Recent work has shown that young children can use fine phonetic detail during the recognition of isolated and sentence-final words from early in lexical development. The present study investigates 24-month-olds' word recognition in sentence-medial position in two experiments using an Intermodal Preferential Looking paradigm. In Experiment 1, French toddlers detect word-final voicing mispronunciations (e.g., buz [byz] for bus [bys] “bus”), and they compensate for native voicing assimilations (e.g., buz devant toi [buzdəvɑ̃twa] “bus in front of you”) in the middle of sentences. Similarly, English toddlers detect word-final voicing mispronunciations (e.g., sheeb for sheep) in Experiment 2, but they do not compensate for illicit voicing assimilations (e.g., sheeb there). Thus, French and English 24-month-olds can take into account fine phonetic detail even if words are presented in the middle of sentences, and French toddlers show language-specific compensation abilities for pronunciation variation caused by native voicing assimilation.