A long-standing question in (psycho)linguistic theory concerns the nature of morphemes: by definition, the fundamental building blocks of larger linguistic objects (words, phrases, etc.). One basic question in this domain concerns how morphemes relate form and meaning; and, beyond this, whether morphemes involve information of a further type, what can be called a purely morphological representation. In this talk, David Embick will review experiments from our behavioral laboratory that use different types of priming paradigms to explore this question. It is well-established that priming can diagnose relatedness between words, both semantic (e.g. dog primes cat) or phonological (e.g. rink primes pink). To probe for purely morphological representations, it is therefore necessary to ensure that semantic and phonological relatedness between prime and target are controlled for. He will review the results from experiments that do this in different ways, and discuss their implications for another general question, concerning how words are decomposed into morphemes. A recurring theme throughout the presentation will be the ways in which the purely morphological representations implicated in processing relate to developments in theoretical linguistics: in particular, to theories that posit a syntax that operates with abstract morphemes that are connected to form and meaning representations only after they have been assembled into complex objects.