Book code was John Jay's preferred form of secret communication for some period during the American Revolution (see here for a perspective of various codes and ciphers used by him).
A book code encodes a word by the page and the line where the word occurs in a book (a dictionary is typically used because it facilitates finding a word to encode). When a page consists of two or more columns, the column is indicated by a letter ("a", "b", ...), an underline, an overdot, an overline, or the word position may be reckoned in the whole page without regard to columns. Occasionally, correspondents arranged that some offset should be added to the page and/or line numbers. This provides many variations of book code.
One group of book codes was based on Entick's New Spelling Dictionary. The 1777 edition used by Jay had 468 pages. Presumably, each page usually included 38 lines (this is assumed because the 1790 edition I have includes usually 38 and sometimes 39 lines per page and this assumption is largely consistent with the line number reversal discussed below.)
At least the seven letters listed below use Entick's Dictionary. All these letters are with translation.
In all of these letters except <AUG14>, the code is written in the form 17 a 243, where "17" indicates the line, "a" the column, and "243" the page. These are designated as variants of WE080 by Ralph E. Weber. <AUG14> puts it in the form 243 17 with an overdot indicating the column (an overdot on the first letter indicating the first column, etc.) and receives designation WE083. Jay proposed another scheme of the form 17 224 with an overdot, designated WE082, but, apparently, this was never used.
In this article, a standard form of 243a17 (page-column-line) will be used, regardless of the convention followed in the manuscript.
The seven letters are as follows.
<FEB15>15 February 1780 (Carmichael to Jay) ... PCC Roll 116, Page 51
The page ranges from 38 to 426 and the line from 2 to 37. The column is indicated by "a" or "b".
<FEB18>18 February 1780 (Carmichael to Jay) ... PCC Roll 116, Page 55, Jay Papers ID 90261
The page ranges from 218 to 606 and the line from 12 to 48. The column is indicated by "c" or "d".
<FEB26>26 February 1780 (Carmichael to Jay) ... Jay Papers ID 12758
The page ranges from 41 to 426 and the line from 1 to 38. The column is indicated by "a" or "b".
<FEB29>15 February 1780 (Carmichael to Jay) as enclosed in 29 February 1780 (Jay to the President of Congress) ... PCC Roll 135, Page 39
The page ranges from 219 to 606 and the line from 12 to 49. The column is indicated by "c" or "d".
<FEB25>25 February 1780 (Jay to Carmichael) as enclosed in 29 February 1780 (Jay to the President of Congress) ... PCC Roll 135, Page 43
The page ranges from 220 to 606 and the line from 12 to 49. The column is indicated by "c" or "d".
<MAR03>3 March 1780 (Jay to the President of Congress) ... PCC Roll 135, Page 50
The page ranges from 216 to 606 and the line from 12 to 49. The column is indicated by "c" or "d".
<AUG14>14 August 1782 (Bingham To Jay) ... Jay Papers ID 7493
The page ranges from 65 to 446 and the line from 11 to 47.
Observations of the range of page and line numbers suggest that the codes used in these letters form roughly three groups: a first group (<FEB18> <FEB29> <FEB25> <MAR03>) seems to have its page numbers offset by about 180 and line numbers by about 10 relative to a second group (<FEB15> <FEB26>). <AUG14> seems to have its page numbers offset by about 20 relative to the second group.
<FEB15> and <FEB26> have several common codes (e.g, 099a17 for "Congress", 108a10 for "court", 233b29 for "majesty", 243a17 for "minister").
<FEB25> and <MAR03> have common codes (e.g., 241c35 for "because", 244c48 for "between", 255c18 for "by").
<FEB18> and <FEB29> have common codes (e.g., 368d20 for "he", 377d39 for "I", 395c35 for "introduce", 400c44 for "king").
Thus, the seven letters may actually be grouped into four groups instead of three.
