Suboptimist
I’m not quite a pessimist. I’m more of a suboptimist.

I should clarify that there was actually no organized movement against the comparative method called the Dialectologists. Really, I just like the idea of a showdown and I know a certain clergyman who will only read things with ‘versus’ in the title. You should pay attention to what I am saying, since it turns out that I do know what I am talking about, or at least I know enough of what I am talking about to nail my recent historical linguistics test and comparative method assignments.

Anyway, in the last installment of this discussion, I explained briefly what the comparative method is and laid out basic assumptions of it. I also brought up what is often called the Neogrammarian Manifesto, sound laws suffer no exceptions. I left off stating that *k > ∫ is recognized as a regular sound change (as well as a change we see commonly in other languages).

A challenge came from some scholars, many of whom happened to be dialectologists, who didn’t buy the position that sound change is regular and exceptionless. They also opposed the family tree model. They also had their own slogan, each word has its own history. The alternative to regularity hypothesis and the family tree model was the ‘wave theory.’

Stated simply, the wave model assumed that language change was the result of borrowing from languages and dialects. These changes spread outward, like waves, and become weaker the further away from the innovation’s origin. A word’s history could be the result of influence from any direction, so in theory each word could have a history different from any other word. When we take this approach, historical linguistics is reduced to a study of etymologies.

Proponents of the wave model cited an example from the French dialects of Normandy. Proto-romance *k > ∫ before front vowels and a in [standard] French. However, in parts of Normandy, a bunch of words retain the k. The words that did this were (taken from Campbell 2004:213):

chaine < catena 'chain'
chambre < camera 'room'
champ < campus 'field'
Chandeleur < Candelamas (church candle) < candela
chandelle < candela 'candel'
chanson < cantio(n-) 'song'
chanter < cantare 'to sing'
chat < cattu 'cat'

Opponents of the regularity hypothesis took this as a contradiction to the hypothesis. Words like ‘cat’ were more resistant to the *k > ∫ change, while prestigious words like those associated with the church more readily accepted the *k > ∫ change.

Here is the thing: we can’t recognize these words as exceptions to the regular *k > ∫ change if we don’t recognize the *k > ∫ sound change as regular. Still, these are indeed exceptions to the strict exceptionlessness of sound change. Clearly both are needed, and now I don’t know what else to say.

Questions? Comments? Concerns?

I’ve become very interested in competing models of linguistic change in history, most notably the Neogrammarians and, well, everyone else, most notably dialectologists. The Neogrammarians argued that sound change was regular, and held this view very dogmatically as noted in their slogan (’Sound laws suffer no exceptions‘). Hell, they even had their own manifesto. Dialectologists typically thought this was crap. Before I can explain further, there are a few assumptions of the comparative method (which I should also explain), which was based on the the noted slogan of the Neogrammarians, need to be discussed.

First, to explain the comparative method. In short, it is a method whereby two or more genetically related languages are analyzed side by side to recover it’s history and hypothetical ancestor language (called a proto-language). The steps in the process are:

1: Assemble cognate sets (and make sure that what you have assembled as cognates are indeed cognates). A good example is Italian[k]apra Spanish [k]abra Portuguese [k]abra French [∫]evre ‘goat’.

2: Assemble sound correspondences. Here, we’re interested in the initial sounds: k- : k- : k- : ∫-

3: Reconstruct proto-sound. Here you need to keep in mind a lot of principles, but now just take my word for it that we should reconstruct *k.

4: Determine status of overlapping data (irrelevant here).

5: Check for plausibility of reconstruction based on universals, phonological inventory, and typological expectations. There isn’t anything relevant to the example above, but it helps in other situations. For example, if a language has b d and g, it will damn near always have p t and k (though not necessarily the other way around).

Now the only way our reconstruction can be accurate is if a few assumptions are meant (though we can’t factually assume they are). In fact, Lyle Campbell (2004: 164) said “What textbooks typically call basic assumptions are really consequences of how we reconstruct and of our views of sound change.” The assumptions are:

1: The proto-language was uniform with no dialect variation. This of course is counterfactual, since every language has variation. It isn’t so much that that the method assumes this, but there is nothing built into the method to deal with this

2: Language splits are sudden. Like, SMASH! BOOM! POW! New languages!

3: After the split up of the proto-language, there is no contact between the two languages. Again, not so much of an assumption, but there is no machinery in the comparative method to deal with things like borrowing.

4: Sound change is regular. To use the example above, we see *k > ∫ in a lot of languages and occurs commonly throughout Romance. This is going to be the most important point in part 2 of this discussion.

Stay tuned, and until next time, watch this.