To OT or not to OT

(OK, this is not, strictly speaking, a phonology-related post. But it’s about Optimality Theory, which is mostly about phonology, and the Language Log post that I discuss here was not, strictly speaking, a language-related post either. So there.)

As the resident Optimality Theorist over at Language Log (or as some folks these days apparently like to say, “optimologist”, as they chuckle to themselves), I feel the need to point out that Bill Poser’s parody-post on different date formats doesn’t even resemble an OT analysis.

First, the proposed constraints (copied below) are stated as yes-no questions of the candidate date formats, rather than declarative statements that establish preferences among those candidates. Without preference establishment among candidates, you’re not doing OT.

  • Monotonicity
    Are the units in a consistent increasing or descending order?
  • Adjacency
    Are the month and day adjacent?
  • Endianness
    Of the first two units, does the larger come first?

Second, the differences between the US (month-day-year), International (day-month-year), and ISO-8601 (year-month-day) date formats is not accounted for by re-ranking of the same constraints. Suppose, for example, that yes-answers to Bill’s question-constraints correspond to constraint satisfaction, and no-answers correspond to constraint violation:

Pattern Attestation Monotonicity Adjacency Endianness
y-m-d East Asia and ISO8601
d-m-y international *
m-d-y US *
y-d-m US Fish and Wildlife Service *
m-y-d unattested? * * *
d-y-m unattested? * * *

ISO-8601 (y-m-d) is the only possible date format given these three constraints, because only that format satisfies all the constraints. (In the parlance of OT, all other date formats are harmonically bound by ISO-8601; none can be chosen under any ranking of the constraints.)

So, the constraints must make preference statements and their re-rankings must select all (and only) the different attested candidates. For example, something like the following would account neatly for the ISO-8601, US, and International date formats (I’ll get to the US Fish & Wildlife Service y-d-m format, which Bill added in an update to his post, in a moment.):

  • Monotonicity (Mon)
    There must be monotonic increase or decrease.
  • Month-b4-Day (Mb4D)
    Month must precede day.
  • Year-at-End (YaE)
    Year must come last.

The ISO-8601 (y-m-d) format satisfies Mon and Mb4D at the expense of YaE, which can be gotten via the partial ranking { Mon, Mb4D } >> YaE:

Pattern Attestation Mon Mb4D YaE
y-m-d ISO-8601 *
d-m-y Int’l *
m-d-y US *

The International (d-m-y) format satisfies Mon and YaE at the expense of Mb4D, which can be gotten via the partial ranking { Mon, YaE } >> Mb4D:

Pattern Attestation Mon YaE Mb4D
y-m-d ISO-8601 *
d-m-y Int’l *
m-d-y US *

The US (m-d-y) format satisfies Mb4D and YaE at the expense of Mon, which can be gotten via the partial ranking { Mb4D, YaE } >> Mon:

Pattern Attestation Mb4D YaE Mon
y-m-d ISO-8601 *
d-m-y Int’l *
m-d-y US *

There are no other distinguishable rankings of these three constraints.

In order to get the US Fish and Wildlife Service (y-d-m) format, something more would be necessary to explain the violation of all three of the proposed constraints. I don’t get the point of that date format anyway; it just seems to be the exact reverse of the US date format. Maybe it’s the US-F&WS folks’ way to spice up what might otherwise be a very dull job.

(Sten Vikner used to have a webpage — I can’t find it anymore — describing how OT constraint conflict and ranking works by calculating when to expect a leap year: every year that’s a multiple of 4, except when it’s a multiple of 100, except when it’s a multiple of 400. The problem in this case is that there’s only one pattern to account for; re-ranking of the constraints involved produces potentially different patterns, but no other patterns are attested — at least, not in the relevant sense.)

Final note: this post is, like Bill’s, not meant to be taken seriously.

3 thoughts on “To OT or not to OT

  1. Bob Kennedy

    Nice analysis Eric. I like how the underlying order of the components doesn’t matter … Richness of the Base at its finest.

  2. Will Emit Candidates for Food

    How about this for a functionalist theory of the US Fish and Wildlife Service date format: since they are men and women of nature, the annual cycle of life is most important to them, and therefore they put the year first. Months, as representatives of seasons, are also important, so they put the month in the prominent final position. The day goes in the only available leftover slot, in the middle.

    If one formalizes this account into a set of functionally grounded constraints (Year-First, Month-Last) and adds them to Eric’s constraint set, m-y-d and d-y-m are still predicted to be impossible: m-y-d is harmonically bounded by y-d-m, and d-y-m is harmonically bounded by y-d-m. Those skeptical of the power of pure reason are invited to verify this at their leisure.

  3. Andrew Nevins

    I had often heard the term “optimologist” before (with and without accompanying chuckles) but was tickled (from a purely morphological productivity point of view) to hear Larry Hyman use “optimitician” repeatedly at the LSA this year. Or maybe it was “optimotician” — ah, the ambiguity of reduced vowels. In any case with the revival of the Transformers movie, perhaps Optimus Prime tie-ins are now on their way (just think of Coetzee’s runner-up candidates as Optimus-Prime?).

Leave a Reply

Your email address will not be published. Required fields are marked *