Jay's explanation to Bingham on 8 September 1781 (see below) tells us that 20 should be added to the page numbers and 10 to the line numbers. Subtracting these offsets from the codes of <AUG14> gives many matches with codes of <FEB15> and <FEB26>. For example, "court" is encoded as 128a20 in <AUG14> and 108a10 in <FEB15> and <FEB26>. "force" is encoded as 186b42 in <AUG14> and 166b32 in <FEB15>. Thus, the page and line numbers of <FEB15> and <FEB26> can be considered to include no offset.
The letter of 15 February (<FEB15>) is encoded differently in the enclosed version (<FEB29>) to the President of Congress. Apparently, Jay re-encoded Carmichael's letter for correspondence with the President of Congress. This allows one to compare the original letter and the re-encoded letter to conclude that the page offset is 181 and the line offset is 11 for <FEB15>.
The same offset relation holds between <FEB26> and <FEB18>. For example, the word "latter" is encoded as 223a34 in <FEB26> and 404d45 in <FEB18> and the word "told" is encoded as 378a07 in <FEB26> and 559d18 in <FEB18>. Both instances nicely fit the page offset of 181 and line offset of 11.
Page numbers are the same between the <FEB25><MAR03> group and the <FEB18><FEB29> group. For example, "the" is encoded as 554d34 in <FEB25> and 554d27 in <FEB29> and <FEB18>. "Their" is encoded as 554d28 in <MAR03> and 554d33 in <FEB18>.
When closer observation is made, the difference in line number between "the" and "their" is 6 in both cases but the order is reversed. In the <FEB29><FEB18> group, the code for "the" is lower by 6 than the one for "their", as expected. In the <FEB25><MAR03> group, however, the code for "their" is lower by 6 than the one for "the". This implies that in the <FEB25><MAR03> group, the line number is counted from the bottom of the page.
Even if one is not lucky enough to find such two pairs of instances on the same page, close observation of the word sequence in the <FEB25><MAR03> group would lead to the same conclusion. That is, note the following pairs:
Thus, words on the same page are consistently arranged in the reverse order in <FEB25> and <MAR03>. This would suggest that the line number is counted backwards from the bottom
As noted before, it is believed that there are usually 38 lines on a page. Further, the observation (iv) shows the line numbers of the <FEB29><FEB18> group are offset by 11 and the observation (i) suggests the line numbers of the <FEB25><MAR03> group are offset by the same amount.
Thus, if a line number indication of the <FEB25><MAR03> group is X, it would indicate the (X-11)th line from the bottom. Counting from the top, this translates to 38-(X-11)+1=50-X. Conversely, if a word is the Yth word from the top, the line number indication would be 50-Y in the <FEB25> or <MAR03> scheme.
In case there are 39 lines on a page (or column), the formula would be 51-X instead of 50-X.
By closer observation, subtler disordering is found for the <FEB18><FEB29> group and the <FEB25><MAR03> group.
These instances imply that the "d" column should precede the "c" column in these groups. That is, "c" corresponds to "b" (the second column) and "d" corresponds to "a" (the first column).
Shrewd readers may have noticed this in the examples of "latter" and "told" in (iv) above.
Thus, for example, the word "Congress" is encoded as 099a17 in <FEB15> and <FEB26> but 280d28 in <FEB29>, which, after removing the page offset 181 and line offset 11 and replacing the column designation, reduces to 099a17. In <MAR03>, the same word is represented as 280d33, which, after removing the page offset 181 and applying the line number reversal, reduces to, again, 099a17.
An earlier letter dated 3 January 1780 from Gouverneur Morris to Jay includes code, which looks dissimilar to what has been described up to now but close inspection would suggest it is a book code based on Entick's Dictionary.
Interlinear translation available on the manuscript indicates words are encoded as a code consisting of two numbers. For example,
Closer inspection reveals the following characteristics:
(i) The ordering of the first number of the code is the same as the alphabetical order of the encoded word.
(ii) The first number ranges from 68 to 491 and the second number ranges from 98 to 170.
These observations suggest that this is a book code, with the first number representing the page (possibly with some offset) and the second number representing the position of the encoded word in a page. This implies one page contains at least 73 (=170-98+1) words. This suggests that the word positions are counted regardless of columns of a dictionary. Further, observations of words bearing the same page number show that the word position is counted backwards.
