In this thesis we used an artificial language learning paradigm to investigate the learning and generalization of different types of rules based on linear and hierarchical syntactic dependencies. We also explored which factors (i.e., type of units, phonological, prosodic and semantic cues) influenced the learning. In all experiments participants were presented with grammatical sentences and their ability to extract syntactic regularities was studied during test phases. In the first two experiments, the rule was based on the linear position of the elements in the sentence. In the other experiments (from experiment 3 to 6), the rule was based on the position of the elements in the hierarchical structure of the sentence. While experiments 1, 2, and 3 did not revealed learning, the subsequent experiments showed that the presence of specific phonological and prosodic cues (e.g., syllables number, pauses) facilitates rule learning. Interestingly, also semantic cues seem to play a role when high level processing was required to generalize the knowledge of the rule to novel syntactic structures.