11.9 Conclusion

Just as we would never write a prompt such as, "Please to say now you PIN," which is ungrammatical syntactically, we should never produce recordings that are ungrammatical prosodically. Recorded messages are prosodically ungrammatical when, for example, they exhibit inappropriate or alien intonation contours, inappropriate or missing pauses, stress and rhythmic patterns not found in human-to-human discourse, and physiologically impossible sound sequences. To avoid these problems, design choices need to reflect the systematic and context-sensitive structure of language at the level of prosody.

Every spoken natural language has its own highly structured grammar of prosody, and so does every language variety, such as African American Vernacular English (popularly known as Ebonics), Moroccan Arabic, and Sicilian. Prosody is an essential component of the way humans structure their intent and mediate interpretations in context. There is no reason to believe that people suspend prosodic assumptions and expectations when interacting with a speech application. After all, prosody is an integral aspect of spoken language, as essential to our system of verbal communication as syntax or semantics. When you record and implement messages, you should respect familiar, natural patterns of stress, intonation, and phrasing, mirroring authentic language use as much as the limits of technology allow.