logo  ELT Concourse teacher training
Concourse 2

Teaching troublesome sounds

trouble
something doesn't sound right

For adult learners, in particular, any ambition to sound like a native speaker is probably unrealistic but all learners need to be comprehensible and none wants to put a strain on people's listening comprehension by the false production of the sounds of English.  Even minor adjustments to, for example, the place in which /t/ and /d/ are pronounced or the way a schwa is produced can have a noticeable, immediate effect on helping learners lose the most obvious elements of a foreign accent.

A good source for specific languages and their sounds systems is Swan and Smith (2001).  That reference takes languages one by one and sets out which sounds are most troublesome in English for speakers of those languages, although it does it in a frustratingly inconsistent manner.  (If you are looking for issues affecting a particular language on this page, use Ctrl + F to search it.)
This guide takes a different approach and sets out to look at a range of the phonemes of English, explain what makes them difficult for speakers of certain languages and then suggest some simple teaching techniques which may help you help your learners pronounce things more accurately.  It also considers some languages lacking from Swan and Smith and excludes others they cover.

There is no attempt to be exhaustive here either in the sounds selected or in the languages mentioned.  A completely exhaustive list would fill many books so you will need to rely on some research of your own into what phonemes cause your learners the greatest difficulties.  If you are a speaker of your learners' first language(s), of course, that's easy.  Teachers working with multi-lingual groups face more of a challenge.  This guide is here to help.
Some of the solutions suggested are applicable beyond the specific focus here, e.g., the suggested solutions to the issue of voicing can be applied to a wider range of consonant sounds than are considered in what follows.

In what follows, we are considering the standard forms of the languages mentioned.  Individual speakers of any language will vary their pronunciation from the standard form and may, in fact, be fortunate that their accent or dialect actually does have a sound equivalent to the English phonemes.


