Chain Letter Evolution

Abstract: Apocryphal letters claiming divine origin circulated for centuries in Europe. After 1900, shorter more secular letters appeared in the US that promised good luck if copies were distributed and bad luck if not. Billions of these "luck chain letters" circulated in the the next 100 years. As they replicated through the decades, some accumulated copying errors, offhand comments, and calculated innovations that helped them prevail in the competition with other chain letters. For example, complementary testimonials developed, one exploiting perceived good luck, another exploiting perceived bad luck. Twelve successive types of paper luck chain letters are identified which predominated US circulation at some time in the twentieth century. These types, and their major variations, are described and analyzed for their replicative advantage. In the 1970's a luck chain letter from South America that touted a lottery winner invaded the US and was combined on one page with an indigenous chain letter. This combination rapidly dominated circulation. In 1979 a postscript concluding with "It Works" was added to one of these combination letters, and within a few years the progeny of this single letter had replaced all the millions of similar letters in circulation without this postscript. These and other events in paper chain letter history are described, and hypotheses are offered to explain advances and declines in circulation, including the near extinction of luck chain letters in the new millennium.

Perhaps the most dramatic event in chain letter history was the advent of money chain letters. This was spawned by the infamous "Send-a-Dime" chain letter which flooded the world in 1935. The insight and methods of its anonymous author, likely a woman motivated by charity, are examined in detail in a separate article titled "The Origin of Money Chain Letters." This constitutes Section 4.1 below, where its link is repeated. It can be read independently from this treatise.

TABLES

I could not have conducted this study without the assistance and friendship of Dr. Michael J. Preston, University of Colorado English Professor and folklorist. He obtained scores of letters, gave me copies of his files and put me up in his home while I worked in the CU Boulder library. The help of Dr. William F. Hansen, folklorist and Head of the Department of Classical Studies at Indiana University was also indispensable. He provided many useful chain letters and translations, and his interest and encouragement have been sustaining.

Special thanks also go to Alan E. Mays, who sent many chain letters, his bibliography on chain letters and the Himmelsbrief, and archived chain email. Paul Smith also provided scores of letters and an extensive bibliography. Anna Guigne sent a stack of chain letters and answered questions. Steve Glickman helped with microfilmed Denver Post articles at UC Boulder. Carol Petty copied local newspaper articles in Springfield, Missouri, where chain letters rampaged for a few days in 1935. John Burkhardt shared his thoughts early in the project and emailed digitized letters. James H. Patterson has provided photocopies of many rare chain letters from his collection of "unmailable" items. Sandy Hobbs sent photocopies of every chain letter that has appeared in the publications Dear Mr. Thoms and Letters to Ambrose Merton.

I have received much needed help with foreign language chain letters. Prof. Sarah E. Winter translated several chain letters and an entire article from French into English. Dr. Yana VanArsdale found several Russian chain letters and articles, and translated published letters in Polish and Russian to English. Dr. Jean-Bruno Renard has sent chain letters from France and Brazil, and a bibliography of French publications. Natalia Kasprzak sent two Polish articles on chain letters and translated a Polish letter into English. Bill Clark translated some chain letter Tagalog. Martinovich Vladimir Aleksandrovich provided Russian chain letters he collected, and has translated a Russian version of the Romance Game chain into English.

Though I am solely responsible for the approach and presentation here, this effort was sustained because a few people expressed interest. I am especially thankful for the encouragement of Richard Dawkins, who suggested I write "a book on chain letters, with all your detailed examples and analyses." This is not a book, but likely it is enough detail for most readers.

A partial list of those who provided one or more paper chain letters appears on an information page for the archive.

Seeking paper chain letters.
If you have any information on where I may obtain paper chain letters please email. Any chain letters sent should be dateable, as by a postmarked envelope. Are there any paper luck chain letters still circulating, perhaps distributed by hand? The last one I collected was almost ten years ago, in 2008. Foreign examples, clippings, obscure or foreign references, beliefs and rumors about chain letters, stories of receiving unexpected money in the mail, or other personal experiences with chain letters are welcome.

Overview.
Texts that appeal to superstition to encourage their copying or publication have circulated for over a thousand years. For English language letters, beginning around 1905, copy quotas and deadlines appeared and claims of divine authorship and magical protection were removed. These innovations probably began in other languages and were translated into English. The resulting "luck chain letters" eventually spread worldwide, and in over four thousand generations of copying (with variation) they accumulated ways to sustain and increase circulation that challenge our understanding.

Using a collection of over 900 dated paper chain letter texts, I have identified types and variations that appear and disappear over the years. Unexpectedly, it was discovered that, repeatedly, a single letter bearing some new innovation had propagated so abundantly and rapidly that within just a few years its descendants replaced all similarly motivated letters in circulation.

Auxiliary Files and Conventions.
Listed here are files in the directory /chain-letter/ and sub-directories /archive/, /e-archive/ and /photo-archive/ which support this essay and are publicly available.

A chain letter explicitly asks a recipient to make or purchase copies of itself and distribute them. It may also instruct the reader to make some modification of the letter, such as updating a list of senders. In this treatise I will use the term "chain letter" exclusively to refer to paper chain letters.

Examples reveal that the form and content of chain letters are highly correlated with the principal motive to distribute copies. I have classified each paper chain letter in the archive into one of nine motivational categories which I define here. Three of these categories (Protection, Luck, and Money) are described in detail in following sections and hence only briefly here. The order of the categories here is the chronological order that English language examples first appeared.

Protection.
The Letters from Heaven (German: Himmelsbrief) claim to have been written by God or some divine agent. They often command Sabbath observance and promise the bearer magical protections.

Himmelsbrief have circulated in Europe and elsewhere for many centuries. They do not exactly fit the above definition of a chain letter since most do not ask that copies be made, but instead ask the reader to "publish" the text. I discuss them later (> Heaven).

The filenames for the Letters from Heaven begin with the letter "h" in the Paper Chain Letter Archive.

Charity.
A charity chain letter requests money or some item be sent to a fixed address, ostensibly for charitable, political or memorial purposes.

Charity letters were common from 1888 up into the 1920's, and influenced early luck chain and money chain letters. Apparently 1888 was a boom year for them, judging from newspaper reports. There was even a parody that circulated [1888]. A June 1887 newspaper article found by Patrick Davison describes a "remarkable scheme" for collecting donations by personal contact which uses a pyramid of 6,144 persons to collect $17,412. Participants were assigned one of the six letters A through F depending on their role in the scheme. Early charity letters may have been influenced by such schemes.

A December, 1888 letter in the archive solicits dimes for the education of "the poor whites in the region of the Cumberlands." This letter states it is an adaption of a previous solicitation, and asks that four copies be sent to friends. For compliance ". . . you will receive the blessing of Him who was ready to die for us". Excluding the Himmelsbrief, this may be the oldest chain letter physically collected. An older charity chain letter from the summer of 1888 is described by Paul Collins, and likely some others circulated previously. A report of an 1881 charity chain letter in the Washington Post is apparently false. In an 1892 example, an American college student solicited dimes and ten copies. This letter, like most early charity chains, claimed to be self-terminating: recipients were asked to increment a generation count at the top of the letter until it reached some preset maximum at which time the donation was to be made, but not more copies. This practice continued at least through 1916 [Billy]. Usually, a few years after a letter was launched, only those circulated which had inflated this maximum (NYT 1917). For example, there are two examples of a solicitation for used postage stamps to build a children's ward in Australia (OED). The first is from 1900 and is numbered 173 of 180 maximum. The second, highly modified, was still in circulation ten years later [1910] and is numbered 375 of 480 maximum. Many chain letters exaggerate the loss if there is a single break in transmission [1895]. Apart from intimidating recipients to comply, this may have been influenced by certain mail frauds of the time (Thomas 1900). Chain letters that did not state a termination number were called "endless" for a few decades, and this language still appears in some laws.

In 1989 the Craig Shergold appeal requested get well cards for a dying child (since recovered), intending to break a Guinness world record that existed at the time. It was launched by FAX, email and chain letters. By December 1990 a record 33,000,000 cards had been received (Guigne). Despite efforts to stop the appeal, hundreds of millions have now been sent. Charity chain letters were an influence on early luck chain letters, and in 1935 enabled the advent of money chain letters. They are common on the Internet but most of these are hoaxes {Jessica Mydek}. A revealing item in the archive is a nine page chain solicitation for one dollar contributions to the 1950 campaign of anti-union Ohio Senator Howard Taft. These were rescued from the discarded files of the Atlantic Coast Line railroad police.

Religion.
Religious chain letters promote religious beliefs, causes or practices, but do not ask for money. If they do they are classified as Charity chain letters.

In English speaking countries, religious chain letters circulated in small numbers throughout the twentieth century. Most of these have Roman Catholic themes. There is a single example in the archive of a chain letter which is titled "A Prayer to St. Joseph" which dates back to 1898. The text follows (format shortened, slightly edited):

Note that only five copies are requested, but the prayer is to be repeated on nine successive days as in a Novena devotion. In the last paragraph it is revealed that the sender may have made a personal "request" of Joseph, who is described as the "patron of despairing souls." This and other features, including the claim that "it has never been known to fail," suggest this letter may be a distant source for personal appeals to St. Jude that appear in the classified ads of present day newspapers in the U. S. (>jude) St. Jude also appears on subsequent English language luck chain letters beginning around 1987 (much later than some have supposed). One appositive for St. Jude is "Saint of things almost despaired of."

Other religious chain letters that have been collected include a solicitation for prayers [1905], and Catholic devotional themes, one of which is called "A Little Flower of Jesus" and claims to be approved by "the sisters of St. Francis" [1937, see also 1951].

Filenames for religious chain letters begin with an "r" in the Paper Chain Letter Archive.

Advocacy.
Advocacy chain letters promote some cause other than religion, and do not ask that money be sent. Often they involve a petition. Also included in this category are announcements and invitations.

A 1903 postcard, as well as asking that copies of itself be distributed, asks that recipients send their name and address to the "U.S. Moral Society" to be added to a petition to Congress to prohibit the sale of cigarettes to minors. In subsequent examples the initial communication itself could be a petition, as in an attempt to draft Calvin Coolidge as the Republican nominee for President [1927]. An example not involving a petition is an August, 1940 letter advocating Republican Wendell Willkie for President and asking that ten copies be sent. A 1917 chain letter with detailed instructions for establishing conscientious objector status is a rare example of anonymous advocacy. Other chain letter causes include Czech independence [1949], nuclear disarmament [1985], protests of apartheid [1988], and a libelous call for a boycott of Proctor & Gamble [1986] alleging satanism. Recipients are invited to a party, and possibly a suicide, in a 1937 chain letter. Advocacy chain emails are also common, such as a perennial appeal to support National Public Radio [e1996].

Luck.
Luck chain letters appeal primarily to superstition, promising good luck if the letter is copied and distributed and bad luck if it is not. They are often called "prayer" chains because many prior types started with a prayer or Bible verse.

Luck chain letters may have developed either from a requirement to distribute a prayer in a Roman Catholic Novena devotion [1898], or as a secularization of promises and threats in the Letters from Heaven [1906], possibly in a preamble. The English language paper luck chain letters of the twentieth century will be my principal topic. Most examples in the last few decades are highly traditional, having gradually accumulated varied devices to promote circulation. The lists of prior senders that often accompanied luck chain letters have at times motivated replication in order for one to display to others that a high status person sent them the letter. Since this motive is not catered to by any language in the host chain letter, I have not listed it as a separate motivational category. Luck chains have also been common on the Internet. Though originally these were simply digitizations of paper letters, they subsequently specialized to the email medium [e1995].

Filenames for paper luck chain letters begin with the letter "l" in the archive.

Money.
Money chain letters urge the recipient to send money to one or more prior senders, claiming that one can likewise benefit if sufficient copies are distributed.

The key innovation of money chain letters was a list of names and addresses with the instructions to remove the top entry, move the others up one slot, and add one's own name and address at the bottom. I call any list with these instructions a controlled list. Money chain letters originated in the United States in the spring of 1935 with the "Send-a-Dime" letter, also called "Prosperity Club" [Denver]. A prior luck chain letter [1933] was used as a model for Send-a-Dime. These and other details of the advent of money chain letters are presented in the article The Origin of Money Chain Letters which can be read independently of this treatise, or read in sequence (section 4-1). Money chain letters have influenced the content and distribution of luck chain letters up into the 1950's and possibly beyond (sections 4-2 and 4-3). Also included in this category are pyramid schemes, which we define as not using the mails to recruit (but they may, or may not, use the mails to make payments). Money chain letters continue as an omnipresent nuisance to this day, both in paper [2002] and as E-mail [2001]. Money chain letters and pyramid schemes violate Federal and State (West's CA) laws.

Parody.
Parody letters mock the style and methods of circulating chain letters. The request for copies may not be serious, but parody letters have often circulated in the mails.

There is a single example of an 1888 letter mocking charity chain letters which had just appeared in large numbers at that time. This letter purports to seek "brutes in pantaloons" to wed "old maids" in Massachusetts. It was not until the money chain letter craze of 1935 that parodies appeared in large numbers and many varieties. These mocked both the language and geometrical progression of the Send-a-Dime letter, as well as the exchange letters it had inspired. Examples mentioned in the press include the "Liquid Assets Club" [1935] (which may have actually been used to exchange liquor, as was possibly the "Send-a-Pint" letter) and the "Drop Dead Club" (shoot the first person on the list). I have collected several complete texts of early parodies, including some scatological examples [1935]. The familiar "wife exchange" [1953] was very common in the 1950's, and I recently found a bare bones example from [1935] using newspapers.com. These wife exchange letters illustrate how punch lines can be topped successively. The early 1935 example simply states that one may receive 15,125 women for its humorous effect. Then a 1939 example introduces the quip that one man broke the chain and got his own wife back. Though illogical, this disappointing result was the final punch line up into the early 1950's. A mimeographed 1953 letter notes in a postscript that at the funeral of a friend who received 183 women, everyone remarked that "he had a smile on his face for the first time in years." This in turn was topped in 1954 by an account of the difficulties that three undertakers had in removing that smile. The "Fertilizer Club ("go to the top address on the list and crap on the front lawn") [1971] also very likely goes back to 1935, but it is unlikely it would have been published in a newspaper. The wife exchange parody was commercially produced as a postcard [1954], and an undated matchbook advertisement suggests even earlier commercial production of chain letter parodies [1940?]. The wife exchange parody itself fell victim to parody in an imitative husband exchange letter [1949]. Despite commercial publication, chain letter parodies circulated in different versions like photocopied office humor. There is no serious request for copies, thus technically they are not chain letters. Parodies have probably served to educate the public on the fallacies of money chain letters, and have influenced the content of luck chain letters. They are very common on the Internet [St. Paul].

Paper parodies of chain letters appear in the archive with filenames beginning with "j" (for joke).

Exchange.
The exchange chain letters ask that an item of small value be sent to one or more prior senders, promising that if a specified number of copies are distributed the sender will in turn receive many such items.

Within weeks after the proliferation of the first money chain letter, Send-a-Dime, letters appeared which utilized its controlled list method to exchange items other than money. [1935] By 1937 the text in these chain letters, as well as the number of names on the list, had been reduced. Unlike luck chain letter types, the copy quota on exchange chain letters varies considerably, as does the number of names present in the controlled list. In chronological order, items exchanged on archived chain letters are: recipes, quilt patches, handkerchiefs, stamps, tea towels, postcards, dish towels, aprons, wash rags, Turkish towels, earrings, QSL cards, Tshirts, new panties, paperback books, dog toys, collectibles, grocery coupons, lottery scratchers and children's books. Exchange chains were still circulating in paper in 1996. Only one example in email form has been collected (a used paperback book exchange).

World Record.
The world record chain letters motivate replication exclusively by claiming distribution of copies will likely set a world record and that participants will be acknowledged. They circulated primarily among children after the new millennium, having developed from a lineage of postcard exchange letters.

In 1976 a postcard exchange letter claimed that it was approved by the US Postal Service as an "educational game for children". It also claimed that it had never been broken in over three years, and that just to delay sending copies beyond three days constituted breaking the chain [1976]. A 1985 cognate, said to have been started by "kids in Germany", asserted that if the letter continued unbroken for a little longer it would be in the Guinness Book of World Records. Later other such letters promised that each person who participated in the chain would get their name in the Guinness record book. But should the recipient not send copies, or even delay doing so for more than three days, the record would be spoiled and all the children "would have to wait another nine years to be in the record book" [1996-08]. This descent into absurdity had became inevitable when an innovation that promoted the total exclusion of adults replicated. On a 1999 letter the recipient is instructed to "... send it to six kids." Soon this restriction to kids was strengthened ("KIDS ONLY"), and was justified by saying it meant "kids will do the longest chain letter" [2001-04]. Distributions to adults may not have changed the text of the most irrational versions, but increased discard may have curtailed their circulation.

A letter mailed in the new millennium [2000-11] drops all mention of postcards and declares that "it is an attempt to get into the world records." So a new motivational category is necessary to cover this chain since postcards are no longer exchanged. I call this motivational category "world record". Our earliest example also claims that "the post office is keeping track". Further, perhaps to make this seem more plausible, the list of names and addresses, which previously directed the flow of postcards, had now migrated to the outside of the envelope. This in turn nurtured a grave fear: the post office could determine "who broke the chain" [2005-04]. This is no small matter: "it has never been broken so please don't spoil it for every one." An additional feature of this letter was the claim that it would be delivered without a stamp. Cognates collected in the next few years, most of them claiming to have started in Australia, dropped this feature but added the instruction that one should write on the envelope: "This is the official Guinness Book of World Records chain letter" [2001-04], or something similar. Presumably this would allow the Post Office to "track" the chain. This requirement of an external declaration continued on most letters of the lineage, and on these we see again the claim that a stamp was not required for delivery. One only had to write the declaration where the stamp would normally be affixed [2005-09]. This curious feature also appeared on "Lottery-Death" type luck chain letters in 1974 ( > no stamp ), as well as in certain French chain letters. The list of names was soon dropped in the lineage [2001-07], but the claim of Post Office tracking continued without it.

The exchange of postcards is the most logical use one can imagine for a paper chain letter. This is because the invitation to participate can itself be a collectible postcard. Thus it is ironic that a variety of postcard exchange letter gave rise to this most absurd of all chain letters. Most of the propagative innovations on the "kids" type letters are likely accidental or naively motivated, but many recipients must have believed them. A letter from a mother describes her daughter's fear of being identified as one who broke the chain [2007-01]. These "world record" paper chain letters may have been one of the most abundant English language paper chain letters in the first decade of the new millennium. But recently (2012) their numbers may have been greatly reduced by computer searching on text. As for all chain letters here, children's names and addresses have been obscured in online transcriptions.

Chain Email.
The primary focus of this treatise is on paper chain letters. But it is sometimes useful to examine copying behavior on the internet, particularly frequently forwarded email ("chain email"). This has a large and growing number of motives for replication. Hoaxes, humor and expressions of friendship are prominent. The following is an alphabetic list of some of the many topics observed since 1993: admonitions (duty to friends, sobriety, safe sex), anti chain letters, aphorisms, ASCII art and scrollers, communication experiments and demonstrations, consumer warnings, friendship, hoaxes (virus warnings, charity, giveaways, false quotations), human rights alerts, humor (single jokes and lists, office humor items, stories), inspiration, Internet protection (modem tax, phone charges, anti-censorship), good luck (often in sex or romance), missing children, money chains, number guessing tricks, parodies, patriotism, personality tests, petitions, photographs and videos, poems, political commentary, practical jokes (especially April Fools Day), prayer requests, protests, rumors, school & exams, seasonal (Christmas, St. Valentine's Day, Halloween, Thanksgiving Day), speeches, surveys, tag (snowball fight, mooning), urban legends (warnings, humor), voting recommendations, and Web page suggestions. Many of these topics appear in combination, such as a humor item with a short luck chain attached.

