Repetition and reading tasks are commonly used to evaluate primary progressive aphasia (PPA). We hypothesized that the ratio of reading to repetition errors can distinguish PPA variants, and due to floor and ceiling effects, ratios of errors with short sentences with common words distinguish some individuals, while ratios of errors with lengthier sentences with longer and less common words better distinguish others. A total of 146 individuals completed the two (reading and repetition) simple sentence tasks and 19 completed the two new sentence tasks. There were no significant differences between PPA variants in age or education (by ANOVA; Table 1), or sex (chi squared). Using multinomial regression, the simple plus new ratio of reading:repetition errors explained more of the variance between PPA variants (pseudo R2 = 0.32; p=0.03; n=15), than either the simple sentence ratio (pseudo R2=0.03, p=0.02, n=146) or the new ratio (R2=0.22; p=0.01; n=19). Both svPPA and lvPPA patients made significantly more total errors on the simple repetition task than the simple reading task, but the mean difference was greater for lvPPA (p<0.00001 vs. p=0.01). Only the lvPPA patients made significantly more total errors on the new repetition than the new reading task (p<0.00001, vs. p>0.1 for the other variants). The ratio of reading:repetition errors in the new sentence task better discriminates lvPPA from the other variants than the same ratio in the simple task, but they provide complementary information. Either pair of tasks discriminates between lvPPA and nfavPPA, which is generally the hardest diagnostic distinction to make (Tippett, 2020). However, the combination of short and long sentences improves classification and distinguishes lvPPA from svPPA. SvPPA participants actually made more errors on reading than repetition in the new task. Further investigation may determine if word frequency, word length, or sentence length effects account for the differences.