Connected Speech Phenomena with Assimilation, elision, and insertion

This section of the English Phonetics Academy designed to provide guidance and practice in English pronunciation through the use of phonetics. The exercises and drills included will aid in reducing accents and improving overall speech sound production. These lessons are particularly beneficial for non-native speakers looking to improve their English-speaking skills.

13. Assimilation, elision, and insertion

In connected speech, neighboring segments may affect each other in various ways, so as to facilitate articulation. In the discussion of the realization of individual phonemes in the preceding chapters we have already looked at numerous examples of this type of interaction within words. Some of the main processes involved are assimilation, elision, and insertion.

13.1 Connected Speech

In connected speech, neighboring segments may affect each other in various ways, so as to facilitate articulation. In the discussion of the realization of individual phonemes in the preceding chapters we have already looked at numerous examples of this type of interaction within words. Some of the main processes involved are assimilation, elision, and insertion.

We speak of assimilation when a segment becomes more similar to a neighboring segment, of elision when a segment is left out, and of insertion when a segment is added. Since all three processes generally serve to save articulatory effort, they tend to be more common in casual styles and in rapid speech than in formal, deliberate styles.

Although they are here discussed under the heading of connected speech, this should not be taken to imply that these processes primarily affect segments situated at word boundaries. They are equally likely to occur within the word. The point, rather, is that they are less likely if a single word is pronounced in isolation, in what is called its citation form.

This form of pronunciation is usually the only form given in (pronouncing) dictionaries. For example, the words statement, postcard, just and facts are transcribed – depending on the notation used in the dictionary one happens to consult – something like /ˈsteɪtmənt, ˈpoʊstkɑrd, d͡ʒʌst, fækts/, and that is also how one would expect them to be pronounced in isolation.

However, in connected speech, the pronunciations /ˈsteɪpmənt, ˈpoʊskɑrd, d͡ʒəs, fæks/ are much more likely. Don’t be misled by the fact that native speakers are frequently inclined to dismiss assimilations as ‘slipshod’ or ‘not proper’.

Such judgments are largely based on the erroneous assumption that words should be pronounced in conformity with the spelling. The fact of the matter is that the use of citation forms in connected speech will easily create an impression of hypercorrectness and unnaturalness in all but the most formal speech styles.

Turning to your role as a hearer, it is important to see that a knowledge of some of the main features of connected speech combined with adequate exposure to spontaneous speech will improve your listening comprehension.

It may come as a surprise that native speakers use an utterance like /aɪˈkæn ˈænsər ðæt kwɛstʃən/ with two diametrically opposed meanings: (1) I can’t answer that question (i.e. I’m unable to do so) and (2) I can answer that question (i.e. ‘You may think you’ve got me there, but I am able to answer that question’. Yet it is ultimately no stranger than the kind of ambiguity found on the lexical level where words like bank and face have several distinct meanings.

13.2 Assimilation

Assimilation occurs when a sound comes to share one or more features with a neighboring sound. The neighboring sound may either appear to the left or to the right of the segment that is affected. When the affected sound precedes the one that causes the assimilation, we speak of regressive or anticipatory assimilation.

An example is /n/ in ten books, which tends to become bilabial /m/ under the influence of the following bilabial /b/. If the sound that is assimilated follows the sound that it becomes more similar to, we speak of progressive or perseveratory assimilation. This occurs for example in disguise, where /g/ will be initially devoiced under the influence of the preceding voiceless /s/.

The assimilation may be complete or partial. Complete assimilation occurs when a sequence of two identical sounds is created, as in ten minutes or goodbye, where /n/ and /d/ may be replaced with /m/ and /b/, respectively. Such sequences are sometimes simplified so that only one sound is retained. The assimilation of /n/ in ten books is of course only partial, since the bilabial nasal /m/ still differs from the bilabial stop /b/.

