Monday, 7 May 2012

graded meaning of gradation

The term gradation and the concept behind it is fairly old. It dates back at least till the 19th c., when it was used as a synonym for ablaut (denoting the variation of a root vowel to indicate a morphological change). We're all familiar with the root vowel changes in ride, rode, ridden or swim, swam, swum. The term ablaut was coined (or, at least, popularised) by the German philologist Jacob Grimm1. The phenomenon itself had been described a few millenia before Grimm.

It was Pāṇini, known for his Sanskrit grammar, who described it in his 'Eight Chapters'.  The concept referred to the length of vowels. In Indo-European linguistics it came to be known as zero grade (= no vowel), full grade (= short vowel) and lengthened grade (= long vowel). An often cited example is pater, the Greek word for father. The singular accusative is patera (with a short stressed<e>), the sing. nominative is pater (with a long stressed <e>) and the genitive is patros (with no e-vowel). So patros illustrates zero grade, patera exemplifies full grade and pater demonstrates lengthened grade. These forms of the Greek noun pater display quantitative gradation. But there's also qualitative gradation as in sing, sang, sung and song.

The gradation types discussed so far serve morphological purposes, they denote a change in case (Greek pater), tense (sing, sang, sung), transitivity (rise, raise) or word class (sing, song). By the end of the 19th c., however, the term had acquired an additional meaning. In Henry Sweet's Elementarbuch des gesprochenen Englisch of 1885 we find a chapter on "Abstufung" (p. xxix ff.). In the English version of that book called A Primer of Spoken English of 1890 the term Abstufung is translated as "gradation" (p. 13). It is in this book that Sweet characterises gradation as a "totally new subject" (p. x).  In his Handbook of Phonetics of 1877 he had made no mention of the term.

Elementarbuch Primer

This 'new' meaning of gradation does not imply that the traditional meaning has come out of use - quite the contrary, see e.g. W. W. Skeat in his Principles of English Etymology, vol. I, of 1892, where he uses the term in its traditional sense (see his ch. X).

Daniel Jones took up the concept of gradation in the Sweetian sense, e.g in his Paris lecture which he delivered in January 1912 to the Guilde Internationale. The lecture to an audience of phonetic laymen was entitled "The teaching of English pronunciation to French students". Jones  applied the concept pretty much in the same sense as Sweet had done when the former talked about "the treatment of small words like have, from, but, should, the, them, of and a large number of others." Jones went on to say: "[...] the weak forms are used when the words are unemphatic or in unimportant positions. [...] The weak forms all have the weak neutral vowel [
ə]." The term gradation, however, was not used until later in his Outline of English Phonetics (thanks to Jack Windsor Lewis (= JWL) for drawinɡ my attention to the book). In the following table you find all major editions and the paragraph in which the term is first mentioned.

ed.   year   paragraph
1st   1918   Ø
2nd   1922   Ø
3rd   1932   467
4th   1934   467
5th   1936   467
6th   1939   467
7th   1949   467
8th   1956   467
9th   1960   467

Here's a scan of the relevant paragraphs taken from the 9th ed. of his Outline:

In addition to schwa Jones records the use of weakforms with /i/ (e.g. /ði/) and /u/ (e.g. /tu/) (I use Jones's set of symbols here).

All of us are aware of the fact that, more often than not, one and the same English word may be pronounced in more than one way. This applies not only to the so-called 'small' words like and (/ænd, ənd,
ən, nd, n/) but also to 'bigger' ones such as particular (/pəˈtɪkjʊlə, pəˈtɪkjələ, pˈtɪkjələ, pˈtɪk, .../) or police (/pəˈliːs, pliːs/). The terms strong form and weak form (overwhelmingly spelled as two words) are widely accepted for describing the changes of /ænd/ to, e.g.,  /ən/. But what are we to call the comp'r'bl changes observable in /pəˈtɪkjʊlə/ -> /pˈtɪkjə/, in /əʊnli/ -> /əʊni/ or in comp'r'bl ?

Very often the phonetic concept of weakforms is tied to the grammatical concepts of lexical and function (or grammatical, structural, synsemantic) words in the literature (see e.g. Cruttenden (20087), Gimson's Pronunciation of English, p. 266). Cruttenden writes:

Lexical words [...] generally have in connected speech the quantitative pattern of their isolate form and therefore retain some measure of prominence based on the occurrence of a full vowel even when no pitch prominence is associated with them.

However, many function words have two or more qualitative and quantitative patterns according to whether they are unaccented (as is usual) or accented (in special situations or when said in isolation). As compared with the accented realizations of these words (the strong forms), the unaccented weak forms of those words show reductions of the length of sounds, obscuration of vowels towards /ə, ɪ, ʊ/ [Cruttenden's symbols] and the elision of vowels and consonants."
The list that follows contains not only function words, but also 2 lexical words: saint and sir.  

