When should one use /fubar/ and when [fubar] when transcribing in IPA? What are the differences?

  • 9
    Related: What's the difference between Phonetics and Phonology?
    – Alenanno
    Commented Sep 20, 2011 at 7:27
  • Note that when people are considering both the answer is pretty clear, but in certain contexts one form of bracketing will be used irregardless. I've definitely seen [ ] in dictionaries pretty often but general purpose dictionaries are always at the phonemic level. Also Wikipedia seems to use [ ] for all IPA currently from what I can see, whether phonetic, narrow, broad, or phonemic. Commented Sep 20, 2011 at 17:30

9 Answers 9


Square brackets ([fubar]) are generally used for what is known as narrow transcription - this includes as much detail as the transcriber feels is necessary. Slashes (/fubar/) represent the broad transcription, which does not include "predictable" information.

For example, in English, voiceless plosives are aspirated word-initially and in stressed onsets. Thus, a narrow transcription of "cool" might be [kʰul], while a broad transcription would be /kul/. Similarly, "lack" could be represented as [læk] and /læk/ - note that broad /k/ can become narrow [k] or [kʰ], depending on its position in the word and surrounding sounds. Because this information is predictable by the above rule, the aspiration is left out of the broad transcription.

In linguistic description, the "broad" and "narrow" designations are defined somewhat loosely, generally according to whatever convention the linguist reporting the language feels is most useful. Under the Generative Phonology framework, however, these are assumed to represent two distinct stages of phonological processing, each with a psychological reality. Square brackets denote the final stage of processing (which is sent to the articulators), called "phonetic transcription", while slashes denote the form stored in the mental lexicon (stripped of all predictable information), called "phonemic transcription".

To perform this kind of analysis, first you must determine which sounds are contrastive in the output. For example, in English, [pʰ] and [p] are not contrastive, because [pʰæt] and [pæt] "pat" are judged to be "the same" word by (most) native speakers (even though [pæt] would be ill-formed). However, in Hindi, [kapʰi] "coffee" and [kapi] "copy" are two separate words. The generative hypothesis is that each set of non-contrastive sounds is stored as a single unit in the brain, called a phoneme (in slashes), which is transformed into a final form (in brackets) passed to the articulators by a series of serially-ordered rules or simultaneous constraints on the possible output forms.

  • Regarding your [pʰæt] and [pæt]. What you say seems to hold for when we speak normally, but when whispering, I think [pæt] is how the word bat is realised. Commented Feb 18, 2020 at 14:08
  • 6
    Your starting sentence seems to be a bit misleading; [brackets] are used for what is known as phonetic transcription which can either be broad or narrow. Commented May 26, 2021 at 11:04
  • I think the broad/narrow and slashes/brackets distinction depends on the context and which framework you're operating from. Your definition is correct in many cases, but the asker didn't specify either context or framework, so I took the most neutral position at first, and delved into generativity halfway through.
    – Alek Storm
    Commented Jun 1, 2021 at 17:33

Yes, /fubar/ is typically used for phonemic transcriptions, and [fubar] for phonetic transcriptions. But, just to clarify the terminology, phonemic vs. phonetic is not necessarily the same thing as broad vs. narrow transcription.

Many linguists talk about using both broad and narrow phonetic transcriptions, which just refers to the level of detail used in representing the actual speech sounds. A narrow phonetic transcription would represent every tiny little characteristic of the speech sounds as they were produced in an utterance of Language X, while a broad phonetic transcription would indicate some of the most salient characteristics of the transcribed phones, without being exhaustive but also without necessarily making claims about which segments are phonemically/phonologically contrastive.

Phonemic transcription, on the other hand, can really only be broad, in that it only represents the sounds that are purported to be contrastive in the given language, without any detail that is not directly relevant to forming these contrasts. Phonemic transcription does not describe how an utterance actually sounds when produced by a particular speaker speaking in a particular style in a particular situation - phonemic transcription is the 'idealised' representation of the speech sounds, and supposed to represent the underlying contrasts that are meaningful to speakers. You can only do phonemic transcription when you have already done quite a bit of work on Language X (collected a range of lexical items, found some minimal pairs and/or worked out the patterns of allophonic variation, etc etc) because phonemic transcription implies that you/someone has decided which phones relate to contrastive phonemes (different sounds that can occur in exactly the same environment) and which phones occur as the result of free/conditioned variation.

The examples given by @Alek Storm above illustrate the differences between [phonetic] and /phonemic/ transcription - just remember that phonetic can be broad and narrow.

