The "long i" sound in English, as in "fight" is usually written /aɪ/, so fight = /faɪt/.

But /a/ is the sound in "hat", and /ɪ/ is the sound in "hit". When I say the two together it doesn't sound anything like the diphtong in "fight". In fact, it doesn't sound like any kind of sound I've ever heard anyone say in any language, except maybe English with an exaggerated upper class British accent. If I had to guess, I'd say the long i sounds more like /ai/.

Is this a dialect issue, or am I just not paying close enough attention to what my mouth is doing? How can I convince myself I'm wrong?

  • For English "fight", "eye", "I", etc., they should be transcribed using short /ɪ/ not long /i:/ so /aɪ/ not /ai:/ (usually denoted by /aj/ instead). I'm not aware of any English variant using /aj/ till now. In my first language (not English), the sound inventory does include both /ai/ and /aj/ and they sound quite different.
    – Tran Khanh
    Dec 23, 2023 at 15:21

4 Answers 4


In American dialects, the sound in 'hat' is generally /æ/, not /a/. The use of /i/ (meaning /ai/) in the 'fight' vowel would be too long/close; you probably do not pronounce 'might' exactly the same as the phrase 'ma eat'. The vowel in 'might' is shorter/opener, and so /ɪ/ (meaning /aɪ/) is appropriate. (See https://en.wikipedia.org/wiki/Diphthong#Transcription)

  • 9
    It should also be kept in mind that when talking about phonemic representations, /aɪ/ is just a conventional representation of a diphthong that can be a number of things depending on dialects, and crucially to this question, those things may or may not be related to the way /a/ and /ɪ/ are pronounced individually. That's because /aɪ/ is its own phoneme and speakers don't really see it as /a/ and /ɪ/ joined.
    – LjL
    Jan 24, 2018 at 4:33
  • @LjL Are you telling me there's a convention in IPA wherein a sequence of symbols can stand for a sound that isn't the sequence of the sounds of each of those symbols?
    – Jack M
    Jan 24, 2018 at 19:10
  • 3
    @JackM: In phonemic transcriptions that use IPA letters, even a single symbol can represent something that isn't the sound associated with that symbol in "canonical" IPA for phonetic transcription. For example, the symbol /e/ can be used to represent the phone [ɛ], and in fact the IPA recommends doing this unless you are transcribing a language that has a phonemic contrast between a phoneme pronounced [e] and another phoneme pronounced [ɛ]. Jan 24, 2018 at 20:57
  • @JackM I think sumelic's comments should answer the question; I'll just add that it might be "more proper", even in phonemic IPA, to use a conjoiner (as in /a͡ɪ/) to show that /aɪ/ is being used to represent a diphthong which is its own phoneme, and not simply the juxtaposition of /a/ and /ɪ/. That's seldom done, at least for English, since there isn't a real need to disambiguate, as the /a/ + /ɪ/ sequence barely exists independently. IPA doesn't really mandate or even specify these conventions, when it comes to phonemic representations; those develop naturally for individual languages.
    – LjL
    Jan 25, 2018 at 20:16

As user6726's answer mentions, IPA letters (like [i], [e], [ɪ], [ɨ], [a]) refer to ranges of sounds (phones); for vowel letters, these can be thought of as regions of the "vowel space" (which in turn can be thought of as either an acoustic space (defined by certain frequencies), or a physiological space (defined by tongue position)).

And phonemic transcriptions (conventionally indicated by the use of slashes rather than brackets) don't even technically use IPA letters to refer directly to the associated ranges of phones. Rather, traditions for phonemic transcription establish conventional associations between IPA symbols and particular phonemes in a language, regardless of the actual phonetic realization of the latter. If there gets to be too much of a difference between the conventional transcription and the phonetics of some particular accent, a new convention may be proposed/adopted, but often the old conventions still stick around for a long time. You can see some of the diversity in transcriptions of English vowels on the web page "IPA transcription systems for English" by John Wells.

The phonetic realization of the phoneme /aɪ/ is variable between accents, and between different utterances made by a single speaker (some of the variation can be predicted fairly easily from the context, and some can't).

