As easy as a, b, see?

Showing and saying are quite different forms of communication. But if pictures spoke louder than words, the "visual dictionary" would have been invented ages ago. This one rashly promises "a breakthrough in language learning", based on "over 80,000 translations matched to stunning illustrations". Four languages are covered: English, French, German and Spanish.

Interesting as a compilation of this kind may be, its practical utility as an interlingual lexicon is extremely limited. Ordinary bilingual dictionaries would be a far better investment. Moreover, unless an ordinary dictionary is used to check the equivalences here supplied, they can be highly misleading, since no supplementary information whatever is given about the range of application of the terms.

As a potential pedagogic tool, it initially looks more promising. The introduction speaks optimistically of "vocabulary building". Now many people would agree that a picture of an elephant will give anyone a better initial idea of what "elephant" means than the verbal gloss "large pachyderm with proboscis and long ivory tusks". But the odd thing is that the Visual Dictionary has no index entry for "elephant"; nor for many other species that are natural candidates for illustration. No giraffes, no penguins, no zebras. Its section on the "animal kingdom" is disappointingly meagre.

This is symptomatic of a more general deficiency. The compilers do not seem to have asked themselves the crucial question: when is visual information essential for the ready comprehension of a word? An obvious first answer would be: in the case of words that themselves refer to visual phenomena, such as colour and shape. One might have expected at least some guidance in the rather complex matter of colour terms, which by no means match in the four languages concerned. For here is an area where verbal dictionaries notoriously flounder. In most cases, it is far more helpful to point to colour samples than to start talking about wavelengths to explain colour words. But anyone who consults the Visual Dictionary in the hope of discovering, say, the difference between "mauve" and "violet" will look in vain. A search for the difference in shape between an ellipse and a circle will likewise prove fruitless. "Triangle" appears only as the designation of a musical instrument and, more bizarrely still, "rectangular" only as one of several types of sleeping bag for campers.

When it comes to subtleties such as the differences in shape, size and colour between various types of bread, the picture we are given is such a jumble (presumably because the artist thought it was more important to have a pleasingly decorative still-life composition than visual clarity) that it is quite impossible to tell the difference between a "baguette parisienne" and a "pain parisien". The slices of pumpernickel look like slate roofing tiles. As for a "flute", a "ficelle", a bagel, or the humble bread roll, they are nowhere to be seen.

In the introduction much fuss is made of consulting specialists in many subject areas, but there is no mention of consulting possible users or of consulting linguists. One suspects that what lies behind the resultant verbovisual hotchpotch is bad basic theory. And for this hypothesis the introduction provides ample evidence, without the help of pictures. There we are told that the Visual Dictionary focuses exclusively on nouns because nouns are "the most significant words in the language". This is the kind of statement that would risk getting a fail mark if made in an examination by any first-year undergraduate studying languages.

Furthermore, it is not in fact the case that the Visual Dictionary focuses exclusively on nouns. Many of its lexical entries are noun phrases in which an adjectival qualifier plays the key role in distinguishing one type of object from another. Many are nouns derived from verbs. But these connections remain unexplained. Thus expressions such as "batting glove", "protector de la garganta", "air conditioning" and "Wohnkuche" are in effect presented as unanalysable lexical tabs arbitrarily attached to physical objects. Far from representing a breakthrough in language-learning, this is a retrogression. Propping up the whole enterprise is the long discredited theory of languages as nomenclatures.

There is indeed something inescapably biblical about the whole style of presentation. It is an updated version of the naming in the Garden of Eden. (What do you call this gadget, Adam? Well, Lord, I think we'll call it "windscreen wiper", "essuie-glace", "Scheibenwischer" or "limpiaparabrisas".) And to go with the Adamic nomenclaturism there is an idealisation of the objects. Here we see human bodies of immaculate form (by western standards), plants and flowers without blemish, clothes and furniture from a Celestial Department Store.

The accompanying semantics are just as implausible. The introduction boasts that the Visual Dictionary is "the only dictionary that allows users to find a word from its meaning". This presupposes that users think of meanings as little pictures in the mind. But the boast is hollow. For the user who knows what something looks like and wants to find out what it is called will be stumped by the lack of any visual equivalent of an index. In practice, what has to be done in such a case is to look up the table of contents and then search through a likely section to see whether the object is illustrated. Which is not finding the word from its meaning but proceeding from verbal classification to verbal item by way of a picture.

But even if languages were nomenclatures and meanings were mental images, and even if for some curious reason anyone wanted a dictionary restricted to nouns designating visible things, this particular publication would fall down on the job. Its designers have not given enough thought to the complexity of the relations between our verbal world and our visual world. It needs to be made clear whether the words listed refer to what is shown by the whole picture or only to part of it. In the latter case, to which part? Despite its state-of-the-art computer-generated images, this book gets no further in solving this problem than the old dotted line. So, for example, a dotted line on p.453 appears to indicate that "office", "bureau", "Kasse" and "oficina" are words for a window overlooking the forecourt of a petrol station. On p.101 a single dotted line leads straight to a point in the middle of the upper belly of a rather handsome brown horse. On the other end of this dotted line we find "flank", "flanc", "Flanke" and "ijar". Only illustrators who believe in telepathy could imagine that this makes it clear to everybody which portion of the equine anatomy is designated by these words. A comparable dotted line on the picture of a cat turns out to be labelled "fur", "fourrure", "Fell" and "piel". From which it would be not unreasonable to conclude that a horse's coat is called "flank", or that a cat's side is called "fur". But deciding which would still be a problem.

Likewise, when a picture includes a number of roughly similar objects, sometimes just one is singled out by the dotted line. You have to work out for yourself whether the words listed apply just to that one or to all of them. A related puzzle is knowing how much of the information given on one picture is "transferable" to another. Thus what on one picture of the human body apparently indicates the nipple seems on another to indicate the surrounding areola. (Women have both. Men, it seems, have one but not the other; although which is visually unclear.) Inevitably such a publication projects the simplistic view that different languages are alternative sets of labels for the same inventory of things. Cases of semantic anisomorphism it cannot handle. So it tries to get round the fact that English "chair" is not the lexical equivalent of French "chaise" by having one page of illustrations of armchairs and another of chairs without arms; but the word "chair" itself is not listed in the lexicon. But even this subterfuge comes unstuck. Inadvertently the phrase "types of chair" has slipped into the classification scheme, where its French equivalent is "types de chaises". And what the illustrations imply is that while a rocking chair can be with or without arms, a folding chair is armless, as is a chaise longue.

In short, what visual lexicography shows it cannot cope with is a problem familiar to philosophers under the rubric of "ostensive definition". Which is why the idea of a dictionary replacing verbal glosses by pictures was doomed to failure from the start.

Roy Harris is emeritus professor of general linguistics, University of Oxford, and editor of Language and Communication.