Beyond Plain and Extra-Grammatical Morphology: Echo-Pairs in Hungarian

Abstract

This paper presents an investigation of echo-pairs in Hungarian. Echo-pairs are formed by duplicating a base with an altered initial consonant and have diminutive, playful, or intimate connotations (e.g., cica [t͡sit͡sɒ] “cat” → cica-mica [t͡sit͡sɒ-mit͡sɒ] “cat.dim”). Echo-pairs are commonly seen as an example of extra-grammatical morphology in the literature. Our goal in looking at this phenomenon is to gain a better understanding of the morphological mechanisms underlying extra-grammatical phenomena and shed new light on the distinction between plain and extra-grammatical morphology. We analyze data from (a) a collection of echo-pairs extracted from a large corpus of online texts and (b) a large-scale online nonce-word experiment with close to 1,500 participants. Our results reveal two key phonological patterns in the data and some additional systematic variation across words and experimental stimuli. We compare two different models of morphology, the Minimal Generalization Learner and the Generalized Context Model, in terms of their ability to capture this variation. We find that echo-pair formation is best captured by lexicon-oriented models such as the Generalized Context Model, but only when they rely on a structured similarity metric that encodes broader generalizations about the data. Our results do not support a clear-cut distinction between extra-grammatical and plain morphological processes, and we suggest that some of the peculiar characteristics of extra-grammatical phenomena such as echo-pair formation may simply follow from their special function and the limited set of contexts in which they appear.