The present study examines the effect of interlocutor type – second language (L2) learner versus heritage speaker – on interaction and phonetic accuracy during the completion of a dyadic map task in Spanish. Twenty intermediate-level, native English-speaking learners of Spanish and 10 heritage speakers of Spanish participated. Data collection occurred over two sessions; at each session, participants completed a dyadic map task on which pronunciation had been made task-essential via street names that constituted vocalic minimal pairs (e.g., Calle Poca vs. Calle Pica). At one session, L2 learners interacted with another L2 learner; at the other, they interacted with a heritage speaker, with the order of interlocutor interaction conditions counterbalanced across learner participants. Dyads were audio-recorded during the completion of the task, and the recordings were analyzed in two ways: (a) with regard to the number and type of language-related episodes (LREs; Swain & Lapkin, 1998, 2001) present in the interaction, with a focus on pronunciation-related LREs, and (b) with regard to the phonetic accuracy of learner Spanish vowel productions as measured by first and second formant frequencies. Results suggest little change in LRE production by interlocutor type but significantly less targetlike realizations of /o/ and /u/ during interaction with a heritage speaker than with another L2 learner. The findings are discussed in light of previous research on interlocutor effects on interaction and linguistic outcomes and with special attention given to how and why pronunciation as a target feature may differ from other foci.