What characterises functions words phonetically is, according to the author, this:
  1. ≥2 qualitative and quantitative sound patterns [I would say 'qualitative and/or quantitative'];
  2. the selection of a pattern depends on whether the function word is accented (= the strongform is chosen) or unaccented (= the weakform is used);
  3. characteristic features of a weakform sound pattern are:
    1. sound length reduction,
    2. vowel quality obscuration, and/or
    3. elision of vowels and/or consonants.
Cruttenden's features just mentioned can easily be applied to and, from, shall and many other function words. But they are also applicable to only as /əʊnli/ or /əʊni/, although the adverb is not a function word:
  1. there are 2 qualitative sound patterns;
  2. the form /əʊni/ is selected when the word is unaccented; 
  3. as a consonant is elided, feature 3.3 applies here.
Likewise, particular, police and comparable can be considered as having a strongform and one or more weakforms.

I'd like to get rid of the grammar-centred terminology and rather concentrate on phonetic distinctions. The best ensemble of such features that I've detected in the literature so far is to be found in an online article by Jack Windsor Lewis with the title "Weakform Words and Contractions for the Advanced EFL User" and in his book A Guide to English Pronunciation (1969). The following features try to describe the differences between strongforms and weakforms:
  1. there are ≥2 sound patterns of a word with different sets of phonemes;
  2. one of the forms (= strongform) is used when sentence rhythm or tempo call for its use or the speaker wants to accentuate the word, else one of the weakforms may be used;
  3. the phonemic difference between the forms must not be affected by contextual effects such as assimilation (/kən ɡəʊ/ -> /kəŋ ɡəʊ/) or elision (/kɑːnt teɪk/-> /kɑːn teɪk/);
  4. the strongform represents the canonical2 pronunciation of a word; all alternant weakform pronunciations are derived from the canonical version (JWL via personal communication - slightly re-phrased);
  5. there's a subgroup of forty-odd words with one or more grammatico-stylistical variant pronunciations; this subgroup is called "special weakform words";
  6. the selection of a weakform or a strongform of a word which belongs to the subgroup is more or less obligatory.
This latter characterisation of weakform words not only enlarges their number considerably, but also cuts across the traditional fairly tight classification (with one or two exceptions such as saint and sir in GB) into lexical and grammatical words. We not only have articles, prepositions, auxiliaries etc. in the set, but also adverbs, adjectives, full verbs and nouns.

Shocking, is it not?

P.S.: See also JWL's blogs no. 399 and no. 400.
1 Grimm, J. (18222), Deutsche Grammatik, vol. 1, (Göttingen). 
2 By canonical pronunciation is meant
  • a) the isolated, non-embedded, heavily stressed/accented, prominent, generally accepted pronunciation, which is often used in definitional, metalinguistic discourse.
  • b) If the canonical pronunciation is sententially embedded, then 
    • all the pronounceable sounds of the word are in fact pronounced and
    • the particular pronunciation of the word is not confined to being embedded in a particular slot of the cotext and
    • the pronunciation is not the result of an assimilatory or elisional process.


  1. Following the above definitions I find that 'only' without /l/ is not a weak form for me. I'm happy to use my l-less form in stressed and accented positions. For me the /l/ is a matter of style - I only use it when I'm making an effort.

    Surely I'm not all alone in this...

    1. In the speech of an individual, if he/she uses an ell in 'only' in any circumstances at all in which the word occurs, the ell-less forms may surely be deemed to be derived from his/her fuller form. Any GB speaker like yourself can be expected regularly to use the ell form before any complete break in rhythmic flow. The on'y exception to this is when the speaker uses a weakform of 'only' which has an articulation reduced from the speaker's ideal normal/lexical/canonical/ version by pre assimilation of the /l/ to the following /n/ resulting in the form /`ounni/.

  2. Would you agree that, for a foreign learner to sound "correct", it would be words like "and", "to" etc that should be "weakly spoken" in connected speech, whereas words like "only" and "particularly" should rather keep their full articulation?

    1. Basically - yes. But as an advanced speaker of English you should also try to be an advanced listener and be prepared to be 'confronted' with reduced pronunciations of words like comp'r'ble, p'tic'lar, p'lice etc.

    2. That's true. Even for beginners, I should say (They also have a right to know what's going on!).

  3. Please define "canonical pronunciation".

    1. canonical ~ isolated, non-embedded, heavily stressed/accented, prominent, generally accepted canonical ~ isolated, non-embedded, heavily stressed/accented, prominent, generally accepted pronunciation - often used in definitional, metalinguistic discourse. One may add that the pronunciation of a word can be called a strongform, if all pronounceable sounds are in fact pronounced and if the particular pronunciation of the word is not confined to being embedded in a particular cotext. This is no definition of canonical pronunciation in the strict sense of the term 'definition', it's just an ensemble of characterising features.

  4. Thanks, Petr.

    I was gonna make some comments, but I see JWL is engaged in more post on the subject of wfs, so I'll remain stumm for the time being.

    1. Don't do that for long or are you practising some spiritual exercises? ;-)

  5. When Kraut sez "In addition to schwa Jones records the use of weakforms with /i/ (e.g. /ði/) and /u/ (e.g. /tu/)" adding
    « (I use Jones's set of symbols here) », it's worth noting that Jones is indicating what is today more usually shown as /ðɪ/ and /tʊ/ ie Jones was recording what he classified as changes of phoneme.