A Rational Approach to Language Learning

The materials on this site promote what might be called the rational approach to language learning. Instead of presenting users with samples of the language in use and leaving them to figure out how it works, a rational approach takes a language to pieces and shows learners how to build utterances in it for themselves.

A combinatory system

This surely is what language learning is all about. A language doesn't consist just of the things that people currently say in it, or even all the things that anybody has said in it hitherto: it consists of all the things that can ever be said in it. So a language can't just be a collection of words, but must provide a way of combining elements to make new meanings, otherwise nobody could understand a new sentence like Was Daisy trying to toast a cabbage? Different languages arrive at radically different solutions to this problem. For example, the French je n'ai rien vu (literally I not have nothing seen) is structured very differently from the English I didn't see anything. So learning a new language means learning a different combinatory system, and that's what makes learning languages interesting and challenging.

All languages contain three subsystems, and a rational approach needs to analyse each of these sub-systems, so that learners can use the elements of them to create their own utterances. The three sub-systems are the speech-sounds, the grammar and the vocabulary. Learning to manipulate these three elements is not as large a task as it might seem. Most languages use between twenty and forty speech-sounds, and have a grammar whose main features can be covered in about thirty pages. Vocabulary is messier and less predictable, but even here all languages have a core of about two hundred 'function words' (prepositions, conjunctions, etc.) which are quick to learn and highly productive. So whereas the task of learning a language may require persistence and attention to detail and freedom from ingrained ways of thinking, it is not overwhelming in size.


The first of the three sub-systems of a language is its speech-sounds, and in these materials I deal with the language on a phonetic basis. That is, I describe the behaviour of the sounds that constitute the spoken language, rather than the letters that make up the written language. This reflects the natural view that speech is more deep-seated than writing: children learn instinctively to speak, but not to write, and while there are many societies and individuals that can speak but not write, there are none that can write but not speak. The approach is not new: it follows a tradition developed in the 1930s by the Phonetics Department of University College, London (though less popular lately). It means that the language must be represented by means of a phonetic alphabet - one that shows its particular speech-sounds - instead of by its normal spelling.

Textbooks that work with the normal spelling of the language to be learned - what is called its 'orthography' - are not good for learning to speak it. This can be seen from the many learners who have used such textbooks and find themselves with a reasonably good reading knowledge, but no ability to speak or to understand what is said. One reason for this is that the ordinary spelling is, for many languages, a complicated and unreliable guide to the pronunciation, and one whose conventions may be very different from those of the learner's own language - the way in which the letter-sequence oi is pronounced oy in English but wa in French is a case in point. The result is that learners find it difficult to work out what the pronunciation is, and when they have worked it out they cannot be sure whether their answer is correct or not. Another difficulty with orthographic textbooks is that they introduce an additional stage between understanding the structure and speaking the utterance: the learner must select the appropriate elements, modify them in their orthographical form and put them in the right order, then convert that to a stream of speech-sounds. It requires less effort to start with the elements in their speech-sound form, and build the utterance with those.

But there is a further problem that is more serious than these. This is that orthographic textbooks don't usually distinguish those speech-sounds in the language that make a difference between one word and another (technically, its phonemes) from those that don't. In English, for example, you can say 'coo' with a plummy sound like an old-fashioned actor, or with a flat sort of sound like a contemporary teenager, and it makes no difference which you use. But in French there is a difference - as a cousin of mine found to her cost - between Merci beaucoup! with the actor-ish sound, meaning 'Thanks a lot!' and Merci. Beau cul! with the teenager sound, meaning 'Thanks. Nice bum!'. Audio recordings don't solve this problem. In all languages, native speakers intuitively bend the sounds in ways that are characteristic for that language, and this makes it virtually impossible for the learner to recover the essential distinctions. (I once got a telling off from an Istanbul taxi-driver for getting the name of my destination wrong in just this way.) Add to this the fact that the actors doing the recordings may adopt 'character' voices and different regional accents, and the result, for the learner who is trying to analyse and copy it, is like trying to cook a meal from a handwritten recipe that has been left out in the rain.

With a phonetic transcription, these problems disappear. The language represented on the page is the language that is heard and spoken, and there is no separation between the elements that must be manipulated to create an utterance and the elements that must be spoken. The sounds that must be kept distinct are clearly identified, and the phonetic make-up of each word clearly shown. This means that you can find out the pronunciation of a word simply by finding it in the text - a much more practical proposition than listening to endless recordings.

These materials, then, tell you in detail what the phonemes are, how to produce each one, and how to modify these sounds in connected speech so that you sound more like a native speaker; each sound is described in both non-technical and technical language, with a glossary of technical terms. As far as the phonetic notation is concerned, the only sensible choice is the International Phonetic Alphabet (IPA), which has been in existence for 120 years and is widely used in language-learning texts (though more widely in continental Europe than in Britain and America). Other phonetic systems exist, but the merit of the IPA is that it uses the same set of symbols for all languages. Learning it is therefore an intellectual investment: given that you have to learn a phonetic system anyway, it makes sense to learn a system that can be at least partly reused when you learn another language. Of course a textbook has to cover the orthography at some stage, but this is best done (in my view) when the spoken language has been fully described. It's easier to learn to write when you know the inventory of speech-sounds to be represented, and are working with familiar words.


Grammar is often thought of as a set of rules that tell you when to use its and it's and not to say it were uz if you come from the north of England; its role in this context is to promote a national standard by deprecating usages that deviate from it. The kind of grammar used in language learning, however, is a more fundamental sub-system than this: it provides the blueprint for constructing any well-formed utterance. In other words, it is the combinatory system that joins words together to make phrases and sentences. It has been called a device for generating all and only the correct utterances in a language.

