1

A Castle in the Air

C’est un langage estrange que le Basque …

On dit qu’ils s’entendent, je n’en croy rien.

Basque is really a strange language …

It is said that they understand one another,

but I don’t believe any of it.

Joseph Justus Scaliger (1540–1609)

 

Everyone knows that the words of a language, from its aardvarks to its zucchini, lend meaning to our utterances, and allow us to understand one another. And it is because foreign languages use so many strange words that we cannot understand them without years of labour. Even Joseph Scaliger, the most erudite scholar of his day, a polyglot not only fluent in Latin, Greek and most of the modern languages of Europe, but also self-taught in Hebrew, Arabic, Aramaic and Persian, still had to give up on Basque, because it used completely different words for absolutely everything. The effort of memorizing many thousands of words so overwhelms our perception of what language learning is all about that it may easily lead to the impression that knowing a language just comes down to knowing its words. Surely, if one could only recognize the meaning of each word, all one would need to do is add all these meanings up somehow, in order to grasp the sense of a whole sentence. But if this is so, and language ultimately amounts to just words, then isn’t the quest for the origin of structure merely an intellectual wild goose chase?

On reflection, however, it soon becomes clear that language is much more than the sum of its words. In fact, a language with only words, and no structure to prop them up, would be a poor instrument of communication. Words may be the bricks in the language edifice, but when we want to convey subtle thoughts, involving intricate relations between different concepts, we need to combine words into proper sentences. The structure of language is what can turn a pile of word-bricks into a palace of expressions – a castle in the air.

As a simple illustration, consider the following example:

Head vizier Sultan troops the of to the his the brought

If the meaning of a sentence is nothing more than the sum of its words, then why doesn’t this one amount to any substance at all, even though the meaning of each word is perfectly familiar? The reason is that there is an essential feature missing from this sentence, and exactly what that is becomes clear as soon as one takes the very same words and arranges them in a different order. Suddenly, they leap into sense:

The Sultan brought his vizier to the head of the troops.

In this arrangement, the words convey a detailed event involving various participants, and now describe not only who these participants are, but also exactly who is doing what to whom. And to remove any lingering suspicion that the choice of words by itself dictates the meaning of a sentence, consider what happens when the same words are once again juggled into a different order:

The troops brought to the Sultan the head of his vizier.

There are many well-turned aphorisms which play on exactly such word permutations: ‘better to lose a moment in life than to lose life in a moment’; Mae West’s ‘a hard man is good to find’; or the definition of ‘foreign aid’ as the transfer of money ‘from poor people in rich countries to rich people in poor countries’. One of my favourites is Kermit the Frog’s rearrangement of a well-worn cliché: ‘time’s fun when you’re having flies’ (although note that he allows himself some poetic licence by sneaking in an additional ’s). But most famous, perhaps, is Alice’s conversation at the Mad Hatter’s Tea Party:

‘Then you should say what you mean,’ the March Hare went on. ‘I do,’ Alice hastily replied; ‘at least – at least I mean what I say – that’s the same thing, you know.’ ‘Not the same thing a bit!’ said the Hatter, ‘Why, you might just as well say that “I see what I eat” is the same thing as “I eat what I see”!’ ‘You might just as well say,’ added the March Hare, ‘that “I like what I get” is the same thing as “I get what I like”!’ ‘You might just as well say,’ added the Dormouse, who seemed to be talking in his sleep, ‘that “I breathe when I sleep” is the same thing as “I sleep when I breathe”!’

Clearly, then, the sense of a sentence depends not only on the meaning of each word but also on the particular arrangement in which these words are joined. The choice of meaning matters, but just as much the order of the combination. (Or you might as well say that ‘the choice of order matters just as much to the meaning of the combination’.)

Now a natural reaction to all this might run along the following lines: of course it matters in which arrangement words are combined, but don’t we simply put the words in the natural order? Doesn’t the order of words in the sentence simply follow the natural order of ideas? To see why things are not so simple, consider another variation on the Sultan theme, in the example below:

Sultan vizier his troops his of head their to brought.

