Some attempt has been made, too, to evolve cipher machines which will produce effective transpositions, but our understanding is that these have never been accepted as worthwhile. The accomplishment of transposition by mechanical means is far from new. In fact, the oldest transposition cipher of which we have any record was accomplished by means of the Lacedaemonian scytale. The Spartan general, departing for foreign conquests, carried with him a rod, or scytale, of exactly the same diameter as one retained by the administration. When it was desired to communicate matter of a confidential nature, the sender, using a narrow strip of parchment, wound this carefully around his scytale with edges meeting uniformly at all points, and wrote his message lengthwise of the rod. When the strip was unrolled, the message appeared as a series of short disconnected fragments, one letter, or two letters, or portions of one or two letters. It was presumed that no person would be able to read the message without being possessed of a duplicate scytale on which to rewind the strip. We are left to suppose that this presumption was justified by fact, though the decryptor of today would make short work of such a system. The scytale, we believe, is the oldest known cipher of any kind, and is still serving today as the emblem of the American Cryptogram Association.
Before leaving types, it should be mentioned that any of the transpositions ordinarily used for disarranging single letters can also be used for the transposal of entire words. The popular name for this is “Route Cipher” — possibly because it is rather cumbersome to accomplish by any other than a “route” transposition.
We have said little concerning decipherment. This, in practically all cases, is a mere matter of performing inversely the two encipherment operations. For either process, the operator begins by setting down his key or design, or adjusting his mechanical device in the agreed manner. The encipherer “writes in” a plaintext, and “takes off” a cryptogram; the decipherer “writes in” a cryptogram, and “takes off” (or reads) a plaintext. If the encipherer, by agreement, has written the text in rows and taken it off by columns, then the decipherer must do the reverse: write his text by columns and take it off by rows.
Before entering into the subject of decryptment, the student should acquaint himself with the significance of the various tables appended to this text, in order that he may consult these or similar tables for information as to frequencies, and sequence. Every written language has its individual characteristics in these two respects, and, to learn just what these are for each language, various cryptologists have, from time to time, counted the letters, the short words, the combinations, and so forth, often on extremely long texts, afterward arranging these data in the form of charts, or tables, or lists. Two such counts are never duplicates, and there may be a noticeable difference, say, between results obtained from literary text and those obtained from military or telegraphic text; yet results for any one language are surprisingly uniform. Finding, for instance, an unexplained cryptogram in which a count of the letters shows that about 40% of these are vowels (with or without Y), we may classify it, not only as a transposition, but as one enciphered in English or German, since one of the Latin languages can hardly be written with so low a vowel percentage. Then, if we note the occurrences of the letter E, and find that this makes up about 12% of the total number of letters, we may discard the possibility of German, in which the letter E is far more likely to represent 18% of the text. Or, if the vowel percentage is high enough to point to one of the Latin languages, French would be distinguished from the others by the outstanding frequency of its letter E, sometimes as great as that of the German E, while the Spanish, Portuguese, or Italian language will not always show it as the leading letter, its place having been taken by A. In the Serb-Croat language, the letter A always predominates, and in Russian the letter O.
As to sequence, and considering English combinations only, certain digrams, such as TH, HE, AN, etc., very consistently predominate over all others. These almost never show identical percentages in any two digram counts (as the single letters sometimes will), and seldom, if ever, are ranked in exactly the same order, aside from the fact that TH invariably comes first. But in all counts, the same fifty to sixty digrams (out of 676) are always found at the top of the list. Thus the Meaker digram chart differs from similar charts made by many others; yet any digram chart is the most valuable weapon we have for attacking a cipher. The Carter contact chart contains the same general information expressed in another way for special use in transpositions. (This was not figured from the Meaker chart, but from an earlier one by Ohaver, made on the same kind of text.)
One very useful phase of frequency data is seen in the group percentages. Single letters, especially in short texts, may vary greatly from their normal percentages, while certain classes, taken as a whole, maintain a fairly constant percentage no matter how short the text. Such classes, or groups, listed under the general heading of English Frequency and Sequence Data, can be memorized as having roughly approximate percentages: Vowels, 40%; selected high-frequency consonants, 30%; extreme low-frequency group, 2%; the five most frequent letters, mixed, 45%; the nine most frequent letters, 70%. This final group of nine letters, E T A O N I S R H, hardly ever varies appreciably; the shorter groups will sometimes vary as much as 5% one way or the other.
Very useful in code decryptment is a list of the commonest words. Trigrams have also been investigated, the favorite positions of individual letters in their own words, average word-length, patterns, and endless other information, some of which is indispensable, and some merely convenient. It will not be possible, in the space at our disposal, to point out all of the uses to which this kind of information can be put; the student is urged to take his cue from the occasional short references made in connection with examples.
All ciphers are decrypted by the general methods suitable to their type, and a transposition cryptogram may involve factoring, examination of the vowel distribution, and anagramming, either singly or in combination. These are best explained in connection with examples, which may themselves have special methods, and we have selected for general discussion four ciphers, two belonging to the complete-unit type and two to the irregular. A careful study of the methods used in individual cases should furnish the student with a basis for analyzing other ciphers and evolving other special methods to suit particular cases.
Concerning the paper work, which, admittedly, is onerous in most forms of cipher investigation, much reference may be found, in the matter which follows, to “paper strips.” These are old stand-bys. Most decryptors prefer to do all of their work on cross-section (quadrille) paper, since the writing of the letters into cells enables them to obtain an accurate spacing both laterally and vertically, and this paper is easily cut apart along the separating lines. But for the kind of cryptograms we are likely to see here, many persons prefer to work with a set of anagram blocks. These can be prepared at home from cardboard squares, or may be bought in sets with frequent letters represented in approximately the correct proportions.