The present article lists some telegraph codes, with a focus on formative years of such commercial codes.
In the following, colors are used to indicate some specific topics about codebooks such as telegraph regulations (see another article for details of the development), superencipherment (manipulation of code groups to provide secrecy), representation of numbers, two-letter differential, termination orders or mutilation tables (see another article), and error checking.
Codebooks before Electric Telegraphy
Earliest Telegraph Codes: 1845-1869
Some Three-Letter Codes: 1874-1880
Some Cotton Codes: 1871-1872
ABC Code: 1873-1936
Meyer's Codes: 1871-1888
Ager's Codes: 1877-1895
Hartfield's Codes: 1877-
Harvey's Codes: 1878-1899
Codebooks in Latin
Codebooks in Print in 1888
Lieber's and Western Union Codes
Whitelaw's Telegraph Cyphers
Peycke's Codes: 1897-1918
Codebooks in Print in 1911
World War I and Thereafter: Golden Age of Commercial Codes
Three-Letter Codes in the 1930s
World War II and Thereafter
References and Further Sources
There were codebooks before electric telegraphy came into being. Besides diplomatic secret codes, there were codes for naval signalling and optical telegraphy (semaphore).
(Google (1803); See also here) This is the code used by Horatio Nelson, when he encoded the famous message "England (253) expects (269) that (863) every (261) man (471) will (958) do (220) his (370) d(4)-u(21)-t(19)-y(4)." before the historic battle of Trafalgar in 1805.
This small book was not for naval signalling or large scale optical telegraphy but for private persons. It listed words and syllables in alphabetic order, which were to be numbered by the correspondents. By omitting one number in every ten or so, the correspondents could have their own numbering. (Kahn p.192)
An interesting page here has images of some actual postcards written in cipher (apparently from the early 20th century).
Modern optical telegraphy was first put into practice in the 1790s in France.
Extant decoding sheets from 1794 shows that, at first, variable numbers of one to four symbols (i.e., semaphore patterns) indicated words etc. (Hier et Aujourd'hui No.11, p.142-143 (p.10-11 in pdf); Le télégraphes Chappe). The symbol scheme underwent several changes. The telegraphic correspondence between Paris and Lille was decoded with the codebook of Claude Chappe from 30 April 1794 to 24 October 1794, and thereafter, with the "Vocabulaire du Comité de Salut Public" (partial reconstruction on sale here) until 1803, when a new codebook ("vocabulaire") was created with 92 symbols (Yoichi Uriu, "Appropriation révolutionnaire de la télégraphie aérienne en 1795", p.9-10 (pdf)).
The new vocabulary included 8464 entries (92 pages x 92 lines=8464 combinations). (Wikipedia in French has a photo of two pages).
In addition to the original vocabulary of words, the Chappe brothers prepared a phrase vocabulary, which required an additional symbol to indicate the phrase vocabulary as well as two symbols for the page and the phrase. Further, a third, geographic vocabulary had to be created. Guyot says this numerical system had not been used or proposed before Chappe. In 1830, the telegraph administration prepared a new extended vocabulary based on a 61952-word work prepared by Chappe. (Jules Guyot, De la Télégraphie de jour et de nuit (Google) (1840), p.61-62).
Lieutenant Colonel John Macdonald (1759-1831) took interest in telegraphy immediately upon his return to England after his sixteen-year service in India (p.17, Memoir of Lieut.-Col. John Macdonald (1831) (Google), a biography by a friend).
In 1808, he published A Treatise on Telegraphic communication (Google, Google), in which he proposed an improvement of the telegraphic system. It was substantially the same as what had been rejected by the Admiralty in 1806 (p.xxi). He believed the current alphabetical system of transmitting letter by letter was inefficient and proposed a numerical system, whereby words, names, phrases, etc. were to be transmitted as figures 1-999 under various classes. (Apparently, he did not know of the vocabulary of Chappe, which was kept secret.) He submitted that such a system could be effected by modifying the six-board telegraph in 3-by-2 arrangement (see Fig. 1 after p.52) to 4-by-3 (see Fig. 2 after p.106), in which the three vertical columns each would represent a figure. (In the British board-telegraph (as opposed to the French semaphore), each board was held horizontally ("up") by default and information was represented by combination of boards pulled "down" to show its face.)
He presented a specimen of a dictionary to implement his numerical system. A class of small and frequently occurring words (p.67-77) also includes letters a-z assigned to 1-26 in order to spell words not in the dictionary.
He made a point of including every inflected variant (e.g., Sail 673, Sails 674, Sailed 675, Sailing 676) as well as elements for compound tense (e.g., Would 978, Would be 979, Would have 980, Would have been 981, Would have had 982).
In addition to the above basic set, specimens of common words are given, in which Class I was to contain 999 words beginning with A, Class II was to contain 999 words beginning with B, and words beginning with C might form two classes. (p.79-86) Additional classes dedicated for Navy terms, names of officers, place names, etc. are also contemplated (p.64).
It was also pointed out that a secret key might be used to alter the numbers transmitted.
In 1817, Macdonald published A Treatise Explanatory of a New System of Naval, Military and Political Telegraphic Communication of General Application: In which a Comprehensive Numerical Dictionary ... is Applied ... to Different Descriptions of Terrestrial and Nocturnal Telegraphs (Google). Having had submitted a complete telegraphic dictionary in 1809 (p.8) and composed a second, including syllables as well as letters, around 1811, he published an enlarged third edition in this volume and made further proposals on the telegraph system in view of the advancement of the science of communication by fixed and portable military telegraphs. Further, when Popham had revised the numerical part of his codebook for marine telegraphic communication and granted Macdonald's wish to see it, he had felt it conformed to the spirit of his previous treatise.
The dictionary, titled A Naval, Military and Political Telegraphic Dictionary, Numerically Arranged on a Very Comprehensive Scale, and Calculated to Express, in the Most Easy and Obvious Manner, All the Simple, Compound, and Potential Inflections of the Verb ...., is included in the volume. Macdonald tried to include every likely combinations of words in the dictionary: "it became necessary to compare such words with every word in a common Dictionary, selecting such combinations as struck the mind's eye to be of common recurrence and use." (p.9). For example, phrases beginning with "how" include "How are" (299), "How are they" (300), "How are you" (301), ..., "How would your" (471) in Class 46. Class XLII for naval sentences of service includes "What Telegraphic Dictionary do you use?" (19), "Sir Home Popham's" (20), "Colonel Macdonald's" (21), "We use the Military Hand-Flag Telegraph" (22).
The first set of classes (1-99) (the last entry "Then so") is followed by the second set of classes (X-LII) (the first entry "Then sorry"). Signalling the distinction is explained after Class 99 for the Navy, the Army, and the board telegraph for land service. "Aaron 119" in Class XL begins the appendix for classified entries.
A circumstantial and explanatory account of experiments ... (1819) (Google) about various military technology includes an extensive appendex titled "An Exposé of the Present State of Telegraphic Communication in This Country" (p.121).
Some short arguments and plain facts shewing that the civilization and instruction of the natives of India furnish the surest means of upholding the stability of our Oriental empire and of the introduction and speedy progress of Christianity (1820) (Google) includes "A Simple Alphabetic Cipher-Table for Secret Correspondence" (p.47-49), which is little more than the simplest Caesar substitution.
A treatise on telegraphic communication by day and night, for naval, military, and commercial purposes (Google) by Joseph Conolly includes an "inflected dictionary" for use in telegraphy. This simply gives various inflected endings in one entry to help decoding (p.31-32) (e.g., 28 for "Abolish-es-ed-ing-ion-ment"). Such a feature is also seen in the Drake-Nelson Code of 1796 (see another article). The entries run from 1 for "Aback" to 10935 for "Zoology."
His later work, An essay on universal telegraphic communication (1817) (Google) discusses secret communication (p.4-5).
(Gallica; 1821 edition)
(10th edition (1847) at Google) This is the first general system of signalling for merchant vessels, as opposed to naval signalling system (Wikipedia). Later editions titled The Universal Code of Signals for the Mercantile Marine of All Nations include those of 1854, 1858, 1864, 1866, 1869.
In Boston in 1833, this was used with two other codebooks: an extensive Appendix to Elford's Book (1832) by Parker (Google, Hathi Trust) and the Boston Harbour Signal Book (cf. an 1848 publication at Google). They assign numbers to questions and answers likely to occur between vessels at sea or at the telegraph stations on the coasts or names of vessels, countries, ports, etc. The book used was specified by an indicator, i.e., two positions (out of six) of a small upper arm of the semaphore. The combination 6-4 indicates Elford's Book, 6-5 indicates the Boston Harbour Signal Book, and other combinations referred to subdivisions of Parker's Appendix. (A Lecture on Telegraphic Language (Google) p.16-21)
Telegraph codes for electric telegraphy appeared right after the service started in 1844 in America. Early telegraph codes mentioned use in signalling or mail in the title or the direction to the user.
(Google, Austrian National Library)
The nature of this codebook, seemingly dated too early, need yet to be confirmed. Catalogs date this in 1830. Indeed, an electromagnet powerful enough to operate a telegraph had been invented shortly before this (Wikipedia) but eletric telegraphy was not in use at this time. From the names of offices in the address section, at least it seems certain that this codebook was before 1848.
The volume is divided into I (grammatic part) and II (lexicon part). Part I includes addresses, small words and grammatical words ("ach", "als", ..., "alle", "allem", ...), verb conjugations ("haben", "zu haben", "um zu haben", ...), which appear to be represented by combinations of section numbers 000, 222, 333, ..., 999 and line numbers 044, 045, ..., 990. Part II includes letters, syllables ("ai", "ao", "abge", ...) and words ("Aal", "Abandoniren, -ung", ...), which appear to be represented by combinations of section numbers 381, 382, ..., 638 (one section covering two pages) and line numbers 101, 102, ..., 378 (with 444, 555, etc. reserved in Part I being skipped).
(Google) This is the first code directed to electric telegraphy and published by Morse's lawyer and promotional agent (Kahn p.189, Friedman p.7). It replaces a word with its initial letter followed by a number. Thus, "abstract" and "abstracted" are turned into "a.247" and "a.248", respectively, while "barefooted" is replaced with "b.247". In order to secure secrecy from even those who possess a copy of the book, it proposes to add/subtract a certain number prearranged between the correspondents to/from the code number.
(Google) This associates a word or a phrase with a number and also proposes addition/subtraction of a predetermined number for secrecy. Translation of a phrase such as "Has declined accepting" into a single number "275" reduces telegram fees. The book is also designed for maritime signaling and the words/phrases may also be represented by letter codes (e.g., "JV") for flag signals.
(cotton codes, catatan) This is a so-called cotton code, a code adapted to cotton trade. It combines a root word with prefixes ("Re", "Pre", "Un") and terminuses ("or", "able") to pack information, represented by the root, prefix, and terminus, into one "word."
(Internet Archive) This employed code words rather than figures. The code words were numbered and it was proposed that a code word with a terminus "s" or "es" represents that number. Further, it proposed error checking by writing the number after the code word for a very important phrase.
(catatan) This employed code words rather than figures.
(Google) This assigns code numbers 1-8466 to phrases and sentences.
(Google) This assigns five-figure code numbers to words and syllables. For secrecy, it proposes addition/subtraction of a prearranged number to/from the code numbers as well as prearranged transposition of digits in code numbers. It also describes "la méthode japonaise", whereby code numbers (after addition/subtraction) are written in vertical columns instead of in horizontal rows (to be transmitted row by row).
(Google, jmcvey) This lists numbers, letters and syllables, words, place names, and a few phrases but the column for code groups is left blank, apparently supposed to be filled in with numbers by the user.
This employed English dictionary words as code words (Friedman, p.8). Listed in the Library of Congress Online Catalog as well as in Catalogue of the library of the Minnesota historical society (1888) (Internet Archive). LANAKI (incorrectly citing as "Brewell") says it had "a fairly complete vocabulary, arranged under captions."
(Google) This assigns four-figure code numbers to words. The appendix reproduces the (first) International Telegraph Convention of 1865. The second edition (1869) (Gallica) has an appendix of the 1869 Convention. A version (probably misdated) referring to the 1872 Convention is mentioned here
According to the preface of 1867, since the figures were counted at a rate of five figures to a word, if the address and the signature only contained 6 words, there would remain 14 words or 14x5=70 figures that could be transmitted without additional charge, which could be exploited to pack at least 17 four-figure code words.
For secrecy, it proposes transposition of figures, addition/subtraction of figures, and combination of both and explains the schemes at great length. Such ingenious schemes were reviewed with wonder in "La cryptographie militaire" (1883).
This is said to have used code words such as CABUHUC or FEDIXIB, where each syllable had a meaning (Friedman p.24).
(Bayerische StaatsBibliothek; see also here or archive) This assigns four-figure code numbers. Each page contains words and phrases numbered 00-99 and a blank to fill in a page number on the top. Thus, users may assign their own page numbering. Superencipherment with addition or transposition of digits is also proposed.
This codebook sold widely (Kahn p.839), which led to production of later editions including those of 1872 (2nd), 1879 (4th), 1882 (5th), and 1883 (6th). It was used by Boulanger (Wikipedia) some time after the Schaebele Affair (1887) (Bazeries, Les chiffres secrets dévoilés, p.111).
The author explains that since word counting is based on dividing the total number of figures by five (i.e., five figures per one word), 25 code groups (100 figures) would be counted as 20 words.
(Google) This provides numerical codes for letters ("100" for "A"), bigrams ("112" for "Af"), words ("0010" for "Advise"), and phrases ("0011" for "Advise by mail"; "1995" for "They would not have had"; "00000" for "Afraid"; "01399" for "Will shortly leave for").
(University of Oxford) The pages look like those of Sittler (1868) above, with numbers 1-65 on top (each number appearing on two pages) and numbers 1-40 and 41-80 assigned to (English) words arranged alphabetically.
The titles says this is to represent numbers with words rather than to represent words with numbers but then numbers ending in 81-99 cannot be represented.
This is a work by an author who served in the US War Department Telegraph Office during the Civil War (Weber1993 p.209).
(Google; 1888 Edition at Internet Archive) This provides five-digit numbers for words from A (00001) to Zouave (24000) and for names (in several sections) from Aaron (24001) to Zurich (25000).
For privacy, it is proposed to manipulate the numbers (by not only additions and subtractions but also various transpositions and regrouping) and re-translate the resulting numbers back into words. When this procedure is followed, the transmitted message is composed of words, not of numbers. Thus, messages "can be sent either in the form of unconnected words or of series of figures" (p.iv). (The above procedure may result in use of long code words such as Misinterpretation (14454), which was within the "seven syllable" maximum length for one word applicable as of 1870 but must have been counted as two words as of 1888.)
This codebook was also printed in 1906 (Telegraphic, Signals & Cipher Code Books), 1916 (Google), 1923 (Telegraphic, Signals & Cipher Code Books) and lasted until at least 1938 (Bellovin p.15).
Slater's code appears to have been widely used. The US War Department code dated 1885 but distributed in 1886, compiled under the direction of Lt. Col. J. F. Gregory to replace rather belatedly the route transposition used since the Civil War, was little changed from Slater's code (Weber1993 p.98, p.212). When the US presidential election of 1876 left 20 votes in dispute which would tip the balance to either Tilden or Hayse and agents of both parties were sent out to secure as many disputed votes as possible (ibid. p.169, 170), Tilden's nephew, to whom was addressed a number of telegrams mentioning dubious machinations (ibid. p.174-175) and encoded with The Household English Dictionary (1876, London) (a word being replaced with the word at the corresponding position in the eighth column back) (ibid. p.173; p.150 in the third edition available on line as of March 2014), purchased numerous copies of Slater's Code (ibid. p.177). (By a political compromise, Hayse acquired the disputed 20 votes and was declared president in early March 1877.)
Similarly directed for secrecy rather than economy was The Simplex Cryptograph (1902, Providence, R.I.) (HathiTrust), which lists words from "00001 a" to "52999 zouaves". Allegedly, the code conforms with the New Official Vocabulary published in 1899-1901 but, as with Slater's, words longer than then-allowable 10 letters would appear in the encoded text.
(code scans: telegraphic codes; Notes on books p.57, p.96)
This provides word codes and letter codes besides numeral codes as in its predecessor Bolton's Patent Code (1868). These are described as follows:
Word Code, by which common words, and pre-arranged sentences, are expressed by one type word, which type words have been selected from the best English Dictionaries, Directories and Gazetteers.
Letter Code, by which the same words and sentences are expressed by groups of letters, which never exceed 4 in number, and which have been selected in such a manner as in no case to form a word.
Number Code, which equally expresses the above-mentioned words and sentences, by the application of the ten numerals, on a system of page and line, forming the Code Signal.
Thus, the word "Hesitated" may be represented as either a code word "Hazardries", a letter code "JBHB", or a number code "48100" and the phrase "Hesitated to buy without orders" may be represented as "Hazardry", "JBHC", or "48101."
Bolton noted in his preface his belief in superiority of numeral codes.
This is listed in the Library of Congress Online Catalog. WorldCat lists Pacific Cryptograph: for the use of operators in mining stocks, mining superintendents, bankers and brokers (2nd ed. 1874 and 3rd ed. 1878).
(1886 reprint at Google) This assigns code words ("cipher") and serial numbers to words and phrases: "4 Abase: About", "5 Abash: About the", "7414 Qualify: Your letter is received". "7427 Quito" to "7700 Roger" are left blank. (Code words "Abase" and "Abash" are a one-letter difference pair.)
Serial numbers are intended for superencipherment and only "cipher words" were supposed to be telegraphed because "These words are more accurate and less expensive than figures in telegraphing." though the author admits the numbers "may be used on postal cards."
The author proposes two interesting schemes for secrecy besides simple addition/subtraction of a prearranged number ("counting a certain number of words forward or backward" in selecting a code word) and simple transposition of words.
One is "a holocryptic cipher", which involves addition of a prearranged number, which varies from word to word. Thus, "to the first word may be added 42; to the second, 17; from the third may be deducted 71, etc." This is similar to stream cipher today.
The other is "double index", whereby a user changes association between plaintext words/phrases and code words by filling in blank columns. By default, a phrase "Buy at seller's option" is given a number "2175" and a code word "Dispossess" but the user may fill "2228" in the first blank column, which tells that the phrase should be encoded into a code word opposite the number "2228", which is "Doctor." For decoding, the second blank column opposite the code word "Doctor" should be filled in with "2175".
At least the word "holocryptic" may have influenced the Red Code of 1876 of the US Department of State, which adopted various new features of commercial codes (see another article).
