No escape from morphemes in morphological processing

Any approach to understanding morphological processing must begin with assumptions about the role of morphemes in linguistic representations. Contemporary linguistic theory proposes that such representations are centred on a syntactic organisation of morphemes, where a morpheme is an abstract syntactic unit that finds an interpretation in form and in meaning. From the linguist's perspective, then, morphological processing falls together with syntactic processing, and both involve the central exploitation of a grammar of morphemes. Recent computational work has challenged this perspective, proposing instead that morphological structure emerges from the acquisition of form/meaning connections. Here we show that these challenges to the role of morphemes in word and sentence processing rest on a misunderstanding of the morpheme within linguistic theory. Once we understand the notion of a morpheme and its explanatory role within linguistic theory, we can see that apparent challenges to the morpheme, rather than dispensing with the morpheme in the architecture of linguistic representations, always in fact presuppose a syntactic structure of morphemes as the unlearned input to language learning models. Since computational modelling is indispensable for work connecting theories of linguistic representation to theories of (morphological) processing, it is crucial that those constructing computational models appreciate why there is no escape from morphemes in morphological processing.