Can there be family-specific trends, or family-specific causal factors, in linguistic evolution?

Over the last decade, there have been repeated suggestions that causal factors in diachronic change (or linguistic evolution) might be family-specific (or “lineage-specific”, using a term from biology). For example, Dediu & Levinson (2012) say:

“We show here that the cultural evolution of structural features is simultaneously shaped by universal tendencies, language family-specific factors and deep genealogical and areal processes acting across language families.“

I find this suggestion very puzzling, because causal factors must be either universal (having to do with universal properties of human communication and cognition), or restricted to particular historical situations. Examples of the latter might be the influence of writing when there is widespread literacy, or the influence of Arabic on languages spoken by Muslim populations, or large-scale geographical patterns such as tone in African languages or SOV order in northern Eurasian languages (which must be due to contact influences over extended time periods).

Could there be family-specific causal factors, and thus family-specific trends? (I am assuming that a “trend” is a series of changes occurring at more or less the same time, caused by the same factor(s).) That is difficult to imagine, because the languages belonging to a family share a common origin, but nothing that could be related to causal factors for change. Speakers of Austronesian or Indo-European languages do not know (even unconsciously) that their language belongs to a family, and whatever change their languages are undergoing cannot be related to this earlier history, it seems to me.

So maybe when the authors of these recent works talk about family-specificity, they merely mean absence of universality? This does not seem to be the case, because sometimes the reference to causal factors seems clear, e.g.

“Contrary to the Greenbergian generalizations, we show that most observed functional dependencies between traits are lineage-specific rather than universal tendencies.“ (Dunn et al. 2011)

“Our approach to test the relative merits of a universal versus lineage-specific dynamics regarding word-order features takes more sources of information into account than just correlation patterns.” (Jäger & Wahle 2021: 12)

“…shed light on the question of whether the causes of the correlations in Indo-European are due to family-specific factors or whether these are more general historical trends of natural language“ (Goldstein 2022: 695)

“Here, we assess if the potential evolutionary constraints on the diversity of gender systems are universal or family-specific.“ (Shcherbakova & Allassonnière-Tang 2024: 3)

“Functional dependencies”, “dynamics” and „evolutionary constraints” seem to be clear references to causal factors, which can hardly be epecific to a family. Likewise, reference to “trends” can only be interpreted as a set of changes that are due to common factors, so it is puzzling that some authors make reference to trends that are specific to certain families, e.g.

“Evolved structure of language shows lineage-specific trends in word-order universals” (Dunn et al. 2011)

“this paper [provides] a diachronic analysis, using a quantitative evolutionary framework to assess processes of change and the evolutionary trends of grammatical gender in Indo-Aryan languages.“ (Allassonnière-Tang & Dunn 2020: 2)

“On the basis of trends within Indo-European, we confirm broader hypotheses regarding crosslinguistically universal tendencies, namely, that factors such as economy and frequency interfere in the processes of language evolution.“ (Carling & Cathcart 2021: 589)

In some of these papers, the hypothesized trends are not observed in all families, so there is absence of universality. That certain expected (or unexpected) patterns are not found everywhere but only in some families is of course possible. Some references to family-specificity in the literature only talk about observed patterns (“linkages”, “relationships”):

“These family-specific linkages suggest that evolutionary processes of language diversification explore alternative ways to construct coherent language systems unfettered by tight universal constraints.“ (Dunn et al. 2011)

“These findings show that the relationship between semantic rules and agreement patterns is lineage-specific“ (Shcherbakova & Allassonnière-Tang 2024: 11)

This is completely unproblematic, because there is no causal claim. Observing some (apparent) patterns only in some data subsets or some geographical areas is a common experience in linguistics (and in science more general), and one might entertain a range of hypotheses in such circumstances. So should we perhaps understand some of the other claims about “dynamics”, “constraints” and “trends” as being merely about observed patterns? I wish these papers were more precise about the claims that they want to make.

One possibility that has been suggested to me is that the languages of a family might share certain features because they were inherited from the common ancestor, e.g. grammatical tone in Bantu, or prepositions in Austronesian, or case-number cumulation in Indo-European. If a set of languages have certain typological patterns, they might be prone to undergo specific changes that other languages are much less likely to undergo. Now that is certainly true, but in such cases, the dynamics or the trends would be type-specific, not really family-specific. Most relevant types occur in other families as well and should be tested there, too – and of course, some typological features that seem characteristic of particular families do not occur in all languages of the family (not all Indo-European languages have case-number cumulation; not all Bantu languages have grammatical tone; not all Austronesian languages have prepositions). So it seems to me that at best, a family could be treated as a “rough proxy” for a particular type – but since we can investigate types directly, this is what we should do.

Another possibility is of course that the languages of a family are still spoken in a very compact area and are undergoing similar changes because of geographical proximity (to each other, and/or to other major languages), e.g. Dagestanian languages, or Munda languages. Again, “family” would then be correlated with something else, but it is really this other factor that is causal. Some families have geographical outliers that behave in special ways, confirming that the true causal factor is language contact, not belonging to a certain family (e.g. Sri Lanka Malay (Austronesian), which has case suffixes rather than prepositions; or Ossetic (Iranic, Indo-European), which has number-case agglutination instead of cumulation).

