This thesis begins with the observation that portions of words may be separated out and interpreted across a distance, as illustrated by the following examples:
(1) These text tiers had boundaries that corresponded with the onset and offset of pre- and post-boundary syllables as well as the onset and offset of consonants and vowels that made up these syllables. (COCA, 2011: ACAD: JSpeechLanguage)
(2) My distinction between the pre- and the post-paradigm periods in the develop- ment of a science is, for example, much too schematic. (Thomas Kuhn, The Structure of Scientific Revolutions)
This observation plays into a longstanding debate within the community of linguistic researchers concerning the nature of words. Lexicalists argue that words are formed in the lexicon and are opaque to the syntactic component of the grammar. Alternately, non-lexicalist approaches to morphology and syntax contend that both words and larger phrases are generated in a single component of the grammar by the same means. It is therefore fruitful, in seeking to resolve this debate, to examine closely sentences in which elements which are smaller, in some sense, than words appear to undergo syn- tactic manipulation. Early investigations of examples like (1) have suggested that they might not be as revealing on this point as they look, as it is not obvious that such ex- amples involve syntactic displacement. If we concede that words and larger phrases are superficially very similar, than the fact coordination appears to be present in both domains may simply be another superficial similarity. It is quite difficult, however, to extend this argument to examples like (2), because the elements on either side of the coordinator and are not constituents. Something beyond coordination must be occurring.
Sentences like (2) exhibit what is known in syntactic analysis as Right Node Rais- ing (Ross, 1967). There are numerous, highly contentious analyses if Right Node Rais- ing, ranging from coordination plus deletion (ellipsis), to coordination with multiple constituents that share one or more elements (multidominance), to “across-the-board” syntactic displacement (movement). Hankamer (1971), looking at Right Node Raising in Turkish, observes several ways in which Right Node Raising and ellipsis may be distinguished in the data. Turning to English and German, Barros and Vicente (2011) suggests that at least some of these distinctions cannot be upheld for apparent instances of Right Node Raising. At least some of the time, they conclude, Right Node Rais- ing and ellipsis may be two instances of the self-same phenomenon. This matters in part because some analysts, like Hartmann (2000) and Chaves (2008), view ellipsis as a purely phonological (i.e., extra-syntactic) phenomenon. If discontinuous affixation instantiates ellipsis and ellipsis is extra-syntactic, then evidence of Right Node Raising below the word says nothing about the nature of word-formation and its relationship (or lack thereof) to the syntactic component of the grammar.
This thesis begins with a very close look at the phenomena at hand and carefully catalogs and details the various properties of sentences like (1) and (2). The findings of that investigation are then used to evaluate the viability of an analysis of sentences like (2) as ellipsis. I show that while many of the insights from Barros and Vicente (2011) remain intact, the overall conclusion must be false: It is not the case that Right Node Raising and ellipsis overlap analytically, but rather that certain constructions are analytically ambiguous between the two analyses. Turning to the sublexical domain, I demonstrate that for at least some cases, an analysis in terms of ellipsis is not possible. Separately, I consider non-syntactic approaches to ellipsis like those offered in Hartmann (2000) and Chaves (2008), and conclude that such approaches lack the clarity and definition that might allow them to account for differences in ellipsis above and below the word. We are left with the conclusion that a non-syntactic analysis of examples like (1) is not possible, and consequently, such examples do provide evidence of syntactic manipulation below the word.