  • +1 So can a certain variant of broad phonetic transcription be identical to the phonetic description of a word, such that all words are transcribed like /yxy/ and [yxy], where yxy is the same in both description for each word? Or does this never happen in practice, and will linguistics always choose a variant of broad phonetic description that is different from the phonemic description?
    – Cerberus
    Commented Jun 5, 2012 at 20:14
  • 1
    @Cerberus I'm not a linguist, but it seems that "atoms" of phonemic and phonetic descriptions are different (even though IPA is often used for both cases). E.g., Toki Pona (a minimalistic conlang) doesn't discriminate [p] from [b], so the phonemic description of pona is only /pona/ (if we agree that both [p] and [b] map to /p/, not /b/ or some /B/) while phonetic descriptions may include [pona] and [bona] (more broad), ['pona] and [po(pause)na] (less broad), ['po(pause)na], etc. So: (1) p in /p/ and [p] are from different sets; (2) different levels of description may be regarded as broad. Commented May 15, 2018 at 10:53

In addition to slashes and square brackets, sometimes also used are double-slashes, pipes, and angle brackets. Their uses are:

  • Angle brackets — ⟨cats⟩ or cats or "cats" or cats — orthography

    Indicates a linguistic entity, like a word or grapheme, written according to a language's orthography. Alternatively, the orthography is often given in italics or quotes, or simply not indicated.

  • Square brackets — [ˈkʰæʔt͡s] — phonetic transcription

    Indicates a transcription that records the phones (speech sounds) that are spoken, without attempting to classify them into phonemes. The amount of phonetic detail transcribed can vary; the example here is a fairly narrow (detailed) transcription. A broader transcription might be [ˈkʰæts].

  • Slashes — /ˈkæts/ — phonemic transcription

    Indicates a transcription that records only information asserted to be contrastive. This records a sequence of phonemes (as well as any phonemic suprasegmental features like stress or tone), ignoring allophonic differences. In this example, the fact that the initial /k/ is pronounced aspirated ([kʰ]) is ignored, because /k/ and /kʰ/ do not contrast in English.

  • Double-slashes or pipes — //ˈkæt z// or |ˈkæt z| — morphophonemic transcription

    Indicates a transcription that attempts to record the underlying sounds of morphemes, before they are combined to form words. This ignores sound changes conditioned on location, which may result in multiple phonemes, that get applied in actual speech. To illustrate, here, the final sound is written ⟨z⟩ (rather than ⟨s⟩ as in its pronunciation) because it represents the English morphophoneme //z//, used as the plural marker. //z// can be realized as /z/ (as in pigs or kangaroos), /s/ (as in cats), or /ɪz/ (as in horses). The posited morphophoneme cannot be an actual phoneme, because English distinguishes /z/ /s/ /ɪz/ in other contexts (like whores, horse, horiz(ontal)). (The choice of ⟨z⟩ for the transcription of this morphophoneme is arbitrary, but makes sense because it is realized as /z/ in the absence of the effect of an adjacent voiceless segment.)

More examples

  • Morphophonological processes are especially prominent in French. Example phrase: On a laissé la fenêtre ouverte. 'We left the window open.':

    • orthographic ⟨On a laissé la fenêtre ouverte.⟩
    • phonetic [ɔ̃.na.le.se.laf.nɛː.tχu.vɛχt]
    • phonemic /ɔ̃naleselafnɛːtʁuvɛʁt/
    • morphophonemic //ɔn a les e la fənɛːtʁ uvɛʁt ə//
  • French petit 'small': orthographic ⟨petit⟩, phonetic [pɵt̪ʲi], phonemic /pəti/, morphophonemic //pətit//

  • Standard Russian vowel reduction

    In the standard (Moscow-based) Russian accent, five vowels /i e a o u/ are distinguished in stressed syllables, but at most three /ɪ~ɨ ɐ~ə ʊ/ in unstressed syllables. /i e/ reduce to /ɪ/, /a o/ to /ɐ/, and /u/ to /ʊ/. Russian also has word-final obstruent devoicing and assimilation of voicing and palatalization. Russian orthography is roughly morphophonemic. Examples:

    • meaning: poppy, mage, poppies, mages, earth (nom. pl.), earth (gen. sg.), transport (infinitive), transports (present, 3sg)
    • orthographic ⟨ма́к⟩ ⟨ма́г⟩ ⟨ма́ки⟩ ⟨ма́ги⟩ ⟨зе́мли⟩ ⟨земли́⟩ ⟨вози́ть⟩ ⟨во́зит⟩
    • phonetic [ˈmak] [ˈmak] [ˈmakʲɪ] [ˈmaɡʲɪ] [ˈzʲe.mʲlʲɪ] [zʲɪˈmʲlʲi] [vɐˈzʲitʲ] [ˈvo.zʲɪt]
    • phonemic /ˈmak/ /ˈmak/ /ˈmakʲɪ/ /ˈmaɡʲɪ/ /ˈzʲe.mʲlʲɪ/ /zʲɪˈmʲlʲi/ /vɐˈzʲitʲ/ /ˈvo.zʲɪt/
    • morphophonemic //ˈmak// //ˈmaɡ// //ˈmak‿ʲi// //ˈmaɡ‿ʲi// //ˈzʲe.mlʲi// //zʲeˈmlʲi// //voˈzʲitʲ// //ˈvo.zʲit//
  • Excellent answer. Is it possible that silent ⟨gh⟩ in English is still morphophonemically significant? This would account for the different sound in expedite → expeditious versus right → righteous. Commented Feb 3, 2014 at 23:58

[fubar] is used for phonetic transcriptions (emphasizing what is actually articulated) and /fubar/ for phonological transcriptions (emphasizing what is phonologically contrastive)1.

