Crossing the boundary: Categorization of morphed figures

Abstract

In our daily life, we are surrounded by objects. To know how to act upon them we need to identify them, a mechanism that is called categorization. Object categorization operates in a discrete manner. As Harnad once wrote “something is a 100% bird, or something is a 100% not-bird”. However, how will we deal with perceptually uncertain objects, or in other words, objects that are not easily categorized? Morphed figures are typical examples of perceptual uncertainty, resulting from interpolating one object into another object with small changes. Morphed figures are predominantly perceived as their dominant object. For example, when a turtle is changed into a car, the morphed figure consisting of 60% turtle and 40% car is preferably categorized as a turtle. The preference of categorizing a morphed figure as its dominant object is referred to as categorical perception. Categorical perception suggests that the nondominant object (car) is not perceived despite the fact that a morphed figure contains information from both objects (turtle and car). In the current thesis, it was investigated whether the nondominant object is processed as well to provide more insights in the categorization process of visual objects in general. First, it was found that the nondominant object indeed was processed. When asked to, observers were able to recognize the nondominant object in a morphed figure. In addition, the categorization process of morphed figures was slowed down by a prime word that was similar to the nondominant object. Second, the question was addressed to what level the nondominant object was processed. The findings indicate that the dominant object was processed up to a perceptual and semantic level, whereas the nondominant object was only processed up to a perceptual level. This suggests a process in which the dominant and nondominant object are both considered as possible response options, but only the dominant object will reach the final response stage that is includes semantic knowledge. Third, it was found that the perceptual similarity between a morphed figure and its nondominant object increased with increase of processing time of the morphed figure. Based on this finding, it was suggested that in a later stage of the categorization process the nondominant object became more strongly activated. At last, it was investigated whether the processing of the nondominant object was reflected in the gaze behaviour. A strong relation was observed: if a morphed figure was interpreted as a turtle, observers preferably looked at the location of the head of the turtle. If the same morphed figure was interpreted as a car, however, observers preferably looked at the location of the driver’s window. Hence the uncertainty observed in the categorization process was not directly observed in the gaze behaviour. Although categorization of objects seems to be a discrete process (something is a member of a category or not), the results from this research suggest that consideration and competition precedes the final interpretation of the visual object.