Process-Specific Constraints in Optimality Theory


() Logical Relation Between RTR/Hi and RTR/Lower-VT



Download 178.73 Kb.
Page6/7
Date02.06.2016
Size178.73 Kb.
1   2   3   4   5   6   7

() Logical Relation Between RTR/Hi and RTR/Lower-VT


RTR/Hi RTR [–high]

RTR/Lower-VT (approximately) RTR [+low]

Chomsky and Halle 1968 [–high] [+low]

RTR/Hi RTR/Lower-VT

In this case, the constraints themselves are in a subset relation. So imposing RTR/Hi on long-distance rightward-harmony, as the subset criterion requires, does not change the effect of RTR/Lower-VT on the same process.

Putting these results together with the schema for process-specificity in (), we get the following ranking for Northern Palestinian:

() Schematic Ranking for RTR Harmony in Northern Palestinian Arabic


RTR/Hi >> Mi >> RTR/Lower-VT >> Mj >> F

The ranking Mi >> F characterizes the process of local rightward spreading, which is blocked only by RTR/Hi (see ()). The ranking Mj >> F characterizes the process of long-distance rightward spreading, which is blocked by both RTR/Lower-VT and RTR/Hi (see ()). As I just argued, top-ranked RTR/Hi neither adds nor detracts from the effect of RTR/Lower-VT on Mj, so the subset criterion for process-specificity is met without much ado.

We have seen, then, that the Northern Palestinian data, far from arguing against OT, support the more restrictive view of process-specificity that ranking gives. To reiterate the point of section 3: a parametric rule-based theory says nothing about which constraints might limit two different processes in the same grammar, whereas OT, because of ranking, insists that the constraints impinging on one process must be a subset of the constraints impinging on another.

Though this is all I will have to say about process-specificity, there is an analytic detail to be settled: what are the actual constraints Mi and Mj in ()? The constraint implicated in long-distance harmony, Mj, is just RTR-Right, as in the Southern Palestinian dialect. The constraint Mi is involved in the local harmony process, which spreads RTR no further than a following (C)V sequence. Unlike long-distance harmony, local harmony processes like this one have been little studied in the phonological literature, before or since OT.10 It is not unexpected, then, that the rule-based local harmony process must fall back on descriptive stipulation, iterating a local harmony process “to a following syllable nucleus” (Davis 1995: 487–8). In OT, though, the stakes are higher; descriptive stipulation is a much less appealing option in an Optimality-Theoretic grammar, where each constraint is universal and, through ranking permutation, will participate in determining a range of typological options.

What is required, then, is a cross-language examination of similar local harmony phenomena — and this is a research project in itself. Until then, we can only speculate as to what constraint fills the role of Mi in (). One possibility is that the responsible constraint requires any RTR span to end on the vowel a. This makes sense because a is exactly where RTR is most salient perceptually, since it produces an obvious distinction between [æ] and [ ].11 Obedience to such a constraint will yield the desired local rightward harmony in forms like Sabaa or a lam, and it is inoffensive to long-distance harmony cases like Sa aaha, where the RTR span also ends on a.

For explicitness, I will formulate this constraint in alignment terms, like RTR-Left and RTR-Right:

() RTR-to-a


Align(RTR, Right, a, Right)

Plugging it into the hierarchy (), together with the other constraints discussed, we obtain a fairly full picture of RTR harmony in the Northern Palestinian dialect:

() Hierarchy for RTR Harmony in Northern Palestinian Arabic


Stay-RTR, RTR-Left >> RTR/Hi >> RTR-to-a >> RTR/Lower-VT >> RTR-Right >> Stay-ATR

The top-ranked constraints can both compel violation of RTR/Hi, which itself is in a position to block RTR-to-a (as in (c)). Satisfaction of RTR-to-a is not limited by lower-ranked RTR/Lower-VT, which nevertheless does control the effect of RTR-Right. At the bottom is the faithfulness constraint Stay-ATR, violated in every case of RTR harmony.

Compare this constraint hierarchy with Southern Palestinian. The gross ranking structure is the same in both dialects, with undominated Stay-RTR and RTR-Left at the top, Stay-ATR at the bottom, and RTR-Right in between. One point of difference is that RTR/Lower-VT dominates RTR-Right in Northern Palestinian, but the ranking must be just the opposite in Southern Palestinian. Domination by RTR/Lower-VT significantly limits the scope of RTR-Right in the Northern dialect, leading to long-distance harmony only over strings of low segments. This is a small example of the inherently typological character of OT (Prince and Smolensky 1993), in which differences in constraint ranking, rather than differences in the presence of a rule or the value of a parameter, are what distinguish languages. Another point of difference is that RTR-to-a is visibly active in Northern Palestinian, because it dominates RTR/Lower-VT. In Southern Palestinian, however, there is no observable activity by RTR-to-a, presumably because it too is ranked below RTR-Right.

