Elsewhere and parenthesis notation

This is one of those posts where I just assume you’re a phonologist.

Suppose you have a set of rules that can be collapsed into a single rule A using SPE parenthesis notation, and another (nonabbreviated) rule B. Is it possible for A and B to be disjunctively ordered via the Elsewhere Condition — that is, is it possible to meet these two conditions?

  1. The structural changes (SCs) of the two rules conflict.
  2. The structural description (SD) of one of the rules properly includes the SD of the other.

The real difficulty is with (2). For starters, what is the SD of the abbreviated Rule A? Is it equivalent to the SD of its longest expansion, or something else?

An example might help.

Here is a simple example. The two rules that Rule A abbreviates are spelled out individually for perspicuity; two examples of a Rule B are considered to highlight the issue.

Rule A: [+high] → [-back] / [-back] ( C ) __

  1. [+high] → [-back] / [-back] C __
  2. [+high] → [-back] / [-back] __

Rule Bx: [+high] → [+back] / i [+dors] __
Rule By: [+high] → [+back] / i __

The SC of Rule A conflicts with the SCs of both of the Rules B. The SD of Rule Bx is properly included in the SD of Rule A-1 but not that of A-2, and the SD of Rule By is properly included in the SD of Rule A-2 but not that of A-1. What is the Elsewhere Condition to do in either of such cases?

3 thoughts on “Elsewhere and parenthesis notation

  1. Bob Kennedy

    Seems like B applies regardless of whether B is Bx or By:

    If the language has Bx, and Bx’s description is properly included in A1, Bx applies and A1 is skipped. A2 could not apply anyway because any sequence subject to Bx ( e.g., I in […ikI]) does not fit A2’s SD.

    Conversely if the language has By, and By’s desacription is properly included in A2, By applies and A2 is skipped. Nothing that By applies to could have A1 apply to it.

    Maybe another way of looking at it is that the three rules — A1, A2, and whichever B is active are in a three-way disjunctive relationship.

    Alternatively you meant that the language has all four rules. Did you?

  2. Eric Bakovic

    Not necessarily — but I didn’t intend the example to be so easily defeasible, either. Guess I was pretty tired when I posted that.

    If the language had all four rules, then it would have two rules abbreviated by parenthesis notation (assuming there is no evidence that the rules in question are not adjacent in the ordering). But things would still work out as Bob noted: application of either subrule of B would block application of either subrule of A, and the other subrules would be inapplicable anyway.

    I guess what might be interesting is to construct an example in which the SD of one subrule of B is properly included in the SD of one subrule of A, but the SD of another subrule of A is properly included in the SD of another subrule of B. Is that even possible to do? And if so, would the Elsewhere Condition trump abbreviation by parentheses?

  3. Bob Kennedy

    I might have found the SDs Eric was looking for. It relies crucially on some standard feature co-ocurrence restrictions … there’s actually two pairs of abbreviated rules that would fit the description.

    Rule A’s SD is [+cor](+nas)___
    Rule B’s SD is [+cor, +son] (-nas) ___

    Both A and B are abbreviated.

    So A is shorthand for
    A1 ____[+cor]
    A2 ____[+cor, +nas]

    And B is shorthand for
    B1 ____[+cor, +son]
    B2 ____[+cor, +son, -nas]

    B2 is a proper subset of A1, while A2 is a proper subset of B1. (Only if [+nas,-son] is impossible). So the environments are possible, but I don’t know if the Elsewhere Condition countervenes.

    Here’s another set of environments … this time using curlies rather than parantheticals:

    A: ____ {+voi/-voi}
    B: ____ {+son/-son}

    Which is abbreviation for
    A1 ____ [+voi]
    A2 ____ [-voi]

    B1 ____ [-son]
    B2 ____ [+son]

    Again, assuming the impossibility of [+son,-voi], A2 properly included in B1, while B2 is properly included in A1. (Or am I not being fair? The rules are equally specific at face value, but the set of segments to which they apply is divisible into subset/superset relationships).

Comments are closed.