Mind in a Bind: Treisman Shows How We Bind Images and Why We Often Don’t

If during Anne Treisman’s William James Award address your attention was on the rusty Powerpoint projector, or the lady in the front wearing a lavender belt around the waist of a green suit, or the precarious waver of an overhead light fixture, that’s all right with her.

In fact, you may help prove her point, because on stage, APS Fellow Treisman was busy illustrating the object binding problem – why human perception sometimes has difficulty simultaneously reconciling every detail of a vast scene.

“At any moment of time the scene around us is filled with objects differing along many dimensions, which we see from particular angles, and which may themselves move and transform,” Treisman said. “I’ve been concerned with how we manage these tasks that seem so effortless, yet actually depend on sophisticated perceptual mechanisms.”

Sensory information arrives in parallel, as a heterogeneous melange of shapes, colors, motions, smells, and sounds, encoded in partly modular systems distributed through the brain. To represent the world in useful form we must collect the features, bind them into the correct spatial and temporal bundles, and then interpret them to specify their real world origins.

For example, the idiosyncrasies of a strawberry – its oblong shape, red color, green tuft, and freckled seeds – are combined to form an integrated perceptual object. If attention is diverted or overloaded, we may bind features incorrectly, as is often shown when subjects are asked to focus on black digits flanking three colored letters. The outcome, as predicted, is the confusion of a colored number, or a black letter; the correct bindings may have been lost.

Accoring to Treisman, our seemingly intrinsic ability to decipher, say, a red square from a flock of blue circles actually relies on a series of hints, a collection of data that requires the human mind to create a system of resource folders, somewhat akin to files stored on a computer.

“I suggest that we set up temporary representations that we call object files, initiated simply by the presence of something at a particular place and time,” Treisman explained. “If the object moves or changes, the object file is updated and its earlier state erased.”

In patients suffering from Balint’s syndrome, an illness due to bilateral parietal lesions, this process can go very wrong. These patients have difficulty with visual space, and the size of their attention window – the capaciousness of which is crucial to successful binding – is usually quite limited. “[The patient studied] was unable to see more than one object at a time,” Treisman said, “although he could identify the one seen object perfectly well.”

When shown a window with multiple elements of different colors, the subject responded that only one color had been present. However, when these same elements were joined by lines linking pairs of different colors, the subject was suddenly able to detect the presence of both colors, considering the pair as one singular unit rather than two separate but connected ones.

Though separating triangles from squares and blue from green seems, for most people, like an automatic response rather than an intricate process of cutting-and-pasting, Treisman upholds the procedure’s underlying complexity, explaining that it is not just Balint’s patients who have trouble piecing together the puzzle. “When focused attention is prevented, binding sometimes fails, giving rise to illusory conjunctions,” Treisman said. “Thus, normal people have the same problem as Balint’s patients if we give them brief exposures and overload attention.”

After an object’s presence and features are discerned, the brain undergoes a triumvirate of tasks to complete the binding cycle: shifting the attention window around the object, suppressing extraneous features to prevent illusory conjunction, and binding the selected features together. “Thus,” Treisman declared aphoristically, “we must shift, suppress, and bind.”

Different areas of the brain are active at different times during the process. The spatial aspect of the attention scan is overseen by the occipital, parietal, and frontal cortexes. The binding process seems to involve two separate areas: the medial frontal area, activated normally during moments of conflict, and the left anterior insula, an area participating normally in a broad range of tasks, from pain perception to numeric counting. “I wish I could say that obviously what they all have in common is binding,” Treisman said of the inscrutably versatile left anterior insula. “But I won’t.”

Yet, while the specifics are still obtuse, the general process is well understood, and even quite practical. “It would be wasteful and probably beyond [human] attention capacity to specify each element in a representation we form,” she said. “Instead, we need to quickly characterize the ensemble, facilitating rapid decisions based, for example, on the quality of fruit, the density of traffic, or the rockiness of the terrain. … We need to study both artificially-controlled and natural stimuli if we are to understand the complexity of our apparently effortless and beautifully efficient visual mechanisms.”