Croatian toponyms

[FlatAssembler]

The sections there about semantic flexibility are also important-- you'll notice many Indo-European correspondences are not to words that mean exactly the same thing.

In my calculations, I assumed I allowed myself a semantic drift of 3 words. I think that's reasonable here. It's not like I know thousands of PAN and PIE words to allow myself the semantic drift of 20.

[Daniel]

semantic drift of 3 words.

I'm not sure what you mean. Semantic drift refers to change in the meaning of words and would be measured by the amount that they have changed. For example, "arm" might become "hand" or maybe eventually "foot". The problem with mass comparison at large time depths is that you no longer know which words to compare because you don't know if you should compare "arm" to "arm" or "hand" or "foot". (See the linked article about chance correspondences for more.)

[FlatAssembler]

semantic drift of 3 words.

I'm not sure what you mean. Semantic drift refers to change in the meaning of words and would be measured by the amount that they have changed. For example, "arm" might become "hand" or maybe eventually "foot". The problem with mass comparison at large time depths is that you no longer know which words to compare because you don't know if you should compare "arm" to "arm" or "hand" or "foot". (See the linked article about chance correspondences for more.)

I meant, I couldn't, for example, compare the PAN word for hand to the PIE word for arm, because I don't know the PIE word for arm (if it existed). I could compare it to the PIE words for "to grab" or "five", and those are the only PIE words I know that could plausibly shift its meaning to "hand". So, I can compare a PAN word with only three (hand, grab, five) PIE words. The article you linked me to complains that those who do pseudoscientific language comparisons often compare one word with as many as 20 words in another language, and of course they would then find an apparent cognate for most of the words.

[Daniel]

The article you linked me to complains that those who do pseudoscientific language comparisons often compare one word with as many as 20 words in another language, and of course they would then find an apparent cognate for most of the words.

But the problem is that you actually do need to consider 20 or more possible words as potential cognates. "Arm" might become "hand" or "foot" or so many other things. Just consider how "foot" became a measurement of distance! Or how "nice" used to mean stupid/foolish, and now means (more or less) "good". And how sometimes "bad" can mean "good" (as in "that's so bad [ass]!").

What makes cognates convincing is when they share multiple corresponding sounds consistently derived from identifiable sound changes. (And that rarely relies on meaning at all, except as a way to guess which words might be cognates-- after a while, though, you can start identifying them based on sounds alone, and then infer how the meanings changed.) Short of that, any 'patterns' you find are probably just coincidental.

It's the regularity of sound change that makes it so helpful in determining linguistic relationships. Find patterns, then figure out how they developed. Not just correspondences.

[FlatAssembler]

I've always thought the patterns of changes in meaning were derived from modern languages. We know that "foot" sometimes comes from the word "to walk", because of the Croatian word "stopalo" (foot) obviously coming from the word "stopati" (to walk). Or the word for "spring" sometimes coming from the word for "to boil", because of the Croatian word "vrelo" (spring) obviously coming from "vreti" (to boil). Or the word "book" coming from the word for a tree because of the Latin word "liber" meaning both. You get the idea. The problem with the toponyms is that folk-etymologies are so common, that, if a toponym doesn't appear to be descriptive, it's most likely a folk-etymology.

[FlatAssembler]

So, you think that "Bellum a nulla re bella." is usually not a good principle?

[Daniel]

What do you mean?

[FlatAssembler]

I mean, words in the same language that sound similar, but mean something completely different (like Latin "bellum" and "bella"), usually don't share the same root.
Do you think that this would be a valid reasoning: "German has the word 'Bach' meaning 'stream'. Its Indo-European root could be *bheh2kj. Its Croatian reflex would be 'bos'. There is a word 'bos' in Croatian. It actually means 'barefoot'. Therefore, a semantic shift occurred from 'stream' to 'barefoot'"?

[Daniel]

Sometimes coincidences are just coincidences. Other times it's more than that.

An interesting question (I'm not sure about) is what is considered reasonable evidence in the field of toponymy. In general for etymologies I think what I've seen is that the most explanatory/plausible/consistent theory is generally accepted among existing theories. That is, the best current guess is tentatively accepted as correct, until additional information suggests otherwise. But it's really hard to be confident in something like this.

The problem is, as I said before, that there is no way to get a larger statistical sample than one: maybe it's etymology A, maybe it's etymology B, and there is (in general) no direct evidence for either, and if both are consistent with the data then no clear reason to reject one over the other, and most importantly, no way to check in the end if you're right or not-- because every etymology is independent. Unless you find patterns (like most cities in an area being named after rivers, which still isn't a "rule" just a trend), you can't support it.

