In this experiment, we applied the method of elicited speech errors to the question of the morphological structure of the mental lexicon. Specifically, we investigated whether there is a 'grey area' between inflection and derivation, or whether there is only one way to produce either forms.
It has been shown by Humphreys (submitted) that inflected and derived words behave differently when subjects have to produce them in a task designed to elicit speech errors. The task we used here (similar to the one used by Humphreys earlier) involves the subsequent reading and production of two word pairs (eg, 'freshman cheaper - city cooler') under a REPEAT or SWAP condition. In the latter condition, the critical words 'cheaper' and 'cooler' have to be swapped. For other subjects, the critical words were replaced by their stems 'cheap' and 'cool' and also appeared in either a SWAP or a REPEAT condition.
Next to adjectives, we used a condition with deverbal nouns or their verbal stems: 'zero voter - sanction lover'. The '-er' suffix of these nouns is more closely bound to the stem than the '-er' suffix of the adjectives, because the meaning of 'voter' is not as predictable as that of 'cooler'. Linguistically, agentives are often considered derivations, although the behaviour of a large group satisfies common criteria for inflection (fully predictable in form and meaning).
The familiar morphologically opaque forms were used as a baseline, we used pairs like 'income crater - hollow clover', and pseudo-stem variants 'crate' and 'clove'.
Opaque words and the deverbal nouns behaved very similar on overall counts: Compared to adjectives, they are, surprisingly, produced in error more often. If we look at the restricted group of affix errors after correct stem, however, deverbal nouns and adjectives group together, having more errors than opaque items.
Critical-word omissions and replacement of the 'stem' with another
unit were highest for the opaque items, with deverbal nouns being in
the middle on both counts. We will suggest a way to structure the
inflectional encoding system such that this can explain our results,
using a notion of degrees of 'inflectionality' that ranks adjectives
highest, predictable deverbal nouns in the middle, and opaque items
off the scale.