APPENDIX E

The Turkish Mirror

In Chapter 7 I mentioned the mirror-image effects in the order of words between English and Turkish, which is well illustrated by the Turkish sentence-word şehirlileştirernediklerimizdensiniz, literally ‘you are one of those whom we can’t cause to become someone from town’. From an English perspective, Turkish arranges the elements almost exactly back to front (the only element that is slightly out of synch is the ‘we’):

From the Turkish point of view, however, it is of course the English that is completely the wrong way round. If this reversal in word order were peculiar to English and Turkish, one could perhaps just dismiss it as the result of some bizarre coincidence, which made the ordering principles in one of the languages go completely haywire. But it turns out that neither English nor Turkish are at all unusual in their ways. They are in fact representatives of two vast opposing camps, into which the world’s languages seem to be divided roughly half and half Arabic, Thai and Mixtec (an American-Indian language spoken in Mexico), just to take a handful of examples, arrange their elements more or less like English, whereas Japanese, Greenlandic Inuktitut (spoken by the 40,000 Inuit in Greenland) and Kannada (a Dravidian language spoken by 40 million people in southern India) arrange the elements roughly as in Turkish. This means that English tongues and minds can easily get into a terrible twist when trying to learn Turkish, but the Japanese, who have a reputation for struggling with English, often astound teachers of Turkish by the ease and speed with which they learn that language, even though there is no family relation between Turkish and Japanese whatsoever – Turkish and Japanese are as far removed from each other as either of them is from English.

But why is it that languages from all over the world, and even without any family affiliation, seem to converge into two opposing word-order camps, and gravitate towards two diametrical poles? I hinted at the beginning of the book that the mirror-image effect in the ordering of elements is largely a consequence of just one basic choice that languages make at some stage during their history, between two equally natural alternatives. And in Chapter 7 we started to get a glimpse of what it is that stands behind this mirror, when the basic choice in the positioning of one particular couple, the verb and the object, was shown to determine whether prepositions or postpositions will emerge in a language. In a language with the order verb-object (or VO for short), as in ‘take stone, cut meat’, the verb ‘take’ can develop into a preposition, standing before its noun: ‘with stone cut meat’. But in a language with the order object-verb (OV) ‘stone take, meat cut’, the same verb will turn into a postposition instead: ‘stone with meat cut’. In short, when prepositions or postpositions develop from verbs, they inherit their alignment with respect to their noun from the alignment of the verb with respect to its object.

But this is only the beginning. In the 1960s, the linguist Joseph Greenberg made the rather startling discovery that the basic choice in the alignment of the verb and the object correlates across languages not just with the appearance of prepositions (as in English) or postpositions (as in Turkish), but with the order of a whole series of other elements. It seems that the choice between VO and OV can ripple throughout the structure of language, and have far-reaching repercussions on the order of many other linguistic pairs. But why?

For some grammatical elements, such as auxiliaries like ‘will’ or ‘must’, the reason for the correlation is not too hard to fathom, as it has to do with direct inheritance, just as in the case of prepositions and postpositions. We have seen that auxiliaries originate from normal verbs: ‘will’, for instance, started out in life as an independent verb meaning ‘want’. Now, in a language like English, where the basic order is VO (as in ‘want coffee’), it is only natural to say also ‘want (to) drink’. But in a language which has the OV word order (‘coffee want’), it would also be natural to say ‘drink want (to)’. So when a verb like ‘will’ loses its original sense (‘want to’) and becomes a future auxiliary, in a VO language like English the auxiliary would naturally end up before the verb (‘will drink’), but in a language like Turkish with OV word order, the auxiliary would end up after the verb: ‘drink will’. Indeed, in the Turkish monstrosity above, the auxiliaries leş ← tir ← eme ‘become ← cause to ← can’t’ appear in precisely the opposite order from English ‘can’t→cause to→become’. So it is not only prepositions and postpositions, but also auxiliaries (and a range of other elements) which can inherit their alignment directly from the position of the verb in the couple verb-object.

But even that is not the end of the story, since the basic choice between OV and VO also seems to correlate across languages with the alignment of other pairs in the sentence, such as the head-noun and the appendage noun in a possessive construction (‘sonof the ruler’), which couldn’t possibly have inherited their order directly from the alignment of the verb and object. In theory, there are two ways in which the head noun (‘son’) and the appendage (‘ruler’) can appear: one option is to have the head noun first, as in the English construction ‘sonof the ruler’, and the other option is to have the appendage before the head, as in the English construction ‘ruler’sson’. Greenberg discovered that languages which have VO order (‘take stone’) strongly prefer the order HEAD-APPENDAGE (sonof the ruler), whereas languages which have OV order (‘stone take’) tend to have a strong preference for the order APPENDAGE-HEAD (ruler’sson).

Of course, English itself immediately shows that even if this correlation is very strong, it is not without exceptions, since English is a VO language, but it has two constructions: sonof the ruler but also ruler’sson, which strictly speaking should belong to the other camp. Now, the complex historical reasons for how English ended up with two constructions are much too muddy to get into here. But recklessly bulldozing over the details, one can say that the ruler’sson construction is a very old relic, and developed at a prehistoric period when the language was actually OV and not VO. (Remember, for instance, that the OV order was still used in the English of Ælfric’s day, around AD 1000, in ic hi worhte ‘I them made’ rather than ‘I made them’.) During the historical period, however, English changed its word order from OV to VO, and the ‘sonof the ruler’ construction is much younger, and dates from a time after the language had already changed. So in some sense, even if English seems at first to flout the rule that VO should go (only with) with HEAD-APPENDAGE, it appears that even English is not such as flagrant exception to the correlation after all, because its two constructions are a testimony to a rather messy process of changing over from OV to VO.

But why should there be such a correlation in the first place, between VO and HEAD-APPENDAGE, and vice versa? Why should the alignment of the verb and object have any bearing on the choice between head noun and the appendage noun, when (as opposed to prepositions and auxiliaries) the possessive construction could not have developed directly from the verb-object couple?

It turns out that even though some pairs do not inherit their position directly from the verb and object, they nevertheless tend to be placed in a compliant alignment in the dance of the sentence, in order to prevent them from tripping up on the feet of either the leading couple, verb and object, or one of its toeing-the-line acolytes. In particular, the two nouns in the possessive construction arrange themselves so as to keep in step with prepositions or postpositions, and thus avoid constructions that are difficult to process. We have seen that possessive markers like ‘of’ can originate from prepositions (or postpositions) meaning ‘from’. And to get an idea of what difficulties can ensue from an ‘inconsistent’ order, let’s consider first what a possessive construction would look like when the possessive marker is a preposition (say ‘of’) stuck before the appendage noun ‘ruler’. The HEAD-APPENDAGE and APPENDAGE-HEAD constructions would look like this:

Notice that in the first order (HEAD-APPENDAGE), the possessive marker ‘of’ is nestled safely in between the two nouns, whereas the second order (APPENDAGE-HEAD) has the possessive marker rather perilously dangling at the edge. Are both constructions as good as each other? It seems that they are not, and that the second order is more problematic, because it can lead to cumbersome constructions that are difficult to process. To see why, consider what happens when we add another noun to the construction, an appendage to the appendage ‘ruler’, so that instead of just ‘son of the ruler’ we want to say ‘son of the ruler of Ruritania’. The two orders will yield:

The first order causes no problems, since the marker of possession comes between the head noun and the appendage, and so one can easily add more and more appendages without much difficulty for comprehension. But in the second order, things quickly get rather convoluted. According to the rules, the appendage ‘Ruritania’ should stand before its head noun ‘ruler’, to produce the phrase [of-Ruritania ruler]. But this whole phrase is the appendage of the head noun ‘son’, so one gets the tortuous ‘of-[of-Ruritania ruler] son’. It’s no wonder, then, that in a language where the possessive marker is a preposition, there is a strong motivation for speakers to prefer the first order, HEAD-APPENDAGE, and thus make sure that the possessive marker is nestled safely between the two nouns.

