PURPOSE: The purpose of this study was to establish a profile of the spelling patterns studied in students from public and private schools and to describe a word spelling task for clinical and educational settings. METHODS: Eighty-two fourth grade students belonging to the elementary school of public and private schools in São Paulo, ranging in age from nine to ten years, took part in this study. The spelling task consisted of a list of ten high frequency words (HFW), ten low frequency words (LFW), and ten pseudowords (PW), in which the typology and number of spelling errors were described. To compare the average number of mistakes on the HFWs, LFWs, and PWs, we used an analysis of variance and Tukey's multiple comparisons (p<0.05). Using a cluster analysis, homogeneous groups were formed based on their performance. RESULTS: Results indicated that the average number of mistakes in the LFWs was higher than in the HFWs (p=0.000) and PWs (p=0.000), and the number of mistakes in the HFWs was lower than in the PWs (p=0.009). The highest number of mistakes was found in the following categories: "others", "rule generalization", "omission", "voiced-voiceless", and "addition". There were no mistakes of the type "ão-am" and "blend-separation" in the HFWs. CONCLUSION: Spelling errors are a part of the process of learning to write, and students can show some variance in spelling performance. Furthermore, students need to be stimulated to analyze words and their aspects of phonology, morphology, and semantics. An analysis from the types of errors is not enough to plan intervention programs, but instead is necessary to understand the strategies that the child uses to write.