My dialect of English

6 Nov 2024

This article is a phonological analysis of American English, focusing exclusively on my own pronunciation (a form of General American without the cot-caught merger), and ignoring all other dialects.

Consonants

There are 24 consonant phonemes:

Nasal m   n   ŋ  
Stop p b   t d tʃ dʒ k g  
Fricative f v θ ð s z ʃ ʒ   h
Approximant     r l j w  

  • /p/ and /k/ are aspirated at the start of a word and at the start of a full syllable (that is, a syllable containing a full vowel rather than a reduced vowel; see the next section for details). The /p/ in ‘aspect’ and the /k/ in ‘(Puerto) Rico’ are not aspirated, requiring us to syllabize these words as /æsp.ɛkt/ and /rik.o/ rather than /æs.pɛkt/ and /ri.ko/. The /k/ in ‘sarcasm’ is aspirated, requiring us to syllabize it as /sar.kæz.əm/.

  • /t/ has the following allophones:

    • Voiced and tapped [ɾ] when preceded by a vowel or /r/ and followed by a reduced or transyllabic vowel other than /ən/. (By “transsyllabic”, I mean “in the next syllable”.)
    • Often, silent [∅] when preceded by /n/ and followed by a reduced or transyllabic vowel other than /ən/.
    • Glottal [ʔ] when preceded by a vowel, /r/, /l/, or /n/, and followed by /ən/, if the previous vowel was a full vowel.
      • Alternatively, [t] is possible if the phoneme preceding the /t/ is /l/ or /n/.
    • Aspirated [tʰ] when preceded by a vowel, /r/, /l/, /n/, or /ŋ/ (and perhaps other consonants), and followed by /ən/, if the previous vowel was a reduced vowel.
      • Alternatively, [ɾ] is possible if the phoneme preceding the /t/ is a vowel.
      • Examples: ‘Washington’, ‘Overton’, ‘Hamilton’, ‘hesitant’, ‘irritant’. I pronounce ‘hesitant’ with either [tʰ] or [ɾ], and the others with [tʰ].
    • Unreleased [t ̚] or glottal [ʔ] when preceded by a vowel, /r/, /l/, or /n/, and followed by a transyllabic consonant, or a hard boundary such as the end of an utterance or sometimes a word boundary.
    • In all other situations, either [tʰ] or [t], with the same distribution as that for /p/ and /k/.

    These rules are pretty complicated, and I might have missed some edge cases. They can stretch both forward and backward across word boundaries; for example, the ‘t’ in ‘a bit of’ is pronounced as a tap, as is the first ‘t’ in ‘how to get there’. In the latter example, the tapping rule overrides the aspiration rule. As with /p/ and /k/, the realization of /t/ can guide syllabization. For example, ‘NATO’ is pronounced with [ɾ], requiring us to syllabize it as /net.o/, while ‘latex’ (with the meaning of ‘rubber’) is a pronounced with [tʰ], requiring us to syllabize it as /le.tɛks/.

  • /d/ is tapped when preceded by a vowel or /r/ and followed by a vowel, /w/, /r/, or /h/, and probably some other consonants, unless at the start of a stressed syllable. It is sometimes also tapped at the end of an utterance following a vowel or /r/.

    In certain contexts, /t/ and /d/ are both pronounced as a voiced tap. However, these can still be said to be different phonemes, for two reasons. First of all, some instances of [ɾ] alternate with /t/, while others alternate with /d/. For instance, while ‘coating’ and ‘coding’ are pronounced the same, ‘coat’ and ‘code’ are pronounced differently, and ‘coats’ and ‘codes’ are pronounced differently. So it makes sense to analyse ‘coating’ as /kotəŋ/ and ‘coding’ as /kodəŋ/. Second of all, the tap allophone of /t/ triggers raising of preceding /aɪ/, while the tap allophone of /d/ does not. So ‘writing’ is pronounced [ˈɹʷəɪɾɪŋ], while ‘riding’ is pronounced [ˈɹʷäɪɾɪŋ].

    That said, within a morpheme, the distinction between tapped /t/ and tapped /d/ is neutralized, unless the tap is preceded by /aɪ/ or followed by /ən/. (In the latter case, it has to be /d/.) There’s no real reason to transcribe ‘ladder’ and ‘latter’ differently; both are simply /læ(t~d)ər/.

  • /r/ is a postalveolar approximant with varying degrees of labialization. It is fully rounded at the start of a stressed syllable (as in ‘red’), and partially rounded in other prevocalic contexts (as in ‘green’, ‘era’). The rounding may also come with a slight degree of velarization or pharyngealization. It is not rounded if there is no vowel after it.

