 1960)
	Tremor	5.6–73.3 s (average 31 s) b	35–65 Hz (average 50 Hz)	–	–	–	No	Hebbard and Marg ( 1960)
	Tremor	0.4 min b	2–40 Hz (unable to record faster rates)	–	–	–	–	Pritchard and Heron ( 1960)
Rabbit	Drift	–	–	–	–	25–50 min/sec	–	Van der Steen and Collewijn ( 1984)
	Drift	=4 deg b	–	–	=1 deg/sec (unable to record faster rates)	1.5 min/sec	–	Collewijn and van der Mark ( 1972)
	Tremor	“small amplitude”	–	–	–	–	–	Collewijn and van der Mark ( 1972)
Birds
Pigeon	Microsaccade	–	“very low”	<20 msec	–	–	–	Nye ( 1969)
	Impulse	Up to 2 deg	–	–	–	–	–	Nye ( 1969)
	Drift	3–5 deg (occasionally exceeding the limit of the recording system: 7.5 deg)	–	–	–	1–5 deg/sec	–	Nye ( 1969)
	Oscillation	Up to several deg (peak-to-peak)	Short bursts with a frequency of 28–35 Hz, usually separated by intervals of 1–2 sec	Up to 0.8 sec	–	–	–	Nye ( 1969)
Owl	Microsaccade	=1 deg	–	–	10 deg/sec	–	–	Steinbach and Money ( 1973)
	Drift	=1 deg	–	–	–	–	–	Steinbach and Money ( 1973)
	Tremor	1 min	20 Hz	–	–	–	–	Steinbach and Money ( 1973)
	Oscillation	1.5 deg	Short bursts with a frequency of 20 Hz, about 4 times a minute	0.25 sec	–	–	–	Steinbach and Money ( 1973)
Reptiles
Turtle	Drift-like	–	–	–	–	4.8 ± 2 min/sec	–	Greschner et al. ( 2002)
	Periodic	Up to 5 min	5 Hz	–	Up to 21.4 deg/sec	–	–	Greschner et al. ( 2002)
Amphibians
Salamander	Tremor-like	About 12.5 min	–	–	–	–	–	Manteuffel et al. ( 1977)
	Respiratory	Up to 30 min	–	–	–	–	–	Manteuffel et al. ( 1977)
Fish
Goldfish	Drift	–	–	0.1–5.5 sec	=1 deg/sec	0.36 deg/sec	No. Simultaneous in both eyes but convergent in the light	Mensh et al. ( 2004)
	Drift	Typically <5 deg	–	–	–	–	No. Simultaneous in both eyes but convergent	Easter ( 1971)
Archer Fish	Microsaccade	–	“rarely” found	–	–	–	–	Segev et al. ( 2007)
	Drift	–	0.2–1 Hz	1.33 sec	=0.4 deg/sec	–	–	Segev et al. ( 2007)
	Tremor	0.1–0.2 deg	5 Hz	–	4 deg/sec	~1 deg/sec	–	Segev et al. ( 2007)

 

^aCalculated from horizontal and vertical components.

 

^bVector values from H-components: a conversion factor of √2 has been assumed.

Fixational eye movements in mammals

Fixational eye movements in the primate

Fixational eye movements were originally discovered and characterized in human subjects (Ditchburn & Ginsborg, 1952, 1953; Ratliff & Riggs, 1950). Human fixational eye movements comprise three main types: microsaccades (called “flicks” in early studies), drift, and tremor (Figure 1). Microsaccades are the fastest and largest of the three types of fixational eye movements (Supplementary Video 1). They travel in a straight trajectory, carrying the retinal image across a range of several dozen to several hundred photoreceptor widths. Drifts are slow curvy motions that occur between microsaccades, and tremor is a very fast (∼90 Hz) and extremely small oscillation (about the diameter of a foveal cone) superimposed on drifts (see Tables 1–3 of Martinez-Conde et al., 2004 for a review of fixational eye movement parameters in humans). 

Figure 1

 

Cartoon representation of fixational eye movements in humans and primates. Microsaccades (straight and fast movements), drifts (curvy slow movements), and tremor (oscillations superimposed on drifts) transport the visual image across the retinal photoreceptor mosaic. From Martinez-Conde and Macknik (2007b).

