BACKGROUND: Working memory. AIM: To verify the performance of working memory abilities and their relation with the severity of phonological disorders. METHOD: 45 children, with ages between 5.0 and 7.11 years, with evolutional phonological disorders (EFD), 17 female and 18 male, were evaluated. All subjects were assessed using the Child Phonological Evaluation proposed by Yavas et al. (1991). The severity of the disorder was determined by the Percentage of Correct Consonants (PCC) proposed by Shriberg and Kwiatkowski (1982), classifying the phonological disorder as severe, moderate-severe, average-moderate and average. After that, subtest 5 of the Psycholinguistic Abilities Test (ITPA--Bogossian & Santos, 1977) and the non-word repetition test (Kessler, 1997) were applied. RESULTS: After analyzing the data according the statistical tests of Kruskal Wallis and Duncan, it was verified that the performance of moderate-severe and severe individuals in the non-word repetition test was inferior to that of average-moderate and average individuals. However, performance results in the digit repetition test did not present a positive correlation with severity. CONCLUSION: The performance of phonological memory has a relation with the severity of phonological disorders. This allows us to accept the idea that the phonologic memory is related to speech production. Regarding the central executor, the results indicate that the performance in digit repetition, used to assess the central executor, did not present a correlation with the severity of the disorder. This can be justified by the fact that the central executor is more directly related to vocabulary acquisition and is responsible for processing and storing information.