vowels

Vowels


Phonemes Examples The issues The advice
/iː/ and /ɪ/ as in beat and bit This is a notorious issue because many languages do not distinguish the short and long sounds represented by these transcriptions or do not lengthen the /i:/ sound sufficiently.
The short sound is slightly more central and lower in terms of tongue position than the long sound.
There are two groups of languages that this affects and the effect of both is that learners fail to produce the short /ɪ/ sound.
  1. Those having only the long vowel, /iː/:
        French
        Italian
        Greek
        Farsi (Persian)
        Chinese languages
  2. Those only having a /i/ sound which is intermediate between /i:/ and /ɪ/ as at the end of happy (/ˈhæp.i/) in English:
        Swedish
        Spanish and Catalan
        Portuguese
        Polish
        Turkish (which also has the short sound)
        Malay / Indonesian
        Japanese
        Korean
If your learners' language falls into the first group, the short vowel needs to be taught from scratch.  Stage 1 is recognition and that can be achieved simply by getting the leaners to notice the shape of the lips, stretched laterally for /i:/ and relaxed for /ɪ/.
The old trick of getting people to smile and say cheese when having a photograph taken usually works well to alert learners to the long sound they can already produce but it is the short sound that causes problems for learners from these language backgrounds.  Start with the long sound (which they can easily produce) and then remove the smile, relaxing the lips (or even rounding them slightly) to make the short sound.
Try getting learners to breathe out completely and empty their lungs before then producing the sound.  With no available breath, it is possible to produce /ɪ/ but not /i:/.
An alternative approach is to get the learners to say the /i:/ sound and while extending it, saying eek!, lower the tongue and move it towards the back of the mouth slightly to shorten it, relax the lips and then cut off the breath.
The second group of learners who actually do have a longer sound /i/ but usually not the short /ɪ/ tend to produce /i/ in both cases, failing to produce the short sound and failing to lengthen the long sound sufficiently.  They can be encouraged to locate words in their first language(s) which have the /i/ and then simply elongate it to produce the /i:/ sound effectively.
For the short sound, try the techniques above.
For both sounds, try some of the ideas for minimal-pair work (see the teaching suggestions below).
/ʊ/ and /uː/  as in put and loose Speakers of languages which have trouble with these sounds in English fall into three groups: those that have /uː/ but not /ʊ/, those that have /ʊ/ but not /uː/ and those that have an intermediate sound like /u/ which is neither as short as /ʊ/ nor as long as /u:/:
  1. Those having only the long vowel, /uː/ but not the short /ʊ/:
        Dutch
        French
        Portuguese
        Farsi (Persian)
        Malay / Indonesian
        Czech
  2. Those only having only the short vowel /ʊ/ and not the longer /uː/:
        Italian
        Spanish and Catalan
        Greek
  3. Those that have an intermediate sound /u/:
        Portuguese
        Polish
        Turkish
        Japanese
        Korean
        Polish
        Chinese languages
        Thai
Speakers of languages which do not have the short sound encounter the most trouble.
The lip rounding is more or less the same for both sounds so it is only the fact that the tongue is raised and placed further back in the mouth which distinguishes /u:/ from /ʊ/.  One way to show the difference is to get learners to produce an elongated /u:/ sound and while doing it, move the tongue downwards and slightly towards the front of the mouth.  It takes a good deal of practice to get it right.
For the second group, it is worth noticing that almost all languages have an expression of disgust or surprise such as Pooh! or Oooh! and that can be dragooned into getting learners to produce the longer vowel /u:/ by exaggeration and then shortening it slightly to be more realistic.
The third group have a tendency to produce the same /u/ sound in all words and they need a combination of the two techniques above to get it right.
First get them to say the surprise sound as in Oooh! and then to move the tongue downwards and slightly towards the front of the mouth while simultaneously cutting off the airflow.
/ə/ the schwa as in about This is the commonest vowel sound in English and it's important for people to get it right or they will sound too precise in many cases or fail to pronounce the final sound correctly.
Languages which do not have the schwa include:
    Scandinavian languages
(speakers of which have something similar but do not reduce the sound sufficiently)
    Italian
    Spanish (but not Catalan)
    Portuguese
    Greek
    Czech
    Polish
    Arabic
    Turkish
(which has something similar but whose speakers tend to make the sound too high in the mouth, tending to /ɪ/)
    Japanese
Speakers of languages which do not have the schwa rarely have trouble making the sound because it is completely central.  A beginning is simply to get the sound right as in the beginning of words like alive, asleep, around etc. or at the end of words like father, doctor, flavour etc.
It is slightly more troublesome to get learners to produce the sound in the central position in words such as roundabout, photographer, gardener (/ˈraʊnd.ə.ˌbaʊt/, /fə.ˈtɒ.ɡrə.fə/, /ˈɡɑːdə.nə/) etc. where the tendency, influenced by spelling, is to give the letter its full pronunciation and produce /ˈraʊnd.eɪ.ˌbaʊt/, /fɒ.ˈtɒ.ɡræ.fə/ and /ˈɡɑːde.nə/ respectively.
The issue is also one of timing and stress patterns across phrases and clauses and that needs handling in connected speech rather than in isolation because that is where the sound is most frequent.
In particular, the reduction of function words such as and, but, of etc. needs focused attention.
/ɜː/ as in worse There is a range of issues here with some languages pronouncing the phoneme with rounded lips which makes it tend too far towards /u:/.  These include Dutch, Danish, Swedish and Norwegian as well as some dialects of German
Other languages have no equivalent sound, including:
    Italian
    Spanish and Catalan
    Portuguese
    Greek
    Czech
    Russian
    Polish
    Farsi (Persian)
    Arabic
    Turkish
    Japanese
    Korean
    Chinese languages
For lip rounding issues, it is usually enough to demonstrate the lip position with the sounds for speakers of Dutch and the other languages to be able to adjust.
Other solutions follow on from the solutions to the schwa, above, because the sound is slightly longer and the tongue position slightly lower than for /ə/.
One solution is to pair the sounds so we can have about and urban (/ə.ˈbaʊt/ and /ˈɜː.bən/) and once the pronunciation of the /ə/ has been satisfactorily mastered, the pronunciation of /ɜː/ becomes much easier.
Lengthening /ə/ makes an acceptable substitute for /ɜː/.
/æ/ and /ʌ/ as in cap and cut These sounds have no equivalent in many languages and speakers will often produce the nearest familiar sound.
  1. The replacement for /æ/ is often /e/ for speakers of:
        Dutch
        Scandinavian languages
        German
        French (also replaced with /ʌ/ or /ɑː/)
        Italian
        Russian
        Farsi (Persian)
        Turkish (also replaced with /ɑː/)
        Thai
        Chinese languages
        Korean
        Czech
  2. The replacement for /æ/ is often /ɑ/ for speakers of:
        Spanish and Catalan
        Portuguese
        Greek
        Japanese
        Arabic
  3. The replacement for /ʌ/ is usually /ɑː/ for speakers of:
        Swedish
        Norwegian
        German
  4. The replacement for /ʌ/ is usually /ɒ/ for speakers of Dutch and Danish.
  5. French speakers may replace /ʌ/ with /ɜː/.
  6. Italian and Portuguese speakers may replace /ʌ/ with /æ/
These are both low front vowels so tongue height is important to get right.  For speakers of most languages, the sound /ɑː/ is not too problematic.
Getting the learners, therefore, to produce the back vowel /ɑː/ and then move the tongue forward in the mouth should result in correct pronunciation of /æ/.
Once that is mastered, the tongue needs to be pulled slightly back and raised a little to produce /ʌ/.
Speakers of languages which have the /ʌ/ sound are more fortunate because they have merely to lower the tongue and bring it slightly further forward to make /æ/.
/ɔː/ and /əʊ/ as in bought and boat  This is a common confusion and speakers of languages which do not have the /ɔː/ with often substitute the /əʊ/ diphthong or vice versa.  There are three groups here:
  1. Languages with no /əʊ/ so a tendency to produce /ɔː/ instead:
        Scandinavian languages
        Italian
        Portuguese
        Farsi (Persian)
        Turkish
  2. Languages with no /ɔː/ sound so a tendency to produce /əʊ/ instead:
        Spanish and Catalan
        Russian
        Chinese languages
  3. Languages with neither sound so a tendency to produce either sound randomly or influenced by spelling:
        German
        French
        Greek
        Polish
        Arabic
        Thai
        Japanese
The first issue is that the spelling of the /ɔː/ is very variable in English, occurring in, e.g.,
bought: /bɔːt/
short: /ʃɔːt/
or: /ɔː/
core: /kɔː/
law: /lɔː/
sure (for many speakers): /ʃɔː/
so learners need to focus on sound not spelling and not be influenced by the letters in the words.  Do not write the words on the board until the pronunciation has been mastered.
This is a difficult area because the second of these sounds is a diphthong, combining /ə/ and /ʊ/ and many languages contain neither of these sounds.
Until learners can successfully produce the two pure vowels, they cannot be expected to produce the diphthong so the first step is to get those right (see above for some ideas).
The /ɔː/ is a pure vowel but it has no equivalent in many languages.  It is a very low back vowel and similar to the /ɑː/ which many learners can successfully produce.  The difference mostly lies in lip rounding and that can be made clear by miming the sound and getting learners to notice the lip positioning before they try to produce the sound itself.


