Metastability and Eye Movements: A Dynamical‑Systems Interpretation

 

1. Why metastability is a natural language for gaze behavior

The idea that perception, attention, and neural activity evolve through transiently stable states has deep roots in cognitive science and neuroscience. Concepts such as attractors, basins, noise‑induced transitions, and escape times appear in work on perceptual switching, decision‑making, and neural population dynamics (Kelso, 1995; Rabinovich et al., 2008; Deco & Jirsa, 2012).
Eye movements, especially the alternation between fixations and saccades, fit remarkably well into this metastable picture. The Freidlin–Wentzell theory of rare events (Freidlin & Wentzell, 2012) provides a rigorous mathematical language for these intuitions.

2. Fixations as metastable states 

Consider the gaze position $X_t$ as the state of a stochastic dynamical system. During a fixation the gaze remains confined to a small region, microsaccades and noise generate small fluctuations, and the gaze tends to return toward a local center.

This is exactly the behavior of a trajectory near a stable attractor (Engbert & Kliegl, 2003; Martinez‑Conde et al., 2004). 

The key property is time‑scale separation: fast local fluctuations vs. rare transitions. This is the hallmark of metastability (Bovier & den Hollander, 2015).

3. Saccades as escape events

A saccade corresponds to a rapid transition:

$$\text{Fixation A}⟶\text{Fixation B}\mathrm{.}$$

In metastability, this is a noise‑induced escape from one basin of attraction followed by rapid relaxation into another. The most probable escape route is the minimum‑action path predicted by Freidlin–Wentzell theory (Freidlin & Wentzell, 2012; E & Vanden‑Eijnden, 2010).

This analogy captures the abruptness of saccades, the stereotyped nature of saccadic trajectories (Bahill et al., 1975), and the dependence of fixation duration on attentional factors (Henderson, 2003).

4. The quasipotential landscape of attention

Attention can be understood through the concept of an effective landscape, denoted as V(x). In this landscape, salient objects are represented as potential wells, while fixation points correspond to local minima. The depth of these wells reflects attentional priorities.

In this framework, deeper wells are associated with longer fixations, whereas shallower wells lead to shorter fixations. Additionally, saccades can be viewed as movements that allow individuals to escape from these wells.

Freidlin–Wentzell theory predicts:

$$E[T_{\text{fix}}]\propto \exp \text{\hspace{0.17em}⁣}\left(\frac{\mathrm{\Delta }V}{\epsilon }\right),$$

the same structure governing chemical reactions (Kramers, 1940), neural transitions (Deco & Jirsa, 2012), and metastable stochastic systems (Bovier & den Hollander, 2015).

This connects eye‑movement behavior with a broad class of escape‑time phenomena.

5. Why this interpretation is psychologically meaningful

Fixation is not a mechanical pause. It reflects visual saliency (Itti & Koch, 2001),  task demands (Yarbus, 1967), memory and cognitive goals (Henderson, 2003), and competition among candidate targets (Findlay & Walker, 1999).

A fixation is therefore a temporary winner in a dynamic competition. As evidence accumulates or priorities shift, the current attractor destabilizes and the system transitions to a new one. This is precisely how metastability is used to describe neural dynamics (Rabinovich et al., 2008).

This perspective aligns with stochastic cognitive models in which latent states evolve in a metastable landscape and observable behavior reflects transitions between hidden attractors. In particular, the two‑layered diffusion model (Piu, Fargnoli, Rufa, 2014) provides a concrete example of how cognitive processes can be represented as layered stochastic dynamics with noise‑driven transitions between decision alternatives.

6. A dynamical‑systems model of gaze

A simple stochastic model is:

$$d{X}_t=-\mathrm{\nabla }U(X_t,t)dt+\sqrt{\epsilon }d{W}_t,$$

where $U$ is an attentional potential. In this model, minima correspond to fixation locations, while saddles indicate the boundaries between competing fixation regions. Saccades occur when crossing these saddles. The most probable saccadic path is identified as the minimum-action path, as noted by Weinan & Vanden-Eijnden in 2010.

This model provides a principled abstraction capturing essential features of gaze dynamics.

7. Beyond spatial attractors: the hidden attentional state

The true state may be:

$$Z_t=(x_t,y_t,a_t),$$

where $a_t$ is a latent attentional variable (task set, object identity, semantic relevance). Then, we can consider fixations as metastable regions in a higher‑dimensional cognitive space, and saccades as projections of transitions in that hidden space.