As it turns out, the page numbers are consistent with the pages of Entick's Dictionary with +23 offset.
Since Entick's Dictionary has usually 76 words in a page, the second number has an offset of at least +94 (=170-76) and at most +97 (<98). Comparison of common codes suggests the line offset of 95.
If a word is on the Ath line in the first column, the second number of this letter would be 76-A+1+95(offset)=172-A. Conversely, if the second number is x (>=134), A=172-x.
If a word is on the Bth line in the second column, the second number of this letter would be 38-B+1+95(offset)=134-B. Conversely, if the second number is x (<134), B=134-x.
Applying these formulae, the codes used in this letter are largely consistent with the book code based on Entick's Dictionary described so far.
Jay's letter to Gouverneur Morris on 2 March 1780 (see below) includes instructions of a book code. According to this scheme, the page offset is 200 and the line offset is 50. The page is placed after the line number. The column is designated by a bar (-) over a figure (a bar over the first figure representing the first column, etc.).
It is noted that these offsets 200 and 50 are simple compared to the induced offsets of 23 and 95 above. Jay may have been tired of complicated maths and proposed a simpler scheme.
Similar change may have occurred in a channel with the President of Congress. As seen above, the code used in Jay's letter of 3 March 1780 to the President of Congress (<MAR03>) used a page offset of 181 and a line offset of 11. In this very letter, however, Jay wrote "I do not like the cipher in which I write, and shall therefore defer further particulars till Mr. Thomson shall receive the one now sent him." The code sent to Thomson (based on Boyer's French Dictionary; see below) involves a page offset of 5 and a line offset of 10.
Arthur Lee is also known to have used book code based on Entick's Dictionary. He used Roman numerals to indicate line numbers. Thus, his code 243,a,xvii corresponds to 243a17 in our standard form.
Some code numbers used by Lee fit those used by Jay (after removing Jay's offsetting etc.). From the examples as below, one may consider Lee used page and line numbers without offset.
At least, the identification of 059a08(base) may be supported by the fact that the word base is 13 words before basis in the 1790 printing of the dictionary I have.
(Note: The author afterwards realized that Lee's letter of 27 August 1778 uses the same book code and is accompanied by translation. Its codes such as 082b05(channels), 099a17(Congress), and 229a37(loan) are the same as in Jay's letters. While there are some mismatches, most of the other codes are consistent with Jay's. Thus, Lee is considered to have used the same Entick dictionary without offsets.)
Apparently, Lee's correspondents in Philadelphia did not know the code. On 17 September 1779, Lovell wrote to Lee "Something in your letters to the Committee was in ciphers. Merryweather [Meriwether] Smith deciphered it readily. Otherwise I must have sent it on to Virginia to your brother. I know that R H L [Richard Henry Lee] thinks that mode is now as useless as if Deane knew it [Smith was a Deane partisan]; need I say more? I think it would be clever to have a cipher in Committee, but not to be also used to your brother or your friend; those should be distinct ones.".
Back in June 1776, Arthur Lee had proposed another book code (see below), in which line numbers were indicated by letters of the alphabet. When the line number is greater than 26, doubled or trebled letters should be used.
Lee gives the following examples.
This letter of Lee was written before 1777 and Lee's codes do not fit Jay's scheme based on the 1777 edition of Entick's Dictionary.
Apparently, Lee, a London correspondent of the Congress, gave this book code to William Carmichael, who was to be Jay's secretary in Spain. Carmichael was living in London at the time and, with a pocket dictionary given by Lee, went over to Paris as an assistant to Silas Deane. He returned to America in February 1778. Carmichael wrote of the book two years later.
Francis Dana, American minister to Russia, proposed to John Adams a book code based on the 1772 edition of Entick's Dictionary in his letter dated 18 October 1782.
Like Jay's, Dana's proposed use was not straightforward. A page number should not indicate the actual page but that of the opposite page. The columns should be numbered 1 to 4 from right to left over two facing pages. The line number was counted from the bottom.
