How coding efficiency explains cross-linguistic asymmetries: A reply to Song (2018: Ch. 7)

Jae Jung Song (1958-2017) published two typology textbooks, one in (2001) and another one earlier this year (which must have been finished just before his death), containing mostly new material. In particular, Chapter 7 of the new book deals with grammatical coding asymmetries and other “typological asymmetries”, as well as “markedness” and explanations in terms of iconicity, economy and frequency. Song makes extensive reference to some of my earlier work (Haspelmath 2006; 2008; Haspelmath et al. 2014), so it may be worth putting together a blogpost with some reactions.

I was really happy to see that Song takes up some of the important observations made by Greenberg (1966) (in his book “Language universals”), because they are still not widely enough known (even though Croft (1990; 2003) also has an entire chapter about “typological markedness”). In a wide variety of contrasts in grammar and the lexicon, we find cross-linguistically consistent patterns, with one meaning preferably expressed overtly, and a contrasting meaning zero-expressed:

singular/plural,  present/future,   nominative/accusative,   positive/comparative,   cardinal/ordinal, …

What explains these patterns? Song rehearses the complicated history of “markedness” (since Trubetzkoy and Jakobson) and then invokes a competition between “iconicity” and “economy”. Patterns that have zero-expression are supposedly motivated by economy, and patterns that have overt expression are motivated by iconicity – so one could say, for instance, that languages like Pirahã (with no singular or plural marking) are wholly economical, languages like Bayso (with overt singular and plural marking, e.g. luban-titi ‘lion’, luban-jool ‘lions’) are wholly iconic, and languages like English are a compromise. But then why don’t we find languages like English-prime with overt singular marking, and zero plural marking? In other words, why do all these pairs behave asymmetrically in that one of the four logically possible patterns does not occur (or hardly)?

Song notes that I have argued against markedness (Haspelmath 2006) and against some types of iconicity (Haspelmath 2008), and that I proposed explanations in terms of frequency (or economy) instead. But he does not note that the iconicity I argued against was iconicity of complexity, not iconicity of explicit marking (as one could call it – in the 2008 I called it “iconicity of syntagmatic isomorphism”). Many people had argued that plurals are overtly coded because the plural meaning is somehow “more complex” than the singular meaning, and I argued that instead, the plural is marked overtly because it is less frequent and thus less predictable. This is a kind of coding efficiency, i.e. the motivating factor is efficiency. I would thus say that Pirahã-type languages are motivated by parsimony (minimal encoding effort), Bayso-type languages are motivated by explicitness (minimal decoding effort), and English-type languages are motivated by efficiency (optimal tradeoff between encoding and decoding effort). (“Economy” is an unclear term like “iconicity” – it sometimes means ‘parsimony’, and sometimes ‘efficiency’).

Song explains my arguments in some detail, including the argument from causative constructions, which exist alongside anticausative patterns in a fascinating regularity that has long escaped linguists (Haspelmath et al. 2014). Quite generally, differential coding is poorly understood, and the diverse coding splits in the domain of argument coding and beyond (Haspelmath 2019) have rarely been seen together (Song treats differential object marking in the following chapter under the heading of “prototypes”, as if it had nothing to do with coding efficiency). Song realizes that differential-coding effects as in causatives and anticausatives are strong argument in favour of my view, but then he warns that Greenberg had good reasons for not replacing markedness by frequency:

“Frequency is itself but a symptom and the consistent relative frequency relations which appear to hold for lexical items and grammatical categories are themselves in need of explanation.” (1966: 70)

Song says that Greenberg “clearly does not regard the role that frequency of use plays as a causal one”, but this is wrong, as becomes clear when one continues to read Greenberg:

“… Such explanations will not, in all probability, arise from a single principle… This topic is left for future exploration.” (1966: 70)

Thus, Greenberg says that frequency asymmetries lead to uniform consequences (coding asymmetries), but have diverse ultimate causes. These ultimate causes are interesting to consider, but whatever we find will not be as general as the frequency-based explanation of the coding asymmetries.

It is not so important, of course, that Greenberg agrees with my position here, but Song has no other real arguments against the view that the coding asymmetries are due to frequency-based predictability and a preference for efficient grammatical coding. Of course, “frequency is merely a symptom of something deeper” (p. 168), but Song has no specific proposal of his own, and thus we have not made progress over Greenberg’s original insights here.

But there are two respects in which we should probably retreat to a more conservative position than the one taken by Greenberg. On the one hand, Song continues Greenberg’s (and Croft’s) practice of treating phonological asymmetries and coding asymmetries together. But they probably cannot be explained by predictability considerations, so something else seems to be going on. And on the other hand, Song treats effects of “behavioural markedness” (or inflectional differentiation) together with coding asymmetries, i.e. phenomena such as the collapsing of gender distinctions in the (“marked”) plural in German and Russian, or fewer irregularities in the Spanish Imperfect tense (which again is said to be “marked”). These phenomena are much more diverse than the coding asymmetries, and they are not nearly as well-understood. Song says (p. 165) that I regard them as economically motivated, but in my 2006 paper, I say (somewhat vaguely) that “the reason why more frequent categories show greater allomorphy or more irregularities is that they are more easily remembered when the category occurs more often (2006: 48). I don’t regard these phenomena as well-documented or well-understood (though see Schwartz (1980) for some interesting discussion), but I have never related them to economy.


Croft, William. 1990. Typology and universals. Cambridge: Cambridge University Press.

Croft, William. 2003. Typology and universals. 2nd edition. Cambridge: Cambridge University Press.

Greenberg, Joseph H. 1966. Language universals: With special reference to feature hierarchies. The Hague: Mouton.

Haspelmath, Martin. 2006. Against markedness (and what to replace it with). Journal of Linguistics 42(1). 25–70. doi:10.1017/S0022226705003683.

Haspelmath, Martin. 2008. Frequency vs. iconicity in explaining grammatical asymmetries. Cognitive linguistics 19(1). 1–33.

Haspelmath, Martin. 2019. Role-reference associations and the explanation of argument coding splits. To appear.

Haspelmath, Martin, Andreea Calude, Michael Spagnol, Heiko Narrog & Eli̇f Bamyacı. 2014. Coding causal–noncausal verb alternations: A form–frequency correspondence explanation1. Journal of Linguistics 50(3). 587–625. doi:10.1017/S0022226714000255.

Schwartz, Linda J. 1980. Syntactic markedness and frequency of occurrence. In John Perry (ed.), Evidence and argumentation in linguistics. Berlin: De Gruyter. doi:10.1515/9783110848854-018.

Song, Jae Jung. 2001. Linguistic typology: Morphology and syntax. Harlow & London: Pearson Education.

Song, Jae Jung. 2018. Linguistic typology. Oxford: Oxford University Press.

OpenEdition suggests that you cite this post as follows:
Martin Haspelmath (October 10, 2018). How coding efficiency explains cross-linguistic asymmetries: A reply to Song (2018: Ch. 7). Diversity Linguistics Comment. Retrieved July 19, 2024 from

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.