Philip makes two separate claims in his recent confessions post, conveniently juxtaposed in one sentence.

1. […] any definitions offered for the putative verbs counterfeed and counterbleed in any phonology textbook are arbitrary stipulations by the authors themselves

2. and are not grounded in any actual facts of phonology […]

On the first claim, I agree — at least insofar as Philip means that the statement A counterXeeds B (where X = bl or f) refers to situations where A would Xeed B for some authors and B would Xeed A for other authors (were the order of A and B reversed). (I have yet to find examples where someone arbitrarily defines the general concept of a counterXeeding relationship, but see note 1.)

On the second claim, though, I have to disagree. For example, I disagree with Philip when he writes of counterXeeding:

The class of counterfeeding [and presumably also counterbleeding?–EB] rule relationships does not identify any known natural class of phonological objects.

I’m not sure what Philip means by “phonological objects”, but this strikes me as simply untrue. Counterfeeding, for example, encompasses the description of chain shifts, which are well-known phonological occurrences. I’m sure Philip doesn’t intend to say that chain shifts don’t exist, but it’s not otherwise clear to me what he does intend to say with the above. (See note 2.)

Just before the above, Philip writes:

Responsible phonologists should not use these terms. […] I really don’t know why one needs these terms, and I wouldn’t bother to teach them to students.

I disagree with this, too, mainly because I think there was some serious progress made in our understanding of phonological process interactions in the 1960s and 1970s that has been sadly ignored in a lot of mainstream work since the representational boom(s) of the 1980s. Perhaps ironically, the overall framework shift from Derivational Theory (DT) to Optimality Theory (OT) in the 1990s has given us a unique opportunity to revisit that older literature to see where we stand with respect to the progress that was made, given that (one of) the fundamental distinction(s) between DT and OT is in how process interaction is understood (i.e., rule ordering in DT vs. constraint ranking in OT). In order to read that literature and understand any results from it, it’s important to know what the terms are and what they mean. In order to gauge what those results mean for the shift in frameworks, it’s important to understand what the concepts are and what about them is and is not independent of the larger framework of DT assumptions in which those results were couched.

---

And now, Part the First of a Short History of Feeding and Bleeding, Being an Attempt to Clarify the Importance of These and Related Notions for Phonological Theory and the Like.

Philip’s right that the concepts/terms feeding and bleeding originate in the work of Paul Kiparsky; specifically, they appear to have been birthed in his 1968 paper “Linguistic Universals and Linguistic Change” (in Universals in Linguistic Theory, eds. Bach and Harms, pp. 170-202). Here is the first relevant passage (pp. 196-197, bold emphasis added):

One way in which two rules, A and B, can be functionally related is that the application of A creates representations to which B is applicable. That is, the application of A converts forms to which B cannot apply into forms to which B can apply; schematically:

A. [   ] > [φ] B. [φ] > [   ]

[…] In such a situation, call A a feeding rule relative to B […]. Call this relationship between rules a feeding relationship […] and the linear order in which the feeding rule precedes a feeding order […].

In other words, A and B are in a feeding relationship regardless of their order, but they are only in a feeding order if A precedes (and feeds) B. This is somewhat different than how the terms are used now, of course — nobody (AFAIK) uses feeding relationship to refer to the above rules in the B-precedes-A order; that’s a counterfeeding relationship.

Here’s the second relevant passage, about bleeding (p. 198, bold emphasis again added):

Another possible functional relationship between two rules is that A removes representations to which B would otherwise apply:

A. [   ] > [~φ] B. [φ] > [   ]

[…] Call A a bleeding rule relative to B, the relationship between A and B a bleeding relationship, and the ordering in which A precedes B a bleeding order.

This is again somewhat different than how the terms are used now, as noted above, mutatis mutandis.