Here we go again, I can hear you thinking. No doubt this is just another meaningless list, a jumble of words waiting to be juggled into some sensible order to create yet another meaning. But in fact – the sentence above already makes perfect sense. Or, to be more accurate, it would make perfect sense if you happened to be born in the Sultan’s own city and spoke Turkish. For the string above is simply a word-for-word gloss of an entirely respectable Turkish sentence:

Let’s not bother for now with the fact that one Turkish word (like ordularinin) can express what English conveys with various independent words, and just concentrate for a moment on the order in which the Turkish elements are arranged. Clearly, ‘natural’ is very much a matter of geography: what is perfectly natural to a Turk does not seem even remotely natural to an English speaker. An even more striking example of the discrepancy between the ordering rules of the two languages is provided by the sesquipedalian Turkish word from the previous chapter: şehirlileştiremediklerimizdensiniz. The gloss below gives an approximate translation of each of this word’s components, in the order in which they appear. Still, the gloss looks almost as much like gobbledegook as the Turkish word itself:

But now try a simple trick, and read the parts in reverse order:

This reverse order almost precisely matches the English translation. We now only have to make one slight alteration (move the ‘we’ two places along) to get a perfectly comprehensible English sentence: ‘you are one of those whom we can’t cause to become someone from town’, or more idiomatically, ‘you are one of those whom we can’t turn into a town-dweller’. Incidentally, this almost exact mirror-image in the word order of English and Turkish has nothing to do with the direction of writing – both languages are written from left to right. So which order is ‘natural’? Is it the English or the Turks who spend their lives talking back to front?

It is only to be expected that the habits of one’s own language should seem utterly natural, while those of other peoples much less so – ‘natural’, after all, is what one is used to. But when one overcomes the biases of familiarity, it becomes clear that neither the English nor the Turkish order is any more natural than the other. Both orders are just cultural conventions, and conventions, by their very nature, can vary across time and space.

All this is not to say, of course, that the conventions of word order are completely arbitrary in every respect, and that different languages can order their words entirely at whim. In fact, in Chapter 7 I will argue that there are a few simple principles for ordering words, which are indeed ‘natural’, and which are common to all languages. Nevertheless, as will become clear later on, these natural principles still leave considerable scope for various choices, and there is one particular choice that languages make at some stage in their development, which can then ripple throughout their structure, and ultimately result in the mirror-image effect between English and Turkish. For the moment, however, the important thing is to note that a sentence makes sense not because its words are simply placed in some ‘natural’ and universally valid ‘order of ideas’. Rather, the feat of transforming a pile of words into a complex coherent whole is achieved through the mediation of an elaborate system of structural conventions, which can vary greatly from language to language.

The conventions of word order are probably the oldest element in the structure of language. Chapter 7 will suggest that back in the ‘me Tarzan’ stage, speakers who were trying to string words together had nothing but some simple ordering principles to go on. And even in today’s languages, it is fair to say that arranging the bricks in a particular order is still the most important element in the art of sentence construction. Nevertheless, word order is no longer the only means speakers can rely on when combining words, since languages have also developed a range of other techniques to help make the bricks stick, such as the use of various adhesives which facilitate the construction of much more complex edifices. The following pages will survey some of these other features in the structure of language – from the mortar, bolts and nails to the grand principles of design. By the end of the chapter, I hope it will be clear that searching for the origin of linguistic structure is nothing less than an attempt to discover how we acquired the ability to build bridges between minds.

A WASTE OF BREATH?

