Diagram Composition

In the early days of the crossword the diagram was so constructed that every letter contributed to two words, one across and one down. This meant, unless the composer was exceptionally ingenious, or lucky, or persistent and prepared to spend an immense amount of time, that there must be few long words and many short ones, and that he would only be able deliberately to choose a small proportion of his words; many would drive themselves into the diagram as being the only ones possible. Hence the spate of emus and gnus, of ores and odes, of ohms and ergs, that occurred in early crosswords. Even two-letter words and abbreviations were admitted, and the sun-god Ra achieved a new fame and became almost a household word.

This didn’t matter much as long as the clues remained mere definitions: one just read “flightless bird” and wrote in EMU, though it did tend to pall after a year or two. But as the cryptic clue gradually came in and more than a definition was sought, it became clear that the composer must have more choice. He must be able to use long, interesting words, or phrases (though these only came into common use later) and to avoid two-letter words altogether and to reduce three-letter words to a minimum. The result was that in many papers, quite early on, blocks were used to shut in some letters and save the composer from making them contribute to two words. At first this was only done in parts of the diagram: wide open spaces still appeared here lull there (see The Times Crossword No. l at the end of this book).

Torquemada, with his bars and lack of symmetry, still stuck to full, or nearly full, “checking”: this is the technical term for making a letter contribute to two words, letters contributing to only one word being called “unchecked”. The difficulty of Torquemada’s clues and the rarity of some of his words made him do this, and some straight-definition composers still do it today. But the use of many unchecked spaces grew in most everyday crosswords, until nowadays the normal thing is for roughly every other space to be unchecked.

Now this can be done in two ways; either the outside rows and columns can be free of blocks, with the even-numbered rows and columns blocked, thus:

or conversely the diagram can be constructed thus:

I very much prefer the first of these two methods, because it leads towards what I want to make the first principle of crossword diagram composition, that at least half the letters of every word should be checked. I consider that a five-letter word with only its second and fourth letters checked is unfair to the solver, though it is very common practice today. In some diagrams the two methods are combined; sometimes, also, small wide open spaces appear, as if to compensate for the lack of checking in other places. But do they compensate, especially when, as sometimes happens, the less-than-half-checked words have none too easy clues?

Another practice in diagram composition is the unchecking of two or even three consecutive letters, thus:

or thus:

The second arrangement, it is true, is generally used for long words or phrases (twelve to fifteen letters); but even so I think the composer is saving himself too much trouble at the solver’s expense. Many solvers are only seeking diversion and don’t mind if they fail to finish a puzzle; but there are many too who don’t like to be baffled unless they feel it’s their own fault. If, having filled in all the interlocking words, they are left with something like -A~E-, there are so many possibilities that even a fairish but somewhat elusive clue may fail to suggest the answer in reasonable time, and the solver has fair grounds for feeling dissatisfied. After all, interlocking, so that the solution of one clue will help towards solving another, is of the essence of the crossword; we have seen that there are reasons why it should not be complete, but it should be adequate.

There is another feature of diagram construction that we composers ought to study and often neglect. In very early crosswords the diagram was sometimes so arranged that the four quarters were completely cut off from each other: in effect, there were four small crosswords instead of one large one. This was soon felt to be unsatisfying, and one never sees it nowadays.

But one does meet diagrams still—I have used them myself, I admit —in which the quarters are very nearly isolated. A popular type is one like this (the north-west corner is illustrated) :

A solver may make a good start in this corner and complete it, perhaps, but be baffled by the second part of the ten-letter phrase at 4 down; then what does he find? He has no letter to help him to continue! After solving six clues, and part of another one, he may well feel entitled to some guides to continuation. Such a diagram has the advantage that it eliminates words of less than six letters. This frees the composer for once from short words that tend to become hackneyed, STIR and STOP and STEM and so on. But the disadvantage to the solver must outweigh this: the composer’s job is to entertain and to avoid irritating.

Nothing has been said yet about symmetry in diagrams. It is curious how obligatory in nearly all crosswords this has become.

Torquemada disregarded it, and in a few crosswords, where bars instead of blocks are used to separate words and to uncheck spaces, it is still sometimes disregarded, especially in very difficult puzzles with a complicated twist, where fairly full checking is considered necessary. But symmetry is the regular thing, certainly in diagrams where blocks are used. Is this entirely for the look of the thing? I think it is, very largely. Afrit argued that symmetry ensures that an even distribution of unchecked letters -Not more in one quarter than another-is maintained in the diagram. He thought, probably, that without its restraint a composer might give way easily to something like this:

North-west corner:

South-east corner:

We should then have a three-letter and a five-letter word with less than half their letters checked. But if he were a conscientious composer, he could still depart from symmetry without doing this, and he might be able to work in a word or phrase, lending itself to a sparkling clue, which symmetry would exclude. No, I think it is the look of the thing that does it; an unsymmetrical diagram would affect many people as a clashing color scheme does. It certainly does that to me, though very occasionally, in a puzzle involving special complications, like the inclusion of a large number of Fixed words, I have deviated slightly from it, and I believe most solvers didn’t notice it. At least hardly any said they did; perhaps they were too polite.

The technique of filling the diagram can best be shown by practical demonstrations, which follow later. But a few general observations may interest solvers who may have tried their hands, perhaps only occasionally and without much success: I have met many of these. And I might even manage to say something which has escaped actual practitioners.