The full hierarchy for the Northern dialect is applied to some choice examples in tableaux 5 through 9. Tableau 5 shows a case of local harmony, in which RTR spreads to the next vowel (an a) but no further. Candidate (b), with no harmony whatsoever, fails on RTR-to-a, since the right edge of the RTR span is not on the vowel a. Candidate (c), with long-distance harmony, incurs an avoidable violation of RTR/Lower-VT, because it contains an RTR b that no other candidate has. All of (a–c) have at least one violation of this constraint, though, because of the underlying RTR S. Form (d) avoids this violation, and satisfies all other RTR-controlling constraints to boot, by eliminating the feature entirely, but that violates top-ranked Stay-RTR. (Having made the point, I will not consider candidates like (d) again.) As in Southern Palestinian, undominated No-Gap bars “skipping” configurations like Sabaa .

Tableau 5

/Sabaa / Sabaa (see (a))



/Sabaa /

Stay-

RTR


RTR-

Left


RTR/HI

RTR-

to-a

RTR/

Lower-VT


RTR-

Right


Stay-

ATR


a. Sabaa













*

***

*

b. Sabaa










* !

*

****




c. Sabaa













** !

*

***

d.. sabaa

* !


















Tableau 6 is another case of local harmony, but with inclusion of the intervening n in the RTR span. In (b), n blocks harmony, satisfying RTR/Lower-VT but violating higher-ranked RTR-to-a.



Tableau 6

/Snaaf/ Snaaf (see (b))



/Snaaf/

Stay-

RTR


RTR-

Left


RTR/HI

RTR-

to-a

RTR/

Lower-VT


RTR-

Right


Stay-

ATR


a. Snaaf













**

*

**

b. Snaaf










* !

*

***



Tableau 7 considers the form aTšaan, where rightward harmony is blocked by š. The basis of the blocking effect can be seen by comparing (a) and (b). In the latter, high-ranking RTR/Hi is violated because the high consonant š has assimilated to the RTR feature. This is fatal, because a candidate is available in (a) that avoids the offending segment just by violating lower-ranked RTR-to-a. Incidentally, (c) shows a case where leftward harmony has failed. Candidates like this were comprehensively examined in Southern Palestinian (section 3) and need not be considered further here.



Tableau 7

/ aTšaan/ aTšaan (see (c))



/ aTšaan/

Stay-

RTR


RTR-

Left


RTR/HI

RTR-

to-a

RTR/

Lower-VT


RTR-

Right


Stay-

ATR


a. aTšaan










*

*

***

**

b. aTšaan







* !




**

*

****

c. aTšaan




* ! *




*

*

***



In tableau 8, where the higher-ranked constraints are irrelevant because both candidates satisfy them, we see the pure effect of rightward harmony. Thus, RTR-Right is decisive, just as in Southern Palestinian, but here RTR/Lower-VT ensures that RTR-Right is effective only in sequences of Lower-VT segments.12



Tableau 8

/Sa aaha/ Sa aaha (see (a))



/Sa aaha/

Stay-

RTR


RTR-

Left


RTR/HI

RTR-

to-a

RTR/

Lower-VT


RTR-

Right


Stay-

ATR


a. Sa aaha













*




*****

b. Sa aaha













*

* ! ***

*

Finally, tableau 9 is also analogous to the Southern Palestinian system. RTR/Hi blocks rightward harmony in the expected way, though satisfaction of lower-ranked RTR-to-a could otherwise be achieved.



Tableau 9

/Si a/   Si a (see (b))



/Si a/

Stay-

RTR


RTR-

Left


RTR/Hi

RTR-

to-a

RTR/

Lower-VT


RTR-

Right


Stay-

ATR


a. Si a










*

*

***




b. Si a







* !




**




***

In summary, I have shown that the phonology of RTR harmony in Northern Palestinian Arabic shows a pattern of process-specificity that meets the subset criterion introduced in section 3. A reasonably complete Optimality-Theoretic analysis has been presented, in which the limitation of processes by constraints as well as the processes themselves are obtained from the essential element of OT, constraint ranking.



Process-specificity is the main point of Davis’s article and of this reply. Nonetheless, he concludes by raising a very different challenge against OT. It rests on an observation about how epenthesis influences RTR harmony. As we have seen, i blocks rightward harmony. Blocking occurs whether the i is underlying (as in Tiinak) or epenthetic (as in baTinha from /baTn+ha/).13 The problem, according to Davis (p. 496), is this: “In standard Optimality Theory, ...the site of epenthesis is viewed as an empty syllabic position lacking featural content. The features of a[n epenthetic] vowel are supplied by a later interpretive component...” That is, an epenthetic vowel is supposed to be phonologically featureless, but it acts like any other high vowel in blocking rightward harmony. (Interestingly, Herzallah (1990: 109–110fn., 190f.) argues that epenthetic i in Northern Palestinian does act as phonologically featureless.)

