Systematic languages are easier to learn: Evidence from artificial language learning (A paper written in 2011)

This study examines whether and how the shape of the linguistic input influences language learning and regularization in individual adult learners. To test this question, we used a set of miniature artificial languages with different degrees of structural systematicity, from fully random to fully systematic, and two types of frequency distributions, skewed and uniform. The results from 80 adult native speakers of English show that systematicity facilitates learning: the mean transmission error decreases as the input systematicity increases. Frequency distributions play an important role in the systematicity of the output language, but, contrary to some previous findings, this factor alone does not drive regularization. Learners increased the systematicity of the learned language only when they were exposed to a high degree of regularity in the input. Systematicity is, therefore, critically determined by the patterns in the language input, and not by some independent pressure to regularize.