Fixational eye movements in old-world monkeys are very similar to human fixational eye movements (Martinez-Conde, 2006; Martinez-Conde et al., 2000, 2002; Skavenski, Robinson, Steinman, & Timberlake, 1975; Snodderly, 1987; Snodderly & Kurtz, 1985). Several macaque species have been studied to date: Macaca mulatta (Goffart, Quinet, Chavane, & Masson, 2006; Horwitz & Albright, 2003; Leopold & Logothetis, 1998; Martinez-Conde, 2006; Martinez-Conde et al., 2000, 2002; Motter & Poggio, 1984; Skavenski et al., 1975; Snodderly, Kagan, & Gur, 2001; Steinman, Haddad, Skavenski, & Wyman, 1973), Macaca fascicularis (Snodderly, 1987; Snodderly et al., 2001; Snodderly & Kurtz, 1985), and Macaca nemestrina (Bair & O'Keefe, 1998), with no substantial differences in the spatiotemporal characteristics of fixational eye movements (rate, velocity and magnitude profiles, etc.) between them. Microsaccade velocities are parametrically related to microsaccade amplitudes, following the “main sequence” in both macaques (Martinez-Conde et al., 2000) and humans (Engbert, 2006; Martinez-Conde, 2006; Martinez-Conde et al., 2006; Zuber & Stark, 1965). Fixational microsaccades have also been observed in the baboon (Papio papio), with amplitudes up to 2.5 degrees (Marchetti, Gauthier, & Pellet, 1983). 

It is important to note that microsaccades cannot be defined according to their amplitude alone, because small exploratory or voluntary saccades can be the same size as microsaccades. Microsaccades can be defined only operationally, as the involuntary saccades that are produced while the subject attempts to fixate (Martinez-Conde, 2006). This “attempt to fixate” includes both holding the eyes still and static while foveating the visual target and also orienting the fovea toward a target. It is worth noting that microsaccade production may require the presence of a visual and/or attentional target (Otero-Millan et al., 2008). 

Thus, there is no known physical parameter (or combination of parameters) that separates saccades from microsaccades. Mounting evidence points toward a common neural generator of saccades and microsaccades (Engbert, 2006; Martinez-Conde et al., 2004; Otero-Millan et al., 2008; Rolfs, Kliegl, & Engbert, 2008; Rolfs, Laubrock, & Kliegl, 2006; Zuber & Stark, 1965). However, one should note that most of the available data in support of a shared microsaccade–saccade generator are indirect. To date, only two physiological studies have directly addressed the question of the oculomotor mechanisms leading to the generation of microsaccades. In those studies, Van Gisbergen and colleagues found that putative motoneurons in the primate abducens nucleus and burst neurons in the nearby pontomedullary reticular formation were similarly active during saccades and microsaccades (Van Gisbergen, Robinson, & Gielen, 1981; Van Gisbergen & Robinson, 1977). Future research should further explore the neural bases underlying the generation of microsaccades and other fixational eye movements. 

Simultaneous recordings of microsaccades and neural responses have been conducted at multiple levels of the primate visual pathway, including the lateral geniculate nucleus (LGN), area V1, and the extrastriate visual cortex (Bair & O'Keefe, 1998; Leopold & Logothetis, 1998; Martinez-Conde, 2006; Martinez-Conde et al., 2000, 2002; Reppas, Usrey, & Reid, 2002; Snodderly et al., 2001). Microsaccades are predominantly excitatory at all these levels, leading to firing rate increases in both early and higher visual areas. The firing rate increases following microsaccades in the LGN and V1 are of visual origin (Figure 2) and result from the microsaccade-induced displacement of visual receptive fields over stationary stimuli. Microsaccade-driven increases in firing rate tend to be clustered in tight bursts of spikes (Martinez-Conde, 2006; Martinez-Conde et al., 2000, 2002). A recent study has suggested that primate microsaccades may improve the efficient sampling of fine spatial detail (Donner & Hemilä, 2007). 

Figure 2

 

Microsaccades drive neural activity in primate area V1. Microsaccades lead to firing rate increases in V1 neurons (n = 308) of the awake primate. In the absence of visual stimulation, microsaccades do not generate spikes in area V1 (n = 37 neurons). The results—shown here for the average of all neurons—were consistent with responses from individual neurons (data not shown). Modified from Martinez-Conde et al. (2000, 2002).