Many e-mail chains began as digitizations of paper chain letters. A very early example is an exact transcription of a circulating paper luck chain letter [e1982 - note archaic address formats]. Paper office humor items were also put online [e1995]. Once established, chain emails rarely surge in replication due to an offhand change or copying error, as we will see occurs within the paper medium. This is because an email is usually reproduced exactly, and thus there are few if any variations. However both luck chain emails and money schemes quickly developed adaptions to the new medium through a series of deliberate hoaxes or calculated modifications. A new restraining factor manifested when email chains were posted on various lists and group venues, provoking critical analysis and ridicule. Recipients of a chain email (and chain letters) are now likely to search the web on key text, particularly if money is solicited. Such a search will discover naive postings and attempts to recruit participants in money schemes. However, high in the list of matches, one will also encounter critical comments and disarming analysis, such as on some of the money chain emails in the archive associated with this essay [me2009]. Email screening criteria by Internet Service Providers have, in recent years, also become a significant factor in the survival of email replicators.

The Collection of letters.
I began collecting chain letters in 1973 with the hope they would reveal an evolutionary sequence. This effort was renewed several years later after discovering the folklore literature, particularly Michael Preston's 1976 article "Chain Letters" (Preston). This documented chain letters in a state of flux and presented variations of the same letter. Subsequently I placed ads for chain letters in collectibles magazines. Collecting large numbers of more recent letters began in June 1995 when Dr. Preston solicited chain letters for me from folklorists. In recent years I have also purchased old chain letters on eBay, the immense on-line auction. Sometimes copies were provided free by the seller or buyer, or a transcript could be made from auction photographs. I renewed collecting efforts in Dec. 2013 by subscribing to the commercial online newspaper databases newspapers.com and newspaperarchive.com. These have provided over a hundred chain letters for the archive and many entries for the bibliography.

All of the dateable chain letters (except for some foreign examples and recent money chain letters) have now been digitized in HTML format and each is accessible on-line as a separate file in the Paper Chain Letter Archive. An index for the archive lists clickable file names of all items in the archive, each with a one-line annotation. There are now (2017) over one thousand items in the archive, the vast majority being chain letters in the English language. These are ordered by (1) motivational category, (2) language, and (3) date of circulation. This index provides an easy way to browse the archive. Transcriptions preserve the errors in the original letter unless otherwise noted.

Table 1 - Contents of the Paper Chain Letter Archive.
English language chain letters presently (April, 2015) in the Paper Chain Letter Archive are tabulated below by year of circulation and motivational category. Himmelsbrief and religious chain letters are excluded. Scores of additional published letters, especially early luck and charity chains, can be easily obtained from existing online newspaper archives.

Years	Luck	Charity	Advocacy	Money	Parody	Exchange	World Record
1885 - 89		4			1
1890 - 94		2
1895 - 99		6	1
1900 - 04		5	6
1905 - 09	54	3	2
1910 - 14	61	3
1915 - 19	35	9	3
1920 - 24	42	1	3
1925 - 29	38		2
1930 - 34	24		2
1935 - 39	12	1	4	59	12	20
1940 - 44	21		3	11	1	19
1945 - 49	15		2	2	3	13
1950 - 54	15	2		2	8	7
1955 - 59	12		1	1	5	2
1960 - 64	5	1			2	1
1965 - 69	11		1	1	1	2
1970 - 74	16				1	3
1975 - 79	28			6	2	6
1980 - 84	37			3	2	4
1985 - 89	35	1 (b)	11	2	6	7
1990 - 94	53	1	1	3	1	4
1995 - 99	49	1		2		16
2000 - 04	5			1		1	5
2005 - 09	2			2		1	9
TOTALS	569	40	42	95 (a)	45	106	14
	Luck	Charity	Advocacy	Money	Parody	Exchange	World Record

(a) Over 100 money chain letters have been collected since 1975 but most have not been digitized.
(b) The Craig Shergold appeal circulated widely beginning in 1989. Many are published (Guigne); only two are archived here.

The numbers in the table may not be reliable measures of relative circulation. Newspapers were much more likely to print the text of a chain letter prior to 1960. The large number of Ancient Prayer examples collected is because it circulated largely on postcards, many of which were saved and eventually offered for sale by dealers on eBay. Recent correspondence is rarely offered for sale. Time gaps in the number of money chain letters in the archive reflects a lack collecting effort rather than circulation.

Foreign Language Letters.
Presently there are over thirty English translations of foreign language chain letters in the archive. Most of these are also presented in their original language as well. There are several foreign language letters that have yet to be translated.

Because of the ease with which letters are transmitted internationally, chain letters are, and have always been, an international phenomenon. Only by the extensive collection of foreign language examples can an accurate genealogy of chain letters be constructed. It is also revealing to see how chain letters vary from one culture to another. Sub-directories have been established in the archive for chain letters in French, German and Russian.

In 2006 I was contacted by Martinovich Vladimir Aleksandrovich, head of the Center of New Religious Movements Studies in Belarus. He has collected many chain letters in the Russian and Ukrainian languages. Transcriptions of some have been entered in the sub-directory /archive/russian [content-ru].

Publications.
Of the 900+ letters in the Paper Chain Letter Archive, 230 were found in publications. Early in the project the New York Times Index located many texts of chain letters, and a mention of a McKinley Memorial chain before it was collected (NYT 1906). As mentioned above, I have found over a hundred texts of chain letters using newspapers.com and newspaperarchive.com, online archives of digitized microfilm images. This has filled in many blanks in chain letter history, particularly with the luck chain letters of the 1920's and 30's. Newspaper transcriptions destroy formats and rarely report lists of names adequately. Some French (Le Quellec) and Polish (Robotycki) publications contain many chain letters that have yet to be entered into the archive or translated. Newspaper articles are also frequent in the Annotated Bibliography, which currently contains over 375 entries, most of them from newspapers.

Web Sites.
There are many thousands of WWW sites that match a search on "chain letter." The vast majority of these are about "email" chains, which are not my topic here. A useful list of annotated links appears in Watrous, and I will not duplicate this. To find the texts of luck chain letters one can search for traditional text, such as "Dolan Fairchild" or "Dalan Fairchild." A few transcriptions of paper luck chain letters found this way have been entered into the Paper Chain Letter Archive [1998]. Others are present on the WWW, but it is difficult to judge if they are complete and unedited. An article by Charles Bennett, Ming Li and Bin Ma, titled "Chain Letters & Evolutionary Histories" appears in the June 2003 issue of Scientific American (Bennett). This uses phylogenetic inference algorithms to construct a cladogram for 33 DL type chain chain letters. These are available on the web, and if dated I have copied them to the archive here.

Interviews.
I have obtained some information about chain letters and people's attitudes toward them by informal questioning of acquaintances. Several inquiries about foreign circulation have been made on USENET newsgroups. Much more could have been learned by systematic interviewing. However, people who send out chain letters, for luck or money, are often reluctant to reveal their activities and motives. Nevertheless, some interview material in newspapers and popular magazines has been very useful for understanding replication (Marilyn Bender, New York Times, 1968).

Ancient documents that advocate their own perpetuation.
Many ancient texts survive which provide diagrams, incantations or prayers that claim to benefit those who learn them. Some come close to our definition of a chain letter by urging that a personal copy be made. The Ancient Egyptian "Book of that which is in the Underworld" states (of a picture it provides):

Another Buddhist text, the Diamond Sutra, is the oldest (868 AD) extant book printed by wood block reliefs. It promised great merit to those who "observe and study this Scripture, explain it to others and circulate it widely . . ." (Goddard, p. 96)

The Letters from Heaven.
The "Letters from Heaven" (often called by the German "Himmelsbrief") claim to have been written by God or some divine agent. Many authors restrict the term to apocryphal Christian letters. These often claim miraculous delivery to Earth, magical protection for the possessor, blessings to those who "publish" them, and divine punishment for disbelief of their claims. The original copies are often claimed to have been written in gold letters, or with the blood of Jesus. Many published versions were illuminated. An early and frequent feature is the command for extreme Sabbath observance, as in the Madgeburg Himmelsbrief [text].

A German authority on the Himmelsbrief, H. Stube, said the letters long predated Christianity (Oda). Examples in Greek, Arabic, Armenian, Syrian, and Ethiopic have been published with German translations. Jewish and Islamic Himmelsbrief are also reported (Hand). These may all derive from an early Greek source (Bittner). A letter which was said to have fallen from heaven existed in the third century AD (Hippolytos, Refutation of All Heresies). The oldest Letter from Heaven for which we have a full text is the Latin "Letter from Heaven on the observance of the Lord's Day," the original of which dates from the close of the sixth century (Priebsch). St. Boniface denounced this as a "bungling work of a madman or the devil himself." Eckehard (1115 AD) wrote that it had spread over the whole globe then known to man. It has circulated in English in many versions [1795 text, image].

Jacob, organizer of the Crusades of the Shepherds, claimed (ca. 1251) the Virgin Mary appeared to him and gave him a letter. While in public he always carried it in his hand. A cult of uniformed flagellants appeared in Germany in 1261 claiming to possess a heavenly letter that had descended upon the altar of the Church of the Holy Sepulcher in Jerusalem before a multitude. The text has survived: God, angry at human sin, has decided to destroy all life, but the Virgin intercedes and God grants humanity one last chance to reform. Any priest who refused to pass on the divine message to his congregation would be eternally damned. During the Black Death (1348-9) the same letter, with a paragraph on the plague added, was used as a manifesto by a revived flagellant movement. At gatherings the manifesto was read publicly, the audience being "swept by sobbing and groaning." (Cohn)

Some Letters from Heaven specialized in protection, and accumulated long lists of weapons by which the possessor could not be harmed. The Count Philip Himmelsbrief [1895] granted protection against "spear, sword, sabre, cutlass, knife, tomahawk, rapier, helmet, burdon, . . . , and everything prohibited by holy writ, that is from all kinds of weapons, artillery, cannon, musket, rifle, gun or pistol." A preamble mentions its use in the American Revolution and claims that Count Philip of Flanders sponsored it after he was unable to execute a condemned prisoner who had secreted a copy on his person. Various Letters from Heaven in German were printed in Pennsylvania during the 19th and early 20th century (Oda), [1887 image1 & image2].

Letters claiming divine authority are also reported from India. Chain letters circulated in Shahabad in 1864 that condemned the breeding of pigs and consumption of alcohol. They were said to be from Heaven. In North Tirhut, 1872, cow protection was advocated by "strange papers" which "warned that Jaganath (Lord of the World) would curse any one who did not pay heed to this message and would burn down the house of any one who failed to pass it along to other people." Letters advocating cow protection in 1893 mandated recipients "make and then issue copies to at least five villages" - an early example of a copy quota. (Yang)

An email chain posted to an Islamic coins mailing list [1999] consists of: (1) an Islamic "Letter from Heaven," which likely first circulated in paper, and (2) a reduced version (testimonials only) of a paper luck chain letter I call the Lottery24 type. In II Chronicles 21:12 it is said that Elijah sent a letter to King Jehoram. It has been determined by scholars that Jehoram did not reign until 14 years after Elijan's death and the text has been interpreted by some clergy to mean that the letter came from Heaven. (1947)

It may be thought that the Letters from Heaven were a phenomenon of centuries past. But searching online newspaper databases reveals that probably hundreds of Jesus' Sabbath Letter have been published in local newspapers in the United States in the last two centuries, continuing up to the 1960's. Searching on the text "fast five fridays" produced 25 matches using newspapers.com and 72 using newspaperarchive.com. Most of these printings were responding to requests by faithful possessors of the letter, heeding its command to "publish" it. One columnist revealed: "It used to be sent to newspaper editors, demanding that the passage be published in the paper and setting out all sorts of dire consequences if the editors failed to acquiesce." (1939) Usually a brief succession of possessors is given, some of whom had bad luck after they did not publish a letter in their possession. Such claimed lineages may go back to the original legendary possessor of the letter [1910]. The Holstein Himmelsbrief, which features protection from weapons, has gained favorable newspaper testimonials for its use in both World War II and the Vietnam war: "He kept track of those to whom he sent a copy of the letter and every one of them returned unharmed from the war." [1968]

Transition to chain letters.
Edwin Fogel, writing in 1908, assumed that a luck chain letter [1908] was a new version of a Letter from Heaven (Fogel). There is little similarity in the texts, but perhaps Fogel was familiar with transitional forms now lost. Speaking of the apocryphal Letter from Jesus Christ [1915], Edgar Goodspeed wrote "it is sometimes sent through the mail with a request that the recipient send copies of it to three others, as some great misfortune is likely to befall him if he does not" (1931). Such a practice must have long predated 1931. Thus luck chain letters may have evolved from the preambles and postscripts to Letters from Heaven. At some stage the divine communication may have been replaced by a less pretentious "prayer," followed by entreaties to copy it. This is the form of the "Ancient Prayer" type [1905 - 1925] discussed in the next section. Some versions of Ancient Prayer promise deliverance "from all calamities" and threaten "eternal punishment" [1906] - as do some Letters from Heaven [Madgeburg]. Folklorists have generally followed Fogel in presuming that luck chain letters derive from the Himmelsbrief tradition (Elli s), though transitional examples have yet to be found.

More collecting should clarify the transition to chain letters. The first luck chain letters may also have been influenced by early charity chain letters [1888], which likely introduced the idea of a copy quota.

In this section I list characteristic features of English language luck chain letters, identify certain kinds of statements that are frequently seen on them, classify most of them into 12 sequential "types", and give a complete text and further information for each of the 12 types.

Features of 20th century luck chain letters.
After 1900 chain letters were influenced by increasing literacy, international mail and postcards, and changing attitudes about religion and miracles. Also chain letters themselves accumulated new technologies for increasing replication. Whereas the prior Letters from Heaven usually urged the reader to "publish" the letter, chain letters gained more circulation by relying on individual copying with specific copy quotas and deadlines. The following features characterize luck chain letters of the 20th century.

No.	Type	Sample Size	Predominance Range	Circulation Range	Standard	No. Words	Copy Quota	Deadline (days)	Wait (days)	List?
1	Ancient Prayer	166	1906-21, 1924 (a)	1906-25, 1938	Leeds	119	9/7/10	9/7/10	9/10	None
2	Good Luck	34	1922-23, 1925-26 (b)	1922-26	Sanders	66	9, 4/5	1	9/4	Most: X to Y
3	Flanders	28	1927-31	1927-31	Davenport	114	4/5	1	4	None
4	Prosperity	14	1932-35	1932-37	Hyatt	102	9/5 (c)	1	9/4	Most: controlled
5	Flanders-Prosperity	9	1939-40	1939-40	Shelby	157	5 (d)	1	4	All: controlled
6	Blind13	17	1941	1936-45	Kingsport	94	13	soon	13	None
7	Luck of London	9	1942-44	1942-48	McKnight	119	5	1	4	None
8	Chain of Good Luck	8	1949	1949-52	Burma	184	12	1	4	All: controlled
9	Luck by Mail	31	1950-64	1948-67	Halpert	132	5	1	4	Most: controlled
10	Death20	18	1968-73	1959-77	Bloomsbury	193	20	4	4	Most: controlled
11	Lottery-Death (LD)	13	1974-75	1974-75	Maryland	383	24 & 20 (e)	4	9 & 4	All: controlled
12	Death-Lottery (DL)	181	1977-2005	1973-05	AFC	351	20	4	4	Early: Some (f)

(a) Circulation in the US in 1924 was dominated by quota ten versions of Ancient Prayer. Two items from England, and one each from Australia and the US, had quota, deadline, and wait all seven [1916, 1925, 1923, 1920].
(b) Augmented versions of Good Luck predominated from 1925-26.
(c) A 1937 reduced Prosperity chain on a postcard asks for ten copies [1937].
(d) "Send this copy and four others" is on Flanders-Prosperity, Luck of London and Luck by Mail types.
(e) On early examples 24 is the quota in the Lottery block and 20 in the Death block.
(f) Some early examples of DL had a senders' list between the D and L blocks.

To recognize copying when there is high variability, and to simplify descriptions of chain letter text, it is useful to identify and name certain non-essential yet common types of statements that appear on various luck chain letters. I will capitalize these names to distinguish them from conventional uses of the same word, and allow them to be both nouns and adjectives.

Linkage. A statement on a chain letter which describes one or two of the latest transmissions of the letter in hand. If present, Linkage statements usually appear at the start of a chain letter, and can function as a declaration that the letter is a chain letter (Dundes). They may also be inserted when a list is removed. Linkage statements appear on some Ancient Prayer examples and are near universal on the Flanders type. Examples:

Circumnavigation. A request that the letter is to go all over the world, or that it is to go around the world, perhaps more than once. Or a claim that the letter has already gone around the world some number of times. Examples:

Expectation. A suggestion that the reader should "see what happens" after a certain number of days, implying that some joyous event or good fortune will happen. Examples:

Affirmation. A statement which, speaking as an observer, affirms the validity of the claims in the letter. It may attempt to explain how the letter works, or restate a claim with different words. Affirmations are highly variable and are often corrupted, rewritten, doubled or deleted. They are universal on the Flanders and Prosperity type letters. Examples:

    "It is positively remarkable how many times this prediction has been fulfilled since this chain was started." [1926]
    "The theory is to set up a definite and positive thought. [1933]
    "Here is infinite proof of this progress" [1940]
    "That's proof for you." [1942]
    "It works!" [1979]

Recycle. A statement which warns the reader to get rid of the letter (often within a certain amount of time), or to distribute it along with the copies that are to be sent. Recycle statements first appeared on the Flanders letters. If there is a list requiring updating, the received copy is no longer a candidate for being sent out again and a Recycle statement will usually not be present. A Recycle warning has become universal on the mainline since 1940. Examples:

    "Do not keep this letter in the house more than 24 hours." [1927]
    "Send this and four others within 24 hours." [1930]
    "Do not keep this letter. It must leave within 96 hours after you receive it." [1959]

1. Ancient Prayer.
Based on what has been collected so far, the "Ancient Prayer" letter was the first "luck" chain letter to circulate in the US, and this started abruptly in 1906. It likely circulated in other countries many years prior. There is a mention from France that it was denounced by the Bayonne Diocese in 1905. The earliest US example is a letter postmarked in Leeds, Maine on January 6, 1906.

The "dreadful accident" and the false attribution to Bishop Lawrence have been dropped and will never return. The advantages to replication of "all over the world" is discussed later (> circumnavigation). The reward of "great joy" for compliance is present on nearly all examples of Ancient Prayer I have discovered (for Russia, see Viola, note 59). Around 1909 the playful suggestion to copy the letter and "see what will happen" was introduced. This "Expectation" became common (but not universal) on Ancient Prayer and persists in the mainline to the present day [2005]. Early versions of Ancient prayer reveal an influence from the Letters from Heaven. For example, a 1909 letter claims that its rewards and punishments were spoken of in "Jerusalem." This was subsequently replaced by in "Jesus' time", perhaps originating as a copying error.

An interesting feature in the above 1910 text is the word "stating", seen to be a copying error for "starting" by comparison to other examples [1908, 1911]. A recipient has responded to this error by writing the date (Oct. 6). An abundant variation was soon established which contained "stating", and the date of the prior receipt [1912, 1914, 1915]. The advantage to replication of this variation was probably that it reminded the recipient of the impending deadline, whereas postcards lacking the date of receipt notation could be more easily ignored until the recipient realized the deadline had passed with no ill effect. The role of copying errors in chain letter evolution can be overestimated, as compared to deliberate innovations. But for any copying error to produce a successful variation is remarkable, and I will investigate other possibilities of this below (> LD).

Some Ancient Prayer examples are self titled "The Endless Chain" [1911], or "The Endless Chain of Prayer" (Fogel, 1908) [1923, 1925]. Chain letters as we know them were originally called "Endless chain letters" (NYT, 1906) to distinguish them from the then familiar self-terminating charity chains. The title "Ancient Prayer" did not appear on American chain letters until around 1909.

With U.S. entry into World War I in 1917, Ancient Prayer proliferated and differentiated. Some were exclusive within various fraternal organizations; some prayed for "peace" and others for "victory." An unmarried woman in Ohio received at least three of the Victory postcards just in October of 1917. [1917A , 1917B] The chain was so numerous that the editors of the New York Times proposed that it originated as a German plot to clog the mails (NYT, 1917d). A wartime postage rate increase, from one to two cents for postcards, may have cooled the chain off and foiled the Huns. The same chain postcard with substituted titles had also served the martial spirit of the Central Powers. A German language version, postmarked in Austria a year before the start of World War I, begins "We Germans fear God, and Nothing else on Earth!" [1913]. Immediately after the war Ancient Prayer declined in the U.S. and England. Some resented that "during the First World War they and many people they knew had received letters threatening death or horrors to their loved ones in the trenches of France if the chain was broken." (Simpson 2000). In 1924 Ancient Prayer revived in the US with a copy quota of ten and a new prayer. One such letter has been collected which was written in a fancy script [1924, image].

