A research paradigm is suggested that combines the perspectives of the humanistic scholar and the behavioral scientist: After differentiating the popularity of actual aesthetic products using archival indices and then subjecting these compositions to objective computer content analyses, further statistical treatment may divulge the intrinsic properties responsible for differences in impact. This approach is illustrated by an analysis of the 154 sonnets attributed to William Shakespeare. Each sonnet was partitioned into four consecutive units (three quatrains and a couplet), and then a computer gauged how the number of words, different words, unique words, primary process imagery, and secondary process imagery changed within each sonnet. Taking advantage of a previous objective measure of the relative aesthetic merit of the sonnets, and implementing a statistical search for interaction effects, it was demonstrated that Shakespeare's lexical choices adopt a discernible pattern in the highly popular creations that is not found in the more obscure poems. Perhaps the most fascinating aspect of this pattern shift is the distinct manner in which the poet modifies his vocabulary when composing the concluding couplet in his best sonnets.