2. The importance of stress in English

Brown (Reference Brown1977, 48) recalls having been asked by a second language learner of English about /əˈnɪmɪzm/ in King Lear. She failed to understand, thinking the student must have meant anaemia, which did not make sense but had the appropriate stress pattern. It took an appreciable length of time to reach the intended animism (with stress on the first syllable). Incorrect stress, therefore, can mean that native speakers fail to understand the message, hence the importance of getting the stress right (just what ‘right’ means will be taken up below).

Benrabah (Reference Benrabah1997) provides many more examples of instances that have been misunderstood by speakers of British English. Some of these are presented in Table 1. As Benrabah (Reference Benrabah1997, 161), like Brown, points out on the basis of his data, native speakers of British English can hear the stress pattern as defining the intended word, even where the result fails to make sense.

Table 1.

Words misunderstand by speakers of British English

No doubt, some of the misunderstandings illustrated in these examples arise from a mixture of stress and segmental inaccuracy, although Benrabah does not provide segmental information and focuses on the question of stress. My favourite example (not one discussed by Benrabah) is that of a student with a background in a syllable-timed language proposing a vote of thanks to Professor Sir Randolph Quirk, many years ago. The student said something to the effect of Quirk being the most important authority on the English language today. Unfortunately, they stressed the first syllable of important giving /ˈɪmpətənt/.

The examples given here, although notable, should not be taken to imply that all use of incorrect stress patterns will lead to unintelligibility, however. The fairly typical pronunciation of development as /diːvelɒpment/, heard from speakers of many syllable-timed languages or in syllable-timed varieties of English, generally remains perfectly comprehensible to native speakers of British English or other stress-timed varieties. Incomprehensible output is an extreme result of misplaced stress, not an inevitable result. What it is that makes misstressed words incomprehensible remains an open question at this point.

3. How close must puns be?

The canonical pun involves a pair of homonyms, that is, the pronunciation of each of the lexemes involved is identical, at least at the phonemic level, as in (3).

However, absolute homonymy is not required for the pun to be successful. The lexemes involved can differ in consonants, vowels, stress and vowel quality or stress alone, as illustrated in (4).

More subtly, they can also differ in degree of explicitness of pronunciation, as shown in (5).

This brief survey is not intended to provide a typology of puns but merely makes the point that puns can move some distance away from the canonical instance of homonymy by quite some distance. That being the case, the joke in (1) is not necessarily significant in showing a difference in stress. Nevertheless, it illustrates the fact that stress variation may be overlooked by native speakers. Because the imperfect pun demands that listeners work harder to find the point of the joke, it may even be that an imperfect pun is more enjoyable that a perfect one.

4. How accurate is stress in usage?

Talking about ‘wrong’ or ‘inaccurate’ use of stress seems to imply that there is always a ‘correct’ stress which will be used by native speakers themselves. This far from being the case. Stress is often not fixed for speakers of English, in the sense that more than one stress pattern may be considered acceptable (in the same variety or in different varieties) or in the sense that speakers may make one-off errors (meaning that if challenged, they would agree that what they said was not ‘right’) or in the sense that individuals or groups of individuals may use a stress pattern which is not generally considered to be standard.

4.1 Variable stress

There are many words in English which have apparently random alternative stress patterns. Although these are not necessarily listed as a type, they can be found in multiple sources (e.g. Fudge Reference Fudge1984, Cruttenden Reference Cruttenden2001, Wells Reference Wells2008, Bauer Reference Bauer2024). Some illustrative examples are provided in (6).

4.2 Motivated stress variation

As well as there being some fairly random variation in stress patterns, there are also various kinds of quasi-motivated cases of variation. Two distinct patterns will be considered here, although more than one may apply to a given example, and in some cases the patterns may clash. Much of this material is discussed in Bauer (Reference Bauer1994).

The first pattern is a pattern of change towards antepenultimate stress, which appears to have started several centuries ago, but to have become particularly prevalent in the early twentieth century. Some speakers (or some varieties) still retain the conservative pattern, but the change can be seen in reference works, as set out in Bauer (Reference Bauer1994). Individual speakers will typically mix stress patterns found in the ‘conservative’ and in the ‘innovative’ columns; ‘innovative’ simply indicates relative diachronic recency. Some examples are given in Table 2.

Table 2.