Though Ancient Prayer continued to circulate for many years after the end of World War I, and even had a boomlet in 1924, the postwar worldliness was not a good fit for its piety. The last Ancient Prayer chain letter to appear in the archive was a much reduced version on a postcard mailed in 1938.

By 1995 the Ancient Prayer chain letter was nameless and all but forgotten. But the chain was preserved on postcards and letters, and these were old enough that they were offered for sale. Of the 165 examples of Ancient Prayer in the archive, about 50 are physical postcards or letters purchased on eBay.

According to some reports (1948, 1968) the Good Luck letter was started by an American soldier during World War I. However our earliest examples come from 1922, a boom year for the chain both in England and the U.S. Thorough searches and inquiries have failed to date the letter prior to 1921. The text was short and secular, and retained the request for nine copies as on Ancient Prayer. Many examples had long lists of paired names ("X to Y") at the top, sender to receiver [1922]. There is a physical example in the archive with 113 names [1926], and a newspaper report of 214 [1925]. Below is a prototypic example, a typed letter mailed from Birmingham, Alabama on June 8, 1922. The X to Y list had 30 entries (I have deleted 27 of them here). Though "Claude Sanders" leads the list, he was not the author of the letter, though recipients who had not seen this chain before may have presumed so.

Good Luck Augmented.
The 1922 Good Luck chain letter was by far the shortest of all our predominant types (< Table 2). This seems to have invited the placement of additional text both at its start [1924] [1926] and end [1926]. The following example was published by syndicated columnist Helen Worth in 1925.

4. Prosperity.
Folklorist Harry M. Hyatt reported in 1935 that "during the latter part of 1933 a 'chain letter' fad appeared" and he gave a complete text except for two towns and two names in the list that he withheld to protect privacy.

"The Luck of London" chain letter was said to have originated during the blitz (1940) and continued to circulate in Europe and America even after the war. (DeLys, 1948). A letter published in the Neosho Daily News on March 16, 1942 is our earliest example. Columnist Robert McNight described it as a "new type of chain letter".

Clearly this chain letter is close to the Flanders-Prosperity type with "London" replacing "Flanders". Both types still have the leading Linkage, the same copy quota five, four day wait, 24 hour deadline, a Recycle command, the pecuniary testimonials followed by an Affirmation, then the "Do not send money" command. And the two names in the testimonials above are cognate to the names on the Flanders-Prosperity text we gave: Grace Fields vs. Mrs. Gay Field and Dr. Arcrose vs. Mrs. Ambrose.

Considering these similarities one could classify the Luck of London letters as a variation of the previous Flanders-Prosperity type. But there is a fundamental difference, besides the updating from World War I to World War II. All of the Flanders-Prosperity letters have a controlled list of names and often towns also. None of the nine Luck of London letters in the archive bear a list of any kind. Also the prior type promised prosperity as well as luck. The Luck of London letters have dropped the mention of prosperity and focus solely on luck. Luck was more needed than money during the war. The new chain letter, with its tribute to a city that survived an onslaught of the German air force, must have appealed to many who had family members at risk in the armed services. I rank the Luck of London chain letters as a new type, as columnist McKnight judged them to be in 1942.

8. Chain of Good Luck.
The letter below was handwritten and mailed from Sandoway, Burma on June 17, 1949 to A. Logozorie at a Roman Catholic Mission in Gold Coast, British West Africa.

There are just eight complete examples of the "Chain of Good Luck" (COGL) in the archive, but this international chain letter seems to have dominated the luck genre in the US in the year 1949. They all attribute their origin to a French officer serving under General DeGaulle in Africa. Other universals for the type are: (1) the title "Chain of Good Luck", (2) a leading Linkage statement, (3) a declaration that the letter is to go around the world the first time, (4) two Recycle declarations, (5) testimonials featuring a Philippine army private, President Roosevelt, and two victims of sudden death, (6) a controlled list of varying length.

In the leading Linkage statements, all but one COGL reads like the standard example above, claiming the chain was sent via "United Press Dispatch", or "United Dispatch", etc. But a 1952 example, published in Syracuse, New York, reads: "This chain of good luck was sent to me by Ronald Service, Essex, ...". This may tell us that "United Dispatch", and similar business names on the other examples of COGL, may have started as a corruption of a personal name. COGL has structural similarities to the Flanders type described above. And on a 1928 Flanders example the Linkage reads in part: "The Flanders Chain of Good Luck was passed to me by A. E. Blandfield ..." . So there is a precedent for personal names in Linkage, and the Syracuse COGL example may derive from one. Having a senders list makes a Linkage statement redundant, so if there ever were a personal name in the COGL Linkage it may not have been updated, and instead subject to many generations of unguided copying and corruption until finally someone miscorrected it to a more familiar name - of a business.

Note also that the 1952 example of COGL gives the city, Exeter, that the sender once removed lived in. None of the four newspaper examples of COGL in the archive give the contents of the list, but here we get a hint that the deleted list on some published COGL examples may have contained both names and towns. If a controlled list had enough entries - twenty would be more than enough - one could prove that a chain letter had actually gone around the world if the locations of senders were on the list. The prototype example above contains only names and initials, yet one might still infer that it was going around the world in a westward direction, perhaps from mission to mission.

If there is one prime reason why the Chain of Good Luck gained so much sudden compliance in the United States it was likely because it contained a potent death threat. "Detective Segundo B. Villanova . . . met instantaneous death in an accident of June 14, 1948." Such detail! This looks like a news item that came over the wire from "United Press Dispatch".

The Luck by Mail type also introduces "this is not a joke" and the qualification that you will receive your luck "by mail." These are now mainline universals, and I judge the latter to have been the innovation most responsible for the predominance of this type in the 1950's. This hypothesis involves a possible relationship with money chain letters (> Luck Follows Money). The declaration "this is not a joke" is discussed in section 3-4. Around 1954 the geographical attribution to "the Netherlands" first appears and became near universal in the mainline. Lists are highly variable on the Luck by Mail type - those present are often trailing controlled lists of prior senders.

Luck by Mail continued to circulate well into the 1960's, in many variations. This is surprising since a potent innovation appeared in 1959.

It is reasonable to suppose that chain letter copy quotas have increased because of the availability of photocopying. But in 1959 copiers were not readily available - this is the same year that Xerox introduced its first plain paper copier (the Xerographic 914).

The Death20 chain still circulates, but an entire chain letter has been added to it.

Apparently in the early 1970's a quota twenty-four chain letter was translated from Spanish into English and put into circulation in the U.S. or Canada. Abundant copies of this letter exist combined with Death20, but no examples of it in English as an independent letter have been collected. There are cognate forms in other languages, such as the French 1979 with a grisly testimonial. I name this type "Lottery24" because of the original copy quota and its introduction of the "Boss Wins Lottery" testimonial:

Around 1973 Lottery24 (L) letters were combined with Death20 (D) on single pages in the two orders LD and DL. This event was documented with unedited multiple examples by Michael Preston (1976). With the appearance of these two high copy quota types in the 1970's, the use of photocopying as a means of reproducing paper chain letters totally dominated. Hand copying all but disappeared. Perhaps a motive for initially combining two chain letters was to reduce photocopying costs after some one received both at about the same time. Our earliest example of the combination Lottery-Death (LD) is a letter mailed from Maryland in 1974.

The above device, "Write F.E.G.E. in the right hand corner of the envelope instead of a stamp," appears on many LD chain letters. Various initials were recommended (some without the instruction to omit the stamp), and examples also come from France (Bonnet and Delestre) and the USSR. The instruction to omit a stamp seems severely counter-replicative. However, in the US the original initials may have been "F.M.B.H" standing for "Free Matter for the Blind and Handicapped." Current postal regulations allow free postage for legitimate purposes if the quoted sentence is written where normally a stamp would appear. Presumably the initials suffice, though I have not verified that. Someone in the early 1970's may have used the privilege to mail chain letters for free. Most recipients would be baffled by the suggestion above, but if the letter they received had no stamp many would try it since they could easily convince themselves that all their stampless letters also got delivered. After all, with no return address there was no way to ever find out otherwise. Since the initials were meaningless to almost all copiers, they would quickly be corrupted. In disbelief, some copier dropped the instruction to omit a stamp and advised the initials be written on the upper left hand corner of the envelope. These versions may have benefited by being opened more often than a letter with nothing at all where one expects a return address. Meaningless initials ("cryptoids") often appear on grimoires and chain letters. Dr. Jean-Bruno Renard has collected an interesting chain letter in France that revives the use of initials as a substitute for a stamp [2000]. Posting without a stamp is also a feature of many of the recent (2006) World Record chain letters that circulate among children. Post Office automation, rather than deliberate indulgence, may explain why many of these stampless envelopes were delivered. Yet such delivery supports the absurd claims in these letters of Post Office involvement with verifying a world record, and even with identifying a person that broke the chain.

The LD type was prolific in 1974 - 1975, and also circulated in the U.K (Times, 1974). Some Hungarian chain letters [1983], though much reduced, reveal descent from an LD source. By 1980 the Lottery-Death letters had been completely replaced in North America by our final mainline type, the "Death-Lottery" letters.

The following letter was collected by the American Folklife Center, Library of Congress, in October, 1974. It was typed, except for the last three names in the second column.

This is a Death20 letter placed above a version of the Lottery24 letter (without a title) - the reverse of the LD concatenation. This Death-Lottery (DL) type first appears in the archive with an example [1973] published by the well known Canadian author John Robert Colombo (1975). However, since that example is missing a testimonial I have chosen the above letter as a standard for the type. Like LD, the DL type was first described by Michael Preston (1976).

One feature of the above letter is atypical - the list starts at the end of the letter. Most DL letters up to 1978 had a list of prior senders like the above, but they were "internal" in the letter, since they originated with the Death20 block and were bounded below by the added Lottery letter. Superstitious recipients may copy a list with the same diligence that they give to the text of a letter. With little room to expand, the internal lists on early DL letters may have been exactly copied for a few years. But by 1979 both these and all the LD letters stopped circulating. As photocopying had became more frequent, there was greater reluctance to comply if one thought some modification of the letter, such as updating a list, was required.

Though I make little use of formatting to infer relatedness, the most common paragraphing of a DL letter trespasses on the unity of the Lottery24 block, placing the last sentence of the Death20 block ( "Please send 20 copies of the letter and see what happens in four days") as the first sentence in the new paragraph starting the Lottery block (right before "The chain comes from Venezuela and was written by . . .") [1983]. This may aide circulation by disguising the compound nature of the letter and its resulting redundancy and contradictory claims of origin.

The early DL type was temporarily eclipsed by LD letters during 1974-75, but a hyper-competitive DL variation captured the entire luck chain letter niche before the end of the decade (the It Works postscript described in > Section 4-5). Thus all mainline luck chain letters since 1980, certainly over a billion, have been the DL type. Within this type are variations that compete with each other for the attention and resources required for replication. The advantages of some of these variations are explained in the sequel (> Section 4-6).

The DL type luck chain letter not only dominated circulation in the United States for decades, it also took hold in many foreign countries. That it originated in the US or Canada, around 1973, is fairly certain since this region nurtured the circulation of the Death20 component as an independent chain letter, and also spawned many early variations, including the unsyncretized 24 copy quota in the Lottery24 component (in LD letters only, as < above). From North America it has spread to many countries. Examples so far collected are listed below; most of the foreign language texts cited are supplied with an English translation.

There are over ten "types" of English language luck chain letters in the archive that did not dominate circulation in the US in any year. Some of these types are represented by only one example. I briefly discuss these outlier types here in their chronological order.

1. Cross Crossings Cautiously
A 1926-01 quota nine advocacy chain letter is titled "Cross Crossings Cautiously" (CCC) and states: "I have resolved that from now I will practice 'safety first', preach 'safety first' and do all in my power to save life or prevent any injury to my fellow men." The CCC slogan and the resolve also appear above a 1926-12 Good Luck block that warns against breaking the chain.

A 1930 luck chain letter has retained CCC but dropped all other mention of safety. It concludes with an X to Y list of 15 names, mostly famous, beginning with Sen. Heflin, Bernard Shaw, Henry Ford, and Colonel Lindbergh. James Thomas Heflin was a white supremacist Senator from Alabama, 1920 - 1931.

2. Chain Letters from France.
The following handwritten letter, titled "The Fortune Chain", was mailed in Okeechobee, Florida in 1931.

3. The Luck of St. Thomas
The following luck chain letter was published in an Iowa newspaper in 1949: "The luck of St. Thomas has been sent you -- it has been around the world four times. Copy the letter and forward it to five other people. Do not keep the letter in your possession. It must leave your home within 24 hours after receiving it. You will have good luck four days after receiving the letter."

4. Sick Girl Performs Miracles
The following luck chain letter was published in an Ottawa, Canada newspaper on Jan. 21, 1970: "This is a chain letter from Landers. Someone sent me this and now I am sending it to you. Do the same for people you know. This letter comes from a little, sick girl in Landers. Whoever breaks the chain will have neither luck nor happiness. This has already happened. The sick girl performs miracles unexpectedly. Happiness will befall you in 48 hours. This letter must not be destroyed or lost. Copy it seven times and send it to seven people. Don't put a stamp on it. Observe what happens to you in 48 hours." [1970]

5. "Every one will get thousands of copies."
A composed chain letter from December, 1975 declares: "... the ironclad science of Mathematics demonstrates conclusively that on the last Good Thursday before Christmas 1976 ... 882,922,240,000,000,000,000,000,000,000,000,000 copies of this letter will be in the mails. [1975].

7. Chain Letters from Mexico.
English translations of Mexican letters circulate in the U.S. in low volume. A 1984 example from Oxnard, California has a brief Tagalog addition at the end, and a comment on this in English. The letter has a quota of twenty-four copies, a deadline of nine days, and a thirteen day waiting period. A recent related letter has two blocks of Tagalog and much transformation of the testimonials [2004].

An English only example [1995], from North Carolina, represents a separate tradition. It states: "This chain would be sent with five cents which will be donated to the church." There was a nickel taped head up on the letter. This request was also present in an untranslated Spanish language letter mailed from Pasadena [1980]. A dime was taped on this letter. This sending forward of money seems to be unique with Mexican luck chain letters and is a striking contrast to U.S. mainline letters since 1939 which instruct "Do not send money." This command functioned to differentiate luck chain letters from the money chain letters that flooded the mails in 1935 (> Section 4-2). Thus this forwarding of a small coin may date from the 1930's also, and may be a different solution to the same discard problem that English language luck chain letters faced. Hopefully older Mexican chain letters will be recovered that can explain the origin of this feature.

8. The Media Chain Letter
Beginning in 1989, a quota five luck chain letter much like the then long extinct "Luck by Mail" type was revived by providing a pretext for a status display. I describe this widely reported "Media Chain Letter" in Section 4-4.

9. Romance Game.
I have four English language examples of a classroom note typically passed between young teenage girls. The following was intercepted from a 13 year old girl by a teacher in California in 1995.

3-1 POPULATION DYNAMICS
Circulation Generation Time Exponential Growth Billions of Chain Letters The Great Advantage of a Small Advantage
Immunization Exponential Decline The One-in-a-Hundred Rule

For accuracy, and for generality, care must be exercised in defining features of a population of chain letters progressing through people and through time. I will often speak of a variation. This refers to a set of related letters that all bear the some innovation, or perhaps to an entirely new type such as a translated foreign letter. The changes in population of a variation will be of concern. The host of a chain letter is a person that reads at least some of the letter and who may transmit copies to other hosts.

Circulation.
In counting how many chain letters of a variation V there are at a certain time, should we count a letter that is crumpled up in a trash can? Criteria are required to describe only those letters that are still "active".

DEFINITION: The "circulation" of variation V at time t, designated V(t), is the number of V existing at time t that will be conveyed to a new host.

Time, t, will always be measured in days, though in some contexts it would be more numerically convenient to choose a longer interval. For the start of a period use t = 0. Recall a "host" both reads some of the letter and may or may not choose to distribute copies. So we are not counting letters that are awaiting destruction, such as a letter that has been mailed to an undeliverable address. A letter which has been placed under a car windshield wiper is counted, provided someone takes it and reads at least a few words. A single physical letter may be counted at different times if it is passed on to a new host, as some chain letters instruct. [1927] The above definition of circulation has the defect of requiring we know the destiny of each letter that exists at time t. This is of course impossible, especially if we are inquiring about the circulation of a chain letter at the present time. But estimates can still be made, and the following discussions show how circulation, and other concepts here, provide useful theoretical tools.

Often I will not estimate the actual circulation of a variation, but rather judge whether it is larger or smaller than another circulation. A basic assumption is that such a comparison can be made by counting frequencies of the dated letters in the Paper Chain Letter Archive (index). For luck chain letters prior to 1960, the number of mentions located in newspapers is also helpful. However there are many biases in such comparisons.

old chain letters are more likely to have been lost or destroyed than recent ones
accumulations of personal letters mailed after 1950 are rarely sorted and sold by collectibles dealers
postcards are more likely to be saved than letters
hoards of chain letters received in a narrow range of time may become available
for recent letters, my effort at collecting has declined since around 2000
newspapers are most likely to give the text of a new type of chain letter that is being received by many
newspapers are unlikely to give the text of a familiar or poorly written letter

Generation Time
Usually, a circulating chain letter has often been (1) received, (2) copied, and (3) distributed. If the time between such receipts were always the same, say g days, we would have a periodic replicator with generation time (period) equal to g. Such regularity does not occur for chain letters, but for certain estimations I may presume it does with generation time equal to the arithmetic average of all elapsed times from receipt to receipt.

Surprisingly, a good estimate of this average is available for the abundant "DL" type luck chain letters. There are a few of these which bear a long list of dates received. Apparently these developed after a single date was placed at the end of a copy and a recipient behaved the same without removing the prior date. The growing list of dates suggested to downline recipients to add the date also, probably most not realizing there was no such instruction in the letter. In one example there are 72 dates from Aug. 7, 1979 to Dec. 23, 1980 [Bloomington]. This gives an average interval from receipt to receipt of 7.0 days. A 1982 example has 34 dates with an average interval of 7.8 days [Wenatchee]. Probably some senders did not add a date, so the average generation time for these two letters was likely somewhat less than one week. Mainline luck chain letters since 1970 have specified that the letter "must leave your hands in 96 hours." Adding three days in the mail gives g = 7 days total from receipt to receipt, supporting the validity of the listed dates.

Exponential Growth
Many perceptions of chain letter population booms are reported, and the dated letters in the archive reveal rapid changes in circulation. I will use a familiar model of population change called "exponential growth". The reader may skip the mathematics in this subsection with no loss of comprehension in the sequel.

For any population of entities in which all individuals can replicate, it is fundamental to surmise that the ensuing change in the population will be proportional to the number of individuals present. If that proportionality is constant throughout the history of the population, the "exponential growth" model expresses this mathematically. If V(t) designates the circulation of variation V at any time t, then the derivative dV(t)/dt gives the rate of change of V(t) with time.

DEFINITION: The population V(t) has "exponential growth" if dV(t)/dt = kV(t), where k is a constant (the growth constant).

This differential equation can be solved to give:

(1a) V(t) = V₀e^kt where V₀ = V(0) is the population at time t=0 and e is the constant approximately equal to 2.71828.

The reader can see some proofs, examples, and other information in the Wikipedia article on "exponential growth". If a chain letter has just appeared in a large population which is homogeneous (in the sense that all subsets of the population are equally susceptible to distributing copies) then during the time this homogeneity prevails the growth will be exponential. For the growth constant k: if k < 0 the circulation V(t) declines, if k = 0 the circulation is constant, and if k > 0 the circulation increases. Of course in a real situation exponential growth can not be sustained indefinitely because so many chain letters will have been received that the willingness to distribute more copies of the letter will be exhausted. Then this immunization phenomenon will interrupt exponential growth and circulation will decrease and possibly converge to zero. Growth defined by a replicator consuming limited resources can also be modeled mathematically (logistic growth). But for chain letters, potential recipients will always be hosting other chain letters that compete for attention. This, and network factors, likely make the logistic model inapplicable. We do, however, assume that the change in circulation of a chain letter, V(t), throughout its history, can be approximated by a piecewise exponential function, where within intervals of time the change is exponential but with differing growth constants. I often use the term propagation for such a local growth constant. A typical chain letter population might be modeled as going through three phases of exponential growth: (1) expansion, with the growth constant k>0; (2) stability, with k near 0; and (3) decline, with k<0. It is not unrealistic to presume these phases are well defined, since changes such as a postage rate increase, the beginning or end of a war, the onset of a recession, or the appearance of a virulent new chain letter are common and fairly sudden, and hence likely to produce sudden changes in the propagation of a resident chain letter.