Taking “family” (or “lineage”) as a proxy for a type or an area may seem innocuous, but I don’t think it is. Language types and language families have often been confused in the past, and it was only with Greenberg’s work (e.g. 1957) that it was fully established that they must be kept apart quite strictly. As late as the 1920s, it was normal to argue for (un)relatedness of languages on the basis of typological features, such as gender in African languages (e.g. in Carl Meinhof’s work, which was widely respected at the time; e.g. Meinhof 1936). By seemingly talking about families when type or area is intended, linguists run the risk of blurring a distinction that constituted major progress in comparative linguistics in the 20th century.

So I am confused by some of this recent literature (in the wake of Dunn et al. 2011), and I hope that future work in this tradition will be clearer about the kinds of causal connections and factors that are intended.


Allassonnière-Tang, Marc & Dunn, Michael. 2020. The evolutionary trends of grammatical gender in Indo-Aryan languages. Language Dynamics and Change 11(2). 211–240. (doi:10.1163/22105832-bja10011)

Carling, Gerd & Cathcart, Chundra. 2021. Reconstructing the evolution of Indo-European grammar. Language 97(3). 561–598. (doi:

Dediu, Dan & Levinson, Stephen C. 2012. Abstract Profiles of Structural Stability Point to Universal Tendencies, Family-Specific Factors, and Ancient Connections between Languages. PLOS ONE 7(9). e45198. (doi:10.1371/journal.pone.0045198)

Dunn, Michael & Greenhill, Simon J. & Levinson, Stephen C. & Gray, Russell D. 2011. Evolved structure of language shows lineage-specific trends in word-order universals. Nature 473(7345). 79–82. (doi:10.1038/nature09923)

Goldstein, David. 2022. Correlated grammaticalization: The rise of articles in Indo-European. Diachronica. John Benjamins 39(5). 658–706. (doi:10.1075/dia.20033.gol)

Greenberg, Joseph H. (Joseph Harold). 1957. Essays in linguistics. Chicago: University of Chicago Press. (

Jäger, Gerhard & Wahle, Johannes. 2021. Phylogenetic typology. Frontiers in Psychology 12. 2852. (doi:10.3389/fpsyg.2021.682132)

Meinhof, Carl. 1936. Die Entstehung flektierender Sprachen: Eine Untersuchung. Berlin: Reimer.

Shcherbakova, Olena & Allassonnière-Tang, Marc. 2024. Evolutionary pathways of complexity in gender systems. Journal of Language Evolution lzae001. (doi:10.1093/jole/lzae001)

Cite this blog post
Martin Haspelmath (2024, May 14). Can there be family-specific trends, or family-specific causal factors, in linguistic evolution? Diversity Linguistics Comment. Retrieved May 30, 2024, from

3 thoughts on “Can there be family-specific trends, or family-specific causal factors, in linguistic evolution?

  1. An excellent post and I agree with virtually everything in it. Just one minor remark. Sometimes we do observ leneage-specific trends which are truely leneage-specific (i.e. not type-specific) and truely a trend. We usually call such cases ‘conspiracies’ because we don’t understand their causes. The best known case is the so-called Open Syllable Conspiracy in Slavonic (a sub-brahcn of Balto-Slavonic, which is a branch of Indo-European). The term Open Syllable Conspiracy referes to a series of sound changes which all the numerous Slavonic languages must have experienced partly independently of each other. The peculiar thing about this is the fact that these superficially very different sound changes (C-drop at word-ends, C-palatalisations, C-degemination, metatheses, emergence of nasal vowels, etc.) replace inherited closed syllables by open syllables. It is true that at the time the changes occurred Slavonic languages occupied a continuous geographic area. Accordingly, we can’t exclude an areal trend which just happened to affect all of Slavonic. The problem is, the trend affected only Slavonic and no other language of the region (i.e. no Baltic, Romance, West Germanic or Uralic language spoken in the neighbourhood). Besides, as early as 800 CE – i.e. approximately at the time the sound changes occurred – the area inhabitated by Slavonic languages already reached from Southern Greece to Lake Ladoga, and/or from the Oka river in the east to the Elbe in the west. This is too big an area to be affected by some hypothetical substrate, adstrate or superstrate having only open syllables. In effect, we can just describe the conspiracy but know nothing about its causes.

  2. A related assumption is used in computerized phylogeny. Autocorrelated relaxed clock models and local clock models both take the rate of evolution (molecular, lexical, etc.) to have some dependence on the lineage within which it occurs. As in the examples you give, these seem to be empirical approaches, not necessarily backed with a functional explanation.
    This also relates to Sapir’s drift, often noted but seldom explained.

  3. Great post! Can this be related to the notion of drift (“independent parallel developments”) and to Cristofaro’s source-oriented approach (that was discussed also in your blog)?

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.