(Google) Various code words (English dictionary words) are provided for expressions used in banking. For encoding numbers (including consecutive numbers up to 585), two code words are assigned to each number, representing "Pounds or Cardinal Numbers" and "Foreign Monies or Ordinal Numbers", respectively. When money is in question, the author says it may be found desirable to use both these words for accuracy's sake. (If this plan is adopted, it must of course be previously mutually agreed, e.g., that the first word shall specify the currency.) (p.12) Shillings and pence are assigned words relating to animals (p.68-69) and calender dates are assigned proper names (p.70-71).
For secrecy, the author proposes a substitution scheme, in which the letters of the alphabet is arranged in a matrix of 3 rows and 9 columns and is represented by the row and column numbers. To indicate that a series of figures should be interpreted with this substitution scheme, a code word "Cadmus" should precede the figures. (p.13) (Apparently, disadvantages of transmitting figures are not recognized.)
While Slater's Code of 1870 did not include phrases and was intended solely for secrecy, economy is also featured in this work.
Moreover, the author makes a point of authenticity in the Introduction and proposes several schemes for preventing a fraud by a dishonest employee of a business firm (p.3-10). They range from the recipient's requesting repetition to inclusion in a telegram of a word indicating the number of words and/or the date of transmission in some prearranged way. When payment of money is involved, a further precaution of including one or two extra words representing the amount to be paid is proposed. In order to represent the amount, the author proposes assigning consonants to numbers (e.g., 1: B/P, 7: D/R, 0: G/N/T) and translating a number to a word by inserting arbitrary vowels. For example, 17,000 may be represented as "abrogation" or "pregnate".
(Google; Internet Archive; a 1886 edition listed at Library of Congress Online Catalog)
While the date of the first edition could not be identified, this is an old-style codebook similar to those in this period, assigning numbers and code words to words and phrases: "1 Aam: Abandon", "4 Ab: I (we) shall be compelled to abandon.", "24881 Zebra: Degrees below zero, Centigrade." Addenda consists of blank numbers and code words "24899 Abettal" to "27240: Decimate" (some of which are filled with dollars and cents). Entries such as "Ab", "Abb", "Abba", "Abbe" indicate not very careful selection of code words. The ten-letter maximum introduced for extra-European messages in 1875 and for any code word in 1885 appears to be observed.
The remarks in the preface for secrecy also remind one of codebooks in the early 1870s: "With regard to secrecy in the use of the code, this object can be reached easily, ordinarily by the methods generally employed in the use of all codes. Two or more correspondents can arrange to read so many lines up or down or so many pages backward or forward. Many ways can be devised to vary from the ordinary method of sending and receiving. While these methods would ordinarily be sufficient, in the event of absolute secrecy being important the use of the duplicate cypher column* would be the only certain safeguard." According to a footnote, "The duplicate cypher column is a patented design owned by the Anglo-American Code & Cypher Co., and consists of a duplicate column which can be altered by shifting the relative position of the words and thus making a purely arbitrary cypher between two or more correspondents ".
This uses English code words such as "Dawn" (If sold buy back), "Dasher" (If unsold on receipt of your answer), "Darken" (If you can), etc.
WorldCat lists tens of codes related to bankers/brokers. Some of the earliest are as follows:
Additions to telegraphic cypher with London & San Francisco Bank, L'd (1870, manuscript)
Wm. T. Coleman & Co.'s telegraphic code (1878)
Private telegraphic code from Wm. T. Coleman & Co., San Francisco (1882)
Gourdin, Matthiessen & Company telegraphic code book (ca. 1880, manuscript)
Telegraphic cipher for the use of Story & Ward, Grain Commission Merchants, New York, and their correspondents (1881)
(Google, Europeana) This is a small codebook of 16 pages, which assign codewords Abatina to Zythum to phrases arranged by subjects (Absence from Home, Appointments, Betting, ...). At the end, 24 secret code words are provided to "fill up by arrangement with friends."
This is counted as one of the important codebooks together with Sittler and Brunswick here. According to "La cryptographie militaire" (1883), this is similar to Brunswick's. This comprises 17,576 (=263) three-letter code groups (de Viaris (1893) p.45, Galland (1945)), including three-letter code groups representing a group of four figures (La cryptographie militaire). Various schemes of secrecy is described as with Brunswick's (La cryptographie militaire).
(Google) The author observes that the number of English words is about 26,500, while the 26 letters of English afford 17,576 permutations in groups of three letters from aaa to zzz, which number can be augmented to 27,000 by adding four letters to the alphabet: æ, œ, φ, and θ.
Thus, the words from the whole English vocabulary (plus some phrases) are assigned three-letter groups and serial numbers. The author notes that the three-letter groups may also be used to send any of the numbers 1-27,000. The code has an additional column listing proper names, which could also be expressed by the three-letter groups. The author considers the context would tell whether the three-letter groups should be interpreted as numbers or proper names rather than common words.
For secrecy, the book proposes independent substitution of letters for the three letters in a code group by "the four movable circles of letters attached to the board of the book." The outermost circle indicates the code letter, which was to be replaced by a letter of the first (or second or third) inner circle for the first (or second or third) letter of the code group. The position of the four circles may be memorized with a four-letter keyword.
As simpler alternatives, the three-letters of the code group may be represented by the next following letter etc. or transposed in a prearranged order.
The author, whose knowledge of cryptography relied on the article of cipher in Ree's Cyclopaedia (another article), is so confident as to say that his scheme would provide secrecy for statesmen or politicians or messengers in wartime.
This code is intended to provide economy by reducing the number of letters per word from about six in average to three (p.iii). Arbitrary sequences of letters were regarded as cipher and were counted according to the number of letters.
(Internet Archive) This is a three-letter code as with E. Erskine Scott's (1876) above. It represents words and some phrases with three letters. Under each letter A-Z (the first letter of the code word) of the alphabet, most pages provide six columns for A-F, G-L, M-R, and S-X (the second letter), each column including words/phrases for A-Z (the third letter).
Code words beginning with Z are numbers (including consecutive numbers up to 100 among others) and blanks for the user's supplement (ZQA-ZZZ). In addition, there is provided at the end of the book a table of 20 columns (A-T, representing the first letter of a code word) and 25 rows (A-Y, representing the third letter). With the second letter fixed to Z, this provides code words for consecutive numbers up to 500, which do not interfere with the other code words in the body. By varying the second letter representing a number to be added (Y=500, X=1000, W=1500, ...) or a number to be multiplied (F=1,000 etc.), a wider rage of numbers could be expressed, the distinction between the figures or words/phrases being indicated by the context.
For secrecy, a pre-designed substitution of letters is proposed (e.g., a codeword ESY is changed to FSY as in Caesar cipher).
There is a reference from 1880 to Slater's three-letter code.
Since three-letter combination of the 26 letters of the alphabet affords only 263=17,576 patterns, these available patterns had to be fully exploited to provide a vocabulary of a meaningful size. However, such a code including every possible combination is vulnerable to transmission errors because any single error in a letter renders a code word into another that has some other meaning. Unbeknown to the authors of these codes, the US Department of State Code of 1867, which employed similar (but more desultory) three-letter code groups, ended up in a disaster (see another article).
As the value of electric telegraph was recognized, there appeared telegraph codes suited for various specific subjects including ones for a fireworks firm, the mackerel industry, the sausage industry, tourists, and the press as well as private codes for use of big companies (Kahn p.839). Cotton trade was one of the most prolific subjects with respect to telegraph codes. One very early example of such a cotton code (1846) is listed above. The years 1871-1872 saw several additions to this genre.
(Google) An early edition (probably the first) of The Cotton Telegraph Code by Henry Robert Meyer appeared in 1871. The fact is mentioned here with other cotton codes published about the same time. See below for Meyer's various codes.
(Google, cotton codes) A cotton code represents various information in cotton trade with code words. For example, in this codebook, "Absorbable" represents "Number of Bales: 100, Price: 8 1/4".
This contemplated use of Bolton's Code as a supplement and had code words such as "Nitric", representing "Read the following words as per Bolton's Code." and "Nitrogen", representing "Read the following words as per Bolton's Code -- words forward." (p.29)
Noting that five figures may be transmitted as one word, it proposes using the last digit for error checking (p.25-26).
According to an example given (p.26), a transmitted message would be something like "Mercury 1, 2, 3." The code word "Mercury" represents the highest class for which it is wished to quote and its price. According to the usage of this code, the figures "1" and "2" represent the price of lower qualities by the difference in eighths. That is, "1" represents a price lower than the price indicated by "Mercury" by 1/8 and "2" represents a price lower than the price indicated by "1" by 2/8=1/4. The last digit "3" is the sum of these differences 1+2=3. With a five-figure code group, four additional quotations (plus one check digit) can be represented. (Such condensing of quoting information in a five-figure word is also explained in Meyer's Cotton Telegraph Code (33rd ed., p.v) but its 1871 edition does not appear to provide such an explanation.)
(Google, cotton codes) This also represents various information in cotton trade with code words (dictionary words and names, called "ciphers").
(Google, jmcvey) This is also for use in cotton trade. This proposes to represent information about price, quality, quantity, etc. with a four-digit number and, "in order to guard against mutilation in transmission" (p.iii), translate every number to a word with a table listing from 100 (aachen) to 9999 (tippecanac). It was clearly recognized that figures were prone to transmission errors.
There are long code words such as 4711 (mucososaccharine) but they appear to be within the "seven syllable" maximum applicable at the time.
The author also admits sending the numbers themselves if the sender is using some other codes, since the telegraph companies accepted a number containing up to five figures as one word (p.iv). The appendix provides code words (called "cyphers") tiraboschi ("1/16") to Zytomir ("The following documents were missing in your letter of the --").
The 1875 revision (Google) of this code reduced the code words to the length of ten letters or less, in compliance with the 1875 revision of the international telegraph regulations. For example, Triacontrahedral and Tshirnhausen were replaced by Triad and Tscherkask.
Watts produced International Telegraphic Code (Google) in 1880. He says the success of the 1875 code induced the compilation of this work on the same principle on a more extended scale.
Code words increased ten-fold and now ranged from "100 aalbaum" to "99999 untertasse". Additional code words "untervogt" to "verdict" are for fractions and "verdieping" to "verpligt" are for numbers up to 6,000,000 including consecutive numbers up to 100. Code words "verpligten" to "wortfolge" are left blank.
The ABC Universal Commercial Electric Telegraphic Code, known as The ABC Code, by Clauson-Thue, with its enormous vocabulary, was a huge success (Kahn p.838), which may be attested by many media reviews proudly quoted in later editions (e.g., ABC2 p.xi-xiv; ABC3 p.xi-xiv; ABC4 p.xix-xxii). It went through many editions in the decades to come. This section follows its revisions along with development of telegraph regulations (for which see another article for details).
The first edition, ABC1 (1873), (Google, Internet Archive) associated some 13,000 words and phrases with "code words" and five-figure "code numbers."
While code numbers may be sent instead of code words ("Most of the Telegraph Companies regard five ciphers as equal to one word" (p.v)), in case of a secret message, it proposes to translate each digit of a number into a letter by some prearranged substitution. For example, with a keyword "March winds", figures could be associated with letters as: 1(M) 2(A) 3(R) 4(C) 5(H) 6(W) 7(I) 8(N) 9(D) 0(S). Apparently, susceptibility of numbers (or arbitrary sequences of letters for that matter) to errors was not recognized. (This is substantially the same as what Friedman (p.24) called an "amateurish scheme" with respect to the semblance of the result to real words.)
The second edition ABC2 (1874) (Google) was revised to use exclusively words found in Webster in order to meet the requirement of the official announcement this year (1874), which stipulated that code words must be ones that could be found in dictionaries. The first part used ordinary English words and the second part, listing expressions for specal fields (stocks; companies; products and commodities; ship's gear, machinery; prices and rates; numbers (including consecutive numbers up to 500), quantities; etc.), used proper names found in Webster's appendix. ABC1 also had such a division into two parts but did not mention the origins of the code words.
Further, "to avoid as much as possible mistakes in telegraphy, no two words of closely similar construction are used, and no word is used as a Code word that is likely to be used in an ordinary commercial telegram." Thus, "Absterge" in ABC1 was replaced with "Abstergent" in ABC2 because there was a similar code word "Absterse".
ABC3 (1876) (Google) complied with the new international regulations made in 1875 that words must be 10 letters or less for extra-European communications and 15 letters or less within Europe. This was substantially a reduction from "seven syllables" up to then and a further provision was to be introduced in 1885 that code words had to be 10 letters or less. The publisher had offered a free Supplement to the previous edition including changes to comply with the new rule. ABC3 incorporated such changes. Thus, code words such as "Abstraction" (11 letters) and "Accoutrement" (12 letters) were gone. The code word "Account" was removed probably because this term may be used in plaintext.
ABC4 (1881) (Internet Archive (this particular copy has typewritten supplements from p.312 pasted over a portion of Part II)) underwent an extensive revision in response to the new rule introduced by the 1879 Regulation, which allowed code "words" to be taken from any of English, French, German, Italian, Dutch, Portuguese, Spanish and Latin for extra-European communications. (Within Europe, one and the same of these languages had to be used, which restriction, however, was lifted in 1885.) ABC4 included words of French, Spanish, etc. Added words include Ablemar, Ablette, Abnormity, Abocinar, Abodement, Aboiment, Abolir, Abolsado, Abomasum, Abonder, Abordage, Abound, Aboundeth, Abrading, Abrazar, Abrensin, Abreption, Abreuver, Abreyer, Abridgement, Abriter, Abrochar, Abromado. This allowed increasing the size of the vocabulary to as many as 25,000. The author went so far as to propose "All languages should be allowed, and any proposal to make a difference between code words and other words should be rejected." (p.iv) It would be in 1903 that any pronounceable artificial words were admitted and in 1928 (in 1909 in America) that pronounceability was no longer a requirement.
Since use of proper names was prohibited by the 1879 Regulation, Part II for special vocabularies used foreign words as well as some English words such as Axiomatic.
Considering that, for European communications, all the code words in a message had to belong to one and the same language (until the 1885 revision), it was clear that the codebook was intended for use in communications with outside Europe. It was such communications, which used expensive submarine cable, that required codebooks for cost reduction much more than communications within Europe. (One observed that "code language was almost exclusively employed in the extra-European correspondence and almost not at all in the European regime." (Friedman p.26))
ABC4 adopted the recent trend as employed by Ager's Shipping Telegram Code (1877) etc. to use the numbering of the code words to represent numbers with code words and provided new entries from: "09321 Nustle: The next word to be read as a numeral" to "09331 Nutriments: The next -- words to be read as numerals".
After the 1885 revision of the regulations, Clauson-Thue published A1 Universal Commercial Electric Telegraphic Code (1887, 1888). This tripled the number of entries to almost 88,000 (jmcvey). British Books in Print for 1888 (Google) lists A1 at a price of 25s. and ABC at 15s. Every word had a superscript indicating the language from which it was taken: 1 (French), 2 (German), 3 (Italian), 4 (Portuguese), 5 (Spanish). This could show authenticity of these words against any claim of being artificial.
ABC5 (1901) (Internet Archive) appears to have further expanded the vocabulary. ABC5 lists as many as 42 phrases from "00001 Abagangay: Can you abandon" to "00042 Abalorios: Had to abandon her on account of terrific weather" under "00000 Aavora: Abandon", while ABC4 had only eleven. The total number of entries was now about 103,000 (the last serial number being "103084") but there was not large change from A1 (jmcvey).
Substantially the same means for secrecy as in ABC1 is explained.
ABC did not follow the worldwide trend of using artificial five-letter code words, which was allowed by the 1903 Regulation, for more than ten years. The inactivity may be because the author William Clauson-Thue was in his late years (to die in 1907).
A five-letter edition (1915) of ABC5 "improved by the addiiton of five-letter codewords opposite the full codewords and Part III" is published by American Code Company.
ABC6 (1920) (jmcvey), now crediting "Edith Mary Maria Clauson-Thue, Florence Charlotte Elizabeth Faulkner" in addition to William Clauson-Thue, caught up with the trend. It listed more than fifty phrases such as "00007 ABAJK Abandon altogether", "00008 ABAKL Abandon for the present", etc. under "00006 ABAIJ Abandon". As was now common, such artificial code words were chosen such that no two code words differed only in a single letter (the idea called "two-letter differential"). In addition, a great effort was made to avoid code words prone to transmutation of two-letter Morse groups. (More than twenty pages were spared for listing two-letter Morse similarities.) Further, it avoided use of five-letter code words having a commercial meaning in English, French, German, Portuguese, Spanish, Italian, Dutch, Swedish, Danish, and Norwegian or coinciding with five-letter names of major ports and places of the World.
While ABC5, taking after A1, had Part I (vocabulary) with 44,794 entries and Part II (stocks etc.) with 55,770, ABC6 reverted to the one-part structure as in ABC1-ABC3, with 88,370 entries.
As a means for error correction, it had 142 pages of terminational order ("terminal index"), which listed code words in the order of backward spelling. This was for finding a correct code word when an early letter in the code words was in error.
ABC7 (1936) (jmcvey) continued to use five-letter code words. By the 1932 Regulation (effective from 1934), pronounceability was no longer required, while the new maximum length of five letters precluded combining two code words into one ten-letter group. Code words such as MEMYJ, MENBC, MEOFV may owe their existence to this change (need to check ABC6).
ABC7 was again divided into two parts: Part I (Shipping and Insurance) with 53,168 entries and Part II (Trade and General Phrases) with 53,269. The first page of Part II includes code words such as "53215 MEOND Abandon", "53216 MEOPF Abandon altogether", etc. The new version allowed use with earlier editions and it included code words such as "53175 MELKU Please use "ABC" Codes 6th & 7th Editions" and "53184 MEMEP Using "ABC" Telegraphic Code 5th Edition."
It had a mutilation table, a more advanced device for error correction than the simple terminal index of ABC6 (see below).
(Prices below are as of British Books in Print for 1888, unless otherwise indicated.)
The 1871 edition, probably the first, is found at Google. So is the 1873 edition (Google). For later editions, see below.
Henry Robert Meyer's codebooks started with this cotton code, which was to represent various information related to cotton trade with code words. The code words were English words including long ones such as "earthenware" (11 letters), "extraordinary" (13 letters), etc.
