back home

Strengthening in unstressed position:
We happy?

Péter Szigetvári
We happy? Vincent, we happy?
Pulp Fiction, 00:18:03–00:18:11

Ádám Nádasdy has written about the vowels of unstressed syllables in British English on several occasions (Nádasdy 1989, 1993, 1994, 2013). This squib, written in honour of Ádám’s birthday,Ádám’s share in my professional and personal development is so immense that I will not even try to estimate it here. Even harder would it be to properly thank him. selects the same topic. We are going to address the unexpected lengthening happening to unstressed vowels in certain positions. The lengthening of a vowel is expected in a stressed syllable, but in English we find this phenomenon in unstressed position, which is odd.

English vowels do not all occur in both stressed and unstressed syllables. Two short vowels, ɪ and ə undoubtedly occur in unstressed syllables, for example, the last syllables of comic kɔ́mɪk and comma kɔ́mə. A third short vowel, ʉ was also common in unstressed position, but has mostly been replaced by ə, as in accurate ákjʉrət > ákjərət, and we also often find ʉw unstressed, as in statute státʃʉt or státʃʉwt. Two further vowels that occur in what seems to be an unstressed syllable are ɪj, as in the last syllable of easy ɪ́jzɪj, and əw, as in the last syllable of photo fə́wtəw. Our examples show for ɪj and əw that the same vowels may also occur stressed. The other unstressed vowels also occur stressed: we find stressed ɪ in kit, stressed ʉ in foot, stressed ʉw in goose, and stressed ə in strut.This may sound surprising, some details to come below. That is, vowels occurring in unstressed position all occur stressed too, but not vice versa, there are vowels that only occur stressed.

What is the happy vowel?

The occurrence of ɪj in unstressed syllables seems to be a new development in English, the result of so-called happy-tensing (Wells 1982:165f). In word-final and prevocalic unstressed syllables where older speakers of standard British English had kit, younger speakers have fleece. Today, kit in this position is more or less a regional feature in England.

The result of happy-tensing is symbolized by i in dictionaries. This symbol was meant to be a shorthand for “either kit or fleece”, ie it is a typographical convention, not a member of the vowel inventory. However, it is often taken to be a third vowel in addition to ɪ and ɪj, an interpretation that Nádasdy (2013) also accepts.There are other similar conventions, like əɪ for the second vowel of comet, as opposed to ɪ in comic and ə in comma. Nobody proposes that we are dealing with three vowels — əɪ, ɪ, and ə — here.

This, however, is not tenable. The happy vowel does not contrast with either kit or fleece: it is in complementary distribution with both. happy occurs only word finally and prevocalically, kit occurs only preconsonantally. happy occurs only unstressed, fleece — if it were a different vowel – would only occur stressed. Although happy may be phonetically shorter than fleece, this fact is no more significant than that the vowel of fleece is shorter than that of fleas, they are still considered identical vowels. It is just natural that in unstressed position a vowel is less prominent (eg shorter) than in stressed position.

Lexical sets are selected in such a way that it is possible to characterize vowels in different varieties of English. No variety has all 27 vowels that the lexical sets proposed by Wells define. Furthermore, these sets do not cover all the variation that there exists in English: there are futher lexical sets, eg poor, which has merged with force in BrE, but remains with cure in AmE. Wells proposes a lexical set bath because in some varieties of English this set has the same vowel as trap, while in others the same vowel as palm. But in this case again we are dealing with only two, not three vowels.

The existence of the lexical set happy represents the fact that these words belong to kit for some speakers and to fleece for others. So Current British English has two vowels — fleece=happy and kit — and other varieties of English also have two vowels, not necessarily these two, but fleece and kit=happy.

How to tell a stressed syllable?

Before proceeding any further, we must take a short detour to ponder on how we know if a syllable (or vowel) is stressed or unstressed in English. For example, why do we think that the second syllable of statute is unstressed? It is obvious that the first syllable is stressed, since the vowel in this syllable is more prominent than that in the second, and when the tonic of an utterance falls on this word it will be on the first syllable, not the second. But the second syllable could still be either stressed or unstressed.

English words with more than one syllable may have stress on any, potentially all of their syllables. Two-syllable examples are given in (1), where we use “1” for stress that is a potential tonic, “2” for stress that is not a potential tonic, and “0” for an unstressed syllable.

  1. 10: moment mə́wmənt
  2. 01: lament ləmɛ́nt
  3. 12: comment kómɛnt
  4. 11: torment toːmɛ́nt

In the above examples syllables labelled 0 contain ə, but vowel quality is not an unambiguous indicator of stress. We have seen that ɪ may occur in an unstressed syllable, but also in a stressed syllable in words of the lexical set kit. In fact, there are analyses of English in which ə is also ambiguous in this respect, occurring stressed in the lexical set strut — or goat for that matter, as in moment in (1a). We have already seen that not only əw, but also ɪj, ɪ, ʉ, and ʉw may occur both stressed and unstressed.Analyses in which some members of the vowel inventory do not occur in stressed syllables must provide some explanation of this: it is not expected to find a segment in unstressed but not in stressed position. One common explanation is that schwa is not a phoneme, ie it is not a member of the vowel inventory. In fact, one reason for the popularity of the symbol i, why it is retained in transcriptions even today, when the transition from kit to fleece is practically complete, is that it provides a convenient way of representing unstressed fleece. Just like ə is used for unstressed vs ʌ for stressed strut, i is unstressed, ɪj (or, more commonly, ) is stressed fleece.

But even without the trick of using different symbols for stressed and unstressed vowels, there are certain symptoms that are suggestive of whether a syllable is stressed or unstressed. The short unstressed vowels other than ə, namely ɪ and ʉ, often alternate with ə: eg accurate ákjʉrət or ákjərət, return rɪtə́ːn or rətə́ːn, ticket tɪ́kɪt or tɪ́kət. In a stressed syllable this type of alternation is vanishingly rare.Mostly found across accents: eg strut northern strʉt, southern strət in England. There are also segmental phenomena that distinguish the two types of syllables. Palatality after coronal consonants, before ʉ is often lost in stressed syllables, but retained in unstressed ones: eg value váljʉw vs volute vəlʉ́wt (hence we know that the last syllable of absolute ábsəlʉwt is stressed, although it is not a potential tonic) or statute státʃʉwt vs astute əstʃʉ́wt, American əstʉ́wt. Because constitute kɔ́nstɪtʃʉwt is -tʉwt in American English, we may conclude that its last syllable is stressed, similarly to that of absolute. Of course it may be the case that constitute has stress on the last syllable in AmE, but not in BrE, but it is more likely that the difference in the presence vs absence of palatality follows from the fact that in this context BrE keeps palatality in stressed syllables too (eg tune tʃʉ́wn).

The last phenomenon mentioned here is flapping and glottal replacement, which occur to t before an unstressed vowel, but not before a stressed one (Harris & Kaye 1990). This makes us think that the last vowel of city AmE sɪ́ɾɪj, BrE sɪ́ʔɪj or sanity AmE sánəɾɪj, BrE sánəʔɪj is not stressed, but that of suttee or manatee, where neither flapping nor glottal replacement is possible, is stressed, even for those speakers who can have the tonic only on the first syllable.Szigetvári & Törkenczy 2011 contains further discussion of the issue.

A paradox

We have seen above that statute has two pronunciations, one with a short vowel in the unstressed syllable, státʃʉt, and one with a “long” vowel — actually diphthong — státʃʉwt. It is not obvious what the chronology between these two forms is, which of them is more conservative and which is the newer development.Kenyon & Knott (1953), for example, have statute with a short vowel (foot) in the second syllable, Jones (1924) has a long vowel (goose) here. But there is another similar change, happy-tensing, some examples of which are listed in (2).

  1. happy hápɪ > hápɪj
  2. create krɪɛ́jt > krɪjɛ́jt

As we have seen above, the direction of the change is clear in this case: ɪ > ɪj. The change then is rather surprising: why should the short vowel of an unstressed syllable lengthen? Cross-linguistically stress is characterized by loudness, higher pitch, or increased vowel length, it is odd that we should find a change resulting in a characteristic property of stress in unstressed syllables in English.

Vowel phonotactics

The vowels of Current British English (CUBE) fall into three groups according to the contexts they may and may not occur in. Any vowel occurs before a consonant, but only a subset of the vowels occur at the end of words, and an even smaller set occurs before another vowel. We show the three categories in (3).

kit, dress, trap, strut, lot, foot
near, square, start, nurse, force, letter
fleece, face, price, mouth, choice, goat, goose

The set of vowels that is excluded from word-final (and also prevocalic) position contains the vowels represented as short monophthongs in the British transcribing tradition.These vowels are the lax vowels in the American tradition. These are commonly referred to as checked vowels. There is only one short vowel that is allowed word finally, unstressed ə, that is, the lexical set letter (as well as comma, the two are merged in British English). This property of letter groups it with the so-called R vowels, which, more often than not, are the result of the influence of an earlier r — itself often lost — on the preceding vowel. Although these vowels do occur word finally, an r appears after them if the next word or suffix is vowel initial (eg nearest nɪː r əst, drawing droː r ɪŋ, comma and colon kɔmə r ən kəwlən). Therefore, R vowels are not available before another vowel. Apart from ə, R vowels are phonetically long monophthongs in CUBE, although the mainstream transcription still uses some obsolete centring diphthong symbols to transcribe some of them (ɪə for near, ʊə for cure, sometimes also for square, though not ɔə for force anymore).