What kind of features are involved in assimilations? Typically, the articulatory features that one segment may pass on to a neighboring one are:

    1. The place of articulation. In ten books it is the place of articulation of /b/ that is transferred to /n/ (assimilation of place).

ten books

    1. The state of the glottis. In disguise it is the state of the glottis that is transferred from /s/ to /g/ (causing it to be [g̥]). (assimilation of voicelessness). Likewise, in AN asbak, the /b/ may voice the /s/, causing it to be /z/ (assimilation of voice).



    1. The position of the velum. In moon, the /n/ may pass on its nasality to the preceding vowel (assimilation of nasality).


13.3.1 Assimilation of Alveolars

The assimilation of GA /t,d,n,l/ to the place of articulation of a following consonant is probably the most common of the assimilation in GA. They can become bilabial, labiodental, dental, postalveolar and velar, depending on the place of articulation of the consonant that follows. The assimilation under b) and c) below will also occur in formal styles, and should definitely be adopted by the foreign speaker.

a) Alveolar /t,d,n/ become bilabial /p,b,m/ before /p,b,m/.

/t/ becomes /p/ footpath, football, treatment, got better, that pen, that book, that man, it missed



/d/ becomes /b/ deadpan, goodbye, headband, admit, good pens, good books, good mengood people



/n/ becomes /m/ input, sunbath, inmate, unpleasant, one pen, one book, one man, one blank



b) Alveolar /t,d,n,l/ become dental [t̪,d̪,n̪,l̪] before [θ,ð].

/t/ becomes [t̪] eighth, wet through


wet through

/d/ becomes [d̪] width, hide that


hide that

/n/ becomes [n̪] enthusiasm, in there


in there

/l/ becomes [l̪] although, well then


well then

c) Alveolar /t,d,n,l/ become post-alveolar [t̠,d̠,n̠,l̠] before /r/. Note that /tr,dr/ are post-alveolar affricates.

/t/ becomes [t̠] entry, ratrace



/d/ becomes [d̠] dry, headroom



/n/ becomes [n̠] Henry, in red


in red

/l/ becomes [l̠] ultra, he’ll run


he’ll run

d) Alveolar /t,d,n/ become velar /k,g,ŋ/ before /k,g/. Of course, if /t/ is glottaled, as in [naɪʔˈkæp͜ʔ], there is no assimilation.

/t/ becomes /k/ nightcap, shotgun, that car, that guy



/d/ bcomes /g/ redcoat, mudguard, good cars, good guys



/n/ becomes /ŋ/ income, engage, one car, one guy



e) Alveolar /s,z/ become palate-alveolar /ʃ,ʒ/ before /ʃ/.

Examples are this ship, these ships: /ðɪʃ ʃɪp/, /ðiːʒ ʃips/.

this ship

these ships

Less commonly, this happens before /j/, as in this year, these years, where you may not want to apply the assimilation yourself, but should be prepared to hear it in the speech of native speakers. If you do use it, be careful not to turn /zj/ into /ʃʃ/ or even /ʃ/ in cases like these years, which at best would assimilate to /ðiːʒ jɪrz/.

this year

these years

13.3.2 Assimilation of Bilabials

Bilabial /m,p,b/ become labiodental [ɱ,p̪,b̪] before /f,v/. The alveolar /n/ may also become [ɱ] in this context: comfortable, infant, shopfloor, obvious, in vain, on fire.



13.3.3 Assimilation of /θ,ð/

The assimilation of GA /θ,ð/ to /s,z/ is characteristic of informal speech:

    1. Dental /θ/ becomes alveolar /s/ before /s,z/: both sides, both zoos

both sides

    1. Dental /ð/ becomes alveolar /z/ before or after /s,z/: Who’s that? What’s the score? How’s that?

who’s that

13.4 Assimilation of Nasality

The alveolar stop /d/ becomes nasal /n/ before /n/ as in wooden, didn’t, had not. Thus, We didn’t have a wooden table may be pronounced /wi dɪnː hæv ə wʊnː ˈteɪbl̩/ in casual speech.

We didn’t have a wooden table

13.5 Palatalization

Alveolar stops /t,d/ + Palatal approximant /j/ in a weak syllable become palatoalveolar affricates /t͡ʃ,d͡ʒ/.This kind of assimilation, in which a single new sound arises from the combination of two sounds, is called coalescent assimilation: don’t you(r), can’t you(r), did you(r), would you, situation, educate.

don’t you

would you



13.6 Assimilation and Elision

Although all types of assimilation could be said to involve a saving of articulatory effort, it is not the case that all languages make these economies in the same way. In other words, we should not expect the same sound combinations to undergo the same assimilations in all languages.

Indeed, it is quite common even for two accents of the same language to differ in this respect. For example, speakers of Eastern accents will generally voice a word-final stop when followed by a vowel, as in Hoe laat is ’t?, ook al, spitsuur, where /t,k,ts/ may become /d,g,dz/.

This type of voicing may affect fricatives, as in Pas op!, huisarts, hoefijzer, where /s,f/ may become /z,v/. Assimilations, in other words, are variety-specific. This is not to say that different languages or accents will not share any assimilations. In fact, quite a number of the assimilations discussed above, especially those involving /n/ and /s/. But there are also important differences and some of these are discussed below.

The reason for this brief discussion of assimilation processes lies in the ease with which assimilations are transferred from the native language to a foreign language. Assimilations and elisions are largely automatic and subconscious. The same applies to the transfer of these processes to a foreign language.

A speaker in whose accent has /z/ will not be particularly sensitive to the oppo- sition /s/ – /z/ in similar contexts in English, and will not only tend to say /ˈlɛz ˈiːzi/ and /ˈðɪz ɪz/ for less easy and this is, but may also find that this particular assimilation is difficult to avoid. In fact, it is usually the case that in order to avoid undesirable assimilations in English, you first need to be able to hear them in in your native language.

13.6.1 Regressive Voicing of Obstruents

Voiceless obstruents are voiced before voiced stops, i.e. /p,t,k; f,s,ʃ,x/ become /b,d,g; v,z,ʒ,(ɣ)/ before /b,d,(g)/.

/p/ becomes /b/ opdracht, abdij

/t/ becomes /d/ voetbal, bloedbank

/k/ becomes [g] zakdoek, ik ben

/f/ becomes /v/ leefbaar, afdak

/s/ becomes /z/ frisdrank, huisbaas

/ʃ/ becomes /ʒ/ finishdoek, douchebak

Voicing of /x/ (becoming /ɣ/) only occurs in Southern fewer accents.

Instead of regressive voicing, many speakers have progressive devoicing of /d/ in function words like de, die, deze, dat, daar. Examples are /ɔp ˈtɪ mənir, ʋɑt ˈkɔs tɑt, ʋɑt ˈmutɑtaˑr/ for op die manier, wat kost dat?, wat moet dat daar?.

In some areas, the fricatives /f,s,ʃ/ may also be voiced before sonorants, as in wijsneus, mislukt, afrekenen /ˈwɛiznøˑs, mɪzˈlʏkt, ˈɑvreˑkənə/.

Advice for learners

Remember that fortis stops tend to be glottalized before a following consonant in GA, and that lenis obstruents are initially devoiced after a voiceless sound. Compare AN /ˈfudbɑl/ (or /ˈvudbɑl/) with GA /ˈfʊtbɑːl/, which is variably [ˈfʊt͜ʔb̥ɑːɫ, ˈfʊp͜ʔb̥ɑːɫ] or [ˈfʊʔb̥ɑːɫ]. So make sure you pronounce a glottalized (or, for /t/, a glottaled) fortis stop in words like football, background, and a strongly articulated voiceless fricative in words like baseball [ˈb̥eɪsb̥ɑːɫ]. This will normally cause the following lenis stop to be initially devoiced, as it should be.




13.6.2 Progressive Devoicing of Fricatives

Voiced fricatives are devoiced after voiceless obstruents, i.e. /v,z,ʒ,(ɣ)/ become /f,s,ʃ,x/. Note that, with the possible exception of a loanword like exact /ɛgˈzɑkt/, syllable-final obstruents will always be voiceless.

/v/ becomes /f/ houtvuur, advise

/z/ becomes /s/ opzij, afzien

/ʒ/ becomes /ʃ/ ’t journaal, dit genre

The rule also applies to /ɣ/, if present initially in Southern english

Speakers from the West of the country will frequently avoid progressive devoicing of /z/ and /ʒ/, and retain a voiced sound. Since /s/ for /z/ in non-assimilating contexts, as in gezakt (i.e. /xəˈsɑkt/), is heavily stigmatized, the failure to devoice /z/ in an assimilating context like niet ziek, op zee, may be looked upon as a case of overgeneralization, whereby what is essentially a spelling-based rule is applied across the board to any /s/ which is spelled z, regardless of the phonetic context.

A similar explanation can be adduced for pronunciations like /ət ʒurˈnaˑl/ for het journaal (standard /ət ʃurˈnaˑl/).

Advice for learners

You may find that your efforts to pronounce a voiced rather than a voiceless fricative in phrases like this village, if they, eight votes result in pronunciations like */ðɪz ˈvɪləd͡ʒ, ˈɪv ðeɪ, ˈeɪd ˈvoʊts/, where the preceding obstruent is voiced as well. Try to avoid this kind of overcompensation by making sure that a syllable-final fricative remains voiceless and a syllable-final stop is glottalized.

Note, on the other hand, that in words like advice, obvious, exam both obstruents are fully voiced in English.




13.6.3 Word-Final Voicing

Word-final voiceless fricatives preceded by a voiced sound are voiced before a vowel. That is, AN /f,s,ʃ,x/ become /v,z,ʒ(ɣ)/ between a voiced sound and a word-initial vowel.

/f/ becomes /v/ hoefijzer, vijf uur

/s/ becomes /z/ mensaap, pas op

/ʃ/ becomes /ʒ/ ’n broche op, cash-and-carry

Again, /x/ becoming /ɣ/ will only occur in the South, as in mag ik. Speakers from the East of the country, from Groningen to Limburg, tend to voice all obstruents in this context. For example, /t/ becomes /d/ in kwart over (drie), einduit(slag); /p/ becomes /b/ in (de) trap op; /k/ becomes [g] in (jij) ook al. And for some, it may apply to final clusters like /ts/

Advice for learners

With the notable exception of flapping of ambisyllabic /t/, this kind of assimilation is not common in GA. Remember that in English liaison commonly applies in phrases like this evening, if ever, but that the fricatives remain voiceless.

When trying to pronounce these phrases, you may find that you either pronounce them as two words, i.e. with a pause, or with liaison and voicing of the fricative. In practicing these sequences, it may help to pretend that the fricative is initial rather than final, as in /ðɪ ˈsiːvnɪŋ, ɪ ˈfɛvər/.

this evening

if ever

Within words, AN /s/ frequently alternates with /z/ between voiced sounds, as in consonant, defensie, sensatie. Make sure you pronounce /ns/ in English consonant, defensive, sensation, and /s/ in words like oasis, crisis, basis, freemason.






13.6.4 Palatization of Alveolars

Sequences consisting of alveolar /t,n,s/ + /j/ are realized as pre-palatal [c,ɲ,ʃ] respectively.

/tj/ becomes [c] liedje, weet je

/nj/ becomes [ɲ] franje, kun je

/sj/ becomes [ʃ] pasje, was je

This coalescent assimilation also occurs in sequences like /ns + j/, /nt + j/, /st/ + /j/ and /nst/ + /j/, as in dansje, handje, kistje, kunstje, which are pronounced [ˈdɑɲʃə, ˈhɑɲcə, ˈkɪʃə, ˈkʏɲʃə], with elision of /t/ in the last two words. Palatalization of /t,n,s/ does not occur in the pronunciation of speakers from Limburg and Flanders.

Advice for learners

Note that GA /j/ is normally only coalesced to /t͡ʃ,d͡ʒ/ when it occurs initially in a weak syllable, as in get you, did you, situation, education. In sequences like those in not young, that year, /t/ may be glottalized, glottaled or voiced, but /j/ remains fully voiced in all cases.

