Phrase level characteristics of F0 and their interactions with duration patterns for a set of contrastively emphasized digits in elicited dialogues are examined. Phrases consisting of three digits plus a street name were elicited in a dialogue format designed to have the speaker repeat the same correction on one of the digits up to five or six times. Perception tests determined which phrases where ``best perceived'' and ``worst perceived'' as to emphasis. Onset, offset, and peak F0 within the sonorant portion of each digit of those phrases ``best perceived,'' ``worst perceived,'' and the reference phrases, with no corrective emphasis, were measured. Results suggest that speakers produce well-perceived contrastive emphasis by lengthening the duration of the emphasized word, shortening the duration of the other words within the phrase, and producing an extensive pitch drop between the emphasized word and the following words in the phrase. In an elicited dialogue situation in which the speaker is forced to produce repetitively the same item with contrastive emphasis, different combinations of these two cues are used, presumably in an attempt to maximize the chance that emphasis will be perceived.