consonants

Consonants

There are four main issues with the pronunciation of consonants in English:

  1. Some languages do not contain consonants which are common in English and some contain more that English does not have.  There is a natural tendency for learners to substitute near equivalents which may be, but are often not, English consonants.
  2. Consonants may occur with one or more allophones in English which are full phonemes in other languages.
  3. Consonants which are full phonemes in English may be allophones in other languages so learners may not see the need to pronounce them differently or perceive the difference at all.
  4. Phonotactic rules often mean that certain combinations of consonants are possible in English but forbidden in other languages and vice versa.  For more, see the guide to the syllables and phonotactics.
The issue The description The advice
aspiration The phonemes /p/, /t/ and /k/ appear in English in at least two main forms: aspirated (/pʰ/ /tʰ/ and kʰ/) and non-aspirated (without the /ʰ/).
For example:
    take: /tʰeɪk/
    pike: /pʰaɪk/
    kite: /kʰaɪt/
    start: /stɑːt/
    pickle: /ˈpɪk.l̩/
    blacken: /ˈblækən/
The allophones carry no semantic information in English.  Other languages do things differently and in these the aspirated and non-aspirated forms are phonemic, carrying a distinction in meaning.
In most Chinese languages, including Mandarin and Cantonese, the aspirated and non-aspirated versions of /t/, /k/ and /p/ are full phonemes distinguishing minimal pairs.
Similar considerations apply to Korean.
In Thai, too, the two forms of /p/ and /t/ are also phonemes.
In other languages, these sounds are never aspirated or always aspirated so there will be a tendency by their speakers to produce the same sound in all positions, rendering, e.g.:
    cake as /kʰeɪkʰ/ or /keɪk/
    pip as /pɪp/ or/pʰɪpʰ/
    tot as /tɒt/ʰ /tʰɒtʰ/
etc.
Speakers of Japanese, Polish and Russian, for example, may do this.
A number of languages do not produce aspirated versions of these consonants in the initial position making /p/, /k/ and /t/ sound closer to /b/, /ɡ/ and /d/, respectively.  This affects speakers of:
    Dutch
    Spanish
    Portuguese
    Russian
Learners whose first languages have both the aspirated and non-aspirated forms of these consonants have no difficulty producing them but some in knowing where to do so.  Practising the initial-position aspirated allophones is a starting point and being aware of when the learners produce the inappropriate sound is helpful.
Learners whose first languages do not have aspirated forms need more careful treatment.
Holding a thin piece of paper in front of the mouth will determine whether a consonant is produced with or without aspiration.  In the former case, the paper should move slightly as breath is released.
Learners whose first languages have aspirated versions but who do not use them in the initial position as English usually does can also be trained this way and the upshot is often that they sound considerably less foreign.
voicing and de-voicing English does not routinely de-voice final consonants so, e.g.:
    pad is pronounced as /pæd/
    tub as /tʌb/
    fog as /fɒɡ/
    gaze as /ɡeɪz/
    ridge as /rɪdʒ/
    smooth as /smuːð/
    love as /lʌv/
Other languages routinely de-voice the final consonant so, e.g.:
    tub sounds like /tʌp/
    fog sounds like /fɒk/
    gaze sounds like /ɡeɪs/
    ridge sounds like /rɪtʃ/
    smooth sounds like /smuːθ/
    love sounds like /lʌf/
Learners from the following language backgrounds may do this:
    Dutch
    German
    French (although this may be because French speakers do not usually lengthen the vowel before voiced consonants)
    Spanish and Catalan
    Turkish
    Czech
The issue is usually quite easily fixed because these languages do have the voiced equivalents, making it simple for the speakers of them to pronounce the sounds.
However, unless their attention is drawn to the issue many learners persist, even at higher levels, in de-voicing final consonants with a consequence that they contribute to sounding more foreign than they should.
The first issue is for learners to be able to distinguish between voiced (lenis) and unvoiced (fortis) sounds.
One way is to start with a sound that can be extended and get learners to say firstly SSSSSSSSSS and then the voiced sibilant ZZZZZZZZ with their hands on their throats.
In the first case, there should be no vibration of the vocal cords and in the second, the learners should perceive the vibration.  Then they can go on to try something like:
    SSSSSSZZZZZZSSSSSSZZZZZZSSSSSS
This is trickier with stops like the distinction between /b/ and /p/ or fricatives such as /f/ and /v/ but can still be done with a little practice.
allophonic differences Sounds which are full phonemes in English may be allophones in other languages and learners may simply not see the need to insert the correct phoneme and may also be unable to perceive it.
In Arabic, for example, /b/ and /p/ and /f/ and /v/ are allophones so differences between pat and bat and fat and vat may simply be inaudible.
For many Japanese, Korean and South Chinese speakers /r/ and /l/ are distinguishable, especially in very slow, careful speech, but are allophones rather than phonemes.  See below for more on /l/ vs. /r/.
Some care needs to be taken to distinguish the sounds and get learners to notice the semantic effect of replacing /b/ with /p/ and/f/ with /v/ in English.
The fact that the sounds are allophones means, that they can be produced but may be in free variation depending on the learners' dialect and personal accent.  It is, therefore, more of an awareness-raising issue than one of production.
θ and ð
as in thank and this
These two sounds are not unique to English but few languages have both and most have neither.
The usual effect is that /θ/ is rendered as /f/ or /s/ and /ð/ is as /d/ or /z/.
Greek has both sounds and this causes speakers few problems in English.
Arabic has both sounds but only in literary forms; colloquial speech renders them as /t/ and /d/.