Conservative and innovative stress

* In my experience, abDOmen is now used only by medical people.

The second pattern is one that leads to morphophonemic transparency. Either the base changes in line with a derivative or the derivative changes in line with the base, with the result that there is less stress change (see also Cutler Reference Cutler, Reed and Levis2015, 113). There may also be changes in vowel quality as a result. Some examples are given in Table 3, where the first two columns show earlier stress patterns.

Table 3.

Stress pattern variations

* The stressed vowel may be /eə/ or /æ/.

^† See also (8), where the more usual version of contribute is listed.

4.4 ‘Wrong’ stress

Whatever it means to get the stress ‘wrong’, stress can change and what was once considered the only possible stress becomes old-fashioned, odd, and eventually ‘wrong’ itself. The word exquisite used to be stressed on the first syllable, but is now more usually pronounced with second-syllable stress (Wells Reference Wells2008 says that 69% of his panel prefer second-syllable stress, but that figure is now over 30 years old, and things must have changed; moreover, the panel was largely made up of people who were professionally involved in language matters or self-selected, and probably rather more conservative than the population as a whole).

However, there are errors of stress. We can recognise them because they are:

1. (a) rare (see Cutler Reference Cutler, Reed and Levis2015, 114)

2. (b) recognised by the speaker as insecure (guessed) or wrong if the speaker is challenged

3. (c) not consistent in the speech of a given individual

4. (d) or found only in the speech of a single individual

Some such words that I have heard in the media (both from professional broadcasters and interviewees from the general public) and from conversations around me over the last few years are given in (7). These examples are all from native-speakers of stress-timed English, and do not involve contrastive stress.

The source of such errors is not usually clear. With transfer we can assume that the noun with the same spelling is a spoiler, but ignominy, for instance, seems inexplicable. What is interesting, though, is that some of these errors remain comprehensible, while others do not. Innovative can still be understood, while therapeutics becomes incomprehensible, even though both involve putting the stress on the antepenultimate syllable. Cutler (Reference Cutler, Reed and Levis2015, 115) gives some factors which are likely to lead to lack of comprehension, most notably changing the vowel quality in a stressed syllable. She also notes (2015, 116, 119) that swapping main stress and subsidiary stress will not lead to misunderstanding – presumably because the vowels remain unaffected.

It seems likely that many such errors (including segmental ‘errors’ as well as those involving stress) arise because speakers use a word that they have previously encountered only in written form, and the orthography does not give the speaker sufficient information to predict the standard pronunciation. For example, the spelling of the word lever does not provide information on the length of the initial vowel (short in the US, long in the UK) and neither does it provide information on stress (although initial stress is more common than second-syllable stress and if the -er is seen as a suffix, that is typically unstressed). Although there is a tendency to stress the fleece vowel in a final syllable, lichee is stressed on the first syllable, and settee is stressed on the second syllable despite the fact that the double <t> in the spelling seems to require first-syllable stress. In some cases, the orthography is not accurately interpreted (though this seems more relevant with vowel quality, as in lever cited above where both pronunciations are consistent with English spelling rules). In other cases, clashing models from different varieties of English lead to indecision (e.g. the words advertisement, croissant, marshmallow and stalactite usually differ in British and American Englishes [Berg Reference Berg1999; Wells Reference Wells2008]).

There are some categories of error which are particularly common from newsreaders and appear to arise from the fact that the material is being read rather than spoken in conversation. I mention two here, two which are related, although one is, in principle, unpredictable and the other is predictable.

The unpredictable series involves compounds (usually, but not exclusively, noun + noun compounds) where the English rules for stressing the sequences are obscure. Some unusual examples observed in the media, again with no contrastive stress, are given in (8).

The type that is predictable has an -ing form in the first word. Where the -ing form is an adjective, the stress falls on the noun. This is part of a larger pattern where, for instance, FRENCH teacher (‘the teacher teaches French’) has French as a noun and initial stress, but French TEACHer (‘the teacher is French’) has French as an adjective and stress on the noun. When the -ing form is a noun, it takes the stress, and the righthand noun is not stressed (as much). So, for example, we find driving RAIN (driving is an adjective) but DRIVing lesson (driving is a noun). Where the -ing form is verbal, it follows the pattern of the adjectival examples. Occasionally we find instances where it matters (again, attested examples with no contrast).

