IGRA 01

summary

We continued our discussion of the paper by Assmann et al.

homework

Please read the paper by Assmann et al.

1. List the assumptions they make
   • Try to formulate each one in as general a way as possible
   • Identify which assumptions are necessary for the analysis, and which are superfluous
2. Extend the analysis by Assmann et al

   • to include antipassivization:

     A₁ V O₂ → S₁ ap-V

3. Read the paper by Polinsky
   • Identify properties of ergative languages
   • Which of these can be accounted for by (natural extensions to) Assmann et al's analysis?

We went over the homework assignment. We discussed the role of typology in theory choice (and in the context of a 'guessing game'). We then discussed the paper by Koutsoudas et al on doing away with extrinsic rule ordering.

summary

We discussed the derivation of sentences with ergative-absolutive case marking, in particular given the typical 'mixed' behaviour of A and O arguments with respect to subject properties.

readings

• Aldridge (2008) Generative Approaches to Ergativity
• Polinsky (2017) Syntactic Ergativity
• Müller (2008) Ergativity, Accusativity and the Order of Merge and Agree
• Assmann et al (2012) Ergatives Move Too Early

summary

We discussed the rule interaction within the auxiliary system of Basque, as analyzed in Arregi & Nevins (2012). In this theory, the post-syntactic morphology is split into several strata, thus inducing an intrinsic rule ordering between different kinds of operations that belong to these strata. The rules discussed and their relevant ordering was

Absolutive Promotion > Participant Dissimilation > Metathesis/Reduplication

We also spent some time on the question whether it is possible that two rules A and B are such that both A feeds B and B counterbleeds A. This was claimed to be the case in Basque with A = Absolutive Promotion and B = Participant Dissimilation in the Ondarru dialect, where one the one hand Absolutive Promotion feeds Participant Dissimilation (by creating an ergative clitic), but on the other hand Participant Dissimilation does not bleed Absolutive Promotion although it deletes the dative clitic (in the context of an ergative clitic!), on the presence of which Absolutive Promotion depends.

The reason why one might think that this double relationship (feeding and counterbleeding) cannot hold is the following. First, if A feeds B, then it is clear that B does not apply unless A has previously created the context for the application of B (unless this context was given independently). Second, by definition, B counterbleeds A if

1. there is the expectation that B should bleed A, and
2. this expectation is not borne out (i.e. de facto there is no bleeding of A).

Third, for B to bleed A, B has to apply before A. Here, the question arose how B can be expected to bleed A (thus, B precedes A) if, due to the independently established feeding relation, it is clear that A must precede B.

The answer (at least for me) is that the assumption that B precedes A is hypothetical and independent of any other previously established orderings/interactions between A and B. Thus if under the hypothetical assumption that B applies before A, there arises the expectation that B bleeds A, and if this expectation is not borne out, then B counterbleeds A. If so, then A can be said to feed B and B can be said to counterbleed A.

summary

Ezer delved deeper into sequential interactions in phonology, by looking at a Yawelmani dataset. He began by characterizingof a Theory of Phonology as involving two components, following Chomsky and Halle's (1968) Sound Pattern of English:

• a specification of how grammars can be written (in our case, grammars consisted of a lexicon and a list of rules that can be written using a small set of primitive symbols)
• a specification of how grammars are interpreted (e.g., how rules apply, what symbols like C₀^infinity indicate)

We then developed an analysis of the Yawelmani data, beginning with a vowel harmony rule. We discussed how rules should apply to more than one position in a string, ultimately rejecting a model where a rule applies simultaneously to all possible targets on the grounds that it did not allow us to account for (descriptively speaking) a rule feeding itself. After carefully arguing for a particular ordering on the rules we ultimately arrived at, we~~somewhat ironically~~noted that, at least for one form which crucially required a particular ordering between vowel harmony, long vowel lowering, and vowel shortening, a 'simultaneous application' model of rules would have worked just as well. We raised the question of whether simultaneous application can work more generally, a question that we will return to in future sessions.

next time

We will hear more from Fabian about sequential interaction in morphology.

summary

Greg gave examples of opacity in syntax ranging from the straightforward (wanna-contraction, reflexivization in imperatives) to those which depend on more abstract assumptions (remnant movement in German, wh-movement from ECM complements). We ended with a case study on syntactic copying constructions, in which head movement counter-feeds copy-identity.

extras (?)

Here are the slides from the lecture

next time

Please read the following document:

• Kenstowicz and Kisseberth (1979) Generative Phonology chapter 4

summary

Ezer gave examples of opaque rule interactions (Catalan final-n deletion and post-nasal consonant deletion on the one hand, and English schwa epenthisis and sibilant voice assimilation). We then discussed Bakovic's argument that underapplication and overapplication should not be identified with counterfeeding and -bleeding respectively, with examples.

next time

There is no next time! Or rather, next week there is no class so that you can attend the DISCO workshop. The real next time is the time after that.

summary

Fabian introduced the concepts of two processes (counter-) feeding and bleeding one another, and went through some examples from the domain of morphology. The difficulty in finding convincing counter-bleeding phenomena in morphology, he suggested, was due to the configurations which would give rise to such being independently dispreferred due to violating either cyclicity or recoverability.

extras

In class I claimed that rule ordering need not have any temporal connotation. The formal justification for this claim can be found in the paper below.

• Kaplan & Kay (1994) Regular Models of Phonological Rule Systems

The basic observation is that phonological rules can be implemented as finite state transducers, and that rule ordering can be implemented via transducer composition. Thus, a set of ordered rules can be implemented as a single (big) finite state transducer. The same is true for Optimality Theory (see e.g. here).

The moral of the story is that notation can be misleading!

next time

Please read the Bakovic paper linked below. The McCarthy reading is optional.

• Bakovic (2011) Opacity and ordering
• McCarthy (2007) Hidden Generalizations section 2.2