Spanish (European but not usually American varieties) and Catalan have only /θ/ so the tendency is to use it when /ð/ is more usual.
There is no quick fix for this because the sounds are unfamiliar and it takes a good deal of practice to produce either and even more to produce both appropriately.
The key point is that the sounds are both described a dental fricatives but /ð/ in particular can be produced as an alveolar sound.
The usual distinction between voiced and unvoiced versions can be made by getting learners to place a hand on their throats and feeling the vibration on the voiced sound as one would for the more usual and common /s/ and /z/ sounds.
/t/ and /d/ This unvoiced-voiced pair exists in many languages but English is slightly unusual in producing them as alveolar rather than dental stops (with the tongue touching the gum ridge rather than the top teeth).  Learners with these first languages may produce the sounds dentally and that contributes to a foreign accent:
    Dutch
    German
    Italian
    Russian
    Polish
Speakers of some south Asian languages often produce a retroflex sound (made with the tip of the tongue curled back) in place of /d/ and /t/ (transcribed as [ɖ] and [ʈ]).
Swedish and Norwegian speakers use the retroflex [ʈ] after /r/.
Cure is simple once learners have had the alveolar and non-retroflex nature of the English sounds explained with the aid of a diagram, where necessary (see below).
Lots of practice is, however, required to make the process automatic.
It is rare for the confusion between dental and alveolar or retroflex pronunciations to be the cause of communication breakdown but both contribute to the perception of a foreign accent in English so it is worth taking a little time with this.
/l/ and /ɫ/ English is slightly unusual in have both allophones of /l/, the dark and the light or clear.
Speakers of languages which do not have a dark form (/ɫ/) will produce a light one instead.  They include speakers of:
    Scandinavian languages
    German
    French
    Russian
    Farsi (Persian)
    Thai
    Japanese
    Korean
    Some Chinese languages
The solution is not simple and the issue is often ignored because using the wrong allophone will not, by definition, be a cause for non-comprehension in a hearer.
However, the non-production of the dark form will make speakers sound foreign and the clear sound at the ends of words is often followed by /ə/ and that is particularly noticeable in Italian speakers.
Simply explaining how /ɫ/ is achieved by raising the velum to block the air flow and then practising it can pay dividends.
/l/ and /r/ The non-distinction between these two phonemes is most significant for speakers of Southern Chinese dialects, Korean and Japanese.
The issue is that the sounds, even when they occur are allophones, not phonemes.
The dark /ɫ/ usually causes fewer problems, in fact.
The /r/ sound is troublesome for many because it is not evident in BrE at the end of words like father unless it is followed by a vowel.  For this reason, some learners find an American accent easier to understand in this respect.
Speakers of most European languages may have a much stronger pronunciation of /r/, rarely like English, and tend to over pronounce it.
The problem is serious but not wholly intractable.
Producing /ɫ/ instead of /l/ will not, in English, lead to any confusion for a hearer but the speaker will sound a little odd.
One way to get them to do this is to ask learners to produce /ɫ/ at the beginning of a word like love by extending the consonant sound unnaturally and producing /ɫʌv/ instead of /lʌv/. Then put the word into a short phrase such love letter and get them to start by saying it very slowly, producing /ɫ/, and slowly speed up when it should lead them to making /l/.
In English, both /l/ and /r/ are approximants so there is, in fact, only a small difference between them with the air being slightly obstructed so that it flows at the sides of the tongue for /l/ and between the tongue and the roof of the mouth for /r/.
Here are some ideas:
  1. Get the learners to rest the tip of the tongue against the alveolar ridge but the tongue should not touch the teeth at the sides of the mouth.
    To test this, get the learners to breath in sharply and make sure they can feel cold air passing down the sides of the tongue.
    Now they can start with the tongue in the right position and produce, e.g., love and then lovely and then lovable.
  2. Now you can move on to the /r/ sound by getting the learners to extend the tongue to touch the side teeth and withdrawing it from the alveolar ridge.
    To test this, again get the learners to breath in sharply and they should feel cold air on the roof of the mouth rather than at the sides of the tongue.
    Holding the tongue in that position, the learners should be able to produce run (/rʌn/).
  3. Another trick is to insert a vowel before /r/ which forces the tongue into the right position. Because the sound /u:/ is a high, back vowel, inserting that before a word like run (/u:rʌn/) means that the tongue is already in the right place to produce the sound.
  4. For the /l/ sound, use a low vowel such as /ʌ/ and insert that so that the word lack is pronounced as /ʌlæk/ and that forces the tongue to lie flat in the mouth. You can also use the English word ‘alas’ for this exercise if it is pronounced without the usual initial schwa as /æ.ˈlæs/ rather than /ə.ˈlæs/.
  5. Once they can do this, the learners can stop producing the two intrusive vowels at the beginning and simply imagine them so that the tongue moves to the right place.
  6. Whichever examples of words you choose to use as practice, avoid consonant clusters with the /l/ and /r/ (such as pr, dr, pl, sl and so on) sounds because these are also difficult for most speakers of these languages who will often insert vowels to break them up (see below).
/h/ This phoneme, when it exists, can be troublesome because in English it is always unvoiced and often dropped in connected speech.
Speakers of the following languages may ignore it or pronounce it wrongly, often as a velar continuant /x/ as in the Scottish loch:
