How does nonconcatenative morphology of the Semitic type (consonantal roots, vocalic templates + affixes) arise diachronically?

It's pretty easy to see how a nonconcatenative inflectional system could morph into a more typical, concatenative system: sound change could mess up the templates in different ways for different lexemes, leading to analogical extensions and leveling of the vowel alternations. But it's harder to envision a process that goes the other way, starting with a linear synthetic language, or an isolating one, or an agglutinating one, and ending up with a root-and-template system. Probably this is an unanswerable question for lack of data, but are there any plausible speculations?

  • 3
    FYI the same thing applies to Indo-European languages; Semitic languages are like lexicalizing the agglutination of multiple ablauting roots together. This is just a wild thought off the top of my head, but perhaps this could have arisen through vowel harmony with a suffix/prefix that was phonologically eliminated. Trique for instance has cases (e.g. several inflections) where a morpheme was affixed then eliminated, leaving only a change in tone at the word boundary to show its existence. Commented May 19, 2014 at 0:28
  • 1
    Note: When I mention vowel harmony I'm merely talking in theory, not about any particular language. Commented May 19, 2014 at 0:30
  • @JustinOlbrantz That's an interesting idea, and of course it's basically what you see in Germanic umlaut, e.g. foot:feet, except there it isn't vowel harmony but partial assimilation.
    – TKR
    Commented May 19, 2014 at 0:38
  • 1
    I don't think it's vowel harmony. In fact, vowel harmony would restrict the number of possible vowel patterns. But one should also be careful with the assumption of easy changes from root-and-pattern languages onward. The opposite seems to be the case, namely that this type of language is relatively stable. If changes come, they seem to come from language contact and borrowing. Commented May 20, 2014 at 14:28
  • 2
    @ThomasGross I don't see how the assumption that language begins with nouns (which strikes me as both implausible and unprovable, but I haven't read the article you cite) is relevant -- my question isn't about the original genesis of language, but about how one type of language might change into another.
    – TKR
    Commented May 21, 2014 at 18:58

4 Answers 4


I can only offer some information obtained from two references I have recently come across when looking for answers to a similar question. Unfortunately, I am neither a Semitist nor an Afrasianist, hence my understanding of the issues is probably imperfect. In spite of this, here is what I have found so far.

Let me begin with a few quotations from an interesting dissertation by Simpson (2009), available here as a PDF (emphasis mine), such as this one (p. 5):

The system of root-and-pattern morphology incorporates elements of phonology and syntax. Morphology in general holds a position between these two domains and as a result reflects the results of changes originating in both domains. The unique properties of Semitic morphology exhibit the results of changes from many different domains including morphologically specific ones. Understanding the processes in morphology, syntax and phonology and how they have contributed to this unusual system of word formation provides important insights into the patterns that occur and those that do not. While the patterns and configuration of Semitic morphology are unique, they reflect processes and underlying mechanisms which are not by themselves unusual. These changes rely on basic phonetic and cognitive mechanisms, as well as the inherent ambiguities in language and the opportunities these ambiguities present for reanalysis. This principle can be seen as underlying many of the changes in language and more specifically those encountered in this study. The results of reinterpretation may result in cross-linguistic patterns but these patterns are considered here to be the results of, not the driving forces behind, the changes.

Also (p. 9):

From the history of Arabic it might seem that the system of ablaut patterns is characterized by a slow process of decay, whereby the system of non-linear morphology represents the oldest, or even the original, stage of the language with a subsequent break down of the system over time. However, once the Semitic family is considered as a whole, this scenario does not hold up so well. The history of the Semitic language family and its subfamilies includes many cases of patterns being lost, but also many cases of new patterns emerging. Many of the non-linear morphological patterns of Classical Arabic are likely innovations not originally found in Proto-Semitic. The modern history of most Semitic languages is characterized by both the loss of older ablaut patterns and the formation of new patterns, giving us a very dynamic picture of developments in the domain of ablaut patterns.

And further (p. 68):

While many patterns characteristic of the earliest varieties have been lost or become obsolete, there are at the same time a number of new patterns and newly productive forms found in many Semitic languages. The types of processes that introduce new alternations or expand the use of existing patterns are proposed to be essentially the same in non-Semitic languages as they are both in the earliest stages of the Semitic and Afroasiatic families and in later varieties of Semitic.

