It is often assumed that when we attend to one part of an object, such as the handle of a mug, our attention spreads involuntarily to all other parts and features of it, even if irrelevant (e.g. the body of the mug, and the logo inscribed on it its colour). This spreading is assumed, in popular theories of object-based attention to be involuntary and facilitatory, improving our perception of parts and features belonging to an attended object.

Much experimental evidence supports this:  ignored visual stimuli can be easier to process, or more distracting, when perceptually grouped with an attended target stimulus. But in many studies, the ignored stimuli share potentially relevant features or locations with the target, so spreading of attention could be strategically beneficial. In this fMRI study, published in Frontiers in Integrative Neuroscience, we measured the brain’s response to entirely task-irrelevant events, inside and outside the focus of spatial attention, and belonging to objects that were either grouped together or separate.

Two checkerboards were displayed each in opposite hemifields, while participants responded to check-size changes in one pre-cued hemifield, which varied between blocks. Grouping (or segmentation) between hemifields was manipulated between blocks, using common (versus distinct) motion cues (see movies above). Task-irrelevant transient events were introduced by randomly changing the colour of either checkerboard, attended or ignored, at unpredictable intervals.

The above assumptions predict automatic and facilitatory spreading of BOLD signals across hemifields from attended to ignored stimuli, when these are bound by grouping (see below).

We actually found more rather than less modulation of the overall response associated with attending or ignoring one moving checkerboard, for the grouped stimulus. This effect was significant across most early visual areas receiving stimulation.

BOLD response to transient events (red flashes) in attended and ignored hemifields should also be predicted to be more similar under grouping, if they automatically receive facilitatory attentional resources. We actually found the opposite pattern, in primary visual cortex. For ungrouped stimuli, BOLD signals associated with task-irrelevant changes were lower, not higher, in the attended versus ignored hemifield; furthermore, attentional modulation was not reduced but actually inverted under grouping, with higher signals for events in the attended versus ignored hemifield.

These results challenge the assumptions underlying object-based attention. We consider a broader biased-competition framework: task-irrelevant stimuli are suppressed according to how strongly they compete with task-relevant stimuli, with intensified competition when the irrelevant features or locations comprise the same object.

Below is a gravitational analogy of a ‘biased competition’ interpretation of our results, incorporating the above assumption. Top illustrations: balance of competition between stimuli, depicted as if each is a weight suspended on a toy mobile. Green and red balls represent relevant targets, and irrelevant transient flashes, respectively. Left and right pairs of red and green balls represent stimuli in attended and unattended hemifields, respectively. In the grouped condition (a), we assume intensified competition between hemifields. Bias is applied to the attended hemifield (depicted as movement in upwards direction), including irrelevant transients, to help segment the stimulus and offset the increased distraction from the irrelevant hemifield. Without grouping (b) competition between hemifields is reduced, and bias is applied selectively to relevant targets, offsetting competition from the transients. (c) The model predictions are re-drawn to match the format of the for block-related results above, under the two grouping conditions; (d) predictions for the transients, rendered for comparison with the event-related results above.

This assumption of intensified competition within groups can explain some classic findings, illustrated below. The two-object cost (Duncan, 1984) and invalid-cueing effects within objects (Egly et al, 1994) may occur because attention can move more efficiently within group by biasing competition.

Intensified competition within objects could also cause greater distraction from flankers when grouped with target (Baylis & Driver, 1992).