The Centum-Satem division is an isogloss of the Indo-European language family, explaining the evolution of the three dorsal consonant rows reconstructed for Proto-Indo-European, *kʷ (labiovelars), *k (velars), and *ḱ ; (palatovelars). The terms come from the words for the number "one hundred" in representative languages of each group (Latin centum and Avestan satəm).
The Satem languages include Indo-Iranian, Balto-Slavic (Baltic and Slavic), Albanian, Armenian and perhaps also a number of barely documented extinct languages, such as Thracian and Dacian. Although Albanian is treated as a Satem language, there may be some evidence that the plain velars and the labiovelars were not completely merged in Proto-Albanian.
The Centum group is often thought of as being identical to "non-Satem", i.e. as including all remaining dialects. More specifically, in the sense of Brugmann's "languages with labialization", the Centum group includes Italic, Celtic, Germanic, Greek, and possibly a number of minor and little known extinct groups (such as Venetic and the ancient Macedonian language and probably the Illyrian languages).
Tocharian, on the other hand, combined all rows into a single velar row, and is therefore typically considered "Centum", although the relative chronology of the change is unknown. Likewise, the proto-language of the Anatolian languages apparently did not undergo either the Satem or the Centum sound change. The velar rows remain separate in Luwian, while Hittite may secondarily have undergone a Centum change, but the exact phonology is unclear.
The Centum-Satem isogloss explains the evolution of the three dorsal rows reconstructed for PIE, *kʷ , *gʷ , *gʷʰ (labiovelars), *k , *g , *gʰ (velars), and *ḱ , *ǵ , *ǵʰ ; (palatovelars) in the daughter languages. A division into a Centum and a Satem group does only make sense with a view to the parent language with the full inventory of dorsals. Later sound changes within a specific branch of Indo-European that are similar to one of the changes, such as the palatalization of Latin k to s in some Romance languages or the merger of *kʷ with *k in the Goidelic languages, have no effect on the grouping.
"palatal explosives". Brugmann terms the Centum languages "languages with labialization" or "u̯ -languages" and the Satem languages "languages without labialization", and he opines that
- For words and groups of words, which do not appear in any language with labialized velar-sound, [the "pure velars"] it must for the present be left undecided whether they ever had the u̯
-afterclap. (trans. J. Wright) By the 1897 edition of his work, Brugmann changed his mind, accepting the centum vs. satem terminology introduced by von Badke in 1890. Accordingly, he denoted the labiovelars as q u̯ , q u̯ h , g u̯ , g u̯ h (also introducing voiceless aspirates).
The presence of three dorsal rows in the proto-language is still not universally accepted. The reconstructed "middle" row may also be an artifact of loaning between early daughter languages during the process of Satemization. For instance, Oswald Szemerényi (e.g., in his 1995 Introduction), while recognizing the usefulness of the distinction *kʷ , *k , *ḱ as symbolizing sound-correspondences does argue that the support for three phonologically distinct rows in PIE is insufficient and prefers a twofold notation of *kʷ , *k . Other scholars who assume two dorsal rows in PIE include Kurylowicz (1935), Meillet (1937), Lehmann (1952), and Woodhouse (1998).
The Satem languages show the characteristic change of the so-called Proto-Indo-European palato-velars (*ḱ, *ǵ, *ǵʰ) into affricate and fricative consonants articulated in the front of the mouth. For example, *ḱ became Sanskrit ś [ʃ], Avestan, Russian and Armenian s, Lithuanian š [ʃ], and Albanian th [θ]. At the same time, the protolanguage velars (*k, *g, *gʰ) and labio-velars (*kʷ, *gʷ, *gʷʰ) merged in the Satem group, the latter losing their accompanying lip-rounding.
The Satem shift is conveniently illustrated with the word for '100', Proto-Indo-European *ḱm̥tóm, which became e.g. Avestan satəm (hence the name of the group), Sanskrit (śatam), Persian sad, Lithuanian šimtas, Russian sto, etc., as contrasted with Latin centum (pron. [kentum]), English hund(red)- (with /h/ from earlier *k, see Grimm's law), Greek (he)katon, Welsh cant, etc. (The Albanian word qint is a loanword from Latin centum.)
The status of Armenian as a Satem language as opposed to a Centum language with secondary assibilation like e.g. French (i.e. the collapse of the velars with labiovelars rather than with the palatovelars) rests on the evidence of a very few words.
In the Centum languages, the palato-velar consonants merged with plain velars (*k, *g, *gʰ). Most of the Centum languages preserve Proto-Indo-European labio-velars (*kʷ, *gʷ, *gʷʰ) or their historical reflexes as distinct from plain velars; for example, PIE *k : *kʷ > Latin c /k/ : qu /kʷ/, Greek κ /k/ : π /p/ (or τ /t/ before front vowels), Gothic /h/ : /hʷ/, etc.
and qairþra q) and Celtic (Ogham ceirt Q). Thus, while usually reconstructed for PIE, the labiovelar quality of this row may also be an innovation of the Centum group, causally related to the fronting of the palatovelars. The chief witness for this question is Anatolian, the phonology of which is for orthographical reasons not known in detail. Hittite (and Luwian) in any case chose not to use the existing cuneiform q- series (which stood for a voiceless uvular stop in Akkadian), but represents reflexes of PIE labiovelars as ku. Opinions on whether this represents an Anatolian single phoneme, or a group of /k+w/ are divided. The likelihood of three dorsal rows has also been disputed on typological grounds, but that argument has little merit, since there are, indeed, languages with such a three-row system, for example the Yazgulyam language (an Iranian language, but its system of dorsals is unrelated to PIE phonology). It is still true that such languages are rare, and one of the three rows is typically quite marginal. It is, then, in any case no surprise that no extant Indo-European language has preserved the full system, even if the parent language really did feature three rows at some point.
Origins of the sound change
In the 19th century, it was sometimes assumed that the centum-satem isogloss was the original dialect division of the Indo-European languages. However already Karl Brugmann, and in particular Johannes Schmidt regarded the Centum/Satem sound changes as an areal feature.
Incomplete Satemization in Baltic, and, to a lesser extent, Slavic, is taken as an indication of the diffusion of the satem sound change, or, alternatively, due to loans via early contact of Proto-Baltic and Proto-Germanic speakers. Examples of remnants of labial elements from labiovelars in Balto-Slavic include Lithuanian ungurys "eel" < *angʷi - , Lithuanian dygus "pointy" < *dʰeigʷ -, Fewer examples of incomplete Satemization are also known from Indo-Iranian, such as Sanskrit guru "heavy" < *gʷer -, kulam "herd" < *kʷel -; kuru "make" < *kʷer - may be compared, but it arises only post-Rigvedic in our attested texts.
Whether areal or dialectal, the centum/satem distinction was long considered to represent a division of Proto-Indo-European into western and eastern zones. The example of Tocharian, though, has led to a competing view of the satem sound change as an innovation radiating outward from the central Indo-European language communities, but largely failing to reach the western (European) or eastern (Tocharian) peripheries.