Microsaccades counteract visual fading and increase visibility during fixation in human subjects (Martinez-Conde et al., 2006; Troncoso, Macknik, & Martinez-Conde, 2008). Numerous studies have also reported that human microsaccade rates and/or directions are modulated by cognitive processes, such as the allocation of spatial attention (Betta & Turatto, 2006; Engbert & Kliegl, 2003a, 2003b, 2004; Galfano et al., 2004; Hafed & Clark, 2002; Laubrock, Engbert, Rolfs, & Kliegl, 2007; Otero-Millan et al., 2008; Turatto, Valsecchi, Tamè, &, Betta, 2007; Valsecchi, Betta, & Turatto, 2007; but see also Horowitz, Fine, Fencsik, Yurgenson, & Wolfe, 2007 and Tse, Sheinberg, & Logothetis, 2004). 

Neural responses to drifts have received considerably less attention than neural responses to microsaccades. This may be due to the fact that drifts are more difficult to characterize objectively than microsaccades (which are more easily detected by automatic algorithms that combine amplitude and velocity thresholds). Thus drifts are usually identified indirectly as the eye position changes that occur between microsaccades (Snodderly et al., 2001). However, this method has the potential flaw that one may unintentionally attribute non-drift-related activity (for instance, undetected tremor) to drifts. Drifts may increase firing in a subset of primate V1 neurons (Snodderly et al., 2001). However, they generate less variability in neuronal responses than a combination of drifts and microsaccades (Gur, Beylin, & Snodderly, 1997). No studies to date have recorded primate neuronal responses in correlation to tremor (but see Hennig & Wörgötter, 2007 for the responses of a model macaque retina to simulated drift and tremor). 

Fixational eye movements in the cat

The cat's fixational eye movements have been investigated in a handful of studies, and their conclusions remain somewhat controversial. In 1960, Hebbard and Marg (1960) found fixational eye movements in the cat to be quite comparable to those in humans, although cat microsaccades were scarcer and smaller than in humans. Pritchard and Heron (1960) also found that the cat's fixational eye movements included microsaccades, drifts, and tremor. Microsaccades were much rarer in the cat than in humans, however, possibly related to the fact that the cat lacks a well-developed fovea. Delgado-Garcia, del Pozo, and Baker (1986) also observed some putative microsaccades in the alert behaving cat. Other studies found no microsaccades in the cat (Conway, Timberlake, & Skavenski, 1981; Winterson & Robinson, 1975). However, in these last studies microsaccades were defined as “saccades smaller than 10 minutes of arc”, which is well below the average microsaccade amplitude found in recent human and primate studies (see Martinez-Conde et al., 2004 for a review of microsaccade parameters in humans and primates). More recently, several studies have addressed the contribution of fixational eye movements to neural responses in the cat's visual system. Hennig, Kerscher, Funke, and Wörgötter (2002) found increased responses from neurons in cat areas 17 and 18 during stimulus motions that mimicked fixational eye movements. Using a model retina, they also found that:

1.  
   simulated fixational eye movements increased neural responses, and
2.  
   simulated tremor improved spatial acuity.

Hennig and Wörgötter ( 2004) later reported that a combination of tremor and microsaccades improved the performance of simulated ganglion cells in a hyperacuity task in the retinal periphery (but not in the central retina). Miller, Denning, George, Marshak, and Kenyon (2006) also investigated the effects of simulated tremor on a model of the cat's retina and suggested that tremor may selectively enhance the processing of large stimuli.

