On Liquid

For Christmas, I asked for and received a copy of W. Sidney Allen's Vox Latina: The Pronunciation of Classical Latin.  It's fairly short, but a good read, packed with all sorts of interesting historical and linguistic tidbits.  The meat of the book is Allen's examination of each Latin phoneme one by one, in which he discusses the various contexts where it occurs and the evidence for its pronunciations in those contexts.  In the section titled Liquids, he discusses the origin of the term liquid, which it had never occurred to me to wonder about.

A liquid, in phonetics, is usually (and rather loosely) defined as an l- or an r-sound of some kind, but precisely which phones are included depends on who you ask.  Let's look at some definitions.  The online Lexicon of Linguistics says:

a cover term for laterals ( [l] ) and various r-sounds.

This is only half of a definition, unfortunately.  Laterals are well-defined, but there are a lot of r-sounds out there.  (pop quiz, hotshot: the voiced uvular fricative [ʁ]—liquid or not-a-liquid?)

Wikipedia defines liquids a little more precisely:

Liquid consonants, or liquids, are approximant consonants that are not classified as semivowels (glides) because they do not correspond phonetically to specific vowels (in the way that, for example, the initial [j] in English yes corresponds to [i]). The class of liquids can be divided into lateral liquids and rhotics.

English sounds [l] and [ɹ] are typical liquids.

So liquids are approximants that aren't semivowels.  That's tidy and consistent, but it's a little suspicious because it excludes [r] and [ɾ] (a trill and a flap, respectively), which are more common r-sounds than [ɹ] in the world's languages.

The OED defines liquid as:

A name applied to the sounds denoted by the letters l, m, n, r, or (by some writers) only to those denoted by l and r.
The name (L. liquidæ, sc. litteræ) is a literal translation of the Gr. ὑγρά (sc. στοιχει̑α) applied to λ, μ, ν, ρ, on account of their flowing and easy sound as compared with other consonants, or perh. as having an indeterminate or unstable character between consonant and vowel (cf. the application of ὑγρός to a vowel of variable quantity; also the term ἡμίφωνα 'semi-vowels', applied to the 'liquids' and σ). A somewhat analogous term is the F. mouillé lit. 'wet', used to denote the palatalized pronunciation of l and some other consonants.

First of all, note that they say some writers include n and m as liquids—that wasn't a usage I'd heard before.  Also notice that they've dodged the phonetic question by talking about sounds denoted by letters rather than phones.  However, this definition starts to get at the origin of the term—it's a translation of a Greek term that they say could refer either to the liquids' "flowing" sound or their "unstable character" (call the cops!).

OK, you've waited long enough.  Here's what Allen says about liquids:

This title is commonly given to the r and l sounds of Latin (and indeed generally).  It is ultimately derived, through the Latin translation liquidus, from the Greek ὑγρός 'fluid'; this rather peculiar term was applied by the Greek grammarians to the consonants r, l, n, and m, in reference to the fact that when they follow a plosive (as, for example, tr), they permit the quantity of a preceding syllable containing a short vowel to be 'doubtful'—as in Greek πατρός, Latin patris, etc.  In Latin, however, this does not apply to n and m, and so the term 'liquid' has come to have a more restricted sense.  (Allen p. 32)

Aha!  All becomes clear.  The term liquid originally referred, not to some quality of the sound of certain consonants, but to their flexible interpretation in the assignment of heavy and light syllables in Latin and Greek.  If I understand what he's getting at, the example Latin word patris could be syllabified either as pa.tris or pat.ris, in which the first syllables are light and heavy, respectively.  This also explains why m and n were mentioned by the OED as being included "by some writers"—whether they're included depends on whether they were writing about Latin or Greek metrical phonology.  (Is this because Greek allows clusters like pn, or is it just an independent fact about Greek meter?)

It's worth pointing out, though, that while Allen tells us its origin, the meaning of the term liquid has drifted over time.  In his A Course in Phonetics, Peter Ladefoged (pbuh) says liquid is "used simply as a cover term for the consonants /l, r/" (p. 58), and that really does seem to be the modern usage.  Fortunately, these days we have much more precise ways (like the IPA) of referring to the phones we want to talk about.  The rather language- (and alphabet-) specific term liquid probably ought to be avoided, unless you're very careful how you define it.

Comments

[Marcos]

Also notice that they've dodged the phonetic question by talking about sounds denoted by letters rather than phones.

... which is, to be fair, pretty much how the term rhotic is defined these days: "sounds denoted by <r> in Latin transcription". :)

[includedmiddle]

Perhaps an uvular fricative could be called a superfluid? Although it would be an oxymoron for a superfluid to be a fricative, the irony makes the term all the more compelling.

[Bridget]

That's really interesting! I've always been uncomfortable writing 'liquids' even in my own notes to mean 'l, r, + nasals' because I've had professors use the term in both ways. Ditto when I abbreviate with R in a phonological rule. I often look back and wonder whether I meant that nasals are included or not.

[Chris W.]

Very interesting. I haven't thought about the term for well over ten years.

My first concept of "Liquida" (in German) was, I'm quite sure, around the age of 12 (it might have been in Latin class), and it meant "l, r, n, m". We didn't make any clear distinction between letters and sounds, let alone phones or phonemes. I explained it to me as "the consonants you can say for as long as your breath holds" -- but then discovered f, s and v. Some time later, after secondary school, it seemed to mean just l and r. But the term wasn't ever used for anything important during my education.

[Chris W.]

Sorry, copy and paste (from "someone explained it to me")... "to myself".

[Robert Port]

What this discussion is REALLY about, in my view, is that the major classes of segments cannot be defined. The simple idea 100 years ago was that speech comes in segments and every segment is either a consonant or a vowel. But, oops, there are also these danged `semivowels'. What are they? C or V? Well, I guess they are `half-vowels'. But we cant really say. So terms like `liquid' are just additional attempts to patch up the category muddle initiated by distinguishing C vs. V. The REAL problem is that speech does NOT come in the form of and words cannot really be spelled out of cognitive consonants and vowels. That is, speech does not really come in the form of segments at all. It comes in the form of gestures and auditory/acoustic trajectories. Segments are a convenient device for writing speech down (if you know the language) and, due to lifelong training using letters to describe language, letter-like units are intuitively very persuasive to us. But evidence is astonishingly absent showing that we store words in memory (long-term or short) using letter-sized pieces.

[Claire]

But Robert, there is also considerable evidence for humans abstracting categories out of the speech signal. We produce speech as a continuous stream, but that's not how we process it. Phonemes weren't just invented by a linguist with too much time on their hands.

wow! man!!!!

interesting discussion!

[agm]

What does this have to do with shooting the hostage?

[visitor]

There are languages, which have vowels corresponding to "l" and "r"! Take for example Sanskrit, in which these two are called semivowels (together with "y" and "v"), and there are vowels (in transcription "l with a small circle under" and "r with a small circle under") pronounced somwhat similar to l in "slurp" and ri in "rig". There are also contemporary languages, like some of the Slavic (not all of them), using these vowels.

[Chris W.]

I just came across this in the stub for "Consonne liquide" in the French Wikipédia: "Une consonne liquide est une consonne spirante qui n'est pas une semi-voyelle car elle ne correspond pas phonétiquement à une voyelle. Les consonnes liquides peuvent être latérales ou rhotiques." This is the same definition you already list above, i.e. an approximant that isn't a semi-vowel.

But interestingly, the writer continues: "Le français ne comporte qu'une liquide : [l]." Which means that at least for some people over here, [ʁ] is definitely not a liquid (as it's not an approximant in the first place).