The only category that occurs even before another vowel, that is, in every possible context, contains vowels that are phonetically diphthongs, though again two of them, fleece and goose, are traditionally transcribed by monophthongal symbols (as and ).Sweet (1900) had ij and uw for these vowels, but later Jones adopted the monophthongal symbols. The inappropriacy of these two symbols is clear: not only are these two vowels commonly pronounced as diphthongs (Jones 1960:65f, 85, Gimson 1989:101f, 121), but they also pattern together with all the other diphthongs of CUBE.

Are there diphthongs in British English?

Standard descriptions of English are almost unanimous in classifying at least price, mouth, and choice as diphthongs, that is, two-mora vowels occupying a branching nucleus. Descriptions of British English also include face and goat, and — as we have just seen — ought to include fleece and goose in their diphthong inventory.

However, convincing evidence for analysing these sequences — ɪj, ɛj, ɑj, oj, aw, əw, and ʉw — as diphthongs is not easy to come by. In fact, there are a number of facts that strongly argue that English does not have diphthongs at all, that is, the seven “vowels” listed above and in the last line of (3) are not vocalic sequences, but the same short vowels as those in the first line of (3), the checked vowels, followed by consonants, j and w (Szigetvári 2016). The reasons for treating these sequences as diphthongs are mostly historical (many instances of almost all of these sequences — with the exception of choice — come from Middle English long vowels), as is clearly reflected in the spelling of English: apart from choice and mouth, all “diphthongs” may be spelled by a single vowel letter (eg scene, sane, sine, stone, rune).

If free vowels are not diphthongs, but vowel+consonant sequences, ie there are no free vowels at all, then it follows that English does not have hiatus. R vowels may occur word finally and before a consonant, checked vowels only before a consonant. The sequences formerly called free vowels are checked vowels followed by one of the two glides, j and w.

Putting it together

If the vowel of happy, which in CUBE has merged with fleece, is not a long vowel or a diphthong, but a (short) vowel followed by a glide (a consonant), then this vowel has not undergone the unexpected lengthening or “tensing”. The change we experience in create from krɪɛ́jt to krɪjɛ́jt is simple hiatus filling. Recall, neither R vowels, nor checked vowels are allowed in prevocalic position. We assume that this constraint did not hold for unstressed ɪ and ʉ earlier, but it has been extended to all vowels, long or short, stressed or unstressed in CUBE (also cf Wells 1982:291).

Word finally, however, we can talk of hiatus filling only when the following morph is vowel initial, as in happier or happy ever after, but not when it begins with a consonant, as in happiness or happy prince, or when the word is utterance final, as in I’m happy. So why do we have the glide — a consonant(!) — epenthesized at the end of a word? We are used to seeing consonants deleted from the end of words, but not to seeing them inserted there.

When the *V# constraint came to exclude all short vowels, but unstressed ə, words like happy hápɪ had to be amended. The obvious option of simply lengthening the vowel was apparently not available, because this would have been an indication of a historical r, as if the words were “happeer.” The next best repair mechanism is inserting a homorganic glide, and this is what has happened: hápɪj. Note that the same strategy is applied to loanwords ending in a short vowel in their donor language: eg Italian spagɛ́tːi > spəgɛ́tɪj, Polynesian tabu > təbʉ́w, French kafe > káfɛj, Italian putːo > pʉ́təw,British English əw is the result of the fronting of ow. American English is conservative in this respect, it does not (yet?) have this fronting. etc. We find a similar glide insertion repair mechanism at the other edge of words in Slavic languages: eg Proto-Slavic edinu ‘one’ > Polish jeden, Proto-Slavic osmⁱ ‘eight’ > Russian vosemʲ, Proto-Slavic ona ‘she’ > Belorussian vona. The only vowel that is repaired by real lengthening in English is the low vowel: eg Yamaha jáməhɑː.Although here the reason for lengthening may well be to save the h.

The vowels of unstressed syllables

If we represent the short/checked vowels of English in a vowel chart, abstracting away from the quality differences between the short–long pairs aɑ and ɔ, we see that in unstressed syllables we find the more closed one of each (front, central, and back) pair.

    ɪ        ʉ    unstressed and stressed
    ɛ        ə        o    only stressed

As we have seen above, three “diphthongs” also occur in unstressed syllables: ɪj in easy, əw in photo, and ʉw in value. This requires absolutely no explanation if we suppose — as has been supposed here — that these are in fact short vowels followed by a glide. It automatically follows from (4) that these sequences are also possible in an unstressed syllable.


happy-tensing occurs word finally (eg in happy hápɪj) and prevocalically (eg in create krɪjɛ́jt). The distribution of unstressed ʉw, however, is wider. Like unstressed ɪj, unstressed ʉw occurs word finally (eg in value váljʉw) and prevocalically (eg in evaluate əváljʉwɛjt), but we also find this sequence preconsonantally before a stressed syllable (eg in stimulate stɪ́mjʉwlɛjt;We have free variation here: stɪ́mjəlɛjt is also possible. although not before an unstressed syllable: stimulus stɪ́mjələs) and in the last syllable (eg in statute státʃʉwt).

