This study reports on listener perception of contrastive emphasis in natural language utterances varied for emotional modality. Length and F0 correlates of well-perceived corrected digits indicate that length but not a L+H* rising contour cues listeners that the digit is emphasized. Length of vowels poorly perceived for contrastive emphasis but agreed by listeners to be emotional are compared with well-perceived corrected digits. Utterances which listeners agree are emotional are shown to have a high misperception rate for the corrected digit, and those with low listener agreement for emotion have a low misperception rate for the corrected digit.