Velar consonants are the most common type of dorsal consonants, pronounced with the dorsum of the tongue raised towards the soft palate (or velum). Velarization utilizes the same gesture, but with another primary articulation. Velar or further back closure is used to produce the airflow in clicks, but their POA are generally defined by the front articulation (this also implies the impossibility of a velar click; not that it stopped IPA from creating a symbol for one).
Back vowels may also be characterized as velar, tho for those in the [ɔ] to [ɑ] region "uvular" or "pharyngeal" may be more accurate descriptions.
The voiceless velar stop /k/ is nigh-universal in languages of the world and commonly a high-frequency consonant. When aspiration or ejectiveness are contrasted, the same applies to these counterparts. Languages lacking /k/ have generally previously shifted it to /ʔ/ (examples include Tahitian and perhaps Xavante) or have other voiceless dorsal stops, usually at least /kʷ/ (sometimes /kp/ (Klao) or /q/ (Hupa)). A third , more academic possibility is a language not making any voicing distinction in stops and having [k] as an allophone of a consonant best described as /g/.
By contrast, the voiced velar stop /g/ (or /ɡ/, if you are a stickler for proper typography) is considerably more unstable than its other "basic" counterparts /b/ and /d/; languages with a hole at /g/ are found randomly scattered around the world, with little areal relevance. Known pathways of elimination include → /ɣ/, also a likely intermediate on the road towards loss. Arabic has freak unmotivated unconditional palatalization to /ʤ/ (while in many dialects leaving /k/ be). (I think this is one of a number of hints for Classical Arabic /q/ having been [ɢ] --ed.) The implosive /ɠ/ is subject to a similar limitation.
A curious exception of stop distribution is (Xalxa) Mongolian, which lacks /k/ but has /g/.
Voiceless velar affricates occur mostly as an areal feature in Southern Africa, commonly furthermore being ejective and in some cases (Zulu, Archi) lateral. Voiceless velar lateral affricates tend to vary with clusters of a velar stop and an alveolar lateral (Zulu, Archi (might just be notational difference there)). It is also fleetingly found in some High German dialects, demonstrating its role as the intermediate in the change /kʰ/ → /x/.
Voiced velar affricates are completely unattested, both as phonemic and as a predictable allophone of any other consonant.
Lateral velar fricatives are possible, but rare.
The velar nasal /ŋ/ is even more commonly lacking as a phoneme than /g/, tho for some reason its distribution is more tied to areal factors. Additionally, among languages that do have /ŋ/, there is also a strong tendency to disallow the word-initial position (examples include the Germanic languages; exceptions include Austronesian and most languages of Siberia) and sometimes any onset position (Mandarin).
The labial-velar approximant /w/ is more common than the plain velar version /ɰ/. This is not surprising, since both are high back semivowels, and the high back rounded vowel /u/ is also more common than the high back unrounded vowel /ɯ/.
A velar lateral approximant /ʟ/ is very rare (a velarized alveolar /ɫ/ is somewhat more common), to the point of being about as common as its obstruent counterparts.
Velar trills or flaps are impossible (but see velopharyngeal trill)
Sound changes involving velars
Velar consonants may commonly be palatalized into palatal consonants near front vowels (which may then decay via assibilation or similar pathways). This change is also attested as spontaneous in languages that contrast velars with uvular consonants, especially where there is a simultaneous rounding contrast, in indigenous languages of the Pacific coast of North America. This is also hypothesized of Proto-Indo-European in some reconstructions, considering that there are (AFAIK --ed.) no clear examples of palatal stops becoming velar stops.
Another typical POA change is retraction to uvular near back vowels (a much rarer spin on this occurs in Lahu, where velars become uvular before all vowels). Spontaneous backing is attested for the fricatives in many Afro-Asiatic languages.
Labialization can occur near rounded vowels, and resulting labialized velars are capable of turning to labials (eg. Greek). The change of *w → /v/, ubiquitous in continental European languages, also falls under this.
Sound changes that systematically create velars are rare. Regardless there are attested examples, including
- labialized labials → labialized velars (in Oceanic)
- linguolabials → labialized velars (in Athabascan; via a similar intermediate as the previous?)
Sound changes creating some individual velar sound are somewhat more numerous:
- in languages with small consonant inventories originally lacking velar stops, /p/ → /k/ (Arapaho), /t/ → /k/ (in Oceanic, generally part of a pull chain with /k/ → /ʔ/ previously)
- /tʰ/ → /tx/, in some languages further → /kx/ (Athabascan); preaspirate stops → /x/ + stop (Celtic)
- /ʃ/ → /x/ (in Slavic (ruki), Spanish, Pashto, possibly Finnic, reportedly certain dialects of Japanese)
- /ʒ/ → /ɣ/ (Pashto)
- /f/ → /x/ (some Salish languages, a few Italic languages)
- /h/ → /x/ (probably somewhere)
- /h/ → /ŋ/ (rhinoglottophilia - Avestan, Nyole)
- /n/ → /ŋ/ (Khalkha Mongolian, word-finally)
- /ɫ/ → /w/ (common, eg. French), /ɰ/ or /ɣ/ (Armenian)
- the same, starting from /r/ (English seems to be on this road)
- creation of /w/ from (back) vowels by diphthongization
- ∅ → /w/ epenthetically near rounded vowels
- ∅ → /ŋ/ word-initially (a Samoyedic idiosyncrasy)
- ∅ → /k/ epenthetically between vowels (example needed)
- Ugaritic appears to have very strange changes from original *θʼ / *ðˤ (emphatic interdental) and *tɬʼ / *ɬˤ (emphatic lateral) to *ɣ.
- /ð/ → /ɣ/ (Irish and Scottish Gaelic)