Three separate studies were successively carried out to investigate the usefulness of intensively training children with dyslexia with daily exercizes based on the temporal processing theory of dyslexia, according to which these children would be specifically unable to process brief and rapidly changing auditory stimuli. The speech modification, similar for the three studies, was close to that proposed by Merzenich et al. (1996) and Tallal et al. (1996), including both artificial slowing of natural speech stimuli and amplification of brief, unstable portions of the speech signal. In the first study, 12 children, aged 10-12 years, received either such modified speech or normal speech for 1 h a day, 5 days a week, over 5 weeks, and they were assessed on phonological tasks before, during and after training. A significant advantage for the modified speech group was found both in pre-post-training improvement and in day-to-day progression on phonological performance. In a second study, 29 children with dyslexia, aged 5-12, received a similar training, but for only 15 min a day, 7 days a week, over 6 weeks, part at the speech therapist office, part in their own homes. The finding of comparable improvement in a more 'natural' environment and in children over a wider age range indicates both the efficacy and feasibility of the method in usual clinical practice. However, this study also showed that one of four children did not improve as expected, prompting a third study where 23 other children underwent specific tasks presumably exploring various aspects of temporal processing in order to find predictors of training efficacy. A 'temporal order judgement' (TOJ) task was found best correlated with post-training improvement, suggesting that one use this task for selecting the best candidates for temporo-phonological training. Moreover, such correlation provided further argument in favour of the temporal deficit theory of dyslexia, not only by showing a link between a purely temporal task and ultimate phonological performance, but also by demonstrating that TOJ performance itself improves after phonological training. Finally, and taken together, these studies provide further justification for a rational, indication-based temporo-phonological treatment of dyslexia. Possible neural substrates of the relevant mechanisms are discussed in the light of recent experimental and brain-imaging studies.