Fixational eye movements in the rabbit

In contrast to foveate species, such as primates (including humans) and cats (which have an area centralis, if not a proper fovea), the rabbit ( Oryctolagus cuniculus) displays few spontaneous eye movements, especially in the absence of head motion (Collewijn, 1977; Collewijn & van der Mark, 1972; Fuller, 1980, 1981; Van der Steen & Collewijn, 1984). In normal visual conditions, the rabbit's eyes are very stable but not perfectly immobile. Small amplitude tremor can be observed, as well as very slow drift. Microsaccades have not been observed. However, drift is occasionally corrected by fast saccadic movements (Collewijn, 1977; Collewijn & van der Mark, 1972; Van der Steen & Collewijn, 1984). These differences may be related to the fact that the rabbit does not have a fovea but an elongated horizontal streak with elevated ganglion cell counts. Olveczky, Baccus, and Meister (2003) recorded from ganglion cells in the rabbit and salamander retinas while presenting visual stimuli that were jittered to simulate fixational eye movements (based on parameters from previously published studies by Manteuffel, Plasa, Sommer, & Wess, 1977 and Van der Steen & Collewijn, 1984). In both species, fixational eye movements helped segregate a moving foreground from a complex background in a population of ganglion cells. 

Fixational eye movements in birds

Fixational eye movements in the pigeon

There are some remarkable differences between fixational eye movements in birds and mammals. Nye (1969) identified four types of fixational eye movements in the pigeon (Columba livia): flicks (i.e., microsaccades), impulse movements, drift, and oscillations. No tremor was detected. Microsaccade frequency was very low relative to humans. Impulse movements occurred more frequently, usually between bursts of oscillations, with amplitudes ranging up to 2 degrees. Drift velocities ranged from 1 to 5 degrees per second. Both microsaccades and drifts occurred in response to moving stimuli. Oscillations occurred in short bursts at rates of about 30 Hz and could be quite large (up to several degrees of amplitude). Bloch, Rivaud, and Martinoya (1984) reported microsaccades, drifts, and oscillations “similar to those described by Nye,” but none of these were analyzed in detail. 

Cyclotorsional oscillations are an integral part of avian saccades (Wallman, Pettigrew, & Letelier, 1994). Thus one might speculate that oscillations are associated with avian microsaccades as well, especially if saccades and microsaccades share a common generator (Engbert, 2006; Otero-Millan et al., 2008; Rolfs et al., 2008, 2006; Zuber & Stark, 1965). However, this possibility has not yet been addressed in the literature. The function of oscillations is unknown (Carpenter, 1988), but one hypothesis is that they serve to facilitate the delivery of arterial blood to the ocular capillary bed (Pettigrew, Wallman, & Wildsoet, 1990; Wallman et al., 1994). 

Fixational eye movements in the owl

Steinbach and Money (1973) recorded fixational eye movements in the great horned owl (Bubo virginianus), a species previously thought incapable of moving its eyes. The owl's fixational eye movements include: flicks or microsaccades, drift, tremor, and oscillations. Microsaccades seemed correlated to visual events of interest to the owl, they were 1 deg or less in amplitude and had peak velocities of about 10 deg/sec. Drifts were slow motions of the eye in between fixations, with amplitudes of 1 degree or less. Tremor amplitude was about 1 arcmin, with frequencies around 20 Hz. High-frequency oscillations (25 Hz) of short duration (0.25 sec) were very similar to those previously found in the pigeon (Nye, 1969). No studies to date have examined the perceptual or physiological effects of fixational eye movements in any bird species. 

Fixational eye movements in reptiles

Fixational eye movements in the turtle

Greschner, Bongard, Rujan, and Ammermüller (2002) identified two types of fixational eye movements in the turtle (Pseudemys scripta elegans), which they called “drift-like” and “periodic.” Drift-like motions were small and slow (with velocities comparable to human drift) and were superimposed by larger and faster periodic components. Periodic motion had retinal amplitudes on the order of the diameter of a photoreceptor, which are comparable to those of human tremor. On the other hand, periodic motion's frequency (5 Hz), velocity, and orderly relationship between velocity and amplitude are closer to those of human microsaccades (see Martinez-Conde et al., 2004 for a review of tremor and microsaccade parameters in humans). 

Simulated periodic motion generated strong synchronous firing in the turtle's retina, whereas simulated drift-like motion had little effect ( Figure 3). These results agree with previous physiological and modeling studies in the primate visual system, in which strong neural transients were observed in response to microsaccades (Donner & Hemilä, 2007; Martinez-Conde, 2006; Martinez-Conde et al., 2000, 2002). Such neural transients may underlie the behavior of cortical neurons as coincidence detectors (Shelley, McLaughlin, Shapley, & Wielaard, 2002; Williams & Shapley, 2007). Moreover, neural transients to stimuli onsets and terminations (similar to those produced by microsaccades in the primate visual system; Martinez-Conde, 2006; Martinez-Conde et al., 2000, 2002) have been related to target visibility in visual masking paradigms (Macknik & Livingstone, 1998; Macknik & Martinez-Conde, 2004; Macknik, Martinez-Conde, & Haglund, 2000). 

