Three eye-tracking experiments investigated how phonological reductions (e.g., "puter" for "computer") modulate phonological competition. Participants listened to sentences extracted from a spontaneous speech corpus and saw four printed words: a target (e.g., "computer"), a competitor similar to the canonical form (e.g., "companion"), one similar to the reduced form (e.g., "pupil"), and an unrelated distractor. In Experiment 1, we presented canonical and reduced forms in a syllabic and in a sentence context. Listeners directed their attention to a similar degree to both competitors independent of the target's spoken form. In Experiment 2, we excluded reduced forms and presented canonical forms only. In such a listening situation, participants showed a clear preference for the "canonical form" competitor. In Experiment 3, we presented canonical forms intermixed with reduced forms in a sentence context and replicated the competition pattern of Experiment 1. These data suggest that listeners penalize acoustic mismatches less strongly when listening to reduced speech than when listening to fully articulated speech. We conclude that flexibility to adjust to speech-intrinsic factors is a key feature of the spoken word recognition system. (Contains 1 footnote, 2 tables, and 6 figures.)