It is interesting to note that the British army sometimes used a book code based on the same Entick dictionary (WE084). They simply put the page, the column, and the line in this order: 115.1.4.
They are also known to have used Nathan Bailey's English Dictionary (21st and 25th editions) and Blackstone's Commentaries on the Laws of England (5th edition).
Aaron Burr used a Wilmington edition of Entick's Dictionary of 1800. In his scheme, the first number indicated the page and the second number indicated the line on which the word appeared. To indicate the second column, a mark above or below the second number or "2" over the second number was used.
Jay's letter of 3 March 1780 (<MAR03>) also uses simple substitution cipher. Such an auxiliary scheme may be necessary when a word not found in a dictionary need to be spelled. However, this letter uses the substitution cipher for enciphering normal words such as courts and my.
From this letter, the following substitution table can be reconstructed.
Asterisks (*) denote where cipher could not be identified because corresponding letters do not occur in this letter.
Bingham's letter of 14 August 1782 (<14AUG>) uses similar simple substitution (WE076).
a b c d e f g h i j k l m n o p q r s t u v w x y z
n m l k i h f i e d c b a l y v x t u r p w s z o y
It is noted there are two i's, two l's, and two y's but no g, j and q in this substitution table. Apparently, this is not a printer's error. From Bingham's letter, it can be verified that both h and e are encoded as i (e.g, Ntript(Arthur), Lntbiryl(Carleton)), both z and o are encoded as y (e.g., Eyntk(Izard), Inuidlfryl(Washington)), and both c and n are encoded as l (e.g., U. Lntybdln(S. Carolina)).
Further, it is noted that plaintext i is encoded as d or e (e.g., Wdtfdldn(Virginia), Lntbiryl(Carleton)). At the time, letters i and j were not always distinguished.
Gouverneur Morris' letter of 3 January 1780 employs single numbers to represent single letters (see the above image). For example,
209 (H) 69 (A) 411 (V) 70 (A) 275 (N) 278 (N) 72 (A) 210 (H)
It is immediately noted that a single letter is given several different numbers. Upon close inspection, these numbers are seen to fit the sequence of codes arranged in the alphabetical order of the coded words.
Thus, page numbers without line numbers represented single letters. To distinguish such use from usual two-number codes, a dot is placed under the code representing a single letter.
Arthur Lee's letter of 3 June 1776 explicitly states endings such as ing, ed, and s must be added "when necessary". Such convention is also observed in Jay's letter to the President of Congress dated 28 January 1781. It is also seen in letters in code other than book code.
Another dictionary used for coding was Abel Boyer's French Dictionary. The following letters use book code based on this dictionary.
<NOV06> 6 November 1780 (Jay to the President of Congress) (PCC, Roll 135, Page 309) *This is a very long letter and the link refers to the page containing the code, which is part of the included note about a conference with Floridablanca on 23 September 1780. Printed text of Sparks is here.
<JAN28> 28 January 1781 (Jay to the President of Congress) (PCC, Roll 117, Page 361; Jay Papers ID 7752 (with translation), ID 11908, ID 13296)
<JUN16> 16 June 1780 (Committee for Foreign Affairs to Jay) (PCC Roll 72, Page 127, Wharton (plaintext))
The <NOV06> and <JAN28> letters share common codes such as 371a12(of), 458b12(secret), 510a40(the) and the <NOV06> and <JUN16> letters share common codes such as 557a24(would) and 366a14(not). While there are some mismatches such as 514c11 <JUN16>, 514c12 <JAN28>, 514c13 <NOV06> (to) and 555a11 <NOV06>, 555a12 <JAN28> (without), the codes used in these three letters are largely consistent with each other. They may be considered to follow the same scheme except that <NOV06> puts pages before lines whereas the order is reversed in <JAN28> and <JUN16> as in WE079. Thomson's draft in May below may also follow this adapted scheme of his.
There is another manuscript (by Charles Thomson, dated 19 May 1780) that may follow the same scheme (see the image below). (More accurately, the overline indicates it follows the variant by Thomson's 7 June letter.) Occurrence of the underline at the third figure suggests the dictionary used in this letter had three columns in a page. The date and the author suggest that this letter uses the same WE079 as above, which was proposed by Jay to Thomson on 29 February 1780 and of which Thomson acknowledged receipt as of 7 June (Letters of Delegates).