Figure 3

 

Fixational eye movements increase retinal activity in the turtle. (A) Continuous recording from 54 retinal ganglion cells in the turtle, during darkness, after switching on a stationary grating, and during wobbling of this grating. The grating was wobbled to simulate periodic (i.e., tremor-like) eye movements. (B) Responses of a single ganglion cell to a drifting contrast border of velocity comparable to ocular drifts. (C) When the drift is superimposed with wobbling, the responses of the cell shown in (B) increase dramatically. From Greschner et al. (2002) and modified by Martinez-Conde et al. (2004).

In the turtle retina, periodic motion-driven neurons with receptive fields located along contrast borders were moreover synchronized and reliably indicated the preceding motions. The authors proposed that this synchronization of retinal activity could be used by the brain to improve the estimation of stimulus features, such as spatial frequency (Greschner et al., 2002). This finding could have a parallel in the primate visual system. Donner and Hemilä (2007) recently proposed that microsaccades significantly “re-sharpen” the image and improve spatial resolution in primates. Rucci, Iovin, Poletti, and Santini (2007) also found that fixational eye movements enhance fine spatial detail in human subjects. 

Fixational eye movements in the chameleon

Fixational eye movements in the chameleon have not been studied in detail. However, Gioanni and colleagues found that the chameleon ( Chamaleo chamaleo) makes saccades during fixation. The authors speculated that these probably correspond to fixational microsaccades found in other foveate species (Gioanni, Bennis, & Sansonetti, 1993). 

Fixational eye movements in amphibians

Fixational (and more generally, spontaneous) eye movements vary widely among amphibians. The frog (an afoveate amphibian) shows a general lack of eye and head movements, except for those that stabilize the retinal image. Dieringer and colleagues (Dieringer & Daunicht, 1986; Dieringer & Precht, 1982; Dieringer, Precht, & Blight, 1982) found that the grass frog (Rana temporaria) produces no spontaneous eye movements in the absence of head movements, with the possible exception of tremor (similarly to the rabbit, another afoveate species; Van der Steen & Collewijn, 1984). It is likely that the frog's lack of eye movements leads to fast visual fading of all stationary objects in its field of view (Ditchburn & Ginsborg, 1952; Yarbus, 1967). Lettvin, Maturana, MsCulloch, and Pitts (1968) stated that a frog “will starve to death surrounded by food if it is not moving” (but see Dieringer & Daunicht, 1986 for a counterpoint). Small head motions, or eye movements correlated with respiratory motions, may diminish fading to some extent (Dieringer & Precht, 1982; Schipperheyn, 1963, 1965, 1967). 

As a side note: the frog's respiratory eye movements play no role in the respiratory function: “ in the course of each respiratory cycle, increase of pressure in the frog's buccal cavity protrudes the eye and causes the optical axis to deviate upwards over 25–35 minutes of arc” (Schipperheyn, 1963). That is, respiratory eye movements are merely synchronized with respiration. 

Autrum (1959) found no evidence of respiratory eye movements in the frog. Ewert and Borchers (1974) observed that stationary stimuli elicit no responses from visual neurons in the common toad (Bufo Bufo L.), regardless of respiratory pressure changes in the buccal cavity. 

Fixational eye movements in the salamander

Manteuffel et al. (1977) identified two types of fixational eye movements in salamanders: tremor-like and respiratory eye movements. Tremor-like is a rapid movement of small amplitude. Torsional eye movements result from the animal's breathing (also found in the frog by Schipperheyn, 1963). Additionally, flicks are occasionally observed, which rapidly change the position of gaze. It is unclear whether such flicks may generally correspond to what we now call microsaccades and/or to voluntary saccades. Based on these observations, the authors proposed that unlike frogs and toads (Autrum, 1959; Ewert & Borchers, 1974; Lettvin et al., 1968), salamanders can continuously perceive stationary objects (Manteuffel et al., 1977). Fixational eye movements in the salamander (Tiger salamander, Ambystoma tigrinum) may also help segregate object and background motion in the retina (Olveczky et al., 2003; see also Fixational eye movements in the rabbit section) and even work to emphasize novel visual features (Olveczky et al., 2007; see Baccus et al., 2008 for details on the retinal neurons and synapses that may underlie the computation of differential motion between an object and its background). 