Preconsonantal epenthesis of w cannot be explained by hiatus-filling (as in create krɪjɛ́jt) or avoiding the occurrence of a word-final short vowel (as in happy hápɪj). So we suspect that in the preconsonantal examples above, we do not observe lengthening/tensing: these vowels were always long.This is certainly true historically, as the spelling of these words also suggests. This implies that the vowels of value and evaluate are also long and potentially shorten in eg stimulate stɪ́mjʉwlɛjt > stɪ́mjʉlɛjt > stɪ́mjəlɛjt. The shortening was obligatory before unstressed syllables: stimulus *stɪ́mjʉwləs, except, of course, in prevocalic position: actual áktʃʉwəl.


Preconsonantal happy-tensing, dubbed presume-tensing by Nádasdy (2013), is also difficult to explain in the current framework. There is no obvious reason for inserting a glide between a vowel and a consonant.

If we take another look at unstressed ʉ(w), we may notice that in the environment of presume-tensing, that is, in the first syllable of words the distribution of palatality is the same as in stressed syllables. That is, few speakers would have a yod (or palatality) before the first vowel in words like leukaemia l(j)ʉwkɪ́jmɪjə (cf value váljʉw), superior s(j)ʉwpɪ́ːrɪjə (cf issue ɪ́ʃʉw), Thucydides θ(j)ʉwsɪ́dədɪjz (cf Matthew máθjʉw), and American English has no palatality here in deuterium, neurosis, nutrition, tuition, tutorial, etc, although it would if the first syllable of these words were unstressed.

Nádasdy (2013) calls this phenomenon “Initial-Pretonic Tensing”. However, the distribution of palatality suggests that this is not simply “tensing”, which is not expected in unstressed position, but stressing. So if presume-tensing is indeed real,We have watched dozens of British pronunciations of presume on Youglish without coming across a single instance of prɪj-. it does still not exemplify strengthening in an unstressed syllable. Which means that we can answer with Vincent: “Yeah, we happy!”

Manee happee reeturns, Adam!


Gimson, Alfred Charles. 1989. An Introduction to the Pronunciation of English (4th ed., revised by Susan Ramsaran). London: Edward Arnold.

Harris, John and Jonathan Kaye. 1990. A tale of two cities: London glottalling and New York City tapping. The Linguistic Review 7: 251–274.

Jones, Daniel. 1924. An English Pronouncing Dictionary (3rd ed.). London and Toronto: J. M. Dent & Sons.

Jones, Daniel. 1960. An Outline of English Phonetics (9th ed.). Cambridge: CUP.

Kenyon, John S. and Thomas A. Knott. 1953. A Pronouncing Dictionary of American English. Springfield, Mass.: Merriam-Webster.

Nádasdy, Ádám. 1989. Unstressed vowels today and tomorrow. In: Lehel Vadon (ed.), Studies in English and American Culture. Proceedings of the Conference on English and American Studies, 5–7 April 1989. Eszterházy Főiskola, Eger, 259–267.

Nádasdy, Ádám. 1993. Unstressed and partially stressed syllables in English words. Doctoral dissertation, MTA, Budapest.

Nádasdy, Ádám. 1994. Partially-stressed syllables and unstable /i/ in English. The Even Yearbook 1: 57–86.

Nádasdy, Ádám. 2013. presume-tensing” and the status of weak /i/ in RP. In: Péter Szigetvári (ed.), VLlxx: Papers Presented to László Varga on his 70th Birthday. Budapest: Tinta, 363–381.

Sweet, Henry. 1900. A New English Grammar: Logical and Historical. Oxford: Clarendon Press.

Szigetvári, Péter. 2016. No diphthong, no problem. In: Jolanta Szpyra-Kozłowska and Eugeniusz Cyran (eds.) Phonology, its faces and interfaces. Frankfurt am Main: Peter Lang. draft version

Szigetvári, Péter and Miklós Törkenczy. 2011. A thistle, or how to tell a reduced vowel. In: Bálint Gárdos, Veronika Ruttkay, and Andrea Timár (eds.), Rare Device: writings in Honour of Ágnes Péter. ELTE Papers in English Studies. Budapest: Eötvös Loránd University. 503–513.

Wells, John Christopher. 1982. Accents of English. Cambridge: CUP.