The relationships between computational texture features and human touch feeling were investigated. Twenty four tactile plaques were manufactured. The textures for the plaques were copies of visual textures in which the grey scale was converted to height, or were copies of existing tactile textures. Textural features of the plaques' topographies were extracted using the most common statistical analysis techniques used in machine vision. Human touch affect was measured by asking 107 participants to touch and rate the plaques against a set of adjectives in a psychological experiment. Partial least squares regression and wrapper methodology were used to predict a set of dependent variables (20 human touch feeling features) and to select a significant subset of variables from a large set of independent variables (115 computational features). The results identify a subset of features that appear to have the most important effect on human touch feeling. These results will be used to synthesize plaques with the required human touch feeling features.