This had a table for "numeral check code to check price and number of bales." Various combinations of the number of bales and the fractional part of the price are assigned numbers (called "cyphers") 001 to 320. When it is desired to corroborate an order for 250 bales @9 1/16, the number of bales ("250") and the fraction ("1/16") are used as indexes to the table to find a number "025", which is placed after the integer part of the cents or pence ("9"). Thus, in this example, a checking number "9025" is used. While use of this check code was optional, it could provide confirmation by taking advantage of the trans-Atlantic tariff of charging five figures as one word. This is particularly mentioned here to be compared with the check sum method and encoding of numerals in the subsequent codes.
(Google) Price 10s. 6d. The code words were 15 letters or less. Thus, there were code words such as "cheerlessness" (13 letters). The same code words from The Cotton Telegraph Code (apparently not the 1871 or 1873 editions) were used (p.iii).
The code words appear to be English words, which were "carefully selected; compound words, which some Cable Companies charge as two or more words, are, with one or two exceptions, not used; and due regard has been given to prevent any similarity of different ciphers." Since its code words of length 11-15 letters counted as two words by the 1875 Regulation, which limited the length of one word to 15 characters within Europe and 10 characters for extra-European messages, this was designated as "for use with the Continent" (i.e., within Europe) in an advertisement page of The Liverpool Cotton Telegraph Code (1882)..
The author proposes a "checking number", the sum of numbers 1-26 corresponding to the letters of an encoded message. The author observes that figures were "very liable to be telegraphed incorrectly" and recommends sending the checking number by a code word corresponding to the number. In a later work (Numeral Code (1874) p.16), the author says this checking number method was used for the first time in this codebook.
(Google) Price 25s. This contained 15,000 sentences for business telegrams. The code words (called "ciphers") were 10 letters or less in compliance with the 1875 Regulation and were all English words found in Walker's Dictionary.
The check sum is explained as in The General Telegraph Code.
Two-letter differential was not enforced and code words with one-letter difference were included (e.g., minding/winding, detraction/retraction, designed/resigned, deservedly/reservedly).
(Google) Price 25s. This was re-compiled edition of the International Telegraph Code to suit the 1879 Regulation. The code words were 10 letters or less.
The author says the new regulation prohibited use of proper names "used adjectively" ("adjectively" is not found in Article VIII of the Regulation; see a quotation in another article), which, to his regrets, compelled him to issue this new edition, "without any other changes than the needful alterations" in a small number of now inadmissible code words. Thus, code words such as "babel", "babylonian", "babylonish" were gone and "babish", "baccated", "backing" were introduced instead.
The author explained that the United States was not a party to the International Telegraph Convention and explained for such countries "there need be no change whatever; and, therefore the old edition of the "International Code" will remain in print just as heretofore. In order that there may be no confusion between the old and new editions, it has been decided to publish the latter under the name of the "Commercial Telegraph Code," and print on toned paper, with a distinctive binding, so that even where the two books are in use in the same office simultaneously, no risk of confounding one with the other need to be apprehended."
The check sum is explained as in The General Telegraph Code.
The Cotton Telegraph Code went through many "editions". The 33rd edition of 1878 (Google), which was issued when the 32nd edition became "almost exhausted", was "entirely remodelled" to suit modern requirements. The code words were now ten letters or less "partly with a view to making it applicable to such countries as are included in the St. Petersburg International Congress, and partly in order to anticipate the adoption of the 10 letter rule by the Atlantic Telegraph Companies (as no doubt will be the case at some future day when the traffic increases)". With a view that it was common that code words were rejected by telegraph companies, the author boasted that the code words were the same as were used in The International Telegraph Code, "very large numbers of which have been now three years in use in all parts of the world, without a single complaint having reached the author respecting them."
Compared to the 1871 edition, among others, this 1878 edition includes a section of numerals for encoding consecutive numbers 1 ("smockfaced") to 100 ("soles") as well as numbers up to 1,000 ("sonata") in 25 intervals and some round numbers up to 5,000,000 ("sovereign"). The General Telegraph Code (1874) and The International Telegraph Code (1875) had a more extensive numeral section.
The check sum is explained as in The General and The International. In these works, the author recognizes that figures were "very liable to be telegraphed incorrectly" and recommended sending the checking number by a code word corresponding to the number. But the numeral section of The Cotton Telegraph Code of the 33rd edition may be insufficient for that purpose.
Concurrently with the 33rd, the 34th edition was sold, which was the same as the 33rd (which was for American cotton trade) except that "the qualities are left blank for adapting it according to the purchaser's requirements to the Indian, Brazil and Egyptian Cotton Trades." In these editions, about 2,500 code words in "General Phrases" were left blank for definition by the user. The General Phrases in the International was available unbound at 5s. per copy. (Advertisement in the front matter of the 33rd edition.)
The 33rd edition declared it would remain henceforth permanently in type and no alteration would be made in it, unless changes in the mode of doing business in Cotton or alterations in the telegraph companies regulations should render it unsuitable. The author's mortification was imaginable when he had to issue the 35th edition as The Anglo-American Cotton Telegraph Code (price 42s.). Probably, as with the change from International to Commercial, code words of the 33rd edition such as "babel", "babylonian", etc. were replaced. It was also sold in a form (price 50s.) containing 24 "Tables of Cyphers" with quantity, quality, and price combined. for cabling firm offers and orders in one word.
This code went through further editions under the title of Meyer's "Atlantic" Cotton Code (37th ed., 1895, and 38th ed., 1901, being listed at Google; an image is found at jmcvey). The 39th edition was one of the few codes allowed under the US government's censorship during World War I (see below) and the 40th edition was listed in Bentley's Second (1929) (see below).
(Google, Internet Archive) Edition A contained tables with 80,000 code words. "Two complete firm offers or orders to buy or sell Cotton can be transmitted, giving price, quantity, quality, and terms, in one word!" This Code could be used either alone or with the Cotton Telegraph Code (the 33rd edition), the code words being carefully selected not to clash. Edition B contained tables with 32nds and 64ths and the headings left blank. Both editions were priced at £5 5s. per copy.
(Listed as The 'Numeral code' supplement to the 'Cotton telegraph code' at Google) Price 20s.
This is a supplement to The Cotton Telegraph Code and allowed transmission of orders, firm offers, advice of purchase and sale, etc., with price, quantity, quality, mode of shipment, and the market in two words. This uses numerals as code groups.
The author, who had believed figures were liable to transmission errors, explains the motivation for this code as follows: "The compiler has until now, considered the use of numeral messages as dangerous, in consequence of the disastrous effects of a possible error in the transmission; but enquiries from firms who use numeral codes, and from the telegraph authorities, would seem to prove that there are as few errors in the transmission of figures as of letters." The author declares "The great objection to the user of figures for Telegrams, viz., the liability to error and the difficulty in detecting it, is here overcome; a simple but thoroughly efficient check being provided without any increase in the length of the message." On the other hand, he admits "Only in certain cases will this system present advantages over the ordinary method of cabling by the "Cotton Telegraph Code", still it is probable that it will effect great economy."
Trade information is expressed in two five-figure code numbers. The last two figures in the second code number are the sum of the preceding eight figures for checking.
This was included in the advertisement page of The Cotton Telegraph Code (1878) and The Appendix Telegraph Code (1880) but not in that of The Globe Commercial Telegraph Code (1882) or The Liverpool Cotton Telegraph Code (1882). Nor is it listed in British Books in Print for 1888. The author's piquant remark in the Globe seems significant: "all figure and other complicated combinations are dangerous in the extreme."
(Google) Price 25s. This contains blank tables and blank cyphers for firms to suit to their own requirements. This would be obtainable either separately bound or also bound in at the end of the International, the Commercial, and the Cotton Telegraph Code (33rd ed.) (notice in the Commercial). In order to be compatible with the other codes, the code words were German words.
Since the code words were numbered, this could be used to encode numbers up to 0001-4379.
(Google) Price 42s. This contains a large number of code words (with ten letters or less, taken from French, Spanish, Latin, Italian, Portuguese, and Dutch) in tabular form for 60 different articles with blank headings. This also includes a second part with English code words, which was identical with the Commercial Telegraph Code.
Price 30s., this was a code with all the phrases in French.
Price 30s. This was a translation of the above.
Price 30s. This was a translation of the above.
This consists of 20,000 phrases. Price 35s. (Principles of Banking (1887) (Internet Archive)). Listed as The Grain Trade Standard Code in British Books in Print for 1888.
(Prices below are as of British Books in Print for 1888, unless otherwise indicated.)
(Google) This assigns a code word and a five-figure number to words and phrases. The code words were taken from the author's Telegram Code with revision and also supplemented by other code words of nine or ten letters as well as by some of the more common Latin words to provide the required number of code words.
The author's remarks on the prospect of the telegraph conference are interesting: "As it is anticipated that a future "Telegraph Conference" may reduce the facilities afforded for employing "figures" in Telegrams, the author has given special attention to this point and flatters himself that he has succeeded in suggesting a method that would not only obviate any inconvenience such a change would entail but one also that will be conducive to economy." For example, a phrase "Report the arrival of the -- as soon as she reaches --." is given a code word "Animation" as well as a number "00777", the number 777 may also be represented by "Animation", since usually the context indicates whether the code word represents the figure (called "cypher") or the phrase. For accuracy, some of the code words are provided with an alternative form marked "(C)". If such an alternative form is used, the next code word should be interpreted to be a figure.
(Internet Archive) (1880) This contains nearly 56,000 words from the languages admitted in the 1879 Regulation. The number was nearly four times that of the first edition and more than twice that of the second. The code assigns a code word and a five-figure number to common phrases.
Some 45,000 code words down to the letter "T" did not exceed eight letters. Price was £2.15s. in 1888, £2.2s. in 1900 and 1911 (still the Third Edition).
The author says he reverted to the plan of the first edition in leaving alternate pages with blank code words throughout the greater part of the book. (The particular scan at Internet Archive is of interest in that many blanks are filled with words and phrases by the user.)
The author says "The Cyphers or figure groups accompanying every Code Word afford facilities for transmitting in one word Quantities, Prices, and Qualities, or other particulars of an Offer, Order, Contract, Purchase, Sale, &c." That is, a code word may represent not only the phrase but also the number, whose digits represent various trade information.
Ager's Standard Telegram Code of 100,000 Words (1879) compiled from the languages sanctioned at the telegraph convention. Price was £5 5s. in 1888, £3 3s. in 1900.
Ager's two-volume code, one with numbers 00000-99999 and the other with 100000 to 125000 is described in a Japanese book (1895) (Ueda p.21 ff.), which may refer to this Standard and two of the following supplements. (The book also mentions Whitelaw's code consisting of a volume with five-figure groups and another with six-figure groups.) The cited Ager's code has code groups "00000 aafsch", "00001 aakste", "00002 aalbes", ..., "99999 Iravail". The first two digits indicate the kind of commodity and the last three digits indicate quantity, price, and shipment according to a table whose rows represent combinations of quantity and shipment and columns represent price. The same authors inspired a Japanese codebook (1896) by Kobayashi Keishi, who cites Ager as comprising 200,000 entries (probably the Duplex below plus supplement) and Whitelaw as comprising more than 300,000 entries (probably the "Comprehensive Edition").
Ager's 10,250 Extra Code Words followed in alphabetical and numerical sequence those in the Standard Code. Price 15s.
Ager's Standard Supplementary Code for General Merchants had 10,250 words with sentences. Price 21s.
Ager's General Merchant's Telegram Code could be used separately or could be used as a supplement to the Telegram Code (Third Edition), no single word being contained in both. Price 31s. 6d.
The Duplex Combination Standard Code (1886, 600pp.) consisted of 150,000 words. It had a double set of figures for every word, thus "affording opportunity for each Figure System of Telegraphing to be used". The words were carefully chosen so as to avoid both literal and telegraphic similarities. Price was £6 6s. in 1888, £4 4s. in 1900.
The Extension Duplex Code of about 45,000 more Words ("just published" in 1888) was to be used in connection with the "Duplex" or for "special arrangement with the Figure System for Private Codes by agreement." Price £1 1s.
The Complete Duplex Code ("just published" in 1888) was the above two codes bound together priced at £7.7s. in 1888, £5.5s. in 1900 and 1911. It consisted of 195,000 words in alphabetical and double numerical order.
Ager's Alphabetical Code of Nearly 50,000 Sentences (1887, 527pp.), where the code words were in sequence to the 150,000 Words in the Duplex Standard Code above. Price 21s. in 1900. This could be also obtained bound up with the "Duplex or Prefix Code".
Ager's Telegraphic Primer, with Appendix consisted of about 19,000 English and 12,000 Dutch words from Webster's and Picard's dictionaries. Price 12s. 6d.
This may be an augmented version of Ager's Telegraphic Primer, Or Skeleton Telegram Code, Consisting of 16,000 Good English Telegraphic Words (1880) (Google). The latter lists English words taken from dictionaries of Webster and Nuttall, to which the users could fill in counterpart words or phrases. Each word is accompanied by a number (called a "cypher") 000-999, the digit indicating the thousands being omitted.
Ager's General and Social Code (1887, 320pp.), in which "the Code Words and Sentences (about 20,000) will be Alphabetical, not only in name but in fact." This was for travellers, brokers, bankers, and mercantile agents. Price 10s. 6d.
The Law Relating to General and Particular average (1900) (Internet Archive) lists two additional codebooks.
The Simplex Standard Telegram Code (1893) included 205,500 code words. Price £5 5s. A reference to a version of 100,000 words (1890) is also found on the Web.
The AYZ Telegram Code (1895) included nearly 30,000 sentences, prices, etc. The code words were carefully compiled from the "Official Vocabulary" (1894) by eliminating many thousand objectionable words. Price 16s.
John C. Hartfield was one of the great code compilers in America and continued to produce various codebooks, with his son from 1890 (Kahn p.839).
Listed in the Library of Congress Online Catalog. This included 15,000 dictionary words.
This contained 61,000 carefully selected words. While it is estimated that there are 160,000 candidate words of at most ten letters in dictionaries of the eight languages allowed, the condition imposed on code words that they must differ from each other by at least two letters reduced the number to this (Friedman p.21, Bauer p.76).
The following are titles listed in an advertisement in The Declaration of London (1911) (Internet Archive).
"Wall Street" Code Specially adapted for General Banking, Stock Brokerage, Arbitrage, and Exchange Business. Price 42s. Google lists Hartfield's "Wall Street" Code: Contains about 467,000 Cypher Words, All Conforming to the Telegraphic Convention Regulations (1905) and Hartfield's New "Wall Street" (Newwallst) Code: Contains 156,563 Cypher Words, Numbered from 00000 to 156562. Also Roots and Terminals Forming Millions of Artificial Words, Conforming to the Telegraphic Regulations (1912), for which see jmcvey. While roots and terminals took advantage of merging two five-letter words into one, the main vocabulary used real words (or words resembling real ones). This shows that not all the codebooks switched to five-letter code words after 1904.
Bankers', Brokers' and Stock Operators' Telegraphic Code Price 40s.
Bankers' and Brokers' Pocket Code Price 20s.
The New Leviathan Code (1896), containing 226,700 words from the Official Vocabulary. Price per copy £15.
The Leviathan Cable Code (2nd ed.), containing 120,000 carefully selected words. "Published at £10 per copy, now offered at £6 net."
Roots and Terminals, 36 Millions Price per copy £3.
Two Millions of Roots and Terminals Price per copy £2.
Atlantic and Pacific Cable Code, containing 36,370 words. Particularly adapted to a general business. Price 40s.
Alpha Beti Cal Telegraphic Cypher Code, containing 100,000 phrases arranged alphabetically. Price per copy £3.
Hartfield's 124,000 Selected Words This can be used with the Alpha Beti Cal Code. Price 47s. 6d.
South American Cable Code (Spanish Edition) Price 30s. There were also an edition in English and an edition in English and Spanish.
Central American Cable Code (No. 4) For use with Alpha Beti Cal Code by which any Two of the 100,000 Phrases or Groups can be sent by One Word. Price 30s.
Hartfield's Ten Figure Code: Ten Thousand Million Words numbered 0,000,000,000 to 9,999,999,999 with a Check on each Half Word of Five Letters. This must be a publication after Whitelaw's in 1904 (see below). Price per copy 10s., per pair 20s.
The Merchant's Code, containing 15,000 words. Suitable for Bankers, Manufacturers, etc., who desire to insert their own phrases. This must refer to the 1877 work above. Price 3s.
Henry Harvey was one of the great code compilers in America and published 21 codes or lists of code words between 1878 and 1899 (Kahn p.839). The Library of Congress Online Catalog lists the following.
My code: 2500 words. Harvey's selections, from Webster's dictionary (1873)
Prefix to my code: 4,400 words and meanings ... (1878)
International cable code for general business matters ... (1878)
Missing link of my code series: 7500 words, C to E, in tables and blanks ... for general business use (1878)
Harvey's quadraginta (1879)
Cosmopolitan cable code (1881)
Harvey's A.B.C. Domestic Code (1881), which is not found in the Library of Congress Online Catalog but is at Google. The second edition (1885) is found at Google. The second edition provides 14,000 code words of six to ten letters taken from Webster's Dictionary, with 11,000 filled and 3,000 blank. The second edition was "just a larger and handsomer re-print of the original 360 pages, with a 60-page Appendix of additional matter (including Sterling Money)."
E prefix & my code, 8500 words ... (1883)
Harvey's compact code: nine thousand code-words, selected from Webster's dictionary (1883)
Harvey's oblong pocket code; 3,250 code-words ... (1884)
Six thousand selected words, A to W ... (1884)
Harvey's ... operis corona: pure Latin code-words of not over ten letters (1886)
Harvey's stock brokers' code (1886, 1899)
Harvey's abecedary code, 1888...on the plan of alphabetical arrangement of subjects (1888)
Harvey's abstuv code. (so named because the code-words are in those letters) ... (1888)
Harvey's mining code, on the A. B. C. (1889)
Addendum to Harvey's cosmopolitan code (1890)
Harvey's vanguard code, 1892, on the A. B. C. (1892)
Harvey's second Berne-word code (1897) The "Berne-word" refers to the official vocabulary (1894) of admissible code words.
(Prices below are as of British Books in Print for 1888, unless otherwise indicated.)
This includes "basis 195,000 words." Apparently, the "basis" refers to a mere list of numbered code groups (like Whitelaw's and other figure codes), which "affords facilities for the compilation or reconstruction of any Mercantile Code in a few days at a comparatively trifling expense." Price £5 5s.
British Books in Print for 1888 lists several tools published by Hawke, all "on the figure system".
