More on the evolution of word order and rare word orders

Oct 21, 2011 by

In the previous posting, I discussed one set of problems with Gell-Mann & Ruhlen’s theory on the origin and evolution of word order (also discussed in an earlier posting); these problems concern the possible misclassifications of various languages and the issue of free word order. As pointed out at the end of that posting,

“the order of Subject, Object and Verb may not be a primitive, as it is considered by G-M&R, but rather an artifact of several factors, perhaps ones that order the verb with respect to the subject and the object separately, as well as ordering the subject and the object with respect to each other, or ordering the verb with respect to other clausal elements (e.g. negation, adverbs, auxiliaries, tense, etc.).”

There are several good reasons to believe that the parametric approach, that is analyzing sentential word order as an artifact of several factors rather than as a single primitive, is a better approach.

First, as noted in Mark C. Baker’s Atoms of Language (on which the discussion below relies heavily), understanding sentential word order as a result of several distinct parameters provides a way of handling statistical generalizations about word orders in languages of the world. Thus, it has long been noticed that similar figures on the cross-linguistic frequency of the six major word orders are achieved regardless of sampling procedures. For example, a number of studies have reached the following generalization: SOV languages are only slightly more common than SVO languages, but both are much more common than VSO languages. For instance, according to the World Atlas of Linguistic Structures, 47.5% of languages are SOV, 41% are SVO and 8% are VSO. According to G-M&R’s figures (see their Table 2), 50% of languages are SOV, 38% are SVO and the same 8% are VSO. Note, however, that the higher percentage of SOV and lower percentage of SVO in G-M&R’s data may have to do with the classification issues discussed in the previous posting.

While G-M&R’s theory can potentially explain the predominance of SOV order (if we assume that languages tend to preserve features inherited from their parent language rather than to innovate), it is harder to see how the proportion of VSO languages, hovering around 8-12% depending on the study, would follow. Given that VSO languages, according to G-M&R, can arise only from SVO languages and not directly from SOV languages (as indicated by their Figure 1 reproduced below), one would expect fewer VSO languages than SVO, but why exactly 8-12%? Also, given G-M&R’s claims, one would expect that there would be fewer SVO and VSO languages together than SOV languages; their figures are in agreement with this expectation, but WALS figures are not, as there are 583 SVO and VSO languages and only 565 SOV languages.

So how can a parametric theory, which relies on multiple factors (called “parameters”), handle the proportion of VSO languages?
According to Mark Baker, three distinct parameters must be used to account for the VSO order:

  • Head Directionality Parameter, which determines whether heads appear to the left or the right of their complements (for our present purposes, we are interested in the ordering of the verb with respect to its complement, the direct object): in an OV language the head/verb appears to the right of its complement/object, while in a VO language the head/verb appears to the left of its complement/object
  • Subject Placement Parameter, which determines the placement of the subject with respect to the verb/auxiliary: in VSO languages like Welsh or Zapotec the subject appears to the right of the verb, while in SVO languages like French or Spanish the subject appears to the left of the verb
  • Verb Attraction (or Verb Raising) Parameter, which determines the position of the verb with respect to certain adverbs and negation: with respect to this parameter French, Spanish, Welsh and Zapotec pattern the same, with the verb preceding the relevant adverbs and negation, while in English, Indonesian, Edo, Khmer the verb follows those elements (compare e.g. the French order Jean mange souvent le chocolat literally ‘John eats often the chocolate’ with the English John often eats chocolate)

(There is also a certain dependency relationship between these parameters, so that the application of these parameters depends on the settings of other parameters in a hierarchical fashion. I will not discuss this matter here; see Baker’s Atoms of Language for a detailed discussion.)

To get a VSO language these three parameters must all be set a certain way: heads must appear to the left of their complements; subjects must appear to the right of the verb; and the verb must appears to the left of the relevant adverbs and negation. If we assume that each (binary) parameter has a 50% chance of being set one way or another, the chance that all three parameters will be set just a certain way is 1 in 8. Therefore, we expect about 1 in 8 languages to be VSO. Indeed, this prediction is not far from the observed frequencies.

By the way, the assumption that each binary parameter is set at random, with a 50% chance of being set this way or that, is supported by the facts concerning the Head Directionality Parameter. Recall that it determines (among other things) whether a given language has OV or VO order (it also happens to be the highest word-order-related parameter in Baker’s hierarchy, meaning that its application is not dependent on the setting of any other word-order-related parameter). According to WALS figures, among languages with a dominant OV or VO order, 713 languages are OV and 705 are VO — this is about as close to a 50-50 split as we can hope to get!

And what of the even rarer object-initial orders (OVS and OSV)? Note that G-M&R have no account of why these orders are so rare (according to WALS figures, only 0.9% of languages are OVS and 0.3% are OSV). Since, according to G-M&R’s theory of how word orders evolve, object-initial orders derive directly from SOV, as does SVO, everything else being equal, both OVS and OSV should be more frequent than SVO (from which some languages develop VSO and VOS; see the diagram above). Clearly, not everything else is being equal! But G-M&R propose no explicit account of the relative rarity of object-initial orders (I would think they would rely on functional explanations, but that’s just a guess).

In contrast, Baker claims that object-initial orders are rare because they are not a result of setting certain parameters a certain way, but are derived via displacement (movement) of certain constituents from their underlying positions. For example, he claims that the OVS order in Hixkaryana is derived from a more basic order SOV order by fronting the verb phrase (i.e. O+V). Thus, a Hixkaryana sentence Otweto yımyakonı rohetxe totokomo wya meaning ‘My wife gave a hammock to (some) people’ (but literally ‘hammock gave my-wife people-to’, with the order Object-Verb-Subject-IndirectObject) is derived from the more basic order Subject-IndirectObject-[Object-Verb] by fronting the bracketed part:

Underlying word order: my-wife to-people [hammock gave]
Derived order: [hammock gave] my-wife to-people

This view is supported by the fact that cross-linguistically languages tend to place both direct and indirect object on the same side of the verb (according to WALS figures, only 14% of languages in their sample violate this generalization, and these may well be languages with derived orders, like Hixkaryana, or languages in the middle of language change, as has been suggested for Mandarin by John Cowan in a comment to an earlier posting).

Similarly, OSV orders must be derived from some other underlying order, according to Baker, since they are in violation of a putative universal stating that the verb and the object form a constituent and hence (all else being equal) must appear adjacent to each other.

To be continued…


Related Posts

Subscribe For Updates

We would love to have you back on Languages Of The World in the future. If you would like to receive updates of our newest posts, feel free to do so using any of your favorite methods below: