Is a new feature learned behind a newly efficient color-orientation conjunction search?

It is well known that feature search is efficient, whereas conjunction search is usually inefficient. However, prior studies have shown that some conjunction search could become very efficient through perceptual learning, behaving like a traditional feature search. An unanswered question is whether a new feature is learned when an inefficient conjunction search become efficient after extensive training. A popular view is that the trained conjunction has been successfully unitized into a new feature and thus could pop out from neighboring distractors. Here, by using stimulus specificity and transfer of perceptual learning as an approach, we investigate whether a new feature is learned when an initially inefficient conjunction search becomes highly efficient after extensive training. In two experiments, we consistently found that long-term perceptual learning over days could induce an inefficient-to-efficient pattern change in a color-orientation conjunction search. Moreover, the learning effect of the conjunction target could partly transfer to a new target that shared a same color or a same orientation with the trained target. Remarkably, the total amount of the learning effect was approximately equal to the sum of the transfer effects of individual features. Such additive learning pattern could last for at least several months, although the learning of separate features showed different patterns of persistence. These results do not support that the trained conjunction is unitized into a new and inseparable feature after learning. Instead, our findings point to a feature-based attention enhancement mechanism underlying long-term perceptual learning and its persistence of color-orientation conjunction search.