This much lays the foundation for the discussion of how OT obtains the process-specific blocking effect. The rest is quite straightforward, once the role of constraint ranking is grasped. The ranking between two structural constraints, like the ranking between M and F, determines which is obeyed and which is violated in case of conflict. The interesting conflicts are between the RTR alignment constraints and the featural constraint RTR/Hi&Fr ( *[high, front, RTR]). Leftward harmony proceeds with total disregard for the creation of offending segments, indicating that RTR-Left dominates RTR/Hi&Fr, as tableau 3 shows. Under this ranking, RTR/Hi&Fr is irrelevant to the satisfaction of RTR-Left.
RTR-Left >> RTR/Hi&Fr
* ! **
On the other hand, the ranking of RTR/Hi&Fr with respect to RTR-Right is just the opposite, as shown by tableau 4. Under this ranking, satisfaction of RTR/Hi&Fr blocks the process of rightward RTR harmony.5 Playing on a terminological ambiguity, one might also say that the high front segments are blockers, in the autosegmental sense, of RTR harmony — because they cannot become RTR themselves without violating high-ranking RTR/Hi&Fr. (Nor can they be can they be skipped over, as in *Sayyaad. Outcomes like this are excluded by a high-ranking constraint No-Gap (Kiparsky 1981, Levergood 1984, Archangeli & Pulleyblank 1994), just as in Davis’s analysis.)
RTR/Hi&Fr >> RTR-Right
In summary, the constraint hierarchy motivated by these arguments is as follows:
() Full Ranking for Southern Palestinian
RTR-Left >> RTR/Hi&Fr >> RTR-Right >> Stay-ATR
This ranking yields the desired pattern of process-specific blocking. The constraint responsible for rightward harmony is dominated by RTR/Hi&Fr, limiting its influence. The constraint responsible for leftward harmony itself dominates RTR/Hi&Fr, so leftward harmony can affect all segments, even those that are high and front. Thus, RTR/Hi&Fr is process-specific, in just the required way. Process-specificity follows from the ranking of constraints, in an entirely unremarkable application of OT.
The constraint hierarchy in () can be used to give a general picture of process-specific constraint limitation, thereby fully answering Davis’s assertion that process-specificity is impossible in OT. Suppose a grammar has two structural constraints Mi and Mj, each of which crucially dominates some appropriate faithfulness constraint. Suppose too there is a structural constraint C that stands between them in the ranking:
() Ranking for Process-Specific Constraint Interaction
Mi >> C >> Mj >> F
Under this ranking, the activity of Mj is mitigated by higher-ranked C, while the activity of Mi is unhindered by C. The constraint C, then, is “process-specific” to the Mj >> F interaction, but specificity is achieved through language-particular ranking rather than language-particular parametrization. The ranking for Southern Palestinian RTR harmony in () is simply a specific instance of this general schema.
From this schema and the analysis of Southern Palestinian, it emerges that nothing fancy is required to obtain the effect of process-specific constraints in OT. The only mechanism invoked is constraint ranking, which is the sole essential and fundamental element of the theory. Process-specific constraints are no challenge to OT — on the contrary, they emerge from its most basic assumption.
What leads Davis to conclude otherwise? He asserts (p. 495) that “general constraints on the language” are the only alternative to parametric process-specificity as in (). In his view, OT posits only “general constraints on the language,” whose scope cannot be limited to one process or the other.
A basic misconception about OT is at work here. It holds that constraints in OT are on or off categorically, so that all on constraints must be generally true of the language. The on/off notion makes sense in rule-based parametric theories, but not in OT. Because constraints are ranked, there is no privileged class of constraints that are activated — there is only the constraint hierarchy, which may support or blunt the force of a constraint in one circumstance or another.
This misconception about OT is an instance of what McCarthy and Prince (1994) call “the fallacy of perfection.” Succinctly, the fallacy says “optimality=perfection”; in the particular case McCarthy and Prince discuss, the fallacy leads to the assertion that, if OT were correct, all words in all languages would be pronounced as ba (Chomsky 1994), since ba achieves supposed perfection on dimensions of unmarkedness. But unmarkedness stands in fundamental conflict with faithfulness, as the discussion above has emphasized (also see in particular Prince and Smolensky 1993: Chapt. 9). The ba-fallacy disregards the role of faithfulness constraints in the evaluation of forms, promoting unmarkedness above all else. Moreover, it disregards the possibility of conflict among the structural constraints and among the faithfulness constraints. Thus, the ba fallacy wrongly presupposes that perfection can be achieved on every dimension — by ignoring constraint conflict and constraint ranking, the very essence of the theory.
In Davis’s interpretation of the Palestinian case, the fallacy puts constraints like RTR/Hi&Fr at the pinnacle of the hierarchy, requiring perfection on that dimension. This misconceived version of OT also fails, because it too ignores the consequences of constraint ranking.
We have now seen how OT obtains the process-specific constraint effect without building the constraint into the formulation of the process, as the parametric rule-based approach does. In the next section, we will see that this property of OT leads to a more restrictive, and hence more interesting, view of process-specificity than the parametric theory. First, though, a few details must be mentioned. They are tangential to process-specificity, but they are necessary in a complete OT account of Southern Palestinian RTR phonology. The details include: (i) constraint conjunction and RTR/Hi&Fr; (ii) additional faithfulness effects; and (iii) RTR in the phonemic inventory. I will briefly consider each of them in turn.
(i) In Archangeli and Pulleyblank (1994), the interaction of tongue-root and tongue-body position is seen in terms of two distinct constraints: RTR/Hi ( *[high, RTR]) and RTR/Fr ( *[front, RTR]). Davis’s rule () conjoins these two constraints in a single parametric slot. Conjunction of the two constraints is also important in the OT analysis ().
Constraint conjunction is somewhat unexpected and arbitrary in a parametric theory with rules like (); why can these parameter-values combine, but not others? As precedent, Davis cites Archangeli and Pulleyblank’s (1994: 412) analysis of Lango, which also invokes constraint conjunction. Significantly, though, Archangeli and Pulleyblank work out their analysis of constraint conjunction in Lango within OT. Constraint conjunction is a natural extension of OT and has been much studied in the OT literature (Smolensky 1993; Prince and Smolensky 1993: Chapt. 5; Hewitt and Crowhurst 1995).
It makes sense to conjoin RTR/Hi and RTR/Fr because they both say that it is hard to constrict the pharynx when the tongue body is being pulled in the wrong direction (superiorly or anteriorly). Constraint conjunction is synergistic: if combining RTR with high orfront is prohibited, then a fortiori combining RTR with high and front is prohibited. This yields the following universal ranking: