Through the Language Glass: Why the World Looks Different in Other Languages

by Guy Deutscher

  More generally, when communicating with intimates about things that are close at hand, you can be more concise. The more common ground you share with your hearer, the more often you will be able merely to “point” with your words at the participants and at the place and time of events. And the more frequently such pointing expressions are used, the more likely they are to fuse and turn to endings and other morphological elements. So in societies of intimates, it is likely that more “pointing” information will end up being marked within the word. On the other hand, in larger societies, where a lot of communication takes place between strangers, more information needs to be elaborated explicitly rather than just pointed at. For instance, a relative clause like “the house [where they used to meet . . . ]” would have to replace a mere “there.” And if compact pointing expressions are used less frequently, they are less likely to fuse and end up as part of the word.

  Another factor that may explain the differences in morphological complexity between small and large societies is the degree of exposure to different languages or even to different varieties of the same language. In a small society of intimates everyone speaks the language in a very similar way, but in a large society we are exposed to a plethora of different Englishes. Among the throng of strangers you heard over the last week, many spoke a completely different type of English from yours—a different regional dialect, an English of a different social background, or an English flavored with a foreign accent. Contact with different varieties is known to encourage simplification in word structure, because adult language learners find endings, prefixes, and other alterations within the word particularly difficult to cope with. So situations that involve widespread adult learning usually result in considerable simplification in the structure of words. The English language after the Norman Conquest is a case in point: until the eleventh century, English had an elaborate word structure similar to that of modern-day German, but much of this complexity was wiped out in the period after 1066, no doubt because of the contact between speakers of the different languages.

  Pressures for simplification can also arise from contact between different varieties of the same language, since even minor differences in the makeup of words can cause problems for comprehension. In large societies, therefore, where there is frequent communication between people of different dialects and speech varieties, the pressures toward simplification of morphology are likely to be higher, whereas in small and homogeneous societies, where there is little contact with speakers of other varieties, the pressures to simplify are likely to be lower.

  Finally, one factor that may slow down the creation of new morphology is that ultimate hallmark of a complex society—literacy. In fluent speech, there are no real spaces between words, so when two words frequently appear together they can easily fuse into one. In the written language, however, the word takes on a visible independent existence, reinforcing speakers’ perception of the border between words. This doesn’t mean that new fusions ain’t never gonna happen in literate societies. But the rate at which new fusions occur may be substantially reduced. In short, writing may be a counterforce that retards the emergence of more complex word structures.

  No one knows whether the three factors above are the whole truth about the inverse correlation between the complexity of society and of morphology. But at least there are plausible explanations that make the relation between the structure of words and the structure of a society less than a complete mystery. Unfortunately, the same cannot be said of another statistical correlation, which has recently been demonstrated in a different area of language.

  SOUND SYSTEM

  Languages vary considerably in the size of their sound inventories. Rotokas from Papua New Guinea has only six distinct consonants (p, t, k, b, d, g), Hawaiian has eight, but the !Xóõ language from Botswana has forty-seven non-click consonants and seventy-eight different clicks that appear at the beginning of words. The number of vowels also varies considerably: many Australian languages have just three (u, a, i), Rotokas and Hawaiian have five each (a, e, i, o, u), whereas English has around twelve or thirteen vowels (depending on variety) and eight diphthongs. The overall number of sounds in Rotokas is thus only eleven (six consonants and five vowels), whereas in !Xóõ it amounts to more than 140.

  In 2007, the linguists Jennifer Hay and Laurie Bauer published the results of a statistical analysis of the sound inventories of over two hundred languages. They discovered that there is a significant correlation between the number of speakers and the size of the sound inventory: the smaller the society, the fewer distinct vowels and consonants the language tends to have; the larger the number of speakers, the larger the number of sounds. Of course, this is only a statistical correlation: it does not mean that every single language of small societies must have a small inventory of sounds and vice versa. Malay, spoken by more than seventeen million people, has only six vowels and sixteen consonants, so twenty-two sounds in total. Faroese, on the other hand, has fewer than fifty thousand speakers but sports around fifty sounds (thirty-nine consonants and more than ten vowels), more than twice the number in Malay.

  Still, as far as statistical correlations go, this one seems pretty robust, so the only plausible conclusion is that there is something about the modes of communication in small societies that favors smaller sound inventories, whereas something about large societies tends to make new phonemes more likely to emerge. The problem is that no one has yet come up with any compelling explanation for why this should be so. One factor that could be relevant, perhaps, is contact with other languages or dialects. As opposed to word structure, which tends to be simplified as a result of contact, a language’s sound inventory not uncommonly increases due to contact with other languages. For instance, when sufficiently many words with a “foreign” sound are borrowed, the sound can eventually be integrated into the native system. If such contact-induced changes are less likely in smaller and more isolated societies, that fact might go some way toward explaining their smaller sound inventories. But this clearly cannot be the whole story.

