Fermilab’s Strange Letter – Interlude

Hi Slashdot! You guys can be proud to be the first stress test of my new host – and they appear to have passed with flying colours (yes, flying Canadian spelled with a u colours).

That being said, progress on the Fermilab letter hasn’t been as good. I’ve tried an absolutely ridiculous number of things, and I get little. So I thought I’d try to present the letter in a more usable form along with the progress so far and any unresolved questions.

For your reference, my cleaned up data can be found in this CSV file. I figured if Slashdot linked to it, I better provide something multiplatform. Please inform me if you find any errors.

Also, please don’t phone/email/stalk Frank Shoemaker! The poor guy is retired after a distinguished career, has been contacted far too many times about this, and say he has no involvement. Pierre Piroue has also been contacted, and has claimed no knowledge of this whatsoever.

If anyone has talked to CF please email me (sorry for being vague, privacy issues).

Ternary Paragraph

Top Grid View

Here’s the first ternary paragraph. If we decode it as in my previous post, we get “FRANK SHOEMAKER WOULD CALL THIS NOISE”. Note that the grey shaded areas are double spaces – if only the I, II, and III symbols mattered, what are these random extra spaces doing in there? A transcription error seems unlikely. To me, it seems like this was written on graph paper in a specific way, then transcribed to a blank sheet of paper. Effort was taken to ensure the symbols reflected the original alignment – while the whole paragraph may be out of overall “grid” alignment, each individual “tic” is well-oriented in relation to it’s neighbours.

Is it on a grid:

  • for simple boring organization’s sake? If so, why the double spaces?
  • to create some sort of bitmap which maps to the symbols? Alone it seems to provide little help, perhaps it is combined with another section?
  • so that “windows” are cut in the grid according to some specification and it is then laid over another section/combination of sections?

Hexadecimal Section

The hexadecimal section consists of two lines of 12 symbol/letter pairs, seen below.

Note that this transcription uses the same images if a hex/symbol pair repeats, for the sake of lazy html/Photoshop and in an attempt to weed out “noise”. Note this will backfire horribly if the point of the letter is in fact noise – from my preliminary analysis I didn’t see anything too significantly different between duplicated hex/symbol pairs, feel free to correct me.

Note that 1 and A are not included in these 24 pairs. 24 is evenly divisible by 3, I’m not sure if this is relevant, but interesting since the other decodings are based off triplets.

After this is a “signoff” with one symbol we’ve never seen before. I hesistate to call it “undefined” since we are not confident that the hexadecimal digits in fact “define” the symbols.

What are the meanings of the symbol/hex pairs?

  • They belong to three/etc “groups” like in an IQ test, and are used to map hex digits to other digits which will create a new message
  • They are a distinct message by themselves. The hex digits were added later to translate an employee number (see Binary Paragraph) out of the “signoff”.
  • The sixteen hex digits map to musical notes and the symbols mean nothing – Update: this has been attempted, and unless Timbaland produces it and the video involves a lot of nudity, it’s far from a number one hit.
  • The symbols are a convoluted mathematical equation, and the hex digits and signoff allow us to decode it somehow

There’s a million more, but there’s a few to start. If you’ve disproved any/have any new ones, post in comments.

Binary Paragraph

Bottom Grid View

I realize this is impossible to read, but the overall view is what we’re after. Grab the raw data at the top if that’s what you want. Note again that the grey spots are “double spaced” and everything is in a grid, leading to the same questions as before.

If we decode this as described in the previous post, we get “EMPLOYEE NUMBER BASSE SIXTEEN”. The spelling of BASE is off, and could be a reference to the French word for low, although I suspect simple repetition of a triad by accident

If we look at how the message is decoded, this has to be a single “I”, however it appears to be significantly out of place compared to all the other marks. A minor transcription error, or a clue? I think it’s a transcription error – because it’s part of the second S in BASSE. I think he accidentally transcribed S (201) twice, then realized his error at the end when the spacing started to go off.