First, the composer should be fairly ambitious but not too ambitious. Long words and phrases lead more readily to amusing clues than short ones, and the ideal 15 x 15-square blocked diagram should, I think, contain twenty-eight words rather than thirty-two or more, with four long words OI’ phrases, usually of twelve to fifteen letters, and two words, not three, in each of the other open rows and columns. There is no reason why all four long words or phrases should be of the same length, since the sort of symmetry which only makes the diagram look the same when turned upside down is just as legitimate as the sort which makes it look the same from all four aspects. Suppose you have thought of two which you can clue pleasantly, one of thirteen and one of fourteen letters. If they won’t fit into each other, put them at two edges of the diagram: if they will fit - and they will somewhere, more often than not - let them cross each other, and fill in the other necessary blocks for symmetry. If they have to go at the edges, you will have to consider first the thirteen-letter one: are its odd letters more helpful as beginnings or as endings of interlocking words? The odd letters are the vital ones: the even ones will be unchecked. For if you expose the even ones by putting a block in each corner of the diagram, you will have only six checked letters and seven unchecked, and this we have already decided to avoid as a rule. Letter-frequency at the beginnings and ends of words is a thing which the composer learns gradually: the following conclusions are based on experience, not on scientific statistics, but I feel pretty sure about most of them.

R, S, and T are friends at either end: so, on the whole, is G, especially in longish words, because of -ing. Most other consonants are more helpful at the beginning, except, emphatically, N, which is much more helpful at the end, and to a lesser extent, though definitely, I think, D and L. E is certainly our greatest friend at the end: S is a rival, if we are going to admit a lot of plurals or 3rd persons singular, but to do that is weakness. A is probably the best vowel at the beginning; the others are not bad for long words, thanks to compounds, ex-, in-, over-, un-, under-, etc., but are apt to involve one in tiresome repetition with short ones. A and O are tolerable at the end, I and U much too rare to to used there if we can help it. Y is far more helpful at the end.

And of course you hardly need to be told not to put J, Q, or V at the end, or X or Z at either end if you can help it.

I would sum up the whole matter thus: the only letters that are more helpful at the end than at the beginning are D, E, L, N and Y, while G, R, S, T and, perhaps, H and K are indifferent.

So, if your thirteen-letter word is BACK-SCRATCHER, put it at the top of the diagram or in the extreme left-hand column: all its odd letters are favourable for beginnings, B, C, and, perhaps, H are less favourable for endings. But if your word is NONRESISTANCE, it is five to one the other way (only the I is unfavourable); so put it at the bottom. Often the choice won’t be so obvious: if it isn’t, you must take a chance; but very often the scales will be tipped one way or the other.

Let us suppose your word was BACK-SCRATCHER and you have put it at the top, and that the extreme north-west corner of the diagram looks like this:

Let us also suppose that you wanted to use VERISIMILITUDE. It at once becomes obvious that you mustn’t use it in this puzzle. If you put it in the left-hand column, every word across on that side will start with a vowel, which is sure to be a nuisance; and you can’t put it in the right-hand column, which must begin with R.

So scrap your VERISIMILITUDE and keep it for a later puzzle.

If, when your four long spaces are settled, you have admitted a difficult ending such as U or I, continue your composing there and fix that word. And all the time, as you proceed, avoid putting blocks in places which will give you few possibilities to choose from, and put them where you will have many. Suppose there is an S somewhere in the middle: try to get a block immediately before it rather than give yourself -s to proceed from.

Whenever you put in a word, look to see if you are facing yourself with a very restricted choice or even an impossibility. If the choice somewhere is between _ H_ E and _ A_ E at the end of some word not yet Fixed, choose the latter. Only so will you avoid tiresome wholesale scrappings of whole quarters of the diagram.

You have to think of two things (at least) at once: word-possibility and word-clueability. The latter is desirable, but it may have to give way sometimes to the former, or you will never have a diagram to clue! And later you may find that an unlikely-looking word isn’t so difficult to think of a clue for after all.

If you are using the type of symmetry which only makes the diagram look the same when it is turned upside down (not if it is given a quarter-turn), it is important, when putting in blocks,

not to make two quarters much more difficult for you than the other two. The preceding diagram will make this clear.

Consideration of various points I have made may have led you to produce something like that. You may even have filled the difficult north-west corner with its many longish words: remember that the north-west corner, where you are working from beginnings, is the easiest. The south-west and north-east corners will be child’s play; but the south-east corner will probably be the very devil! Much better to close up the north-west and southeast corners a bit more with blocks, and include more longish words in the other two. You won’t like scrapping your beautiful north-west corner; but the odds are very long that you will have to do so in the end, and you’ll hate it much more in a couple of hours’ time! Why is the north-west corner the easiest to fill and the best place in which to start? Because one thinks forwards and not backwards: it is easier to proceed from beginnings than up to ends.

Rhyming dictionaries are a help in the latter process, but that isn’t enough to compensate. So as soon as you have settled your long spaces, and any specially difficult endings, like the I or U suggested above, go to the north~west. Go next to the north-east or south-west, whichever looks harder to settle: there you will at least be working half the time from beginnings. The south-east will be the hardest struggle, nearly every time. “Then why not start there and get it done?”, you ask. Because if you do, you can’t take advantage anywhere of working from beginnings, and the whole thing will take longer. This is the normal principle, but circumstances can make it wise to violate it. You will find that my colleague in Chapter IX and I myself in Chapter XII have both found particular reasons for working on the south-east corner before going to the north-west. But these are particular reasons, caused by apparent impending difficulties; they don’t affect the validity of the principle in cases where there are no such reasons.

Now for my final word on diagram composition, and the most vital. Don’t be obstinate. When things look black, scrap and start again! I only wish I always followed my own advice. So often you hit together a combination of words easy to clue, but the last one or two words in the corner just won’t come. You spend ages, and at last you have to give way. You change two or three words to duller ones, and the whole thing is finished in a few minutes. And remember that you may well be able to use those nice words that have been scrapped in your next puzzle. It simply doesn’t pay to go slogging on with the luck spitting at you; but I shall probably go on doing it from time to time!