Most words we use, like ‘table’, ‘kick’, ‘walk’ or ‘rabbit’, have a simple solid meaning, so they are often called ‘content words’. Of course, philosophers may dispute that anything to do with meaning is simple, not even the meaning of ‘rabbit’. Over the years they have fought long and bitter wars – with thousands of pages as casualties – over the difference between ‘denotation’ and ‘connotation’, ‘reference’ and ‘sense’, ‘symbol’ and ‘sign’, and have busied themselves with such grave questions as whether, when a speaker of an exotic language points at a rabbit and says gavagai, he means ‘rabbit’ or ‘undetached rabbit parts’. But in the end, surely anyone with a healthy dose of common sense will reach the conclusion that a rabbit is a rabbit is a rabbit, and there’s no need to make a dog’s dinner out of that.

Nevertheless, there is also a group of words in language whose meaning really is quite a lot less obvious. These are ‘grammatical words’ such as a, the, of, so, that, which, or, than. Think about it: what is the meaning of a, for instance? Can you point at a the, or close your eyes and imagine a than? In dictionaries and grammar books, such unassuming words appear under a variety of titles: conjunctions, prepositions, articles and so on. But there is one basic property that is common to them all: they cannot boast their own independent meaning. They don’t refer to objects, actions or properties, or to any other concepts that can be imagined in their own right.

In fact, language is also populated by beings that are even humbler than grammatical words: various splinters which cannot even stand up on their own. Think of prefixes like un-, or suffixes (endings) like -ly, -er, -s. Not only do these fragments have no meaning of their own, they don’t even have an independent existence, and in order to make any appearance at all, they must find other words to latch on to: un-like-ly, long-er, piece-s. In later chapters, the properties of grammatical elements will be probed in more depth, but for the moment, the profusion of such empty vessels raises an obvious question: why should language be crowded with these meaningless hangers-on in the first place? If grammatical words and elements don’t add any independent meaning to the sentence, aren’t they just a waste of breath, mouthfuls of airy superfluity, excess freight?

But it would be rash to start clearing the decks and chucking them all overboard just yet, for far from being redundant, grammatical elements are indispensable for keeping the hull of the sentence together. Consider, for instance, the following sentence:

This agreement is about a principle compromise, not a principle compromise.

Not a terribly good example, you might feel, since this ‘sentence’ doesn’t make any sense. But that is exactly my point – as all it takes is a modest grammatical fragment, the ending -d, and the meaningless lines become profoundly sensible.

So even though -d has no meaning to call its own, and even though it is a mere chip of a syllable, the whole sense of the sentence nevertheless pivots on it. The ending welds the two content words principle and compromise together, and determines the hierarchy between them, by marking which one is the head of the phrase, and which is only an appendage (the word that -d clings on to):

Similar examples are not hard to come by: a group of strikingdancers is quite different from a group of strikersdancing; doing something with disgracefultaste is certainly not the same as doing it with distastefulgrace; and to be uncommonlylucky is a different thing altogether from being unluckilycommon. So even though grammatical words and elements may not have any meaning of their own, they play a crucial role in the administration of the sentence, and help to determine the hierarchy and precise relations between content words.

The valuable role of grammatical words can become evident precisely in those genres that are traditionally parsimonious in their use, for instance, in newspaper headlines that economize on the article ‘the’. The drawback of such a spare style is that it can fail to indicate the correct hierarchy of the sentence, and this can lead to such proverbial headline gaffes as ‘Lawyers Give Poor Free Advice’. What the writer intended to convey here, of course, was the newsworthy information that lawyers were giving free advice to poor people (that is, ‘giving the poor free advice’). So ‘poor’ was conceived as an independent participant in the sentence (the ‘indirect object’, to be precise). But as it happens, there can be quite another reading of the headline, where ‘poor’ is no longer an independent participant in the sentence, but only a hanger-on, another appendage to ‘advice’. This different role transforms the meaning of the sentence into something rather less newsworthy, namely that the free advice given by lawyers is poor:

The hierarchical organization of the sentence can create many similar pitfalls for headline compilers, in such well-known examples as ‘Fund Set Up for Beating Victim’s Family’, or ‘Juvenile Court to Try Shooting Defendant’ (both of which would become unambiguous with the judicious placement of one little ‘the’). Sometimes, playing on the internal structure of a sentence is also a source of deliberate puns, as in the children’s joke-question: ‘How do you get down from an elephant?’ The answer, of course, is that ‘You don’t get down from an elephant. You get down from a duck!’ The trick here is revealed when the role of the word ‘down’ in the question and the answer is represented graphically:

*   *   *

The role of hierarchy in language is considerably more important, however, than merely a source for verbal wit or headline gaffes. In fact, the multi-tier organization of the sentence is one of the most fundamental design-features of language. Even though on the surface a sentence might seem like just a string of words in linear succession, underneath the floorboards there is a lot going on, with words hierarchically organized on different tiers. And even though we may not be consciously aware that we are speaking or listening ‘on different levels’, we are all keenly attuned to the tiered structure of the sentence. It was Noam Chomsky who first stressed this point by illustrating how even the simplest everyday linguistic operations, such as forming questions, are sensitive to the hierarchy of the sentence. Consider a sentence like ‘the seal was eyeing a fish’. There is a simple way of turning this into a question, as any native speaker knows. If the first verb in the sentence is an auxiliary (a support-verb such as was, has, will), then all one has to do is move it to the front of the sentence:

Now suppose we want to turn a slightly longer sentence into a question: ‘the seal that was eyeing a fish has picked a fight with a walrus’. Applying the simple rule above, we find the first auxiliary in the sentence – it happens to be ‘was’ again – and move that to the front:

Why did this operation produce such gibberish? The reason is that the rule outlined above was not entirely accurate. When forming a question, it’s not the first auxiliary in the sentence that needs to be moved, but rather the first auxiliary on the main level of the sentence. And in this case, there is an entire phrase which is not on the main level, but only dangles as an appendage of the noun ‘seal’:

All speakers of English intuitively know that when posing a question, the appendage ‘that was eyeing a fish’ should be passed over in favour of the first auxiliary on the main level, in this case ‘has’, which should be moved to the front:

In linguistic parlance, the appendage ‘that was eyeing a fish’ is called a ‘relative clause’. But no English speaker needs to sign up for a crash course in syntax to know that this phrase lies low and doesn’t get involved when the sentence is rearranged to form a question. When learning the language as children, speakers have intuited that a phrase starting with ‘that’ or ‘which’ is subordinate: it doesn’t participate in the real action on the main level of the sentence, and sticks to the participant it modifies. So the hierarchical structure of the sentence is not just a graphic game of subscripts, but a fundamental feature of language, which we all take into account when producing or processing sentences.

All this raises an obvious question: why design language in this way? Why not have a system where all words work on the same level? The simple answer is complexity. Later on, I will suggest that at the ‘me Tarzan’ stage, words were indeed combined on only one level. And as long as all that was involved were just two- or three-word sentences, this flat set-up was perfectly adequate. But when a system grows in complexity, hierarchical structure becomes a more efficient way of doing things. In a large army, for instance, chaos would prevail if the commander-in-chief had to worry about how to place each individual soldier in battle. Instead, the commander only has to think about how to arrange his divisions, the commander of each division then has to decide how to arrange his individual brigades, the commander of each brigade arranges his individual regiments, and so on. In language, a similar hierarchical principle allows us to undertake complex manoeuvres with little difficulty. Consider the following sentences, arranged in order of increasing complexity:

• The seal has picked a fight.

• The seal that was eyeing a fish has picked a fight with a walrus.

• The quarrelsome seal that was eyeing a disenchanted but rather attractive fish has picked a fight with a phlegmatic walrus.

• The quarrelsome seal that was eyeing a disenchanted but rather attractive fish that was jumping in and out of the icy water has picked a fight with a phlegmatic walrus that was innocently passing by.