Automated speech systems are particularly prone to errors of these types, because they have only the spelling to go by, and the spelling does not provide sufficient information to give the ‘correct’ stress patterns.

4.5 The effect of verse: prosodic dissonance

Because of the effect of the prosodic structure of verse, the stress of individual words may be shifted to an unexpected syllable, and this can lead to the wrong stress pattern being demanded. Some examples are given in (10) mainly from Smith (Reference Smith2024), who also provides sound clips, and who sets the question in a wider framework.

5. A wider perspective

Although this investigation began by looking at questions of stress, many of the questions that are raised here are much wider than that. There is a question of how much variation listeners are willing to tolerate before they simply give up, and a wider question still of production errors in phonetic and phonological output, and whether these are caused at the level of planning of an utterance or whether they are caused by articulatory errors. Such questions are far too complex to be answered in a superficial approach to questions of problems for language learners.

One minor extension of the subject, however, may help. A parallel with tone in tone languages seems relevant. Although I have not conducted a serious experiment to check comprehension, speakers of Mandarin, Thai and Vietnamese have all told me informally that they think they would be able to understand a foreign speaker talking their language without tones, although it would be awkward. They all said they would expect the context to allow them to resolve potential ambiguities. (Many thanks to the speakers of these languages who answered my query.)

This observation raises the question of whether the problems identified here with stress also extend to other areas of pronunciation, including segmental errors and intonational errors. I have observed a Spanish speaker not being understood because of using a Spanish [β] in place of a [v] (the [β] was interpreted as [b] or not understood at all). Because Spanish speakers tend to use the same vowel sound to represent English lot, thought and goat vowels, they are often not understood using words that contain these vowels. At the same time, although the same Spanish speakers may not distinguish the kit and fleece vowels, this does not usually cause major problems of comprehension for English speakers. It is not immediately obvious why there is this difference for English listeners, but again we see some errors leading to a greater loss of comprehension than others. Errors of intonation demand a good level of fluency to become relevant, but Dutch and German learners of English are sometimes thought rude when they use their native intonation patterns in speaking English.

6. A suggestion for pedagogical progress

If we all allow context to help disambiguate ambiguous or unclear input (as surely must be the case, because we notice this in noisy surroundings, in people who suffer from some degree of hearing loss and in learners of a second or subsequent language), the question arises as to what factors make some words easy to understand and others hard to understand when we hear (or fail to hear) what we need. To return to the question of stress, what factors influence whether we take stress as a crucial factor in interpreting a word or whether we use segments and are happy to ignore the perceived stress? Why, in other words, is stress not important some of the time, and vital for comprehension some of the time? If learners can avoid disruptive stress patterns, errors in non-disruptive patterns may not matter.

Based on no more than fundamental considerations of well-known principles in psycholinguistics and language teaching, we can consider some potentially relevant factors.

• • The listener’s familiarity with the word concerned, and thus the ability to predict the occurrence of the word, both grammatically and lexically.

• • The frequency of the word, which provides one type of input to the familiarity of the word.

• • The extent to which the stress of the word is predictable for the individual listener because of its general structure (e.g. the presence of suffixes).

• • How much variation there is the stress of similar words (whatever ‘similar’ means in this context for learners).

• • How many words exist in the learner’s vocabulary which are phonological neighbours of the relevant word and how consistent the stress is on those neighbouring words.

• • What vowel is normally in the syllable which receives the stress; if it is normally [ə] the wrong stress may be more disruptive than if the same full vowel is retained.

• • The spelling, since English orthography does not have a way of specifically marking unstressed vowels.

Empirical research into the value of such features in the classroom situation holds out possibilities for improving outcomes (reducing disruptive stress patterns from learners), and may be generalisable to other aspects of pronunciation practice. In practical terms, the teaching of stress and stress patterns may be of less importance than we tend to believe because, as long as other features of pronunciation are not impenetrable, listeners will be able to understand words despite errors of stress in most instances.

If, as was reported above, variation of stress on syllables with full vowels is not disruptive for native listeners, it may be, however counter-intuitive it may appear, that learners should be taught where the unstressed syllables are, rather than where the stressed syllable are.

These proposals, if considered as serious options for helping with the reduction of problems in pronunciation by second or foreign language learners, rather than just suggestions based on superficial observations, could lead to new approaches to pronunciation teaching.