I am currently rebuilding this website and reorganizing everything. Sorry for the mess in the meantime. The information below is outdated now.
I am a Linguistics PhD candidate at the University of Chicago, focusing primarily on morphology and its interfaces. I have worked on and am interested in Vietnamese and other SE Asian languages, Ndebele, and Modern Hebrew.
My dissertation broadly addresses the issue of morphological learning and by extension what we consider morphemes to be: specifically, I propose more gradient notion of correspondence between form and meaning. The issue of morpheme learning is broken down into two primary questions: (1) what are non-trivial phonemic sub- parts of words?; and (2) what semantic meaning should these subparts be associated with? To answer these questions, I propose a decompositional, computational framework based on inducing generalizations from statistical distributions in a speaker’s linguistic experience, using the lexicon as an instantiation of that experience. Contra traditional compositional frameworks of morphology, I assume that decomposition of a word (form), defined in the Bloomfieldian convention as a minimal utterance, can be non-exhaustive.
I demonstrate this non-exhaustive decompositionality to be the case for word truncation (i.e. mayo from mayonnaise) and phonaesthemes (i.e. gl– in glitter, glimmer, glisten, …), two phenomena that have historically lay on the outskirts of morphological theory. In the model I propose*, truncation provides an example of how phonemic subparts of a word can be identified by optimizing recoverability of the original word given the preserved material and deletion of the deleted material based on its informativity. The case study in phonaesthemes shows how phonemic substrings can be assigned meaning to varying degrees of confidence based on the semantic similarity of words sharing that substring. Crucially, the principles of my framework apply to conventional morphemes as well, and as a result do not require an ontological difference between conventional morphology and so-called ‘extra-grammatical’ morphology, such as truncation and phonaesthemes (c.f. Mattiello 2013 for an overview of other morphological phenomena that have been argued to lie beyond some standard sense of grammaticality).
While non-exhaustive decompositionality is possible, exhaustive decomposition is as well, yielding not only the familiar compositional structures seen in morphosyntax and morphosemantics such as dog-s, but also morphological reanalysis such as alco-holic. Pressure for exhaustive decomposition leading to morphological reanalysis can come from generalizations induced from other grammatical domains – i.e. syntax, semantics, phonology – which I show to be the case for Vietnamese compound heads being reanalyzed as classifiers.
The perspective I propose here allows for standard linguistic theory forming a stable core of the grammar, while still allowing us to capture the mechanisms underlying relatively variable and seemingly arbitrary properties of morphological phenemona such as truncation (and by extension blends and other subtractive processes), phonaesthemes, morphological reanalysis (i.e. backformations, folk etymologies, etc) – extra-grammatical morphology is grammatical after all. By re-examining the conception of a morpheme as a more gradient correspondence between form and meaning, induced from distributions within a speaker’s linguistic experience, I hope to provide framework that can naturally be extended to account for speaker variation and diachronic change.
*work in conjunction with Jackson L. Lee