The change of /s/ to /ʃ/ before /j/, as in this year, is somewhat more common, but by no means as regular. In words like menu, onion, a palatal [ɲ] may frequently be heard in rapid and casual speech, but not in more deliberate styles.

To avoid undesirable palatalization, make a firm alveolar contact for /n/ and /s/, and glottalize, glottal, or voice /t/, making sure that the following /i/ is a fully voiced approximant. For /s/ before /j/, practice the unassimilated form. Here are some examples for you.

not yet ten years Is this you?

courtyard Spaniard misuse (n)

last year in Europe misuse (v)

not yet

ten years


13.6.5 Nasalization and n-Deletion

Before all consonants other than plosives, AN /n/ tends to be elided after having nasalized the preceding vowel. As a result of this /n/-deletion, words like zonlicht, ongeluk may be [ˈzɔ̃lɪχt], [ˈɔ̃χəlʏk].

Advice for learners

Although /n/-deletion also occurs in GA (see the section on nasal deletion), it does not apply in the same contexts. In words like sunlight, unrest, unwise, the nasal is not normally elided, even though the vowel is at least partly nasalized. So try to make a firm alveolar contact for /n/ in words like:

dancing mention inch lunches

only unrest onwards insensitive

moonlight unsure unlit sunshine




13.7 Simplification v. Gemination

In GA, a sequence of two identical consonants as in club badge, midday, pine nut tends to be about 1.5 times as long as a single consonant. Such ‘double’ consonants are therefore essentially pronounced as single consonants with a longer articulatory contact. This is called gemination.

club badge


pine nut

Sequences consisting of identical fortis stops like those in pep pill, nighttime, bookcase will be realized as single glottalized, aspirated voiceless stops with a lengthened compression stage. That is, [ʔ͜kːʰ] will occur in bookcase, etc.

pep pill



How can a single [kː] trigger off both glottalization and aspiration?

In double consonants are reduced to single consonants through simplification. There is no difference between the double /tt/ in (hij) gaat toch and the single /t/ in (ik) ga toch. Words like looppas, achttal, doellijn are pronounced /ˈloˑpɑs/, etc.

Simplification also affects sequences of identical consonants which result from the application of various assimilations. In words like kopbal, bloeddruk /bb/- and /dd/-sequences arise as a result of regressive voicing. Subsequent application of simplification yields /ˈkɔbɑl, ˈbludrʏk/. Progressive devoicing creates a /ss/-sequence in leeszaal, which through simplification becomes /ˈleˑsaˑl/. Similar cases are roofvogel, jaszak, zeepbel.

Be careful not to apply these AN rules to GA words. Pronunciations like */ˈʃɔbɛl, ˈbɛkraʊnt, ˈfaɪˈfoʊts/ for shopbell, background, five votes are strikingly foreign. A lack of familiarity with gemination may also cause perceptual problems as when phrases like I‘d do it, we’ll leave and they’ve found it are interpreted as I do it, we leave, they found it by speakers of English.

13.8 Elision

Like assimilation, elision varies with style: elisions are very common in rapid informal and casual styles, but much less so in more monitored and deliberate styles.

13.8.1 Schwa-Elision before /r,l,n/

Schwa tends to be elided when it occurs before the sonorants /r,l,n/, followed by a weak vowel. Schwa-elision is more common in frequent words than in infrequent words. In very common words like general and different schwa is readily elided even in careful styles: /ˈd͡ʒɛnrəl, ˈdɪfrənt/. In fact, the word every is normally /ˈɛvri/ in all styles.




In addition to this, the frequency of elision depends on the phonetic context. It is most common between an obstruent or nasal and /r/, as in separate, history, elaborate, nursery, century, camera, memory, scenery, though after /k,g/ it is less common, as in bakery, mockery, vagary /ˈveɪgəri/.

