R.M.W. Dixon, Are some languages better than others?
(Oxford) (extracts)
R.M.W. Dixon, Are some languages better than others?, Oxford University Press, 2016
(p.1) The reader will be able to decide for themself whether some languages can be considered ‘better’ than others (taking to be certain what one means by ‘better’).
(p.8) There are some folk who insist that one should never try to compare the relative worth of languages. Why not? What are they afraid of? If linguistics is to be regarded as a natural science then evaluation must be an element in its modus operandi. The doubters say that all languages are ‘equal’, that each language is perfect for the role it plays in the society which uses it. But if a language is perfect, why — indeed how — could it ever change? And each language is in a process of change, all the time.
(p.13) Linguistics is generally reckoned to be the second oldest science (after astronomy); its beginnings were in India, before 500 bce, with Pānini’s magisterial grammar of Sanskrit. Like every other science, it has four fundamental tasks: description, explanation, prediction, and evaluation. We can briefly comment on these.
(p.13) (a) Description. For each language, there should be a full grammar, detailing every grammatical structure, every prefix and suffix, their meanings, and their possible combinations. Each sentence in the language should be providable with a grammatical analysis. And, by applying the rules of the grammar in an appropriate manner, new sentences — which are judged as acceptable by native speakers — can be generated.
The second component is a full lexicon (or vocabulary). (…)
For a little-known language, there is a third component to the description: a collection of texts. (…)
(p.17) (d) Evaluation
Comparing two things and assessing their worth is a natural practice in most disciplines, but it is something which has by-and-large been shunned in linguistics. If linguistics is to be recognised as a science, which is my contention, it has to seriously engage in evaluation.
(p.19) The tribe I lived amongst in the Amazon jungle is called Jarawara by neighbours, but their term for themselves is ‘Ee jokana’, literally ‘We, the real people’. All else is unreal.
(p.20) Stage 1, Racist evaluation. When Europeans used their ships to traverse the world, and their guns to conquer substantial territories, they came into contact with many ethnic groups. These were inferior—to varying degrees—in material culture. On this basis they were judged to be inferior people, and their languages were assumed—without evidence—to be primitive things, with just a few hundred words and at best a smattering of grammar.
The intruders, with their blinkered view, only perceived what was on the surface. In fact, these ethnic groups typically had more finely-tuned social systems than those of the invaders, and languages which were certainly as rich, often richer. It is surely significant that Europeans typically experienced considerable difficulty in mastering the local languages, whereas the conquered people soon exhibited an easy fluency in English, French, German, Spanish, Portuguese, Dutch, or whatever.
Stage 2, Redress. So pervasive was this racist evaluation that, if one was to teach the elements of linguistics, it had to be countered at once. The first pages of textbooks and the first lectures of freshman courses emphasised, as loud as was possible, that ‘no language spoken in the world today is primitive’ and then ‘that all languages are about equal in complexity’.
Stage 3, Scientific evaluation. I suggest that, linguists having now devoted about a hundred years to redress of the racist idea, it is time to fulfill one of our missions as a science and embark on a measured evaluation of the worth of different languages. The present book aims to be a first step in this direction.
(p.25) A language involves two independent but interlocking parts: grammar and lexicon.
/Grammar is like a little city centre; a lexicon is loke a parking lot./
(p.28) The lexical and grammatical elements of a language are represented by sounds, which are articulated by the speaker and then heard, and their meanings understood, by the hearer.
For every language, a set of phonemes can be recognised. There are speech sounds which carry a meaning contrast, and their make-up varies between languages. For example, English has contrasting phonemes /r/ and /l/; substituting one for the other produces a new word, as in /lap/ and /rap/. Jarawara, however, has a single phoneme which can be pronounced as either [r] or [1]. The most usual articulation of the name for the assai palm is [fare], but if [fale] is said, then this is recognised as the same word; [r] and [1] do not contrast in this language.
Dyirbal, in contrast, goes in the opposite direction, having three contrasting phonemes where Jarawara has one and English two. These comprise a regular l Sound, plus two kinds of r; one, which is written as /r/ is like the /r/ in standard British English (p.29) but with the tongue tip turned back a little further, while the other, written as /rr/, is a trill, as in Scottish pronunciation of English. The three phonemes contrast in /bulu/ ‘father’s father’, /buru/ ‘elbow’, and /burru/ ‘rhinoceros beetle’.
