How to teach historical linguistics

It was 1991, and I was an engineering student who was not interested in engineering, and for whom the joy of mathematics had recently been killed by the "Shut Up And Learn The Formula" approach to the proof of Fourier Transforms. I met a girl at a party, and followed her into a linguistics course. It was Historical Linguistics, taught at the time by Ilia Pejros. Twelve years later, I would lecture that course. Sixteen years later, the University of Melbourne would cease offering that course. Aber alles Irdische ist übel, selbst das Geld geht in Verwesung über.

But it was 1991, and I just grinned through the entire course. The linguistics undergrads, leery of the loud engineering student who suddenly turned up among them, assumed that I was grinning at Ilia's accent. I was in fact grinning because I didn't believe a word of it. Historical reconstruction is a special kind of magic, and when you convince yourself it works, it's breezy and amazing. But you need to be convinced that it works first, and I wasn't being convinced by Ilia's assertions that it was obvious.

I convinced myself a couple of years later, but you need some special circumstances to be convinced.

• You need to take a proto-language that you know reasonably well.
  
• And a daughter language that you know at least somewhat.
  
• And the daughter language must in at least some ways look nothing like the proto-language.
  
• Yet the phonological changes from the proto-language to the daughter language need to be reasonably regular,
  
• So that, finding out about one rule, you unlock a whole suite of correspondences where you did not expect them.
  

For me, the stars lined up with Tsakonian, which is one of the reasons I ended up liking Tsakonian so much.

• The proto-language is not quite Doric, as the enthusiasts will tell you; there's bits of Doric in there, but most Tsakonian words, you can reconstruct armed just with v AD Koine Greek; and you can usually get to v AD Koine Greek straightforwardly from the historical spelling of xxi AD Modern Greek.
  
• Tsakonian has dictionaries and grammars, so you can see the words in action.
  
• Tsakonian really looks nothing like Greek should—not just morphologically, but phonologically as well: the clusters have melted away, and the language is pretty much CV syllables (as long as you allow aspirated stops and affricates to be a single C).
  
• But the rules for how the phonological meltdown happened are pretty regular, and in fact continued in productive use into the 19th century: ειρηνοδίκης [irinoðicis] "justice of the peace", a calque from German, ended up quite regularly as ρζινοδίτχι [r̝inoðitɕi], and γραμματική [ɣramatici] "grammar" as γραμμακιτχή [ɣramacitɕi].
  
• And once you know the phonological rules, a lot of opaque words suddenly make sense: /tʰuma/ is /stoma/, /ʎuke/ is /lykos/, /atʃe/ is /aðros/, and so forth.
   

(Btw, I've snuck in there an orthographic reform I'd like to see if Tsakonian does go online with the wikipedia and what-not. Posting on that separately.)

For students in Europe, the stars traditionally lined about around Romance. Everyone knew Latin; everyone knew French (and could be convinced about Portuguese); neither particularly look like Latin in their core vocabulary; but explain the core changes, and suddenly chère does connect to carus, and hui to hodie. In fact, one of my students at the time—who's now my colleague in my day job—told me at the time that his stars-lining-up moment was realising in my final lecture, through a combination of Grimm's Law and velarisation, that French peau was cognate to English fell "the skin or hide of an animal; pelt."

Now, if you're teaching historical linguistics in the University of the South Pacific, as the late Terry Crowley used to, French and Portuguese aren't going to convince your students about the validity of the comparative method. Moreover, given how close the Polynesian languages are to each other, Fijian and Tongan and Maori will convince your students, because after all, you are at the University of the South Pacific. Which means that Terry Crowley writing a historical linguistics textbook based on Polynesian examples was an entirely appropriate and sensible thing to do.

I would also like to suggest that using a textbook based on Polynesian examples in Australia, for students who don't know any Fijian or Tongan or Maori—but a lot of whom have had high school French or German—is paedagogically lame, and deserves all the grins that loudmouth engineering students can muster. Because Tongan will not convince people who know neither Tongan nor Hawaiian that the comparative method works. Sorry, OUP, you got no sales from me; the also late Larry Trask did instead. (I thought it was a bit on the easy side, myself; but apparently my course was the hardest one some of my students ever took in linguistics, so one should err on the side of caution.)

And I can only shake my head at the prof whose very first page of lecture notes on historical linguistics illustrated a phonetic change (I think it was lenition)... with the development of an Australian proto-language into an Australian daughter language. Dude, noone in the room knows Pitjantjara (except for you—we ain't in Bachelor College), and certainly noone in the room knows proto-Pama-Nyungan. INCLUDING YOU. That's just circular reasoning, and its risible to put it in your first lecture. Prove to me lenition happens with languages someone else in the room can nod along to; then get back to me with proto-whatever.

(It's liberating to be able to say this kind of thing, now that I'm no longer anywhere near trying to get a linguistics job...)

That, incidentally, is why dialects are really helpful for convincing yourself of the validity of phonetic change: they're close enough that the cognates are obvious, but still have all the range of phonetic change that languages display. Unfortunately English dialect does more vowel change than consonant change; but its not like you can't find plenty of palatalisation, or cluster reduction, or analogical change...