The Four Letter Word is an electronic light sculpture. It was created by Raymond Weisling in 1973, using Nixie tubes that had just then been removed from the New York Stock Exchange quotation boards, and were available on the surplus electronic market. These tubes are neon-filled cold-cathode devices that have individual cathode segments arranged to form letters when turned on in combination. The letters are 2.5 inches (6.3 cm) tall, and bright orange in color.

Why...  English has a certain love-hate relationship with its four-letter words. Many such words go back to its Anglo-Saxon background, perhaps even earlier, and often represent basic, simple things often with coarse or even vulgar meanings and connotations. One who has mastered the use of four-letter words is not typically viewed with respect and dignity. With this mystique, it was tempting to construct a device that displayed only four-letter words, letting the viewer enjoy the show, waiting for a nice juicy one that might not be kind to utter in mixed company. But in actual fact, the number of four-letter words with negative meanings is probably no more than one hundred or so, out of perhaps a total of around 2000. "Like a dear love, many will seem nice, pure, with a warm, soft glow."

History... The original Four Letter Word was designed using available technology of 1973. Approximately 60 units were produced and sold. It appeared in the June 1973 Playboy Potpourri. Units are known to have been in the collections of Stephen Sondheim and Alan Kaprow.

This is a photo of a small 5-inch black and white video monitor connected to the Apple II computer. The photo was taken in 1980. The characters were proportionally spaced and kerned for the best possible appearance. Reconstruction. By 1994 the historic and nostalgic attraction of these quaint Nixie tubes was a motivating factor in deciding to work on this piece again. A limited number of parts and completed units from the 1973-1974 production were still on hand, so it was decided to rebuild a new set and to use much newer microcomputer technology. The original model used two printed circuit boards, but advances in electronics allowed many more components to be placed on one new board.

As of this time (2002) an entirely new version has been completed and is available in kit form. Words are arbitrarily generated with a statistical weighting that favors English words, but some words are also drawn from an internal database of over 1200 selected words. Different processes of indeterminacy are used to make the selection, generation and display seem randomized and ever-changing.

Late Breaking News
The NEW version of the Four Letter Word
is NOW available as a kit of parts.
For more information click HERE.

Bad Words, Good Words. There are settings provided that allow the selection of words to be unbiased, to censor vulgar or impolite words and the opposite, to favor occasional inclusion of these impolite words at two different rates of occurrence. Personal preference for or against such aspects of languate usage can be set by the owner.

Variations. The original 1973 piece simply diplayed words at a fixed speed, one after another. This was altered so that some words remain on the display longer than others. This is the normal display mode. Two other word variations are provided. One is called "Solitare Letter Hangman". A word is generated internally, but only one letter at a time appears from a blank display, until the word is complete. The observer invariably plays along trying to guess what the word will end up being. The order in which the letters appear is indeterminate, different each time. Another related "guessing game" variation is the "Stick Segment Hangman" where the display also begins fully blanked but each of the straight-line segments is turned on, one by one, showing partial letters until the whole word is fully displayed. One more setting allows these three modes (normal plus two hangman variations) to be mixed, of course, indeterminately.

A Clock, Too. In addition to the word display, a digital clock display mode can be selected. The time (12-hour format only) is displayed approximately every 20 seconds, inserted between words when they change, with the numbers flashing rapidly a few times to attract attention. The clock display is meant to augment but not dominate the operation.

Nixie Tubes. The original source of the tubes had mostly dried up by the 1980's, but in 2001 thousands of the tubes were found in a warehouse, so there should be no problem obtaining tubes for this piece.

The original manufacturer of the type B-7971 Nixie tubes was Burroughs. Click to download JPEG images of the original manufacturer's data sheets from 1969 (about 100 KB each): Page 1 | Page 2 | Page 3 | Page 4.

Original Algorithm

In the original 1972 model, only certain letters appear in each position. A hand-tabulated analysis of several hundred four-letter words was done and the frequency of letters in each of the four positions was counted. The ten most frequent letters of each position were used. For each position, a simple decimal (0-9) counter (IC 7490) ran freely and the count was captured and held about once every few seconds to once per minute, depending on the speed setting. The counter snapshot was then applied to a diode decoding matrix (using about 150 diodes) to form the characters. Since only the ten most common letters were selected for each position, with no regard for letter combinations, many non-words can appear, e.g., FRLR, LREE, LLLL, etc. Note that the second position only had eight letters, since the ninth and tenth letters in the sample had a proportionally low frequency, while the first and second had extremely high frequency—so they were doubled. Thus 10 x 8 x 10 x 10 = 8000 permutations. These are the letters that were programmed into the first 1973 model:

1st Position Letters B C D F H L M P S T

2nd Position Letters A E H I L O R U

3rd Position Letters A C E I L N O R S T

4th Position Letters D E K L N P R S T Y

Revised Algorithm

The 1979 Apple II algorithm, which was again used as the basis of the recent reconstruction, uses lists of letter pair frequencies derived from analysis of over 2000 real four-letter words. Three tables were built, containing the frequency of the first, middle and ending letter pairs. A word's letter pairs are looked up in the table and the three numbers found are multiplied together to form a score. Using a randomizing process, a high-low band of acceptable word scores is chosen from a group of similar bands. Thena "trial" word is generated and the word's score is checked against the existing high-low band. If the score falls into the band it is displayed, and if it doesn't, another word is randomly selected. The bands are chosen to favor scores in the middle range of possible scores since more real words have scores in this area. Certain letter pairs with high frequencies do not necessarily combine with other high-frequency pairs, so high-score words are in fact quite rare. For example, in the table below, RE is the most frequent 3rd and 4th letter pair, but in the middle column only AR would match with RE. Because LL is also high in the 3rd column, and matches IL, AL and OL in the second column, words with high scores having ILL, ALL and OLL endings dominate high scores; for this reason the high-low band that includes high scores is intentionally not selected very often. Several weeks of trial runs and adjustments to band-score parameters were done to optimize the word generator.

Words that contain pairs that have zero frequency in English will never appear, since a score of zero is never in any high-low bands. Hence things like NGIK, AQIV, YIIS, or EGBK will never appear.

The three tables below show the most frequent eleven pairs for English four-letter words.