Phonemes divide into vowels (V), which function as the peak of a syllable, and consonants (C), which corne at the beginning and often also at the end of a syllable. The number of consonants in a language varies from less than ten to several score (found in languages from Southern Africa with a goodly array of clicks), while vowel inventories range from two to several dozen (in languages with nasalisation, a contrast between long and short vowels, and so on).
All languages have syllables of structure CV. The next most common type is CVC. English has one of the most complex syllable structures, allowing up to three consonants at the beginning and also at the end, as in splints. In some dialects there are even a few words with four consonants at the end of a syllable, as in the CCCVCCCC word strengthens /streŋθnz/.
The larger the Systems of consonants and vowels, the more words there can be of a given size. For a language with just 12 consonants and 3 vowels, there are 12 x 3 or 36 possible monosyllables of structure CV, and 3 x 36 or 1,296 CVCV disyllables. Compare this with a language having 45 consonants and 16 vowels; there will be 720 possible monosyllables of type CV, and 720 x 720 or just over half a million disyllables CVCV. The more contrasting phonemes a language has, the shorter its words can be.
We see that there may be many phonemes and lots of short words, or a small number of phonemes, requiring longer words. The ideal phonological System lies between these extremes. The most common vowel System, across languages of the world, has (p.30) five members— i, e, a, o, and u—as in Latin. English has adopted the Roman alphabet from Latin but in fact has, in the standard British dialect, six short vowels (plus unstressed a, as at the end of China), five long vowels, and eight diphthongs. (These are illustrated by din, den, dan, don, put, and donc for the short vowels; dean, darn, dawn, dune, and turn for the long vowels; and deign, dine, (con)done, down, coin, dear, cairn, and dour for the diphthongs.)
(p.30) The average set of consonants, across languages of the world, has 20 to 24 members. Larger Systems include sounds which speakers of other languages find hard to pronounce — apico-dental fricatives such as voiced /ð/ in English though and voiceless /θ/ in thought, plus whole series of ejectives, clicks, pharyngeals, and so on.
(p.31) Around half of the languages of the world go further than this, using differences of pitch, called ‘tones’, to distinguish the (p.32) meanings of words in the same way that vowels and consonants do. The most common System of tones has two members, marked by high and low pitch. Some languages exhibit larger Systems, with up to six (or even more) contrasting tones. There are ‘register tones’, each having a relatively fiat pitch level, and ‘contour tones’ (as in Chinese), which involve pitch movements, such as rising, falling, falling-rising.
(p.41) 2.6 Types of words
The lexicon divides into word classes, the major ones being noun, verb, and adjective. The central members of the noun class refer to things, of the verb class to actions, and of the adjective class to qualities.
(p.57) There are a few languages – such as Newar from Nepal – where a double negative does create a positive statement, and is used to make a strong assertion.
(p.58) ENG : rabbit + ‘S + ear
CHIN tùzi-DE + ērduō
(p.59-60) Finnish, Latvian, Japanese, Amele and Lango have no verb ‘to have’.
(p.67) Some languages have no verb ‘to be’.
(p.88) Slavic languages employ a grammatical system relating to extent in time. This has two members:
- ‘perfective’—an event is regarded as a whole without regard for its temporal composition (even though it may be extended in time)
- ‘imperfective’—this focuses on the temporal make-up of an event
Russian has a system of three tenses—past, present, and future—interrelating with specification for perfective/imperfective.
(p.125) Is it the case that: ‘the more complex the better? Most certainly not.
A language may satisfy many of the requirements we identify as relevant for an ‘ideal’ language either in a straightforward way, or in a convoluted way. The former is plainly preferable.
(p.138) Swahili – spoken as first or second language by well over a hundred million people in East Africa – is one of the few Bantu languages to have lost tones.
(p.149) Like all the pastoral Nilotes, [the Nuer] use an enormous number of words and phrases about cattle and the tasks of herding and dairy-work.
(p.170) The online edition of the compendious Oxford English Dictionary currently November 2014) shows 275,000 entries. However, this deals with the entire history of the language, not just its present state. Many entries are marked ‘obsolete’ or ‘dialectal’. Shakespeare, in his plays, uses about 20,000 words and Milton, in his poems, only around 8,000. By extrapolating from dictionary samples, I estimate myself to have an active vocabulary (words I use) of about 21,000, adding about 2,000 more for passive vocabulary (words I would recognise but am unlikely to use).