This aligns with biological models of attention as a multi‑dimensional dynamical process (Rabinovich et al., 2008; Deco & Jirsa, 2012) and with layered stochastic models of cognitive dynamics such as the two‑layer diffusion framework (Piu, Fargnoli, Rufa, 2014).

When examining oscillatory structures, we can draw several associations. Fixation periods can be likened to metastable plateaus, while saccades can be compared to escape events. Additionally, low-frequency oscillations are related to the modulation of barrier heights, and the timing of saccades influences the probabilities of escape in a slowly varying quasipotential.

This parallels periodically forced metastable systems (Berglund & Gentz, 2006), in which oscillations modulate the likelihood of escape rather than directly causing transitions.

This gives a principled interpretation of oscillatory components in psychological or eye‑movement time series.

8. A rigorous caveat: when is metastability justified?

One should not claim that “a fixation is a metastable state” without empirical support. The defensible statement is:

A fixation can be modeled as a metastable state if gaze dynamics exhibit clear separation between fast fluctuations and rare transitions.

To support this assertion, several empirical criteria should be met:

1. Gaze remains close to fixation centers for extended periods (as shown by Engbert & Kliegl, 2003).

2. Transitions occur quickly in comparison to the durations of fixations (according to Bahill et al., 1975).

3. The durations of fixations exhibit characteristics of escape-time statistics (as outlined by Unema et al., 2005).

4. Transition paths tend to cluster around preferred routes.

If these conditions are satisfied, it becomes mathematically appropriate to apply the concept of metastability.

9. Conclusion

Under appropriate empirical conditions:

Gaze fixations behave like metastable states, and saccades behave like escape events from their basins of attraction. Fixation duration corresponds to a metastable residence time, and attentional saliency defines an effective quasipotential landscape.

This is not merely metaphorical. It is a mathematically rigorous way to model eye movements, grounded in stochastic dynamical systems, supported by cognitive theory, and compatible with modern analyses of oscillatory and multiscale psychological data.

References 

1. Bahill AT, Clark MR, Stark L. The trajectory of saccadic eye movements. Sci Am. 1975;233(4):108–17.
2. Engbert R, Kliegl R. Microsaccades uncover the orientation of covert attention. Vision Res. 2003;43(9):1035–45.
3. Martinez-Conde S, Macknik SL, Hubel DH. The role of fixational eye movements in visual perception. Nat Rev Neurosci. 2004;5(3):229–40.
4. Findlay JM, Walker R. A model of saccade generation based on parallel processing and competitive inhibition. Behav Brain Sci. 1999;22(4):661–722.
5. Henderson JM. Human gaze control during real-world scene perception. Trends Cogn Sci. 2003;7(11):498–504.
6. Itti L, Koch C. Computational modelling of visual attention. Nat Rev Neurosci. 2001;2(3):194–203.
7. Yarbus AL. Eye Movements and Vision. New York: Plenum Press; 1967.
8. Unema PJ, Pannasch S, Joos M, Velichkovsky BM. Time course of information processing during scene perception. Vis Cogn. 2005;12(3):473–94.
9. Kelso JAS. Dynamic Patterns. Cambridge (MA): MIT Press; 1995.
10. Rabinovich M, Huerta R, Laurent G. Transient dynamics and metastability in neural systems. Science. 2008;321(5885):48–50.
11. Deco G, Jirsa VK. Ongoing cortical activity at rest: criticality, multistability, and ghost attractors. J Neurosci. 2012;32(10):3366–75.
12. Freidlin M, Wentzell A. Random Perturbations of Dynamical Systems. 3rd ed. Springer; 2012.
13. Bovier A, den Hollander F. Metastability: A Potential-Theoretic Approach. Springer; 2015.
14. E Weinan, Vanden-Eijnden E. Transition-path theory and path-finding algorithms. Annu Rev Phys Chem. 2010;61:391–420.
15. Kramers HA. Brownian motion in a field of force. Physica. 1940;7(4):284–304.
16. Berglund N, Gentz B. Noise-Induced Phenomena in Slow-Fast Dynamical Systems. Springer; 2006.
17. Itti L, Koch C, Niebur E. A model of saliency-based visual attention. IEEE TPAMI. 1998;20(11):1254–9.
18. Tatler BW, Hayhoe MM, Land MF, Ballard DH. Eye guidance in natural vision. J Vis. 2011;11(5):5.
19. Haken H. Synergetics. Springer; 1983.
20. Friston K. The free-energy principle. Nat Rev Neurosci. 2010;11(2):127–38.
21. Piu P, Fargnoli F, Innocenti A, Rufa A. A two-layered diffusion model traces the dynamics of information processing in decision making. Comput Intell Neurosci. 2014;2014:383790.