Suppose a variation V obeys exponential growth in some interval of time. Instead of knowing the growth constant k for this interval in advance, we may often know its circulation V(d) at some later time t = d in the interval. There is then a unique exponential growth V(t) that will supply the two required values at t = 0 and t = d. The growth constant k of this function is found by requiring V(d) = V₀e^kdand solving for k. This can be done by taking logarithms, and gives the value k = ln[V(d)/V₀] / d, where ln is the natural logarithm function. Define f = e^k = [V(d)/V₀]^1/d. Then equation (1a) can be written in the form V(t) = V₀f^t . The constant f is called the "growth factor" for d days, and is the growth per unit replicator and per unit of time. Note the circulation V(t) declines if f < 1, remains constant if f = 1, and increases if f > 1. A familiar use of a growth factor is in compound interest calculations, where if r is the annual interest rate, the growth factor with the time unit one year is f = 1+r. Summarizing:

(1b) V(t) = V₀f^t where f = [V(d)/V₀]^1/d is the growth factor for d days.

If we are given a target circulation V(t) = C for an exponential growth, we can find how much time t it will take for the circulation to reach C. Solving (1a) for t gives:

(2a) t = ln[C/V₀] / k, which in terms of f is
(2b) t = ln[C/V₀] / ln(f)

Example: Say a variation is launched at t = 0 days with V₀ = 100 copies and after one month (d = 30) it doubles its circulation to 200 letters. Assuming constant exponential growth, this determines a growth factor for 30 days of f = [V(d)/V₀]^1/30= [2]^1/30 = 1.0234. The circulation at time t days is then V(t) = V₀f^t= 100*[1.0234]^t by using equation (1b). To estimate how long it will take for there to be ten million letters in circulation put C = 10,000,000 in equation (2b) to find t = ln[10,000,000/100] / ln(1.0234)= 498 days (over 16 months). If the population of potential hosts is 170 million (as it was in the US in 1980), 10 million is about 6% of that, which suggests that this may be about the maximum circulation for this variation.

With money chain letters, one hopes that some solicitation or possibility of receiving money increases exponentially. Of course not all these opportunities result in a payment, but perhaps a constant ratio of them will. I call such a process of multiplying opportunity exponential feedback. It is (taken in historical order) the goal of pyramid sales, money chain letters, pyramid schemes and multilevel marketing. The actual feedback from each of these falls short of exponential computations. But the hope for exponential feedback has been much used in recruiting for these schemes.

Billions of chain letters.
From one source and a few inquiries, I estimate that a typical American adult received more than 12 luck chain letters in the period 1970 to 2000. If so, for the recipient population of 180 million adults in the U.S., more than 12*180,000,000 = 2.2 billion luck chain letters were received in that 30 year period. Approximations here are on the low side.

This estimate of 2.2 billion received for 30 years (1567 weeks) implies an average of 1.3 million receipts per week. As determined above, the DL mainline luck chain letters during these years had an average generation time of one week. Assuming that all receipts occur within a week, the circulation of the DL type will be 1.3 million at any time. Presume for each letter, three days of the seven days between receipts were spent in the mail. Since 3/7 * 1,300,000 = 550,000, on a typical day during 1970-2000 there were probably over a half million luck chain letters in the mail. We have reduced this estimate some to account for many chain letters being distributed by hand.

The Great Advantage of a Small Advantage.
The conceptual tools introduced above are now used to demonstrate that the appearance of a seemingly minor innovation on a predominant type chain letter can have a significant effect. Suppose the active circulation of a mainline luck chain letter type is stable: for every 100 received, enough of these hosts will comply with the request for copies that about 100 copies of these letters will in turn be sent out and received. Suppose John Doe gets a degenerate photocopy of one of these letters and retypes it before making copies. He happens to add the postscript "Do not send money!" thus creating a variation V, makes 20 copies, and distributes them.

When Jane gets a copy of a chain letter in the mail she habitually glances at it and throws it away without reading the body of text, figuring it asks the reader to send money. When she gets a copy of V she glances at the title and at the bottom of the letter where one might look to see who the sender was. There is no sender listed, but the words "Do not send money!" appear prominently. This communicates at once that this letter does not ask for money. Jane reads the full text and, since she is waiting to hear if she got a desirable job, she decides not to take a chance on bad luck and complies with the demand for 20 copies.

Others may react as Jane did: say the new postscript induces just one additional person per hundred to fully and effectively comply to the demand for 20 copies. So after one week, the circulation will increase from 100 to 120. This is a weekly growth factor of f = [120/100]^1/7 = 1.0264. With this growth factor, the population will more than double every month since using equation (2b) we find t = ln[200/100] / ln(f)= 26.6 days to double. After 18 months out, equation (1b) gives V(540) = 20*(1.0264)⁵⁴⁰ = 26 million active letters. This is 15% of the adult population of the United States in 1980 (170 million), which means this off-hand variation, which started with 20 copies, could have replaced most of the previously circulating luck chain letters.

The scenario employed in this example is based on an actual event in chain letter history. The sentence "Do not send money" did appear on a Flanders-Prosperity type luck chain letter around 1939, and has remained ever since (> Section 4-2). It was repeated in a postscript in 1979 and this likely contributed significantly to the rapid predominance of the new variation that bore it, and to the demise of the prior variation. (> Section 4-5).

Immunization
In an example above we presumed a population of chain letters was doubling every month. Obviously such growth cannot be sustained very long. The number of possible recipients is limited, and there is an immunization effect whereby receiving more than one chain letter of the same motivational category makes one less likely to comply with copy demands. If one variation of a luck chain is abundant, another variation may be deprived of the attention and resources required for making and distributing copies. Eventually the abundant variation will foul its own nest by the same process. Thus for population booms, the exponential growth model applies until a variation has been received by a significant percentage of the subject population. More sophisticated mathematical models of growth with limited resources are available, but it would be difficult to verify their applicability for chain letters. Exponential "growth" may also apply when a variation is in terminal decline, as we discuss next.

Exponential Decline
In the John Doe example above, consider the fate of the unimproved mainline letters (call them variation "U"), which had a stable circulation before the V letter appeared. For U, every 200 hundred letters received were producing about 200 receipts in turn. For simplicity, assume that all these 200 new receipts were the result of 10 recipients who fully complied to U and sent out the quota of 20 copies. Now along comes variation V, doubling every month. Suppose just one of the ten U boosters, Joe, gets the now common V letter a week before a U arrives. Joe is likely to throw U in the trash, having already complied to V. Note that this immunization effect happens even if Joe is one of the 99 out a 100 people who have no preference between U or V. If this spoiling is typical, for every 200 U letters received, now only 180 copies are sent out.

If we assume this declining circulation conforms to the exponential model, we can use the above formulas to estimate how the circulation continues to decline. This weekly growth factor is f = [180/200]^1/7 = 0.985. Since (.985)⁴⁶ = .5 the circulation of U is being cut in half about every seven weeks (46 days). The initial circulation of U could have been quite large when the innovation V first appeared. I estimated above that average mainline circulation from 1970 to 2000 was around 1.3 million letters. Using this, we can calculate the circulation under exponential decline for the next three years.

Circulation of U after 1 year: 1,300,000 x (.985)³⁶⁵ = 5,226
Circulation of U after 2 years: 1,300,000 x (.985)⁷³⁰ = 21
Circulation of U after 3 years: 1,300,000 x (.985)¹⁰⁹⁵ = 0.08

So after one year one might get lucky and collect a U letter from that time period. After two years U is all but extinct. This corresponds to what we see in the archive (> Table 4, > Table 6). These extinctions are due to the advent of new variations, with what appear to be only small advantages. This example is useful in understanding the evolution of luck chain letters.

The One-in-a-Hundred Rule.
The above calculated examples, and tabulations using the Paper Chain Letter Archive, allow formulation of the following rule of thumb.

Consider a stable population of quota 20 mainline luck chain letters with a generation time of one week. If a variation arises that gets just one extra person in a hundred to fully comply, the circulation of this variation will initially double every month. Within a couple or so years it will be the only mainline luck chain letter still circulating.

Such captures of circulation by new variations are a common and striking feature of chain letter history. Analysis of why a new variation predominates may be difficult, especially if several innovations are present on a single letter. Because of the One-in-a-Hundred Rule, this replicative advantage could derive at least in part from infrequent or unknown factors in the recipient population, such as paranoia, minority ethnic identification, or participation in money chain letters.

< Start of above section < Start of Chain Letter Evolution - Contents

I will consider in Section 3.8 how certain chain letter content influences the selection of recipients (> Effective Distribution). Here I speculate about the flow of chain letters through a population, how flow patterns may persist and change, and how this may affect the circulation of variations.

Chain Letter Distribution.
The following individual behavior holds with regard to certain social replicators, and affects their overall pattern of distribution. This applies to photocopied office humor, jokes, rumors, and luck chain letter variations.

For luck chain letters, "single source" is generally true for significant changes except, possibly, deletions. Evidence for "habitual transmission" can be found in some interviews (NYT, 1968). This may begin when an individual correlates some good or bad luck with receipt of "the letter." "Habitual targeting" can be a matter of convenience, and also compliance with targeting instructions in the letter, which may suggest copies be sent to "people who need good luck." In a hoard of nine linen exchange letters received by one person, the lists of senders contain 22 names and addresses but there are only 12 different ones [xe1940]. Finally, "repetition taboo" is in part a restatement of the immunization phenomenon, which explains the cessation of chain letter crazes. Immunization is understandably a refusal to expend one's own time and money on repeated demands for copies. But when transmission is not anonymous a respect for one's recipients will be a factor. This may still operate for anonymous distribution, though not as strongly. From about 1922 to 1977 the majority of luck chain letters contained lists of the most recent senders. After 1978 there is not a single mainline chain letter in the archive that bears such a list, and almost all the envelopes these chain letters were mailed in did not have a return address. So there was a dramatic shift to anonymous distribution. However most transmissions were still probably from friend to friend, with the prior "known friend to friend" networks still active.

Imagine the complete flow of a social replicator V through a population. We can represent this by a network of transmissions whose points (nodes) and directed connections (arrows) between pairs of nodes are specified as follows.

Core Networks.
The core of a chain letter network of transmission can be roughly defined as the largest subnetwork of mutually connected habitual senders. Several formal definitions of the core of a network are given in Doreian and Woodward (Social Networks, 1994), but I have so little data compared to the number of participants that computational methods would be very approximate. I suspect this core, as defined, is more numerous and richly connected than would result from random linkages because:

Efficient Flow.
The "repetition taboo" implies that there will be an avoidance of duplicated arrows (A to B and A to B) in a chain letter transmission network. And short cycles (such as the dyad A to B and B to A) will be less frequent. This especially applies to letters with a senders list. If the list contains the last n senders, all cycles of length n + 1 or less can be avoided if one simply avoids distributing to anyone on the received list. Possibly, competition between networks will also develop this "efficient flow" since receipt of multiple copies by one person within a few weeks is wasteful. Those networks with long cycles should be favored. This could also involve a general westward movement of letters, or a tendency to move between three major cities in the same cyclic order. If cyclic flow is in both directions there will be more duplicate receipts, hence more discards. A very large sample would be needed to check for such patterns, nor should we expect that the repetition taboo and immunization alone could bring them about. But my guess is that more persistent structure exists in the transmission networks of folklore than is presently observable. Perhaps some systematized method of sampling will eventually enable the observation of flow patterns.

Descent.
Until the 1970's most paper luck chain letters were copied by hand or typed. When photocopiers became more common there was some debate if one could use them for chain letters and still receive good luck (NYT, 1968). One chain letter innovator declared "may Xerox" in a footnote [1975]. Predictably, the mainline photocopiers won this debate [but not for one outlier], and almost every letter that has circulated since 1980 is a photocopy, including originally hand written ones. But late generation photocopies must eventually be retyped because of image degeneration. In recent years this retyping is usually done with a word processor.

The word "copy" here allows that there may be errors, deletions, innovations, and even translation. But let us require that most of the parent letter is carried forward on the copy with matching details. On extremely rare occasions a chain letter may not have a parent, for example the founder of 1975. Or it may be the concatenation of two letter, for example the very first Death-Lottery type letter - a hybrid. It is convenient to exclude such letters in what follows, and consider only those chain letters that have exactly one parent. A first generation copy may itself be copied, producing a second generation copy, and so on. A letter M is a descendant of letter L if M is some nth generation copy of L, and then L is an ancestor of M. All descendants of chain letter L, plus L itself, constitute a descent group (or clade), with L the founder. The ancestry of a letter M within a descent group is the sequence beginning with M, then the parent of M, the parent of the parent of M, etc. until the founder of the clade is reached. Any two distinct chain letters in a descent group have a unique "most recent common ancestor." This is the first member of the ancestry of one which is in the ancestry of the other. Because of the convention that a letter begins its own ancestry, this means the unique "common ancestor" of M and its ancestor M' is M' itself.

Two chain letters are regarded as identical if they have the exact same text, character for character, as well as the same text styles and formatting. Usually when chain letter L is photocopied, and possibly when retyped, an identical copy is formed, called a clone of L. It might be possible that two letters could be identical but had different parents, but we can disregard that unlikely possibility. A replicator is not considered to be a clone of itself. For any letter L not a clone, the clone group with founder L is constituted by L and all its clones. Clone groups are the natural unit to consider when describing the descent of variations.

The Paper Chain Letter Archive provides overwhelming evidence that chain letters inherit text from their ancestors. From a "Luck of London" letter we read "It has been around the world four times" [1944]. Over 50 years later we read on an Australian letter "It has been around the world nine times" [1997]. From a letter mailed in 1959 from Bloomsbury, New Jersey we read about money won but life lost in the Philippines [1959], just as we do on the 1997 Australian letter.

Inherited details strongly suggest that the letters in the "types" I have identified are descended from a single founding letter for each type. If we start with the Good Luck letters [1922], these and all subsequent mainline letters form a single descent group that extends to the present, and whose founder may have been written in Europe soon after World War I. This descent group numbers in the billions of letters and some of its ancestries contain thousands of generations (over four thousand, if there is an average of one generation per week).

All luck chain letters since 1900 are probably influenced by the first letter with both a copy quota and a deadline. Present day familiarity with these devices masks their ingenuity: copy quota probably began with a single letter and the concept spread only with the distribution and translation of this letter. The same is likely true for deadlines, dropping claims of divine authorship, statements that the letter is to go around the world and non-miraculous testimonials. Such innovations distinguished luck chain letters from the Letters from Heaven with which early luck chain letters were once identified, though perhaps mistakenly (1908). The Letters from Heaven in turn probably have a conceptual founder, perhaps a Greek letter in the first century, and this in turn a pagan predecessor. If we presume there existed spoken rituals that demanded their own repetition, these may all have begun with a communication that claimed divine origin and contained an instruction for periodic repetition. As if echoing this primal origin, many of the Letters from Heaven emphasized rigid Sabbath observance. [1863]

Variation.
Hand written letters are often difficult to read and thus many variations are introduced as the copier tries to guess what is written. Many errors can also be introduced on chain letters that are typed. For examples of the corruption of names see 1926. With photocopying, after some 15 or so generations the text becomes wiggly, spotted and unreadable in places. Titles and other text at the margins may be lost because of misplacement of a sheet in a photocopier or image expansion [1991]. Thus photocopied chains, to survive, must be retyped periodically, which introduces errors and wrong guesses at illegible or missing words [e.g. "faxed" for "faded" in 1997]. Lines of text are often omitted when copiers lose their place in the source letter [compare Newark1 to the close Newark2 - the later has omitted "of receiving this letter" and "He failed to circulate the letter"]. Or the copier may notice the omission, and enter the missing line in a new position in the letter. For example in 1979 "Do not keep this letter" has been transposed with "It must leave your hands . . ." In 1985 a misplaced period has shifted an important ethnic cue (the Philippines) from one testimonial to another. A few changes are the result of copying what was not intended to be copied, such as a personal comment to the recipient [1906], a date, a casual postscript, or signed name.

In addition to such copying errors, there are many intentional changes. The testimonial of The Unbeliever's Death is often deleted, presumably for ethical reasons [1981]. Attempts to improve the writing style are seen [1995], and reformatting is common [1991]. Often a brief salutation [1989] or postscript [1997] is added, usually never to appear on another letter in our sample. Sometimes a whole new title [1997], sentence [1991], or testimonial appears [1975]. Both with hand copied letters [1939], and photocopied letters (Preston 1976), a recipient of two chain letters may combine them on one page producing a new and subsequently abundant type.

Probably there are thousands of major innovations every year, but most do not replicate sufficiently to find their way into our sample. There are so many variations, accidental and deliberate, that most retyped letters differ from their parent. I have never collected two identical luck chain letters. Paradoxically, ancestors can still be identified after hundreds of generations, and across translations and subsequent cultural modification. Compare the ancestral 1974 to Hungary 1986, or to the second part of 1999.

There is convincing evidence in the archive that on rare occasions, in copying parent letter X to copy Y, text from a third letter (the "donor") is also placed on copy Y. This process, and the text involved, is called a transfer. Here are some examples.

Differential Replication.
Very clear evidence that chain letter content affects replication is present in Table 4 and Table 6. These show that letters bearing certain variations have greatly increased in frequency over a few years, and letters without those features have totally disappeared from the dated collection. The succession of luck chain letter types (< Table 2) is also proof of the effect of content on circulation. The range of years in Table 2 records the earliest and latest year of circulation so far collected. Thus all the Good Luck type letters in the archive were received during the 1920's, and you are no more likely to receive one today than to be asked to dance the Charleston at the senior prom.

For many variations we can be fairly sure that after some initial appearance, all subsequent appearances of this variation within some group of letters under discussion are descendants of the initial example. Or if the variation re-appears as a result of a transfer or re-invention, we may always be able to verify this by analyzing other variations present. I call such a variation a feature (or character). Features are variations that can be used, at least in part, to infer that two letters bearing the variation had a common ancestor that also bore it.

Some variations are not features, or at least their use in diagnosing ancestry poses difficulties. For example, deletion of the Unbeliever's Death testimonial occurs independently in separate lineages. Certain corruptions and varying forms of numbers may also appear and re-appear, such as $755,000, $755,000.00, $755000.00 or $75,500,000. The words "Philippines," "receive" and "ignore" are frequently misspelled in the same way. "St. Jude" may be added to a letter, and also removed.

Descent groups (clades) are often considered when their presence significantly increases or decreases in our dated sample. Members of a descent group are recognized by the presence of shared features. Some of these features may have a positive effect on circulation, others neutral, and some may have a negative result. Features that are neutral or negative in an increasing clade are called riders, since they proliferate without themselves motivating replication. Usually there is one key feature judged to be primarily responsible for the increase of the group. Sometimes it may be difficult to select a key feature from two or more positive features present.

Small copying errors will generally be neutral, but some may have had a positive effect on circulation and increased in frequency as a result. Here are three candidates for this curious phenomenon:

The descendants of a single letter have repeatedly replaced all other mainline letters in our sample. I call such a descent group (or its founding letter, or the key feature) hyper-competitive. For example, all mainline letters after 1983 are the descendants of a single letter that first appeared around 1979! This descent group numbers over two billion letters. The key feature responsible for this spectacular replication was probably a new postscript (> It Works). We say this "It Works" postscript (or the first letter bearing it, or its clade) captured the mainline. Such striking examples of differential replication are surprisingly common in chain letter history.

The term "funneling event" from population genetics may be applied to captures, since they reduce the inheritable variation present in a population. These events not only establish highly replicative innovations, but also reset details of text with the features that happen to be present on the founding letter, for better or for worse. Chain letter evolution is characterized by a succession of funneling events through single letters. If a variation, a "small improvement," is merely increasing in the population of letters, it is subject to total elimination by the next hyper-competitive variation. Nevertheless, small improvements do appear to accumulate on chain letters, for example with the mix of testimonials (> Office) or with instructions on to whom the letter should be sent (> Effective Distribution). This appearance requires explanation.

