In this study we determined whether untrained raters' assessments of fluency in low-proficiency second language speech were related to temporal measures and whether they varied across tasks. We collected speech samples from 20 beginner Mandarin learners of English on picture description, monologue, and dialogue tasks. Temporal measures were made on each sample. Twenty-eight untrained judges rated fluency, comprehensibility, and accentedness. Three trained raters also judged samples for "goodness of prosody." The rating data paralleled the speech measurements: speakers performance on the monologue and dialogue tasks was significantly better than on the narratives; however, listeners judgments of goodness of prosody did not vary across tasks. Comprehensibility and fluency ratings were highly correlated; fluency was more strongly related to comprehensibility than to accentedness.