But now let’s look at what happens in a language where the possessive marker is a postposition, say -s, just for the sake of argument. What would the two orders look like in this case?

Here, we get exactly the mirror-image, since it is the first order (HEAD-APPENDAGE) which puts the -s at the edge, whereas the second order (APPENDAGE-HEAD) places the -s safely in between the two nouns. And so this time, it is the second order which can easily be extended, while the first quickly gets gummed up:

So it’s hardly surprising that speakers in languages where the possessive marker is a postposition tend to opt for the APPENDAGE-HEAD order (ruler’sson), in order to avoid such pile-ups. This is not to say, of course, that there are never any exceptions – there always are, to everything. Some languages do actually use the problematic order, and Sumerian is a good example. While there is no need to get into the historical nitty-gritty of why and how Sumerian ended up with such an awkward state of affairs, Sumerian provides a perfect illustration for why an inconsistent order can be clumsy and problematic. The possessive marker in Sumerian is a postposition -ak, stuck on the appendage noun:

As long as there is just one appendage, this construction hardly causes any problems. Even with two appendages, the Sumerians still managed somehow, and stuck two -ak’s at the end:

But when it came to a chain of three appendages, even the Sumerians themselves couldn’t quite cope. They seem to have got into a twist and lost count, as they never remembered to stick on the third -ak:

Sumerian, then, is an exception which very much proves the rule: when the possessive marker is a postposition, it is natural for speakers to prefer the order ruler’sson (otherwise they would get cumbersome constructions as in Sumerian), and when the possessive marker is a preposition, it is natural to prefer the order ‘sonof ruler’. But since prepositions, as we have seen, tend to appear in VO languages while postpositions develop in OV languages, we now already have a third pair that tends to correlate with the basic choice between VO and OV. VO (as in English) correlates with prepositions, auxiliary-verb order, and HEAD-APPENDAGE order, whereas OV (as in Turkish) correlates with postpositions, verb-auxiliary and APPENDAGE-HEAD.

And in a domino effect, the alignment of various other elements can take the lead from the pairs that have already been mentioned. Additional elements inherit their alignment directly from the verb-object couple, or from descendants of that couple, or from pairs which have aligned themselves according to that couple in order not to trip up in the dance, and so on and so forth. Just as one final example, take the position of relative clauses. The last section of Chapter 7 showed that relative clauses are modelled on, and can develop historically from, the possessive construction, when the appendage noun in a possessive construction is gradually extended into a whole clause. The relative clause thus inherits its alignment from the position of the appendage noun in the possessive construction, so the order HEAD-APPENDAGE would naturally result in relative clauses following the head noun (‘thosewhom…’) as in English, but the order APPENDAGE-HEAD would result in relative clauses preceding the noun (‘… whomthose’) as in Turkish.

And on it goes, with various other elements joining in the fray, ultimately accumulating to create the startling mirror-image effect between English and Turkish. The reason why, from an English perspective, Turkish sentences have to be processed exactly ‘back to front’ is that Turkish consistently chooses to align its pairs the opposite way round from what an English speaker finds natural. Behind this effect is one basic choice which English and Turkish made about the ordering of the verb and object. Turkish chose OV (‘stone take’), whereas English chose VO (‘take stone’). In making the choice English gravitated towards one pole, whereas Turkish moved towards the other, with more and more pairs aligning themselves with the verb-object couple or with one of its already lined-up acolytes. And this is why to English ears the Turks seem to ‘talk backwards’, whereas for the Turks or Japanese it is of course the English language that is consistently the wrong way round.