Mark Liberman, a linguist, has a post on Language Log mentioning some of the features that may make a phonetic transcription like [aɪ] valid for the phoneme /aɪ/ in at least some contexts in some present-day accents of English: The rɑɪt sɑʊnz?

But it isn't necessarily incorrect to use something like /aj/ as a phonological transcription. Geoff Lindsey, a linguist and dialect coach, transcribes the standard Southern British English "price" vowel as /ɑj/. In his blog post "The British English vowel system", Lindsey says:

in SSB, the upper corners of the vowel space function as the endpoints of its two subclasses of diphthongs. I prefer to transcribe them as the semivowels j and w, which are exactly equivalent in terms of the vowel space, but have additional benefits. First, j and w make explicit that the English diphthongs are falling, ie moving from greater to lesser prominence. Second, they reinforce English vowel linking for the foreign learner, eg way out wɛjawt, show off ʃəwɔf, as well as initialisms like those in the previous paragraph. Additionally, they provide by far the best way to teach cardinals 1 and 8 to native English speakers: simply pronounce intervocalic j or w, then “hit the pause button” in the middle of the semivowel.

In final position, especially when heavily accented, SSB diphthongs are sometimes followed by a short, schwa-like sound which highlights the final j or w. Here are me and loo said by Kate Winslet: [sound sample on the original blog post]

Such pronunciations clearly differ from the closing diphthongs of old RP with their distinctively not-quite-high endpoints ɪ and ʊ. These sound very old-fashioned, especially in final accented position, e.g. May in this 1932 newsreel clip: [sound sample on the original blog post]

or the Queen saying Christmas Day in 1957: [sound sample on the original blog post]

or the narrator of the 1948 film Quartet saying 24 plays: [sound sample on the original blog post]

Of course, in the less accented contexts of continuous speech the contemporary j and w endpoints will often not be reached. This kind of target undershoot is a universal phenomenon in running speech.

My summary:

There are phonetic reasons for the traditional transcription of the vowel of PRICE as /aɪ/; some of them are historical but others continue to be relevant. The use of this transcription however does not necessarily imply that /aɪ/ ends in the same sound as the KIT vowel phoneme on either a phonological or a phonetic level.

There is nothing wrong with using another transcription like ai, aj or even ay if you have some valid reason to do so. The choice of symbols for phonemes in a phonemic transcription is basically arbitrary, and most phonemic analyses treat the "PRICE" vowel as a single phoneme.

Transcriptions that identify the last element of the PRICE with the palatal semivowel (as in the start of the word "yam") have some theoretical support and are also supported by a tradition of use that continues to the present day, although it is not as prominent as the tradition that uses /aɪ/.

The FLEECE vowel of course is also phonetically and phonologically similar to the palatal semivowel, but I think transcriptions like ai have not been as popular as ones like aj because using a symbol like j that is explicitly dedicated to transcribing glides makes it clear that we are not talking about a disyllabic sequence.

I'd recommend not worring too much about the details of phonemic transcription unless you're planning on learning or teaching some kind of standardized transcription system (e.g. for use in EFL contexts). Whether you're primarily interested in phonology or in phonetics, IPA transcriptions are fairly superficial*, and the most commonly used transcriptions of phonemes tend to be common because of tradition rather than because of deep theoretical reasons.

*I don't mean this in a negative way. If IPA transcription were more strongly based on phonological theories, it would make it harder to give transcriptions in a relatively theory-neutral fashion; if IPA transcription were more strongly based on phonetic details, it would make it harder to give "broad" transcriptions that are appropriate for describing how a speaker pronounces a phoneme across a range of contexts, or how a group of speakers tend to pronounce a phoneme. But the compromises of IPA mean that there is often more than one theoretically reasonable way to transcribe something, and so the specific way that people actually do transcribe it is based on tradition and somewhat arbitrary from a theoretical perspective.


There are indeed dialect issues.

