The Comparative Method III

So, let's say we've compared several hundred word sets and we've found a likely genetic relationship between a group of languages. This tells us that the languages in the group form a language family, sharing a common ancestor. Now we need to refine our relationships further - within the language family, which languages seem to be more closely related than others? Let's take another look at the European set from the previous post. This time, I've changed the transcription to use the International Phonetic Alphabet, a writing system employed by linguists to precisely transcribe words and speech from any human language. I've also reorganized the languages into two groups based on certain regular phonological correspondences:

English	Dutch	German	French	Italian	Spanish	gloss
aɪ	ɪk	ɪç	ʒə	ˈio	jo	‘I’
mun	maːn	mɔnd	lyn	ˈluna	ˈluna	‘moon’
ˈsɪstəɹ	ˈzʏstəɾ	ˈʃvɛstəʁ	sœʁ	soˈɾɛlla	eɾˈmana	‘sister’
ˈfaɪəɹ	vyɾ	ˈfɔʏəʁ	fø	ˈfwɔko	ˈfweɡo	‘fire’
eɪt	aχt	aχt	ɥit	ˈɔtto	ˈoʧo	‘eight’
fʊt	vut	fuːs	pje	ˈpjɛde	pje	‘foot’
dɹɪnk	ˈdɾɪnkən	ˈtʁɪnkən	bwaʁ	ˈbeɾe	beˈbeɾ	‘drink’

Let's look at initial consonants to begin with. For 'moon', the words in the English, Dutch, and German set (which I'll abbreviate to EDG) all begin with m-, while those in French, Italian, and Spanish (FIS) all begin with l-. For 'drink', the words in the EDG set all start with dr- (don't worry about the various 'r' symbols - they represent different pronunciations, but the fact that they correspond is what matters), while those in the FIS set all begin with b-.
Now look at the 'foot' set. We see the sounds f-, v-, f-, p-, p-, p- in this, which may not seem like a good set at first. However, compare it to these correspondences:

English	Dutch	German	French	Italian	Spanish	gloss
ˈfɑðəɹ	ˈvadəɾ	ˈfatəʁ	pɛʁ	ˈpadɾe	ˈpadɾe	‘father’
fɪʃ	vɪs	fɪʃ	pwaˈsɔ̃	ˈpeʃe	pes	‘fish’
foɹ	voːɾ	fʏʁ	puʁ	peɾ	poɾ	‘for’
fli	vlo	flo	pys	ˈpulʧe	ˈpulɡa	‘flea’

Here we see the exact same pattern across four more sets of words. This is an excellent example of a regular phonological correspondence.
Now, take a look at the words for 'fire'. We see the same f-, v-, f- correspondence in the EDG set, but we then get an f-, f-, f- correspondence in the FIS instead of the expected p-, p-, p-. What's going on here?
When we get apparently inconsistent correspondences like this, there are two likely explanations. One is that we haven't accounted for the phonological environment. The phonological environment is the combination of phonemes around the phoneme we're examining. Depending on whether neighboring phonemes are consonants or vowels (and what types of consonants and vowels), whether the phoneme is at the beginning, in the middle, or at the end of a word, and even if it's in a stressed syllable or not, are all factors that can affect the value of the phoneme over time, and therefore affect the correspondences we're investigating.
However, this is not the case in the 'fire' set. All the phonemes in question are word-initial, and they're followed by a variety of vowels, just like the fvfppp set in the second table, so it's unlikely that a particular type of vowel is throwing off the correspondence. This brings us to the other likely explanation: there isn't meant to be a correspondence. That is, the words for 'fire' in the FIS set are not actually related to the words for 'fire' in the EDG set, despite their superficial similarity.
Words that are related to each other, such as the sets of words displaying the fvfppp correspondence, are known as cognates. Just as genetically related languages share a common ancestor, so too do 'genetically' related words share a common ancestor. When we look for regular phonological correspondences, we are looking for the existence of cognate sets between different languages. In the first table, the FIS-set words for 'moon' form a cognate set - they show regular phonological correspondences, not just in that they all begin with l-, but also that they all contain -n- after the first vowel. In other words, French lyn, Italian luna, and Spanish luna are cognates of each other. The conclusion we draw from this is that, in the language that eventually became French, Italian, and Spanish, a particular word became the three words lyn, luna, and luna over time.
Interestingly, the fvfppp correspondence set suggests that all six languages in the table are actually related, but that they fall into the above-seen EDG and FIS subfamilies. That is, while all six languages appear to share a common ancestor based on the fvfppp correspondence, English, Dutch, and German appear to share a common ancestor that is not shared with French, Italian, and Spanish, and vice versa, as evidenced by the two different word sets for 'moon' and 'drink' in each group, which show no real correspondences.
Next time, I'll explain more about how to recognize likely versus unlikely correspondences, and we'll look at some other complicating factors that need to be accounted for in a comparative study.
Bonus: Look at the word set for 'sister' in all three languages. It should be fairly clear that the Spanish word doesn't correspond well to the other five words. But what about those five? Do they correspond well to each other? Why or why not?