The author then divides the processes into the following main two types:

  • segmentally conditioned alternations (such as long-distance assimilation leading to umlaut-like patterning)
  • prosodically conditioned alternations (basically loss/reduction of vowels in certain contexts, changes in vowel quality/quantity due to the placement of stress, with examples from several languages including Hebrew, Maltese, and the Modern South Arabian languages)

I strongly suggest that you read the text to learn more. I'm not a Semitist and many of the Semitic-specific details are difficult to comprehend without a proper Semitist background, but otherwise the linguistic reasoning seems rather sensible most of the time.

The author himself admits that the ultimate origin of all of the patterning is currently not discernable, but assumes - and I tend to agree - that there is no reason to think it was very different from what we know of the later history of the family and its individual daughters.

Now, as far as Proto-Afrasian or Proto-Afroasiatic is concerned, a concise introduction to the issue can be found in Bomhard's Afrasian Comparative Phonology and Vocabulary (2014:15).[see the note below] Apart from some information on how Diakonoff vs. Orël-Stolbova vs. Ehret reconstructions differ and/or compare, there seems to be some interesting information concerning the reconstruction of Proto-Semitic (e.g. p. 37) and pertaining to the origin of the Semitic "templatic" morphology (p. 39-42), a few lines of which I can quote here as well (p. 39):

There has been much discussion, some of it rather heated, concerning root structure patterning within Afrasian. Until fairly recently, there was strong resistance to look objectively at the data from all of the branches of the Afrasian language family, far too much emphasis being placed on the importance of the Semitic branch alone, which was often uncritically taken to represent the original state of affairs. In the Semitic branch, the vast majority of roots are triconsonantal. It is certain, however, that at one time there were more biconsonantal roots and that the triconsonantal system has been greatly expanded in Semitic at the expense of roots with other than three consonants (cf. Moscati 1964:72—75; Ullendorf 1958:69—72; Militarëv 2005).

And further (p. 40):

Thus, the Proto-Afrasian root may be assumed to have had two forms, either *CV or *CVC. *CVC could be extended by means of a suffix to form an inflectional stem: *CVC-(V)C-. Originally, these suffixes appear to have been utilized primarily as verb extensions. Depending upon when they became separated from the rest of the Afrasian speech community, each branch exploited to a different degree the patterning that was just beginning to develop in the Afrasian parent language, with Semitic carrying it to the farthest extreme.

Hence, it appears that (1) languages such as Classical Arabic have developed the patterning to an utmost extreme, but (2) not only may the system have been much less "extravagant" at the Proto-Semitic stage, (3) it was also much less profound at the Pre-Proto-Semitic or Proto-Afrasian stage, which seems to be supported by both, internal evidence (in the former case) and external comparisons (in the latter).

While I definitely agree with @jknappen's answer in that we don't really know much, an interesting picture is emerging - essentially, one that actually also agrees well with @carsten's speculative proposal.

[Note: Although Bomhard is a Nostraticist, which could be discouraging to quite a lot of people, references are almost exclusively restricted to Semiticist and Afrasianist literature, with the exception of the Introduction and/or Appendix.]



Can't it be something to do with stress-dependent vowel reduction in one of the ancestors of Proto-Semitic? It can go to great lengths. Say, in Russian, NOM of "horse" is lošəť (in some regions: lošť), and GEN PL is ləšɨdej; meanwhile the diminutive of "horse" is lɐšatkə ("little horse"). To an outsider, it may look pretty much like Semitic consonantal roots, but it's not. It' simply predictable cases of vowel reduction thanks to Russian's dynamic stress: phonemes /a/ and /o/ collapse to [ə] in unstressed syllables, except when immediately before the stressed syllable, where it turns to /ɐ/. Additionally, /š/ (also /ž/ and /ts/) can trigger mutation to [ɨ] (soft consonants also trigger mutation to [ɪ]).