This leads me to believe that there probably aren’t images stored in the “dashes” in some manner (otherwise he would have fixed the second S, or all the information is contained before this), and the grid was simply to organize or for another purpose.

As well, if we decode it based on simple Morse code (I=dot, II=dash) it reads EUREKA until trailing off to gibberish (credit Henry H in comments). It’s possible that it isn’t gibberish, but since Morse letters are different lengths decoding this becomes a huge pain. My guess is it’s a red herring with no real meaning, but still something to note.


All I can say is I hope this helps someone, and if you figure out anything, let me know! The only thing I think I managed to figure out of note is why it’s “BASSE” sixteen instead of “BASE”. I’m insanely busy this week so I can’t put as much time toward it as I like, perhaps this weekend will be more illuminating…

73 thoughts on “Fermilab’s Strange Letter – Interlude

  1. I no longer think that there is any significance to the structure (aka. grid) that might have been theoretically applied to the first stanza. Have a look at the following image which shows just how out-of-whack the columns are in the first stanza.

    http : / / synhxd.sourceforge.net / images / stanza_one.jpg

    I took the first stanza into my trusty Gimp and colorized it, drawing boundaries for the following:

    1. A light-blue boundary was first drawn around each tick mark
    2. A llight-red boundary was then drawn around tick-mark groups, grouping them into trinary values
    3. A green box was then drawn around groups of three trits
    4. Lastly, the trinary triplet is decoded and the corresponding values are overlayed

    Just to show that the author probably didn’t intend to strictly align the tick marks, I have done my best to align the columns. To do this, I first rotated the entire page 2 degrees CCW. Then, after I had bounded each tick mark, I drew a dotted-line from the center of each tick to the center of closest tick in the next row. You can see that it didn’t look very good afterwards (which leads me to believe that he/she didn’t even use graph paper; the rows are straight, but the columns are WAY off).

    In the picture, you can see some question marks. Some seemingly apparent columns disappear, introduce double-spacing, and cause strange formations.

    Let me know what you think and if it gives any idea. Also, let me know if you want the same thing done to the last stanza.

  2. Devin:

    Wow, great job on the row & column analysis! It does seem to appear the author didn’t use graph paper, which is a major mark against the grid alignment theory.

    I, for one, would be interested in seeing the same applied to the 3rd paragraph. :^)


  3. Finished producing the overlays for the bottom portion:

    @ 203, Pete,

    You may be right, that a grid was still used. Analyzing the last stanza shows that a grid might have been used after-all. The last stanza appears to be a bit more rigid than the first. I also noticed some key areas that have problems such as double-spacing, lack of indentation, and even what appears to be “a period” (directly after the “R” in “NUMBER”).

    By-the-way, you’ll notice that in my analysis of both images, there are “dots” that have been outlined as if they are a stroke. I did not blindly just decide to outline them, rather I ran several Gimp filters on the image to determine what could be a stroke versus specks of dust. The dots, interestingly enough, appear to have direction to them and under certain filters you can see where the pen initially made contact with the paper (on all strokes including the outlined dots). Just thought that I’d mention that.

  4. (apologies for posting this in both threads – it appears this is the more current one and I should have posted it here instead):

    Devin – I’ve looked at the question of if a grid were applied too and have reached a different way of looking at it:


    I rotated the image to approximate a vertical alignment of the left side, then drew a rough series of verticals between each of the top line marks and the lowest “matching” marks.

    Doing so, shows excellent alignment for the top row and the bottom 3, and most of the 2nd row. The 3rd – 5th rows however each appear to show an approximate slip to the left of one space starting roughly where I’ve placed the pink rectangle. Most of the rest of the marks (from left to right) on those rows appear to be approx 1/2 to 1 space out of line then.

