Purpose: The goal of this study was to determine whether the results obtained from a 25-utterance conversational language sample were as reliable as those obtained from a 50-utterance sample. Method: Robust conversational language samples from 220 children with typically developing language (106 boys, 114 girls) ranging in age from 3;2 to 7;10 (years;months) were collected. The language samples were randomly assigned to one of two conditions: a 25-utterance condition and a 50-utterance condition. Transcripts were examined for three metrics, including mean length of utterance[subscript SUGAR], words per sentence, and clauses per sentence. Results: Data were analyzed using two methods. A linear mixed-model analysis was used to assess absolute and relative reliability, and the Bland-Altman procedure was used to assess absolute reliability and clinical acceptability. Results of the mixed-model analysis indicated that mean length of utterance[subscript SUGAR] and words per sentence demonstrated relative reliability; however, none of the metrics demonstrated absolute reliability. In contrast, results of the Bland-Altman scatter plots indicated that all three metrics demonstrated absolute reliability because 94%-96% of participants' scores fell within the limits of agreement. Taken together, the results suggested that the statistically significant differences indicated by the mixed-model analysis were not clinically significant. Conclusion: These results highlighted the importance of using different methods of analysis in studies of reliability. The findings indicated that reliable language sample results can be obtained from 25-utterance samples. Furthermore, by including practices already in use (e.g., collecting samples [less than or equal to] 50 utterances) and including only minimal changes to current practices, the methods used in this study are feasible for school-based clinicians, could be easily integrated into clinical practice, and could increase the use of evidence-based assessment practices in schools.