The fourth sentence, running to a total of thirty-five words, is pretty complex. And yet, for a speaker of English, this sentence is immediately comprehensible, because its complexity is mitigated by the hierarchy of command. At the highest level, the first and the last sentence have the same simple structure: The seal has picked a fight. All the further intricacies are entirely an internal matter for the phrases headed by ‘seal’ and ‘fight’. And even within each phrase there is an internal hierarchy of command, so that the task remains manageable at all levels:

The hierarchical organization of language is thus quite an ingenious system, which allows us to perform complex tasks with remarkable ease: to produce and understand sentences with many different participants and relations, not to mention hundreds of sounds. The question of how our ancestors might have hit upon such a system will be taken up in Chapter 7.

*   *   *

So far, I have made various claims about the structure of ‘language’, but in practice, the previous examples dealt mostly with one language in particular. Nevertheless, even if the details were drawn primarily from English, the general principles are valid in all languages of the world. All languages are organized hierarchically, all languages rely on some word-order conventions, all languages use grammatical words, and almost all languages use grammatical elements such as suffixes or prefixes. But while the underlying principles are always the same, languages can differ quite radically in the details. There is considerable variation, for instance, in how the burden is divided between these strategies. Languages such as Vietnamese, Yoruba (spoken in Nigeria) and English rely heavily on word order to spell out the roles of the participants in the sentence. In English, for example, the only way of telling who shoots whom in ‘the thief shot the cop’ and ‘the cop shot the thief’ is by the word order: the subject (the one doing the shooting) comes before the object (the one being shot). But in Tamil, Warlpiri (an aboriginal language of Australia) and Russian, word order is much freer.

If not through a strict order of the participants, how can a language manage to signal who is doing what to whom? Hebrew, to take one example, uses a grammatical word for this purpose, and marks the object by putting a preposition, et, before it. As the example below illustrates, the participant immediately after the preposition et is marked as the one being swallowed:

If one wants to swap the roles (and set the story in a sushi-bar), there is no need in Hebrew to change the order of the participants. It is enough simply to move the preposition et to before the whale:

Japanese employs a similar method, only that instead of a preposition, it uses two postpositions, which come after the relevant nouns. Here, the postposition ga marks who is doing the eating, and o marks what is being eaten:

Russian (like Latin) uses another strategy and instead of adding independent grammatical words, tacks endings on to the nouns themselves. In the example below, the ending -a on akula ‘shark’ marks it as the subject, and the ending -u on rybu ‘fish’ marks it as the object:

Because the endings indicate the roles of the participants explicitly, changing the order of the words doesn’t change the basic meaning of the sentence:

Finally, some languages manage to have a flexible word order, but still don’t need to compensate for this by tagging markers on the participants themselves. Their trick is to indicate the role of the participants on the verb instead, as can be seen in the following sentences from a dialect of modern Aramaic spoken in Alqosh, a small town in northern Iraq. The only thing which signposts who is seeing whom in the two sentences below is the shape of the verb: ‘saw-she-him’ means that the girl sees the boy, while ‘saw-he-her’ reverses the roles.

*   *   *

The examples above should have given a flavour of the variety of the strategies languages have come up with to highlight the basic plot of the action. But there is more to life, of course, than who is doing what to whom, and languages have developed various tools for conveying a great deal of information far and above the rudimentary business of imparting the basic roles of the two leading protagonists. English, for instance, may rely on word order to distinguish between the subject and the object, but for marking the function of various supporting roles, as well as circumstantial evidence such as time and place, it uses both prepositions and postpositions:

Languages have also devised a range of different forms of the verb, which can express subtle nuances of the action itself. Consider, for instance, the following variations on the theme ‘the seal ate the fish’. All of the examples below have precisely the same participants, playing precisely the same roles: a seal is doing the eating and a fish is being eaten. But the sentences still vary considerably in their meaning.

So even when the participants all stay the same, and even when the action remains ‘eating’, there is still a variety of finer nuances of the action that speakers can convey. For one thing, we can specify the time of the action (‘will eat’, ‘ate’), and the manner in which it takes place (‘eats’ when it happens regularly, or ‘is eating’ when it happens right as we speak). We can also add our own personal perspective on what is going on, by indicating how much we know about it and what we think about it. If you say may have eaten, you imply that you are not sure whether it happened or not; if you say must have eaten, you imply you are pretty certain; saying should eat means you think it’s a good idea; must not eat means you think it isn’t.