This purported argument against OT rests on a category error, an error that is canonized as “standard Optimality Theory.” As the previous discussion has emphasized, OT is about how grammars are defined by constraint hierarchies; it is not about how to represent epenthetic vowels. Thus, no result about the representation or behavior of epenthetic vowels can possibly bear on the correctness of OT per se. OT simply does not bring with it a commitment to a particular view of vowel epenthesis. How could it, when its fundamental claim is that grammars are defined by constraint hierarchies? This is quite remote from the question of how to represent epenthetic vowels.

Facts about epenthetic vowels will bear on the correctness of any theory of epenthetic vowels, and this truism applies equally well to a theory of epenthetic vowels embedded within OT. One view of epenthesis has considerable support in the literature: epenthesis is a consequence of excessive syllabification, in which otherwise stray segments may be incorporated into syllables with empty onsets or nuclei (Selkirk 1981, Broselow 1982, Piggott and Singh 1985, Itô 1986, 1989; Lowenstamm and Kaye 1986). The phonological values of these empty positions are spelled-out by rules of default feature insertion, which apply extra-systemically (Archangeli 1984: 36; Broselow 1984; Paradis and Prunet 1991; Herzallah 1990) — that is, they are outside the syllabification system, and may be outside the (lexical) phonology proper.14

Some OT work adopts this view of epenthetic segments because it accommodates a purely structural basis for the anti-epenthetic faithfulness constraint. Specifically, the constraint Fill is posited, which militates against empty syllabic positions (Prince and Smolensky 1991, 1993; McCarthy and Prince 1993a). But this treatment of epenthesis is by no means peculiar or intrinsic to OT, since it both predates OT, as we just saw, and is rejected in other OT work (e.g., Smolensky 1993, McCarthy 1993, McCarthy and Prince 1994, 1995), which develops alternative views of faithfulness based on the epenthetic segment’s lack of a morphological affiliation or an underlying correspondent. Davis’s argument is perfectly appropriate for what it tells us about the representation of epenthetic vowels and, by extension, about the constraint Fill, but it is a category mistake to think that this bears on the correctness of OT.

This spurious challenge to OT is one of many category errors that arise in discussions of the theory. Other examples include these: OT requires that deleted segments be present but syllabically unparsed; OT cannot treat floating tones or empty segments; OT is a theory of prosody only, which can’t deal with segmental phonology; OT is incompatible with (or is only compatible with) moraic prosody; OT denies the possibility of distinct lexical and post-lexical phonologies; OT says that morphology can impinge on phonology only at constituent edges, through alignment; OT is inherently non-derivational; OT is a theory of phonology only, without relevance to morphology or syntax; OT has, at its peril, discarded all the insights of feature geometry/skeletal theory/metrical theory/lexical phonology/you-name-it. To repeat the question that anchors this section: how could any of these things be true of necessity, when what OT says is that grammars are defined by constraint hierarchies? These issues have their place in the context of evaluating particular analyses embedded within OT, but there is no such thing as a “standard OT” story on any of them nor, I think, could there ever be one.

Objections to any theory require a response, and sometimes in the responding, a result that was always present but only latent is exposed to view. So it is with Davis’s allegation that OT cannot contend with process-specificity of constraints. We have seen that OT provides a general theory of process-specificity through constraint ranking. Moreover, OT demands that process-specificity meet a subset criterion, whose properties were defined in section 3. The essence of the subset criterion is that similar processes can be ranked for robustness, because the constraints impinging on one must be a subset of the constraints impinging on the other. The OT approach is therefore more restrictive and interesting than alternatives that use rules and parameters, since they make no connection between the constraints on different processes in the same grammar. Furthermore, OT obtains these results from minimal assumptions, consisting of nothing more than the core idea of the whole theory: constraints are ranked.

These results about process-specificity in OT derive empirical support from the Palestinian Arabic material. The Southern Palestinian pattern is straightforwardly compatible with the subset criterion, and the Northern Palestinian pattern is as well, once the constraints involved are properly understood. A hypothetical example that is incompatible with the subset criterion was also discussed, affirming the restrictiveness of the claim OT makes.

Finally, in section 5 I discussed a very different objection against OT, based on a type of category error in which an implementational aspect of some OT analyses is confounded with the essence of the theory itself.




Share with your friends:
1   2   3   4   5   6   7




The database is protected by copyright ©essaydocs.org 2020
send message

    Main page