  • The letters ‘t’ and ‘d’ are pronounced as [tʃ] and [dʒ] before ‘r’ (unless the ‘r’ is in a different morpheme as in ‘footrest’.) I analyze these as /tʃ/ and /dʒ/ — for example, ‘tree’ is /tʃri/ — so this is not an allophonic rule, but rather a spelling-to-phoneme rule.

  • /l/ is velarized [ɫ] in all positions.

  • Contrary to claims made by Sapir and others, I have never perceived [ŋ] as reflecting any kind of underlying /ng/, and recall being confused upon seeing the spelling ‘ng’ for the first time in my early childhood. I believe that any existing perception among English speakers of [ŋ] as /ng/ is a result of spelling, except in the minority of dialects (such as my roommate’s) where it can be pronounced as [ŋg].

Vowels

There are 17 vowel phonemes:

i ĭ   u ŭ
ɪ   ʊ
e ə o oɪ
ɛ   ʌ
æ æʊ a aɪ ɒ

Three of these phonemes (/ə/, /ĭ/, and /ŭ/) are reduced vowels, and the rest are non-reduced (or full) vowels. Of the full vowels in a word, one carries primary stress, and the rest carry secondary stress. The transcriptions above, especially for the “back” vowels, are only approximate. My dialect of English (and American English more generally) stands out among the world’s languages for its near lack of any back rounded vowels. The only fully rounded syllable nuclei are the first parts of /oɪ/, /or/, /ol/, and /ul/.

  • /əl/ and /ər/ are pronounced as syllabic [ɫ̩], [ɹ̩]. Other sonorants may sometimes be syllabic after /ə/, but this is inconsistent. Syllabic liquids can also occur in full syllables, as in ‘bull’, ‘term’. Phonemically, these are best analyzed as /ʊl/ and /ʊr/, in my opinion. (The /ʊl/ analysis is standard; the /ʊr/ analysis is my own innovation.)

  • /ə/, /ĭ/, and /ŭ/ (occuring respectively in ‘comma’, ‘happy’, ‘casual’) are the only reduced vowels. In my dialect, the second vowels in ‘hammock’ and ‘music’ are pronounced the same, as are the second vowels in ‘gallop’ and ‘turnip’. All are transcribed in my analysis as /ə/. The realization of /ə/ in the first pair is farther forward (closer to /ɪ/) than in the second pair, but this is an allophonic variation. Specifically:

    • In morphemes ending in /ə/ followed by a sequence of one or more consonants in which the first consonant is not bilabial, the /ə/ is fronted (made closer to /ɪ/). This, of course, only applies if the /ə/ is actually pronounced rather than syllabicizing the consonant after it.
    • At the end of a morpheme, /ə/ is backed (made closer to /ʌ/).

    The standard transcriptions of ‘music’ and ‘hammock’ appear to be /mjuzɪk/ and /hæmək/. That is nonsense, at least where I am from. It is /mjuzək/, /hæmək/, /gæləp/, /tʊrnəp/, and so on. I even analyze the ‘-ing’ suffix as /əŋ/, with the justification that /ə/ is especially fronted before /ŋ/.

  • /ŭ/ only occurs in the sequence /ŭə/. It could be analyzed as /əw/, with the added rule that /w/, like /l/ and /r/, becomes syllabic after /ə/, but I have decided not to do that.

  • The vowels /ɪ/, /ʊ/, /ɛ/, /ʌ/, and /æ/ do not occur syllable-finally, except in interjections such as ‘uh’. /a/ is rare syllable-finally, occuring only in words of foreign origin such as ‘spa’, ‘bra’, ‘mardi gras’, ‘Fermat’, etc.

  • /i/ is a monophthong. It is never pronounced as a dipthong.

  • /u/ is slightly diphthongized. It is fronter than cardinal [u], becoming closer, backer, and more rounded over its duration. I can’t tell what the exact phonetic value is, so I will simply transcribe this allophone as [u̟]. Before /l/, /u/ is more or less a cardinal [u]. /ŭ/ is closer than /u/ to being a monopthong, as one would expect for a reduced vowel.

  • /ɪ/ is rather lax. Unlike in many other languages, in which /ɪ/ is perceived as a shorter or laxer form of /i/, I have always perceived them as totally distinct vowels. Before /ŋ/, /ɪ/ is a closing diphthong [ɪi].