Price £2 2s. This includes basis 187,000 words and provides for "telegraphing in one word the useful combinations of 33x17x18x18 or every possible variation of 17 subjects." "Firms possessing keys of 200,000 words can extend the combination to 33x18x18x18 without any trouble."
Price £2 2s. This includes 160,000 words and provides for telegraphing in one word any four variations of 42 subjects, every possible variation of 16 subjects, four quotations with a range of 20 to each, or any combination of 20x20x20x20 or 400x400.
Price £2 2s. This includes basis 100,000 words and provides every possible variation of 15 subjects, and combinations of 40x50x50.
Price 2s.6d. This was a pocket book (46 pp.) and specially adapted to social requirements for home use and for travellers.
In 1911, the lineup of Hawke's codebooks was as follows. (Taken from The Declaration of London (1911) (Internet Archive), escept for the 1907 condenser. The prices are as listed in this book.)
(jmcvey) Containing close upon 120,000 Words (from A to M, specially selected from the Berne Official Vocabulary). Price 10s. 6d.
A 100,000 Word Supplement to the Premier Code. Words from M to Z specially selected from the Berne Official Vocabulary. "For special Tables for Offers, Buying, Selling, etc., the Five Figure System, worked in conjunction with Keys of Words, numbered from 00,000 to 99,999, and 2440 Reserve Words for Indicating or Catch Words or Special or Temporary Tables."Price 10s. 6d.
All easily pronounceable and specially arranged to fulfill the conditions of the latest International Telegraph Regulations [i.e., the 1903 revision]. Price 42s.
The 1879 reivsion of the telegraph regulations stipulated that code words must be real words taken from English, French, Latin, etc. In a way, this officially allowed use of words taken from a foreign language (albeit limited to one of the eight allowed languages). While some code compilers took advantage of the new rules to expand the vocabulary of their codebooks, others came up with codebooks in Latin.
Latin had two advantages. First, Latin code words could be used with any existing codebook with code words in English (provided that Latin words similar to some English words are avoided). Secondly, complicated inflection of Latin words meant that many distinct code words could be systematically formed from a single root form.
(Google) Price 21s. (two or more copies, 15s. each.)
The code words were all Latin verbs of the first and third conjugation. So the code could be used as a supplement to any other code that did not use such code words. One code word could represent offers or orders of any one of 56 qualities, with any combination of 10 quantities, 72 prices, and 4 shipments, selected according to the author's own requirements.
Specifically, Part I lists stems of 720 Latin verbs of the first conjugation. These represent the quantity 10-100 combined with various prices. The stems are to be followed by any one of 37x4=148 different inflection endings, which represent 37 different "quality numbers" combined with four shipment options "in two weeks", "in four weeks", etc.
Similarly, Part II lists stems numbered 721-1440 of the third conjugation, with endings representing quality numbers 38-56 combined with four shipment options.
While there were other codes which used such combinations of stems (roots) and endings (terminals), Parker's code used combinations of regular Latin verbs and their endings, which ensured that any combination was an authentic Latin verb.
Similar use of Latin verb conjugations is also seen in Whitelaw's "14,000 Latin Words" (1880) (see below), The "Nonpareil" Code (1884) (see below), and probably Verini's (see the next section).
An advertisement claims it complies with the requirements of the International Convention of 1879, which came into force 1st April 1880. Its code words are formed by combinations of Latin roots and terminals (Bentley's essay). Part I consisted of 100,000 Latin words (£2.2s) and Part II consisted of 50,000 Latin words (£1.5s) (the above-mentioned ad)
British Books in Print (Google) provides a list of codebooks in print in 1888. (A slightly earlier list is also printed in The Structure of the Wool Fibre (1885) (Internet Archive). Other lists may be found in Subject Index of the Modern Works Added to the Library of the British Museum in the Years 1885-1890 (Google) under the heading "TELEGRAPHY", The Principles of Banking (1887) (Google, Internet Archive), etc.
Such a listing may to some extent make up for the bias in my selection dictated by what I came across on the Web.
Price 21s. (two copies for 31s. 6d.; additional copies, 10s. 6d. each).
One code word represents purchases or sales, quotations, firm offers, or recommendations. Adapted to the India, China, Japan, etc. markets.
(Google) Price 30s. This may be used with Ager's, or any code providing code words for numbers up to 100,000. Adapted to the Eastern trade.
Part I is for representing various offers with two five-figure numbers (e.g., 31931 21612), which are to be translated into code words (e.g., Doophek Ciloma).
Part II is for representing various data or phrases used in Eastern trade with numbers 1 to 9999. The number found (e.g., 178) is supposed to be translated into a code word (e.g., Abarbar) by using Ager's Code.
(1889 reprint ("sixth edition"): Google, Internet Archive; a few pages from the 1902 reprint, indicating a few changes; 1912 reprint; There is also a 1920 edition (WorldCat)) Price 2s. 6d.
This used Latin words as code words: Abactus (Abandon the negotiations), Abazea (Am able to), Abdo (Have you been able to). Telegraphically similar words were avoided.
An edition of this codebook was used in a telegram of 19 June 1924 to London, reporting the death of George Mallory (Wikipedia) and his partner Andrew "Sandy" Irvine in their attempt to climb Mt. Everest (the remainder of the party came back safe) (Klausis Krypto Kolumne). The telegram message is as follows. The code group NOVE and ALCEDO can be found even in the 1889 edition.
See below. Price 80s. per one pair.
(Google) This lists no plaintext words or code words. Instead, it lists sequences (called "variations") of letters ("signs"), in alphabetical order, consisting of 1 to 15 letters taken out of 15 letters (A, B, C, ..., O). The sequences are numbered 00001-32767. (The letters in the sequence are in strictly ascending order. So, there are no sequences such as AABC or BCA. For the index system explained below, only combination of different letters taken from A-O is relevant.) This was what had been in constant use in the author's own business (between England and South America) for several years.
"The most valuable application" is said to be the index system, whereby the first (or any other prearranged) word in a telegram was to be an index word (e.g., "ACDF"), of which the first letter ("A") indicates the codebook for the first word following, the second letter ("C") the codebook for the second word, and so on. In the example given, the letters ("signs") represent subjects such as A (Orders), B (Repeats of Shipments), C (Purchases), D (Sales), E (Quotations), F (Firm Offers), etc.
Thus, instead of using a single generic codebook addressing various subjects, several codebooks, tables, or lists, each devoted to a single subject could be used. In particular, the author points out that various information for a specific subject may be conveniently expressed with figures by using tables (though there were codes which gave code words in such tables).
The index ("ACDF") could be transmitted by converting the number assigned to it ("00655") into a code word by using some ordinary code such as Whitelaw's, Ager's, ABC, Stracker's (called "Number Codes" (p.vi)), or even any private code in which a number was written against each word.
Naturally, the full length of fifteen letters would not be needed for many cases. If at most eight signs are used (i.e., numbers up to 22818), a code of 25,000 words would be sufficient. Seven signs (numbers up to 16383) could be covered by a code of 20,000 words. Six signs (numbers up to 09948) could be covered by a code of 10,000 words.
To check the accuracy of the index word, the author "found it in practice amply sufficient" to include in another part of the telegram the last digit of the sum of the digits in the index number.
An appendix explains a method of using a table to pack information into one number, which is substantially a base-32 system. When three groups of options numbered 1-32 are given, a number X selected from a first group of options for a specific subject adds (X-1), a number Y selected from a second group adds 32*(Y-1), and a number Z selected from a third group adds 32*32*(Z-1). Thus, a combination (X,Y,Z) can be uniquely represented by a number 1024*(Z-1)+32*(Y-1)+(X-1)+1.
An interesting variant allows each group to have a different range from each other. Thus, a number X selected from a first group of 1-10 adds (X-1), a number Y selected from a second group of 1-8 adds 10*(Y-1), a number Z selected from a third group of 1-11 adds 70*(Z-1), and a number W selected from a fourth group of 1-14 adds 700*(W-1). Thus, a combination (X,Y,Z,W) can be represented by a number 700*(W-1)+70*(Z-1)+10*(Y-1)+(X-1)+1.
Price 1s. WorldCat and/or Google list Henry Sell's Telegraphic Code (1886 edition and 11th edition of 1939). Possibly referring to one version of this is Telegraphic Code Book quoted by Ishikawa (1908) (p.111), which provides phrases with serial numbers and code words of three to five letters having the same initial letter as the plaintext phrase (e.g., "63936 Lava: Let me know at once").
See above. Price 15s. or, interleaved with plain paper, 20s.
See above under ABC Code. Price 25s.
"Figure codes" as used here were not codes using code numbers as opposed to code words. They refer to codes for encoding figures. Typically, they allow one word to represent an order (or offer) for some article with price, quality, shipment, and mode of transit.
The "Exchange" Code contained 100,000 Latin words under ten letters, each representing five figures. With the whole being condensed into eight pages, it was "the most compact code yet published." Price £2 2s. The British Library catalogue lists codes with similar titles: The Stock Exchange Telegraph Code (1885, London, F.C. Mathieson & Son); McKay's Stock Exchange Telegraphic Signal Code (1885, London) by George MacKay; and M'Kay's 'Acme' Stock Exchange Signal Code [a card] (1886, Dublin) by W. R. MacKay, and Co.
The "Nonpareil" Code (1884, Liverpool) (Google) provided a range of figures greater than any code published: 00,000-99,999 plus 100,000-1,000,000 in 50's, and 1,000,000-10,000,000 in 250's. Since the code words were Latin verbs, it could be "used as a supplement to any Code now in use, and be made available for Continental Messages". For 00,000-99,999, each page contains verb roots representing the first three digits and corresponding 100 terminals representing the last two digits 00-99. (Similar terminals such as "ete" and "ere" are included.) For larger numbers, roots represent the first three digits but terminals represent the following three or more digits 000, 050, 100, ..., 0000, 0250, ..., 9750. Price £2 10s.
The "International" Code (Watt's), probably referring to Watts' International Telegraphic Code (see above). Price £4.4s.
An advertisement page of McNeill's Code (see below) has a similar list.
Whitelaw's 401,000,000 "Without doubt the best." Price $35.00. (See below.)
Hackett's 5,000,000 "The best of those of small range." Price $10.00.
Twombly's 2,000,000 Price $10.00.
Hartfield's 2,000,000 Price $10.00.
Economical 10,000,000 Price $7.50. (Possibly referring to The Economical Telegram Code Vocabulary of 10,000,000 inconvertible cipher words  below.)
Morse 20,000,000 Price $5.00.
Early French telegraph codes by Brachet (1851), Brunswick (1867), Sittler (1868), Mamert-Gallian (1874) are given above. This section lists further non-English codebooks. (For Japanese telegraph codes, see another article. For Chinese telegraph codes, see another article, in particular this section.)
A copy is on sale here, with a photo. Also listed at Google.
(Google; the above title is from the title page, though the front cover gives "Corrispondenza in Cifra"; "corrispondenza" is a singular form of "corrispondenze" in Italian)
It used (variable-length) code numbers rather than words. Single digits (0-9) represented vowels and punctuation marks; two-digit groups (00-99) represented consonants, grammatical forms, and auxiliary verbs; three-digit groups represented syllables (pages 0-9, each containing entries 00-99); and four-digit groups (pages 00-99, each containing entries 00-99) represented words and phrases. A code group is formed by an entry number following a page number. While the page number is printed at the bottom of the page, the user's own pagination might be written by the user on the top of the page.
An edition of this code was used in the famous "Panizzardi telegram" sent in 1894 (see another article). At least, code groups used in the Panizzardi telegram can be confirmed in this 1873 edition. (One slight difference is that "1791 Commenta-re, Commento" is rendered as "commenti" in the decoded telegram.)
WorldCat lists editions of 1873, 1888, 1890, 1896, and 1905. ShichereIT has some photos of editions of 1878, 1888.
Listed in Galland (1945). The title suggests this provides letter code groups and numeral code groups.
Listed in Galland (1945). Also mentioned in "La cryptographie militaire" (1883).
Mentioned in "La cryptographie militaire" (1883). Contains 20,000 words.
The 1884 edition contained 24,000 entries and the 1891 edition contained 30,000 (Galende Diaz, Criptografia: historia de la escritura cifrada). An 1899 edition is listed at Google. The 8th edition of 1912 is listed in Galland (1945) and cited as Clave telegrafica by Bauer.
This work by an English code compiler, listed above, is repeated here in the chronological context. This was followed by translations into English and Spanish: The British and Foreign Trade and Shipping Code (1887) and Código Telegrafico Anglo-Español (1888).
Allegedly these were the first publications in Spanish and Portuguese prepared for general use to fill a long-felt want for codes in these languages (British Books in Print for 1888).
(Google) This uses code words rather than code numbers. According to the preface (p.vi), codebooks for representing phrases by single code words had long been used by the English and the Americans, who were "much more experienced than us in the affairs overseas". Apart from private, specialized codebooks, there were no such codebooks accessible to the French public and covering the subjects such as politics, family affairs, travel, construction, machinery, etc. This codebook aimed to fill these deficiencies.
This assigns some 30,000 numbers to words/phrase/syllables. Code words appear to be artificial words containing at most ten letters that resemble real words (e.g., Amniculus, Amobalia, Amoniarum, Amontalis, ...). As the title shows, this is directed to cost reduction. So words of ten letters or less are not to be encoded. Code words are given serial numbers, which are to be used in superencipherment. That is, the serial number for the intended code word is augmented by a prearranged number and a code word corresponding to the augmented number is transmitted.
(Google) This appears to assign code words (not numbers) to phrases. One-letter difference pairs abound (e.g., Saca/Saja/Sala)
Listed in a WorldCat catalogue. Etienne Bazeries probably referred to this when he said that Katscher was the best known commercial code in German (Bazeries (1901), Les chiffres secrets dévoilés, p.141).
Listed at Google.
Listed in Galland (1945).
This appears to be a code in both Dutch and English. Google lists an 1891 edition, MIT Vail Collection and WorldCat list an 1892 edition.
An 1896 advertisement (search for "mercuur-code" at Leiden Universeity) mentions an enlarged and revised edition, which employs codes words carefully selected from the official dictionary (vocabulaire officiel pour la rédaction des telegrammes en langage convenu).
A 1912 advertisement mentions a third edition (in Dutch and English).
10 pp. Listed at Google and Vail Collection.
A photo of the cover is found here.
A photo is found here. Also listed at Google. Seemingly related works are Nuovo Cifrario Mengarini (1898, Rome) (a photo of the 1913 edition is found here, which shows this assigns five-figure code groups (three-figure page number, hand-written, plus two-figure line numbers 00-99) to words and phrases) and Novissimo cifrario Mengarini (1924, Rome).
Listed in Galland (1945). Contains 59,200 five-figure groups.
(See here) This work is motivated by inconvenience of numeral codes or letter codes and uses a different word to represent a plaintext word. The author points out use of words is less error-prone and more economical than figures, which were at the time counted at a rate of three letters to one word for extra-Europe messages.
This codebook includes two parts, each associating numbers (by page-line number combinations) with 19,700 words and phrases (besides which there were 500 blank lines and 800 lines for special vocabulary, bringing the total to 21,000 distributed over 210 pages). When a number (e.g., 3754 on page 37, line 54) corresponding to a plaintext word (e.g., clouer) is found in the first part, that number (e.g., 3754) is used as an index to find another (Latin) word (e.g., cabala) in the second part. For privacy, the page numbers in the second part is left blank, to be filled in by the user.
(Google) This is a codebook for public use compiled by Commandant Bazeries of the French army, the famous codebreaker (see another article).
Much like Sittler (1868), each page 00-99 contains 100 entries (letters, syllables, words, phrases) numbered 00-99, to form a four-figure code. The entries encompass many variant forms. For example, 1347 is "brûl.ant,é,er,ure", meaning either "brûl", "brûlant", "brûle", "brûler", or "brûlure", while the corresponding entry in Sittler (1868) provides only two variants "brûler, brûlure". Representation of such a variety in one entry is a characteristic feature of French military/diplomatic codes, which Bazeries must have been familiar with.
According to the example given in the annex, four-figure code numbers may be regrouped into five-figure telegraph words, which could save telegraph cost.
In addition to these numerical codes (called "chiffres"), a list of 200 words ("mots de convention") (ordinary French words) is given, which the users could fill with any meaning that may be useful to them.
Each page has three blanks ("ordinaire", "réservé", "spécial") for writing alternative page numbers.
Further, the appendix proposes transforming each figure into another figure for secrecy. For example, if the correspondents agree on a keyword "république", each letter of the keyword is numbered according to the sequence of the alphabet (7158043692) and the figures 0,1,2,3,4,5,6,7,8,9 are translated into 7,1,5,8,0,4,3,6,9,2, respectively. This is a technique sometimes called a mnemonic key.
(Gallica) De Viaris was one of those known for skills in cryptanalysis and once pointed out a weakness in Bazeries' ciphering device (Candela p.3, 7-8).
For de Viaris' published code (1898), see below. In this essay, Part I critically reviews development of international regulations up to then. Part II proposes a new encoding scheme with syllables formed by one consonant chosen from ten (D, F, J, K, M, P, R, S, T, V) followed by one vowel chosen from five (A, E, I, O, U). The vowels and consonants are selected such that they are sufficiently different from each other both in pronunciation and in Morse signals (p.48-49). While three syllables provide 503=125,000 combinations, two or four syllables may also be used. De Viaris' idea was at least partially anticipated by Martin K. Thompson (1867) (see above).
20pp. Listed in Galland (1945). Apparently the one referred to by Kahn (p.839).
WorldCat lists an 1899 edition (2nd or 3rd ed.), a 1903 edition, and a 1912 edition (3rd ed.). Also listed are the same author's Racines des mots qui, réunis avec les terminaisons ci-dessous, forment le code Word (1887) [the title reads in English "Roots of words that, combined with the terminations below, form the code word"] and Omnibus: code télégraphique français de poche à l'usage de tout le monde (1905; 1927 3rd ed.). Page images of Omnibus are found at codescans. It appears French words and phrases are represented with seemingly English code words.
Listed at Google.
Listed at Google.
Listed in MIT Libraries' catalog, Vail Collection.
A cover photo can be seen here. 107 pp. According to a photo of the title page, a fuller title is Code-Wörterbuch "Imperial", 1600 Millionen künstliche Wörter von 10 Buchstaben die Zahlen 0 000 000 000 bis 1 599 999 999 bildend .... An English title is given as Code-Dictionary "Imperial", 1600 Millions of Artificial Words, all of 10 letters representing the numbers 0 000 000 000 to 1 599 999 999 in strict accordance with the decision of the London International Telegraph Conference of June/July 1903 and with usual safeguarding measures against mutilations. On the left page, a portion of titles in French and one other language are seen.