Chain Letter Evolution.
I have described the descent and variation of chain letters, and their differential replication depending on copied features present in the text. These processes assure that chain letters "evolve" - that is, they accumulate inheritable features that increase or sustain circulation. It is this evolution that ultimately explains "how chain letters work," and why they worked even as public attitudes and beliefs changed over generations. This success is even more remarkable considering the universal condemnation of chain letters from both secular and religious authorities, and the lack of any real service they provide to their hosts apart from dealing with the false hopes and empty threats that chain letters themselves created.

Richard Dawkins describes the mechanics of chain letter evolution in River Out of Eden, while also emphasizing that chain letters "are originally launched by humans, and the changes in their wording arise in the heads of humans" (1995, pp 146-150). Our collection reveals that there are a great many such changes but very few that significantly increase circulation. And very few of those that do are the result of an innovation designed to work in the way it does. Indeed, some successful changes are the result of accidents in copying. As in biological evolution, successful chain letter "mutations" are rare events that can exploit an opportunity for replication in a variety of unpredictable ways. So chain letters do "evolve," and apart from computer simulations, are probably the best documented and simplest example of evolution known. Yet, unlike computer simulations, chain letters are readable documents that exploit human hopes and fears. This provides chain letter evolution an unlimited palette of invention, and makes their history intelligible in human terms.

Are the similarities between chain letter evolution and genetic evolution worth our attention? In the previously mentioned article "Chain Letters & Evolutionary Histories" (Bennett, Li, Ma), algorithms used for genetic sequences are applied to reconstruct the ancestry of 33 DL type luck chain letters. The authors state " . . . if algorithms used to infer phylogenetic trees from the genomes of existing organisms are to be trusted, they should produce good results when applied to chain letters. Indeed, their readability makes them especially suitable for classroom teaching of phylogeny (evolutionary history) free from the arcana of molecular biology."

The following biological phenomena suggested or prompted guesses about chain letter adaptions.

If every letter in the archive which bears feature H also bears feature G, and visa versa, this does not imply they appear together on all letters ever produced (if so we say G and H were concurrent). It is quite possible that G could have appeared first and later H was added to a letter bearing G, but no example of G without H has been collected. For chain letters there is no way to deduce concurrence of two features solely from texts, unless one had every letter ever produced. If features G and H are both present in a group of letters under discussion, the following table lists the five possible ways they may appear with relation to each other. The symbol {G, GH}, for example, means that within the group of letters: (1) G appears without H on at least one letter; (2) G appears with H, in any order, on at least one other letter, and (3) H does not appear without G, otherwise I would have written {G, H, GH}. The five possible relationships between G and H are all hypotheses, subject to revision depending on subsequent collecting or verification of certain deletions or transfers. None of these relationships depend on recorded dates of circulation for letters, though dates may be used in arguing for or against spoiling exceptions. However, the pre-linkage of feature G to feature H implies that G appeared before H.

Example 1: The DL letters with an early It Works postscript read "I, myself, now forward it to you." But this never appears on a letter with a Kiss or Love title. "I, myself, . . ." is thus unlinked to these titles. When these Kiss and Love titles were introduced they were first added to letters from which "I, myself, . . ." had been deleted.

Example 2: Our earliest Death20 type letter [1959], and all thereafter, bear both the Death and Money testimonial and the demand for 20 copies. No letter has been found which bears either of these features singly. Thus we say these two features are co-linked. They would be concurrent, if, say, they were both transferred from a foreign letter at the same time. If not concurrent, then one appeared without the other. But only if such a letter is collected would we then say the first is pre-linked to the second.

Example 3: Around 1988 a new testimonial, "Car," was added to a letter bearing the title "With Love all things are possible" ("Love"). The Car testimonial proved to be hyper-competitive within the clade of Love titled letters. The earliest example of Car [1988] (and almost all thereafter) also bears a duplication of the admonition against sending money, a feature I call "send no money," which reads:

Example 5: In the early 1980's two titles, called "Kiss" and "Love," captured the mainline (> in Section 4.7). For about 10 years a mainline letter bore either one or the other title, so during this time these features were unlinked. However around 1993 both titles began to appear together on single letters. They now formed the combinations {Kiss, Love, (Kiss)(Love)} which implies a transfer had occurred, since deletions from (Kiss)(Love) had certainly not produced all the single titled Kiss or Love letters. The Kiss title had been transferred to a Love titled letter at least two times. [1991, 1994]. There was also one case in which both the Love title and the Car testimonial were transferred to a Kiss letter. [1997] Kiss and Love were now transfer-linked.

We use the linkage of features to argue (with inherent uncertainty) for or against assertions of the following forms.

Cladistics is a method of classification that utilizes a "sister" relationship between two "taxa" (named groups of organisms), this holding when the two are more closely related to each other (have a more recent common ancestor) than either has to any third taxon. For cladistic analysis, a taxon must be a clade - an ancestor and all its descendants (a descent group). Sister relationships between "nested" taxa (taxa contained in taxa) are expressed by a bifurcating diagram called a cladogram. A sample of taxa appear at the terminal nodes of a tree, and two are connected to a hypothetical ancestor taxon if they are deemed sisters. Often these relationships are determined using characters (what I have called features) present in some but not all of the taxa in the sample. Cladograms are chosen which account for the distribution of characters in the "simplest" way, a principle called parsimony. In one approach, parsimony is defined as minimizing the number of non-inherited appearances (transfers) plus the number of losses (deletions) of features. Instead of using characters, pairs of items may be compared by a numerical measure of their "relatedness," and from these numbers a consistent cladogram is constructed. This later method might apply when instead of taxa we are comparing genomes.

Behavior that Affects Circulation.
In 1966 Alan Dundes described these universal components of chain letters: (1) a proclamation that the letter is a chain letter, (2) an injunction to send a specific number of copies, sometimes within a definite period of time, (3) a description of desirable consequences of compliance with the injunction, and (4) a warning of undesirable consequences if the injunction is ignored or disobeyed (Dundes). Mainline luck letters contain these components, though the method of identification may not be by proclamation.

We first identify behavior of a recipient that promotes the circulation of a chain letter, isolating these six components:

By retention of a received letter we mean keeping it in an undamaged state, accessible for copying. The bane of chain letters is immediate discard. If the recipient just saves the letter, as time passes it may work its will by playing on circumstances such as bad luck. The importance of first impressions for chain letters is revealed by the leading sentence in a version of the classroom Romance Game: "You touched this letter so you have to keep it!" [1998].

Identification.
Retention of a chain letter may depend on its identification as a luck chain letter, and perhaps as a certain luck chain in which the recipient believes. Quick recognition seems largely based on what is at the top: titles, initial text or rarely, images. Reputed authors and places of origin may also serve to identify a letter. A Brazilian letter is titled "Oracao De Santo Antonio" and a capitalized prayer follows [1994]. A 1996 English language letter originally from India has Sai Baba devotional images at the top [text, image]. Two translated Mexican letters have the title "St. Jude Thaddeus," one underlined, the other in a large font [1984, 1995]. Below this title, both these letters begin with about the same sentence:

The final words of a text likewise have added significance for quick identification, as shown by the replicative success of certain postscripts (> It Works). Placing "St. Jude" at the far bottom of a letter, a pretense of authorship, attempts to lock the conclusion [1996]. The Book of Revelations "seals" itself with: "If any man shall add unto these things, God shall add unto him the plagues that are written in this book" (Rev. 22:18). Some interpret this as sealing the entire Christian canon. The apocryphal Jesus' Sabbath Letter attempts to discredit any other letter claiming to have been written by Jesus by having him say: "You shall hear no more from me except through the holy scriptures until the day of judgment." However most versions of this Himmelsbrief then conclude with this addition: "All goodness and prosperity shall be in the house where a copy of this letter is found." [1926] One astute editor observed that this was likely the words of the "glib tongued rascals who sold this rubbish ..." (1884).

If identification as a particular luck letter aids replication, then we may expect a highly adapted chain letter will have multiple identities, depending on who is reading it. Several years may have passed since the recipient received a previous chain, so only one or two highlights may be recalled. These might be leading text, alleged geographical origin, or one of the numerical specifications. New types typically retain some of these, and thus may appear as the traditional letter to some and a novelty to others. Neither recipient is wrong - in recent decades successful mainline innovations are notably conservative, most adding just one or two new features to an existing letter.

A clear advantage results if a letter is identified, rightly or wrongly, by an ethnic group as the "same" letter that circulated in the old country.

The question of where a chain letter originally came from usually has no single answer. Though North America has been a creative source for innovations, especially since 1935, still most of the text on contemporary mainline letters is likely from various other countries. Perhaps the question of origins has an answer if we limit it in a reasonable way. Since 1970 the high copy quota letters (20 and over) have dominated international circulation. Where and when did they first appear? The first appearance of a high copy quota in our collection is on a letter mailed from Bloomsbury, New Jersey [1959]. Its demand for 20 copies is associated with, and enforced by, another innovation that first appears on this letter, the "Death and Money" testimonial. The events in this testimonial are purported to have taken place in the Philippines.

Differentiation.
I have discussed the advantage for a letter to identify itself with multiple traditions. It may also help circulation if a chain letter prominently distinguishes itself from another letter. The advantage of differentiation is very clear for post 1935 luck chain letters (for example, the Flanders-Prosperity type). In section 4.2 I explain how the hyper-competitive innovation "Do not send money" avoided immediate discard by distinguishing the luck chains from the boom/bust money chain of the day (> Divergence). Another innovation of these post 1935 letters was movement of the list of names from the top to the bottom of the letter, where they remained until disappearing in the mid 1970's. This provided a quick visual flag that the letter in hand may not be the abundant Send-a-Dime. Money chain letters retained the list at the top for several years (compare 1935 to 1941).

Woe to scoffers.
We see in testimonials that if one breaks the chain because of tardiness or unspecified reasons, misfortune may follow, but rarely death. However, all behavior that makes subsequent replication impossible is punished by a death.

Devices may be defined with varying comprehensiveness. The three testimonials above and the statement "For any reason, do not destroy or tear" [Mexico / U.S., 1984] might be considered together as exemplifying the single device: "Warn against destruction or loss of the letter."

If the recipient takes the letter as a joke, its promises and threats have no power, and this castrating perception may itself replicate. Possibly some previous types have succumbed to changing attitudes and derision. The pious Ancient Prayer postcard chain, which circulated from 1906 up through World War I, eventually disappeared during the irreverent and fun loving 1920's. The first three text examples below discourage disbelief. The last three discourage a more serious threat to replication - the expression of disbelief.

In this section I presume one has received a luck chain letter and retained it, but has yet to comply with its demands for replication. By compliance I mean that the recipient effectively distributes at least one copy of the letter (perhaps the copy received) within a month of receipt - thus contributing to its circulation. If the entire copy quota is distributed as instructed within the deadline we have full compliance. But partial compliance may be very common. In the previous section on retention I examined features of a letter that affect a recipient's immediate response to it. Here the focus is on how chain letter content may influence the recipient's deliberations on whether to comply, particularly on his or her interpretation of ensuing circumstances.

Motives.
The Letters from Heaven motivated possession and publication by the promise of divine blessing or magical protection from various perils, combined with threats of divine punishment for disobedience or disbelief. These were identifiably Christian in Europe and Hindu in India. After 1900 divine sanctions were downplayed, and by 1922 the mainline had only nonsectarian promises of good and bad luck.

Below are listed motives for replicating luck chain letters. These are based on statements of those who send chain letters, chain letter content, and known motives for sending certain postcards.

Some testimonials probably start as hoaxes (> "Car" below); but I suspect most started as rumors that are subsequently incorporated into the body of a chain letter. Say a chain letter has spread through a town, so much so that most people have received it. This can happen without anyone realizing it, as initially with Send-a-Dime in Denver, until Post Office officials noticed increased mail volume. In such a situation, most people have either broken the chain or complied with it. Now say John Doe is hit by a train and killed. There is a good chance he broke the chain. Suppose the letter is found among John's papers. It may then be said that "He broke the chain and was hit by a train a week later," and this may become a local rumor. The rumor may travel with the chain letter, orally or by telephone, each promoting the replication of the other. In this phase the more effective versions of the oral rumor will be favored. However, the advantage of distant transmittal applies if the rumor is written. First it may be on an attached letter. Next it may be a postscripted note below the chain letter, and finally it could be incorporated into the body of the letter. This could have been the origin of: "one person who failed to pay attention to it met with a dreadful accident" in a 1906 Ancient Prayer letter. Good luck testimonials likely spread in the same way - again, that a lottery winner also complied to a chain letter is not as improbable as it may seem.

Further, events that are not at all remarkable may be perceived by the public as prophesy fulfilled. The familiar "Unbeliever's Death" testimonial (> Woe to Scoffers) states that a person died exactly nine days after discarding the letter in disbelief. A simple estimate reveals that this is certainly true, in fact, it could have happened around 36,000 times just in English speaking countries in the last 25 years. Using the above approximations, a typical person received about 10 DL type luck chain letters in the last 25 years (4 per decade). Let us estimate, conservatively, that 6 of those receipts were discarded in disbelief. Assuming a quarter billion English speaking adults, this gives 6 x 2.5 x 10⁸ = 1.5 x 10⁹disbelieving discards for the last 25 years. According to the The World Almanac and Book of Facts, the U.S. death rate per 100,000 population in year 1999 was 877. This gives the probability that a person will die on a random day as .00877 / 365 = .000024. Multiplying this by the number of times a person discarded a luck chain letter in disbelief gives 36,000 estimated deaths on the ninth day following!

Once a testimonial is established on a luck chain letter, details may vary considerably over the years, such as names and amounts of money won or lost. But the basic structure of the story is surprisingly persistent, suggesting that traditional testimonials play a major role in winning compliance.

Classification of Testimonials.
To analyze how testimonials promote replication, I classify them by the following five structures: Win, Comply-Win, Lose, Win-Lose, and Lose-Win.

(1) Win: person X received the letter and had good luck.
Example: General Patton received $1,600 after receiving it. [1952]
The Win testimonials are consistent with a belief that luck chain letters are a "charm" whose mere receipt brings luck, much as possession of a Letter from Heaven might grant a woman an easy delivery. They suggest a recipient interpret good luck as caused by the letter, creating an obligation to pass the charm on to others. A Win testimonial may thus recruit a previous nonbeliever who has good luck. Alternatively, readers may assume that a Win testimonial is about someone who previously complied with the letter's demands, as in our next structure.

(2) Comply-Win: X distributed the quota of copies within the deadline and received good luck.
Example: Mr. Franklin D. Roosevelt was elected for the third term as president of the United States 52 hours after he mailed this letter. [1949]
Comply-Win testimonials promote the belief that dutiful replication of a chain letter will bring good luck. This particularly appeals to those who hope for gain from some forthcoming event, such as a lottery drawing.

(3) Lose: X failed to circulate the letter and after the deadline passed had bad luck.
Example: Mr. Nevin broke the chain and lost everything he had. [1939]
Lose testimonials promote the belief that only replication of the received letter can save one from bad luck. They particularly exploit those who feel that this is not a time they can risk bad luck. This insecurity could be due to a life threatening illness in the family, a job interview or a son in the military during war.

(4) Win-Lose: X received the letter and had good luck. But X failed to circulate the letter and lost what was gained, or much more.
Example: Dr. F. A. Anderson won $25,000 but lost it because he broke the chain. [1944]
A Win-Lose story promotes the belief that receipt of the letter brings good luck, but in return one must circulate the letter or lose what they received, or much more. It reports two connected events that imply the letter is a causal agent, in contrast to Win and Lose testimonials that can much easier be dismissed as coincidences. Win-Lose is persuasive with those who perceive themselves to have received good luck but have yet to comply. Such good luck could be escaping injury, recovering from sickness, success in an examination or winning a bet (Renard, 1987).

(5) Lose-Win: X received the letter and procrastinated or forgot to comply within the deadline. X had bad luck. X belatedly distributed the quota and received good luck.
Example: Mr. ASC received this letter. He forgot to post. A few days later he lost his job. After that he understood the significance of this letter and he sent 30 copies. He found a new job within 3 days. [India/UK, 1996]
The Lose-Win device encourages the belief that failure to comply causes bad luck, but this can be reversed, even after the deadline, if one complies in full to the copy quota. Like Win-Lose, the reversal of fortune points doubly to the letter as a cause. Lose-Win preys on those who perceive that they have had bad luck since failing to meet the deadline. Such bad luck could be an accident, loss of a bet or sale, illness, car trouble or not being hired.

Early Versions.
On an early Ancient Prayer chain letter we read that "one person who failed to pay attention to it met with a dreadful accident" [1906]. This is the only testimonial, win or lose, that I have collected on the Ancient Prayer type or the following Good Luck letters of the 1920's. Testimonials reappeared in North America during the Great Depression with brief accounts of gain and loss of money. The Prosperity letters usually have three, with the pattern Win / Win / Lose. Most versions of the World War II Luck of London type [1944] introduces the familiar pattern Win / Win-Lose, possibly having combined the last two testimonials from the Prosperity type. From our earliest Luck by Mail type:

Officer Wins. Elliot wins and loses.
With the remarkable Bloomsbury letter of 1959, our early example of the Death20 type, we get the canonical versions of these two testimonials. Likely by some unknown path of corruption, the chain breaker "Gen Allen" has been transformed to the civilian "Don Elliot".

Death and Money.
This Win-Lose testimonial first appeared in North America on the Death20 founder around 1959 and was associated with inflation of the copy quota from five to twenty copies. It was the first implied death threat in the mainline.

The variant "wife" (for "life") first appeared around 1975, possibly as a corruption, though it is curious that this 1975 letter had a list of names of 17 couples. It was present on the Kiss -Love founders in the early 1980's and thus became universal. We have no examples of corruption or correction back to "life." Whether "wife" was simply a rider on the successful new titles, or instead carried some replicative advantage over "life," is a difficult question whose answer could be revealing. The "wife" version of Death and Money has been translated and transferred to a Spanish chain letter.

Win-Lose testimonials like Death and Money exploit those who perceive themselves to have received good luck since receipt of the letter. If a gambler wins big at the track after receiving the letter, he may comply to avoid being jinxed his next time out.

We now consider the three traditional testimonials that first appeared in the Lottery24 (L) block of the "DL" and "LD" compound letters of the 1970's. Like the Death and Money testimonial, we have no prior history of these invaders.

Boss Wins Lottery.
This Comply-Win testimonial provided the first mention of a lottery in an English language letter.

The above phrase "in his country," apparently an early North American addition, disappeared in the 1980's, thus allowing the recipient to believe the lottery was won in his own region. However this deletion was first present with some potent innovations (It Works, Kiss, Love) and hence prevailed in part, if not entirely, as a rider. Other changes have been minor, including the more usual "Diaz" for "Diso" and the syncretization to 20 copies.

The so called "sweepstakes" promotions likely also increase chain letter circulation. These are sponsored by American Family Publishers, Publishers Clearance House and other firms. Tens of millions participate in these, hoping to win a fortune. Promotion is by television advertising in conjunction with a direct mail campaign of incredible magnitude - almost all adults in the U.S. get this pitch (1998). Recipients are led to believe that they have already won a huge prize, and that they only need send in an application to receive it. For example, I received a letter from American Family Publishers that displayed through a cellophane window a formal looking document decorated with eagles on each side. It proclaimed:

In addition to its appeal to gamblers, Boss Wins Lottery makes compliance easy for some by suggesting they have a secretary prepare and distribute copies. And it assures them that the good luck still belongs to the boss. This testimonial, and the next one I discuss, show that Lottery24 was well adapted to an office environment.

Lost Job - Better Job.
The following Lose-Win testimonial appears after Boss Wins Lottery on Lottery24.

From the above scenario, we see that Boss Wins Lottery and Lost Job - Better Job function together in dealing with an office environment. The concerns of both supervisors and subordinates are dealt with by example. Just as quota 20 first appeared in the U.S. in association with Death and Money, perhaps copy quota 24 first developed in Latin America in association with these complementary office testimonials.

The Unbeliever's Death.
After the Lost Job - Better Job story we usually find the following "Lose" testimonial:

Car.
The following Lose-Win testimonial first appeared around 1988 on DL letters with the Love title. Thus it is not a traditional part of the Lottery24 block, but is usually formatted continuous with it.