Kiparsky’s point in this second-to-last section of his paper is the claim that linguistic change via rule reordering follows two principles:

I. Feeding order tends to be maximized. (p. 197)
II. Bleeding order tends to be minimized. (p. 199)

Note that there is no direct reference to the opposites of Xeeding orders; no counterXeeding, no nothin’ (see note 3). Kiparsky’s discussion of how the principles above are unified, however, suggests some terminology (p. 200):

There is a more general principle underlying the two reordering tendencies (I and II) which combines them under a single wider concept of fuller utilization and makes their nature intuitively much clearer:

III. Rules tend to shift into the order which allows their fullest utilization in the grammar.

If I am right that such a principle determines the direction in which reordering proceeds, then it follows that the order toward which rules gravitate in this way is linguistically simpler than its opposite. It is hard to see what other explanation there could be for such a consistent tendency toward a specific kind of order in linguistic change. As a convenient designation for the order types which are shunned and preferred according to Principles I-III, I suggest marked and unmarked order, respectively.

OK, so he doesn’t directly say it, but it seems that what he means here is the following: feeding order, being maximized and therefore preferred, is unmarked feeding order; bleeding order, being minimized and therefore shunned, is marked bleeding order. What we’re used to thinking of as counterfeeding is thus marked feeding order, and counterbleeding is unmarked bleeding order. Confused? Maybe the following table will help.

Kiparsky 1968 rule order typology	unmarked	marked
feeding relationship	A. [   ] > [φ] B. [φ] > [   ] (= “feeding”)	B. [φ] > [   ] A. [   ] > [φ] (= “counterfeeding”)
bleeding relationship	A. [   ] > [~φ] B. [φ] > [   ] (= counterbleeding”)	B. [φ] > [   ] A. [   ] > [~φ] (= “bleeding”)

Still confused? Sorry, can’t really help much more. The most confusing aspect of it, I think, is that we’re used to thinking of opaque rule orders (counterfeeding and counterbleeding) as constituting a natural class — and based on Kiparsky’s later work, no less. They are clearly not a natural class in this table; one is marked and the other is unmarked.

To be continued …

---

Notes

1. I have so far found one example in a published work (though not in a textbook) where I believe the concept of counterbleeding is confused — but with very good reason, because the situation is not one of the simpler cases upon which the definitions of all these terms is usually based.
       Moreton & Smolensky (2002) state that “epenthesis counterbleeds a consonant-deletion rule in Turkish” (p. 10). Their source is Sprouse (1997), who writes: “In Turkish an epenthetic vowel is required to break up certain disallowed coda consonants clusters. Yet this vowel sometimes triggers the deletion of one of the offending consonants” (p. 3). Epenthesis thus feeds deletion. While it’s true that the operation of deletion removes part of the context that made epenthesis applicable in the first place — making the rule interaction opaque — it’s not technically counterbleeding because if you reverse the order of deletion and epenthesis, it’s not the case that deletion would bleed epenthesis. Deletion simply would not apply to the relevant forms because it wouldn’t be fed by epenthesis. (I’m assuming, of course, that we all agree that “reverse the rules and there would be bleeding” is the correct definition of counterbleeding; if there’s another definition out there that would categorize this case as an example of counterbleeding, I’d love to hear about it.)
       In general, the relationships between rules can differ greatly when they affect each other’s foci vs. when they affect each other’s contexts; for some discussion of the relevant distinction, see the Kenstowicz & Kisseberth article cited in note 3 further below. [back]

2. I should probably note at this point that Philip and I go way back — he was a beginning grad student at UC Santa Cruz when I was still an undergrad there, we took grad classes and TA’d undergrad classes together, etc. I would say we have highly similar training in most respects, save perhaps for the fact that I probably made it to my undergraduate phonology class more often because the weather for me in Santa Cruz was better than he describes it being for him at Amherst. (This difference probably does not account for our difference in appreciation of counterXeeding; I confess that I didn’t really understand the terms until much later, after reading more of the relevant literature and eventually teaching phonology classes myself.) [back]

3. I still haven’t found the undeniably original source for the counterXeeding terms. In a paper written in response to this paper by Kiparsky, Kenstowicz and Kisseberth refer to the opposite of a Xeeding order as a non-Xeeding order, but this clearly did not catch on. (The paper is “Unmarked Bleeding Orders”, Studies in the Linguistic Sciences 1.1, 1971, pp. 8-28; reprinted in Studies in Generative Phonology, ed. by Kisseberth, 1973, pp. 1-12.) [back]