Michal Starke - Nanoseminar, Fall 2009

Topics. This semester, we will go back to basics, and try to refine some of the core technology of nanosyntax. Much of the initial discussion will center on pointers, and their roles in derivations. Given that discussion, we will then revisit the tense system of English. And if time remains, we will go into some issues of semantics, picking up on the discussion in the reading group last year.

The discussion will start around a set of unresolved paradoxes in the current formulation of nanosyntax:

• the concept-immutability paradox: given that the "biggest wins", a lexicalised plural such as 'mice' should be able to override a word without a lexicalised plural such as 'elephant', since the latter will always be smaller. But it doesn't. This is a paradox that was noted from the very start of nanosyntax, but has only ugly (and sometimes self-defeating) solutions. (The same paradox can be constructed on the basis of the other competition principle - minimise junk)
• Pavel's pointer-inclusiveness problem: if lexical items are just stored syntactic structures, how come they contain pointers whereas syntactic structures don't have pointers in them?
• phrasal spellout vs spanning: we all agree that lexical item spellout several terminals at once, and that these terminals look like they don't form a constitutent. Some of us are happy just stating that, others try to derive it in a principled way. Which way should we go?
• missing pieces vs identification: we all agree that lexical items can spell out subsets of the terminals they include. But some think that the unspelled terminals must be present in the syntactic structure anyway - the identification principle - (and hence ultimately lexical items must be entirely spelled out every time), whereas others think that the unspelled parts can go genuinely missing.
• phrasal spellout vs remerge
• parameters vs overgeneration

We will then continue our road by examining the two detailed case-studies in nanosyntax - Taradsen's Nguni class markers and Caha's case affixes - and discuss the issues that they raise for the technology of nanosyntax.

Dates. Look at the calendar for the latest news on rest of the menu.

Content. Here is a brief session-by-session summary of what we are doing:

• 2009.09.17: Continuation of the overview of the technology of nanosyntax, and highlighting of the associated paradoxes. Given that a chunk of syntactic structure is matched to a lexical entry by the superset principle, several competing chunks could match, whereby one chunk contains the other or overlaps the other. How are these conflicts resolved? We went over a few cases showing that the descriptive generalisation when one potential spellout chunk contains another is that "the biggest wins". We then saw that we don't need to state either this or the superset principle, as they already follow from phrasal spellout in a bottom-up cyclic merging and spellout. However, as it stands, this picture has a devastating consequence:
  (iv) the plural of 'elephant' is 'mice'. In fact the plural of every noun is 'mice'. Because the struture of 'mice' is bigger than the struture of 'elephant', and therefore wins over it. The fact that the concepts associated to the two lexemes are different is irrelevant, since there are no concepts in syntax and modularity (encapsulation) prevents us from placing a phone call to the relevant module.

• 2009.09.10: We went over the architecture of nanosyntax, in order to show the paradoxes and disagreements lurking in there and in order to familiarise the newcomers with the overall picture of the world in nanosyntax. Main points: the growth of trees, submorphemic terminals, no presyntactic lexicon, phrasal spellout, a principled lexicon consisting of regular syntactic trees. At that point we met two issues:
  (i) if the lexicon only comes into play after syntax/morphology, and parameters reduce to lexical variation, it seems follow that there is no cross-linguistic variation in syntax. Syntax should be invariable. Every language should have the same word-order. Eeeek.
  (ii) There is a difference of terminology, and maybe of opinion, among the nanopeople about spelling out chunks of syntactic structures: one part of this little crowd is happy just stating that a 'stretch' of terminals (or 'span' of terminals) can be spelled out as a single morpheme, another part of the crowd wants to be more principled and only allow spellout of constituents. This is a tension we will need to resolve.
  We then moved on to one aspect of this picture: if spellout is not married to single terminals, how does one decide which node to spellout, and what is the algorithm matching lexical entries to syntactic parse trees? We went over the basics of the algorithm: the superset principle, based on one of the case-studies using it, my syntax of ED/EN in English. At this point we met a third issue:
  There is a clash between the "identification" requirement in Ramchand (2008) and the subset principle used by several nanostudies.