Feature binding across different visual dimensions

The human brain represents different kinds of visual feature dimensions in different ways. For example, surface features exhibit some properties that are very different from contour features, and some feature dimensions may be represented more extensively in either the dorsal or the ventral visual stream. Given such differences, we investigated feature binding across different feature dimensions and whether some feature dimensions might be more easily bound together than others. In Experiment 1, we looked at cross-dimension bindings for all combinations of color, orientation, and shape dimensions, while at the same time controlling for feature discriminability. Rates of correct binding, illusory conjunctions, and feature errors were equivalent in all cases. There was no bias so that some feature dimensions were easier to combine than others. In Experiment 2, we manipulated the difficulty of feature discrimination for the key, target-defining feature and the report feature. Rates of binding errors increased with difficulty of the key feature, but not with that of the report feature. The accuracy of feature discrimination could be dissociated from the accuracy of binding the feature to an object. Across both experiments, the accuracy of feature binding was independent of specific feature dimensions or perceptibility. These findings are discussed in relation to both feature integration and multiple-stage accounts of visual feature integration.