Probably this is something similar to Whitelaw's 400 Millions of Pronounceable Words (see below). The claimed 1600 million artificial words would be ten-letter code words made by combining any two of 40,000 five-letter code words.
Said to be for the Ministry of the Interior. A cover photo as well as one page can be seen here. 111 pp.
The image of page 5, apparently the beginning, lists 100 code words with serial numbers as follows:
11,300 bakad A
1 bakal a
2 bakan à
3 bakar ab
4 bakax abaisser, -ement
5 bakei abandon
6 baken prenez radical
7 baker abandonner
8 bakex abatis
9 bakez abattre, abattoir
Although this bears no date, use of five-letter code words suggests 1904 or later. On the other hand, simple vocabulary and single-letter difference code words are reminiscent of commercial codes in the 1870s in England. A typewritten note suggests either code words or code numbers could be used.
Listed in Open Library.
Mentioned here. Also advertised on p.186 of Stichler, Argentinien (1920) (Internet Archive).
A cover photo can be seen here. 252 pp. Google also lists Code International Lugagne (1914) by Georges Lugagne and Gabriel Lugagne, Lugagne's New Vocabulary (1921), Vocabulaire condensé Lugagne (1933), 7 chiffres Lugagne: 5 lettres (1933), which probably translates 7-figure numbers to 5-letter code words (265 combinations being able to represent up to 11881376 including 0000000 to 9999999), 12 Chiffres à redressement Lugagne 000 000 000 000/999 999 999 999 (10 lettres dont une lettre-contrôle)... (1933), wihch would probably be similar to Whitelaw's work in 1910 (see below).
Listed at WorldCat. Page images are found at codescans. Words and phrases are given five-figure numbers and eight-letter code words with alternating vowels and consonants (e.g., "00001 abababal", "00002 ababedem", ...).
Found in Catalog of Copyright Entries.
Listed at Google. This is a seven-letter code according to Bauer (2002). (Does that mean seven-figure numbers could be represented by five-letter codes, whose 265=11881376 combinations could cover 1-9999999?)
Lieber's Code and Western Union Code were popular in the early 20th century. Some codes were compiled to be used with them (see the references to them in the present article). They were among the nine codebooks allowed during World War I (see below).
Benjamin Franklin Lieber compiled eight codes, one of which was widely used and was even translated into French and German (Kahn p.839).
(Internet Archive) This assigns numbers and code words to phrases ("14475 Amastridem: We send you by to-day's mail our draft for --"). Towards the end is included a table of numerals ("70630 Dwergberk" assigned to 100,000,000). All the code words (of 5 to 10 letters) begin with letters A-E and are selected from the Berne Official Vocabulary under "careful supervision". The numbers are said to be for facilitating "arrangements for combinations now so generally employed by the cabling public" (i.e., packing various information into a series of digits, which is to be expressed by a code word).
Another title is Lieber's bankers' and stockbrokers' code and merchants' and shippers' blank tables (1905). [MIT Vail Collection catalog lists a 1900 edition.] Page images are found at codescans.
The Library of Congress Online Catalog further lists the following:
Lieber's universal telegraphic cipher (1888);
Code télégraphique de Lieber (1901), for which a cover photo can be seen here;
Lieber's five letter American telegraphic code (1915) (Internet Archive; this particular copy has typewritten supplements pasted in the blanks); According to the title page, this was "published also in French and Spanish (and Pocket edition in English)"; this retains code words from the Berne official vocabulary as well as the new five-letter code groups.
Lieber's latest code (1926, posthumous).
The Western Union Telegraphic Code published in the 1890s in New York (WorldCat, 472 pp.) was deemed useful enough by the United States War Department that it was specified for use to supplement the US War Department Code of 1899 (see another article). There is also a 1900 edition.
There is also Western Union Telegraphic Code (Universal Edition) (1901, 804pp., Google, 1903 edition (since the images are blurred, the following transcriptions will be inaccurate)). It assigned a (5- or 6-digit) serial number and a code word (up to 10 letters) to words and phrases, for example:
00175 Behaglish ...Absent
00176 Behangael ... Absent from
00177 Behappen ... Have been absent
41785 Esturbato ... Investment was made in accordance with instructions
78202 Kokarcude ... Zero (p.471)
The above forms the main body. Page 472 provides blanks for addition by the user.
From p.485, there is a table of prices for "5,000 Bushels", "10,000 Bushels", "20,000 Bushels", .... The following is examples for "5,000 Bushels."
0 ...... 79000 Kriagemann
0 1/8... 79001 Kriagamuth
Subsequent pages include technical terms.
154149 Verroehten ... Galvanised iron running rope, -- inch circumference
154150 Verroegern ... Galvanised cast steel running rope, -- inch circumference
157167 Vuurjdaat ... Salmon River Gold Mining Company
According to a 1906 ad, it had 175,000 words and was priced $16.00 (pocket edition $10.00).
There was also Western Union Telegraphic Code, Five-letter Edition .
(Internet Archive) This assigns code words (looking like real words) and consecutive numbers (00001 Kabache to 44759 Typtologe) to words and phrases.
It proposes to use a code word with an additional prefix "O" to represent numbers accompanying it. For example, since there is an entry "19444 Pomatomes: All the very best material", a code word "OPOMATOMES" represents a number "19444." Since all the code words begin with a consonant, there is no fear of ambiguity.
Actually, since there are only code words beginning with letters K to W, Lieber's Code and Appendix, which uses code words beginning with A to E only, could be used with the Westinghouse Code (p.xiv).
There is provided a scheme for combining two code words into one (called "code condenser") (p.xv-xvi). It was "intended to be used for economizing in cable messages where the rate per word is high, and to reduce the cost of long messages, but it is not intended for general use in domestic and low rate telegrams."
In order to combine two code words, the five-figure numbers accompanying them are concatenated to form a ten-figure sequence. Every two-letter pair in the ten figure sequence is translated to a consonant-vowel pair by reference to a table. A consonant-vowel pair may be reversed to vowel-consonant order or may be replaced with an alternative letter pair in order to make the result look more like a word. This latitude is also used to make sure the result begins with a vowel (other than "O") and thus is distinguishable from other code words.
This particular instance is described here as an example of combining code words into a ten-letter code word before Whitelaw (1904).
The China Inland Mission had used a small code compiled by T. G. Willett since 1899. This China Inland Mission Private Code (C.I.M. Private Telegraph Code) was a more comprehensive code published in 1907. The second edition of 1913 was a reprint with typographical improvements.
Tables 001-306 each have entries numbered 00-99 (fifty entries on one page), such that a three-digit table number and a two-digit entry number form a five-digit code word. Tables 307-395 include currencies, names, etc.
As with Westinghouse, a table to combine two code words into a ten-letter group is provided. It is called the G.Y.X. Table because the consonants are arranged G, Y, X, V, R, ....
While Westinghouse describes the vowel and consonant of a two-letter pair could be transposed in order to make the resultant codeword look like a real word, the "real word" requirement was repealed by the 1903 revision of the international telegraph regulations. So, the C.I.M. Code proposes to use vowel-consonant pairs (as opposed to consonant-vowel) to represent code numbers of Chinese characters. Since each Chinese character was assigned a four-digit number (another article), an eight-letter code word could represent two characters and two ten-letter code word could represent five characters.
A similar (but not identical) table to combine two code words into a ten-letter group is also cited by Nakagawa (1916) (p.733). The table is referred to as "Consonant Vowel" and is said to have been widely used at the time.
Nakagawa describes a scheme to use the vowel-consonant ordering for representing a checksum (the last figure of the sum of the ten figures). That is, whereas the consonant-vowel order is the norm, reversal in the second pair represents a checksum of 0, reversal of the second and third pairs represent a checksum of 1, etc. Similar checking is also employed by The ACME Ten-Figure Code attached to the Acme Commodity and Phrase Code (1923), p.902.
The same conversion table is also given in Modern Business Practice (1912, London) according to Sakai (1939) , which describes use of the consonant/vowel pattern to represent a check sum mod 32 (p.71-73) or additional information (e.g., the CCCCC pattern representing "Sample Nos. xxxx and xxxx; check figs. xx" and the CCCVV pattern representing "Order Confirmation Nos. xxxx and xxxx; check figs. xx", etc.) (p.74-75).
Another similar table is given in Kurihara (1920) (here p.81-82).
Friedman (p.23) describes another scheme to translate the first two digits into a two-letter syllable and the last three digits into a three-letter syllable.
This allows converting each five-figure group into a five-letter group which is made of syllables and thus looks like a word.
According to Friedman (p.31), the practice of expressing two code words at the cost of one as above had become very widespread in international telegraphy as of 1896.
(Prices below are as of British Books in Print for 1888, unless otherwise indicated.)
Whitelaw, who started a new generation of five-letter code words in 1904, was already active from 1879. "Whitelaw's English, German, and Latin Ciphers" is mentioned as early as 1881 in Macgregor's codebook (see above).
(Google) The publisher is David Whitelaw and this appears to be the first of Whitelaw's telegraph ciphers.
This volume with less than 100 pages is not a codebook that lists code words for expressing words and phrases. Instead, it merely lists code words with numbers (with 400 words on each page). The preface (dated 20 February 1879) expressly describes its use in expressing numbers with those code words. This formed the baseline of forthcoming Whitelaw's telegraph ciphers.
A total of 25,000 English words are given in columns and numbered 00-99. The user can write the first three digits at the top of the columns (similar to Sittler (1868)). While there are 55,000 to 60,000 words in English, "weeding out those too nearly alike for telegraphic purposes" (two-letter differential being not mentioned here) brought down the number to 24,187, all contained in a "very comprehensive and easily procurable" Nuttall's Standard Pronouncing Dictionary of the English Language. The number 25,000 was attained by employing 813 classical words ending in us, which could be found in the list of classical names at the end of Walker's and Webster's dictionaries.
The latter half of the book provides a terminational order, which lists the words in the order of the terminations. British Books in Print (Google) for 1888 advertises Whitelaw's telegraph ciphers as "the only telegraph cyphers with terminational order".
The same telegraph detector as in 14,400 Latin Words (see below) is included.
This is mentioned in the preface of 25,000 English Words.
Whitelaw's Telegraph Cyphers contains 22,500 English words with 25 words to each page, sold by two or more copies at 60s. each. It had "full width of the quarto page for filling in phrases" (unlike the 1879 book, which merely listed 400 numbered words on each page). Probably the words were numbered. Such association of words and numbers could be used to express numbers with words, which were less prone to errors. Bentley (1906) testifies that "Many merchants use Whitelaw's Telegraph Cyphers for their Figure Codes" (Bentley's Complete Phrase Code p.ix). It had a terminational order and had "two systems of numbering." Probably this is The Telegraph Cyphers with Terminational Order. For Phrases ... Arranged Both in Alphabetical and Terminational Order (1884) listed at Google and British Library.
There were several additions separately sold:
40,000 Dutch Words (1886) Price 50s. each. 200 words on each page. Listed at Google as The Telegraph Cyphers, with Terminational Order. 40,000 Dutch Words Not Exceeding Ten Letters. Arranged Both in Alphabetical and Terminational Order.
14,400 Latin Words (1880, Google) Price 15s. each. This lists 25 columns, each containing roots of 24 Latin verbs (up to five letters) to be combined with any of 24 inflectional endings (also up to five letters), making 24x24x25=14,400 inflected Latin verbs in total. The verbs are of the first and third conjugations, ensuring that every combination is an authentic inflected verb in Latin.
Use of regular inflection of Latin verbs of the first and third conjugation was also used in Parker's Combination Telegraph Code (see above) published in the same year, which had ten times more code words than Whitelaw's. (Whitelaw's preface is dated 18th February 1880.) But Parker's code abounded in one-letter difference pairs such as cant/capt, cel/cen/cer etc. and, among inflectional endings, amus/emus and ata/ate/ati/ato/atu. On the other hand, Whitelaw featured two-letter difference between the code words. Thus, Whitelaw adopted "cant" but not "capt." He boasts that "Consistently with this principle no more words than are given here can be obtained from the five-letter Latin roots." This had not been featured in 25,000 English Words published in the previous year.
Use of inflection of Latin verbs is also seen in The "Nonpareil" Code (1884) (see above).
Each inflectional ending represents 24 letters A-Z (there are no I and U) or numbers 01-24. Of the 25 columns of the codebook, the first includes roots representing 24 letters A-Z or numbers 01-24 and the remaining 24 include roots representing combinations of two letters AA-ZZ or two two-figure numbers 01-01 to 24-24. Thus, this codebook could be used to represent any three-letter group or any three two-figure groups not exceeding twenty-four.
Quantity and Quotation Tables Price 20s. per copy, explaining "figure arrangements of all kinds".
Telegraph Detector Price 2s. 6d. per copy. This is for detection of a mutilated word. This may be the same as the "Telegraph Detector" included in the 14,400 Latin words, which lists Morse signals for the letters and figures and possible misreading thereof by wrong spacing between dots and dashes.
The British Books in Print lists separately from the above Whitelaw's Telegraph Cyphers, Final Revised Editions (apparently 1887 (State Library of New South Wales catalogue)), consisting of 202,600 Latin, Italian, French, Spanish, and Portuguese words, £7.10s each unbound, 25,000 English words, 40s. each, 40,000 Dutch words, 50s. each, and 42,600 German words, 50s. each, totalling 310,200 words. Numbering and binding were to be additional to order.
This edition effected "improvements and alterations suggested after four years' experience and careful watching ... of their actual use in daily long and important extra-European telegrams received and despatched." The code words do not exceed ten letters. With pride, it is stated that the terminational order was "now acknowledged to be the best and readiest detector of words mangled in transmission." It used the principle of two-letter difference and avoided telegraphically similar words.
London Mercury (here and another fragmentary OCR text) of 14 November/19 and 26 December 1891 appears to carry an ad for Whitelaw's Telegraph Cyphers including 338,200 words, with the English words expanded to 53,000.
De Viaris (1893) cites one version of Whitelaw, containing 242,600 words taken from six languages, as "the most complete of the collections of actual telegraphic words."
The 1903 revision of the International Regulation officially admitted artificial code words as long as they were pronounceable. In February 1904, even before the new regulations took effect in July, Whitelaw's Telegraph Cyphers: 400 Millions of Pronounceable Words listed 20,000 five-letter code words (called "cyphers") such as FORAB, LUFFA, LOZOJ, FREAN. Since a code word could contain up to ten letters, the user was supposed to put together every two such words into one code word of ten letters. (The advertised "400 millions" referred to the number of such combinations.) Thus, by combining two code words into one, the telegraph cost could be cut by half. Soon this "new idea" became the standard form of telegraph code. (Friedman p.39)
What seems to be its American version is listed in WorldCat as Whitelaw's telegraph cyphers : artificial words ; 401 millions of pronounceable words all of 10 letters representing 4 complete sets of 8-figure groups ; also an additional 134 1/2 millions representing 12 complete sets of 7, 6, & 5-figure groups and all numbers thereunder (New York). [There are 100 million 8-figure numbers, including ones with leading zero(s). Four sets makes 400 million. There are 10 million 7-figure numbers, 1 million 6-figure numbers, and 0.1 million 5-figure numbers, including ones with leading zero(s), which totals 11.1 million. Twelve sets makes 133.2 million. The one million residual may be additional four-letter code words as in Bentley's One Million (see below).]
The British Library catalogue also has codebooks with different ranges:
Whitelaw's Telegraph Cyphers. Pentasyllabic words representing two series of 10-figure groups from 0,000,000,000 to 9,999,999,999 . "Pentasyllabic" seems to refer to a scheme of converting every two-figure group in a ten-figure number into two-letter syllable, in a similar manner to that described above.
Whitelaw's Telegraph Cyphers: artificial words. 1st sheet, containing 101 million artificial cyphers all of 10 letters ... representing all 8-figure groups 0000 0000 to 9999 9999. (2nd sheet, containing 101 million artificial cyphers ... 1-0000 0000 to 1-9999 9999) .
Between these ranges, "Whitelaw's nine-figure words, ranging from 000,000,000 to 999,999,999" and "Whitelaw's One Thousand Millions [1,000,000,000] of Pronounceable Words" (apparently referring to the same work) are credited by Bentley and Broomhall (see below).
The British Library catalogue also lists a seemingly similar codebook published in Bombay: The Economical Telegram Code Vocabulary of 10,000,000 inconvertible cipher words, each of ten letters, numbered consecutively from 0000000 to 9999999. With terminational order [1904; 2nd ed. 1909] (possibly by a different author). In one advertisement, what appears to be this title was offered at $7.50, while Whitelaw's 401 Million was offered at $35.00 (see above).
This page has some photos of Whitelaw's 1 to 12-Figure Groups 0/9 to 000,000,000,000/999,999,999,999 in 10-Letter Pronounceable Cypher (1910). By relinquishing the two-letter difference principle, the entries could be increased from about 2500 millions "in our largest two-letter difference work" to one billion (i.e., 1012 in the British usage). So it is proposed that its use be confined to subjects "where an error in transmission, turning one word into another, would not be of serious consequences." The author proposes error detection by using the last two figures of the 12-figure group as the addition of the preceding ten figures or the last figure of each 6-figure half-group as the addition of the preceding five figures but remarks that these measures "although greatly facilitating the detection of error, have not been found so safe or so convenient in practice as the two-letter difference rule."
The examples given in the preface includes "1 = vevsaaluab" to "124,573,826,447 = atistsyrbi" (my transcriptions may not be accurate but will give the sense of what they are like). Again, the ten-letter code word (including no "q", probably because of the pronounceability requirement) was composed of two half words. To find a five-letter group for a six-figure number, the first three figures is looked up in the Upper Table of two-letter groups at the top of each page. The second three figures are converted to three-letters according to the Lower Table on the same page.
The code is not alphabetical. The receiver converts each half word to a six-figure number. First, the last three letters of the half word is looked up on pages 1 to 4 (if the first letter of the half word is a vowel (including "y")) or pages 5 to 10 (if the first letter is a consonant) to find out the page to look to. Then, the first two letters are translated to a three-figure group with the Upper Table and the last three letters are translated to another three-figure group with the Lower Table.