For example [pʰɛʔts] vs /pets/ (or even /petz/) with the implication that former is predictable from the latter using phonological rules.


  1. By “phonologically contrastive”, I mean that in English, there are no cases where alternating [pʰ] and [p] changes the meaning of an utterance. Which one is articulated is entirely predictable from the context and so many theories models this as a single /p/. /p/ and /b/ contrast because, for example, /pat/ and /bat/ are different words. [pʰ] and [p] do not contrast at this level.
  • 1
    what do you mean by something that is "phonologically contrastive"?
    – Louis Rhys
    Commented Sep 20, 2011 at 4:53
  • 3
    @LouisRhys in English, there are no cases where alternating [pʰ] and [p] changes the meaning of an utterance. Which one is articulated is entirely predictable from the context and so many theories models this as a single /p/. /p/ and /b/ contrast because, for example, /pat/ and /bat/ are different words. [pʰ] and [p] do not contrast at this level. Commented Sep 20, 2011 at 6:40

A complement to the answers above:

The alphabet used for transcribing and the level of transcription are frequently confused. IPA can, just as any other alphabet, be used for both phonetic and phonemic transcription (the difference has already been neatly explained), or anything in between. One might argue IPA is richer, poorer, more or less readable, flexible, regular, whatever, but in principle there is nothing in it, or any other alphabet, that forces you to use it at any particular level of abstraction. Alphabet and level of abstraction are two independent variables.


This answer from Prof. Kevin Russell (BSc Manitoba; MA, PhD USC) should help.

There is no such thing as the transcription of a word.

Strictly speaking, you can only transcribe a single utterance -- for example, how Kevin Russell pronounced the word cat at 12:58:03 pm on February 4, 2004. You can transcribe this utterance as exactly as possible, within the limits of your hearing and the conventions provided by the IPA.

If you want to go beyond that, to try to describe how Kevin Russell pronounces the word in general, or further still to how English speakers pronounce it in general, then you have to start making abstractions -- you have to decide which details to include and which to ignore.

It's common to distinguish between two kinds of transcription, based on how many details the transcribers decide to ignore:

  • Narrow transcription: captures as many aspects of a specific pronunciation as possible and ignores as few details as possible. Using the diacritics provided by the IPA, it's possible to make very subtle distinctions between sounds.
  • Broad transcription (or phonemic transcription): ignores as many details as possible, capturing only enough aspects of a pronunciation to show how that word differs from other words in the language.

The key factor in a broad transcription is meaning -- if a pronunciation detail can change the meaning of words in a language, it must be included in a broad transcription of that language.

For example, consider the difference between the vowels in [liv] and [lɪv].

  • For Canadian English, a narrow transcription would note the difference between the [i] and the [ɪ]. So would a broad transcription, since leave and live mean different things.
  • For Canadian French, a narrow transcription would note the difference between [i] and [ɪ]. But a broad transcription would not. [liv] and [lɪv] do not mean different things in Canadian French -- they're both ways of saying 'book'. Both [i] and [ɪ] occur in the language, but they never contrast, that is, they never cause a difference in meaning. So a broad transcription would ignore the difference and write both as [liv].

With the symbols covered in this section, we are able to make broad transcriptions of Canadian English.

Some frequently asked questions about broad and narrow transcription.


Professional phonologists need a notation that distinguishes a theory or a perception of pronunciation, for which they often use slashes, from actual facts about the pronunciations, for which they often used square brackets. In part, this is because a phonologist will usually be aware of several different theories about a given interesting pronunciation. Without a systematic way of distinguishing fact from theory, a phonologist's world would be even more confusing than it is.


Typically, use slashes for broad [phonemic] transcriptions and brackets for narrow [phonetic] transcriptions.

I hope that I got that right.

  • 1
    can you explain what is a phonemic and a phonetic transcription?
    – Louis Rhys
    Commented Sep 20, 2011 at 4:52
  • Phonetics (where you would use a phonetic transcription) deals with the actual audio properties of spoken language. Phonology (where you would use a phonemic transcription) deals with how those sounds are perceived in a given language. Commented Sep 20, 2011 at 23:40

It is important to note that many dictionaries use the square brackets instead of slashes, but none of these is actually giving a close or broad phonetic transcription. A dictionary entry is (necessarily?) phonemic. Thus their use of square brackets is (for those who know the difference) incorrect, if not to say annoying.

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.