Using the known codes from the above three letters, one may feel making some sense of it.
... 057b18 [cf. 057b19(bills)] 045c11 [cf. 045x13(be)] immediately 173b18 [cf. 173c16(drawing)] under the direction of 519c18 [cf. 520a36(treaty)] ...
But Franklin 221a11 [cf. 221b11(forbear)] 25000 170b16 [cf. 170b23(dollars)] 024a11 [cf. 024a18(and)] ...
|3 January 1780 (Gouverneur Morris to Jay)|
|15 February 1780 (Carmichael to Jay)
26 February 1780 (Carmichael to Jay)
|18 February 1780 (Carmichael to Jay)
15 February 1780 (Carmichael to Jay) as enclosed in 29 February 1780 (Jay to the President of Congress)
|25 February 1780 (Jay to Carmichael) as enclosed in 29 February 1780 (Jay to the President of Congress)
3 March 1780 (Jay to the President of Congress)
|(instruction) 2 March 1780 (Jay to Gouverneur Morris)|
|(instruction) 19 November 1780 (Jay to Robert Morris)|
|(instruction) 8 September 1781 (Jay to Bingham)
14 August 1782 (Bingham to Jay)
|243 a xvii||page-column-line||+-0||+-0||a/b||WE086|
(instruction) 3 June 1776 (Arthur Lee to the Secret Committee of Correspondence)
4 October 1777 (Arthur Lee to R. H. Lee)
12 May 1778 (R. H. Lee to Arthur Lee)
19 May 1778 (R. H. Lee to Arthur Lee)
27 July 1778 (R. H. Lee to Francis Lightfoot Lee)
27 August 1778 (Arthur Lee to Committee of Foreign Affairs)
31 August 1778 (Arthur Lee to Committee of Foreign Affairs, quoted in Lovell to R. H. Lee)
16 September 1778 (R. H. Lee to Arthur Lee)
27 October 1778 (R. H. Lee to Arthur Lee)
11 February 1779 (R. H. Lee to Arthur Lee)
21 March 1779 (R. H. Lee to Arthur Lee)
29 March 1779 (Arthur Lee to R. H. Lee)
4 April 1779 (R. H. Lee to Arthur Lee)
22 April 1779 (Arthur Lee to the Committee for Foreign Affairs)
27 April 1779 (Arthur Lee to R. H. Lee)
9 May 1779 (Arthur Lee to Schweighauser)
23 May 1779 (R. H. Lee to Arthur Lee)
12 October 1779 (R. H. Lee to Arthur Lee quoted here)
26 February 1780 (Arthur Lee to John J. Pringle)
|242.4.22||opposite page-r_column-r_line||opposite||reversed||4/3/2/1||(p. 52)|
|(instruction) Dana to Adams, 18 October 1782|
|Used by the British|
|(instruction) 19 February 1780 (Jay to Livingston)|
|(instruction) 29 February 1780 (Jay to Thomson)
?19 May 1780 (Thomson to Jay)
16 June 1780 (Committee for Foreign Affairs to Jay)
28 January 1781 (Jay to the President of Congress)
|6 November 1780 (Jay to the President of Congress)|
The manuscript of Thomson's reply on 7 June 1780 (mentioned above) to this letter is at Mount Vernon. The manuscript elucidates Thomson's proposed improvement of Jay's scheme: "I recd. your favor of 29 Feby. Would it not be convenient to count down as well as up distinguishing the former thus 186 and the latter 166." That is, while "16" in Jay's example 16.6 refers to the sixth line from the bottom (augmented by ten), one might also count the word from the top and write 186, with the overline designating the counting down from the top. Thomson actually uses the system in the subsequent portion of the letter, in which is a passage "I am afraid Ompnsub yjil rb 13 245." [Charles town is gone]. The substitution cipher used is the same as that in <MAR03> noted above.
Written in the cover of a dictionary.