Optimal Divinity and Divine Optimality: Key Points of Difference in DUG and OT
By Athanasious Schadenpoodle
Both Divine Unification Grammar (henceforth DUG; adumbrated by Garcia 1990) and Optimality Theory (Prince and Smolensky 1993) represent attempts to simultaneously maintain claims for universals and for meaningful language particulars. DUG does so, of course, by positing a language-transcendent pantheon, the members of which are not always in agreement, and the collective judgments of whom, therefore, can vary over specific languages.1 OT approaches the same problem by positing violable universal constraints, the ranking of importance of which varies from language to language. Essentially (and these theories are nothing if not essentialist), when a given language’s forms contradict a principle claimed to be universal, DUG deals with the situation by claiming that one of the gods did it, while OT claims that an invisible constraint ranking order did it. There are, of course, additional important similarities and differences between the two theoretic models. On the assumption that there is a community of theorists interested in these models, and that Butler probably won’t be doing a comparison of Numinolinguistic theories any time soon, I shall provide a cursory comparative overview.
DUG has no explicit equivalent of OT’s ability to discuss phenomena in terms of constraint rankings. However, DUG’s structure does not prohibit such. One can argue that DUG’s omission of explicit discussion of this topic is motivated pragmatically rather than theoretically; there is ample textual evidence supporting the idea that assigning ranks to deities is the kind of thing that mortals should avoid under any circumstances (see, for example, Troy, the ruins of). OT theorists are presumably not concerned with the possibility of the occasional admonitory thunderbolt or targeted armada. Indeed, practitioners appear to be conscious of very few potential threats whatsoever, whether they be from the gods, from work on phonology in other theories, from philosophy of science, or from anything outside the U.S. or before the 1960s.2
One can argue, however, that there is less difference between the two theories on this issue than is immediately apparent. The symbol commonly used for a constraint violation in OT is <*>, obviously derived from the Sumerian cuneiform DINGIR sign () indicating a god. It is thus possible that OT theorists have solved the “hubris→ate” problem by referring to gods only indirectly, focusing on a god’s judgment without specifying the judgment’s exact origin. Additional support for this interpretation comes from the use of the exclamative <!> symbol for “fatal” violations, as this is exactly the sort of thing one would expect of authors who are obliquely indicating that this * is wrathy, and is not about to be bought off by a couple of plump sheep, or a nice bronzed Chihuahua.3
DUG could incorporate this strategy by separating “whims” from “deities,” yielding statements with a comfortably high PDI, e.g.:4
Somebody’si anti-voicing whim edged out somebody’sj say-everything commandment, if you know what I mean. Just sayin’.5
Importantly, any given whim in DUG can potentially come from any member of the pantheon, although naturally some fall closer to a given god’s usual domain than others. Far from being a disadvantage, this ambiguity serves as an additional layer of indirection. Also, it allows a many-to-one matching between whims and deities, reducing the number of underlying figures needed for explanation. One does not need to add more constraints to explain phenomena; one needs only to posit that a god can have more whims. And one does not want to posit that a god cannot have more of anything that god wants. While this would appear to render DUG formally simpler than OT, the latter can be easily argued to be equivalent, as the various instances of <*> are standing for particular deities—i.e., they have invisible subscripts.
Given a set of constraints and a language-specific ranking order, OT predicts that there should be no unexplained variation in how inputs are matched to outputs. When such variation is observed, the difference can be accounted for by positing that the inputs are actually different—the “richness of the base” principle allows the inputs to have all sorts of additional bits as needed—or that the speakers are switching varieties, and thus adopting a different (but totally-consistent) ranking order. DUG approaches the same issues by positing that one or more of the deities involved changed his/her/its/their mind(s). Again, however, this difference may be more apparent than real. Additional elements of OT inputs can be reanalyzed, within a DUG framework, as state-of-mind or time indices, thus representing a statement about change over time in a deity’s attitude as if it were an element in a phonological sequence, e.g., we could say that /marmət◦/ is acceptable, whereas /marmət•/ is not, with the understanding that /◦/ is to be read as “deity X’s opinion of this form at time Y” and /•/ as “deity’s opinion of this form at time Y+n.”