All languages have the means of expressing such nuances, but again, they vary in how they go about doing it. Suppose, for instance, that you were conversing with an English seal about one of its favourite pursuits, the consumption of fish. Should you try to convey the various ins-and-outs of this activity, you would have to use independent grammatical words: may eat, should eat, will eat. And if you wanted to draw out even subtler nuances, you would generally use different combinations of independent grammatical words: should have eaten, were not being eaten. But were you to hold the same conversation with a Roman seal instead, you would have to employ another method, and rely mainly on different endings on the verb itself:

In Latin, the root of the verb (‘ed-’) gives the basic meaning (‘eat’), while the endings supply all the various nuances of the activity. As it happens, the Latin verbal system is much more complex than the forms above would suggest, since the endings on the root also vary according to who is performing the action (that is, according to the ‘person’):

No wonder that the Latin verb seems so off-putting to learners. As opposed to English, where the various nuances are expressed by combinations of independent words, in Latin each individual ending is a synthesis of different pieces of information: the person performing the action, the time it took place, as well as various other nuances. The drawback of this system is that there are so many different endings, which all have to be memorized individually. But ‘synthetic’ structures such as this do also have their advantages. The architects behind the Latin system made it possible to express a wide range of nuances with admirable brevity. In the form ed-ar, for example, a single one-syllable ending -ar encapsulates all the information which English has to code with the rather long-winded ‘I will be (eat)-en’. And by the time an English fish has managed to spit out ‘I will be eaten’, the seal will have polished him off three times over.

*   *   *

Finally, if the Latin verbal system looked uncomfortably complex, here is an example which makes Latin seem like child’s play: the verbal system of the Semitic languages, such as Arabic, Aramaic and Hebrew. The architecture of the Semitic verb is one of the most imposing edifices to be seen anywhere in the world’s languages, but it is founded on a concept of the sparest design: a root which consists of only consonants. The verbal root in Semitic is not a pronounceable chunk like English ‘eat’ or Latin ‘ed-’, but a group of just three consonants, like the Arabic l-b-s, which means ‘wear’, or s-l-m, which means ‘be at peace’.

But how can a vowel-less group of three consonants ever mean anything, if it cannot even stand up on its own three legs and be pronounced unaided? The answer is that such roots do not have to be spoken by themselves, because the root is an abstract notion, which comes to life only when it is superimposed on some templates: patterns of (mostly) vowels, which have three empty slots for the three consonants of the root. To take one example, the Arabic template forms the past tense (in the third person ‘he’), so if you want to say ‘he was at peace’, you just insert the root s-l-m (‘be at peace’) into that template, to get:

And if you want to form the past tense of another verb, say ‘wear’, you take the root l-b-s, and insert that into the same template, to get labisa (‘he wore’). In order to give a sense of what these templates actually feel like for native speakers of Semitic languages, I will fill the three blank slots with the consonants of the fictitious root s-n-g – let’s pretend, for the sake of argument, that it means ‘snog’ (Americans, read ‘make out’). So in other words, instead of representing a template purely notionally as (PAST TENSE), I will write it as (‘he snogged’). The should thus be taken to represent the first consonant of any root, the represents the second consonant, and the third. The template is just one among many dozens in Arabic, and these express every conceivable nuance of the verb, from ‘he snogs’, to ‘the action of snogging each other’. The table overleaf illustrates a few more templates from Arabic, to give an idea of the complexity of the system.

To English ears, words like Islam, Muslim, Salām, which have hardly any vowels in common, may sound quite dissimilar, but for speakers of Semitic languages, such words, as well as names like Salman, Suliman, Salim, Solomon, (Ab-)salom, are all perceived as closely related variations on a theme: the root s-l-m.

