title: Fixational Eye Movements
status: established
last_updated: 2026-05-31
sources: Rolfs 2009 Microsaccades, Martinez Conde 2004 Fixational Eye Movements
tags: [fixational-eye-movements, microsaccades, tremor, drift, attention, visual-perception, oculomotor]

Fixational Eye Movements¶

Status: established
Last updated: 2026-05-31
Sources: Rolfs 2009 Microsaccades, Martinez Conde 2004 Fixational Eye Movements
Tags: [fixational-eye-movements, microsaccades, tremor, drift, attention, visual-perception, oculomotor]

Summary¶

During attempted visual fixation, the eyes produce three classes of small involuntary movements: microsaccades, drift, and tremor. These fixational eye movements prevent visual fading caused by neural adaptation and, in the case of microsaccades, carry information about the locus of covert visual attention. Their functional significance makes them relevant to both basic vision research and applied contexts where gaze data is collected.

Body¶

Context¶

This article draws on two reviews of the small involuntary movements that occur during attempted fixation: Martinez-Conde et al.'s (2004) Nature Reviews Neuroscience review of their perceptual role, and Rolfs's (2009) Vision Research minireview of microsaccades and their link to attention. The first treats all three movement classes — tremor, drift, microsaccades — and their function in counteracting neural adaptation; the second concentrates on microsaccades, their generation in the oculomotor system, and their relation to covert attention. Within this knowledge base the article supplies the basic-vision account of what happens during a "fixation," qualifying the event-detection abstraction used in Fixation Saccade Detection (where fixational movements are treated as noise to be removed) and complementing the cognitive-load strand in Pupil Dilation Cognitive Load.

Key Points¶

On classification, Martinez-Conde et al. (2004) characterise tremor as an aperiodic, wave-like, high-frequency oscillation of about 90 Hz — the smallest of the three movements and possibly a by-product of the oculomotor system's noise — and drift as a slow, continuous wandering (mean speeds of a few minutes of arc per second, up to roughly 0.5°/s) occurring between microsaccades during the great majority of fixation time (PDF pp. 3, 5, orig. pp. 231, 233). Rolfs (2009) characterises microsaccades as fast, jerk-like movements occurring once or twice per second during fixation, conventionally bounded by an upper amplitude of about 1° (some labs use 2°) (PDF pp. 1, 4, orig. pp. 2415, 2418).

On perceptual function, Martinez-Conde et al. (2004) centre the role of fixational movements on preventing visual fading: when retinal-image motion is eliminated, perception fades within seconds through neural adaptation (Troxler's effect), and microsaccades counteract this by increasing the spike probability of neurons in the lateral geniculate nucleus and area V1, producing transient bursts of firing, while drift and tremor sustain stimulation between microsaccades (PDF pp. 1, 3, 9, orig. pp. 229, 231, 237). Rolfs (2009) adds that microsaccades also help correct accumulated fixation drift, keeping the image near the centre of the fovea, and may contribute to the coding of spatial detail (PDF p. 5, orig. p. 2419).

On attention, Rolfs (2009) reports that microsaccades index covert attention through a characteristic rate signature: after an attentional cue the microsaccade rate drops to a minimum at about 150 ms, then rebounds to a maximum around 350 ms before resettling, and during the rebound the movements are biased first opposite the cue and then toward the attended target (PDF pp. 17–18, orig. pp. 2431–2432). Martinez-Conde et al. (2004) review convergent evidence linking microsaccade direction to attended locations (PDF pp. 9–10, orig. pp. 237–238). Together this makes the microsaccade rate signature and direction a potential non-invasive index of attention even when gaze is nominally fixed.

On neural substrate, Rolfs (2009) notes that microsaccades share the main-sequence relationship — a distinct correlation of peak velocity with amplitude — with larger saccades, suggesting a common generator, and identifies the superior colliculus (SC) and brainstem saccade-related burst neurons as the primary circuitry, the same structures that control regular saccades (PDF pp. 5, 7, orig. pp. 2419, 2421).

For applied eye tracking, the implication of both reviews is that fixational movements, usually treated as noise and removed by event detection (see Fixation Saccade Detection), are not neutral: microsaccade rate and direction carry attentional information that most downstream analyses discard. Reliable measurement, however, depends on the high-resolution, high-sampling-rate recording used in the laboratory studies the reviews summarise; the reviews do not specify the equipment thresholds at which microsaccades become undetectable in remote or head-mounted recordings [synthesis across Martinez-Conde et al. (2004) and Rolfs (2009)].

Conclusion¶

The two reviews are complementary rather than competing. Martinez-Conde et al. (2004) establish that the three fixational movements together keep the visual image from fading, treating microsaccades primarily through their perceptual function; Rolfs (2009) extends the account to show that microsaccades also index covert attention and are generated by the same SC-and-brainstem circuitry as ordinary saccades. They agree on the anti-fading role and on the main-sequence link to saccades; the principal extension is Rolfs's attentional reading, which gives the movements applied value as a possible real-time attention probe where sampling resolution permits.

References¶

Martinez-Conde, S., Macknik, S. L. & Hubel, D. H. (2004) 'The role of fixational eye movements in visual perception', Nature Reviews Neuroscience, 5(3), pp. 229–240. doi: 10.1038/nrn1348. martinezConde2004fixational

Rolfs, M. (2009) 'Microsaccades: Small steps on a long way', Vision Research, 49(20), pp. 2415–2441. doi: 10.1016/j.visres.2009.08.010. rolfs2009microsaccades

Open Questions¶

Can microsaccade direction be used as a real-time attention probe in control room monitoring tasks?
At what eye tracker sampling rate and spatial resolution do microsaccades become reliably detectable in practice? Neither source states quantitative detection thresholds; an earlier draft's figures (> 500 Hz, < 0.05°) were not in the sources and have been removed.
Do microsaccade patterns differ between high-workload and low-workload monitoring states in operational settings?
Earlier drafts contained several figures not verifiable in the sources: tremor amplitude "< 1 arcmin" (the term "arcmin" does not appear in Martinez-Conde et al., 2004), and microsaccade "peak velocity 10–100°/s" (Rolfs, 2009, gives no such range). The attention timing was also imprecise — the rate signature is an inhibition reaching a minimum at ~150 ms and a rebound peaking at ~350 ms, not a "shift within 200–400 ms" with "suppression around 200 ms." All three were corrected against the sources.