Language learners are often taught that cognates are sets of words that look or sound similar in different languages that mean the same thing.
False friends on the other hand, are words that look or sound similar in different languages that mean different things. “They are not cognates,” some language teachers will tell you sternly.
Spanish: constipado -- to have a cold
I understand why language teachers say this. Even when they know better, this explanation is usually much more useful than the real explanation, at least from a language learning point of view.
The problem is, it’s just not true. It’s also much less interesting than the correct definition of a cognate.
Cognates are sets of words in different but related languages that share an etymological origin: that is, they descend from the same word.
In the examples I gave above, English “much” and Spanish “mucho” are in fact false cognates: words that look like cognates and have similar meanings but are not related. They just happen to look like each other. English “much” traces its lineage to Proto-Germanic *mekilaz and to Proto-Indo-European *meg-, while Spanish “mucho” goes back to Latin “multus” (from which English gets “multi”) and Proto-Indo-European *mel-.
Taking a look at some Romance languages:
Meaning: to hurt, to injure
Meaning: to do, to make
Meaning: to flee
These are all cognates, by both the linguistic and “common” definition of a cognate. It's fairly obvious what's happening here, since we know that Latin is the parent language of Catalan, French, Portuguese and Spanish. We have lots of Latin texts, obviously, and we can trace how each of these languages developed from Latin over time. But supposing we didn't know what Latin sounded or looked like, we could still make a pretty good guess -- we could guess that Latin had the initial /f/ that disappeared somewhere along the way many areas of what is now Spain. Once you know this, finding these types of cognates in these Romance languages is a doddle.
If you have enough data, you can use the lexis, grammar and phonology of the child languages to reconstruct the most recent common ancestor of the two languages using the comparative method(http://en.wikipedia.org/wiki/Comparative_method_(linguistics)). Round up all the Romance languages, run them through the comparative method, and you can reconstruct their most recent common ancestor: Latin.
If the most recent common ancestor is not attested, then the reconstructed language is called a proto-language(http://en.wikipedia.org/wiki/Proto-language). We don’t have proof that a given proto-language exists, but we can be pretty sure that at some point, a language (or possibly a collection of languages) like it must have existed in order for its descendants to have the phonology, grammar and vocabulary they do.
Here are some cognates in English, Dutch and German:
Again, there's a regular pattern here, and this is just one of many patterns that you can find when comparing English and German. Gather a large enough corpus across enough related languages, and you can reconstruct Proto-Germanic(http://en.wikipedia.org/wiki/Proto-Germanic_language). (I'm simplifying things a little, mainly because the actual work of linguistic reconstruction is well outside my limited expertise.)
Let’s dig a little deeper. What about cognates in Spanish and German? In theory, with enough data, you could reconstruct the nearest common ancestor of those two languages:
What you’re seeing here is part of Grimm’s Law(http://en.wikipedia.org/wiki/Grimm%27s_law), one of the most famous laws in historical linguistics, both for being the first systematic sound change to be detailed and for the fact it was discovered by that Jakob Grimm(http://en.wikipedia.org/wiki/Jacob_Grimm), half of the pair of brothers who wrote about rather creepy things when you think about it. Among other things, Grimm’s Law says that in Germanic languages, sounds that were voiceless stops in Proto-Indo-European become voiceless fricatives in Proto-Germanic. PIE /p/ becomes /ɸ/ in Proto-Germanic.
Damn. Now I have to finish writing Articulatory Phonetics 101, I say, only slightly upset.
But yes, if you run all these languages through the comparative method, supplemented by other related languages, you eventually get the reconstructed language known as Proto-Indo-European.
Why don’t we compare all the languages of the world and try to reconstruct a sort of Proto-Babel, the language from which all human language descends?
Because if you take, say, English and Mandarin and use the comparative method on them, what you get is — nothing.
Here’s a set of cognates for you:
Wait a second — one of these is not like the others.
Spanish: actual — current
Catalan: actual — current
German: aktuell — current
Dutch: actueel — current
English: actual — actual
Or, more famously:
Just so you don’t make this mistake in Spain or Germany:
Spanish: preservativo — condom
German: Präservativ — condom
English: preservative — preservative
What happened here? These are true cognates but false friends, just like “constipated” and “constipado”: they share an etymology, but at some point, the meaning attached to one or both of the words shifted, leaving very confused language students in their wake.
If you can remember that cognates do not have to have the same meaning, you can have a lot of fun with cognates, though. They reflect a language’s history and the agility of the human capacity for language in a myriad of ways: consider the mental leaps that must have happened for this change to happen:
German: wollen -- to want
Dutch: willen -- to want
English: will -- future auxiliary
Or this one:
German: sterben — to die
Dutch: sterven — to die
English: starve — to starve
To come back to the whole “cognates for language learning” thing: I’m not saying, don’t use “cognates” or “false friends”. I am saying that understanding what cognates actually are is, in my not-so-humble opinion, a much more expansive and stimulating way of engaging with a language.