    Given how well the top and bottom rows match the grid, I’m suspecting this is an example of poor draftsmanship in transferring the code to the blank page before sending rather than hidden additional information, although if sufficient care were taken then it is conceivable that deviations from the exact spacing may in itself contain deeper coded information. This would seem unlikely given the apparent pen man ship and that only part of the code appears to have “slipped”. My best guess would be that the code was initially created on a grid (perhaps even using a computer to print it) and then traced onto the blank sheet to be sent – but that in the process, the sheet being traced slipped a few times slightly. I further suspect that the tracing wasn’t done systematically from top to bottom/left to right, and that the top and bottom rows were traced first, then the left side, and then the central “block” – and the slip happened during the last block. Perhaps the coder was becoming tired of it, or moved the sheet to check their progress and then failed to align it accurately again.

    Given this level of error in the tracing (if true), I further suspect that the speculation that the dots on the page, line lengths or other details of the marks contain further information will prove fruitless as it would appear that the coder does not posses the necessary degree of accuracy to have conveyed details of that subtly.

  5. Nick:

    Great job! I think you and Devin both have great points.

    You may be right about the penmanship and accuracy not reflecting a deeper hidden meaning behind the grid alignments and other visual artifacts.

    Now we just need to solve that damned middle section!


  6. Pingback: ternary form
  7. Nick,

    Good work on the overlay analysis of the columns. However, I believe that there is a discrepancy in your column positions.

    Let’s take a closer look. In all crypto work, we assume one thing: That there are a set of rules that are applied when decoding the message and that those rules are not broken during the decoding process. If a rule must be broken to reach the decoded message, then either the person that encoded the message did not strictly adhere to the rule or the inferred rule does not exist or the inference is not complete (the rule is slightly different than what you think it is). The cryptologist working to break the message should initially assume that there are no mistakes in the encoding process unless there is good reason to believe that the mistake is plausible given any number of factors (such as source, destination and/or medium that the message was encoded in/on).

    For the time being (without knowing anything about the decoded message or analyzing the encoded message yet), we will make the following assumptions:

    1. There are rules to the decoding process.
    2. We do not know the author, it’s origin nor the destination. Thus we can not plausibly deny nor allow for mistakes. We will [initially] assume that there are no mistakes.
    3. If something is considered to be a mistake (beyond a doubt), then it will invalidate any rule(s) that subjugate that mistake. For example, if we attempt to infer that there is a rule against using two-spaces between a ternary word/group (or T-word), we can deny the existence of such a rule based on the undeniable deductive proven existence of a double-spacing in one or more places.

    These are very basic, and fair, assumptions to make. I make no further assumptions. From here on, I will only rely on inductive or deductive reasoning to reach conclusions and form a sort of “proof” to show that the column positions on my overlay is more accurate (which is important, since the more facts we deduce, the closer we can get to 100% decoding of the paper).

    Without making assumptions, we can safely state that the first stanza contains at least one human-readable message:

    “Frank Shoemaker would call this noise.”

    Our first assumption made above, states that there must be a rule (either previously agreed upon between both constituents of the message or to be discovered by the unfamilial) or set of rules that are followed to reach this decoded message.

    Here’s the set of rules that I garner from analysis (which are never broken during the process of decoding):

    1. Let | equal 1
    2. Let || equal 2
    3. Let ||| equal 0
    4. Let [space] delimit groupings of |, ||, and |||
    5. Let three (3) groupings form a trinary triplet
    6. Let each trinary triplet equal a base-10 number from 0 thru 26
    7. Let values of 0 represent a space in the decoded text
    8. Let values of 1 thru 26 represent the LATIN alphabetic characters A-Z respectively.
    9. If a line begins with a tick-mark (instead of a space; id este “indented”) AND the last column of the previous line above is also a tick-mark, then said tick-marks are to be grouped into a single trit having the value of 1, 2, or 0 as stated by inferred rules 1 thru 3 respectively.

    These are the only rules that I can come up with that are never broken during the decoding process. If we can come up with more iron-clad rules, then we may stumble upon new hidden-messages. For now, we’ll just stick to the affirmed message and the above inferred rules.