(p.172) How many words does a language need? How many words do most of the non-major languages actually have? All that can be offered is an educated guess—probably between five and ten thousand. English and other major languages have many more, partly through multiple semi-synonyms, but mainly due to so many particular fields of endeavour: specialised terms used in law, music, art, and all manner of trades and sciences.
(p.198) Other cultures have different conventions. In Russia a middle name is required—the patronymic. This is the father’s first name plus -ovich ‘son of’ for a boy and -ovna ‘daughter of for a girl. Thus the son of Vladimir Veniaminovich Rudov is Mikhail Vladimirovich Rudov and his daughter is Anastasia Vladimirovna Rudova (here the last name bears feminine suffix -a).
Whereas Russian names include information about the father’s first name, in Spanish the mother’s last name (which is her father’s last name) is added after the father’s last name. The full name of the daughter of Miguel Alvarez and Carmen (p.199) Gonzalez will be Maria Alvarez Gonzalez. This is how her name will appear on a passport and other official documents, but on a day-to-day basis she is just Maria Alvarez (and is placed under ‘A’ in alphabetical order). If Maria should emigrate from Peru into Brazil she would become Maria Gonzalez Alvarez on formal documents but remain Maria Alvarez informally; in Portuguese the mother’s last name is placed before the father’s last name.
(p.202) All languages are equally suitable for being learnt (to be spoken) by a child. Any child will acquire reasonable competence in a language to which it is exposed, although some will reach a certain level at a slightly earlier age than others. Relevant factors are (among others) the child’s intelligence, their innate language aptitude, the familial setting in which they are placed, and the complexity of the language.
The spoken mode is everywhere the major manifestation of language. But in many societies reading and writing are also important, and languages do differ a great deal in how easy it is to master the written mode. If the writing system is alphabetic, the ideal situation is for each phoneme to be represented by a single letter in the orthography. With a couple of minor exceptions, this is the case for Spanish. Once you know how to pronounce a word, you know how to write it; and once you know how to write a word, you know how to pronounce it.
The English writing system is quite different, and much harder to master. A single sound may be written in many different ways. For example, each of the following nine words ends in the same diphthong (phonetically /ou/), although written differently: owe, sew, toe, blow, though, cocoa, disco, depot, tableaux. In the opposite direction, the same letters may indicate different sounds; for instance, each of the following words ends in -ood, but this represents three different vowels: good /gud/, food /fu:d/, blood /blʌd/.
(p.205) Consider, for instance, Tok Pisin, which is a lingua franca over most of Papua New Guinea. This is in no way ‘primitive’; it has three numbers in pronouns, generally an inclusive/exclusive distinction, a possessive construction marked by bilong (similar to English of), relative clause and complement clause constructions, much compounding, and a great deal more. But still it is easier to master than the 700 or so indigenous languages, which show demanding complexities (reflecting the culture of which they are a part). Interestingly, children often speak first in Tok Pisin, adding the local language by the age of 5 or 6.
Chapter 1 An ideal language
(p.215) 1. Do not have intonation as the only mark of a grammatical distinction. For example, Mary says, The plumber’s coming at 8 o’clock in the morning? John responds, Is he? Mary then says, irritably, I don’t know, I was asking you. That was a question. There may hâve been some background noise—perhaps a tap running—and John hadn’t picked up the rising intonation of Mary’s (confirmation-expecting) question.
(p.216) 3 Have few or no homonyms.
(p.218) 5 Have a minimum of irregularities.
(p.219) 6 Avoid having an orthography which does not have a simple correspondence between contrastive sounds (phonemes) and letters of the alphabet.
(p.220) Have one or more productive process of reduplication.
The process of reduplication occurs in most languages of the world (but is, surprisingly, rather rare in the familiar languages of Europe). It involves repeating all or part of a word form either before or after it (or, sometimes, in the middle) and carries any of a wide range of meanings.
In Indonesian, full reduplication of a noun carries plural meaning:
rumah ‘house’ rumah-rumah ‘houses’
perubahan ‘change’ perubahan-perubahan ‘changes’
(p.221) Verbs in Dyirbal has full reduplication of a noun, also indicating plurality:
gundulu ‘cassowary’ gundulu-gundulu ‘cassowaries’
(…)
In Mandarin Chinese, verbs and adjectives are rather similar in their grammatical properties. One criterion for distinguishing them is the meanings of reduplication. With a verb this indicates ‘do a little bit’: dòng (to move) > dòng- dòng (to move a little)
In contrast, when an adjective is reduplicated, the semantic effect is intensification of the quality, as in:
hóng ‘red’ > hóng-hóng ‘vividly red’
Reduplication carries many diverse meanings across the world’s languages. With nouns it may indicate ‘collective’ or ‘dispersed’ or ‘diminutive’ and with verbs ‘do repeatedly’, ‘do intensively’, or ‘happen continuously, among other senses
(p.221) Reduplication is a straightforward and easy way of showing meaning, not involving the addition of an affix. It is a highly desirable feature, for any language.
- Have ways of forming augmentatives and diminutives.
Many languages have a set of affixes which, when added to a noun or adjective, indicate ‘a large version of’ and a ‘small version of’. From sapo ‘toad’ in Portuguese can be derived augmentative sapão ‘a great big toad’ and diminutive sapinho ‘a tiny little toad’.
Diminutive may not always refer to size at all but can simply add a warm and endearing nuance. This applies to suffix -ito / -ita in Spanish.
(p.223) Vowels. The most appropriate number of vowels is that which is most common across the world’s languages, five. Prototypically they are /i /, /e/, /a/, /o/, and /u/. This is the vowel system of Latin, echoed in the alphabet for English, but quite different from the actual system of seven short and five long vowels in British English today. (…)
Consonants. An ideal system of consonants consists of those found in most languages, and which adult learners find relatively easy—about twenty or so. Standard places of articulation involve the two lips, the tongue tip against the gums, and the back of the tongue against the soft palate at the rear of the mouth. These give voiceless stops (p, t, k), the corresponding voiced set (b, d, g), and corresponding sets of voiceless and voiced fricatives, and nasals. A single lateral (1) and some variety of rhotic, or r-sound, are pretty standard, as are semi-vowels y and w.
(p.224) 11 Syllable structure
The ideal structure to process and learn: CV(C) and (C)V(C)
(…) As a native speaker, I have always found the word sequence Smith’s crisps, /smiθs krisps/, to be something of a tongue twister. Complex structures are a luxury, which may be of mixed benefit.
12 Tones. Every word consists of a sequence of vowels and consonants, and in pronouncing them the voice must adopt some level of pitch. Why not have contrasting pitch as well as contrasting vowels and consonants? There is nothing to lose and much to gain.
Tones can distinguish lexemes, and also mark grammatical categories, such as gender and tense. A tone choice may apply once for a word, or on every syllable within a word.
(p.225) 10.3 Grammar
Demonstratives. There may be just a two-term system of nominal demonstratives (‘this’, near speaker, and ‘that’, away from speaker) and corresponding adverbs (‘here’ and ‘there’). The ideal is a three-term system, with a further contrast within ‘that’ between (i) ‘that near addressee’ and ‘that distant from speaker and addressee’, or (ii) ‘that mid-distance’ and ‘that far’, or (iii), in hilly country, ‘that higher’ and ‘that lower than speaker’. A more extensive demonstrative system is a definite luxury.
14 Personal pronouns. The minimal acceptable system has separate forms for lst, 2nd, and also 3rd person in singular and plural (here referring to more than one). A three-term number system is ideal, (singular, dual, plural (now referring to more than two)). Larger number systems are, of course, useful but fall within the basket of luxuries.
(p.226) 17 Explicit marking of moods. For maximally efficient communication, there should be explicit marking (by something other than intonation) to distinguish between a statement (declarative mood), a command (imperative mood), and a question (interrogative mood). One possibility is to (p.227) leave declarative unmarked but to employ explicit affixes to indicate the other two moods. (…).
18 Having a full set of content question words. An ideal language will have a separate form for each of the standard interrogative words: ‘who’, ‘what’, ‘which’, ‘where’, ‘when’, ‘why’, ‘how’, ‘how much’,and‘how many’. Note that English has separate forms for the first seven, but how much and how many are simply combinations of how plus an adjective: How many cars are there? is parallel to How clever is she? and to How old is grandfather?
