Although numerous studies have demonstrated that prosodic cues (eg pitch, intensity, rhythm, temporal relationships and stress) are critical to human speech perception, most automatic speech recognition systems used in artificial intelligence research process only phonetic cues. This work seeks to demonstrate the benefits to such systems of using multiple complementary sources, and includes details of extensive performance evaluation that shows an improvement of almost threefold in a phonetic word hypothesizer when used with several prosodic knowledge sources running in parallel. First, several novel algorithms which extract prosodic parameters reliably are introduced. The work then implements and evaluates prosodic knowledge sources that apply the extracted parameters at appropriate processing levels including the lexical, syntactic and sentential levels. To permit large vocabulary capability, the knowledge source designs emphasize a concern for minimizing lexical search, exploiting parallelism and speaker-independent or template-independent operation. The word is aimed at a wide computer science readership; no specific knowledge of linguistics or speech science is presumed.