French, Italian and Greek have no equivalent so the phoneme is often ignored altogether or substituted by /x/, /k/ or even the voiced /ɣ/.
Some Dutch speakers may substitute /x/.
In Spanish, Catalan, Russian, Polish, Arabic and the Chinese languages: the 'h' is pronounced as /x/ and that is often carried over to English.
In Portuguese, there is no initial /h/ sound so the phoneme may be dropped.
The sound in English is a glottal fricative produced in the throat rather than the velar sound made when saying /x/, /ɣ/, /k/ or /ɡ/.
Speakers of languages which have no equivalent clearly have the most problems and they need training in how it is pronounced.
For other learners, the problem is usually the over pronunciation of the letter as /x/ or even /ɣ/.  This leads to the pronunciation of, e.g., hello as /xe.ˈləʊ/ rather than /hə.ˈləʊ/ so with their first word, some learners have already advertised the fact that they are not native speakers.
One trick is to use a diagram to show learners where the sound is produced (at the far back of the throat rather than with the vellum.  Get learners to produce the vowel /ɑː/ (which most can do) and then to constrict the throat slightly to make the air turbulent.
When teaching distinctions between voiced and unvoiced consonants it makes sense to include the /h/ sound and distinguish it from /x/ and /ɣ/, /k/ and /ɡ/.
In connected speech, some learners may be reluctant to omit the sound in, e.g., give him it (/ɡɪv.ɪm.ɪt/), saying /ɡɪv.hɪm.ɪt/ or even /ɡɪv.xɪm.ɪt/.  Usually, this is simply a lack of awareness.
/b/ and /v/ /b/ and /v/ may cause problems for speakers of Spanish and Catalan because the languages have an intermediate phoneme so the letters may be pronounced indistinguishably. The issue here is not very serious although it contributes to a noticeably Spanish / Catalan foreign accent in English.
For the most part, focus on place of articulation is needed to distinguish /b/ (bilabial) from /v/ (labiodental).  Minimal pair work with words like vote, boat, lover, lubber etc. can pay off.
/ɡ/ Some languages, including Dutch, Arabic and Korean, lack the voiced /ɡ/ so /x/, /k/ or /ɣ/ may be substituted.
Greek has the sound but it is rare and the same effect may be noticed.
The sound is a voiced velar stop formed by cutting off the airflow at the back of the mouth and vibrating the vocal cords at the same time.
For learners whose first languages have a /k/ sound, the solution is simply to make the /k/ and add voice (see above).
For learners who substitute /x/ or /ɣ/, the issue is a bit more serious but one approach is to begin with the familiar fricative and then stop the airflow suddenly, making the /ɡ/ correctly.  Learners with the tendency to use /x/ instead of /ɡ/ will then produce /k/ to which they need subsequently to add voicing.  It is, therefore a two-stage process for them but a one-step process for those using /ɣ/ which is already voiced.
/tʃ/, /ʒ/ and /dʒ/ These sounds, are either fricatives (/ʒ/) formed by restricting the airflow to produce turbulence or affricatives (/tʃ/ and /dʒ/) formed as a combination of a fricative and a stop.  They appear in, e.g.:
    chair: /tʃeə/
    measure: /ˈme.ʒə/
    hedge: /hedʒ/
The sounds /tʃ/ and /dʒ/ are an unvoiced-voiced pair.
Some or all do not occur in a range of languages and the usual substitution is:
    /ʃ/ for /tʃ/ and /ʒ/
    /ʒ/ or /tj/ for /dʒ/
Languages in this category include:
  1. Those with none of these sounds:
        Dutch
        Scandinavian languages (including Finnish)
        Greek
        Japanese
        Tagalog
  2. Languages with no /ʒ/ or /dʒ/ (but with /tʃ/):
        Czech
        German
        Thai
        Spanish (not Catalan)
        Korean
  3. Languages with no /tʃ/ or /dʒ/ (but with /ʒ/):
        Portuguese
        French
  4. Languages with no /ʒ/ (but with /tʃ/ and /dʒ/):
        Italian
        Farsi (Persian)
        Arabic
        Turkish
        Malay / Indonesian
  5. Russian has no /dʒ/ (but has equivalents for /tʃ/ and /ʒ/)
  6. Chinese languages have some related sounds but they are different and sound odd to English speakers.  The sounds which come close are transcribed as /tɕ/, and /tʂ/ which are fricatives akin to /dʒ/ and /ʒ/ respectively.
Given the range of issues here, these sounds need careful practice.
Learners whose first languages fall into group one will have the most difficulty because they have none of the sounds.  For these and other learners, the first step is to use a diagram explaining where they are pronounced (see below).
/tʃ/ and /dʒ/ are affricatives produced by pressing the tongue to the hard palate, stopping the airflow and then allowing turbulent air to flow over the tongue by withdrawing it downwards and backwards.  The first is unvoiced (fortis) and the second is voiced (lenis).
Making the sounds requires, for some learners, a good deal of practice but positioning the tongue as if for a high vowel such as /iː/ or /ɪ/ and then following it with the consonant helps because the tongue is already in more or less the right place.
To practice /tʃ/ start with a full stop sound such as /t/ which most learners have little difficulty making and then get them to move the tongue down slowly until turbulent air produces the following /ʃ/ sound.
/dʒ/ can be practised in the same way for this group by starting with a voiced stop /d/ and keeping the vocal cords in vibration while making the sound.
Learners whose first languages have no /dʒ/ but do have the /tʃ/, can also be led to the same sound by adding voice.
To ensure that learners are able to add or remove voice from sounds at will, start by getting them to say SSSSSSSS followed by ZZZZZZZ and then add and remove voice by saying SSSSSZZZZZSSSSSZZZZZ etc.
Many languages have no /ʒ/ and this sound requires different treatment because it is a full fricative (and can be extended indefinitely).
Start with /ʃ/ which most learners can make and add voice to the sound to produce the /ʒ/. Learners who can say mesher (/meʃə/) can, by adding voice to the sound, produce measure (/ˈme.ʒə/) (eventually).