Fixational eye movements in fish

Spontaneous eye movements in the goldfish

Eye movements during fixation have been studied in both afoveate (goldfish: Carassius auratus) and foveate (archer fish: Toxotes chatareus) fish species. The goldfish, a long-standing model organism in vision and oculomotor research (Aksay et al., 2003, 2007; Cabrera, Torres, Pasaro, Pastor, & Delgado-Garcia, 1992; Pastor, Torres, Delgado-Garcia, & Baker, 1991; Salas, Navarro, Torres, & Delgado-Garcia, 1992; Torres, Pastor, Cabrera, Salas, & Delgado-Garcia, 1992), produces three types of spontaneous eye movements: saccades, stretch, and drift (Easter, 1971; Mensh, Aksay, Lee, Seung, & Tank, 2004). We should note that in the motionless goldfish, eye movements are very infrequent. Those that do occur consist mainly of saccades: slow drifts are very rare (Easter, Johns, & Heckenlively, 1974). During saccades (identified by thresholding the eye velocity at 5 deg/sec), both eyes move rapidly, simultaneously, and usually in the same direction (Mensh et al., 2004). Stretches occur every 2–3 min and are characterized by a brief, simultaneous large excursion of both eyes to extreme temporal positions (Easter, 1971; Mensh et al., 2004). The function of stretches is not known (Mensh et al., 2004). Because fixations were defined as those periods that were free of saccadic and stretch movements (Easter, 1971), only drifts (from the three types of spontaneous movements) can technically be considered “fixational eye movements.” However, the goldfish saccadic main sequence includes saccades with equivalent amplitudes and velocities to those found in human microsaccades (Mensh et al., 2004; Zuber & Stark, 1965). Indeed, microsaccades have been identified in other (foveate) fish species, although they are rarer than in humans (Segev et al., 2007). Small saccades in the goldfish tend to be divergent, whereas large saccades tend to be convergent (Mensh et al., 2004). Drift occurs simultaneously in both eyes, usually in the same direction. Drift velocities are almost exclusively ≤1 deg/sec (Mensh et al., 2004). No tremor has been identified in the goldfish. 

Spontaneous eye movements in teleost fish

Spontaneous eye movements have been studied in the New Zealand parore ( Girella tricuspidata; Montgomery, McVean, & McCarthy, 1983). The pattern of eye movements was similar to that previously described in goldfish (Easter, 1971; Hermann & Constantine, 1971). 

Fixational eye movements in the archer fish

A recent study in the archer fish (Segev et al., 2007) has provided the first perceptual and physiological correlates of fixational eye movements in any fish species. Archer fish have remarkable visual abilities: they shoot down insects resting on foliage—as well as stationary targets covering a single retinal photoreceptor—with squirts of water from their mouths, they predict the location at which the insect will splash down in <100 msec, and they deduce the absolute size of an object presented at different viewing distances (Supplementary Videos 2 and 3). Segev et al. (2007) identified three types of fixational eye movements in the archer fish: microsaccades, drift, and tremor. Microsaccades were rare and were not analyzed in detail. Drift velocities were ≤0.40 deg/sec. Tremor was a high-frequency oscillatory movement (the primary component was 5 Hz) with peak-to-peak amplitude of 0.1–0.2 deg (corresponding to an image displacement of 1–2 photoreceptors) and velocity of about 1 deg/sec (Figure 4). Whereas the amplitude and velocity of tremor in the archer fish is comparable to those in humans, its frequency is much lower (5 Hz vs. >40 Hz). Interestingly, archer fish tremor is very similar, both in amplitude (photoreceptor widths covered) as well as in frequency (5 Hz), to the tremor-like periodic motion in the turtle (discussed earlier; Greschner et al., 2002). 