I have mentioned the image degeneration that results from successive photocopying. In my experiments, after 15 generations (all with the same photocopier) there was significant loss of legibility. However, more recently, packets of forwarding letters with the Media chain letter are often legible after 25 or more generations (different photocopiers). In any case, contemporary chain letters are creatures of photocopying, and they must be retyped from time to time. This testimonial is the only one we have seen that explicitly encourages retyping. Evidence that it succeeds in this is present on a DL letter from 1995, which adds to the Car testimonial: "I have retyped it again today in 1995." Another letter gives 20 alleged retype dates, including ten just in the year 1992 [1995]. Presumably, the propagative success of Car is due to this more frequent retyping. Similar letters without it would more often become partially illegible, and therefore more likely to be discarded, or to be retyped with fatal mutations. With a larger sample I could test this by comparing image quality of Love titled letters with and without Car. Our present sample is inconclusive on this.

Though less appealing theoretically, the most effective feature of this testimonial may be its use of the automobile. If someone holds the letter past the deadline, there is a fair chance they too may be "plagued with various problems, including expensive car repairs." Car repairs often occur unexpectedly, always seemingly at the worst time, thus evoking the specter of bad luck. Like any Lose-Win testimonial, Car implies that no matter how late you may be, compliance will turn bad luck to good. By using car trouble as its example of bad luck, this testimonial may be particularly effective in activating late compliance.

Further, for many the desire for a new car is greater than their respect for reason. Suppose John Doe plays the lottery. Then likely he has already given careful consideration to which vehicles he will purchase after winning. The Car testimonial, in its original form, is clever in not specifying how the young woman gets a new car. She may win it, or win the money to buy it, or perhaps receive it as a gift from a man she is to meet. Thus this testimonial may activate compliance by interacting with the fantasies of lottery players and others.

Notes that one has retyped a chain letter occasionally appear as postscripts. This appears on a Brill parody letter:

Faithful copying.
Successful replication of a chain letter requires copies be legible, accurate and complete. I examined in the last section how the Car testimonial suggests retyping, and thus upgrades the legibility of a photocopied chain that contains this testimonial. The following instructions also promote faithful copying of text.

Names and numbers are usually highly variable on chain letters because if partly illegible there is no context for the copier to infer the correct form. But as they vary they may at some time assume a form that allows the whole to be reconstructed from a part. I call such text self-correcting. Consider:

Below is a list of variations on the name of the victim in the Unbeliever's Death testimonial.

Zarin Berrachille [1973], Zerin Berreskilli, Zarin Rurreasville
Zorin Barrachilli, Zerij , Berreskilli Caren Wichile, Zerim Berreball
Zarin Borracbilt [1975], Brian Barbialle, Brian Barabiaila
Dalin Nairchild [1979], Colin Holschild, Darinn Meirchild,
Dalan Fairchild (Kiss-Love founder) [1983], Darin Hairchild, Dolon Fairchild

I pause here to note that the family name "Fairchild" is self-correcting since it is a compound of two familiar English words and a familiar name. Fairchild predominated from 1983 onward. I designate it by "F" below, and continue with the tortuous transformations of the first name, which never settled.

Dalon F, Delan F, Sobon F, Dalah F, Dallan F, Davan F, Galan Pairchild
Bolan F, Blaine F, Olean Lauchild, Dales F, Dallas F, Galan Paircheild
Dian F, Dilan F, Daln F, Darlene F, David F, Delri F, Mr. Fairchild
Deian F, Delea F, Dala F, Karen F, Nolan Foarohald, Darron F, Mellisa Horton [1994]
Carl Daddit (name from the prior testimonial, which was deleted), Colan Fatchild [France, 1997]
Brian Fairchild [2003]

Despite the great initial variation in the family name, and continuing variation of the first name, clearly almost all copiers are trying to get it right. Some are probably working with a highly degenerate photocopy. Perhaps "Fairchild" has more going for it than just self-correction. It is an innocent and virtuous sounding name, yet this seeming virtue offers no protection if one destroys the letter. On a translated Mexican letter: "Isabel Buena lost her copy and lost her life" [1984]. Again the family name is a word in Spanish, and a virtuous one (Buena = good). Such a tactic may boost circulation slightly, but recall that chain letter evolution is characterized by the rise of new variations that eliminate their cousins and establish their incidental features on every letter. If the founder of the Kiss & Love titles [1983] had read "Zarin Rurreasville" instead of something close to "Dalan Fairchild," Zarin would have been everywhere until obliterated by copy errors. And if "Rurreasville" had ever stumbled on a self-correcting form, it would not have been "Fairchild." Self-correcting text does not increase circulation, it only preserves itself from corruption. Its only chance for predominance is to ride a successful innovation, self-correction preserving it where an unfamiliar name would soon disappear in the copying noise. This is how "Fairchild" became near universal - by riding the hyper-competitive Kiss-Love founder and, like a rainbow in a waterfall, transcending chaotic disruption.

The name "Elliot" (as in the Elliot Wins and Loses testimonial) also appears to be self-correcting. This may be due to the scarcity of common surnames that begin with the letters "Ell." Also it seems some letters remain more legible in late generation photocopies than others; upper case "E" for example.

Self-correction applies to most of the text of a chain letter, since there is inherent redundancy in language. Some words and phrases are often corrupted, others rarely. Spoken replicators transform to a more memorable form; written replicators to a more self-correcting form. However, it is difficult to formulate general principles to assess the self-correcting power of any given text. I will rarely make use of self-correction, instead focusing on text whose meaning increases circulation.

Copy quotas.
All known luck chain letters specify a fixed number of copies that the recipient is directed to produce.

Do luck chain letters threaten punishment if you distribute fewer copies than the quota? I have yet to find an example. Often the only implied threat is for "breaking the chain," which could be interpreted as requiring only that one pass on the received letter to avoid bad luck. Partial compliance to the copy quota may account for many distributions. Thus it would probably reduce circulation for a chain letter to explicitly threaten this behavior, since then some may reject the letter entirely as too demanding. Yet it is essential to get recipients to produce the full quota of copies as often as possible. This dilemma has resulted in the survival of chain letters that are ambiguous on this issue. You are told "you must make twenty copies." And after his secretary made twenty copies and sent them out, Constantine Dias won a lottery. But in the same letter, bad luck comes only to those who distribute no copies at all, as in Death and Money where life is lost after one "failed to circulate the letter." An explicit statement of this option appears on a Russian chain letter [Homily, 1990], where just passing it on is considered a "neutral" act.

Copy First versus Copy Later.
So by just passing on the original letter, perhaps one may avoid bad luck. How many copies must one distribute to get good luck? Again chain letters are ambiguous; by one reading you need not distribute a single copy! The contemporary mainline letter is a compound of two differing folk beliefs or susceptibilities: Copy First views the work of replication as bringing subsequent good luck, Copy Later sees the letter as a charm whose mere receipt brings luck.

COPY FIRST text requires one first distribute copies of the letter before receiving good luck, as if in compensation. "You must make twenty copies . . . and after a few days you will get a surprise." One cannot refuse to send copies just because no luck is received: "For no reason whatsoever should this chain be broken!" However, ambiguously, bad luck may be reversed by late compliance to the letter's demands, as in Lost Job - Better Job.

The Copy First orientation places the recipient subordinate to Fate. Hope for good luck and fear of bad luck are about the only motives for replication. Copy First testimonials are of the Lose, Comply-Win, and Lose-Win structure. Much of the text of Lottery24 is Copy First.

The Copy First structure also appears on other social replicators. Devotional messages have been placed in the classified advertisements of U.S. newspapers for many years. Here is an example (my italics) from the "Religious Announcements" category of The Los Angeles Times (Nov. 25, 1991):

In the Copy Later orientation, one is almost bargaining as an equal with Fate. If no good luck is received in the stated interval, the charm has failed, and perhaps no copies need be distributed. If luck is delivered, the motives for replication now include gratitude and benevolent transmission of the charm to another. Hope and fear are less active, though not absent. The great taboo is to receive luck but then neglect your side of the bargain and fail to distribute the letter. You may lose what you received: "Don Elliot received $60,000 but lost it because he broke the chain" - or you may lose much more " . . . before his death, he received $775,000 which he had won." Copy Later testimonials are of the Win and Win-Lose type. The notion of simply passing on the original letter to avoid misfortune is associated with the Copy Later belief (as expressed by Boris Pasternak in 1959, see also lr1990). The chain is a chain of benevolence, not of fear. Equity is maintained by granting it to one more person. In the Death and Money testimonial a life is lost after General Walsh "failed to circulate the prayer (letter)." But this is not punishment solely for not complying, rather it is for not complying after receiving a large sum of money. Thus the letter grants luck by mere receipt, but exacts a dreadful toll from ingrates who do not then pass it on. This view of chain letter magic is outside the mindset of some revisers. Two letters collected by folklorist Paul Smith modify Death and Money to read: "Before her death he received $7,775.00 after circulating it just prior" [England, 1992]. This attempts to cast the events in a Copy First frame, but implies the wife died despite prior circulation of the letter. A North American revision fails in the same way [1991].

Other social replicators display the Copy Later structure. The following example (my italics) was published in The Los Angeles Times classifieds (Feb. 7, 1990):

Copy First and Copy Later may not be folk beliefs, but rather susceptibilities to one or the other side in a branching that naturally develops when supernatural promises compete for replication. A currency chain is a short message on paper money that encourages the reader to copy it on other bills (Olbrys). These varied replicators have appeared on U.S. bills for several years. Here are two examples:

The manner in which chain letters are distributed may significantly affect circulation, but chain letters have little to say about it. Mainline letters have always just said to "send" the copies. In practice they are sent by mail, placed in work mail slots or left where they will be found, such as on car windshields or desks. In Brazil they may be left in elevators and on doorsteps. When a choice of recipients can be made, the persons selected obviously will affect the number of second generation copies produced. Chain letters have evolved instructions for this choice that probably work better than anything a mere mortal could devise.

Targeted distribution.
By targeted distribution I mean any preference or exclusion in the selection of recipients. Chain letters usually have something to say about targeting. The most common recommendation is "send these to your friends" [1902]. This may seem of little help to circulation, but note that at the least it discourages sending to a celebrity - an almost sure waste of a copy. The next major innovation in mainline letters appears on Good Luck: "Copy this out and send to nine people whom you wish good luck" [1922]. Though this opens the door to celebrity distributions, it may also target people who need good luck. During the Great Depression "whom you wish good luck" was changed to "whom you wish prosperity to" [1933]. During the perils of World War II "prosperity" changed back to "luck" [1944]. With the Death20 letter this was improved to read ". . . to friends you think need good luck" [1959]. With this targeting the chain letter will seek out those in a desperate situation and who are thus more vulnerable to its promises and threats. But some recipients may give up on the letter if they have fewer than twenty friends. On the It Works innovation of 1979 "friends" has changed to "people," giving the current and stable reading "Send copies to people you think need luck" [1997].

The Lottery block originally stated: " . . . send it to your friends, parents or acquaintances." With new titles in 1983, "parents" was deleted, and "friends and associates" now appears (> Section 4-6). The classroom Romance Game addresses girls, so "Send this to 7 people, no boys" [1998]. There is, however, an Internet version that is genderless [e1995]. The same letter also specified "You can't send it to the original person" meaning the sender. Others chains specify the copies are to be sent to "diferentes personas" [1980]. When a list of prior senders is present this by itself aids in effective targeting since one may avoid sending a copy to any of these people. This may be explicitly instructed: "Important: Do not address your letter to persons already on the list" [International, 1949].

I should emphasize again that evidence from the dated collection shows that chain letters do NOT evolve by a series of small (secondary) improvements. Instead, a killer variation comes along every few years, swamps all its rivals, and universal details are set by what was linked to it. But small improvements can still accumulate for reasons previously discussed (< in Section 3-3).

All the above targeting instructions likely increase circulation over no suggestion at all. However, the following, which appeared on a luck chain letter, seems counter-replicative: "Limited to Masons" [1954]. A similar restriction arose in the course of the Springfield, Missouri pyramid craze (1935), and during World War I the Ancient Prayer split into specialized versions that circulated within various fraternal organizations (NYT, 1917). The Send-a-Dime craze manifested letters that were restricted to people with last name "Smith," another for "Johnson" (DP, 1935). This "over-specialization" phenomenon probably arises when there is a chain letter boom and a recipient has many versions to choose from. One which then makes a personalized appeal may be chosen over others. But this is a recipe for extinction since targeted groups will be the first to overdose on the letter. The last three examples of over-specialization were definitely during chain letter booms; perhaps "Limited to Masons" is evidence that there was a luck chain boom around 1954, soon after the innovations of the Luck by Mail type appeared.

In Section 2.2 I listed Circumnavigation as a near universal feature on luck chain letters and later gave three examples. Here are some more.

Deadlines.
Chain letter events in the 20th century reveal that for propagative success it is not enough merely to reproduce in quantity - it must also be done quickly. Repeatedly a new variation will flood potential senders, thus starving out competing variations for the attention, energy and respect needed for compliance.

The following four quotes reveal the development of deadlines on American chain letters.

Example (1) is from an early charity letter that solicited a dollar to help the campaign of Benjamin Harrison for President of the United States. It expresses a need for haste, but the history of chain letters reveals that specificity works best, whether it is in requesting multiple copies or rapid execution. No charity letters in the archive state a deadline. In (2), the Ancient Prayer method of a copy a day is prescribed. Most examples of this prolific type ask that this be done for nine days, as in a Roman Catholic Novena devotion. An implicit deadline is set by asking that this copying commence on the day the letter is received. On (3), a Good Luck chain letter, we find a pure deadline of 24 hours to complete nine copies. The nine day period is retained by asking it be counted out in expectation of some benefit. This could suggest to Catholic readers that the letter is still in the domain of a Novena. In example (4), from a Flanders luck chain letter, all trace of a daily duty is abandoned.

Can one still receive good luck, or escape bad luck, by distributing copies after the deadline? The answer is a clear "maybe." Considering first the "Copy Later" Death20 block, the only explicit deadline statement is "Do not keep this letter. It must leave your hands within 96 hours." As noted above, this seems like you only need to pass on the original. And there is no mention of someone suffering misfortune on the fourth day after receipt. Examining the "Copy First" Lottery24 block, there is no explicit deadline statement. Nine days seems to be the implied deadline, judging from Lost Job - Better Job and the promise of a "surprise" in nine days. Thus again, it appears ambiguity is the optimal policy. The letter needs to encourage promptness and does so with a deadline and accounts of bad luck for tardiness: at the same time it needs to encourage late compliance and does this with a Lose-Win testimonial.

Typical Send-a-Dime money chain letter, this mailed May 8, 1935 in Beaumont, Texas. Archive item 1935.

Money chain letters have greatly diversified since 1935 and many billions have been mailed worldwide. I have collected examples from England that are derived from the American Send-a-Dime letter. These ask for a sixpence, bear a list of five names and addresses instead of six, and reassure the reader that there is "no further assessment or catch" to the procedure [1935e1, 1935e2]. A 1988 study describes contemporary letters and attitudes of the "players" (Boles & Myers). Money chain letters have also invaded the Internet in great numbers [2001].

Send-a-Dime crash.
The Send-a-Dime chain letter craze of 1935 peaked in the weeks after it received its first newspaper coverage (April 19 in Denver, April 21 in New York). Denver restaurant owner A. McVittie received 2,363 copies in two days (April 27). However, by May 28 the New York Times was reporting "Chain-letter fad on wane." Hopes to make money using Send-a-Dime quickly collapsed and the letters became a public nuisance. Parodies circulated that mocked the process and expressed intimidating hostility to senders [1935]. Despite the subsequent long term survival of money chain letters, surely there came a time in the mid 1930's when chain letters were quickly discarded with little examination.

Identity crisis.
While money chain letters were being contemptuously discarded by the millions, what was happening to the prevailing Prosperity luck chain letters? Many must have suffered the same fate since they looked so much like Send-a-Dime. After all, Send-a-Dime was modeled on one of them, with the leading list of six names. Here is a challenge for you: for a luck chain letter to survive then, how could it convincingly distinguish itself from the reviled and illegal Send-a-Dime money chain letters?

Identification by command.
You may suggest a prominent disclaimer. But money chain letters often claim that they are not a chain letter [1978]. Probably from thousands of variations there appeared "Do not send money" on a 1939 Flanders-Prosperity type letter. In 1952 an observant reporter, writing about Send-a-Dime, noted that almost all luck chain letters had this sentence (Nelson), as all do now. It mattered little, for replication, whether a recipient sent money or not. The replicative power of "Do not send money" resided in its distinguishing the letter it was on from a money chain. What other sentence could so decisively inform a reader that the letter in hand did not ask for money? This use of a command for identification is a striking example of the creativity of the folk process. The contents of a traditional chain letter are not understood by their literal meaning, but by their affect on circulation.

Once "Do not send money" predominated on luck chain letters, they could not nourish hope of bringing in money by some rational means. And the author of Send-a-Dime, the fabulous Jane Doe, had removed all mention of good luck or bad luck, and removed all testimonials. Thus luck chain letters and money chain letters parted ways, and since this splitting of the motivational stream they have radically diverged. Testimonials would re-appear on money chain letters, but instead of third person tales of good and bad luck we now see first person fictions of riches gained. No luck is involved in these stories - getting rich allegedly follows by cause and effect if you obey the instructions.

A Demon Appears.
A discovery in 2003 reveals that there was an early experiment (1936?) in asking for money and threatening bad luck for noncompliance. The following "actual letter found among some mementos" appeared in the March 1977 issue of the nostalgia magazine Good Old Days (Esther Norman, 1977). The initial list of five names and addresses was not published.

But could such a chain letter long survive, and if so, what might it evolve toward? Surely the most malignant combination of the many motivations for complying with a chain letter is the joining of money with fear. A grim evolutionary potential of a such a letter is extortion, real or simulated. An internet informant told me (1996) that a hundred quota money chain letter existed in India that, as they recalled, contained bad luck warnings against breaking the chain. And there is the curious suicide of Cecil Headlee, 39, who "shot and killed himself because he thought 'a mob was going to get him for breaking the chain' " (DP, 1935). This brief account does not tell us what type of chain letter was involved, but considering the early date, May 15, 1935, probably the fear came from Mr. Headlee's head, rather than through the mails. But there is no reason to doubt the authenticity of the Norman letter given above, and thus likely threatening versions did circulate. However I suspect these would have been brought to the attention of the postal inspectors. A "send money or die" letter would surely have concerned them, and thus any such demon may have been squashed as it emerged from its egg.

Though luck and money chain letters diverged in content after the 1935 Send-a-Dime craze, some features have appeared in luck chain letters that may have had special appeal to people involved with money chain letters.

Money Influences.
Consider the following three lines of text from our standard example of the Luck By Mail type - the 1952 chain letter published by Herbert Halpert.

Money letters setting higher antes probably followed transmission paths of Send-a-Dime in 1935. In 1978, the $1000 ante pyramid scheme "Circle of Abundance" followed the earlier $100 ante "Circle of Gold" (Marks 1978). It is easy to see why this happened. Pyramid schemes are illegal and flow along social contacts in the early going. Thus the lower ante Circle of Gold established a tree of trusted prior participants that was available to the same initial "circle" for the higher ante next round. Of course this second tree was pruned of all those who lost money in the first round. There may also have been a documented attempt to boost the Circle of Gold scheme by utilizing a circulating luck chain letter. On a prevalent Death-Lottery example, this is a final postscript: "May you continue to be encircled in gold." This is added to a published letter which has no date, but by its content I estimate to be around 1980, during which Circle of Gold was still active in some states.

By 1952, the paths of prior money letters had left loosely connected transmission networks totaling perhaps 10 million people, some links going back to 1935. The Halpert letter, and variants, may have surged through these networks, also spinning off letters into the general public in sufficient quantity to kill off all its luck chain letter cousins by immunization.

Design or Self-organization?
Apart from suggestive bits of text, there is no other direct evidence that luck chain letters ever followed money chain letters. If we had all the chain letters received by fifty people over the years, then by examining postmarks the phenomenon might be confirmed. Even if this symbiotic distribution could be proved, that does not prove someone deliberately modified a luck chain letter to serve as a follow-up to a money letter. For if among the thousands of variations present in the chain letter population, some happen to appeal more to money chain letter players, then these will proliferate in the money chain transmission networks. Thus it is difficult to decide if some features are the product of calculating human design, or the product of selection from a vast pool of accidental or uncalculated variations.