The materials presented here use this kind of grammar. They follow a scheme based on phrase structure, and describe how the language forms words, phrases and clauses that will operate as nouns, verbs and qualifiers. Since everything in a language takes the form of word, phrase or clause and performs the function of a noun, verb or qualifier, this scheme must in principle cover all possible utterances. It does however oblige the materials writer to ensure that all significant patterns of noun phrase, verb phrase and so on are represented, and also requires him to deal with all the morphology (the changes to word-endings that typically express plurality or different tenses), which is much more extensive in some languages than others.

A grammar on a phonetic basis reveals features that are often overlooked. Sometimes the description of the way sounds behave is more complicated than the way the orthography behaves: the English past-tense ending spelt ed, for example, must be specified as /t/, /d/, or /id/, depending on what sound precedes it (passed, posed, parted). Spanish verbs become simpler, since much of their complication arises from orthographic changes designed to represent the unchanging pronunciation of the stem; French verbs become harder, because the question (for example) of whether the stem-vowel in nous aimons is pronounced like e-grave or e-acute can no longer be sidestepped. Again, the Portuguese verb dever, which is irregular in that it changes its stem-vowel between 'close e' and 'open e' for different persons, does not normally appear in lists of irregular verbs, because orthographically speaking it's regular. Overall, phonetic treatment of the grammar brings lots of hidden features out into the open, and results in a more accurate and useful account.

Grammatical coursebooks normally contain sets of exercises, one on each form or structure. The materials on this website use a software package that provides the learner with virtually unlimited practice in morphology (changes to the forms of individual words) and syntax (fitting words together to make phrases). These exercises have a number of advantages over textbook exercises. Firstly, they cover all the vocabulary, forms and syntax-patterns in the language, whereas textbook exercises don't do this. Secondly, they are context-free, so the point of the exercises - which is to develop fluency in all the resources of the language - is not obscured by non-linguistic considerations, such as the social decision in French whether to use tu or vous. Thirdly, the exercises can be repeated indefinitely. And fourthly, the exercises are not progressive: learners are not forced to follow a set sequence through the material, but can dip and skip as they wish, maintaining the motivation that comes from curiosity.

The result of the rigour and economy described above is that a workable grammar can usually be kept down to about 30 printed pages, for which the software gives unlimited practice.


The third sub-system of a language is its vocabulary. Vocabulary is the least systematic part of a language, and therefore the least productive when learnt. Learning the speech-sound system enables the learner to pronounce any word or phrase, and learning the grammar enables the learner to construct any utterance; but learning vocabulary just gives the learner more vocabulary.

One way of dealing with this difficulty is to concentrate on 'function words'. These are words such as articles, prepositions and conjunctions, whose role in language is to link things together; this makes them highly useful. In the nonsense-poem from Through the Looking-glass, for example, 'Twas brillig, and the slithy toves / Did gyre and gimble in the wabe..., it's the function words the, did, in and so on that enable us to make some kind of sense of the lines - they signal to us that toves is a noun, no doubt denoting creatures of some kind, and that gyre and gimble are verbs, denoting what the creatures did. Similarly, if you hear the Swedish phrase ett hus, and you know that ett is the indefinite article a, you can be reasonably sure that hus is a noun. Function words are few in number and don't change over time, so once learned they stay learned.

A second way of reducing the burden of learning vocabulary is to focus on the words that are actually needed. There's a core vocabulary - greetings, numbers, food and drink - that is likely to be useful in almost any situation, and these materials build on that notion by adopting a limited vocabulary - about 600 words - of high everyday usefulness. Beyond that, it makes sense to focus on the words that are needed when using the language in real life, and this will depend on what the learner intends to use the language for. A practical way of developing this vocabulary is to look up, before any linguistic encounter, all the words that are likely to be needed, and then to check afterwards for any new words that turned up. It's often remarked that each learner's vocabulary is personal to them, and this necessity to choose a learnable number of words out of the tens of thousands in the language is the reason.

In the materials on this website, vocabulary is grouped by meaning. This makes it easier to remember it - there's evidence that we store words by meaning in our memories. This grouping doesn't prevent the learner from looking up a word quickly - there isn't so much vocabulary that a separate alphabetic list is required. So the learner doesn't need to worry so much about memorisation, and can take a more relaxed approach.


Languages vary from locality to locality, from social class to social class, and from generation to generation (which is why teenagers' speech-habits horrify their parents). This means that learners have to decide which variety they are going to learn, and materials writers have to decide which variety to describe (and say which one it is). There are those who claim that a textbook should contain a mixture of dialects, on the grounds that this makes the learner more flexible, but this claim is unconvincing. If a textbook mixes dialects, no one of them can be described fully or accurately; the resulting account will be inconsistent and incoherent, full of imprecision and muddle, and a muddled learner is anything but a flexible learner. It would be like trying to teach learner-drivers to drive on both sides of the road at once. So the materials writer has to choose a dialect, and arrive at a view (for example) about the pronunciation carr parrk as against cah pahk, the use of outwith or outside, and the choice between kids, children and weans. He has to do this not because one alternative is inherently more correct than the other - the point of a descriptive account is to describe, not to pass judgement - but because one alternative is a more accurate record of what people in a given speech-community actually say, and so makes a better learning tool. In the materials shown here, I aim to present a version of each language that has no unusual features and will pass without comment as ordinary, everyday speech.