19

In Old English, vowel length was phonemic, but stress and certain kinds of consonant voicing were not. In Modern English, that situation is reversed: vowel length is no longer phonemic, but stress always, and consonant voicing in most cases now, is.

My main question is . . .

Is there any connection between losing one phonemic property but gaining another? Is there some sort of “conservation of phonemic axes” principle at work here that requires another axis of distinction to appear when one is lost? Did this cause other effects, such as for example stress or voicing or something else becoming phonemic?

Old English represented 14 simple vowels using just 7 letters (a, æ, e, i, y, o, u) in paired short and long variants. This reminds me of how Latin had 10 simple vowels using 5 letters, again in paired short and long phonemically distinct variants. But in the transformation of Latin into Spanish, that 10-vowel system was lost, leaving only 5 vowels remaining. And yet, Latin’s predictable stress was also replaced with a new phonemic stress in Spanish; for example, término, termino, terminó are a minimal triple.

Did the same thing happen to Old English as happened to Latin, as the short-vs-long vowel distinction was lost but phonemic stress was gained, or is it completely unrelated?

There is one particular similarity between the transformation of Old English to Modern English and the one from Latin to Spanish: in both cases, phonemic vowel length disappeared but phonemic stress appeared. Is this pure coincidence?

In Old English, stress was not phonemic because the stress was always on the first syllable except for verbs like forsacan (forsake) or behydan (hide) with a leading particle. But in Modern English, stress is phonemic (compare record as a noun and as a verb), but vowel length is not. Was this new phonemic stress in any way compensatory for the loss of phonemic vowel length?

Other notes

It was these two postings that got me thinking about the matter:

  • In American English, vowel length is almost always allophonic. Generally stressed vowels before voiced consonants are held longer than those before voiceless consonants (e.g, great ~ grade pronounced [greⁱt] and [gre:ⁱd]), and that perseveres even after voicing neutralization, as in grating ~ grading, pronounced ['greⁱɾɪŋ] and ['gre:ⁱɾɪŋ]. As Neil says, real vowel length is one phenomenon which helps the listener to decode speech. But in school, everyone is taught about "long A" [eⁱ] and "short A" [æ], and this has nothing to do with Modern English speech. – John Lawler yesterday

  • The IPA transcriptions are US /ænt/ and UK /ant/. One can use a macron /ā/ or colon /a:/ for the UK one, depending on transcription habits, but vowel length isn't phonemic in any dialect of English, so a simple /a/ will do. I do think that questions and answers about pronunciation in a written medium should try to use standard English phonemic symbols. Otherwise, how do we avoid confusion? – John Lawler 3 hours ago

I notice that we do use vowel length a bit phonologically, even in diphthongs, at least those of us with Canadian raising, such as in the otherwise phonemically identical pairs ice cream and I scream, or writing and riding, where in each pair, the first diphthong is shorter and higher than in the second.