Before /l/ and /n/, as in family, finally, privilege, passionate, arsenal, schwa-elision may create a syllabic consonant, i.e. /ˈfæml̩i, ˈfaɪnl̩i, ˈprɪvl̩əd͡ʒ, ˈpæʃn̩ət, ˈɑrsn̩əl/, which in more rapid styles may lose its syllabic status, giving rise to disyllabic /ˈfæmli, ˈfaɪnli, ˈprɪvləd͡ʒ, ˈpæʃnət, ˈɑrsnəl/.





In words like artery, cidery, rotary, uterus, watery, malodorous schwa is normally retained: apparently, previous application of flapping blocks schwa-elision. In this connection it is worth noting that elision does not normally occur after /l,r/ either, so that /ə/ is retained in salary, quarreling, colony, mariner.





In very rapid speech schwa may also be elided before a stressed vowel, as in police, believe, marine, collapse, sonata, solarium. Here again, we can distinguish between cases where the loss of schwa is compensated for the creation of a syllabic consonant, as in (disyllabic) /pl̩iːs, bl̩iːv, mr̩iːn/, and the more extreme monosyllabic forms pliːs, bliːv, mriːn/, where no such compensation takes place.

A different type of schwa-elision may occasionally be heard in non-standard speech when /r/ occurs between two schwas. In words like different, dangerous, ignorant, the second rather than the first schwa may be elided, yielding /ˈdɪfərnt, ˈdeɪnd͡ʒərs, ˈɪgnərnt/. Because of the low prestige associated with these pronunciations, hypercorrections may arise like */ˈmɑːdrən/ for modern, i.e. /ˈmɑːdərn/ (Bailey 1985:125).

Advice for learners


Schwa-elision is particularly common in frequent words like different, history, camera, family.

Avoid schwa-elision before non-sonorants: Many speakers are inclined to apply elision in words like difficult, physical, relative, syllable, political, development, pronouncing them */ˈdɪfkʌlt, ˈfɪskəl, ˈrɛltɪf, ˈsɪlbəl, pəˈlɪtkəl, dɪˈvɛlpmənt/. A similar undesirable elision is commonly heard in final unaccented syllables, as in crisis, packet, which may be pronounced */ˈkraɪss̩, ˈpækt/.

Here are some examples of words where schwa-elision is common: reference, prisoner, personal, opera, especially.



13.8.2 t/d-Elision in Coda

GA /t,d/ are elided in the coda when occurring between consonants. For instance, /t/ and /d/ are deleted in postcard, facts, walked back, windpipe, finds, gunned down, but not of course in actress and laundry, where they occur in the onset of the syllable.





These are some constraints on the preceding as well as the following consonant:

  1. Preceding consonant:

If the preceding consonant is /r/, there is no elision of either /t/ or /d/, as in heartbeat, cardboard. If the preceding consonant is /n/ or /l/, /d/-elision is common, but /t/-elision is not. Compare felt bad, stuntman, and cents, where /t/ is normally retained and realized as [t͜ʔ], with held back, landmark, sends, where /d/ is readily elided.



felt bad


held back


  1. Following consonant:

If the following consonant is /r,j,w,h/, elision is less likely, as in left right away, last year, last week, guesthouse, send Ronnie, old yachtsman, cold water, build houses.

left right away

last year

last week


send Ronnie

old yachtsman

cold water

build houses

The elision of /t/ in disyllabic contractions with /nt/ (not) is a special case: (it) isn’t clear, (he) doesn’t know, isn’t it, doesn’t she. Such an elision may also occur in monosyllabic contractions like (I) don’t like it, (I) don’t have one, but not when the contracted monosyllabic form is accented. So we would expect I ˈcan’t ˈdo it to differ from I ˈcan do it, i.e. [aɪ ˈkæ̃ʔ ˈduːɪʔ] vs [aɪ ˈkæ̃n ˈduːɪʔ], much like they’d spent the night there [ðeɪd ˈspɛ̃ʔ ðə ˈnaɪʔ ðɛr] differs from they’d spend the night there [ðeɪd ˈspɛ̃n̪ ðə ˈnaɪʔ ðɛr].