Figure 4

 

Archer fish eye movements. (A) Example trace of eye position as a function of time, as measured by search coil. (B) Example of a single 1.5-s fixation superimposed on a grid representing the photoreceptor mosaic. The back and forth movement is caused by tremor. (C) Distribution of fixation durations (taken as time interval between saccades). (D) Power spectrum of horizontal fixational eye movements. (E) Distribution of tremor amplitude (data taken from 1 animal in a 15-min session). Modified from Segev et al. (2007).

Segev et al. (2007) trained archer fish to perform a size discrimination task: the fish had to distinguish a medium-sized target from larger and smaller distracters (they indicated their choice by shooting a jet of water). To relate this behavior to responses from retinal neurons, the authors recorded from populations of retinal ganglion cells while presenting objects of different sizes, moved to simulate saccades, drifts, and tremor. Drifts and tremor elicited weak neuronal responses that were not very informative about the spatial structure of the stimulus (although they did provide some information about target size, in agreement with the results of Greschner et al., 2002 in the turtle). Moreover, performance was similar for tremor alone and for tremor plus drift, indicating that drift plays a negligible role in revealing spatial structure (also in agreement with Greschner et al., 2002). 

When present, the firing rate was phase-locked to tremor. Saccades were more effective than drifts or tremor in driving informative responses from the ganglion cells (consistent with theoretical and physiological studies of microsaccade-driven neuronal activity in the primate retina and LGN; Donner & Hemilä, 2007; Martinez-Conde et al., 2002). Segev et al. (2007) proposed that in primates, both saccades and microsaccades (which are rare in the archer fish) may contribute to size discrimination tasks as well as enhance spatial detail, whereas drifts and tremor by themselves may not be sufficient to improve visual perception. However, it is important to remember that central vision in humans can be maintained by tremor and/or drift, even when microsaccades are suppressed during strict fixation (Martinez-Conde, 2006; Martinez-Conde et al., 2004, 2006). On the other hand, one cannot rule out the possibility that, if one could completely eliminate drifts and tremor, microsaccades might then suffice to sustain both central and peripheral vision in primates and humans. Future experiments are needed to further explore these ideas. 

Discussion

The vast majority of eye movement studies conducted in non-human species to date have not analyzed eye position data in the range of fixational eye movements. Thus eye movements during fixation have been reported in only a handful of vertebrates. Most of these studies have concentrated solely on the physical parameters of the fixational eye movements, without addressing their physiological and perceptual consequences. Fortunately, this trend has started to reverse in recent years, with a growing list of papers addressing the effects of non-human fixational eye movements in physiology and perception. This is an important development: A large amount of all psychophysical and physiological visual research has been conducted during conditions of visual fixation. Therefore understanding the precise physiological and perceptual contributions of fixational eye movements may be critical to the interpretation of previous and future vision research. 

There are some interesting differences and similarities in the types of fixational eye movements produced by different species. For instance, ocular tremor is possibly ubiquitous to all vertebrates. The fact that tremor has not been observed in all species studied could be due to the technical difficulty of recording tremor accurately (i.e., tremor amplitudes and frequencies are usually in the range of the recording system's noise). Drift has also been recorded in most vertebrates, and it may be as omnipresent as tremor. Conversely, microsaccades seem most important in species with foveal vision (i.e., primates and humans), whereas they are scarce or absent in afoveate species (i.e., rabbits, frogs, goldfish). A link between microsaccades and foveal vision may reflect the two-pronged functional significance of microsaccades in humans and primates: first, microsaccades may be necessary to correct the slow drift of the eye and bring back the object/region of interest to the retinal area where photoreceptor density is highest (Cornsweet, 1956). The corrective role of human microsaccades has been demonstrated in recent studies (Engbert & Kliegl, 2004; Engbert & Mergenthaler, 2006). Second, the dynamics of microsaccades themselves may improve spatial resolution (Donner & Hemila, 2005). 

In reviewing the patterns of fixational eye movements across vertebrates, it is tempting to draw parallels with other sensory modalities. Fixational eye movements refresh the retinal images during fixation, and thus prevent neural adaptation and visual fading. But this mechanism is not circumscribed to the visual system. Sniffs in rodent olfaction discretely sample sensory information every 200–300 msec and thus are similar in their temporal dynamics to primate saccades (Otero-Millan et al., 2008; Uchida, Kepecs, & Mainen, 2006) and microsaccades (Otero-Millan et al., 2008). A similar mode of discrete sampling may also be at play when objects are recognized through tactile information, for instance if we sweep our fingertips over an object's surface with our eyes closed, or when blind individuals read Braille script. 