    You may notice that I did not state in the above rules that there are 47 columns. That is because we can neither confirm nor deny that the author has introduced double-spaces in some areas. We can deduce that there are 47 columns, but will not state it as a hard-driven fact, especially since the persistence of X number of columns is not required to decode the message.

    Since we NEED to know where there is a space versus a tick-mark at the beginning of the line (to know if it is a continuation from the previous line), we can deduce that the first column is in alignment. Otherwise, if first column were NOT aligned, then we could not arrive at the properly decoded message. Thus, we know that the first column has the following structure:

    Column A, Row 1: Tick-mark
    Column A, Row 2: Space
    Column A, Row 3: Tick-mark
    Column A, Row 4: Space
    Column A, Row 5: Space
    Column A, Row 6: Tick-mark
    Column A, Row 7: Tick-mark
    Column A, Row 8: Tick-mark

    We now know the values of each cell in the far-left column and their role in word boundaries.

    It’s apparent when decoding that you must combine the ending ticks of line 5 with the beginning ticks of line 6. Similarly true for the ending ticks of line 6 and beginning ticks of line 7. This tells us that the tick-marks that appear at the end of lines 5 and 6, in fact, do appear in the _last_ column for that row. If the tick-marks at the end of rows 5 and 6 were NOT in the last column (id este, if there was a space in a column beyond the last tick-mark) then it would break our Rule #9 and thus mangle our decoded message (we would not arrive at “Frank Shoemaker would call this noise.” but rather “Frank Shoemaker would ca[…gibberish…]”).

    Now, we know just about where the last column is. We can verify beyond a doubt where the last column is for rows 5 and 6. For the rest of the rows, we can plausibly show arrive at the following makeup for the last column:

    Last Column, Row 1: Tick-mark
    Last Column, Row 2: Space
    Last Column, Row 3: Tick-mark
    Last Column, Row 4: Tick-mark
    Last Column, Row 5: Tick-mark [verified]
    Last Column, Row 6: Tick-mark [verified]
    Last Column, Row 7: Space
    Last Column, Row 8: Space

    Thus, so far, we have shown without making unreasonable assumptions and via simple deductive reasoning based on some concrete data at least SOME of the basic structure of the first stanza.

    Your overlay doesn’t coincide with this information. In fact, you show a space in the last column for rows 5 and 6, which would either produce a gibberish message if you use the above decoding rules or require that the rules be changed in order to arrive at the proper message.

  8. Hi,

    The folks downstairs haven’t been able to find the original yet — they’re still looking.
    I’ll let you know if we find it.

    – B

  9. Devin – I agree with your comprehensive approach to looking at the code – I just think I failed to explain my point correctly. I approached the analysis from an opposite direction, but reached the same conclusion as yourself that accurate alignment to a grid was not present and that hidden meaning in the details is unlikely.

    In looking at the layout ‘naively’ it appears to follow a regular grid alignment, and certain fits the rules you derived from first principles. However, on closer examination, discrepencies from a perfect grid alignment are apparent which raises a question.

    The question is, are these deviations from a perfect grid signs of a deeper structure/meaning in the marks or simply a sign of human innacuracy in the transription process. I.e. is there a rule (or rules) relating to an exact grid alignment with derived meaning yet to be decerned to add to those you outlined.

    In that context, my first view was that additional meaning was certainly possible – especially based on the decoded message implying that meaning could be hidden in noise. So to examine the likelyhood of this prior to trying to drive it, I decided to examine the layout based on the assumption that a human had transcribed the marks to the page and hence would be prone to human foibles rather than assuming perfection – in that respect, my approach was in almost direct opposition to your assumptions that the message would be perfect until we find evidence againts that. My reasoning for this approach was that the marks had every appearance of human made pen marks on paper and that (in my experience 🙂 ) humans have a limited capacity for accuracy in motor control and hence deviations from perfection were not only possible but likely. Once adopting this approach, I decided to treat it more as an exercise in psychology rather than cryptography – and social analysis can of course often be a useful shortcut tool in code cracking of couse.