I can’t do it

I can do it

they’d spent the night there

they’d spend the night there

In nonstandard speech, final -Ct and -Cd clusters are frequently reduced to single -C, so that bold, past, next, left may be pronounced /boʊl, pæs, nɛks, lɛf/, even before a vowel or silence. In the standard language, this type of reduction may occasionally be heard in unaccented syllables, as in England, island, Egypt, perfect, though not in /nt/ clusters, as in different [ˈdɪfrə̃ʔ].

Advice for learners

Try to practice /t,d/-elision between consonants, except before /rj,w,/ and /h/, as in fast race, last year, wristwatch, guesthouse, goldrush, second year, old wine, blind horse.

One further context where /t/ should not be elided is after /n/ and /l/, as in went back, pointless, salt mine, hints. In words like facts, texts, sends, winds, postcard, handbook, /t,d/-elision is the rule.

Here are some further examples to practice /t,d/-elision: costly, perfectly, soft palate, act two, windshield, standpoint, restless, lists.

13.8.3 Other Elisions

Here is a list of some elisions that may occur in informal speech. You are not expected to apply these in your own speech.

  1. Elision of pre-consonantal /v/ is common in the weak forms of of and have, as in (I can take) care of myself /ˈkɛr ə maɪˈsɛlf/, (in my) neck of the woods /ˈnɛk ə ðə ˈwʊdz/, (You) should have seen (her) /ˈʃʊd ə ˈsiːn/. In informal writing, spellings like kinda, shoulda for kind of, should have are sometimes used to reflect this type of elision.

    care of myself

    neck of the woods

    should have seen

  2. In rapid styles, /ð/ may be realized as a dental nasal [n̪] after /m/, as in from there, from this: /frəmˈn̪ɛr, frəmˈn̪ɪs/.

    from there

    from this

  3. Elision of /f/ is common in fifth, twelfth.



  4. Before unstressed me, the verbs let, give lose their final consonant in informal styles: /lɛmi, gɪmi/.

    let me

    give me

  5. In very rapid styles, flap deletion may occur, as in better /bɛər/, greatest /ˈgreɪəst/.



  6. Finally, note that /k/ is normally elided in asked and /g/ frequently in recognize.



13.9 Insertion of Stops

In GA, a fortis stop is inserted in words like nymph, sense, strength. If a nasal and a voiceless fricative occur in the coda, a fortis stop is inserted which has the same place of articulation as the nasal. For the above examples, the result is /nɪmpf, sɛnts, strɛŋkθ/.




The fricative can also be taken into the coda as a result of the Weak Syllable Principle. That is why stop insertion also occurs in symphony, concert, lengthen, where the voiceless fricative is ambisyllabic. The result is /ˈsɪmpfəni, ˈkɑːntsərt, ˈlɛŋkθən/.




Recall that the WSP does not apply when the following vowel is strong: consort /ˈkɑːnsɔrt/, sensation, Bloomfield have no stop insertion.




Stop insertion arises as a result of a ‘premature’ raising of the lowered soft palate in anticipation of the velic closure required for the following oral fricative. When the velic closure precedes the release of the oral closure for the nasal, a stop is produced.

When words like symphony, sense and lengthen have undergone stop insertion, they are subject to nasalization and nasal deletion, just like camp, cant, bank (which, you will recall, are pronounced [kʰæ̃p͜ʔ, kʰæ̃t͜ʔ, bæ̃k͜ʔ]). So sense may be [sɛ̃t͜ʔs] and strength [strɛ̃k͜ʔθ].



Advice for learners


Stop insertion leads to homophony in pairs like princeprints, tensetents and sensecents. In proper names like Thompson and Sampson as well as in words like empty and contempt the inserted stop has made its way into the spelling.







Unlike n-deletion, which tends to apply before consonants other than stops, as in dans, mensen, English nasal deletion applies before fortis stops only, including inserted stops. A word like dance is therefore realized as [d̥æ̃t͜ʔs] rather than [d̥æ̃s]. It may therefore be a good idea to concentrate on pronouncing a (glottalized) stop in words like concert, sensitive, defense, chance, difference, thanks rather than on deletion of the nasal.