Abstract
Conveying emotions in spoken poetry may be based on a poem's semantic content and/or on emotional prosody, i.e., on acoustic features above single speech sounds. However, hypotheses of more direct sound-emotion relations in poetry, such as those based on the frequency of occurrence of certain phonemes, have not withstood empirical (re)testing. Therefore, we investigated sound-emotion associations based on prosodic features as a potential alternative route for the, at least partially, non-semantic expression and perception of emotions in poetry. We first conducted a pre-study designed to validate relevant parameters of joy- and sadness-supporting prosody in the recitation, i.e. acoustic production, of poetry. The parameters obtained thereof guided the experimental modification of recordings of German joyful and sad poems such that for each poem, three prosodic variants were constructed: one with a joy-supporting prosody, one with a sadness-supporting prosody, and a neutral variant. In the subsequent experiment, native German speakers and participants with no command of German rated the joyfulness and sadness of these three variants. This design allowed us to investigate the role of emotional prosody, operationalized in terms of sound-emotion parameters, both in combination with and dissociated from semantic access to the emotional content of the poems. The findings from our pre-study showed that the emotional content of poems (based on pre-classifications into joyful and sad) indeed predicted the prosodic features pitch and articulation rate. The subsequent perception experiment revealed that cues provided by joyful and sad prosody specifically affect non-German-speaking listeners' emotion ratings of the poems. Thus, the present investigation lends support to the hypothesis of prosody-based iconic relations between perceived emotion and sound qualia. At the same time, our findings also highlight that semantic access substantially decreases the role of cross-language sound-emotion associations and indicate that non-German-speaking participants may also use phonetic and prosodic cues other than the ones that were targeted and manipulated here.