In order to learn a language, a listener must extract a complex array of regularities that range from the phonemes forming the words to the syntactic rules organizing sentences. Extensive research has shown that from an early age humans are endowed with the capacity to learn distributional and abstract patterns after brief exposure to linguistic input (Marcus et al. 1999; Saffran et al. 1996). The apparent effortless manner in which statistical and rule learning seems to operate suggests these processes occur in an attention-independent manner. We have demonstrated that attention might in fact modulate them. We showed in a series of experiments that if attention is diverted away from a speech stream, participants cannot segment out the statistically-defined words from it (Toro, Sinnett & Soto-Faraco, 2005). Even more, recent experiments suggest learning of abstract rules might be affected by diverting attention depending on the structures of the rules to be learned. That is, participants readily learn simple rules containing adjacent repetitions even if attention is diverted away from them, but are not able to do so if such rules are defined over non-adjacent repetitions (Toro, Sinnett & Soto-Faraco, 2011). Further experiments will target how is attention directed towards relevant information to be extracted from the linguistic input.