Coherence masking protection (CMP) refers to the phenomenon in which lower thresholds for accurate vowel labeling are obtained when an informative F1 is presented with a constant, and therefore non‐informative, F2/F3 cosignal than when F1 is alone. Tarr and Nittrouer reported this effect for adults and showed enhanced effects for children with synthetic speech [J. Acoust. Soc. Am. 126, S2300 (2009)]. With sine waves, no CMP was found for either group. Two new experiments asked what accounts for these differences across age and stimulus type. Hypotheses focused on the principle of common harmonic structure and expectations of speech‐like signals. Experiment 1 combined sine‐wave F1 with three synthetic cosignals, one that was harmonically related to F1 and two that were not. Experiment 2 used all synthetic speech, but F1 and the cosignal differed in harmonic structure. All signals were speech‐like. Stimuli were presented in white noise for labeling, with and without the cosignal. Labeling thresholds were measured adaptively for F1 only and all three formants. Adults showed no significant CMP in either condition, but children did in both. It was concluded that children perceptually organize auditory signals based on the expectation of speech and do not require harmonicity. [NIDCD grant DC‐00633.]