In this section the origin of the "Media Chain Letter" is described in detail. Many famous people participated in this chain letter around 1990. Its complex history involved hyper-competitive innovations, major deletions, and migrations to new genres. At one phase of its development, two traditions of text copying were joined together: a mainline luck chain letter was placed at the bottom of an office humor item. Since some of the text on the chain letter goes back to the beginning of the twentieth century, I will start with that.

Luck Chain Letter (LCL)
The "Luck by Mail" type of chain letter developed a few years after the end of World War II and dominated circulation on up through most of the 1960's. Toward the end of that decade someone received such a letter and kept it for a few years. It may have read somewhat like the following example which was received by William F. Hansen in 1967.

Scores of published examples of "Food for Thought" appear when one searches newspapers.com using "greatest wheat speculator". Many are identical to that given above from the Billy Rose column. Other syndicated columnists, such as Norman Vincent Peale, later gave other versions. Some appear only locally. For example, a minister might use "Food for Thought" in a newspaper sermon. Variations include alternative titles, such as "Something to Think About" or "The Deceitfulness of Riches". The number of tycoons may be reduced, and the conclusion may be rewritten. If an alleged source is given for FFT, it is inaccurate and usually not dated. The publisher may have obtained FFT from other published versions, or received it as an undocumented clipping, or saw it on a bulletin board.

When an anonymous piece is based on other published versions, and when it is often changed by a publisher (e.g. condensed, rewritten, retitled, etc.), I will call it a "chain publication". Most Himmelsbrief had such a publication history well into the 20th century, though they may not have been changed as often as FFT.

Many of the names on the cover letters were so well known that the packet of these with the chain letter was called: "The Media Chain Letter" (Joseph Nocera, New Republic, Nov. 12, 1990); "The Chain Letter of the Rich and Famous" (Diedre Fanning, New York Times, Oct. 7, 1990); and "The VIP Chain Letter" (Charlie Clark, Washington Post, Nov. 16, 1991). The oldest example of the Media Chain Letter in the archive contains many names of celebrities and executives in the entertainment industry [1990]. But the oldest cover letter in the stack was sent by Pierre Salinger, former Press Secretary for President Kennedy. For narrative convenience, I assume that the celebrity attributions are reliable. However it should be kept in mind that the signatures present are all corrupted photocopies and hence unverifiable. Another Media Chain Letter bundle [1990] circulated in American real estate and investment firms, crossed over the Atlantic, and returned. Many comments are musings on the utility of good luck. "Some of us in the securities and real estate businesses forgot that it's better to be lucky than smart." Another theme, common in published reports on the Media Chain, is admission of fear of bad luck. "A man will do anything out of fear" (Newspaper editor). Most reporters accepted these comments at face value: "Media Barons Knuckle Under Superstition" (AP headline, Aug. 29, 1990). But some observers noted social reasons for the letter's success.

Status motives were present with other chain letters. Many copies of the Good Luck chain contained long lists of "X to Y," so that if you are Y, any downline recipient can conclude that X knows you [1922]. A striking later example has 113 different names. These start out with Japanese naval officers, phase into a European venue, and half way through shift to silent film celebrities such as Sid Graumann, Harold Lloyd and Mac Sennet [1926]. The half serious 1979 Brill chain [1979] had no real threat but many entertainment industry stars in its list of prior senders. Perhaps many of the listed celebrities never actually sent out the Good Luck or Brill letters. But forwarding letters with the Media chain were on corporate stationery and usually signed, thus providing a more convincing display for identification with celebrity.

The Media Chain Letter continued to progress through hierarchies for several years. A note in the British Medical Journal (March 25, 1995) complained of its "wad of memos." Infected organizations included the Ministry of Defence, the Metropolitan Police and the National Health Service. Like chickenpox, the Media chain letter is usually a one-time infection, leaving a trail of immunity. As the paper form of this chain letter worked out its extinction, the text crossed over into the vast new territory of the Internet.

Though I have not located an example, probably sometime in the early 1990's FFT+GS+LCL was digitized and transmitted via E-mail. Many photocopied office humor items have also crossed over to the Internet. A July 1993 chain E-mail consists of a series of ironic questions, later versions of which are titled "Why ask why?" By mid 1995 this item had been pasted on the beginning of FFT+GS+LCL [e1995]. Since then the symbiont chain letter has been attached to other email jokes, including "The Gift" [e1995-12], and "The Lucky Frog":

In 1979 a certain offhand postscript first appears on copies of the prevailing luck chain letter. In a few years it is on all circulating luck chain letters, millions of them. All the letters without it are gone. I examine this event in detail.

The rise of the "It Works" postscript.
In the 1970's four structurally related types of luck chain letters circulated in the US:

The earliest IWP letter in the archive was mailed anonymously in May 1979 to my California address. It was a photocopy of a typed letter. Here are the exact keystrokes - the format has been shortened.

There is much reason to believe that the DL type compound letter was initially formed in the English language. The same is true for the initial appearance of IWP, perhaps months or even a year prior to the above letter. It is unlikely that successive translations into English could have produced the series of letters in the archive. But close translations from the English letters exist in French [1995], Spanish [1996], Polish [1992], Italian and likely many other languages. This itself is evidence that there were no indigenous predecessors in these languages. Descendants also invaded the Internet in nearly word-for-word form, at first still asking for 20 copies. But in this medium the entire postscript was soon deleted and replaced by "You may not sign on this message". [e1994]

The following archive tabulations document changes in the population of chain letters around this time using mostly two year intervals. The first four abbreviations are for the chain letter types listed above. The next five are for key textual innovations within the DL type.

D: All Death20 type letters
L: All Lottery24 type letters (none collected so far).
LD: All Lottery-Death type letters
DL: All Death-Lottery type letters.

DL-N:    Death-Lottery (DL) letters with a list of names and none of the key postscripts.
DL-0:    DL letters with no list of names and none of the key postscripts.
DL-1:      DL letters concluding with "Do not send money" or variants.
DL-12:   DL letters concluding with "Do not send money. Please do not ignore this" or variants.
DL-123: DL letters concluding with "Do not send money. Please do not ignore this. It works" or variants.

Years	D	L	LD	DL	DL-N	DL-0	DL-1	DL-12	DL-123
1959 - 1971	5	0	0	0	0	0	0	0	0
1972 - 1973	2	0	0	1	1	0	0	0	0
1974 - 1975	1	0	13	4	4	0	0	0	0
1976 - 1977	2	0	0	3	0	3	0	0	0
1978 - 1979	0	0	0	7	1	2	1	0	3
1980 - 1981	0	0	0	8	0	0	0	0	8
1982 - 1983	0	0	0	14	0	0	0(a)	0	14
1984 - 2005	0	0	0	138	0	0	0	0 (b)	138

(a) One letter likely had "Please do not ignore this - It works" deleted in its ancestry [1982-01]. It is tabulated as DL-123.
(b) Several letters likely had the two words "It works" deleted [1984-08]. They are also tabulated as DL-123.

Inferred Relatedness Using Text Alternatives
The above counts of occurrences in the archive establish without doubt that chain letters bearing IWP captured the gigantic American luck chain letter niche within a few years of their appearance. But a more thorough understanding of what happened requires looking at details on early IWP letters and their predecessors. For example, note (a) of Table 4 claiming that the letter 1982-01 was likely an IWP descendent despite its shortened postscript can be supported. It is sufficient to show its near identity with a letter bearing IWP in full, and its significant variation from all DL letters that never bore IWP. Thus we seek identifying features of the IWP clade - their presence on a chain letter being very unlikely to have come about by any means other than copying from another member of the clade.

The following "text alternatives" table facilitates assessing the relatedness of chain letters under consideration. Six text alternatives (T1 to T6), listed in the column heads, can be either version A/a or version B/b for each of 16 selected DL letters listed in the rows. An "x" in the table means that deletions or other changes to a letter have nullified the possibility of determining a text alternative. An apostrophe (single quotation mark) after a text indicator, say b', designates a variation of b that is not of significance for the purposes of the table. The chain letters are grouped by variation, and chronologically ordered within each variation. First the definitions of the six text alternatives are set.

The rows of a text alternatives table are condensed descriptions of parts of chain letters, the alternatives and their names being selected here to facilitate human judgments on relatedness. But concatenated strings of these alternative names could themselves be input to a computer program, such as the one employed in Bennett, that calculates a symmetric matrix giving the relatedness of all pairs of letters in a sample. For this purpose: (1) the text alternatives should encompass all text in the letter that has variations in the sample, (2) names of the text alternatives should be all different, and (3) the number of letters in these names should correlate with the unliklihood that the alternative could be independently invented. Then an evolutionary tree could be constructed, by various methods, which would not exaggerate the effect on relatedness of long innovations in a letter.

T1 = a: You will receive good luck within 4 days of receiving this letter.
T1 = B: You will receive good luck within four days of receiving this letter provided in turn you send it back out.
Note: When a text alternative, such as T1 = B above, is deemed to be a feature (very unlikely to have appeared more than once without having been copied from an existing letter), it is designated by an upper case letter.

T2 = a: Send twenty copies of this letter to people you think need good luck.
T2 = b: Send copies of this letter to people you think need good luck.
Note: These alternatives are not "features" since they can easily change a to b, or b to a. But they add some weight to a conclusion about relatedness.

T3 = a: Please do not send money.
T3 = B: Do not send money, for fate (faith) has no price on it.
Note: Either "fate" or "faith" may appear in the added clause.

T4 = a: It must leave you within 96 hours.
T4 = B: It must leave your hands within 96 hours.
Note: In T4 = a, the word "you" may be omitted.

T5 = a: This chain . . . was written by St. Aptine de Cade . . .
T5 = B: This chain . . . was written by Saul Anthony De Cadif . . .
T5 = b' This chain . . . was written by Sol (Soul) Anthony De Cacief . . .
Note: Only the first name (or title) is useful here. The last names are highly variable.

T6 = a: . . . a missionary from South America. Since the chain must make a tour . . .
T6 = B: . . . a missionary from South America. I, myself, now forward it to you. Since the chain must make a tour...
Note: "I, myself, now forward it to you" was deleted around 1984 and is absent thereafter on all DL letters.

First I use information in Table 5 to justify the claims in Table 4 that deletions of postscripts have occurred on letters #12, #13, #14, and #15 (Table 5 designations) . Note that the earliest DL-123 type letter, #8, introduced the diagnostic variations T1 = B, T3 = B, and T6 = B. Thus it seems very unlikely that letter #12 could bear these same B versions without having copied them from some DL-123 letter. So we conclude that letter #12 had "Please do not ignore this. It works" deleted at some time in its ancestry. Similar reasoning suggests that #13, #14, #15 (and other post 1980 DL-12 letters in the archive) have had the entire IWP deleted, perhaps because of image expansion or misaligned photocopying. To claim otherwise would require much independent invention, such as the diagnostic changes T1 = B and T3 = B, and the name "Saul" (T5 = B). Later DL-12 letters do not bear the Linkage statement (T6 = B) that the early DL-123 letters did, but as noted above, T6 = B disappeared completely around 1984 from all DL letters.

Next I will use Tables 4 & 5 to address some questions about the IWP capture of the luck chain letter niche.

Question 1: Did the three parts of IWP appear all at once, or in two or three phases?
It is likely that IWP originated in two phases. First the postscript "Do not send money" was added to a DL letter. One descendant of this letter is present in the archive: le1979-01, chain letter #7 in Table 5. This is not a result of deletions from a DL-123 letter for it contains the variations T1 = a, T3 = a and T6 = a. There is not a single undeleted example of a DL-12 letter in the archive. Thus it is very likely that IWP originated in two phases: "Do not send money" and then "Please do not ignore this. It works." I may continue to use the symbol "DL-12" to represent letters with this text, in spite of its ancestry.

Question 2: Was IWP initially placed on just one circulating letter, or more than one?
Probably just one, because all the many examples collected except one have the exact same "B" options for T1, T2, T3, T4, T5 and T6 (or T6 = X). If IWP had been placed on another circulating letter, very likely some of the features of this letter would differ from the "B" option. The one exception is 1981-04, a DL-123 letter with T1 = a. But this likely resulted from a deletion of the added phrase "provided in turn you send it out" (T1 = B), changing it to T1 = a. This deletion could have been accidental when retyping, or perhaps was a deliberate reversion to the more familiar "copy later" chain letter orthodoxy. In any case, it is hard to imagine why someone would go to the trouble of adding IWP to multiple letters. Unlike, say, a Bible verse or "With love all things are possible", the IWP postscripts have the look of offhand comments to a friend rather than a cause that merits promotional effort.

Question 3: Were there one or more other significant innovations that were launched in the same letter that first bore IWP?
The first DL-123 letter in Table 5, le1979-05, introduces not only IWP but also the diagnostic innovations I have labeled T1 = B, T3 = B and T6 = B. The other tabulated alternatives, T2 = b, T4 = B and T5 = b', appear on prior DL-0 and DL-1 letters. It can never be shown with certainty that any two features appeared concurrently. Likely there was some uncollected letter which contained, say, T1 = B, but not IWP, or visa versa. Concurrence would say something about the motives of an innovator. But it makes little difference to the resulting population of chain letters if two innovations were co-linked, or if one was post-linked shortly after the debut of the other. In any case, it appears the "It Works" innovator may have made two or three other significant changes to the text.

Question 4: Was IWP itself the main reason the letters bearing it proliferated so greatly, or was there another feature on these letters that was the main reason?
Estimating the effect of a feature on the circulation of its chain letter host is largely a matter of informed guesswork. Unknown motives of readers may be involved. We will attempt to answer Question 4 by considering a broader question in the next subsection.

Why did "It Works" Work?
The tabulated data of Table 4 suggest that each of three successive innovations (DL-0, DL-1, and DL-123) increased circulation. Here are likely reasons for each.

DL-0: The senders list on DL-N letters was an awkward presence, especially since photocopying was rapidly becoming the dominant method of replication. When a letter must be typed or hand copied, it is not that much additional bother to update a list of names. But if one is photocopying, instructions to revise the letter in any way may prompt discard. Further, asking for one's name is not consistent with anonymous distribution, which prevailed in subsequent years. Most of the lists on DL-N were internal, between the D and L blocks. Deleting this gives the resulting text the appearance of a single chain letter, making the contradictions between the D and L blocks less likely to be observed.

DL-1: When first taking a letter in hand, a reader often may glance at the end of the letter, looking to see who signed it. This gives postscripts special potency. The DL-1 postscript, "Do not send money", was a highly visible flag signaling that the letter was not a solicitation for money.

DL-23: "Please do not ignore this. It works" together constitute a polite affirmation of the power of the letter, in contrast to the previous bossy conclusion: "For no reason whatsoever should this chain be broken." This was retained, but now was much less prominent. A money chain letter, attributed to "Nelson Robards", was circulating at the time that concluded with "Do it . . . it really does work" [1978]. Perhaps "It Works" luck chain letters "followed" this money chain (< Section 4-3), each increasing the other's circulation. "It does work" is a common variant of the third sentence of IWP, and is the version on the two oldest examples in the archive. DL-23 constituted a revival of the "Affirmation" type statements that were common on luck chain letters from 1927-1940. These functioned as an independent voice affirming the claims of the letter. Many radio and television advertisements use the phrase "It works."

I now consider whether the identifying features T1 = B, T3 = B or T6 = B had a significant affect on the circulation of DL-123.

T1 = B: Could the added proviso to the promise of good luck ("You will receive good luck within four days of receiving this letter provided in turn you send it back out) have been the the major cause of the success of DL-123? This appears early in the letter, and fundamentally changes its operating superstition from a lucky talisman (Copy later) to a fateful obligation (Copy first). There is one, and only one, letter on which it appears without the complete It Works postscript. [1982-01]. Whatever the genealogy of this letter, it did not replicate sufficiently to show up again in the archive, something we should expect if the "provided" clause were highly effective by itself. However, the potency of the full postscript is itself challenged by the rapid increase of a variation that deleted the last two words ("It works"). This deletion, co-linked to a transfer of the Kiss title, appears in the "KCL" variation of the 1990's (> S4-6_kcl). But the KCL letters begin with the Kiss title and conclude with the Love title, thus combining the appeal of both. This may more than make up for the loss of "It works." Another reason for assessing T1 = B to be more effective than IWP is that, as noted above, it stayed on a widespread Internet version of the letter whereas all the postscripts were soon deleted without consequence. [e1994]

T3 = B: This added phrase ("Do not send money, for fate has no price on it) may function like an Affirmation, but readers would not need to be further convinced that money should not be sent. Perhaps some would be impressed by this kitsch flourish, but in any case I do not see how it could affect circulation significantly.

T6 = B. The Linkage ("I, myself, now forward it to you") is awkwardly placed in the L block of DL-123. It is not surprising that by 1984 it had been deleted, perhaps deliberately by the author of a highly replicative new title. It is always present on early DL-123 letters, but likely had no role in the replicative success of its host.

So the answer to Question 4 above is that there was one other innovation on letters with IWP that could have been the main reason for their success - the added proviso that good luck requires compliance first (T1 = B). But it is near impossible to judge if this, or the postscripts, motivated the most compliance.

Introduction.
After 1981 almost all of the millions of luck chain letters circulating in the United States were the descendants of a single founder that had appeared a couple years before bearing the It Works postscript (IWP) and a "copy first" proviso (< Section 4-5). In this final section seven ensuing changes in this population are described, concluding with its near extinction at the end of the millennium. I give the following allegorical names to these events: (1) Trust expires, (2) Belief fizzles, (3) Kiss and Love divide the territory, (4) Kiss gets wife's money, (5) Love gets a car, (6) Kiss jumps on top, and (7) All fall down. Except for the last, each of these events involve one or more of the following named innovations.

The obvious changes in frequency in Table 6, including remarkable extinctions, are the basis for characterizing the post-1981 history of the English language DL letters with the seven events named above. I now consider them one-by-one in the next seven subsections. This analysis is detailed; the reader may wish to skip this part and go directly to the final subsection of this treatise (>4-6all_fall_down).

1. Trust expires.
Proverbs 3:5-6 had appeared near the top of most mainline letters since 1952 in countless corrupted forms until the It Works capture in 1979 fixed its form as "Trust" given above. This header probably had positive replicative effect at first, but by the early 1980's it seems to have become a liability. Perhaps quick recognition of a letter as the nuisance of the prior decade was a factor.

2. Belief fizzles.
Someone replaced Trust with "Belief" in 1980, placing this New Testament verse above the body of the letter. On our earliest version [1980] the verse is identified correctly as Matthew 21:22. "Belief" was immediately successful at the expense of Trust, but within two years its circulation seems to have ended. This was probably due to the appeal of two new secular titles I discuss below. The omission of the Unbeliever's Death was an early post-linked feature [1981] that apparently captured the Belief clade. This may have also contributed to the demise of Belief.

3. Kiss and Love divide the territory.
The earliest Kiss title in the archive is from October 1983, and the earliest Love title was sent from Sweden to England in June, 1983. Both these letters, as well as three letters from earlier in 1983 that had no title [1983-02], have nearly identical text details (other than the titles), yet they differ markedly from the prior DL letters with either the Trust or Belief titles. The table below facilitates comparisons. Ten text alternatives are tabulated for a selection of seventeen chain letters from the IWP clade. Again, text alternatives considered diagnostic are designated by upper case letters.

Table 7. Text Alternatives on DL Title Variations

T1 = a: This prayer has been sent to you for good luck.
T1 = b: This quote has been sent to you for good luck.
T1 = c: This paper has been sent to you for good luck.
T1 = d: This letter has been sent to you for good luck.

T2 = a: The original copy is from the Netherlands.
T2 = C: The original copy is in New England.

T3 = a: The luck has now been brought to you.
T3 = c: The luck has now been sent to you.

T4 = a: . . . provided you in turn send it back out.
T4 = c: . . . providing you, in turn, send it on.

T5 = a: . . . Welch lost his life six days after he received this letter.
T5 = b: . . . Welch lost his wife six days after receiving this letter.

T6 = a: Please send 20 copies to see what happens ...
T6 = c: Please send 20 copies of the letter and see what happens ...