Probably, the structure of the tables is as follows. The upper tables on the 10 pages as a whole list three-figure numbers 000-999 consecutively and assign two-letter groups. Since the 676 two-letter combinations fall short of 1000, the same two-letter group may be assigned on different pages. The lower tables on different pages include the same three-figure numbers but never the same three-letter combinations. Thus, even if the first three figures are translated into the same two-letter group, the resultant five-letter groups are unique to the six-figure numbers.
Essentially, the information of which lower table is used supplements the two-letter group in uniquely identifying the first three figures.
Various schemes providing multiple tables to supplement the information of letter groups were used in many other figure codes. One example is Vollers' 12-Figure System (1906) (see below).
According to a bookseller's offer, the preface is dated August 1918 and William Douglas, the author of the very first 1879 work, is credited as "author and proprietor." It contains 53,000 five-letter code words attachable to each other, which indicates this is indeed a publication after 1904. Pages are numbered with suffixes a and b (i.e., 0a 0b 1a 1b ...). A terminational order of 39 pages is included. In addition to two-letter differential, telegraphically convertible code words and mere transpositions of two adjacent letters such as abdar/adbar were avoided.
This is listed at Google.
(Prices below are as of the reprint at Google.)
(reprint at Google, an earlier reprint at Internet Archive)
Whitelaw (1904) opened an era of five-letter code words but, merely listing code words with numbers, it was used as a figure code to express numbers with words, which are less prone to transmission errors, as also described in Bentley's preface. Bentley (1906) took full advantage of Whitelaw's new idea in a codebook to encode words and phrases and, being a "compact, well-constructed, moderately priced", sold about 100,000 copies. It was "perhaps the best known and most widely used of commercial codes" (Kahn p.843-844; Bentley's Second p.iv).
The five-letter code words represent about 30,000 words and phrases such as "ababd: A", "abajl: Abandon(s)", "abaln: Abandon it (her)", "wuvma: Zone(s)", etc. Some 2,000 code words "wuvne" to "zyzwo" were left blank for supplement.
Any two of these code words were to be combined into a ten-letter code word. (Hence, a five-letter code word was called a "half-cypher" or "half-word".) Thus, even if a fitting phrase was not found in the vocabulary and each word had to be encoded into one code word, 50-percent cost saving was obtained.
Since "most Codes have dictionary words as cyphers", the artificial code words of this code could be used as an auxiliary, indiscriminately mixed with code words of other codes.
The code words were not numbered. Instead, after the entry "lamof: Numeral(s)", many code words were assigned to numerals, including fractions, such as "lamug: 1/64", "layne: 1", "mejiz: 1,000", after which numbers up to "miagz: 2,000" in steps of 10 and thereafter up to "mumwy: 5,000,000" in increasingly larger steps. These numbers could all be expressed with one five-letter word.
Two-letter differential was implemented. (Actually, Bentley overlooked some one-letter-difference pairs such as BEBPY/BEEPY, CIBPY/CIEPY, etc. (Bellovin p.7)) Further, efforts were made to eliminate words that were "telegraphically convertible". To help error correction, it provided a termination index ("Cyphers in Terminational Order").
Bentley's Complete Phrase Code was offered in various bindings: Half Morocco (£4 4s.), Thin Paper, Pocket Edition (£4 4.), Thin Paper, Whole Morocco (Limp) (£4 14s.6d.), Thin Paper, Pegamoid (£4 10s.).
Responsive to the demand of the users, there appeared various derivative works, which are advertised on the backside of the title page of the reprint. Some of these are explained in the preface (1909).
This used the supplement section to include a wide range of mining expressions. Binding was Half Morocco (£4 4s.) or Thin Paper, Whole Morocco (Limp) (£4 14s.6d.). The Supplement was also sold separately at £1 1s. The Mining Supplement (1907, 32 pp.) is listed at Google.
Two bindings: Half Morocco (£4 10s.) or Thin Paper, Whole Morocco (Limp) (£5). The Supplement was also sold separately at £1 5s.
This is mentioned in Bentley's Second (see below).
Printed versions of supplement of 2,000 phrases were made for large private firms and companies, usually costing about £12. For the public, separate Blank Supplement (10s. 6d.), Special Supplement 300 half-cyphers (5s. 6d.), "X" Supplement 872 half-cyphers (5s. 6d.), and "U" Supplement 2,477 half-cyphers (15s.) were offered. (The "U" Supplement (1922) is catalogued at WorldCat.)
The original Phrase Code did not use code words beginning with Q or X or ending with I or U. Probably, the X and U Supplements took advantage of these spaces.
This is said to be "7 & 3 letter combinations with blank lines" and was available for £1.
A skeleton version with 11,000 code words with blank lines was available for £2 10s.
This is appended to the above Google reprint and appears to correspond to the Numbered Telegraph Cyphers One million Edition (10s. 6d.).
Two systems of representing numbers with code words are described.
The first is a list of four-letter code words (0: abev to 999: zuof), having two-letter differential, representing numbers up to 999. By one eight-letter code word consisting of two such code words, one could represent a code word in the range 000000-999999 used in another codebook (or three two-digit numbers). Since these words consisted of eight letters, they could be unambiguously distinguished from the code words from the Phrase Code.
The second system proposes to represent a code number up to 210,000 with one ten-letter code word consisting of two five-letter code words. By using code words representing numbers up to 210 specially reserved for this purpose (i.e., separate from the code words representing numbers) (0: zuker to 210: zyzwo), the ten-letter code word could be the latter half of one ten-letter group transmitted and the first half of the following code word.
It was recognized that code words having two-letter differential could be transmitted with more accuracy than figures. (One error in a digit of a number changes the number to another, which could never be detected from the internal information.) Although the Phrase Code covers numbers up to 5,000,000, not every number could be represented by one half-word. The deficiency was filled by several figure codes.
900 Million Edition (£4 4s.) provided half-words (in alphabetical order throughout) representing 00000 to 29999. This could express numbers up to 2999929999. (So explains Bentley on p.x but, with such juxtaposition, numbers 30000-99999 etc. could not be expressed. Bentley means to use the number for packing data into one code word as described below.) This was also offered bound with the Phrase Code at £8.
400 Million Edition (£2 10s.) provided half-words (beginning with a consonant) representing 00000-19999, expressing numbers up to 1999919999.
300 Million Edition (£4 10s.) could express numbers up to 299999999 or 999929999 by a combination of a half word-beginning with a vowel (0000-9999) and another half-word (00000-29999).
200 Million Edition (£3 10s.) could express numbers up to 199999999 or 999919999 by a combination of a half word-beginning with a vowel (0000-9999) and another half-word beginning with a consonant (00000-19999). ("200 Million" and "300 Million" are more expensive than "400 Million" probably because they combined two lists of numbers.)
100 Million Edition (£1 10s.) provided half-words (beginning with a vowel) representing 0000-9999, expressing numbers up to 99999999.
One Million Edition is described in the previous subsection.
The author also recommends packing several numbered data in one code word. The code words up to 2999929999 could express any of 29 sentences (by the first and second digits), any of 99 prices (by the third and fourth digits), any of 9 shipment options (by the fifth digit), any of 299 articles (by the sixth, seventh, and eighth digits), and any of 99 quantities (by the last two digits) in one word.
While four-letter words from the One Million Edition could be distinguished from the code words from the Phrase Code at a glance, code words from the other figure codes might not.
This led to 600 Distinguishable Cyphers with "2-Vowel Prefixes", for expressing numbers 000-599 with code words beginning with two vowels and ending in an "i" or "u". Such beginning or ending features allows distinguishing these code words from those of the Phrase Code.
Similarly, there were "2-Vowel Suffixes" for expressing 100 numbers 0-99 with four-letter code words ending in two vowels as well as "English Cyphers" for expressing 300 numbers 0-299 with five-letter English (uncommercial) words.
The author also proposes using 999 of the blanks in the Supplement (called "Z Cyphers") for numbers 000 to 999. Using these as a prefix combined with a half-word from any of the figure codes, numbers up to 99929999, 99919999, or 9999999 could be expressed, which could be unambiguously distinguished by the first half-word reserved for this purpose.
The author points out that firms who frequently use figure codes employed the maximum combinations of the 900 Million Cyphers without using such Distinguishable Cyphers but instead using a special address indicator to show that the message is to be read with the Figure Code, not with the Phrase Code.
In addition to the code words assigned to the numbers in Bentley's Complete Phrase Code, by using the Million Cyphers, the user could include code words (represented with the assigned numbers) from other codebooks such as "the A.B.C., the Western Union and Lieber's Codes." (Many codebooks had numbers accompanying the code words.)
Secret cipher cards for rendering the messages secret were available at 2s. 6d. Moreover, exclusive cards different from any other in existence could be compiled for the use of particular firms.
According to the explanation on p.xiv, "The black figures are applied to the Code Words, and instead of the number of the Cypher standing on the Card against the Cypher, it tells you to send another Cypher (according to the number stated) in the same column. The recipient places the red side ("Translating Card") against the Cypher received and he is at once referred to the Cypher conveying the correct translation." Apparently, this card printed numbers in columns (black and red) and placing the card on a page of the codebook provided numbers to the code words, which could be used to change the association of code words with plaintext.
This, priced at £2 2s. contained 153,000 sentences on general subjects. This is "Scott's Code": The Ship Owner's Telegraphic Code compiled by E. B. Scott and enlarged and re-arranged by Bentley (British Library catalogue). (Bentley had been private secretary to a partner of a shipping agency, where he revised a private code, and founded a code company in 1905 (Kahn p.843).)
Since the code words were "whole word cyphers" selected from the Berne Official Vocabulary, i.e., real words officially admitted to belong to any of the eight admissible languages, it could be used as an auxiliary to the Phrase Code.
Listed in the British Library catalogue.
Price £2 10s. Details unknown to the present author.
A cover photo can be seen here (325 pp.). The code words of Bentley's Phrase Code are not numbered. This may be a version with additional numbering.
(1921 reprint listed at Google) This does not mention Bentley anywhere. Pagination is different and the five-letter code words are printed in all capitals (jmcvey).
(1945 third reprint at Internet Archive) The new International Regulation of 1928 admitted any sequence of letters, pronounceable or not, up to five letters as a code word. This allowed a huge increase in the vocabulary. On the other hand, while ten-letter code words were still admitted, the new rule sptipulated that ten-letter code groups must contain at least three vowels, which precluded combination of five-letter code words with only one vowel (e.g., "flond", "flopf", etc.) as found in the Phrase Code or other five-letter codebooks.)
Bentley's Second included about 94,000 words and phrases (e.g., 00000 aabat: A; 93446 ylpde; Yokohama; 93718 ymmgo: Zwickau), more than three times as many as in the original. In addition, it had code words with blanks (93719 ymmif to 91688 yvixf) and a list of code words without blanks (91689 yviyn to 99999 zyzym).
The revision of the telegraph regulations in 1932 (in force since 1 January 1934) did not affect this and Bentley's Second remained unaltered.
While Bentley's Second was intended to replace the original Phrase Code, code words in the Blank Supplement at the end of the Phrase Code were avoided and thus published or private supplements using these blank code words, including Oil, Mining, and Wool Supplements, could be used with the Second.
In addition to the usual two-letter difference between code words, "reversal of any pair of consecutive letters", "reversal of any three consecutive letters" [i.e., transposition of two letters separated by one intervening letter], or "mutilation of any pair of consecutive letters ... by a pause-error in transmitting by Morse" [presumably referring to such pairs as ER ". ._." vs. UE ".._ ."] will result in a pattern not used in the codebook and thus will surely be detected. (Bentley reminds the user that "many modern telegraph instruments do not involve the use of Morse".)
Further, those for numerals, sums of money in Sterling or US dollar were specially designed to be different from each other in at least three letters. Such code words were interspersed throughout the entire volume of the alphabetically arranged code words and marked with a dagger (†). For encoding, those scattered code words were assembled in tables near the end of the book (including numerals up to "98859 zolce: 10,000,000", including consecutive numbers up to 1,000 and fractions). This replaced the entries under "Numeral" of the first Bentley.
Another means of encoding numerals is provided by using serial numbers assigned to code words, which were not present in the Phrase Code. By using code words to represent the serial number assigned to it, consecutive numbers up to 99,999 could be represented with one word. The instruction provided that such use should be limited to use with particular code words marked as such.
For error checking, Bentley's Second had a mutilation table. This is a table to specify the relationship between the first two letters, the middle letter, and the last two letters in a code word. Any code word in Bentley's Second is formed according to this table. This allows unique identification of the mutilated letter provided that the other four letters were correct. (Since any of the five letters may be in error, there would be five candidate code words, from which some might not be actually present in the codebook and others may be excluded by the context.) In addition, the mutilation table could ensure the two-letter difference without manually checking. Thus, the overlooked one-difference pairs in the Phrase Code must have been avoided.
A mutilation table is not found in ABC6 (1920) or Imperial Combination Code for Mining, Company Promoting, and Stock Exchange Purposes (1913), which both had a terminal index, and Lieber's five letter American telegraphic code (1915), which had neither a mutilation table nor a terminal index. There is a mutilation table on the inside back cover of the Universal Trade Code (1921, New York), compiled under the supervision of Herbert O. Yardley and Charles J. Mendelsohn (Internet Archive).
See another article for details about mutilation tables.
Here are some codes by Edmund Peycke.
Mentioned at the back of the 1918 codebook below. Contained 10,000 words in 120 pages.
Google lists The Economy Telegraphic Cipher Code (1900) with 211 pages.
(Edmund Peycke code) Mentioned as Revised Economy Code at the back of the 1918 codebook below. Contained 50,000 words in 343 pages.
This codebook is characteristic in the way it employed to expand the vocabulary while keeping away from foreign languages "from which so many words have of late years been used in cipher codes, most of them being harder to copy and spell, and a great deal worse to telegraph than the worst of our made-up words." It boasted not only the size of its vocabulary but also its use of "made up" words: "The Book before you has a Vocabulary of Code words numbering over 40,000. There is not that number of words in Webster's Unabridged that can be used for this purpose, for a Code-word must have no meaning in the particular line of business in which the Code is used, nor should words be used the spelling of which differs in but one letter. Furthermore, all compound words, as well as words consisting of more than ten letters are barred by recent ruling of the Telegraph Companies. In order to follow these rules, as we have done, not only made-up words (pronounceable groups of letters) have to be used, but also a great many that have the appearance of being badly misspelt dictionary words." For example, this codebook contained 49 "words" composed by adding suffixes to the word "nigh" such as "nighant", "nighbake", "nighcast", etc.; 67 similar "words" composed by adding suffixes to "night", etc., of which only a few were bona fide words (Friedman p.22 n.1).
Listed at the British Library catalogue. According to p.146 of Peycke's Grain Code (1905) (see below), this "fits such Codes as Western Union, Liebers ABC, A No. 1 Postal (new edition) and converts ANY TWO of THEIR Cipher words INTO ONE word." Code words are five-letter artificial but pronounceable words. From these, this "key" may have been something like Bentley's Million Cyphers (see above), which could be used to include code words from other codebooks such as the ABC, the Western Union, and Lieber's Codes. That is, using the key, one could convert the number assigned to a code word in any such codebooks into a five-letter code word, two of which could be combined into one ten-letter code word -- a measure to reduce the telegram cost by half probably invented by Whitelaw (see above) and followed by Bentley (see above) and many others.
The key could also "'repeat' any word or words in a message without extra cost." (It is not clear how this could be done. If there were a code word to mean "repeat xxx", it would add some cost.)
The key could also give absolute secrecy, whereby "Any Public Code is turned into a Secret Code by the use of our Key." Possibly, this involves some manipulation of the number assigned to a code word before translating it into a five-letter code word.
(Edmund Peycke code) This uses five-letter artificial but pronounceable code words as above.
The title page says this is "arranged under Peycke's Abbreviation System." According to p.146, code words are of the same character but "applied in a different way." Although Peycke was fully aware of importance of two-letter differential (see above), there are one-letter difference pairs: kelbo/kello, keldo/kelmu, kelfa/kelna, pibdo/pibmo, pibga/pibpa, ....
Similar to the General Conversation section in the Modern Economy Code of 1908 (Edmund Peycke code).
(Edmund Peycke code) Mentioned as Modern Economy Code as a successor to the Revised Economy Code (1903) in the 1918 codebook below. Contained 125,000 words in 246 pages. This employed five-letter code words.
While Grain Code (1905) seems to pay some attention to one-letter difference (e.g., the one-letter pair kelbo/kello is found in the context: kelab, kelbo, keldu, keled, kelfa, kelge, kelif, kelki, kello, kelmu, kelna, kelog, kelpe, kelri, kelso, keltu, keluk), such consideration is not seen, for there is a sequence of code words such as camib, camid, camif, camig, camih, camij, camin, camip, camir, camis, camit, camiv, camix.
Occasionally, three-letter syllables (e.g., representing the average weight of watermelons) and two-letter terminals (e.g., representing the approximate weight of carload or the number of melons per carload) are listed separately, to be freely combined into five-letter code words.
(Edmund Peycke code) This codebook, puslished soon after the end of World War I, contained 1,310,089 "cipher words", more than ten times that of its predecessor of 1908. But the volume expanded from 246 to only 312 pages. Such an expansion of vocabulary could be allowed because from the end of 1909, cable companies in America decided to charge on a five-letter-per-word basis, regardless of pronounceability. So, there are code words starting with "dc", "dd", "df", etc. At least, the code words have at least one vowel with the exception of the Car Numbers in the Bulletin Table, which have purposely been confined to consonants to distinguish them from the regular code words. Again, one-letter difference pairs abound.
The Declaration of London (1911) (Internet Archive) has a list of codebooks then available.
"A few copies of the Original Edition. Price on application." This is a list of officially admitted code words taken from various languages. See another article. Here is a newspaper report of the publication (The New York Times, 19 December 1894).
It contained 256,740 words of 5 to 10 letters. The code words differ from each other in at least two letters and in three elementary Morse signs. They were all numbered (from 000,000 Aabam to 213,949 Zythogala (jmcvey)), with the three figures representing thousands being given at the head of each page and the lower three figures before each word.
While much enlarged four volumes were published in 1900-1901, their code words were not numbered. The numbering of the code words allowed the 1894 version to serve as a figure code (as noted under Figure Code for Stocks and Shares) below.
Price 20s. See above.
Price £4. See above.
Price £3 3s. A fuller title is Beith's 10-Letter Combinations. 8 Figures. Pronounceable According to the Requirements of the International Telegraph Conference of 1903 (1904, Manchester) (60 pp.). Probably, this was similar to Wihtelaw's 400 Millions of Pronounceable Words published the same year (see above) and provides 10,000 five-letter code words, any two of which could be combined to form ten-letter code words. Since numbers assigned to the code words can represent numbers 0000 to 9999, combinations of two may represent numbers up to 99999999 (8 figures).
