This paper explores the nature of residual literacy and cognitive deficits in self‐reported dyslexic Norwegian adults. The performance of 26 self‐reported dyslexic adults was compared with that of a comparison group of 47 adults with no history of reading or spelling difficulties. Participants completed standardized and experimental measures tapping literacy skills, working memory, phonological awareness and rapid naming. Spelling problems were the most prominent marker of dyslexia in adults, followed by text reading fluency and nonword decoding. Working memory and phoneme awareness explained unique variance in spelling, whereas rapid automatized naming explained unique variance in reading fluency and nonword reading. The moderate to strong correlations between self‐reported history, self‐rating of current literacy skills and outcomes on literacy tests indicate that adults estimated their literacy skills fairly well. Results suggest that spelling impairments, more strongly than reading impairments, make adults perceive themselves as being dyslexic. A combination of three literacy and three cognitive tests predicted group membership with 90.4% accuracy. It appears that weaknesses in phoneme awareness, rapid automatized naming and working memory are strong and persistent correlates of literacy problems even in adults learning a relatively transparent orthography. Copyright © 2014 John Wiley & Sons, Ltd.