T7 = a: . . . send it to your friends, parents, or associates.
T7 = c: . . . send them to friends and associates.

T9 = a: He asked his secretary to make 20 copies and send them.
T9 = c: He asked his secretary to make 20 copies and send them out.

T10 = a: Carlo Raditt . . . forgot it and a few days later, lost his job.
T10 = C: Carla Dadditt . . . forgot it had to leave his hands within 96 hours. He lost his job.

No	Chain Letter File	Type, Variation	T1	T2	T3	T4	T5	T6	T7	T8	T9	T10
1	le1979-05_dl_tw!	DL-123 Trust title	a	a	a	a	a	a	a	a	a	a
2	le1982-01-28_01_m_t	DL-123 Trust title	a	a	a	a	a	a	a	a	a	a
3	le1982-09-20_dl_wt_dates	DL-123 Trust title	a	a	c	a	b	a	a	a	a	a
4	le1983-06-30_dl_wt'	DL-123 Trust title	a	a	a	a	a	a	a	a	c	a
5	le1980_dl_wb!	DL-123 Belief title	b	a	a	a	a	a	a	a	a	a
6	le1980u_dl_B-L29_wb	DL-123 Belief title	b	a	a	a	b	a	a	a	a	a
7	le1981u_dl-_wb	DL-123 Belief title XU	b	a	a	a	a	c	a'	a	c'	a
8	le1982_dl_wb	DL-123 Belief title XU	b	a	a	a	a	c	a'	a	c	a
9	le1983-02_dl_w(k)	DL-123 Untitled	c	C	c	a	b	c	c	x	c	C
10	le1983-04_dl_w(k)	DL-123 Untitled	d	C	c	a	b	c	c	C	c	C
11	le1983-05_dl_w(k)	DL-123 Untitled	d	C	c	a	b	c	c	C	c	C
12	le1983-10-04_d_wk!	DL-12 Kiss title	d	C	c	a	b	c	c	C	c	C
13	le1984-05_dl_wk	DL-123 Kiss title	c	C	c	a	b	c	x	C	c	C
14	le1985-06_dl_wk	DL-123 Kiss title	d	C	c	a	b	c	c	C	c	C
15	le1983-06-13_dl_wl!_e	DL-123 Love title	d	C	c	c	b	c	c	C	C	C
16	le1983_dl_wl	DL-123 Love title	c	C	c	c	b	c	c	C	C	C
17	le1984-05_dl_wl	D-L123 Love title	c	C	c	c	b	c	c	C	C	C

Note that the four Trust letters (#1 to #4) have almost all the text alternatives the same, except for three cases (for example, T3 = c for letter #3). None of these three exceptions are diagnostic, and could easily be "pop ups". The second two Belief letters (#7 & #8) have the Unbelievers Death testimonial deleted (variation XU). This is two sentences long, so this was very likely deliberate - probably for ethical reasons. The absence of this threat was likely negative for circulation and could explain the early disappearance of this sub-type from the archive. These two Belief letters have another variation in common - the innovation T6 = c ( "Please send 20 copies of the letter and see what happens ..."). Since all the remaining letters in the table (#9 to #17), including the three Untitled letters (#9 to #11), also have T6 = c, we can suspect that all these remaining letters have a Belief ancestor. The cladogram of Bennett, Li and Ma (2003) displays the same relationship. Based on the information in the table, the following is a parsimonious hypothesis for the origin of the five title options, and the text alternatives.

(1) In 1980 or before, someone replaced the Trust title with Belief, and changed the self-reference from "prayer" to "quote". (T1 = b, see letter #5)
(2) Not long after, someone casually inserted the innovation T6 = c ( "Please send 20 copies of the letter and see what happens ..."). (# 7 & #8)
(3) In 1982 or early 1983, the title from one of these T6 = c letters was deleted from the top of a photocopy. The self-reference "quote" was retained.
(4) Soon after, a copyist decided to correct the self-reference since there was no longer a "quote". They made the odd choice of "paper". (#9)
(5) Around the same time the Kiss title was added to an Untitled letter. (#12)
(6) Around the same time the Love title was added to an Untitled letter. (#16)
(7) At various times, copyists were not satisfied with the self-reference "paper", and changed this to "letter" on Untitled, Kiss, and Love letters. But "paper" remained the most common self-reference. (#11, #12, #16)

Some other scenarios seem about as likely. The occurrences of the self-references "paper" versus "letter" are the greatest challenge to parsimony. A tempting alternative phylogeny has "paper" a miscopy of "prayer" on an uncollected Trust letter, and this lineage losing its title, accumulating the "B" alternatives and receiving the two new Kiss and Love titles.

Most of the ten text alternatives in the above table, though often corrupted, are identifiable on every mainline letter after 1984. For example, on 1998-07 "Gene Welch" has been corrupted to "George Welch," but we never again see "General" or "Gen." as was present in the 1970's. Even minor changes such as "copy" for "chain" have persisted. This is a remarkable demonstration of 15 years of faithful copying, and proof that the progeny of the Kiss-Love founder completely captured the mainline. This constituted a descent group of over a billion letters.

Why were the Kiss and Love titles so successful? Did some of the above linked features also contribute significantly to the circulation of these letters? Examining the ten text alternatives used in the table, few if any seem as if they could have impacted propagation as much as the five titles (Trust, Belief, Untitled, Kiss or Love). But we will consider three candidates.

(1) T5 = b: " ... Welch lost his wife six days after receiving ...".
The alternative "wife" was present on the Kiss-Love founder and thus became universal in the mainline, never reverting back to "life". The brief success of the pre-war Blind13 type chain postcard, which threatened a family member with blindness, suggests that threatening the reader's family gains more compliance than threatening the reader. But "wife" had appeared in 1975 and probably many other times uncollected, without getting established.

(2) T7 = c: "...send it to your friends or associates."
"Parents" was dropped from the distribution in the Kiss-Love clade. This likely could not have had a major impact on circulation, but it could have had some indirect positive effect.

Consider the participant age of a social activity - for chain letters the average age of those who are replicating it. Participant age may regress (marijuana smoking), remain fixed (school traditions), or advance with time (canasta), perhaps even keep pace with calendar time (class reunions). There is a postcard exchange letter (an ancestor of the kids' World Record chain letter) whose participant age regressed. You be the judge: here is some text from an example received by an eleven year old in Clarkston, WA:

The origin of S$ may depend on a simple mistake. It would be very easy to shift the gender of the last pronoun from "he" to "she" in Death and Money:

Remarkably, this new version of the Kiss letter, S$, captured the clade within a few years, as seen in Table 6 above (< Occurrences). From 1984 to 1998 there are 24 examples of S$ in the archive, but only 8 examples of H$. And H$ completely disappears from the archive by 1990. Careful comparison of letters bearing H$ and S$ reveals no new variation on S$ letters, other than this modification to the Death and Money testimonial, that could explain this. There is the usual alternation between the self-reference "paper' and "letter", and a few non-diagnostic additions like "The luck has now been sent to you." So why did the S$ testimonial get more people to copy a letter than the H$ testimonial?

We have already noted that a threat to a family member may be more fearful than a threat to oneself. But both H$ and S$ have that. Perhaps the conversion to Copy First superstition in S$ made the testimonial more effective, especially to the younger readers that may have disproportionally circulated the Kiss title. But consider someone who has never read the Death and Money testimonial before, or has but does not remember it. There is a key difference between H$ and S$ other than a gender shift. S$ mentions the winning of a lottery - a first for the leading D block in the ubiquitous DL chain letter dynasty. Yes, Mrs. Welch wins and dies, but the second fate is easily averted by distributing copies. And since this lottery mention is much closer to the beginning of the letter than the second in the L block (Boss Wins Lottery), many will read it who otherwise would have stopped reading before even getting to the L block. Further, this winner is an Anglo woman, instead of a Latino male, as "Constantine Diaz" is in the "L" block. Thus with S$ there are more possibilities of a reader identifying with the promise that the letter may help them win a lottery. If a man is reading it, his wife may already be holding a lottery ticket. Recall the "One in a Hundred Rule", all that is needed for a quota 20 chain letter variation to become hyper-competitive is for one additional person in a hundred to become fully compliant.

The Car testimonial is usually self-dated 1987, and first appears in our sample in July 1988, appended near the end of a Love titled letter [1988]. Not a single Love titled letter in the archive thereafter fails to possess it! Despite this quick conquest of the Love subtype, the Kiss titled letters persisted without Car. This supports the above speculations that these two titles had a different audience, perhaps differing by motive, age, or gender. If Kiss and Love had the same motivational niche, Love & Car would have swamped Kiss just like it did Love without Car ("No Car"). I discussed the replicative advantages of Car in Section 3.6 (< Mainline Testimonials).

For the table below I have identified seven text alternatives which may distinguish Love-Car examples from No Car. Thirteen chain letters which date from near the advent of Car are checked for which alternatives are present on them.

T1 = a: The luck has now been sent to you.
T1 = b: The luck has been sent to you.

T2 = a: You will receive good luck within four days of receiving this letter, providing you, in turn, send it on.
T2 = b: You will receive good luck within four days of receiving this letter, providing you send it on.

T3 = a: This is no joke. You will receive it in the mail.
T3 = b: This is no joke. You will receive good luck in the mail.

T4 = a: . . . in the mail. Send copies to people you think need good luck.
T4 = B: . . . in the mail. Send no money. Send copies to people you think need good luck.

T5 = a: While in the Philippines Gene Welch lost his wife six days after receiving the letter.
T5 = b: While in the Philippines, Gene Welch lost his wife 51 days after receiving the letter.

T6 = a: Since a copy must make a tour of the world you must make twenty copies and . . .
T6 = b: Since the copy must tour the world, you must make 20 copies, and . . .

T7 = a: Please don't ignore this. It works.
T7 = b: Do not ignore this. St. Jude. It works.
Note: "St. Jude" is also added to top or at the extreme bottom of some letters.

First, note that the Car testimonial must have been added to a letter similar to #5 (le1987-06_wlj!) since this earlier letter introduces the alternatives T3=T6=T7=b and the diagnostic T4 = B seen later on the first Car in the archive, #7. T1=b and T2=b are also pre-linked to Car (on letters #2 and #4), and the earliest Car letters apparently had T5=a. So the first Car must have looked much like the No Car letter #5. This is fairly strong evidence that whoever placed the composed testimonial Car on a chain letter did not take the opportunity to make other changes in the letter. It also suggests that Car was placed on only one letter, for if it had been placed on two different circulating letters likely at least one of the text alternatives T3, T4, T6 or T7 would have had an "a" value. But there is no descendant of such a letter in the archive.

Did any of the seven text alternatives on early Car contribute significantly to its impressive replication? Alternatives T1, T2, T3 and T6 are very likely neutral for propagation. Text alternative T4 = B above, the early "send no money," is probably slightly positive for propagation since it doubles and advances this universal prohibition. It may have originated accidentally in re-typing, first as an exact duplication of "Do not send money . . .", then edited to "Send no money" for stylistic variety.

The text alternative T5=b, ("Gene Welch lost his wife 51 days after receiving the letter") was discussed previously (< 51 days). It was not on the earliest Car letter in the archive, but appeared shortly after (on #8) and likely became common as a post-linked rider on the new testimonial.

The text T7=b ("St. Jude") first appears in the archive on the No Car #5 [le1987-06_dl_wlj!]. The Car testimonial was first added on a letter close to this letter and bearing "St. Jude". All Car letters thereafter bore "St. Jude" except for a few deletions [1994, 1997] and a Protestant substitution [1993]. There have also been a few independent appearances of "St. Jude" [1991, 1992] but none of these show up more than once in the archive. Thus by itself Jude is unreliable to use for inferring phylogeny. It may have been positive for propagation of the Car clade, or perhaps it was just a pre-linked rider. On Mexican chain letters "St. Jude Thadeus" appears to be an essential component. [1984, 1995] But the appeal of St. Jude to Latin Americans may not be so old. One informant called him the "Patron Saint of Anglos" in the 1950's. (Orsi)

This examination of 1990's DL letters has revealed at least five different transfers that both replicated and were collected for the archive. For the last two years of large circulation, 1996 & 1997, the archive contains 28 DL luck chain letters. Over half of these, 15, are descendants of one of the above transfers. So letters with both titles present were favored. Likely the transfer of text from one chain letter to another was common behavior throughout the twentieth century, and was an important factor in chain letter evolution. If paper luck chain letters had sustained their circulation for another few years, possibly the KLC transfers would have captured the entire North American luck chain letter niche.

Despite the success of the DL type letters in the early 1990's, the circulation of all paper luck chain letters dramatically declined beginning in the late 1990's. The following table documents this with numbers from the Paper Chain Letter Archive.

Table 9. Numbers of English language paper luck chain letters collected per year since 1995.

Year of circulation	Mainline	Outliers
1995	15	3
1996	20	0
1997	9	0
1998	1	2
1999	1	0
2000	0	1
2001	0	0
2002	0	0
2003	1	0
2004	3	1
2005	1	0
2006	0	0
2007	0	0
2008	1	0
2009	0	0
2010	0	0
2011	0	0

The last DL luck chain letter collected was titled "The Financial Blessings Letter" [2008]. It retains all DL testimonials except the Unbeliever's Death, deletes much other DL text (especially at the beginning), and has rewrites to promise a divine monetary reward for replication.

Though I reduced efforts to collect paper chain letters after 1997, still, if their circulation had been even at 10% of previous levels, many more would have been collected than were. The term "chain letter" is now universally applied to examples on the web, and authors often deem it necessary to explain to their readers that this term once meant actual paper letters. The primary cause for the disappearance of paper luck chain letters was the rise of new communications technologies. Here are three suggestions on how this operated.

(1) The flood of email and internet luck chain letters immunized people, especially the young, against the promises and threats of paper chain letters.
This was the basis of a 1995 prediction that "the familiar 'prayer' or 'good luck' type chain letters will totally disappear from the US mail by the year 2000" [e1995-06]. The public is now far more skeptical about chain letters than it was in, say, 1985.

(2) Compliant recipients of a paper chain letter chose to fulfill its demand for copies by sending it out in email form, or posting it on social media.
The effort and cost of complying electronically is minuscule compared to the what is required to mail out twenty paper letters. Many chain messages on the internet began as word for word transcriptions of paper chain letters. A very early example is a standard DL letter with the "Trust" title and the "It Works" postscript that appeared on the ARPA net. [1982] Email forwards used to list all the parties who relayed the message, and comments by individual forwarders were preserved. On this example we find a note stating: "This is the infamous ARPA-net chain letter which caused much havoc a few years ago on the ARPA-net." But it was not until the 1990's that personal computers proliferated and numerous paper chain letters were shunted into the digital realm.

(3) Computerized search technologies, such as Google, exposed recipients to critiques of chain letters.
Prior to the internet, very few people were able to access any comment at all on a chain letter they received. Not so now. My ISP reports give the search strings used that led to someone accessing any of my web pages, including the digitized texts of hundreds of chain letters. These reports confirm that receipt of a chain letter, paper or not, often prompts people to search the Web with some characteristic text from the letter. This was more frequent in years past, especially for luck chain letters. Then the searcher was likely to encounter unkind words about chain letters from a web vigilante. Wasting "bandwidth" was the usual complaint. Both paper and cyber money chain letters still circulate; an abundant example has the title "As seen on Oprah and 20/20". My archive entry for this chain letter shows up second in a Google search using the title. After giving the text, this page presents an analysis of the letter that discourages replication.

A century of denunciation failed to eliminate paper luck chain letters (New York Times: 1916, 1917e, 1931, 1959b). The Internet, email, and smart phones all but ended them in just a few years.

Chain Letter Evolution (THIS FILE): http://www.silcom.com/~barnowl/chain-letter/evolution.html

End: Chain Letter Evolution, Version ~~10/27/1998f~~ 11/04/2002p ~~1/7/2007~~ ~~12/9/2012~~ ~~5/2013~~ ~~6/2015~~
revised 3-1 10/2016

My terminology: G is _____ to H	All known presences of G and H in the clade.	Cladistic terminology: G is ____ relative to H	If H becomes universal, G becomes _______	Possible Origin	Spoilers
1. unlinked	{G,H}		absent	H was first added to a letter with G deleted.	GH exists but is uncollected.
2. pre-linked	{G,GH}	plesiomorphic, or ancestral	universal	G was on the letter that H was first added to.	All G are by deletion of H from GH.
3. co-linked	{GH}	congruent	universal	G and H first appeared on the same letter.	G or H exist but are uncollected.
4. post-linked	{H,GH}	apomorphic, or derived	frequent if GH came soon after H	H was on the letter that G was first added to.	H was transferred to a letter already bearing G.
5. transfer-linked	{G,H,GH}	homoplasitc, or conflicting	frequent if GH came soon after H	G was transferred to a letter bearing H.	All G or H are by deletion from GH.

No.	Chain Letter File Name	Type - Variation	T1	T2	T3	T4	T5	T6
1.	le1973-11p_dl!_prvrbs_q20n21	DL-N	a	a	a	a	a	a
2.	le1974-10_dl_n27	DL-N	a	a	a	a	a	a
3.	le1975_dl_n23_a1	DL-N	a	a	a	a	a	a
4.	le1977-06-08_dl_fgge_q20	DL-0	a	a	a	a	a	a
5.	le1978-04_dl_q20n0	DL-0	a	b	a	a	a	a
6.	le1979-06-21_dl_t_q20n0	DL-0	a	b	a	a	a	a
7.	l le1979-01_dl_xmoney(2)_q20	DL-1	a	b	a	B	b'	a
8.	le1979-05_dl_tw!	DL-123 It Does Work	B	b	B	B	B	B
9.	le1979-07-30p_dl_w't	DL-123 It Does Work	B	b	B	B	B	B
10.	le1980-12-23_dl_wt_dates	DL-123 It Works	B	b	B	B	B	B
11.	le1981-04_dl_wt	DL-123 It Does Work	a	b	B	B	B	B
12.	le1982-01-28_dl_m_t	DL-1 (2 & 3 deleted)	B	b	B	x	B	B
13.	le1984-08_dl_w't	DL-12 (3 deleted)	B	b	B	B	B	x
14.	le1986-05-08_dl_w-k'	DL-12 "	B	b	B	B	B	x
15.	le1991u_dl_w-(kc)!	DL-12 "	B	b	B	B	B	x
16.	le2004-10_dl_wklc	DL-123 It Works	B	b	B	B	B	x

Years	Trust	Belief	Kiss & His Money H$	Kiss & Wife's Money S$	Love & No Car	Love & Car	Kiss transfers to Love & Car
1972-1974	1	0	0	0	0	0	0
1975-1977	5	0	0	0	0	0	0
1978-1980	9	1	0	0	0	0	0
1981-1983	6	4	1	0	1	0	0
1984-1986	0	0	5	4	15	0	0
1987-1989	0	0	3	10	2	3	0
1990-1992	0	0	0	3	1	24	2
1993-1995	0	0	0	5	0	10	7
1996-1998	0	0	0	2	0	11	15
1999-2001	0	0	0	0	0	1*	0

No.	File name	Variation	T1	T2	T3	T4	T5	T6	T7
1	le1983-06-13_dl_wl!_e	Love - No Car	a	a	a	a	a	a	a
2	le1985_dl_wl	Love - No Car	a	b	a	a	a	a	a
3	le1985-12_dl_wl_h2pp	Love - No Car	a	a'	a	a	a	a	a
4	le1986-05-08_dl_w-k'	Love - No Car	b	a	a	a	a	a	a'
5	le1987-06_wlj!	Love - No Car	a	a	b	B	a	b	b
6	le1988_dl_wl	Love - No Car	a	a	a	a	a	a	a
7	le1988-07_dl_wlc!j	Love - Car	b	b	b	B	a	b	b
8	le1988_dl_w(l)cj	(Love) - Car	a	a	b	B	b	x	b
9	le1989-03_dl_wlcj	Love - Car	b	b	b	b	a	b	b
10	le1989-07_dl_wlcj_rec-fate	Love - Car	b	a	b	B	b	b	b
11	le1989_dl_wlcj	Love - Car	a	a	b	B	b	b	b
12	le1990-01_dl_w'lcj	Love - Car	a	a	b	B	b	b	b
13	le1990-08_dl_wlcj	Love - Car	a	a'	b	B	a	b	b

CHAIN LETTER EVOLUTION

TABLES