There are also Beith's 10-Letter Combinations. 9 Figures (1904) and Beith's 10-letter Combinations. 6-figures (1906) (11 pp.).
Price £4 4s. See above.
Specially for the use of tourists. Compact and bound conveniently for the pocket. Price 1s.
Price 50s. (1907, 1909 ("rubber edition" adapted for rubber industry), 1913 ("rubber edition") at Internet Archive) The over 38,000 five-letter code words were taken with permission from Whitelaw's One Thousand Millions of Pronounceable Words. A "terminational order" of the code words is included. Opposite the title page, there is a notice of certification by the telegraph administrations to the effect that "although a certain proportion of the expressions in the said Code give rise to difficulty in pronunciation, the Code must be considered, on the whole, as conforming to the conditions fixed by paragraph 2 of Article VIII of the International Regulations." (The article refers to the new provisions in the 1908 Regulation. See another article.)
Broomhall had authored 89 distinct private codes besides public codes below. The Imperial was intended to be used with the author's such other codes.
(Google lists 1895 and 1898 editions.) Arranged for nearly 170,000 phrases. Price £3 13s. 6d. Cloth. Limp leather. Price £4 4s.
Price 60s. net.
To be used with the "Official Vocabulary," or any similar list of numbered words. Price 42s.
An Immediate Saving of 50 per cent in Cabling Expenses [i.e., combining two five-letter words into one ten-letter word]. Adaptable to all Codes. Price 63s.
Adapted for the special use of the Old Iron and Metal Trades. By George Cohen, Sons & Co. Revised Edition, 1903. Price 42s.
Google gives a fuller title The Ideal Code Condenser: Being a Thirteen Figure Code (0,000,000,000,000 to 9,999,999,999,999) for the Purpose of Enabling a Number of Different Numbered Codes to be Employed in the Same Message at the Rate of Two Phrases to One Code Word, the Recipient Being at the Same Time Informed from which Code to Read Each Part of the Message, with Two-figure Safety Check Against Mutilations. With Full Instructions in English, French and German. Price 21s. It provides a two-figure safety check against mutilations. The Telegraph Administrations of the United Kingdom, France, and Germany have, on behalf of the International Telegraph Union, examined the 'Ideal Code Condenser,' and have issued a certificate in virtue of which the words compiled from the same cannot be queried or rejected."
See another article for more.
Arranged to meet the requirements of Mining, Metallurgical and Civil Engineers, Directors of Mining and Smelting Companies, Bankers, Brokers, Solicitors and others. Price 21s. This may refer to McNeill's Code (1895), though the online copy, having "Preface to Reprint", must have been of a later date because advertisement pages mention ABC5 published in 1901 as well as Whitelaw's 401,000,000 Pronounceable Words offered at $35.00 per copy, the latter suggesting a date in 1904 or later. A 1908 edition is listed as  in Bellovin.
Price £10 per copy (for not less than two copies). (British Library and Google also list similar-titled Parker's Nine Figure Code, or 1,000 Million Pronounceable Words (1906, Manchester) by Alexander H. Parker.)
Code I. The General, Commercial and Mining Telegram Code containing 274,000 Phrases and Sentences. Google gives a fuller title The General Commercial and Mining Telegram Code ...: Preceded by a Complete Index to the Most Important Words in the English Language, with Groups of Words Expressing Ideas Closely Related to Suggest and Facilitate the Compilation of Sentences. Also Including Economical Combination Tables ... . Together with the Official Vocabulary of Cipher Words, Prepared in Accordance with the Decisions of the International Telegraph Conference of Paris (1897). Price £5 5s.
Code II. The Multiform Combination Telegram Code, (1903) with 206,460 Cypher Words, with 960,045 Groups of Numbers. £8 8s.
Code III. The Catalogue Combination Telegram Code, (1903) consisting of 274,979 separate References to Catalogue Numbers. Price £7 7s.
For the Use of Mining Companies, Mining Engineers, Stockbrokers, Financial Agents, and Trust and Finance Companies. Price 21s.
Especially adapted for use in Banking and Investment Business. Price 20s.
Can be had in English, French, Spanish or German. Price 30s. This is advertised in McNeill's Code (see above). For example, it allows sending four code words from ABC5 (e.g., Certifichi Umbrao, Silabear, Drilboren) to be cabled in two words (e.g., Carudlotef, Lafidsifig). Similarly, four code words from Lieber's Code (e.g., Aeravissem, Aerocorum, Bastonerai, Dekenwever) can be sent as, e.g., Bilerbilig, Gupalrazei. Even one code word from ABC and another from Lieber's could be sent as one word without any confusion.
Price 52s. 6d. (cf. 'Scott's code'. The ship owners' telegraphic code (1880) by Edward Benjamin Scott. Reprint with supplement (1882) combined (Google).)
Can be used as a 6, 7, 8, 9, or 10-figure Code (or more). Adaptable to any size of Telegraphic Code, on any subject and in any language. By Arthur Tracey. Price, with patent binder, £3 13s. 6.
(jmcvey; advertised on p.186 of Stichler, Argentinien (1920) (Internet Archive) under the title Vollers 12-Zahlen-System) 1,000,000,000,000 Pronounceable Words, all of 10 letters, in strict accordance with the decisions of the London Telegraph Conference of 1903. Price £2. The claimed combinations are achieved by providing five-letter code words for six-figure numbers (one million) and combining two such half-words into a ten-letter code word (one million times one million). Of the six figures, the first and the second figures 00-99 were converted to two letters by referring to a column on the left margin. The third and the fourth figures 00-99 and the fifth and the sixth figures are each used as indices to the rows and columns of a table running over some pages to look up a three-letter pattern. Pronounceability is allegedly ensured by (i) avoiding jj, xx, iiy, iyi, iyy, yii, yyi, yiy, as well as aaa, bbb, ccc, etc. and (ii) selecting the two-letter combinations on the left margin to form easily pronounceable combinations with the three-letter patterns on the table on each page.
1,000 Millions Pronounceable Words of ten letters. Price £2.
Price £7 7s. This is not found on Google, which instead lists Wilson's Ship-broker's Telegraph Code (1881) (Google). "Watkin's code" and "Watkin's code (new edition)" are entries of Bentley's Phrase Code (1906) (but not in its second edition (1929)) along with A1, ABC, Ager's, Bentley's, Lieber's, Scott's, Standard, and Western Union. Watkins' Universal Shipping Code is mentioned in Nakagawa (1916) (p.723).
Price 65s. See above.
Marconi Dictionary and Marconi International Code (1919) by James H. C. Macbeth, London Manager of the Marconi Wireless Telegraph Company, were used by Marconi wireless operators worldwide (Marconi Apparatus).
The Marconi International Code is interesting in that it covers nine languages (including English, French, Spanish, German, Dutch, Italian, Portuguese, Russian, and Japanese) in four volumes (Bauer, Fig. 40 etc.). Each volume deals with English plus two other languages and assigns five-figure consecutive numbers and five-letter code words (e.g., UVVIM, UVVON, UVWEO, UVWUP, UVWYR, ... for consecutive numbers 19140-19144). Opposite each code word are given words in English and the other two languages. Thus, in theory, one may encode in one language and a recipient may decode in another of the nine languages. Since the alphabetical arrangement of words was preserved only for English, the eight other languages had indices referring to the place in the code. A similar code of 1,048 words for five languages was compiled by Athanasius Kircher back in 1663. (Kahn p.845-846, including a partial page image; LANAKI also has an excerpt)
Some images of Marconi's Wireless Telegraphic Code, apparently the basis of the International, are found here.
According to Wikipedia in German, the code was developed in the company of Rudolf Mosse (1843-1920) (Wikipedia) and was published in 1922. It includes nearly 100,000 entries represented by five-letter code words with two-letter differential (the letter "q" is not used).
It consists of the main Part A (abaal to ukdas for general phrases) as well as Parts B (blanks for addition), C (unoot to yftir for numbers, quantities, currencies), D (yftna to zoear for technical terms), E (zohol to zuzyv for phrases with variable numerical values), F (zwaag etc. for "Drei-Buchstaben-Schlüssel" [three-letter keys]), G (zyeac to zyzsa for organizations inside the Rudolph Mosse company), H (tables for deciphering multilated code words), J (zyzup, zyzyf, ... for secret writing), and K (Mosse-Condenser).
The code was also published in other languages (WorldCat lists English (c. 1926), French (1929), Spanish ). There is a supplement (1929) including 130,000 entries. This supplement is multilingual and entries are given in German, English, French, Spanish, and Portugese. An image is found at jmcvey.
The Mosse Condenser (Part L in the English edition cited) is explained in Baba (1940) p.203. It allows conversion of ten-figure numbers into 10-letter code words. First, the ten figures are grouped into five two-figure numbers. The sum of the five two-figure numbers is looked up in a "control table", which provides five letters. The five letters (e.g., BBBAA) indicate the columns which should be used to convert each of the five two-figure numbers into two letters.
A letter partially encoded with this codebook in 1941 can be seen at Klausis Krypto Kolumne. It was sent by German chemicals group Bayer and decoded by British codebreakers.
From 1914 to 1918, censorship put a stop to international commercial communication in code almost worldwide (Friedman p.59). When joining the war, the United States introduced censorship on cables out of the United States and telegraph/telephone lines into Mexico, as reported by newspapers of 1 May 1917 (e.g., The Oregon Standard). Codebooks still allowed were as follows:
A.B.C. fifth edition;
Scott's tenth edition;
Western Union (not including five-letter edition);
Lieber's (not including five-letter edition);
Bentley's Complete Phrase Code (not including Oil and Mining Supplements);
Broomhall's imperial combination code;
Broomhall's imperial combination code, rubber edition;
Meyer's Atlantic Cotton Code, thirty-ninth edition; and
Riverside code, fifth edition.
Restriction on use of codes was lifted in December 1918 (ASA, Wayne G. Barker (ed.), The History of Codes and Ciphers in the United States during the Period between the World Wars, Part I. 1919-1929 p.75).
In the five or six years after World War I, more codes were produced than in the twenty years before it (Friedman p.59, Kahn p.845). Among them were many great commercial codes: the ABC (6th) (see above), the Acme, the Boe, Farquhar's, the Lombard, the Rudolf Mosse, Peterson's, the United Telegraph, and the Western Union (Kahn p.845). It was the golden age of the commercial codes. A WorldCat catalogue search finds the following titles:
Acme Commodity and Phrase Code (1923) and Acme Supplement (1932) by A. C. Meisenbach (see also jmcvey);
Commercial Phrase and Shipping Code (1925), Appendix to the Boe Code (1930), The New Boe Code: commercial, traffic, and shipping code (1937), of which the mutilation table, including no Q, is shown in Fig. 6 of Bellovin, The 'Q'-list to be used in connection with the new Boe Code (1937) by Conrad Boe;
Cosmo Cipher Words (1913), Farquhar's International Bank(ers) and Commercial Code (1929) by Cosmo Farquhar;
Lombard Shipping & Transport Code (1934?), Appendix Lombard Shipping Code;
Rudolf Mosse Code (see above);
Peterson's International Banking Code (1911), Peterson's International Code (for 2nd ed. (1923), see Fig. 19 of Bellovin, showing bilingual plaintext in English and Spanish; for 3rd ed. (1929) see jmcvey); an actual telegram encoded with this codebook (3rd) addressed to the government of formative Israel in 1948 can be found at Klausis Krypto Kolumne.
The United Telegraph Code, signal "UCODE" (1933), X.Y.Z. Section, 1934, United Telegraph Code (1934); and
Western Union Telegraphic Code (see above)
While two-letter differential was already the norm, the Acme code introduced exclusion of code words which may be formed by transposition of two adjacent letters, which could be attained by constructing the code on the basis of a mutilation table (Kahn p.847-848). An earlier attempt to that effect by manual work is found, e.g., in Whitelaw's Special Cyphers (1918).
Three-letter codes used in the 1870s reemerged in the 1930s. Combination of three code words into ten-letter groups (one letter being for error checking) was officially facilitated by the new regulations of 1928 and 1932, which removed the pronounceability requirement for code words. Although a three-letter code could provide only 17576 combinations, if the necessary vocabulary (e.g., specific to a particular trade) fits in this range, it could provide further economy relative to five-letter codes. Since two-letter differential could not be implemented in order to fully exploit the small number of combinations, some error checking scheme must have been essential.
John McVey's police and forensic codes describes this. I wanted to mention it here because it is one of the few three-letter codes before 1927 and after 1880 known to me.
This 547-page volume appears to have been substantially enlarged from its predecessors Keegan's International Code, Specially adapted for the use of shipowners, shipbrokers, etc. (1913, 163 pp.) and Keegan's International Code. A practical Three-Letter Code (1917, 2nd ed., 348 pp.) listed at Google.
According to the descripton of the third edition by Baba (1940) p.219-221, this appears to expand the limit of vocabulary by using an indicator to switch the code between "oil", "special", "brokers'" etc. For example, IDI means "Read from oil code until indicator changes"; UGU means "Read from "Special" Code until indicator changes"; and "ECE" means "Read from Brokers' Code until indicator changes."
Each three-letter code group ("cypher") is assigned a digit ("value"). The three digits for three 3-letter code groups are summed and the result plus the three code groups form a ten-letter group.
(1930 edition at National Archives)
Before this three-letter code came out, a five-letter police code had been compiled in New York under Richard E. Enright, President of the International Police Conference, in accordance with a resolution of the 9th session of the International Police Chief's Conference in New York in May 1923. It took effect in January 1925. Outside the United States, the code was adopted by police departments in Canada and Siam, Chile, Peru, England, Australia, South Africa, Esthonia, China, Mexico, Guatemala, Cuba, and India. (New York Times, 14 January 1925; JACAR Ref. A05032275400)
Preparation of an international police telegraph code had also been resolved in the International Police Congress in Vienna in September 1923 but "the code already in use in the United States of America" was considered unsuitable for general use in Europe because of the employment of alphabetical arrangement according to English words. The arrangement of this three-letter code based on "a logical sequence of subjects" ensures that knowledge of the English language would not be required. (the 1926 preface of the 1930 edition)
In 1927, a German version of the three-letter "International Police Telegraph Code" was completed (the forerunner of Interpol (Wikipedia), established in 1923, was then headquartered in Vienna; the United States did not join until 1938) and was translated into French in 1928. By 1931, it was also available in Bulgarian, English, and Czech. (Mathieu Deflem, Policing World Society, p.130; his source is some proceedings of Internationale Kriminalpolizeiliche Kommission) Through translation into local languages, "the Code renders superfluous the translation of an incoming telegram ..." (preface)
The code uses three-letter code groups. Under the section "personal and physical description", there are code words ... atz, axa, axb, axd, axe, axf, axg, axh, ..., with "axg" meaning "strikingly tall" and "axh" meaning "strikingly short". Three 3-letter code words (plus an additional "j" to "prevent the cyphers running into each other") are to be transmitted in two (five-letter) words. (So a check digit is not employed.) The letter "w" is reserved for placing before and after names etc. in plaintext.
(Until August 2014, the present article had adopted the year 1926 from Bellovin but I found that his later slides only referred to the 1930 edition. John McVey's reference to the 1930 edition (police and forensic codes) inspired me to rewrite this whole section.)
This proposes to combine three three-letter "ciphers" plus one check letter into one ten-letter code word. (At that time, up to ten letters were allowed in one code word.) For example, GEY ("Refer to your letter of the 1st ult."), GPO ("impossible to operate at your limit"), HEN ("market advanced"), and Y (check letter) are combined into a code word GEYGPOHENY for transmission. Each cipher (ABA-QZY) representing some word or phrase is given a serial number (0000-8917) and a check figure (0-9, which is the last digit of the sum of the digits in the serial number).
To find the check letter for the ten-letter code word, three check figures for the three ciphers being combined are added and the last digit of the sum is converted to a letter by the correspondence: 0(A) 1(E) 2(I) 3(O) 4(U) 5(Y) 6(N) 7(Z) 8(S) 9(L) or alternatively 0(F) 1(B) 2(G) 3(R) 4(T) 5(M) 6(C) 7(K) 8(D) 9(J).
A terminational order (called "multilation detector") is provided, which is a list of "all ciphers grouped according to check figure and arranged in terminational order". If the check letter does not match, the recipient determines by context which of the three ciphers is mutilated. (If the decoding makes sense, it is assumed the check letter is wrong.) If the first letter of the mutilated cipher is wrong, the check letter of the ten-letter code word is used to determine what the check figure of the correct cipher is supposed to be. Then, with reference to the terminational order for that check figure, possible candidates are found by using the last two letters. If the second letter of the mutilated cipher is wrong, with reference to the terminational order for the supposedly correct check figure, possible candidates are determined by using the first and third letters. If the third letter is wrong, with reference to the body of the code, the first two supposedly correct letters allow narrowing down to possible candidates. From the possible candidates, a correct one is chosen by context.
The check letter system also allows distinguishing the code words of the Oriental from those from other five-letter codes. Thus, the author claims "the user can use any code in conjunction with this code" without confusion.
In addition to ciphers representing numerals, association of serial numbers and ciphers may be used to represent numbers with code words.
While the second edition was altered largely in form only, the somewhat expanded introduction suggested use as a figure code. While the author considered "figure codes such as 10-figure code and 12-figure code" (i.e., ones for packing various information into a sequence of 10 or 12 figures) were too troublesome, he admitted they would provide economy if properly used by "those who do not mind troubles". Thus, the three code words in a ten-letter code group can represent 3*4=12 digits, each of which (or groups thereof) could be assigned various meanings.
This is similar to the Oriental above, providing three-letter code words ("ciphers"), each assigned a check digit. The last digit of the sum of check digits for three code words is converted to a letter by the correspondence: 0(A) 1(B) 2(C) 3(D) 4(E) 5(F) 6(G) 7(H) 8(I) 9(J). It also provdes for use of code words to mean a four-digit number ("code number") assigned to it. It also provides tables to represent various information in two or four figures. It further explains use as a "twelve figure code", whereby the three code words in a ten-letter code group represent 3*4=12 digits, each of which (or groups thereof) could be assigned various meanings, with the tenth check letter representing the last digit of the sum of the check figures of the three code words: 1(K), 2(L), 3(M), 4(N), 5(O), 6(P), 7(Q), 8(R), 9(S), 0(T).
