Relationship between color appearance evaluation and categorical color responses of small stimuli in central and near peripheral visual fields

The appearance of color stimuli can be measured psychophysically using two major techniques: (1) elementary color naming and (2) categorical color naming. On the relation between the two naming techniques, a network model (published 20 years ago) autonomously labels colors without retrieving color names from a database and has the flexibility to adapt to individual differences or observing environments. However, this network model has not been recently applied in this context as few works have focused on two types of color naming experiments using the same observers and conditions, mainly because most studies focused either on continuous changes in color appearance or on categorical color perception and did not need to employ both techniques. In our previous study, new datasets of hue and saturation judgments with whiteness and blackness evaluations were obtained. The evaluation of these datasets included elementary color naming and categorical color naming using 11 basic color terms (BCTs) with 0.5°-diameter stimuli presented at the center and 12 locations along the horizontal and vertical meridians in the visual field for three observers. The feasibility of applying the model to our new datasets in the center and near periphery has to be examined before utilizing the model for various applications. Gain factor values for each categorical color response (CCR) were individually optimized, and the same combination of gain factors were used for all locations for each observer. The results of the chi-square goodness-of-fit test indicated that the response distribution estimated by the model (color names and “undefined”) was not significantly different from the experimental data obtained in 35 out of 39 conditions. The overall average of the correct estimation was 77%. The high estimation obtained from the same gain factor values is indicative of an invariant relationship between the two naming techniques, with up to 20° eccentricity. It also shows the model's plausibility in explaining individual differences among normal observers.