Neighborhood Density and the Root-Affix Distinction

Work in phonology and morphology consistently recognizes that affixes show a strong tendency to be less marked than roots crosslinguistically. Within Optimality Theory, this observation has been dealt with through the imposition of a universally-fixed ranking, known as the Root-Affix Faithfulness Metaconstraint (McCarthy and Prince 1995), which states that faithfulness to root material universally outranks faithfulness to affix material. In this paper, we present evidence drawn from psycholinguistic research that provides functional grounding for the root-affix markedness distinction without resorting to any stipulative device or metaconstraint. Research into lexical access processes shows that the frequency and lexical neighborhood characteristics of a target lexical entry have strong effects on its efficiency of access. The Neighborhood Activation Model, which accounts for these findings by modeling access as a process of lexical competition, predicts that because affixes are significantly more frequent than roots, they will tend to contain less marked material. This prediction stems from the proposal that it is effective contrast, a function of both frequency and phonemic contrast, that is optimized relative to markedness, not phonemic contrast alone. Thus, despite the relatively low phonemic contrast of material in affixes crosslinguistically, their high frequency allows them efficient processing and access, leading them to evolve toward less marked structure. Further, we show that the Neighborhood Activation Model predicts that when language-specific factors result in different affix neighborhood relations, such as in Hebrew, the Root-Affix Faithfulness Metaconstraint may be reversed. Hence, the ostensibly anomalous behavior of affixes in languages like Hebrew can be straightforwardly accommodated by abandoning the metaconstraint strategy in favor of a model of contrast based on psycholinguistic evidence that the efficiency of lexical access