I presume you are aware that IPA letters represent regions of vowel articulation and not precise points. There is an informal transcriptional practice for vowels that in transcribing a language, one uses the symbol which most closely matches standard reference performances. But this is not an absolute rule, and in fact usually people write "a" unless there is a contrast such as [a,ɑ]. For most American dialects, one could use either [æ] or [a] for the vowel in "hat", since [æ] and [a] do not contrast. Typically, [æ] is used because that better matches the reference pronunciation.

The treatment of diphthongs in IPA is analogous to the treatment of affricates: they are written with two letters, possible with the tie diacritic to imply a phonological analysis as a single segment. Again, the convention of "closest phonetic match" would be applicable in deciding how to transcribe "fight". There are about 3 reasonable choices for the first vowel: [a, ɑ, ɐ]. Because of the following diphthong element [ɪ], it's difficult to make the decision based just on introspection (hearing [ɪ] right after the low vowel is a distraction), but you can use Praat to extract just the initial 1/3 of the diphthong and compare it to the initial 1/3 of other vowels to see which vowel you think the a-like vowel is closer to.

I suspect that you won't find [ɐ] to be a particularly good match (there are dialects where that is the appropriate vowel, often diagnosed by the pair "rider/writer"). Many dialects have both a low front vowel and a low back unrounded vowel, so [a, ɑ] are plausible choices. Whichever vowel you pick, it won't really match the phonetic quality of the low-vowel part of "fight". This is when articulation-adjustment diacritics are appropriate, for example you might classify it as an advanced ([˖]) version of [ɑ], or a retracted ([ˍ]) or centralized ([¨]) version of [a].

Phonetic closeness is not the only consideration: another, relevant to transcribing diphthongs, is whether the component vowels are individually required. If a language has an ai-like diphthong that sounds most like [æi] but the language has only [a] and not [æ] as a phoneme, then you might appeal to the lack of an [æ] phoneme as justification for the transcription [ai]. Since many American dialects have both [æ] and a lower backer vowel written either [ɑ] or [a], any of [æɪ, aɪ, ɑɪ] would be a possible transcription for that diphthong. [aɪ, ɑɪ] fare better in terms of matching the quality of the initial part of the diphthong to the corresponding reference vowels. Assuming that "father" is transcribed with [a], then you would write [faɪt] if you're not introducing new vowel symbols for diphthongs.

My dialect lacks [ɑ] ("cot" and "caught" are both [kɔt]), so I would opt for [aɪ]. However, the vowel [a] exists only in [aɹ, aɪ] and adult-acquired foreign words like "pasta". If I were minimizing the number of letters in my transcriptions, I would probably write [æɪ, æɹ], but that would be phonetically misleading.

  • You're right, I meant /æ/ for "hat". Is the heart of your answer basically that the /ɪ/ in /aɪ/ is one that's located near the boundary of the /ɪ/ region of the vowel chart?
    – Jack M
    Jan 24, 2018 at 20:20
  • 2
    The essence of a diphthong is transition, so the brain is more interested in detecting the overall slope of the vowel map rather than length or specific end points.
    – amI
    Jan 24, 2018 at 23:28

Bear with me, there are a few errors there.

Error 1, Typically, brackets ([]) typically are used for precise pronunciations and slashes (//) are typically used for broad transcriptions.

*A good example of how vastly different vowels of the same articulation can sound is with the /a/ vowel. In Southern American accents, the diphthong /aɪ/ is replaced with an /a/ vowel.

Error 2, the "american 'a'" in "hat" is the vowel /æ/. /a/ is not that common of a vowel in american english, but it is present.

Answer: /aɪ/ is the diphthong. If you say it, then I recommend holding the ending articulation of it, you will find that it should be probably a high articulation of the /ɪ/, if you pronounce it of a similar articulation to I.

  • I think the vowel of "hat" in North American accents can range from [a] to [eə̯]. It seems to vary a fair amount between different regions and different people Jan 24, 2018 at 20:54
  • I think I agree. I am not very well versed with all accents. If I recall, I think some southerners say /hæət/ Jan 25, 2018 at 4:27

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.