Purpose: To determine whether measures of stuttering frequency and measures of overall stuttering severity in preschoolers differ when made from audio-only recordings compared with audiovisual recordings. Method: Four blinded speech-language pathologists who had extensive experience with preschoolers who stutter measured stuttering frequency and rated overall severity from audio-only and audiovisual recordings of 36 preschool children who were stuttering. Stuttering frequency (percentage of syllables stuttered [%SS]) was based on counts of perceptually unambiguous stutterings, made in real time, and overall severity was measured using a 9-point rating scale. Results: Stuttering frequency was statistically significantly lower by around 20% when made from audio-only recordings. This was found to be directly attributable to differences in the counts of stuttered syllables, rather than to differences in the total numbers of syllables spoken. No significant differences were found between recording modalities for the ratings of overall severity. Correlations between %SS scores in the 2 modalities and severity rating scores in the 2 modalities were high, indicating that observers agreed on data trends across speech samples. Conclusions: Measures of %SS made from audio-only recordings may underestimate stuttering frequency in preschoolers. Although audio-only %SS measures may underestimate stuttering frequency at the start of a clinical trial to a clinically significant extent, posttreatment scores at or below 1.0%SS are likely to underestimate by 0.2%SS or less, which is clinically insignificant. (Contains 1 figure.)