One significant difference from the Oriental is in the mutilation table, which the author claims to be "completely different from the one so far devised." When it has been determined which code word is mutilated and what its check figure should be, three candidates for correct the code word can be easily found in a mutilation table, which shows relation between the first two letters (top section), the third letter (bottom section) and the check digit (left and right to the bottom section). Thus, one does not need to look through a terminational order as in the Oriental.
This is a revision of the Oriental above to take advantage of the new regulations, which took effect in 1934. "In this code, the number of ciphers have been redoubled, vocabulary triplicated." The author claims this revision achieves "raising the Three-Letter Code almost up to the level of the most discreetly compiled Five-Letter Codes, and yet saving more expenses for the users."
A new check system was adopted. Each code word is given a check letter ("Cipher-Check") instead of a check digit. Three check letters are used to determine a check letter for a ten-letter group ("Word-Check") by referring to a table. A table similar to the mutilation table of the Paramount above is used in determining the "Word-Check" as well as for error detection/correction.
Richard Schofield published many codes in Japan, including a three-letter code as early as 1927. The following codes are listed at Google.
Schofield's Eclectic Phrase Code (1914, Kobe). A reprint by C. Bensinger Company in 1920 is also listed. The 6th reprint in 1925 from the original 1914 edition refers to "pirated editions of this work in the United States" and claims international copyright secured in 1919 and United States copyright secured in 1920. The code words are five-letter pronounceable groups. A mutilation table (called the "Analysis for tracing mutilations") is provided as a foldout after the Introduction. Two-letter differential is implemented. It refers to previous works: Schofield's 13-Figure Telegraph Code (2nd ed., 3rd ed.), Schofield's Symbol-Check Cipher Code (13 Figs.), and Schofield's 12-Figure Code.
Schofield's 3-Letter Code (1927, Kobe)
Schofield's 100,000 Cyphers: Being an Extension to Schofield's Eclectic Phrase Code and China & Japan Supplement (1928, Kobe)
Schofield's Brussels Supplement (1929, Kobe). Probably, "Brussels" refers to the conference held in Brussels in 1928 to discuss code words (see another article).
Schofield's 13-Figure Telegraph Code. Second edition. [With Extension Table] (1929, Kobe) (The 2nd edition (2 sheets) is found in the US Catalogue of Copyright Entries (Google) for the year 1916.)
Schofield's 1931 Supplement and 100,000 Cyphers, Etc. (1931, Kobe)
Schofield's Safe-Check 3-Letter Code [With Supplements] (1932, Kobe)
Schofield's Safe-Check 3-Letter Mutilation Table (1932, Kobe)
Schofield's 7-Figure Code, Improved (1933, Kobe)
Schofield's Second 3-Letter Code. Mutilation table (1936)
An article in Japanese (Baba p.195) describes a three-letter code published in 1935, which may refer to the 3rd edition of Schofield's Safe-Check 3-Letter Code found in CiNii. Similarly to other three-letter codes, 3 three-letter groups plus one check letter are combined into a ten-letter group (transmitted as two words because the length of a code word was limited to five letters as of 1935). The check letter is determined with a table according to the sum of the 3 check numbers assigned to the 3 three-letter groups.
It appears many code compilers provided a three-letter edition in this period. Considering Telling's three-letter code was one of the codebooks allowed by the Allies during World War II (see below), it seems to have received wide circulation. The following works by Harry George Telling are listed at Google.
The "Adaptable" Code Condenser and Error Corrector (1914, London)
Marconi Standard Piece Goods Code (1920, Marconi International Code Company)
The Marconi 6-letter (1920, London)
The "New Standard" Half-Word Code (1929, London) Probably "New Standard" refers to the new regulations of 1928.
Supplement to the "New Standard" Half-word Code (192?)
The "New Standard" Tricode (1931, London)
New Standard Three-letter Code (1934, London)
The "New Standard" Seven-Figure Condenser (1933, London)
Check Table So Arranged as to Detect Transpositions Occuring in Codewords (1934, London)
A. C. Meisenbach ("Acme"), the author of the Acme Commodity and Phrase Code (1923), also published a three-letter code among the subsequent works found at Google:
Meisenbach's Fourteen Figure Code, Etc. (1929)
Duo Banking and Commercial Code (Duo Code Directory) (1929)
The Multi-Code (1932)
Duo Master Code (1934) 38 pp.
Meisenbach's Three-letter Code System: Dedicated to the Furtherance of International Trading (1936) (Duo Code Company) 106 pp.
Meisenbach's Three-Letter Code System, Etc. (Revised Edition) (1936) 106 pp.
Meisenbach's Fourteen Workable Figure Code (1937, London)
"Ace" System and Mutilation Table for Five Character Codes (1938) by C.W. Hopkins, A.C. Meisenbach, Wm. J. Mitchel, 1 p.
"Ace" Complete Checking System for Three Character Codes (1938) by C.W. Hopkins, A.C. Meisenbach, Wm. J. Mitchel, 1 p.
Codigo Privado de Café Meisenbach (1938) (Duo Code Company), 14 pp.
Duo Master, Conversion Table 2 (1951)
Duo Banking and Commercial Code. Banking Supplement (1952)
An article in Japanese (Baba p.199) describes a two-letter code School Practice Code Books, Telegraphic Codes for use in schools (1938, Kyoto) by Takeo Hara. Every two-letter code group has a check number and a check letter for 2 two-letter groups is found by a table by using the two check numbers for the two groups as indexes to the rows and columns. The 2 two-letter groups and the check letter are combined to form a five-letter code word for transmission.
This may be the same as School Practice Code Books: Containing General Five-letter Code and Self-checking Two-letter Code (1938) (listed at Google), as well as the two-letter code in A text-book of telegraphic codes for use in schools: containing general five-letter code and self-checking two-letter code (1935).
The General Five-Letter Code (1935) is a small code of some 3000 entries. For secrecy, it proposes to use a five-letter code word ("code") obtained by substituting the five digits assigned to the original five-letter code word according to, e.g., 0(A), 1(B), 2(R), 3(I), 4(D), 5(G), 6(M), 7(E), 8(N), 9(T).
This is also a three-letter code with a checking system. That is, three of such code words were to be combined together with one check letter and sent as two five-letter code words. (From 1934, five-letter code words were the norm.) The check letter was derived from the sum of the last two figures of the numbers assigned to the three code words. The author claimed "Misunderstandings are ABSOLUTELY EXCLUDED because the nearest last two figures of each of the three phrases forward or backward are of at least two different letters out of three." (Bellovin, p.8)
This also combines 3 three-letter code groups plus a check letter into a ten-letter group but, unlike others, places the check letter at the beginning of the ten-letter group (Baba p.200). The author, Theodorus Reumerman, appears to have many related patents (e.g., US2646923). A page image is found at jmcvey.net.
This was published by C. Itoh & Co., now Itochu Corporation (Wikipedia) and seems to be a revised edition of an earlier code.
It contains about 16,000 three-letter code words ("ciphers") in Section 1 (general words and phrases), Section 2 (commodity names, proper names, technicals and blanks for supplementary uses), and Section 3 (numerals, quantities, etc.). In addition, it also has about 6,000 four-letter "code words" with serial numbers "for colour, design No. or any other supplementary uses", apparently for encoding numerals.
The author says "the arrangement of 3 letters has been fundamentally improved" such that ciphers such as NER and NRE do not have the same check figure and that the ciphers for the numbers etc. have at least two-letter differential from each other.
Every three-letter code group has a check number (0-25). A check letter for 3 three-letter groups is found by a table according to the sum of the three check numbers for the 3 three-letter groups. The 3 three-letter groups and the check letter are grouped into 2 five-letter groups for transmission.
This is a fourth edition of an earlier code.
Every three-letter code group has a check number (0-25). A check letter for 3 three-letter groups is found by a table according to the sum of the three check numbers for the 3 three-letter groups. The 3 three-letter groups and the check letter are grouped into 2 five-letter groups for transmission.
The code words are given a serial number ("auxiliary figure") 0000, 0001, ... and thus could be used to represent numbers.
Takeo Hara's two-letter code above was for school practice. But Sakai Kashichi published a two-letter code for actual use. Since the vocabulary is limited, the author intended this for "exporters and importers' daily use only" rather than for general commercial use.
Every two-letter code group has a check number (1-26) and a check letter for 2 two-letter groups is found by a table according to the sum of the two check numbers for the 2 two-letter groups. The 2 two-letter groups and the check letter are combined to form a five-letter code word for transmission.
The code words are given a serial number 000, 001, ... and thus could be used in "numerical sense."
The code (with all blank entries) had been printed in Sakai (1939), p.159-178.
When World War II broke out in September 1939, the Allies prohibited use of any codes. However, commercial codes were already an indispensable part of business and at the end of December it was permitted to use Bentley's Complete Phrase Code, Bentley's Second Phrase Code, the ABC Code (6th ed.), and Peterson's Code (3rd ed.) (Kahn p.516; The Engineer, 26 January 1940, p.98 "Use of Code in Overseas Telegrams"). In April 1940, five more codes were admitted: Acme Code and Supplement, Lombard General Code, Lombard Shipping Code and Appendix, New Standard Half Word Code, and New Standard Three Letter Code. (Kahn p.516)
An actual telegram encoded with Peterson's International Code (3rd ed., 1929) sent from New York to Tel Aviv on 5 June 1948 (only a few weeks after the establishment of the state of Israel was declared in Tel Aviv on 14 May) is presented in Klausis Krypto Kolumne.
After World War II, use of code decreased drastically. Rapid change of the world quickly made codebooks obsolete. After 1950, few commercial codes were compiled. (Kahn p.850) The few exceptions may include a postal banking codebook (1968) issued by the Australian Postmaster-General and an operational codebook (1972) by the Victorian Railways (Bellovin p.27).
telegraphic codes and message practice
Telegraph Code Books lists various codebooks from 1763 to 1957 (accessed in September 2013). The following gives an extract of those from 1877 to 1908 which are not described above. While the present author added the year (and some annotations) and sorted them in chronological order, it should be noted that the cited publications are not always the first edition.
1877: Telegraphic cypher (Julius Büttner) [This is a small code book of 44 pages, which assigns English words to numbers (including consecutive numbers up to 100) and phrases.]
1878: Private telegraphic code (Chesebrough, A)
1878: Cotton telegraph code (Chesebrough, A)
1878: Telegraphic cipher code, especially adapted to the cotton trade (Shepperson, Alfred B) ["renewed 1906"]
1878: Revised and improved edition of the telegraphic cipher code especially adapted to ... (Shepperson, Alfred B., d. 1911.) ["extended 1920"]
1879: Private cable code for the timber trade (Price and Pierce)
1879: J.R. Foster's private telegraphic code, covering general business transactions, for the use ... (Foster, J. R)
1879: Private Telegraphic Code of Heath & Finnemore, Produce & Commission Merchants
1879: Telegraphic Codex (Theodore Hunter)
1879: The Phillips telegraphic code for the rapid transmission by telegraph of press ... (Walter Polk Phillips)
1880: Cypher code compiled by messrs. Phipps & co. for their own use ... (Phipps and co)
1880: Law's mercantile cipher code for forwarding business communications by telegraph, telephone or ... (W.A. Law)
1880: Private telegraph code of Hamilton, Fraser & co (Hamilton, Fraser and co)
1881: Whittingham's skeleton telegraph code (Whittingham W.B. and Co) [It provides "code words representing fractions, weights, numbers, pounds sterling, merchants' orders, every date in the year, &c." as well as "blank spaces and extra lines for 4,148 special messages." Code words are English dictionary words, "revised under the latest International Regulations."]
1881: Private telegraphic code with James Adam, son & co (Adam James son and co)
1881: Wilson's ship broker's telegraph code (Wilson Effingham and co)
1882: Telegraphic code to insure privacy and secrecy in the transmission of telegrams (Frank Miller)
1883: The Globe telegraph code (E. Garsin)
1883: Telegraph code (Preston, Kean & Co) [A small code of 40 pages for representing trading expressions with code words (English words).]
1884: Cipher and secret letter and telegraph code, with Hogg's improvements. The most ... (Larrabee, Charles S.)
1884: Cypher code for telegraphy (Thomas S. Dunn)
1885: Telegraphic cipher (Cameron, Amberg)
1885: Cable code (Crossley John and sons)
1885: The telegram formula and code combiner (Frederic George McCutcheon) [The "Telegraph Formula" provides rules to represent information with letter groups indicating subjects and sentences classified under them. Such letter groups are to be translated to code words ("cipher words"). In one example, there are subjects represented by A, B, C, ...; each subject has three columns (A, B, C) for three aspects related to the subject; and each column has phrases ("sentences") represented by A-Z. The user selects three subjects (e.g., A, B, D) and, for each subject, three phrases for three columns (e.g., M-M-G for subject A, G-L-I for B, and D-B-Q for C). The four triplets ABC, MMG, CLI, DBQ are translated to code words to be transmitted. The three-letter combinations are given code words and serial numbers (AAA Abacero 1 to ZZZ Wyvern 17576). Blank "Combination Code Forms" are available separately from the "Telegraph Formula". The user is supposed to "prepare his list of sentences, exhausting the probabilities with regard to any important event" (p.7). Separate supply of the forms facilitates updating of the code. At the end of the volume, a table is provided to express numbers 1-11,881,376 (i.e., 265) with five letters AAAAA-ZZZZZ, which is substantially a base-26 notation.]
1885: Telegraphic Code (Cornell University)
1886: Telegraphic Code (American florist)
1887: Telegraphic Code, The American Educational Catalogue (Michael Mok)
1887: Telegraph & Cable Code: In Use Between John Paton & Co., New ... (Benjamin Graham)
1888: The Science observer code (Chandler, Seth Carlo)
1888: Telegraphic mining code (Moreing, C. Algernon)
1889: A Brief Sketch (W. H. Miner)
1889: Proposed Cable Chess Code [see p.111]
1890: Chess Telegraphic Codes (Edwyn Anthony)
1890: The dynamic chess notation: whereby any possible move in a game of ... (F. Startin Pilleau)
1891: Private telegraphic code (Williams, Brown & Co)
1892: Cipher code of words, phrases, names of organizations and titles of their ... (Sheahan, W. A.)
1893: The United States Telegraphic Cipher (Joseph H. Wilson)
1893: Police Telegraph Code of England for 1893 (George Wesley Hale)
1894: The Adams Cable Codex (Adams, E.A. & Co)
1894: Coffee. Private cipher code (Castle Brothers)
1894: Inter-state cipher (Pratt, Harmon K)
1894: H. & W. Pataky's telegraphic code for use in obtaining and negociating ... (H. & W. Pataky)
1895: Barnard's universal cipher code (Shock, Floyd; Geo. D. Barnard & Co)
1896: Publications of the Astronomical Society of the Pacific, Volume 8 (Astronomical Society of the Pacific)
1896: The Adams cable codex (F.O. Houghton & Co)
1896: Atlas, The, universal travellers' and tourists' telegraphic cipher code (Thomas Walter Hartfield)
1897: The Robinson telegraphic cipher (Steven Lynn Robinson)
1897: Private Telegraphic Cipher Code (Kerr, Gordon & Co.)
1906: Travellers' Telegraphic Code (Edmund C. Stedman, Thomas L. Stedman)
1906: Telegraphic Cipher Code, Gerrish System (Willard Peabody Gerrish of the Harvard College Observatory) [This was devised for telegraphing numerical data of astronomical announcements. After forming the data into five-figure groups, each figure is converted to syllables according to 1(ba) 2(de) 3(fi) 4(go) 5(ku) 6(am) 7(en) 8(ip) 9(ot) 0(ux), forming ten-letter code words.]
1906: The standard cipher code of the American Railway Association for the use ... (American Railway Association) [This uses dictionary words rather than five-letter artificial words.]
1907: Standard Lumber Reference Book and Code (Benjamin F. Ulmer)
1907: Postal Code (Telegraph-Cable) (Frank Shay, R. V. Dey)
1907: Un Code Telegraphique Du Portrait Parlé (R. A. Reiss) ["Portrait parlé" ("spoken portrait") is a method for describing features of a person. It was invented by Alphonse Bertillon and compiled in 1893. Page images are found at codescans. Reiss had authored Manuel du portrait parlé à l'usage de la police avec vocabulaire français, allemand, italien et anglais (1905) (Gallica).]
1908: Wellcome's Excerpta Therapeutica (Burroughs Wellcome and Company) [Pages 151-153 provide "telegraphic cable code roots" for ordering therapeutic sera, tuberculins, or vaccines. The roots are to be combined with "terminations" for indicating quantities and sizes (e.g., "six phials, regular size"). Both roots and terminations are five-letter artificial words, which form ten-letter code words for sending.]
MIT Libraries' catalog, Vail Collection.
Richard Brisson, Books on Telegraphic Codes and Signals
Jim Reeds, Commercial Code Book Database
British Books in Print for 1888 (Google)
Friedlich L. Bauer (2002), Decrypted Secrets: Methods and Maxims of Cryptology Translation of Entzifferte Geheimnisse: Methoden und Maximen der Kryptologie (2000). Page 75 (of the English edition) was a helpful lead to non-English materials.
Steven M. Bellovin (2009), "Compression, Correction, Confidentiality, and Comprehension: A Look at Telegraph Codes", Preliminary version (pdf; his later slides (August 2009) pdf)
Rosario Candela (1938), "The military cipher of commandant Bazeries" (HathiTrust)
Joseph S. Galland (1945), An historical and analytical bibliography of the literature of cryptology
LANAKI (1996), Classical Cryptography Course, Lecture 20 Codes
"La cryptographie militaire", Journal des sciences militaires, February 1883 (online)
Abe Akira, Gaikoku Boeki Shoyo Ango Nyumon (1949)
Baba Makoto, "Boeki Gijutsu toshite no Denshin Ango", Shogyo to Keizai (1940) p.195 ff. (Digital Library from the Meiji Era)
Ishikawa Bungo, Shogyo Eisakubun Kogi (1908) (Digital Library from the Meiji Era)
Kobayashi Keishi, Nihon Denshin Ango Jibiki (1896) (Digital Library from the Meiji Era)
Nakagawa Shizuka (1916), Shinsho Seikan (Digital Library from the Meiji Era)
Sakai Kashichi (1939), Keiburu Kuraaku Tokuhon (Cable Clerk Reader) (Ditigal Library from the Meiji Era)
Ueda Kotaro, Gaikoku Kawase to Denshin-Ango (1895) (Digital Library from the Meiji Era)
See also another article, "Telegraph Regulations and Telegraph Codes" by me.