Both theories rely on denial or suspension of natural principles, but they do so differently. DUG requires that the pantheon manage all language used by all speakers everywhere simultaneously—something that only gods could do.6 It thus considers the actions of the numinous as manifesting in everyday life. OT requires that GEN create a potentially infinite set of possible forms so that EVAL can instantaneously winnow them—something that only gods could do. It thus instantiates a kind of high-speed eschatology of forms; that which is actually said is that which has been deemed Select—a corpus of utterances is, in a sense, a record of the afterlife of language. Both options are equally believable, equally falsifiable, and fit observable phenomena equally well. We cannot distinguish them on empirical grounds, but they create crucial differences in focus. DUG holds out to the speaker the possibility that any sentence could be pleasing to the pantheon; the fate of a given sentence is not determined at its generation, but is rather shaped by how the gods react to it over time. The speaker is left in a state of suspended decision; hopeful, but on the lookout for the occasional overly-excited swan or suspiciously aureate raincloud. OT, on the other hand, is Calvinist; the myriad candidates for production are for the most part born doomed. The shadow of EVAL is upon them, whether they show outward signs of it or not.
It is in this area that one of the more intriguing differences between the two belief systems appears. DUG, which is most explicit about its theological roots, has almost nothing in the way of documented ritual behavior, priestly regalia, etc., although Garcia specifically suggested that the ability to add such was a positive characteristic of the theory. There is a reference at one point to leaving an empty space for Nullifica, but this is more like a quirk than a practice. After all, it is rather difficult to tell when a space is not being left for someone, unless the someone’s size has already been specified. OT, on the other hand, obscures its nature as a religious account but yet is replete with ritual practices—the carefully formatted grids to track the judgments of the gods, the use of all-caps to mark names as divine, the profligate use of quaint and unnecessary formalisms and abbreviations to separate esoteric from exoteric knowledge, and so on. One possibility is that practitioners are signaling the theory’s Gnostic status, making clear that there is secret knowledge but enjoining the postulant to prolonged effort to gain it. Alternatively, they may have (in a rather clever twist) actually adopted a position in which outward appearances determine essentials—if it looks sciency, it must be science. These are not mutually exclusive, of course; postulants who view the texts as scientific but do not focus on the status of GEN as a rupture of reality would, presumably, be unready for the higher levels of esoteric knowledge the theory makes available.
1 Despite OT’s somewhat later date, it almost certainly represents an independent development. Neither Prince nor Smolensky were on the mailing list for the venue in which DUG appeared, Linguist of Fortune (the author obtained and exhaustively scanned both pages of said list), and P&S did not address the same side-issues as one would expect if OT were a comment on DUG (there is, for example, no independent motivation for the invention of the raincoat presented in OT). And there is just no comparison between the cover art of Linguist of Fortune and that of Prince and Smolensky (1993).
2 What to the outsider appears to be a rather ominous “floating finger of doom” glyph in the theory’s ritual designs turns out, on closer examination, to represent judgments about linguistic forms, rather than potential judgments about the linguists themselves, and so (pace Threadhandle (1998)) cannot be taken as coded exhortatations to exercise caution.
3 Nullifica likes these, reportedly. There is a debate over whether the key point is that the Chihuahua is pleasing because it is bronzed, or whether it is pleasing because bronzing has killed it.
4 Plausible Deniability Index
5 For a full formalization, of course, several of the words of the above would have to be replaced by Greek letters, but this should not prove difficult, particularly as ξ usually needs something to do. The just sayin’ operator is implemented in several fourth-order logics.
6 The rather obvious similarity of GEN to “djinn” should not be taken as a sign that OT theorists are assigning it a sub-deific status. It is, after all, capable of generating an infinite number of forms simultaneously, and there is no evidence that OT views it as satisfying a limited number of, say, wishes on the part of semantics.