The Semitic verbal architecture may already seem pretty scary, but please fasten your seat-belt, because the cells in the table above represent only a handful of around a hundred different such nuances in Arabic. And if all that were not enough, each of these cells can actually contain up to thirteen different forms for the different persons (I, you, she, etc.). Again, merely to give the gist of what’s involved, here are the forms for the different persons in just one of the cells above (the top left corner), the simple present tense.

It won’t come as a surprise that the verb is the cause of some consternation for learners of Semitic languages. But since we do not need to sweat over the details, let’s just sit back and reflect on the principles involved. Think for a moment about all the meticulous planning which must have gone into developing such a system – it almost defies belief that such an algebraic scheme could have been conceived in any other way except through the inspiration of a gifted designer. How else could the abstract idea of a purely consonantal root have been devised? Is it really possible that the templates that produce a whole network of nuances could have arisen of their own accord? Cracking the Semitic verb poses a serious challenge, one which will be taken up in Chapter 6.

SEXED TURNIPS AND OTHER IRREGULARITIES

The previous pages presented a few examples of both familiar and exotic structures from languages around the world. Needless to say, there is a great deal more to the structure of language than what we have seen. There are whole expanses of language that have not been mentioned, and those areas that were touched upon were only sketched in rough outlines. Nevertheless, even the few examples above should have left little room for doubt as to the sophistication of language’s structure and the ingenuity of its designers.

But it would be disingenuous not to mention another side of language, a less appealing aspect that I have so far conveniently overlooked. For wherever one finds impressive edifices in language, one is also likely to find scores of imperfections, a tangle of irregularities, redundancies and idiosyncrasies that mar the picture of a perfect design. English, for example, is renowned for the irrationality of its past tense verbs. Native speakers may be blithely unaware of the chaos that reigns in the English verbal system; not so anyone who has had to learn it at school. Here is a rhyme I wrote in memory of my frustrations:

The teacher claimed it was so plain,

I only had to use my brain.

She said the past of throw was threw,

The past of grow – of course – was grew,

So flew must be the past of fly,

And now, my boy, your turn to try.

But when I trew,

I had no clue,

If mow was mew

Like know and knew

(Or is it knowed

Like snow and snowed?)

The teacher frowned at me and said

The past of feed was – plainly – fed.

Fed up, I knew then what I ned:

I took a break, and out I snoke,

She shook and quook (or quaked? or quoke?)

With raging anger out she broke:

Your ignorance you want to hide?

Tell me the past form of collide!

But how on earth should I decide

If it’s collid

(Like hide and hid),

Or else – from all that I surmose,

The past of rise was simply rose,

And that of ride was surely rode,

So of collide must be collode?

Oh damn these English verbs, I thought

The whole thing absolutely stought!

Of English I have had enough,

These verbs of yours are far too tough.

Bolt upright in my chair I sat,

And said to her ‘that’s that’ – I quat.

Another area where languages often display erratic behaviour is what linguists call ‘gender’, by which they don’t necessarily mean distinctions based on sex, but any classification imposed on nouns according to some of their essential properties. ‘Masculine’ versus ‘feminine’ is indeed one of the most common distinctions, but many languages choose instead (or in addition) to divide nouns into ‘human’ versus ‘non-human’, or ‘animate’ (humans and animals) versus ‘inanimate’, or sometimes even ‘edible’ versus ‘non-edible’. (Which class humans then fall into depends, of course, on local custom.)

While the idea behind such gender distinctions sounds quite sensible, the problem is that in most languages reality doesn’t match up to the theory, and so it is often difficult to discern any logic behind the actual classification. The American author Mark Twain came across such a capricious classification system for the first time when he was trying to master the German language. Like most foreign learners, he was somewhat put out by the arbitrary gender of different objects, and in his book A Tramp Abroad, he vented his frustrations in an appendix entitled ‘The Awful German Language’:

Every noun has a gender, and there is no sense or system in the distribution; so the gender of each must be learned separately and by heart. There is no other way. To do this one has to have a memory like a memorandum-book. In German, a young lady has no sex, while a turnip has. Think what overwrought reverence that shows for the turnip, and what callous disrespect for the girl. See how it looks in print – I translate this from a conversation in one of the best of the German Sunday-school books:

GRETCHEN: ‘Wilhelm, where is the turnip?’