  • /ʊ/ is very lax, and is only slightly rounded. It should be noted that the diphthongs /aɪ/ /æʊ/ and so on do not have as their targets the realizations of /ɪ/ and /ʊ/, but rather [i] and [u]. I write /aɪ/ (as is standard) rather than /ai/ because these diphthongs do not actually reach their targets.

  • /e/ and /o/ are usually diphtongs [eɪ], [əʊ]. (The first component of the latter may be slightly rounded.) They are only monophthongs before coda /l/ and /r/. Examples are given by the following:

    • ‘fair’ /fer/ [feɹ]
    • ‘fail’ /fel/ [feɫ]
    • ‘fate’ /fet/ [feɪt]
    • ‘lore’ /lor/ [ɫoɹ]
    • ‘lower’ /lo.ər/ [ɫəʊ.ɹ̩]

    Note how the syllabic forms of /l/ and /r/ do not trigger monophtongization, since they are phonemically preceded by /ə/.

    The monophthongal realizations of /e/ and /o/ are somewhat lower than the corresponding cardinal vowels, but are not open-mid. The same applies to the first component of /oɪ/. The standard transcription of “fair” as [fɛɹ] has always baffled me, and /e/ and /ɛ/ contrast before /l/.

  • /ʌ/ is fronter than the corresponding cardinal vowel, but not as front as [ɜ].

  • Before /m/ and /n/, /æ/ is pronounced as a centering diphthong /eə/, the start of which is lower than cardinal [e]. Before /ŋ/, /æ/ is a closing diphthong that I will transcribe [ɛɪ]. This dipthong is rather similar to the primary realization of /e/, and I kept changing my mind on whether to assign it to /e/ or /æ/ (a process that involved saying the words ‘sang’, ‘sake’, and ‘sane’ out loud hundreds of times) before ruling in favor of /æ/.

  • /a/, including the first component of /aɪ/, is typically an open central vowel, the same as the Spanish, French, and German pronunciations of the letter ‘a’. This vowel occurs in words such as ‘lock’ and ‘father’. Before /r/, it pronounced further back as [ɑ].

  • Before phonemically voiceless consonants (including the tap allophone of /t/), the diphthong /aɪ/ is raised to /əɪ/. This stretches across word boundaries in certain common phrases, such as ‘high school’, which is pronounced [həɪskuɫ] (One could argue, then, that ‘high school’ is really one word, despite having a space in the middle.)

    An issue arises when looking at raising of /aɪ/ before morpheme-internal taps. I pronounce ‘spider’, ‘cider’, ‘idol’, ‘idle’, ‘Poseidon’, and a number of other words with /əɪ/, despite the tap being spelled with a ‘d’. For most of these words, you can fix the problem by analyzing the tap as /t/, despite the spelling. For example, you can analyse ‘spider’ as /spaɪtəɹ/, which under regular allophonic rules becomes [spəɪɾɹ̩]. However, this doesn’t work for ‘Poseidon’, because /pəsaɪtən/ is realized as [pəsəɪʔən], not [pəsəɪɾən], under regular allophonic rules. I will simply ignore this issue, as it is just one word.

  • /ɒ/, occuring in words such as ‘boss’, ‘fog’, and ‘dawn’, is only slightly rounded. It may be slightly diphthongized, centering somewhat over the course of the vowel.

    My mom, and certain other people I know, pronounce ‘fog’ as /fäg/ but ‘dog’ as /dɒg/, a distinction that seems to have arisen sporadically. However, I pronounce every ‘-og’ word with /ɒ/.

Vowel-consonant restrictions

  • The phonemic dipthongs /æʊ/, /aɪ/, /oɪ/, cannot be followed by a coda /l/ or /r/. I analyze words like ‘mile’, ‘hour’, as /maɪ.əl/, /æʊ.ər/. As mentioned above, /e/ and /o/, which are usually pronounced as diphthongs, can occur before code /l/ and /r/, but in this case they are realized as monopthongs.

  • The only vowels that can precede coda /r/ are /i/, /e/, /a/, /o/, /ʊ/, and /ə/, occuring in ‘fear’, ‘fair’, ‘far’, ‘for’, ‘fur’, and ‘offer’ respectively. (Recall that /ʊr/ is how I transcribe non-reduced [ɹ̩].) I do not have the phenomenon, found in other dialects of English, in which more vowels can occur before coda /r/ provided the /r/ is followed by a vowel. So ‘merry’, ‘marry’, ‘Mary’ are all pronounced the same, ‘mirror’ and ‘nearer’ rhyme, ‘hurry’ and ‘furry’ rhyme, and the first syllable of ‘forest’ sounds like ‘for’. I also do not have any kind of /ur/ sound; the words ‘cure’, ‘pure’, ‘sure’ are /kjʊr/, /pjʊr/, /ʃʊr/ i.e. [kʰjɹ̩], [pʰjɹ̩], [ʃɹ̩], and ‘tour’ is /tor/.