    In this approach, I deliberately disregarded any meaning in the message (assuming that the person performing the transcripion would not be likely to be focused on the meaning while they made so many bland marks) and looked at it purely as an exercise in mechanical human trancription of arbitrary marks to a page and to see how closely this had been done to a regular grid. Again – this could be seen as the opposite approach to yourself which works from derived meaning in the coded message to infer confirmation of the alignment of the right column for example.

    I attempted to overlay a perfect grid over the image with exact regular spacing, and quickly found that this didn’t fit (as had already been identified). However, as the scanning/printing process may have introduced a degree of horizontal distortion in addition to any errors in the original, I decided to create a manual grid instead. I further reasoned that the originator may have created a series of columns manually rather that using perfect grid paper, an to therefore mirror this process.

    I tried aligning the marks to a vertical line on the centre of each mark, but the deviation from these was significant. I then reasons that if I were to be attempting to create this grid, I would do so by putting the marks in the gaps on a grid (shading them in in a way similar to optical mark read multiple choice options). Hence, I shifted the grid I had created to put the marks in the spaces rather than the lines, and saw quickly that with the gaps being wider than the lines, the marks appeared to be fitting very well.

    I therefore applied this manually created grid across the image and began to look for patterns in the discrepencies to see where hidden meaning might lie. However, instead I saw that the left side, top 2 lines and bottom lines stayed within the grid very well, and only the middle lines and mostly towards the right side deviated from matching.

    I applied vector lines (very roughly and with pleanty of my own human errors) to get a visual indication of where the discrepencies were greatest. I had to make a large number of assumptions as to if groups were closer to the left or right spaces in many cases, and again initially I suspected there could be a pattern to some marks being moved left or right.

    However, eventually I realised that with the further assumption of paper slippage happening at some point, all the mrks could be seen as ‘reasonably’ having been shifted in the same direction by various amounts, but seemingly consistent with this assumption.

    Thus I derived the image resulting and decided to post my view that this _could_ be evidence that human foibles were responsible for the discrepencies from a perfect grid layout and that this would indicate that further examination of the double spaces etc could be fruitless.

    I certainly accept that this approach is not without risk and relies on a large number of assumptions, but I feel that in the context of what appears to be a humanly produced series of marks, many of these assumptiions are ‘reasonable’ if not likely.

    In addition, I do not assert any additional rules than those you derived – if anything, my approach mearly adds weight to the case that further rules revealing additional information would seem unlikely.

    Now – if only someone could tell me what the middle part decodes to, I’d be able to stop staring at the thing in my spare time! From my social analysis approach, I keep returning to the idea that it’s a quote or title of something by SFC (Frank Shoemaker)…if the hex decodes to text in a similar way to the other 2 sections then it would appear to be a reasonable length for this, and the format with what appears to be SFC centred would fit that too.

  10. Sorry for the many errors in the previous post – that derives from typing with my thumbs on a PDA – I’m only human after all 😉

  11. subsequently looking at this again – I wonder if the block of marks to the right of where I placed the pink block is simply slightly rotated as I noticed the first line includes a double space jut above/left of the pink block – thus indicating the first line could actually be 1 whole space out from that point across to the right. as the largest change would then be at the top of the block, and a counter clockwise rotation of that block would also appear to align the right most column again, this seems plausible.

    Unfortunately I can’t actually examine this on the PDA I’m typing this on so I’m only infering this possibility from attempting to visualise the transformation and estimate the degree to which this would re-align all of the questionablly aligned marks to a regular grid.

