UC San Diego

The observation that the same coherence relations serving to constrain interpretation at the discourse/intersentential level also operate on constituents within sentences (we label enrichments resulting from such intrasentential coherence relations ’eliciture’) has surprising and far-reaching consequences for our understanding of coherence and coherence establishment. Current accounts construe discourse coherence establishment — and, therefore, discourse level coherence-based enrichment — as resulting from a bottom-up search for ways in which elements of contents expressed might be coherently related to one another, and mandatorily triggered by a requirement to bind components into wholes. But such accounts leave us without an obvious explanation for intrasentential coherence. One problem is that intrasentential coherence establishment (unlike intersentential coherence establishment) is not required for felicity: hence, one cannot see coherence-based enrichments uniformly as the mandatory downstream consequence of a trigger. A second problem is that, because elicitures can arise from complex interactions between any combination of a sentence’s constituents, an account rooted in a search for possible coherence relations between expressed contents quickly runs into trouble: such a search would have to compare the contents expressed by every pair of constituents, then every triple, and so on. This is clearly not computationally tractable. These and related considerations suggest a quite different picture of the inferences arising from coherence establishment – one on which such inferences are not results of triggered searches, but the inevitable upshots from entertaining combinations of linguistically expressed contents, analogous to our recognition of causal and other relations obtaining between components of the non-linguistically presented world. We’ll develop this picture by starting with eliciture, then show how it can be extended to intersentential coherence establishment, and finally draw out consequences resulting from this reconceptualization. Among other benefits, we’ll contend that our account provides explanations for features (such as the preference for causal interpretations) that have required special principles in the more traditional accounts developed with only intersentential coherence in mind.