Abstract
Many linguistic theories focus on producing utterances from an economical set of abstract units and/or devices. There exists however a contrasting perspective: that of comprehension, where the key matter is to discriminate meanings from surface observations. In the case of inflectional morphology, we show that the two perspectives are substantively different: they face different challenges and result in different analyses. In comprehension, the question of exponence can be phrased as the Paradigm Cell Recognition Problem: what recurrent patterns in words could language users use to discriminate inflectional meanings? We provide a formal and implemented theory of exponence from this perspective. The resulting units (formatives) do not coincide with traditional morphemic segmentations, but rather, constitute the smallest discriminative cues which speakers could attend to for the comprehension task. We show that starting from the perspective of discriminative information, very simple principles lead to unambiguous analyses in terms of both segmentation and meaning. This new theory is especially promising for cross-linguistic research as it results in deterministic, comparable analyses across languages, and is defined independently of any model of cognition.