Subjective responses related to touch are described via perceptual and affective adjectives. Understanding relationships among such responses is useful for industrial surface design. We developed a method for investigating relationships among touch-related responses by constructing multilayered adjective models without assumptions regarding model structure such as adjective hierarchy or a definition of layers. The method consists of two processes through which a model structure and parameters are estimated. To acquire the model structure, subjectively evaluated causal relationships among adjective pairs are used. Free parameters, such as effects among adjectives, are then statistically estimated through sensory evaluation and structural equation modeling. To validate the method, two models with different levels of details were developed using 29 adjective pairs and 46 texture samples. For example, the model with three layers, which we categorized into psychophysical, affective, and hedonic layers, was constructed. Moreover, a detailed model with four layers was provided, which was more complex than previously reported models. The present method helps advance our understanding and design of connections among subjective surface ratings.