A folk etymology is a good theory that has been shown to be wrong. Your plausible theories may be folk etymologies (yes, more technical, but still, appealing, plausible stories!) that have not yet been shown to be wrong. Or they might be right. I'm not sure how to tell.

So, no, resemblance is not evidence for (or against) a particular etymology. Your "therefore" (in the last post) is not logical-- more like "maybe". (Again, I'm not saying you're wrong, but I also don't know how to confidently show that you're right either.)

[FlatAssembler]

I've always thought folk etymologies were easily recognised by not making sense logically. One typical example is the village name "Crna Vlast" (black power). It takes a very far-fetched story to explain why a village would be called "black power". And, from historical documents, we know its original name was "Crna Vas" (black village). The word "vas" later became an archaism, so it was replaced by a similar-sounding more familiar word. Or the river name "Vuka" (she-wolf). It takes a very far-fetched story to explain why a river would be called "she-wolf", and, if you assume its name comes from some earlier Indo-European language, the explanation is obvious. Its from *welk (to flow), and Croats replaced it with a similar-sounding Croatian word. But if a place with ancient thermae was called "springs", that's probably not a folk etymology.

Etymology sometimes makes (at least theoretically) falsifiable predictions. An example I could think of is the village name "Sopje". If I claim that it comes from the Illyrian word "Salapia", meaning "salt lake", which I think would actually give "Sopje" by the sound changes from Old Church Slavonic to modern Croatian, you could prove it wrong by showing there was never a salt lake near that.

[Daniel]

Folk etymologies can be plausible or implausible but actually often seem right intuitively so that's why they catch on. Generally they're made by non-experts guessing (and yes sometimes can be easily shown to be false) but often they are clever or apparently correct.

Yes, falsifiability is important and can sometimes be applied usefully but there are still many possibilities that cannot be definitively decided in that way (when evidence is lacking as with ancient toponyms).

[FlatAssembler]

Well, true, there are folk etymologies that make sense logically, but those are rare. If you replace an unknown element with a similar-sounding one, chances are, that new element wouldn't make sense logically. The only counter-example I can think of is the folk-etymology of Poreč being derived from "porječje" (river bank), but it takes only a basic knowledge of Croatian dialectology to understand why it can't be correct ("porječje" would be pronounced "poriče" in the Istrian dialect). Also, we know its ancient name, it was called "Parentium" in antiquity. Though, for all we know, "Parentium" could have meant "river bank" in Illyrian. You know, from PIE *por-h1ey-nt-y-om. Though this etymology is quite shaky. Given the other toponyms, it would probably give, by regular sound changes, something like *porintium or *parintium, and not "Parentium".

[FlatAssembler]

So, what's your strategy for recognizing folk etymologies? I don't see any alternative except asking if it's too semantically far-fetched. You usually don't have enough knowledge to even tell if it makes sense grammatically.

[Daniel]

As I've said, many folk etymologies seem completely plausible. Sometimes having enough specialized knowledge (or even just looking at basic facts) can show them to be false (e.g., earlier usage of the word in recorded texts, or inconsistency with known sound changes). But the only general way to know something is a false/folk etymology is to identify a better one that is more consistent with known information.

Some folk etymologies are laughably silly and easy to rule out. Many others are much more plausible. Some I'm sure are generally accepted today because we don't have better explanations for them. I suppose a "very good folk etymology" could be called something else (e.g., "linguist's theory"), but that difference is not substantive.

I don't see any alternative except asking if it's too semantically far-fetched.

Newtonian physics is wrong. It certainly was convincing for a while.

You usually don't have enough knowledge to even tell if it makes sense grammatically.

Well, you should, if you're trying to evaluate these as an expert. Physics is hard, but people still study it. We use whatever information we have to do as well as we can.

[FlatAssembler]

A question here is what counts as an explanation. Someone who claims to be able to systematically "explain" all the Croatian toponyms (and there are such people, most of them knowing almost nothing about linguistics) is more than likely a pseudoscientist. I think that valid explanations include demonstrating that the same or similar element reappears in some (descriptive) meaning or tracing the elements back to a proto-language, ideally both.

Well, you should, if you're trying to evaluate these as an expert. Physics is hard, but people still study it. We use whatever information we have to do as well as we can.

But the problem is that "experts" on Illyrian languages agree on almost nothing about the Illyrian grammar or phonology. They don't agree even on whether it was a centum or a satem language. And the things they agree on appear completely baseless to me. For instance, most of the "experts" agree that PIE *bh turned to *b in Illyrian. And, if so, how exactly would you explain away the element *puk in the Croatian hydronyms (if not from *bhogj)?