One of the major issues in linguistics is uncovering how complex morphological systems are learned. The learner must make associations between form and meaning, while fitting each form into a complex system. Languages that mark multiple grammatical features (e.g., both number and gender in a single morpheme) pose a learning problem; the more features marked with a single morpheme, the more morphemes that need to be learned. Consistency within the semantics and the phonological features may reduce the learning load by providing additional cues to learning. Phonological consistencies are common within morphological systems; semantically related words often share some similar phonological feature (e.g., German definite articles like /der/, /dem/ and /das/ all start with /d/). Syncretism often follows a logical organization in which semantically related morphemes tend to be merged. The present study provides support for the hypothesis that phonological consistencies and semantic consistencies in syncretism aid in learning morphological systems from two artificial language learning experiments. Participants exposed to a novel language in which both gender (masculine, feminine and neuter) and number (singular, dual and plural) were marked, were more accurate in learning the morphological system when provided with consistent phonological cues (Experiment 1) and consistent semantic cues following syncretism of the nine grammatical markers into three morphological forms (Experiment 2). These results support the hypothesis that languages support structures that are maximally learnable, and that learning biases shape the frequency of structures and patterns found across languages of the world.