One intriguing possibility that merits further experiments and analyses is the potential correlation between the spatiotemporal parameters of fixational eye movements (such as those summarized in Table 1) and the adaptation properties of visual neurons across species. For instance, the greater ocular stability of some species (rabbit, goldfish, frog, etc.) versus others (primates) could be associated with some particular adaptive properties of their visual cells. Thus future comparative studies should pay attention to the response properties of visual cells triggered by fixational eye movements in the presence of stationary stimuli. Other potential factors underlying the variation in fixational eye movements across species could be differences in oculomotor control at the level of the eye muscles, or perhaps differences in retinal organization. 

Future research will further improve our understanding of fixational eye movements and the neural activity that maintains the visibility of stationary visual scenes in human and non-human species. One interesting question that remains to be addressed concerns the role of tremor in vision and visual physiology. Because tremor frequencies are well over the flicker fusion frequencies in humans, it has been argued that tremor of the visual image would be ineffective as a stimulus (Ditchburn, 1955; Gerrits & Vendrik, 1970; Sharpe, 1972). However, recent studies have found simulated tremor to enhance visual processing and improve visual acuity in a model retina (Hennig et al., 2002; Miller et al., 2006). Follow-up research should be aimed to replicate these findings in actual neurons, with correlated perceptual reports. One should also note that early visual neurons can follow high-frequency flickering that surpasses the perceptual threshold for flicker fusion (Martinez-Conde et al., 2002). Thus high-frequency tremor could be adequate to maintain activity in the early visual system, which then may lead to visual perception in higher visual areas. Another significant question that remains unanswered concerns the oculomotor bases of the various types of fixational eye movements across species. To date, there is too little information on this important topic to draw a conclusion. 

Conclusions

We have reviewed the parameters of fixational eye movements in non-human mammals, birds, reptiles, amphibians, and fish, as well as their consequences for physiology and perception. Fixational eye movements, and microsaccades in particular, appear to be most important in foveate vs. afoveate species. Animals lacking fixational eye movements, such as afoveate frogs and toads, may be incapable of seeing stationary objects during visual fixation. Future research should further explore the perceptual effects of fixational eye movements and determine the oculomotor mechanisms responsible for their generation across species. 

Supplementary Materials

Supplementary Video 1 - Supplementary Video 1 

Supplementary Video 1. Microsaccades in the rhesus macaque during fixation of a small spot. Eye movements were recorded monocularly with a search-coil. Each cross indicates a 1-msec eye position sample. Microsaccades were identified with an automatic objective algorithm and are indicated by red crosses (Martinez-Conde, 2006; Martinez-Conde et al., 2000, 2002). The time scale has been slowed down by approximately 8 times, to help appreciate the changes in eye position during fixation. 

Supplementary Video 2 - Supplementary Video 2 

Supplementary Video 2. Archer fish shooting at an immobilized fly mounted in a plastic tube. From (Segev et al., 2007). 

Supplementary Video 3 - Supplementary Video 3 

Supplementary Video 3. Archer fish can distinguish between different target sizes. The archer fish has been trained to shoot at medium-sized visual targets (presented among larger and smaller distracters). Water jet is the faint white line starting at the fish's snout and hitting the correct (i.e. medium-sized) target. Note that while the snout sticks out of the water, the rest of the fish remains under water. From (Segev et al., 2007). 

Acknowledgments

We thank Mona Stewart and Isabel Gomez-Caraballo for technical assistance, Jorge Otero-Millan for programming assistance, and Dr. Xoana G. Troncoso for comments on the manuscript. This study was funded through grants from the Barrow Neurological Foundation (to SMC and SLM), the National Science Foundation (Award 0643306 to SMC), the Arizona Biomedical Research Commission (Award 07-102 to SMC), and Science Foundation Arizona (SLM). 

Commercial relationships: none. 

Corresponding author: Susana Martinez-Conde, Ph.D. 

Email: [e-mail]. 

Address: Barrow Neurological Institute. 350 W. Thomas Rd., Phoenix, AZ 85013, USA.