As illustrated in section 3.1, an interrogative verb ‘do what, do how’ is most useful. It occurs in a relatively small number of languages (although widely scattered) and thus one hésitates to include it in the inventory for an ideal language.
19 Distinguishing between ‘how much ‘and ‘how many’. Many languages have a single quantitative interrogative, ‘how much / how many’ (section 3.1). If one hears, in such a language: ‘That (p.228) company owns lots of wells producing a huge amount of oil each month but I don’t know exactly how much / many’, it is unclear whether the interrogative is asking about the number of wells or the amount of oil. Many languages do have two words here, for instance kìi ‘how many and thâwrày ‘how much’ in Thai. This is clearly a desirable feature.
20 Having ample means of negating. Many languages with intricate word structures show negation just by an affix to the verb. Those with simpler word structures may have a separate negative word, and there can be wide possibilities for its placement. For example, in English one can say:
Surely you couldn’t have not known that he was not telling the
truth
The -n’t negates you could have not known that… ; the first not negates have known that… ; and the final not negates the complement clause that he was telling the truth.
There should be means for negating main clauses, every variety of subordinate clause, and also a noun phrase within a clause, as in English: No sane person could believe that the world is flat.
Most languages have forms ‘no’ and ‘yes’ as single word responses to a polar question. However, the discourse profile of others requires a full clause; a reply to ‘Is she going?’ would have to be ‘She is going’ or ‘She is not going’. It is thus open to argument whether having a single-word negator ‘no’ is desirable for languages of every possible mien.
21 Possession. If there is a single possessive construction, the ideal situation is for there to be an explicit marker, rather than (p.229) just possessor and possessee being placed in apposition.
22 Having a verb ‘to have’.
23 A system of case marking. If an activity involves two core participants, it is important to know which role each has; did Tom punch Fred, or did Fred punch Tom? (section 3.4). Some languages have no explicit mechanism, and the listener has to work it out—as best they can—from co-text and context. Obligatory bound pronouns can be a limited help. Using ordering of words is alright, but this takes away from word order the pragmatic possibilities it can embrace in a language with case marking.
(p.230) 24 Copulas. Some languages show identity and attribution simply by juxtaposition, as in ‘My mother a doctor’ or ‘That pig fat’. Ideally, these relationships are shown by a copula, which can be marked for tense and the like, in a similar fashion to transitive and intransitive verbs.
Some languages have more than one copula; for instance, be and become in English. Others have distinct forms for positive and negative copulas. Both can be considered luxuries.
25 Techniques for linking clauses. For quite a number of languages, the only way of showing that two clauses are related is to juxtapose them within one intonation unit, something like ‘John bought a car on Monday; it broke down the following day’. It is more effective to have an explicit marker of clause linking. In English, one could insert, between the two clauses just illustrated: and, just showing temporal sequence, or but, here indicating something unexpected.
An ideal language will have a range of clause linkers. (…)
(p.231) 26. Subordinate clauses. The ideal language will include a relative clause construction, one or more complement clause constructions, and a set of adverbial clause constructions .
(p.232) 27. Pivots and switch-reference. If two clauses share an argument, which is in a specified function in each, then a pivot condition may allow it to be omitted from the second clause with no loss of meaning.
Consider the following coordination of a transitive and an intransitive clause, where the S argument is omitted from the second clause;
(1) JohnA saw Mary0 (and) —s ran away
(p.233) 28. Genders and classifiers. A system of genders fulfils semantic and grammatical roles. It provides a partial codification of the way in which a society views and categorises its life-style and environment. Gender plays a useful role in anaphora. And if an adjective agrees in gender with the noun it modifies, the noun may be omitted with some indication of its reference being retained in the adjective. For example, bona, the feminine nominative form of ‘good’ in Latin, can be used without a noun and indicates something good of feminine gender. In contrast, good in English includes no such information (and indeed, requires a foliowing noun in most circumstances).
A basic component of an ideal language is a gender system of three terms, two of which relate (among other things) to female and male humans. Larger systems are a luxury (and each language can only handle a limited number of luxuries). English does not have a gender System, as the term is used here. But its (p.234) sex-based 3rd person pronouns (with occasional extensions beyond humankind) are a useful second-best, with a strong anaphoric role.