Because Chinese speakers do have related sounds, the issue is one of adjusting the mouth shape from the Chinese sounds to the English sounds.  Diagrams (see below) will help but practice is needed to get it consistently right.  For Chinese speakers, /tʃ/ can be nearly made by devoicing the Chinese sound /tɕ/.
/ʃ/ Languages without /ʃ/ or an intermediate sound between /s/ and /ʃ/ include:
    Spanish and Catalan
    Greek
    Dutch (some dialects)
    Thai
    Japanese
    Tagalog
Chinese languages (although a similar sound, /ʂ/ exists as it does in Finnish)
Korean has a similar sound but /s/ is often substituted
This is not usually a serious problem although the substitution of /s/ for /ʃ/ contributes to a foreign accent considerably.
The sound is a true fricative and can be extended indefinitely as in, e.g., Shshshshshsh! and is formed by allowing turbulent air to flow between the tongue and the hard palate.
Making the sound is not usually the difficulty but remembering to make it is problematic.  You need to be alert to learners whose first languages do not have the sound that they are not routinely substituting /s/ for /ʃ/.
/v/ and /w/ /v/ and /w/ cause a familiar problem for some German speakers and the influence of spelling is also noticeable.  The German letter 'v' is pronounced as /f/ and the letter 'w' as /v/.  Dutch speakers may also confuse /v/ and /w/.
Scandinavian languages do not have the /w/ and /v/ is often substituted.
Although Farsi (Persian) has both /v/ and /w/, the phonemes are sometimes confused.
The inability to pronounce /w/ often arises from its vowel-like articulation.  Learners can be led to the correct pronunciation by making it a diphthong pronounced something like /u:/ plus /i:/ or /u:/ plus /ɪ/.  Saying it slowly at first helps and then speeding up until the pronunciations of, e.g., whisky (/ˈwɪ.ski./) and wine (/waɪn/) are clear.
Start with /u:.iski/ and /u:.aɪn/ and then go faster.
Simple awareness raising of the spelling issue with /v/ and /f/ is usually enough but this can be a persistent problem for some speakers who are wedded to the written word.
Consonant clusters A number of languages do not exhibit consonant clusters and learners will be tempted, therefore, to insert a vowel, usually a schwa, between the elements of the cluster, producing, for example, /səkəruː/ instead of /skruː/ for screw.
These languages include:
In French, there are no final /z/ sounds in clusters making some plurals and third-person verb forms difficult for French speakers.  The temptation is to replace the sound with /s/, making the plural of hands as /hænds/, not /hændz/, and the third-person verb ending similarly as /həʊlds/, not /həʊldz/ for holds.  The /z/ is not frequent in French but does occur initially so the sound itself is not mysterious.
Spanish and Catalan do not allow /s/ + /p/, /t/ or /k/ initially and speakers may insert /e/ before them, producing, e.g. eschool (/ˈeskuːl/) instead of school (/skuːl/).  Portuguese phonotactic rules lead to a similar phenomenon.
Russian speakers may have problems with certain initial clusters starting with /t/ as they are not permitted in Russian.
Finnish does not allow more than one consonant in the initial position and speakers may tend to reduce such clusters, rendering, e.g., stop as top.
Farsi and Arabic do not have consonant clusters within syllables and vowels may be introduced as above.
Thai and other South Asian languages as well as Japanese and Chinese languages are similar in this respect and the same issue arises.  Thai speakers, in particular will often simplify final clusters, retaining only the first element, so plural endings may be lost.
Korean speakers have fewer problems.
Turkish does not have initial consonant clusters.
Even those languages which allow various clusters vary in terms of what clusters are permitted and where.
For more, see the guide to the syllables and phonotactics.
The issue is the avoidance of an intrusive vowel which may be an initial /e/ or /ɪ/ or an inserted /ə/ or /ɪ/ between elements of unfamiliar clusters.
The solution, which requires lots of practice, is to focus on place of articulation and how the move from, e.g., /s/ to /k/ to /r/ takes place in words such as scream or at the end of marks where it moves from /r/ to /k/ to /s/.
Stretching out the sounds unnaturally is the place to begin so the learners can focus on the individual consonants before gradually making the production faster until the cluster is produced.
Thai speakers encounter particular problems because of the language's phonotactic rules.  It is not the production of the cluster which is usually the issue but the reluctance to place it in the final position.  They need practice starting simply with two-sound clusters such as /ks/, /ts/, /ds/ etc. as at the ends of sacks (/sæks), lots (/lɒts/), lads (/lædz/) etc. before tackling complications with more than two consonant sounds.  English words can end in no fewer than four consonants and that is a real challenge for many learners (not just Thai speakers).
Avoid, therefore, focusing on words like glimpsed (/ɡlɪmpst/) and texts (/teksts/) because that will demotivate and frustrate.
Remember, too, that that in BrE, the 'r' in words like marks (/mɑːks/), carts (/kɑːts/) and lords (/lɔːdz/) is not sounded so these are, in fact, two-consonant, not three-, clusters.
Clusters are also frequently simplified in rapid speech so it unnecessary to trouble learners with the full pronunciation of words like products or camped because the 't' and the 'p' are not usually sounded by native speakers (so we have products as /ˈprɒ.dʌks/ not /ˈprɒ.dʌkts/ and camped as /kæmt/ not /ˈkæmpt/).
The troublesome /ð/ in clothes is also often ignored by native speakers and learners can take the same route (say /kləʊz/, not /kləʊðz/).
Lastly, the /d/ sound before /z/ is often elided by native speakers so the pronunciation of, e.g., sends is /senz/ not /sendz/.  Learners who are tempted to simplify the cluster by removing the last element rather than the middle or first sound run the risk of being misunderstood because the final element is usually more significant.