Community
  • 1
tchrist
  • 134,759
  • 3
    This is a really interesting question, but what exactly do you mean by consonant voicing not being phonemic in Old English? Every source I can get my hands on suggests that voice was phonemic for stops, at least. – Branimir Ćaćić Jan 02 '13 at 21:20
  • 2
    @BranimirĆaćić I am thinking of /f/ and /v/, and in fact, consonants at the end of the word, which devoiced. Also, the two versions of what we now spell as th. – tchrist Jan 02 '13 at 21:22
  • 4
    Consonant devoicing at the ends of words is a very common feature of many unrelated languages (also a phenomenon of early child language). You might want to ask this at Linguistics.SE since such alternations have occurred in other languages diachronically. – Mitch Jan 02 '13 at 21:39
  • 5
    I think this should be migrated to Linguistics.SE since it's clearly not localized to English. – Charles Jan 02 '13 at 21:52
  • @tchrist. The versions of th are still phonemic, aren't they? Try thy thigh. – TRiG Jan 02 '13 at 23:05
  • @TRiG Sure, they are now. But I do not think they were in Old English, were they? – tchrist Jan 02 '13 at 23:10
  • 8
    Voicing was allophonic for English fricatives in ME, but phonemic for stops. Resonants and vowels were all voiced, as usual. That's where the irregular plurals like thief/thieves come from; it's also the reason why /θ/ and /ð/ have such a low functional load -- there's widespread free variation (with), there are only two minimal pairs (thy/thigh and either/ether), and they're both pretty flaky. Since they're spelled the same, most English speakers aren't even aware of the distinction. – John Lawler Jan 02 '13 at 23:46
  • @JohnLawler Did we get something in return for losing phonemic vowel length? Phonemic stress perhaps then? I do here some people who make plurals without adding voicing to certain words that I do (houses, moths) but it isn’t like there is a minimal pair there. – tchrist Jan 03 '13 at 00:13
  • 3
    I don't know, but the GVS made vowel length moot, since all the long vowels changed but the short vowels didn't. Now we have tense and lax but not short and long. – John Lawler Jan 03 '13 at 00:16
  • 2
    In the GVS, we lost phonemic vowel length and in exchange, we got a language which is impossible to spell. I think it was a bad bargain. – Peter Shor Jan 07 '13 at 05:16
  • 2
    Only because the installed base of Caxton Beta spelling was so big. It's the same reason we'll be using Windows three centuries from now. – John Lawler Jan 07 '13 at 18:28
  • 1
    An interesting question - why not move it to Linguistics SE? I could share my thoughts on this there. – Alex B. Jan 07 '13 at 19:44
  • @AlexB. Well, for practical and mechanical reasons, it can’t be moved now: it has a bounty. That blocks deletion. But the original point was just related to English, which is why it got asked here. Is there some reason you can only answer there? – tchrist Jan 07 '13 at 19:51
  • I see. No, I can write it here - in a couple of days though (need some time to put things together, insert links etc). Meanwhile, you may want to take a look at Welna 1998 http://www.degruyter.com/view/books/9783110804072/9783110804072.471/9783110804072.471.xml (not exactly what you're interested in but closely related) – Alex B. Jan 07 '13 at 19:56
  • 1
    @AlexB. Thanks much; no hurry. I also looked at Chapter 3 on OE spelling and pronunciation in Dennis Freeborn’s From Old English to Standard English, which is promising but doesn’t quite get there. If you’re bored, you might also see this question on periphrastic tenses for another that might have been better placed on Linguistics.SE, since this place here has become something of an ELL hole of late. – tchrist Jan 07 '13 at 21:36
  • @AlexB. If you would like to answer, you might want to do so before the bounty runs out in a couple days. – tchrist Jan 11 '13 at 00:23
  • The relaxation of any phonemic property makes it easier for non-native speakers to be understood. Perhaps useful for a language to become a lingua franca. –  Jan 11 '13 at 09:23
  • While phonemic vowel lengths are clearly dead in AmE, are they really dead in BrE? It seems to me (although I'm not an expert) that in some dialects, length is the main distinction between bid and beard, between shed and shared, and between tusk and task. – Peter Shor Jan 11 '13 at 12:23
  • @PeterShor I actually wondered that, and was wondering whether someone might bring it up. It seems possible that non-rhotic speakers mark r-deletion by adding length, but I am no expert on such folk, since for me, such speakers in the Am. South sound like they are actually adding a new “uh” syllable, which changes the equation. About tusk/task, I am unsure that is an actual length distinction even for those on the other side of the TRAP/BATH split. Seem like height is involved, too. – tchrist Jan 11 '13 at 12:29
  • @tchrist: See this vowel chart for at least some evidence tusk and task overlap. More evidence is that for at least one British singer (I can't remember who), his /ʌ/ and /ɑː/ sound the same for me; in singing, you lose vowel length distinctions. However, I would bet that most RP speakers have different qualities for /ʌ/ and /ɑː/. – Peter Shor Jan 14 '13 at 14:02
  • @PeterShor Wow, look how much more overlap British vowels have compared with American ones in those charts! – tchrist Jan 14 '13 at 14:05
  • @tchrist: but nowhere near as much overlap as in German, which truly has phonemic vowel length. – Peter Shor Jan 14 '13 at 14:25
  • @PeterShor I notice how nicely spread out Italian’s 7-vowel system is in those charts. I’ll bet Spanish’s 5-vowel system is even further separated. This makes for a goodly number allophones that non-native speakers often throw into the wrong phonemic buckets. – tchrist Jan 14 '13 at 14:33

2 Answers2

10

Yes, there is a connection between losing one phonemic property and gaining another. Most approaches to phonology conceptualize words as having double lives: on the one hand, they are made of a particular sound sequence which you have to pronounce correctly; on the other hand, the sounds in sequences are only recognized as discrete parts because they contrast with other sounds.