Some languages include a set of classifiers, which may have a wider semantic role than genders but lack their grammatical possibilities.
29 Definiteness. Although found, in explicit form, in a minority of languages, a definite/indefinite distinction is so useful that it should be considered basic (section 4.2). If one hears Max was the founder of the firm, it is clear that he did it all by himself, whereas Max was a founder of the firm States that he was one of several people involved. A language lacking definite/indefinite articles would just have ‘Max was founder of firm’, which is vague.
30 Tenses and modalities. All languages make explicit reference to place, but some are vague with respect to time. An ideal language has at least a basic tense system; it may distinguish past from present from future. However, past and present are known, whereas what has not yet happened can only be seen in terms of prediction, obligation, necessity, possibility, desire, and so on. An alternative to having a future tense is a system of modalities, dealing with features such as those just listed.
Some languages have several past tenses, and a few have several futures. However, the ways in which they divide up time vary considerably. These fall under the heading of luxury, as do various varieties of aspect, and the like.
31 Evidentiality. People who don’t have it in their languages often wish they did—an obligatory grammatical system whereby how a speaker knows a certain thing must be specified. The (p.235) nature of the evidence has to be stated . Was the speaker just told about it, or did they see it for themself, or know about it in some other way?
Around a quarter of the world’s languages include such a system—of varying sizes—in their grammar. One of the basics for an ideal language should be an evidentiality system of modest size, with perhaps three or so terms.
32 Comparative constructions. (…) many small egalitarian societies do not think in terms of competitiveness between people. They have no need for lexemes such as ‘win’ and ‘lose’, or for a comparative construction in their grammar.
33 Passive constructions.
(p.236) The various roles for a passive make it a desirable feature for an ideal language.
34 Reflexive and reciprocal constructions. In al societies, people do things to themselves. Some languages have no special way of describing this. Saying ‘I cut me’ is clear enough, but if one hears ‘Betty cut her’, this is ambiguous between ‘her’ referring to Betty or to someone else.
(p.237) Causatives. Every language has intransitive clauses describing activities and states. And there is generally a grammatical technique deriving a transitive clause specifying who was responsible for the activity or state.
(p.238) 10.4 Vocabulary
36 It is best to have a minimum of double duty words.
(p.239) 37 Simple lexemes are preferable to descriptive labels.
So ‘blue’ could be expressed by ‘sky-coloured’.
‘kiwi’ < Kiwifruit relabelled for Chinese gooseberry (cultivated in New Zealand)
(p.240) 38 Having a goodly array of abstract nouns. Surely no one would deny that it is beneficial to have a set of abstract nouns such as ‘time’, ‘colour’, ‘size’, and ‘age’ (section 7.4). They enable us to ask ‘What time is it?’ rather than ‘Where is the sun in the sky?’, and ‘What colour is it?’ rather than ‘Is it red, or green, or blue, or yellow or what?’
39 Distinguishing process from result. After a process of covering has been applied to some object, it can be described—using the past participle of the verb—as ‘covered’. However, there are different ways of being covered—with a blanket too thick or too skimpy or just right, fully or partially, and so on. Dyirbal has adjective ngulgun (quite separate from verb dadil ‘cover’) meaning ‘covered in just the right manner’. And similarly with other result adjectives. This is plainly an advantageous feature.
40 All kinds of names. In an ideal language, every person has a different name, and these contain useful information; for (p.241) instance, including the first name of the father, as in Russian, or the last name of the mother, as in Spanish and Portuguese.
Labels for social roles should be unambiguous; for example, a distinction is needed between ‘queen’ as ruler, and ‘queen’ as wife of a male ruler.
41 Kin terms. Much confusion is avoided by having separate terms for distinct classes of relative. For example, not lumping together ‘father’s mother’ and ‘mother’s mother’ as just ‘grand- mother’. And actually having a term for ‘son or daughter’s mother-in-law’, similar to ‘co-mother-in-law’ in Indian English. Speakers of other varieties of English simply don’t know what to call such a relative, and frequently fret over it.
42 General vocabulary. Each language deals with a set of universal concepts—things, qualities, States, activities—but these will not have identical ranges of meaning. For example, one language has separate lexemes for ‘lower arm’, ‘wrist’, and ‘hand’, while another has a single lexeme covering all three.
Classifying languages
(The Economist, 25/08/2018)