Names of places in Croatia

[FlatAssembler]

Then what is "ad-hoc hypothesis"? It is inventing reasons why an experiment wouldn't work. Saying, in response to my paper, "Maybe the collision entropy of the nouns in the Croatian language is significantly lower than the collision entropy of all the words in the Aspell word-list.", seems like an obvious ad-hoc hypothesis. What magic would make it do that? How do you know it is not higher, rather than lower? It is not an ad-hoc hypothesis if you said that about the Swahili language, as, because of the noun classes, some two-consonant prefixes which are possible for verbs aren't possible for nouns, so it seems plausible that the collision entropy of consonant pairs is indeed lower in nouns than in all Swahili words. Suggesting that for a Swahili-like language is not baseless, but saying that for Croatian seems like a baseless ad-hoc hypothesis.

No. An ad hoc hypothesis is inventing reasons to render an hypothesis unfalsifiable. Like this:

Hypothesis: Bigfoot exist.

Criticism: Show me one.

Ad hoc #1: Can’t - they can render themselves invisible.

Criticism: Well, then show me evidence - fur, scat, bones.

Ad hoc #2: No can do - Bigfoot are transdimensional beings and are able to make their evidence undetectable to humans.

Criticism: Photographs, videos, audio recordings?

Ad hoc #3: Those are all government fakes to hide the truth.

And so on. Ad hocs are meant support an hypothesis, not disprove it.

Boru

I think it is essentially the same thing. The mainstream linguistics considers the river names that start with k-r (Krka, Korana, Kravarščica, Krapina, Krbavica, Kravica, two rivers named Karašica) to be unrelated, that this k-r pattern is accidental. My paper says: "That would imply that there is a significant probability of that k-r pattern occurring by chance. There isn't. Here is a simple experiment that shows that probability has to be somewhere between 1/300 and 1/17...". Neuralbeans says, in response to that, "Maybe that experiment is flawed because the collision entropy of the nouns is significantly lower than the collision entropy of all the words in the Croatian language.". Neuralbeans is essentially modifying mainstream linguistics to make it harder to falsify it. Measuring the collision entropy of only the nouns is a lot harder than simply measuring the collision entropy of all words in Aspell word-list.