The aim of the present study was to evaluate processes of lexical access, selection and early semantic access in young native Hungarian students as well as in dyslexics compensating successfully for their reading problems of developmental origin. The present study made use of the well-known lexical decision paradigm in which event-related potentials (ERPs) elicited by words, number-words and pseudowords were measured. Subjects had to judge whether the letter strings seen were meaningful or meaningless. Our results suggest that in good readers additional activity occurs in the sensory or selection stage of lexical access when words of low sight frequency, e.g. number-words are read. Significant processing differences for words vs. number-words were found in the later stage of processing. Based on our ERP data we do not suggest number-words for judging general features of lexical processing, especially when developmental dyslexia is the focus of study. Our results show that young adults may develop a particular compensation strategy for reading words of different frequency. We found that: (1) Lexical access is fast and accurate in good readers and the early components elicited by words and number-words do not differ. (2) Attentional effort is reflected by enhanced early components to number-words. (3) Dyslexics may compensate for the weakness of sight word vocabulary, characteristic for frequent words as well, during lexical selection and at a later stage of processing. (4) Dyslexic adults, who compensate well for reading difficulties, differ significantly in this later stage when words have to be read. (5) The late positive component of ERPs reflects additional activation allocated to word reading when low frequency words such as number-words are read. Good readers show this effect as well, therefore, the largest difference found between dyslexics and controls is found for frequent words. (6) The early semantic access is absent in dyslexics when pseudowords are read and this process may be one of the strategies used by dyslexics in a transparent orthography.