  SUBORDINATION

  Finally, there is one area of language whose relation to the complexity of society may after all correspond to the considered opinion of the man in the street: this is the complexity of sentences and, in particular, the reliance on subordinate clauses. Subordination is a syntactic process that is often touted (by syntacticians, at least) as the jewel in the crown of language, and the best example for the ingenuity of its design: the ability to subsume a whole clause within another. With subordination, we can produce expressions of increasing complexity that nevertheless remain coherent and comprehensible:

  I must have told you about that seal

  I must have told you about that seal[which was eyeing a fish]

  I must have told you about that seal[which was eyeing a fish[that kept jumping in and out of the icy water]]

  And there is no need to stop there, because in theory the mechanisms of subordination allow the sentence to go on and on for as long as there is breath to spare:

  I must have told you about that quarrelsome seal [which was eyeing a disenchanted but rather attractive fish [that kept jumping in and out of the icy water [without paying the least attention to the heated debate [being conducted by a phlegmatic walrus and two young oysters [who had recently been tipped off by a whale with connections in high places [that the government was about to introduce speed limits on swimming in the reef area [due to the overcrowding [caused by the recent influx of new tuna immigrants from the Indian Ocean [where temperatures rose so much last year [that . . . ]]]]]]]]]]

  Subordination makes it possible to convey elaborate information in a compact way, by weaving different assertions on multiple levels into one intricate whole while keeping each of these levels under control. The paragraph above, for instance, has just one simple sentence at its primary level: “I must have already told you about that seal.” But from there downward, more and more information is interlaced using different types
of subordinate clause.

  There are no reliable reports about any language that lacks subordination altogether.* But although all known languages use some subordination, languages vary greatly in the range of subordinate clauses they have at their disposal and in the extent to which they rely on them.

  For instance, if you have nothing better to do with your time than pore over ancient texts, you will soon notice that the narrative style of ancient languages such as Hittite, Akkadian, or biblical Hebrew often seems soporifically repetitive. The reason is that the mechanisms of subordination were less developed in these languages, so the coherence of their narrative relied to a much greater extent on a simple type of “and . . . and . . .” concatenation, in which the clauses merely followed the temporal order of events. Here, for instance, is a short Hittite text, a report by King Murshili II, who reigned in the fourteenth century BC from his imperial capital of Hattusha, in what is today central Turkey. Murshili is describing in dramatic tones how he came to be afflicted by a severe illness that impaired his ability to speak (a stroke?). But to modern ears the vivid substance of the report contrasts starkly with the monotonous staccato of the style:

  This is what Murshili, the Great King, said:

  Kunnuwa nannaun

  I drove (in a chariot) to Kunnu

  nu aršiarši udaš

  and a thunderstorm came

  namma Tarunnaš atuga tetiškit

  then the Storm-God kept thundering terribly

  nu nun

  and I feared

  nu-mu-kan memiaš išši anda tepawešta

  and the speech in my mouth became small

  nu-mu-kan memiaš tepu kuitki šar iyattat

  and the speech came up a little bit

  nu-kan aši memian arapat paškuwnun

  and I forgot this matter completely

  maan-ma ur wittuš appanda pir

  but afterwards the years came and went

  nu-mu wit aši memiaš tešaniškiuwn tiyat

  and this matter came to appear repeatedly in my dreams

  nu-mu-kan zazia anda keššar šiunaš araš

  and God’s hand seized me in my dreams

  aišš-a-mu-kan tapuša pait

  then my mouth went sideways

  nu . . .

  and . . .

  Today, we would tend to use various subordinate clauses and thus would not need to follow the order of events so punctiliously. For example, we might say: “There was once a terrible thunderstorm when I was driving to Kunnu. I was so terrified of the Storm-God’s thundering that I lost my speech, and my voice came up only a little. For a while, I forgot about the matter completely, but as the years went by, this episode began to appear in my dreams, and while dreaming, I was struck by God’s hand and my mouth would go sideways.”

