Vowels can be produced as long, isolated, and steady-state, but that is not how they are found in natural speech. Instead natural speech consists of almost continuously changing (i.e., dynamic) acoustic forms from which mature listeners recover underlying phonetic form. Some theories suggest that children need steady-state information to recognize vowels (and so learn vowel systems), even though that information is sparse in natural speech. The current study examined whether young children can recover vowel targets from dynamic forms, or whether they need steady-state information. Vowel recognition was measured for adults and children (3, 5, and 7 years) for natural productions of /dæd/, /dUd/ /æ/, /U/ edited to make six stimulus sets: three dynamic (whole syllables; syllables with middle 50-percent replaced by cough; syllables with all but the first and last three pitch periods replaced by cough), and three steady-state (natural, isolated vowels; reiterated pitch periods from those vowels; reiterated pitch periods from the syllables). Adults scored nearly perfectly on all but first/last three pitch period stimuli. Children performed nearly perfectly only when the entire syllable was heard, and performed similarly (near 80%) for all other stimuli. Consequently, children need dynamic forms to perceive vowels; steady-state forms are not preferred.