pair

Minimal pair work

Nearly everything in this guide has focused on pairs of phonemes so, clearly, the place to start in their teaching and practice is with some minimal pair work so that learners can notice:

  1. How the sounds differ
  2. How their own production differs from a model

Raising awareness

Linking the perception of the sounds to physical movements often works well and has the advantage that you can easily see who has identified the right sound and who is still confused.
For example:
    If the sound is long, point at the door.  If it is short, point at the window.
    beat | bit | keep | kip | sleep | slip

and so on.  The same technique can be used for receptive training for any of the pairs of sounds discussed here which cause problems.  For variety, you can ask people to remain seated on one sound and stand for another and so on.
The next obvious step is to hand the activity over to the learners and get them to give the commands and the examples.  This involves them both in production and perception of the sounds and can be done with any pairs of sounds which are troublesome for the group or individuals within it.

Pronunciation teaching is often confined to the realms of modelling and drilling but there is no reason why some deductive focus on place and manner of articulation cannot be effective.  For many learners, it is helpful to know where they should be placing the articulators and which ones they should be using in what way to make sounds that are unfamiliar to them because their first language(s) do not have them.
Most learners need to know this and cannot be expected to infer it simply by being drilled, any more than they can be expected to learn any other elements of the language through repeating them.
Using a diagram such as this one is often the way to do this:
articulators

Projecting this allows you to be able to point out to learners where, for example, they should be putting the tongue tip to make the difference between /t/ and /θ/.  It can also be used to explain which parts of the vocal tract to block to make stops such /k/ and /b/ etc.
For vowels, the position of the tongue (front to back and floor to roof of the mouth) can be demonstrated.

Vowels, in particular, as you will know if you have followed the guide to them, can be classified by four parameters:

  • tongue height
  • tongue position
  • length
  • lip roundedness

A productive exercise is to take vowels which differ in only one characteristic and use them for practice.  That way, learners can focus only on height, position, length or roundedness and are not distracted by having to make multiple changes to distinguish the sounds when they speak.  Use this table as a reference to do that.

Vowel height position length roundedness minimal pairs
/iː/ high front long stretched bit / beat
/ɪ/ high front short stretched
/ʊ/ high back short rounded full / fool
/uː/ high back long rounded
/e/ mid front short neutral errand / around
/ə/ mid central short neutral
/ɜː/ mid central long neutral hurt / hot
/ɒ/ mid back short rounded
/ʌ/ low central short neutral cup / cap
/æ/ low front short neutral
/ɔː/ low back long rounded caught / cart
/ɑː/ low back long neutral

The same kind of exercise is simple to devise for consonant sounds that cause difficulty for your learners. Again, the trick is to contrast the sounds which differ by only one characteristic: voicing, place of articulation or manner of articulation.
Following this principle, contrast, for voiced / unvoiced pairs:

