How does the Comparative Method make falsifiable predictions and how are they tested?

Question

The previous question was closed as predicted despite it being edited to only focus on the method itself. The current question was suggested as a substitute, so lets try this one then.In order to justify whether the comparative method is scientific as is being claimed, it need to be shown how it is scientific by showing what is universally accepted since the 20th century in all sciences, including social: does it make falsifiable predictions that can be empirically tested?

On PIE. It seems so far to me somebody sits somewhere looking at a list of cognates. Then they deduce from the list of cognates the proto word, then they proceed to derive from the proto-word the sound laws by which it changed in the various attested IE languages. Then from the sound laws it is decided which word is older and which is younger. Then it is decided how the word migrated across Indo-Europe even if it contradicts native histories of those lands or the archaeological record of them. If it does contradict the native history or archaeological record, then an auxiliary hypothesis is added to make the main hypothesis coherent with the contradictory data. In this way falsification never happens because any n number of auxiliary hypothesis can be added multiplying unknowns ad-infinitum.

As per the modern and universal scientific method the first move in the steps above deducing the proto-word is a testable hypothesis. The others moves will not be allowed. First the proto word needs to be tested. So now my question, how do you test it? I will present one example of cognate lists and can you show me how you derive a testable theory from it and the how do you go about testing it in your discipline.

The word for four: Gothic: fidwor; Latin quattuor; Ancient Greek: téssares Sanskrit: catvā́ras Avestan: čaϑwārō OCS: četyre; OPrus: ketturei; OIr: ceth(a)ir Armenian: čʿorkʿ; Albanian: katër; Tochariyan A: śtwar

Now please walk me through all the steps by applying your method how you derive the proto word and how you test your hypothesis.

Note: I am recording all these conversations as source material for future papers on the topic of Epistemology of the Comparative Method in Historical Linguistics.

Answer 1

On PIE. It seems so far to me somebody sits somewhere looking at a list of cognates.

Yes.

Then they deduce from the list of cognates the proto word, then they proceed to derive from the proto-word the sound laws by which it changed in the various attested IE languages.

Other way around. The point of the comparative method is to figure out regular correspondences between cognates in different languages, that apply across a wide variety of different words (ideally the entire inherited lexicon). Then from those you can reconstruct a proto-form.

The fact that these correspondences need to be regular is what makes them falsifiable. "In inherited cognates, Latin /p/ corresponds to Germanic /f/" is a testable hypothesis: you can test it by finding some more Germanic and Latin words and seeing if they correspond (pater~father, piscis~fish, etc).

If you test this hypothesis enough, you'll find that it is, in fact, wrong: Latin spuō ~ English "spew", Latin septem ~ English "seven" (and Gothic sibun), etc. So we revise the hypothesis: Latin /p/ corresponds to Germanic /f/ with a couple specific exceptions (not when it's preceded by /s/, for example, or not when it's preceded by an unstressed syllable). And this version of the hypothesis has held up extremely well to scrutiny, and has reinforced connections with other languages—for example, the "not when it's preceded by an unstressed syllable" part suggests that some predecessor of Proto-Germanic had the same stress patterns as Sanskrit.

Then from the sound laws it is decided which word is older and which is younger.

Not generally, no. It doesn't really make sense to talk about whether English or German is "older", for example: they're both modern, living languages. But we can say that Classical Latin is "older" than modern English because the type of Latin we're talking about comes from thousands of years ago. This isn't the comparative method, it's just history/archaeology.

Then it is decided how the word migrated across Indo-Europe even if it contradicts native histories of those lands or the archaeological record of them.

Some historical linguists try to use the comparative method to figure out where a proto-language would have been spoken, but this isn't the main purpose of the method, and there's disagreement on how useful this actually is. Look at all the controversy about the PIE urheimat.

If it does contradict the native history or archaeological record, then an auxiliary hypothesis is added to make the main hypothesis coherent with the contradictory data.

Yes, that's how the scientific method works. If the evidence contradicts your hypothesis, you alter the hypothesis.

In this way falsification never happens because any n number of auxiliary hypothesis can be added multiplying unknowns ad-infinitum.

Falsification doesn't generally mean giving up on an entire discipline because of contradictory evidence. Falsification instead generally means that the hypothesis needs to be altered. The scientific method is an incremental process.

Answer 2

The question appears to carry misconceptions over "the scientific method", in particular favoring a popularized interpretation of Popperian epistemology. The valid essence of the "falsifiable predictions" part is that the theory makes objective logical claims as to "what is" versus "what is not". That distinguishes science from various schools of literary criticism where there are no existential claims, there are simply ways of "interpreting" a text. It is not always possible to currently verify all scientific claims, owing to various technological limitations. Various physical theories such as string theory are not unscientific because they predict the existence of magnetic monopoles yet we cannot prove that magnetic monopoles failed to be observed in a circumstance where they are predicted.

Second point, the comparative method is not a theory, it is a method. The method can be used to do some things. Sticking a thermometer in a pot of fluid is a method, not a theory – you have to have a theory that tells you what the method produces. This, incidentally, is a basic problem with Popper's emphasis on disproof, that the method of disproving is valid only if the underlying theory of method is valid, which by his account cannot be proven, it can only be disproven.

Now on to what the comparative method is or does. It is very similar to comparative anatomy, an even older science, indeed the logical differences between comparative anatomy and linguistics are minor. They distinguish homologous structures from analogous structures: the former are the structures from a common ancestor that correspond to inheritance from a proto-language in linguistics. Linguistics and biology both have very similar theories of "common ancestry" and "change".

The theory that languages change over time is directly validated in many ways (there are actual records of languages from before the present generation, and we can observe that languages have indeed changed). There are differences between the level of validation possible in biology vs. linguistics, since we can induce multi-generational changes in fast-reproducing species, but we are forbidden to experimentally induce changes in languages (pesky IRB problems) which would take hundreds of years to result in "different languages", but that doesn't make the underlying theory that languages change over time "unscientific".

As for how reconstructions are done, it is important to recall that scientific hypothesis formation and testing is an iterated process – an initial hypothesis is advanced, consequences of the hypothesis are adduced and compared with observation, then the hypothesis may be adjusted depending on how well the initial hypothesis matches observation. We do not, therefore, reconstruct "4" from a handful of descendant words as a single isolated inference. Thus we originally did not distinguish e,o from a in reconstructions of PIE because Indo-Iranian did not preserve that distinction; and then it was realized that Indo-Iranian had a vowel neutralization. Now, it is true that linguistics suffers from a technological impediment that we cannot travel back in time to directly observe PIE as spoken in the year 5000 BP. Comparative anatomy suffers from that limitation to some extent, since soft tissues are not well preserved in the fossil record. Once in a while, we uncover linguisitic fossils, such as the laryngeals of Anatolian which instantiate hypothesized "sonorant coefficients" hypothesized by Saussure using the comparative method.