The present study examined the persistency of phonological deficiencies over time. The participants were 40 pupils in grade 2 with documented reading and writing difficulties and a comparison group of 30 pupils. The participants were followed over a 10-year period by word- and non-word-reading tests and tests of cognitive ability. The persistence of phonological deficits was indicated by a high correlation between non-word-reading tests in grades 3 and 12 in the reading-disabled group. A dyslexia cut-off definition based on phonological ability was the most consistent definition over time compared to a word-decoding definition or multiple cut-off definition based on IQ. Phonological decoding abilities were remarkably stable over time, and non-word-reading was found to be a valid instrument in diagnosing and discerning dyslexia both in children and adults. (Contains 2 tables and 4 figures.)