Agglutinating structures are often predicted to be easier to learn than fusional structures due to compositional transparency. In a series of artificial language learning experiments, we test the importance of transparency and compositional structure as well as offline decomposition for the learnability of fusional and agglutinating structures of words expressing two grammatical features.
Fusional morphemes typically express more than one meaning and therefore exhibit less compositional transparency than agglutinating morphemes which can be combined to words in a more transparent way. It is often argued that this transparency renders agglutinating structures more learnable (Brown 1973, DeKeyser 2005, Dressler 2012, Goldschneider & DeKeyser 2001, Hengeveld and Leufkens 2018, Narasinham & Gullberg 2011, Penke 2012). However, the benefits of agglutinating compared to fusional structures can only be exploited when the learner initially segments words into each of their morphemes. While it has been found that L1 speakers decompose words that appear morphologically complex around 170 ms after stimulus presentation, a time before lexical access becomes available (Rastle, Davis, New 2004, Zweig & Pylkkänen 2009), the case of L2 speakers is less clear. However, various studies on morphological processing have suggested that L2 speakers rely more on declarative memory than L1 speakers even after ample L2 experience (Clahsen et. al. 2010, Gor 2010, Kimppa et al 2019). In a series of artificial language learning experiments, we investigate whether there is a difference in learnability between fusional and agglutinating structures of nouns expressing the grammatical features of animacy and number. While learners may not decompose words in either condition in real time, they may still find the agglutinating condition easier to learn due to offline decompositionality. Results from 280 native speakers of English suggest that there is no difference in learnability overall when the inflectional system is relatively small and that offline decomposition is successful even without color cues indicating morpheme boundaries. This suggests that learners decompose offline, but that transparency does not boost the learnability of agglutinating structures under all circumstances. Instead, transparency may only be advantageous when the inflectional system is larger or it may correlate with other factors in inflectional systems.
