The effects of orthographic legality, contour integrity and spatial complexity on Chinese-character crowding

Abstract

Visual crowding is a ubiquitous experience for almost everyone, for example, when you are reading this text. Objects within foveal vision could be recognized without effort. Those falling outside foveal vision appear cluttered if they are close to one another, but their presence remains detectable. This phenomenon is called crowding, commonly considered as a bottleneck of object recognition. In this study, Chinese characters were used to examine both low-level and high-level factors that could affect crowding. Chinese character is a special class of visual objects with two outstanding characteristics: First, they are not only defined by the composition of constituting stroke patterns but also their configuration. Second, they are also highly varied in spatial complexity. The current study tested the effects of orthographic legality, contour integrity and spatial complexity on the crowding of Chinese characters. All experimental conditions intended to induce crowding involved a trigram of Chinese characters, with the target surrounded by two flankers aligned horizontally on the horizontal meridian. Experimental results showed that the orthographic legality of flankers, defined by the configuration of stroke patterns, had no effect on crowding strength; contour integrity of flankers, manipulated by perturbing the phase spectrum of images, modulated crowding strength. Target-flanker similarity in spatial complexity surprisingly only modulated the crowding of low-complexity as well as mildly-complex targets but not high-complexity targets. These results were all obtained from luminance contrast threshold elevation method, which measured the elevation of luminance contrast required for target identification due to crowding. As an attempt to further examine the effects of spatial complexity, confusion matrix was proposed and tested as a new method. It focuses on the change of confusability, on top of recognizability, as a result of crowding. This might be more relevant to the study of crowded percepts. The method was tested on Latin letters, and it demonstrated that the baseline confusability of a letter (i.e. when it is uncrowded) was disrupted by crowding. Specifically, a confusion matrix would show more randomness or higher dispersion in errors. It was able to differentiate conditions where same-category flankers (i.e. letters) and different-category flankers (i.e. numerals) were presented, even when they induced similar contrast threshold elevations. However, it was not found very promising in the study of the complexity effects on Chinese-character crowding. Overall, Chinese-character crowding is not likely to involve object-level representations but mainly the character contours. Target-flanker similarity, particularly with regards to spatial complexity, is not as robust as previously thought; it depends on the target complexity and the form similarity of characters. Target characteristics are likely to affect such theorized integrative mechanisms as pooling and substitution occurring between the target and flankers, resulting in crowding, and therefore modulate the target-flanker similarity effect in complexity. Other high-level factors such as attention could also partly explain the crowding of high-complexity characters.