This dataset is a large-scale collection of human perceptual ratings for material appearance, designed to study how reflectance, surface geometry, and illumination jointly affect material perception. It contains more than 215,680 responses covering 42,120 distinct combinations of material, shape, and lighting conditions, with ratings for appearance attributes such as glossiness, contrast and sharpness of reflections, metallicness, and lightness. Its purpose is both analytical and practical: to examine how geometry and illumination influence perceived material properties across diverse appearances, and to train a deep learning model that predicts perceptual material attributes from 2D images for applications including appearance reproduction, BRDF editing, illumination design, and material recommendation.