I’ve gotten a handful of replies to my query earlier this week about the meaning of “A counterbleeds B”, and I’ve also polled a few people personally. Given the nearly-even split in the replies (4 for my bet-losing definition, 5 for Colin’s bet-winning one), I’m confident that there’s something weird going on with this “counter-” morpheme, whatever it is.

Just to set the record straight from the beginning: I’m not on some sort of mission to change this terminology. (I also don’t care about my $10; I lost ’em fair and square.) Sure, the concepts are complex enough on their own without the difficulty with the terminology, but I’m not convinced the concepts would be made any easier to understand if the terminology were unambiguous. Some people simply internalize it one way (in terms of the example I gave, “Lengthening counterbleeds Devoicing”) and others the other way (“Devoicing counterbleeds Lengthening”), but the ordering relation itself (“Lengthening and Devoicing are in a counterbleeding relationship”) is not under dispute, and that’s really all that matters.

Besides, this issue just doesn’t seem to come up all that often, if at all. For all I know, Colin and I are the first to notice that there was something to notice here. Sure, I’ve only discussed this with fewer than a dozen people and phonoloblog has not (yet) reached all phonologists, but nobody’s written to tell me that this issue has been the topic of discussion somewhere, whether in writing or between some stumbling-drunk phonologists at a party (or whatever).

OK, so are you ready to find out the answer? Read on.

Colin’s bet-winning definition was that the rule that would have bled is also the one that counterbleeds. (Similarly, the rule that would have fed is also the one that counterfeeds; the terms are perfectly parallel in this respect.) So, in terms of my example, “Devoicing counterbleeds Lengthening” (and, mutatis mutandis, “Devoicing counterfeeds Shortening”). The source we agreed on as arbiter was Koutsoudas et al. 1974 — mostly because that’s what we had available — and on p. 2 (which I’ve cited on phonoloblog twice before!) it says:

One e-mail respondent (Lev Blumenfeld) “found this very clear statement on p. 97” of Kenstowicz (1994), Phonology in Generative Grammar (and correctly assumes that the same goes for the parallel ‘counterbleed[s]’):

‘counterfeed[s]’ means ‘fails to feed’

This “counter-x = fails to x” way to put it is also what two other respondents (Kie Zuraw and Gunnar Hansson) independently volunteered as their reasoning in favor of this definition.

Three respondents who chose this definition (Lev, Kie, and Nathan Sanders) also pointed out that when two rules are in a counterbleeding relationship, the one ordered first is not (necessarily) doing anything to the second; the interest is in what the second would do to the first if the order were the reverse of what it is. (I add the “necessarily” here to account for more complex interactions, such as between height-parasitic Harmony and Lowering in Yawelmani/Yowlumne Yokuts.)

My bet-losing definition was that the rule that comes first in the ordering is always the agent/subject of the verb, whether the verb is feeds, bleeds, counterfeeds, or counterbleeds. When I was forced to think about why this reasoning should apply to the ambiguous “counter-” cases, I thought of a “preemptive strike” sort of interpretation: in terms of my example again, Devoicing would bleed Lengthening, so Lengthening preemptively applies to thwart the bleeding. (Bob Kennedy, Geoff Nathan, Sharon Rose, and “C Callosum” all agreed with this definition; C Callosum also volunteered a version of the “preemptive strike” argument.)

The problem with this reasoning is that Lengthening doesn’t itself bleed Devoicing; the prior application of Lengthening prevents the Devoicing from bleeding Lengthening. This is inconsistent with the definition of “counter-” in words like “counterpunch”, “counterstrike”, “counterattack”, etc. Even though you can counterpunch without having been successfully punched yourself — say, because you block the punch — if you counterpunch, then you necessarily punch. (Note that you can counter a punch with a kick, but that wouldn’t be a counterpunch.) So, if A counterbleeds B, then A should bleed B — but it doesn’t (at least, not necessarily). (Thanks to Strang Burton for discussing this with me.)

Google searches for the unambiguous terms “counterbleeding” and “counterfeeding” results in about 200 hits each (your results may vary). Searches for “counterbleeds” and “counterfeeds”, by contrast, results in fewer than 40 hits each. The use of these terms is also divided there, though perhaps skewed a bit more in the direction of Colin’s correct definition. My guess at this point is that the terminology was used as Colin defines it for some time after the terms were invented, and that it got muddy somewhere along the way. I’m going to sort of passively collect a set of references and post a summary at some point later this year; if anyone else wants to help me in this almost completely pointless project, write to me (phonoloblog#gmail|com) with the following information:

• The source and page number (plus link to an electronic copy, if available).

• A paraphrase of the two rules involved in the counterbleeding or counterfeeding relationship.

• Whether the author writes “Preceding Rule counter{bl/f}eeds Following Rule” or “Following Rule counter{bl/f}eeds Preceding Rule”.

Examples of references in which only the safe, unambiguous terms “counterbleeding” and “counterfeeding” are used are also welcome. For example, I found that McCarthy (1999) “Sympathy and Phonological Opacity” (in Phonology) only uses these -ing forms.