This study investigates whether words differing in a single contrastive semantic feature (positive/negative) can be discriminated by auditory evoked responses (AERs). Ten right-handed college students were provided with auditory stimuli consisting of 20 relational words (more/less; high/low, etc.) spoken with a middle American accent and computer modified to match peak intensity level and duration. A microcomputer simultaneously presented the recorded words and provided visual stimuli consisting of pictures of eight configurations of two rectangles, shown with the larger or taller rectangle on the right and left sides of the picture an equal number of times. Scalp electrodes monitored brain and eye activity. Analysis of the results showed the subjects reliably discriminating between words differing in the semantic feature in question, supporting the semantic feature hypothesis that suggests words may be decomposed into more basic elements. It appears that both hemispheres in the parietal region are sensitive to the kinds of features exemplified by these relational word pairs, and that the AER procedure could be quite useful in studying the semantic component of language. (MSE)