Secondary articulation occurs when the articulation of a consonant is equivalent to the combined articulations of two or three simpler consonants, at least one of which is an approximant. The secondary articulation of such co-articulated consonants is the approximant-like articulation. It "colors" the primary articulation rather than obscuring it. Maledo (2011) defines secondary articulation as the superimposition of lesser stricture upon a primary articulation.
For example, the voiceless labialized velar plosive [kʷ] has a stop articulation, velar [k], with a simultaneous [w]-like rounding of the lips. This is in contrast to the doubly articulated labial-velar consonant [k͡p], which is articulated with two overlapping stop articulations.
There are a number of secondary articulations. The most frequently encountered are labialization (as with [kʷ]), palatalization (as with the Russian "soft" consonant [tʲ]), labio-palatalization (as in the name Twi), velarization (as with the English "dark" L [lˠ]), and pharyngealization (as with the Arabic "emphatic" consonant [tˤ]). It can be difficult to distinguish primary and secondary articulation. For example, the alveolo-palatal consonants [ɕ ʑ] are sometimes characterized as a primary articulation of their own and sometimes as palatalization of postalveolar fricatives, [ʃʲ ʒʲ] or [s̠ʲ z̠ʲ].
The most common method of transcription in the IPA is to turn the letter corresponding to the secondary articulation into a superscript written after the letter for the primary articulation. For example, the w in ⟨kʷ⟩ is written after the k. This can be misleading, as it iconically suggests that the [k] is released into a [w] sound, when actually the two articulations are often pronounced more-or-less simultaneously. Secondary articulation often has a strong effect on surrounding vowels, and may have an audible realization that precedes the primary consonant, or both precedes and follows it. For example, /akʷa/ will not generally sound simply like [akwa], but may be closer to [awkwa] or even [awka]. For this reason, the IPA symbols for labialization and palatalization were for a time placed under the primary (e.g. ⟨k̫⟩ for [kʷ] and ⟨ƫ⟩ for [tʲ]), and a number of phoneticians prefer such unambiguous usage (with ⟨kʷ⟩ and ⟨tʲ⟩ only used for off-glides) despite the official policy of the IPA. In the official IPA there remains an alternate symbol for velarization/pharyngealizaton that is superposed over the primary (e.g. ⟨ɫ⟩ for dark L), but it has font support for a limited number of consonants and is inadvisable for others, where it can be illegible.
There is a longstanding tradition in the IPA that one may turn any IPA letter into a superscript, and in so doing impart its features to the base consonant. For instance, [ʃˢ] would be an articulation of [ʃ] that has qualities of [s]. However, the features are not necessarily imparted as secondary articulation. Superscripts are also used iconically to indicate the onset or release of a consonant, the on-glide or off-glide of a vowel, and fleeting or weak segments. Among other things, these phenomena include prenasalization ([ᵐb]), prestopping ([ᵖm, ᵗs]), affrication ([tᶴ]), pre-affrication ([ˣk]), trilled, fricative, nasal, and lateral release ([tʳ, tᶿ, dⁿ, dˡ]), roticization ([ɑʵ]), and diphthongs ([aᶷ]). So, while ⟨ˠ⟩ indicates velarization of non-velar consonants, it is also used for fricative release of the velar stop (⟨ɡˠ⟩). Mixed consonant-vowels may indicate a transition: [ᵇa] may be the allophone of /a/ with the transition from /b/ that identifies the consonant, while [fʸ] may be the allophone of /f/ before /y/, or the formants of /y/ anticipated in the /f/.
The following superscript variants of IPA letters are supported by Unicode. Dots mark unsupported IPA letters. Cells in grey are articulations that do not have IPA letters in the first place, either because such articulations are judged impossible, or because they are indicated with diacritics. For the latter, secondary diacritics must be used with superscripts just as they are with full letters. (For example, a superscript dental nasal would be ⟨ⁿ̪⟩.)
|Stop||ᵖ ᵇ||ᵗ ᵈ||. .||ᶜ ᶡ||ᵏ ᶢ||. .||.||ˀ|
|ᶲ ᵝ||ᶠ ᵛ||ᶿ ᶞ||ˢ ᶻ
|ᶴ ᶾ||ᶳ ᶼ||ᶝ ᶽ||ᶜ̧ ᶨ
|ˣ ˠ||ᵡ ʶ||. ˁ/ˤ||ʰ ʱ|
|ᶭ ʷ .
There are no superscript implosive, click or ExtIPA letters, except accidentally with ⟨ꜝ⟩, ⟨ꜞ⟩. With a properly designed font, combining diacritics (such as the bridge for dental consonants, ring for voicelessness, etc.) will work with superscript letters, as in ⟨ᵑ̊ǃ⟩. Spacing diacritics, however, as in ⟨tʲ⟩, cannot be secondarily superscripted through hard coding: ⟨ᵗʲ⟩.
The three superscript mid central vowels correspond to the complete set of mid central vowels in the IPA immediately after the Kiel Convention.
The precomposed rhotic vowels are not supported, but the rhotic spacing diacritic works fairly well despite not being superscripted: ⟨ᵊ˞ ᵌ˞ ᵋ˞ ᵓ˞ ᵅ˞⟩. So does the ejective spacing diacritic: ⟨ᵖʼ ᵗʼ ᶜʼ ᵏʼ⟩. Combining diacritics work as normal, though they may be oversized in some fonts: ⟨ᵓ̃⟩