Underlying representations in phonology


Martin Krmer

University of Tromsø / CASTL



Underlying representations are at the heart of modern phonological theorising. Most contemporary, especially generative, phonological theories are mechanisms that map assumed more or less abstract underlying representations to much less abstract phonological representations, which are either regarded as instructions to the articulators or translated into such in a phonology-phonetics interface component. The degree of abstractness, as well as the primitives of phonological representations in general and in particular of underlying representations have been subject to a long debate starting with the introduction of underlying representations into modern linguistics by Bloomfield (1933). Before this, the Prague School, based on Saussurean work, introduced a revolutionary degree of abstractness with the archiphoneme (Jakobson 1929, Trubetzkoy 1939), restricting the phonological content of phonemes to distinctive features. With the introduction of a rule component in generative grammar (Halle 1959, Chomsky & Halle 1968), a tool was available to strip underlying representations from all predictable or otherwise redundant information and arrive at fully specified segments at the exit from grammar, leading to an ever higher degree of abstractness.


Underspecification and abstractness had its downs (Stanley 1962) and ups (Kiparsky 1981, 1985 and most of the phonological literature of the 1980s) and its downs again (with the emergence of Optimality Theory (OT) and of Exemplar Theory (ET) in the early 1990s and late 1980s, respectively). In these lectures we will concentrate on three issues: Which evidence is there for contrastive phonological features and for underspecification and other forms of abstractness? How are phonological features defined/grounded? Finally, we will have a look at the role of underlying representations in OT and discuss the theory's predictions for underlying forms.