This property of phonology was termed double articulation by the French phonologist Martinet. One also speaks of form vs. substance: form being properties of speech sounds which they have by virtue of being in contrast with other sounds, or undergoing meaning-preserving alternations involving other sounds, and substance being the physical phonetic details of their pronunciation. (Quite confusingly, one also hears of function vs. form, where form coincides with substance from the other, roughly equivalent, dichotomy.)

These two factors frequently come into play in the evolution of vowel systems. To characterize it approximately: Consider a set system of 14 vowels, with 7 basic vowel qualities and two lengths. For such a system, the pertinent formal properties are that each vowel is contrastive with every other, and that the vowels may also be divided into pairs (e.g., e and e:) on the basis of meaning-preserving alternations. The fact that the pairs are differentiated by length and not by quality instead relates to the substance.

When a language undergoges sound change such that the phonetic substance is altered, but the formal relations between the sounds are preserved, it is usually referred to as transphonologization. Such a process is quite common historically, because languages do have a tendency of conservatism in form, if not in substance.

For further reading and numerous examples, consider a 2008 paper by the phonologist Larry Hyman and references included therein. For further reading treating English phonology more specifically, see various studies by Roger Lass, esp. English Phonology and Phonological Theory, and Old English Phonology.

  • Marvelous!! Gimme a few hours in case AlexB chimes in, but I do like this answer. – tchrist Jan 12 '13 at 17:05
  • I don’t know whether this matters, but in English, the long vowels mutated under the GVS, while the short ones didn’t. So we kept a large number of vowels, but length was no longer the distinguishing property. On the other hand, when the Latin ten-vowel system collapsed into the Spanish five-vowel system, they had an actual reduction in the number of phonemically distinct vowels. – tchrist Jan 12 '13 at 17:36
  • @tchrist the English change could be characterized as transphonologization since the contrastive properties stayed the same more or less, while the Latin-Spanish change is a case of loss of contrast. –  Jan 12 '13 at 17:54
6

(first draft, part 1, to be edited)

You have asked so many questions, and many of them don’t have simple, irrefutable answers.

So, I’ll try to address some of your questions below - immensely simplifying things and ignoring exceptions and minor cases.

At first, some caveats (the following is mostly based on Hogg 2008).

  • Vowel length was rarely marked in the Anglo-Saxon manuscripts.

  • Another problem is that there was considerable dialectal variation in OE. When we talk about the OE sound system, we actually mean one dialect, Late West Saxon.

  • Another problem is that “the core antecedent” of Standard English wasn’t West Saxon at all – Hogg argues that West Saxon had a marginal influence only.

  • Also, since there are no sound recordings of Old English speech, all we can do is to reconstruct the OE phonemic system. And, as usual, there is no consensus among linguists on many issues there.

Hogg 2008 reconstructs seven short vowels and seven long vowels for OE.

Q1. How do we know that there were short and long vowels in OE?

Q2. Were OE short and long vowels also quantitatively different, like in Latin?

Since there is no hard evidence – we have no recordings or native speakers of Old English – we don’t know for sure.

However, Hogg 2008 and Lass 2008 argue that short and long vowels began to diverge qualitatively before the end of OE.

Q3. Was vowel length sensitive to postvocalic environment in OE?

Lass 1994 argues that vowel length was free in OE. However, in late OE and early ME, things changed:

  • homorganic lengthening: OE cild (hence, child);
  • pre-cluster shortening: originally, before clusters of three consonants, e.g. OE cildru (hence, children); later before clusters of two consonants, e.g. OE cepte (hence, kept) vs. OE cepan (no shortening, hence keep), or OE mette (met) vs. OE metan (meet).

Q4. What about stress in OE?

It was always on the root. Noun and adjective prefixes were also stressed, unlike verb prefixes, which were usually unstressed (for more complex cases, see Lass 1994: 92). Hogg 2008 reminds us that what we usually know about OE stress is mostly based on poetry, and there is good reason to believe that in normal speech stress could have worked differently.

Also note that long vowels were always stressed in OE - in unstressed syllables, long vowels got reduced.

Alex B.
  • 4,210