Consonant pair minimal pairs
/p/ and /b/ Peter / beater
/tʃ/ and /dʒ/ chump / jump
/f/ and /v/ fat / vat
/s/ and /z/ sue / zoo
/k/ and /ɡ/ luck / lug
/t/ and /d/ tent / dent
/θ/ and /ð/ breath / breathe
/ʃ/ and /ʒ/ mesher / measure

For manner of articulation, contrast:

Consonant pair minimal pairs
/t/ and /s/ taint / saint
/d/ and /z/ do / zoo
/ʃ/ and /tʃ/ shock / chock
/ʒ/ and /dʒ/ leisure / ledger
/dʒ/ and /j/ jet / yet
/k/ and /ŋ/ sink / sing
/m/ and /w/ mix / wicks

For place of articulation, contrast:

Consonant pair minimal pairs
/p/ and /t/ paint / taint
/p/ and /d/ part / dart
/p/ and /k/ part / cart
/p/ and /ɡ/ pardon / garden
/dʒ/ and /j/ jet / yet
/θ/ and /s/ thing / sing
/ð/ and /z/ do / zoo
/s/ and /ʃ/ sue / show
/z/ and /ʒ/ baize / beige
/m/ and /n/ man / nan
/n/ and /ŋ/ sin / sing

 

Images, too, can be used to raise awareness and aid memories so, for example:

You say:
Which picture is right?
    He is hitting the metal
    He is heating the metal
hit heat heating
hitting the metal heating the metal
Which picture is right?
    We loved the pub
    We liked the pup
pup  pub
pup pub
Which picture is right?
    What's that tag for?
    Who put that tack there?
tag tack
tag tack
Which picture is right?
    That's an interesting lock
    Why did you take a photograph of a rock?
rock lock
rock lock

and so on.
Finding suitable images is not difficult.
Once the learners can recognise the sounds using one of these techniques, they can take over the activity and test each other by making up their own minimal pair sentences.
That requires them both to produce the correct phoneme and to recognise it when their colleagues do.  It is a self-correcting feedback loop, in other words.

Tempting though it may be, the advice is to avoid tongue twisters such as:
    Peter Piper picked a peck of pickled peppers
    Red lorry, yellow lorry, red leather, yellow leather

and so on because they are too hard and tend to demotivate (being designed to make life hard even for native speakers).
A better way is to practise the relevant sounds separately in short phrases such as:
    red robin
    lamplights
    laugh loudly
    liking the pub
    liking the pup
    bought a boat
    have a happy Easter

etc.

A way of practising a range of sounds is to use a grid like this:
minimal pairs activity
Here's the explanation and procedure:

  1. Along the top row and down the left side of the grid are the target phonemes set in minimal pairs (hate, ate, bit, beat etc.)
    You dictate a word by using two of the target words: the first one from the top row and the second from the left-hand column.  For example:
    The word came is identified by selecting bit in the top row, followed by ate in the left-hand column and week is selected by saying bait + pup.
  2. Demonstrate the activity with a simple phrase such as
        on the journey
    To dictate the phrase, you use the words along the top row, followed by the words in the left-hand column so the phrase is dictated as:
        bait plus pet
        bit plus hate
        bait plus ate
  3. When you have checked that the learners have understood the task and got the phrase right, you can move on to dictating longer clauses and sentences such as:
        I went to London last week and I saw her sister on the journey.
  4. Once the idea is clear and you have practised the perception of the target phonemes, you can hand over the activity to the learners.  They then make an accurate sentence using the words in the grid.  For example:
        We saw her mother on the train
  5. In pairs or groups, one learner dictates the sentence to the other(s) but only by using the grid co-ordinates.
  6. Each word in the sentence is selected individually so, for our example a learner will say something like:
        bat plus ate
        boat plus bet
        but plus pub
        bait plus pub
        bait plus pet
        bit plus hate
        bat plus pet
  7. If the production and reception of the sounds has worked, the other learner(s) will have written:
        We + saw + her + mother + on + the + train
  8. The speaker then reveals the original sentence and the learners compare what they have, identifying any errors and noticing which sounds were either incorrectly produced or incorrectly perceived.
  9. Learners take it in turns to dictate their sentences using the grid co-ordinates.
  10. You can amend the grid, of course, in any way you like, putting the minimal pairs for the current target phonemes in the top row and left-hand column and placing vocabulary or function words that you want to revise in the main grid, focusing on other elements of the language as well as practising pronunciation.
    It will mean that your learners are using their lexical, structural and phonological knowledge simultaneously to understand what they are hearing as well as using their knowledge of what is in the main grid to predict what they might hear and that, of course, simulates a real-life listening / speaking event.

All pronunciation activities need to be repeated often because the aim is to automatise the production of the correct sounds.  That takes time and patience both from the teacher and the learners so mixing up activity types to maintain motivation is important.
Little and often is probably more effective than focusing on the sounds for lengthy periods in a lesson.
A combination of the ideas above involving images, information-gap activities and physical movements can help to keep learners motivated.  Using each one of these ideas to focus on, say, two or three troublesome phonemes for your learners each week will pay dividends.



Go to the index for the pronunciation section of the in-service guides


References:
Campbell, GL, 1995, Concise Compendium of the World's Languages, London: Routledge
Swan, M and Smith, B (Eds), 2001, Learner English, 2nd Edition, Cambridge: Cambridge University Press