  • The only vowels that can precede /ŋ/ are /æ/, /ʌ/, /ɪ/, /ɒ/, /a/, and /ə/, occuring in ‘sang’, ‘sung’, ‘sing’, ‘song’, ‘bongo’, and ‘running’ respectively. Of these, /a/ is the rarest, but it does occur in words other than ‘bongo’, such as ‘(in)congruous’, ‘Congo(lese)’, and ‘franc’.

  • /a/ can only occur before /l/ if there is a vowel afterward, as in ‘holler’, ‘volley’. Note that ‘holler’ /halər/ does not rhyme with ‘taller’ /tɒlər/.

  • Most consonants have a preference for either /a/ or /ɒ/ occuring before them. For example, /ɒg/ occurs in many words, but /ag/ occurs only in marginal words like ‘Prague’ and ‘Häagen-Dazs’. On the other hand, /atʃ/ occurs in many words (such as ‘watch’ and ‘notch’) but I cannot think of any word with /ɒtʃ/. However, /t/, /n/, and /k/ occur after both vowels fairly frequently.

Transcriptions and pronunciations of specific words

Vowels before /r/ and /ər/

fear /fir/ [fiɹ]
fair /fer/ [feɹ]
far /far/ [fɑɹ]
for /for/ [foɹ]
fur /fʊr/ [fɹ̩]
hour /æʊ.ər/ [ˈæʊɹ̩]
fire /faɪ.ər/ [ˈfäɪɹ̩]
marry, merry, Mary /mer.ĭ/ [ˈmeɹi]
spirit /spir.ət/ [ˈspiɹə̟t]
hurry /hʊr.ĭ/ [ˈhɹ̩i]
cure /kjʊr/ [kʰjɹ̩]
tour /tor/ [toɹ]

/e/ and /o/

fair /fer/ [feɹ]
player /ple.ər/ [ˈpʰɫeɪɹ̩]
fail /fel/ [feɫ]
fate /fet/ [feɪt]
lore /lor/ [ɫoɹ]
lower /lo.ər/ [ˈɫəʊɹ̩]
cole /kol/ [kʰoɫ]
coat /kot/ [kʰəʊt]

/a/ and /ɒ/

cot /kat/ [kʰät]
caught /kɒt/ [kʰɒt]
don /dan/ [dän]
dawn /dɒn/ [dɒn]
Prague /prag/ [pʰɹʷäg]
dog /dɒg/ [dɒg]
fog /fɒg/ [fɒg]
waffle /wafəl/ [ˈwäfɫ̩]
off /ɒf/ [ɒf]
wasp /wasp/ [wäsp]
toss /tɒs/ [tʰɒs]
water /wɒ(t~d)ər/ [ˈwɒɾɹ̩]
volley /valĭ/ [ˈväɫi]
fall /fɒl/ [fɒɫ]

Schwa

comma /kamə/ [ˈkʰämə̠]
gallop /gæləp/ [ˈgæɫəp]
America /əmerəkə/ [əˈmeɹəkə̠]
American /əmerəkən/ [əˈmeɹəkə̟n]
Rosa’s /rozə.z/* [ˈɹʷəʊzə̠z]
roses /rozəz/ [ˈɹʷəʊzə̟z]
music /mjuzək/ [ˈmju̟zə̟k]
hammock /hæmək/ [ˈheəmə̟k]
running /rʌnəŋ/ [ˈrʌnɪŋ]
lever /lɛvər/ [ˈɫɛvɹ̩]
level /lɛvəl/ [ˈɫɛvɫ̩]

*The dot here represents a morpheme boundary rather than a syllable boundary.

Other

real, reel /ril/ [ɹʷiɫ]
mile /maɪ.əl/ [ˈmäɪɫ̩]
sang /sæŋ/ [sɛɪŋ]
tree /tʃri/ [tʃɹʷi]
actual /akʃŭəl/ [ˈækʃuɫ̩]
picture /pɪkʃər/ [ˈpʰɪkʃɹ̩]
issue /ɪʃju/ [ˈɪʃju̟]
Latin /lætən/ [ˈɫæʔ(ə̟)n]
Aladdin /əlædən/ [əˈɫæɾə̟n]