Periodic Reporting for period 4 - INDIVISUAL (Individual differences in human gaze behaviour and the visual system)
Reporting period: 2024-07-01 to 2024-12-31
Our ability to see clearly and resolve visual clutter is much higher at the so-called fovea - the center of gaze - than in the periphery. Therefore, human vision is a process of selection – we constantly move our eyes and place the fovea on different parts of the scene in front of us. This poses a challenge to the visual system. At each moment in time, it needs to decide where to move the fovea next, often based on poor peripheral input. But how does it achieve this feat? How can we select the most relevant parts of a scene before we look at them directly? For a long time, vision scientist modelled this process as a function of low-level image properties. For instance, these models may predict the next eye movement to go to the part of the scene with the highest color contrast. However, more recent work has shown that eye movements are much better predicted by semantic properties of a scene, like the location of objects, faces or text. Furthermore, eye movements differ systematically between observers – they are not just a function of the scene in front of us, but also the person looking at it. INDIVISUAL aims to better understand how eye movements differ from one person to the next, what the sources and consequences of these differences are for individual perception and whether we can use individual eye movements as a diagnostic marker. We also aim to use individual differences as a tool to better understand the general mechanisms allowing the human visual system to efficiently select relevant parts of a scene for eye movements.
After five adventurous years, INDIVISUAL came to several conclusions. Systematic individual differences in gaze predict individual differences in perception. Gaze traits can predict how two people will describe a scene and even how the representation of a movie will differ in their brains. Moreover, how we look at faces can predict how well we do at recognising them and distinguish between 'super recognisers' and controls. Our differences in this regard seem to go back to basic features of the visual system. They are not specific to face, but extend to inanimate objects: Someone who avoids looking at the eye region will also tend to look lower in objects. Individual differences in the visual brain are at least partially heritable (identical twins have more similar visual brains than non-identical ones), but where we move our eyes is also shaped by development. A massive dataset we colected in a museum allowed us to trace the development of eye movements across thousands of children and adults and revealed that gaze behaviour takes surprisingly long to mature, with changes lasting well into teenage and beyond.
After five adventurous years, INDIVISUAL came to several conclusions. Systematic individual differences in gaze predict individual differences in perception. Gaze traits can predict how two people will describe a scene and even how the representation of a movie will differ in their brains. Moreover, how we look at faces can predict how well we do at recognising them and distinguish between 'super recognisers' and controls. Our differences in this regard seem to go back to basic features of the visual system. They are not specific to face, but extend to inanimate objects: Someone who avoids looking at the eye region will also tend to look lower in objects. Individual differences in the visual brain are at least partially heritable (identical twins have more similar visual brains than non-identical ones), but where we move our eyes is also shaped by development. A massive dataset we colected in a museum allowed us to trace the development of eye movements across thousands of children and adults and revealed that gaze behaviour takes surprisingly long to mature, with changes lasting well into teenage and beyond.
For our experiments we use dedicated ‘eyetracking’ cameras which can track eye movements with up to 1,000 images per second. Observers typically see naturalistic images or movies of complex scenes. Testing hundreds of volunteers in the lab, we found highly systematic individual differences in the way we look at people, with opposing individual tendencies to either look at the face or body. Furthermore, the individual way we look at faces is predictive of our abilities to recognise them - so called ‘Super Recognisers’ tend to look just below the eyes, a region predicted to be most informative by computational models. Additional brain imaging experiments show that individual eye movements lead to more idiosyncratic representations of visual events in the parts of the visual brain processing objects and people. These results show that individual gaze biases generalise between images and movies and have real-world consequences for perception outside the lab.
We further found that pre-literate children look much less at text than adults and have a strong tendency to focus on hands. This qualifies previous results on the heritability of individual gaze biases – massive visual experience, like learning to read, can permanently modulate individual gaze. At the same time, we found that the strong propensity of faces to attract fast saccades does not extend to artificial features we have much experience with, like glasses or masks, but rather seems driven by the eye region. This suggests the mechanism steering eyes towards faces uses simple pictorial cues and its plasticity is rather limited.
Finally, we developed a ‘quick test’ of individual gaze and implemented it in a clinical setting, as well as in an ‘eyetracking booth’ in a public museum. So far, over 2,000 museum visitors, as well as children with ADHD and autism volunteered to take the test.
We further found that pre-literate children look much less at text than adults and have a strong tendency to focus on hands. This qualifies previous results on the heritability of individual gaze biases – massive visual experience, like learning to read, can permanently modulate individual gaze. At the same time, we found that the strong propensity of faces to attract fast saccades does not extend to artificial features we have much experience with, like glasses or masks, but rather seems driven by the eye region. This suggests the mechanism steering eyes towards faces uses simple pictorial cues and its plasticity is rather limited.
Finally, we developed a ‘quick test’ of individual gaze and implemented it in a clinical setting, as well as in an ‘eyetracking booth’ in a public museum. So far, over 2,000 museum visitors, as well as children with ADHD and autism volunteered to take the test.
The museum data establish the largest benchmark sample of individual gaze variation to date, using a practical, low-demand test. We aim to collect a total of 10,000 datasets and to use these data to probe the diagnostic potential of our test. We further plan to test the relationship between individual gaze and image understanding in more detail. Finally, we will probe whether individual gaze biases can be traced to the functional anatomy of the individual visual brain.