  Here is another example, this time from Akkadian, the language of the Babylonians and Assyrians of ancient Mesopotamia. This document, written sometime before 2000 BC, reports the result of a legal proceeding. We are told that a certain Ubarum proved before the inspectors that he had told a Mr. Iribum to take the field of Kuli, and that he (Ubarum) didn’t know that Iribum, on his own initiative, had instead taken the field of someone else, Bazi. But while this is the gist of what the document says, the Akkadian text doesn’t put it quite like that. What it actually says is:

  ana Iribum Ubarum eqel Kuli šlu’am iqbi

  Ubarum told Iribum to take Kuli’s field

  š libbiššuma

  he (Iribum) on his own initiative

  eqel Bazi uštli

  took the field of Bazi

  Ubarum ula de

  Ubarum didn’t know

  mahar laputtî uknšu

  he proved (this against) him in front of the inspectors

  The difference between the Akkadian formulation and the way we would naturally describe the situation in English lies mainly in our pervasive use of constructions such as “he didn’t know that [ . . . ]” or “he proved that [ . . . ].” This particular type of subordinate clause is called “finite complement,” but although the name is rather a mouthful, the construction itself is the bread and butter of English prose. In both written and spoken registers, we can take practically any sentence (let’s say “Iribum took the field”) and, without altering anything in the sentence itself, make it a subordinate part of another sentence:

  He didn’t know that [Iribum took the field]

  And since it is so easy to set up this hierarchical relation once, we can do it again:

  Ubarum proved that [he didn’t know that [Iribum took the field]]

  And again:

  The tablet explained that [Ubarum proved that [he didn’t know that [Iribum took the field]]]

  And again:

  The epigrapher discovered that [the tablet explained that [Ubarum proved that [he didn’t know that [Iribum took the field]]]]

  The Akkadian report does not use such finite complements. In fact, most of its clauses are not hierarchically ordered but simply juxtaposed according to the temporal order of the events. This is not a coincidence of just one text. While we may take finite complements for granted today, this construction was missing in the oldest attested stages of Akkadian (and of Hittite). And there are living languages that do not have this construction even today.

  Not that linguistic textbooks will divulge this information, mind you. In fact, some will ardently profess the opposite. Take that flagship of linguistic education, the Introduction to Language by Fromkin and Rodman that I mentioned earlier, and its twelve articles of faith that constitute “what we know about language.” The second affirmation, as you will recall, is that all languages are equally complex. A little further below, affirmation eleven asserts:

  Syntactic universals reveal that every language has a way of forming sentences such as:

  Linguistics is an interesting subject.

  I know that linguistics is an interesting subject.

  You know that I know that linguistics is an interesting subject.

  Cecelia knows that you know that I know that linguistics is an interesting subject.

  Is it a fact that Cecelia knows that you know that I know that linguistics is an interesting subject?

  Unfortunately, the textbook does not disclose the precise identity of the “syntactic universals” that have revealed that every language has such constructions. Nor does it specify when and where this revelation was handed down to mankind. But is the claim actually true? I have never had the privilege of communing with a syntactic universal myself, but the evidence from more mundane sources, namely descriptions of actual languages, leaves no doubt that some languages do not have a way of forming such sentences (and not just because they don’t have a word for “linguistics”). Many Australian aboriginal languages, for example, lack a construction equivalent to the finite complements of English, and so do some Indian languages of South America, including one, Matses, that we will meet in the next chapter. In such languages, one simply cannot form sentences such as:

  It is a fact that many students don’t realize that their linguistics textbooks don’t know that some languages do not have finite complements.

  Instead, this state of affairs would have to be expressed by other means. For example, in the early stages of Akkadian, one would do it along these lines:

  Some languages do not have finite complements. Some linguistics textbooks don’t know that. Many students don’t realize their textbooks’ ignorance. This is a fact.

  While systematic statistical surveys on subordination have not yet been conducted, impressionistically it seems that languages that have restricted use of complements (or even lack them altogether) are mostly spoken in simple societies. What is more, ancient languages such as Akkadian and Hittite show that this type of “syntactic technology” developed at a period when the societies in question were growing in complexity. Is this just coincidence?

  I have argued elsewhere that it is not. Finite complements are a more effective tool for conveying elaborate p
ropositions, especially when less information can be left to the context and more explicitness and accuracy are required. Recall the sequence of events described in the Akkadian legal document found here. Of course, it is possible to convey the set of propositions described there just as the Akkadian text organizes it, with a simple juxtaposition of clauses: X told Y to do something; Y did something different; X didn’t know that; X proved it in front of the inspectors. But when the dependence between the clauses is not explicitly marked, some ambiguity remains. What exactly did X prove? Did he prove that Y did something different from what he was told? Or did X prove that he didn’t know that Y did something different? The juxtaposition does not make that clear, but the hierarchical structure of finite complements can easily do so.