This study aims to examine whether there is behavioral evidence for the linguistic and acquisition relevance of the Subregular Hypothesis (Heinz, 2010), which predicts that patterns with specific subregular computational properties are much more readily learned. Specifically, we are testing whether linguistic constraints that are activated in laboratory situations are directly "channeled" into incremental, real-time phonological predictive processing. We compare the learnability of two phonotactic patterns that differ computationally and typologically. The study uses the oddball paradigm incorporated within an artificial grammar learning experiment as a design parameter and Signal Detection Theory (SDT) to measure sensitivity to the signal. The results reported in this paper support the Subregular Hypothesis that learnability of phonotactic patterns depend heavily on specific subregular computational properties. The merit of this paper is that it replicates a previous finding by using a very different methodology and demonstrates that artificial language learning happens at or is implemented at the level of the phonological parser, in a way consistent with implicit learning of linguistic information.