On the Relationship between Prosodic Features and the Lexical Content of Speech
Henry I. Soron · The Journal of the Acoustical Society of America · 1964
The test material for this experiment consisted of 40 simple sentences. The words in 10 of these expressed fear; 10 others expressed anger; 10 connoted joy; while the remainder were neutral. These were read 4 times by two speakers, who attempted to convey one of these emotions by his manner of speaking, even though the lexical content and the emotion might not correspond. The speakers' voices were recorded simultaneously by a high-quality condenser microphone and by a throat microphone whose output was filtered to lie between 50 and 200 cps, thus eliminating intelligibility while preserving most of the prosodic features. These sentences, both high-quality and filtered, were presented to groups of male and female listeners, who were asked to give forced-choice impressions. Results show that, for stimuli containing both text and prosodic features, the listeners responded to the emotions conveyed by the prosodic features 62.3%, 84.2%, 59.5%, and 88.5% of the time for fear, anger, joy, and neutrality, respectively. The responses to emotions conveyed by the words were 36.3%, 42.0%, 39.9%, and 41.6%. For filtered sentences containing only prosodic features, correct identification was made 77.6%, 79.7%, 73.7%, and 93.3%, respectively.