So it's possible that, in one of the ancestors of Proto-Semitic, something similar could be happening. Then, at some point, the vowel reduction law stopped functioning for some reason, basically fossilizing all those vowel mutations. The type of stress must also change, I guess. After the fossilization, trying to figure out all that mess, the language then would try to find certain post-hoc "patterns". Note that there are still affixes without problems. Add also (over the course of hundreds, thousands of years):

  1. possible sound losses after that (imagine some sound that would trigger a very specific vowel mutation, but then it would collapse with another sound which had never triggered mutation, and voila, synchronically now there's no way to explain why we have /CV₁/ in one word form, but /CV₂/ in another word form)
  2. various contractions; for example, imagine a hypothetical language where [əw] > [u]; and later [w] dissappears altogether, leaving forms like CuC ~ CaC, which stem from CəwCV ~ CawCV

etc. -- and you can get something totally different from the original language

Vowel reduction in Russian is still active and alive, but there are some cases of this kind of fossilization already happening (colloquially), for example:

pla "payment" > plɐtiť "to pay", and then the first /ɐ/ in the verb is misrecognized as stemming from /o/ (hypercorrection, as both /o/ and /a/ collapse to /ɐ/), so in 3p sg. we get plotit instead of platit (the stress change is required by this type of word). Voila, a new kind of pattern has just evolved, ex nihilo, basically, where plat- suddenly alternates with plot-, something which never happened in Old Russian.

This is just one of possible explanations, I don't pretend it to be the real answer to how it really formed in Proto-Semitic.


  • Apophony (ablaut) seems to be a different phenomenon than root-and-template morphology system of Semitic languages. The key factor is pattern: in Arabic/Semitic, for a fixed arbitrary vowel template and various consonant roots, one may construct words, even not knowing them from a dictionary. However, in Russian, ablaut is rather a phonetic tool used for phonetic harmony. Consider: diminutive лошадь→лошадка (different stress pattern), but печка→печечка (no change of stress pattern). Contrary, SNG GEN лошадь→лошади (no change), but печка→печи (change). Commented Sep 3, 2015 at 19:08
  • In Russian, the actual "ablaut" as it is can construct new words too, without knowing them from a dictionary. If a verb contains -o- in the root, you can predictably construct frequentative forms by replacing -o- with -a- (even in roots where etymologically it does not make sense): ходить-расхаживать, носить-разнашивать, топить-растапливать etc.
    – carsten
    Commented Sep 4, 2015 at 17:19
  • The examples with stress patterns different from I cited don't prove anything, however. There's the tendency for words in Russian to adopt stress patterns from other words, and it's moving towards the stress patterns where stress is more differentiated in various word forms. It's basically to combat vowel reduction, as vowel reduced endings make word forms ambigious (schwas all over the place), so the language tries to fix it by adopting such stress patterns as the standard where dynamic stress is more differentiated, so that endings received different degrees of vowel reduction (less ambig.)
    – carsten
    Commented Sep 4, 2015 at 17:26
  • носить→разнашивать did not convince me either. :) It rather looks like a chain of раз- prefix (носить→разносить) and -ив- suffix which denotes progressive tense (носить→нашивать). Only then the vowel harmony takes place to modify root vowel о into а. But again, this happens only because о is in weak position. Can it be treated as a morphology tool? Commented Sep 5, 2015 at 2:54

Accadian verbs had 4 types of vowel patterns of which only one type, the a-u-class, had apophony of significance. The other 3 patterns mostly used stable vowels a, i and u. If this is an old situation or an innovation in Accadian, I do not know.

On the other hand, ancient Egyptian, seems to have had the same type of non-concatenative morphology as Semitic, and that seems to indicate that this type of morphology is rather old.

The origine of such a morphology could simply have been umlauts due to influence of neighbouring affixes, possibly combined with vowel reductions due to stress patterns. If the language in the first place had only a few vowels, like a, e, u, i, this could easily result in a situation where the flectional forms or derivative processes totally determined the vowels of the stem.


Since you ask what is known, here is the answer that may be a little disappointing to you:

Even in the oldest attested Semitic language (Akkadian ca. 2500 BC) the typical Semitic morphology is already fully developed. Also Proto-Semitic (dated ca. 3500 BC) has the typical morphology.

Anything before that is subject to speculation (AFAIK, there is no reasonable hope for a scientific reconstruction of Proto-Afro-Asiatic).

  • What should we infer about your position on Ehret's work?
    – user6726
    Commented Aug 28, 2015 at 18:33
  • @user6726 AFAIK, not even Ehret dared reconstruct a proto-Afro-Asiatic morphology (which could shed some light on the evolution for the Semitic morphology). Essentially, we have enough evidence to establish the Afro-Asiatic family, but not enough for something I'd call a "reconstruction". Commented Aug 29, 2015 at 10:43

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.