We examined what determines the typicality, or graded structure, of vocal emotion expressions. Separate groups of judges rated acted and spontaneous expressions of anger, fear, and joy with regard to their typicality and three main determinants of the graded structure of categories: category members' similarity to the central tendency of their category (CT); category members' frequency of instantiation, i.e., how often they are encountered as category members (FI); and category members' similarity to ideals associated with the goals served by its category, i.e., suitability to express particular emotions. Partial correlations and multiple regression analysis revealed that similarity to ideals, rather than CT or FI, explained most variance in judged typicality. Results thus suggest that vocal emotion expressions constitute ideal-based goal-derived categories, rather than taxonomic categories based on CT and FI. This could explain how prototypical expressions can be acoustically distinct and highly recognisable but occur relatively rarely in everyday speech.

When adults talk to infants, they dramatically change the prosodic and acoustic structure of speech. Recently, new insights have been gained on those changes, especially on the vocalic and temporal structure of speech which are described as being more variable than in adult conversations. In the present contribution, we examine formant and fundamental frequency characteristics of different infant-directed registers, notably infant-directed speech and singing, the latter not being investigated so far. We present data from 14 German-speaking mothers singing a playsong and reading a story to their 6 months old infants, or to the experimenter. Infant-and infantabsent versions of speech and song were compared on the formant characteristics of the primary vowel triangle (/i, a, u/) and on general fundamental frequency changes. Our results show that vowel space did not differ in infant-and infantabsent versions of speech and song. However, vowel dispersion, i.e., formant variability, was higher in both infantdirected song and speech than in infant-absent versions. Consistent with previous findings, f0 was higher in infantthan infant-absent versions of speech and song, with song showing generally higher f0. These results are discussed in light of current approaches to the variability of infant-directed registers, and their attractiveness to infants.