Recent research has shown that characteristics of a talker’s voice are preserved in memory representations of spoken words. These findings suggest that aspects of surface form that co‐occur frequently with a word’s phonological form and meaning may influence lexical processing. To determine if surface form is linked to word meaning, this study examined the production of emotional homophones including one emotional (happy or sad) and one neutral meaning, each with distinct spellings (e.g., bridal/bridle). It was hypothesized that each meaning of these homophones would be linked to different characteristic surface forms, with emotional prosody linked to the emotional meaning and no systematic prosody linked to the neutral meaning. Participants read aloud either the emotional or the neutral spelling of each homophone and their utterances were recorded and acoustically analyzed. For happy homophones, average fundamental frequency and rms amplitude were higher when participants read aloud the happy spelling than when they read the neutral spelling of the same homophone. For sad homophones, shimmer and jitter measures were greater when reading the sad then the neutral spelling. These results suggest that emotional prosody is linked to emotional meaning in spoken words and this link appears to guide speech production.