We report on an experiment aimed to test the hypothesis that listeners orient to canonical forms when judging the tempo of reduced speech. Orientation to canonical forms should yield higher tempo estimates than orientation to surface phone strings when canonical phones are deleted. We tested the hypothesis for English, capitalizing on the fact that the non-realization of schwa in an unstressed syllable (e.g. support) may result in a surface phone string associated with a different word than the intended one (sport). We presented listeners with sentences containing ambiguous surface realizations, along with orthographic representations which convinced some that they were listening to disyllabic words (support etc.) and others that they were listening to monosyllabic ones (sport etc.). Asking listeners to judge the tempo of the sentences allowed us to assess whether the difference in imposed lexical interpretation had an impact on perceived tempo. Our results reveal the predicted effect of the imposed interpretation: sentences with a ‘disyllabic’ interpretation for the ambiguous word form were judged faster than (the same) sentences with a ‘monosyllabic’ interpretation.