WILHELM: ‘She has gone to the kitchen.’

GRETCHEN: ‘Where is the accomplished and beautiful English maiden?’

WILHELM: ‘It has gone to the opera.’

Twain could not understand why, for instance, German rain should be a ‘he’, a German fishwife should be an ‘it’, and a German fish-scale a ‘she’. So after a few more pages of rant, he went on to recount the following touching ‘Tale of the Fishwife and its Sad Fate’, purportedly translated literally from the German:

It is a bleak day. Hear the rain, how he pours, and the hail, how he rattles; and see the snow, how he drifts along, and of the mud, how deep he is! Ah the poor fishwife, it is stuck fast in the mire; it has dropped its basket of fishes; and its hands have been cut by the scales as it seized some of the falling creatures; and one scale has even got into its eye. And it cannot get her out. It opens its mouth to cry for help; but if any sound comes out of him, alas he is drowned by the raging of the storm. And now a tomcat has got one of the fishes and she will surely escape with him. No, she bites off a fin, she holds her in her mouth – will she swallow her? No, the fishwife’s brave mother-dog deserts his puppies and rescues the fin – which he eats, himself, as his reward …

Twain was venting his anger at German, because German was the language he happened to be learning. But despite his protestations, there is really nothing special about German in this respect. French, for instance, with la pluie, la grêle, la neige, would not cut a much better figure: ‘hear the rain, how she pours, and the hail, how she rattles; and see the snow, how she drifts along…’ And if Twain had tried wrestling with Russian, Latin or a long list of other languages, he would have encountered similar idiosyncrasies. A stone, for instance, may be an ‘it’ in English, but it is most definitely a ‘he’ in German, Norwegian, Polish, Albanian, Russian or Lithuanian, and unquestionably a ‘she’ in French, Italian, Irish or Hebrew. Classical Greek and Akkadian (the language of Ancient Babylon and Assyria) came up with something even better, since in these languages, a stone was a ‘he’ or ‘she’ depending on one’s fancy.

A comprehensive survey of all the different types of irregularities in all languages would make for a very hefty tome indeed. So I will mention just one more example of particularly eccentric behaviour, from a North American Indian language of the Kiowa family, called Jemez, spoken by about 2,000 people who live near Albuquerque in New Mexico. Jemez has an ending -sh which is placed on nouns in order to change their number, as can be seen below:

It seems, then, that the Jemez ending -sh performs exactly the same function as the English plural ending -s. And what could be more sensible than that? But now consider what happens when the ending -sh is added to a different group of nouns in Jemez:

On these nouns, the ending -sh has quite the reverse effect, as instead of marking plurality, it indicates a reduction in number. When it is added to nouns like weeds or trees, which tend to come in quantities, the ending -sh marks them as few (one or two). But how can the same ending function as a plural marker with some nouns, but as a ‘paucal’ marker with others? It seems that even Jemez speakers themselves were not entirely comfortable with this polarity between the two groups, so on a third set of nouns they decided to opt for the middle ground:

It would appear, then, that language strongly bears out Napoleon’s dictum that it is but a small step from the sublime to the ridiculous. On the one hand, the designers behind the structure of language have somehow managed to erect magnificent palaces of sophistication, but for some mysterious reason they failed to clear away the piles of ramshackle irregularities and irrationalities that lie just a stone’s throw away. To understand what has brought about this mix of grandeur and folly, we will have to uncover much more of the forces that shape, batter, and renovate linguistic structures. The following chapters will set out to do exactly that, and the first challenge will be to solve a simple-sounding problem: what is it that makes language change?