A blog discovery today, that I however find a bit too tangentially related to add to my main histling blogroll: The Genealogical World of Phylogenetic Networks. They mostly discuss phylogenetics in general, with most examples drawn from biology — but for once, linguistics also has its share, covered by Johann-Mattis List (who’s also the author of some interesting papers on the same topic). A few example posts, and some comments of mine:
This highlights something I’ve noted for a while now: there seems to be no properly illustrative generally accepted way of presenting wave theory models of language history. Most variations are indeed pretty much ahistorical (the local Uralic studies iteration of this would be Salminen’s “ball graphs” ). I’m even aware of at least some related models that explicitly admit this fact and do not claim to represent language change, such as van Driem’s “fallen leaves” model.  I’m not sure though if I would call the theory itself ahistorical as much as often poorly presented … but this might not be the time to defend the concept in detail.
This focuses on the inexactness of current linguistic terminology regarding the relatedness of words. We largely make do with a single term ‘cognate’, but there does not quite seem to be consensus regarding if this means 1) words related only thru inheritance, or also words related thru loaning; and 2) only words with the same meaning, or also words that have changed their meaning along the way.
List’s proposed amendments don’t particularly resonate with me, though. I for one am happy to extend the word ‘cognate’ also to words related in a way that includes loaning. This is especially since I think it’s quite difficult to rule out entirely the possibility of loaning having occurred during a word’s transmission. Any linguistic innovation starts at a point, and diffuses across its speaker community effectively by loaning. Since the language/dialect distinction is somewhat arbitrary (more sociolinguistic than objective), it follows that distinguishing loaning from inheritance is somewhat arbitrary as well.
For example: in various other languages of the world we can easily call jazz a loanword from English. But shall we also call it (at least in the sense of the music genre) a loanword from American English in British or Australian English? Or a loanword from Chicago English in New York English? More relevantly, supposing that a 30th-century linguist examines this word’s descendants across the Anglic languages of their times, would they be justified calling the American and Australian words cognate — or should that label be reserved only for words deriving from pre-colonial Early Modern English? And finally: how well do we know that our alleged “cognates proper” deriving from prehistorical protolanguages do not have this kind of a background?
There’s also the fact that in a certain fundamental sense, all linguistic transmission from one generation to the next can be said to be comprised of loaning, though usually in a fashion faithful enough that we categorize it off as “inheritance”.
Similarly I do not think the lack of emphasis on semantic identity is a major flaw. Again, while clear-cut cases of words with dissimilar meaning exist, it’s a matter of taste where to decree that two words now do have “the same” meaning. Is the fact that English head and Finnish pää ‘head’ are used in several different idiomatic ways an obstable to claiming that they’d have the same meaning? Or the fact that pää in some usages is better translated as e.g. ‘end’, while head in some usages is better translated as e.g. johtaja? Or, even if we supposed two words in two modern language varieties to be completely identical in meaning: would we have to again count them semantically non-homologous if we could show that their common proto-form had also some additional meaning that was in both languages later lost?
I do agree that having more detailed terminology would be helpful… but I’d perhaps start adding it from the other end. Since it’s loaning and semantic change that are visible, positively identifiable processes, it might be more productive to have terms denoting “cognate via loaning” and “cognate but semantically changed” specifically, rather than attempting to circumscribe the negative categories “appears cognate purely by inheritance” and “appears cognate without any semantic change”.
 As in e.g. Salminen, Tapani (1999): Euroopan kielet muinoin ja nykyisin. In: Fogelberg, Paul (ed.): Pohjan poluilla. Suomalaisten juuret nykytutkimuksen mukaan. Helsinki: Suomen Tiedeseura.
 Outlined in van Driem, George (2001): Languages of the Himalayas: An Ethnolinguistic Handbook of the Greater Himalayan Region. Leiden: Brill.