    I’ve also been looking again at the last section, and began to wonder if the basse…etc part could have had a simple error which could have then propigated through the remaining parts producing a different text. I became excited when I realised that a slight re-interpretation of the double bar breaks (and allowing the author to have put a duplicate double bar break) could turn this into barco…thinking it may lead to “barcode” etc…but there’s no way I can re-interpret the breaks/numbers to make that actually work…pity though as I’d started to think of all sorts of ways the middle section might be an encoded bar code number! 🙂

  12. (just wondered if anyone attacked/solved the cyphertext posted back above too:

    “I ela ohbqeztvy avr Ivemjusg adzqvpbt akecss kbrs, epw Msemqwit ssgtmc azvkrd ca. I dhfte kscfr cs tpp twahrr ozb vlqbdmo dwym duqnsdf, pht bsmjl wf sbttd h ueoca wx”

    I started trying a simple manual frequency analysis on that tiny snippet but that fell apart quickly – not least because of the small sample size, but I can’t try manually the full blog entry that obvious way or code for it on this PDA…)

  13. hi all. None of you know me, or so I hope. Just a quick thought when I read that jazz/noise post somewhere above…

    We might try to establish if the message has potential meaning thanks to that phrase of

    Let’s start from somethign obvious…
    Frank Shoemaker was a scientist.
    Scientists hold beliefs. Those beliefs mark the direction of their scientific proceedings, but also (far too) often make them blind for contradictory information, especially in elderly age.

    This message may be a reference to F.S. misinterpreting some data during his cereer as noise, or ‘background noise’, or assumption he would should such results be produced. This may be mathematical data, but more likely data obtained by empirical experiments

    Sending a message of this kind to the target (scientists at Fermilab) assumes they have enough background knowledge to decypher it, especially that the first part, raw text, was decyphered without much problems (much problems= the amount of problems breakers are faced in the middle part).
    The coded message might be arranged to fit a standarized form or a chart used at fermilab in certain experiments… or even a precise, existing chart, presented here with every/most/relevant only data on it, including the ID of the person who filled in the documents.

    Dots could mark places on a graph, or results on a chart if matched to the cyphered message.

    If my predctions are correct, there should be a day/month/year written somewhere… perhaps even precise time. This may give background to decyphering numericals in the message, which, indeed, could be meaningless unless set against a proper chart…

    And if all this is false, then there must be any other way in which this message should be important to a person sharing F.S. field of interests, but NOT sharing his scientific beliefs.

    Otherwise it just as well migh be an ASCII goatse or a raging lunatic getting freaked over strange matter in backwards klingon haiku.

    All typos are [sic]

    Enjoy my cancer.

  14. In this “random-sounding” msg by “Shonaili” there seems to be a “moving” solution. Ie. “Rqnpsyr Segyusu” might be Richard Feynman mentioned in the link just before it, so one can’t use direct mapping á la X -> A and so on, but the key changes. Like in first letter A -> B, next one A -> C etc. Also “Msemqwit ssgtmc” seems to refer “Fermilab letter”, since the links point there, but you can’t replace the letters directly.

  15. Hi all. I only found out about this last night but I’ve read everything I could on it and tried a few things. Looking at it overall, there are some conflicting clues. The fact that the first stanza is quite narrow on the page, and that some symbols are split across two lines when they could so easily finish on one suggests that the shape and/or layout of the grid is highly significant and contains additional encoded information. Then again the sloppy alignment of the grid suggests the layout is not important.

    I tried looking for more character encodings in the first and last stanzas running vertically in columns, in the coder’s trinary encoding and for other numeric encodings. There’s nothing obvious. And frankly, it’s *really hard* to encode multiple messages in one piece like that. The spurious S in BASSE is also a dubious clue because it could be vital for lining up another message within the grids, or it could be an amateur’s mistake. The way the first two tick marks of the second S are mis-spaced casts doubt on the coder’s concentration at that point, but again, it could just be a requirement for alignment or another obscure encoding, designed to look trivial.

    I agree with EV in comment #42 that the last stanza’s positioning casts some doubt on how integral it is to understanding the message. Even the sFC thing looks a little like an after-thought.

    But regardless, I’m fairly sure the middle stanza contains the most absolutely important part of the message. There’s a whole 24 hexadecimal digits worth of information there that hasn’t been deciphered at all…..

  16. Since I haven’t seen this information posted anywhere, I thought I’d help out. Frank Shoemaker’s Fermilab ID number is 1586V. The “V” is code for “visitor”. Regular employees have an “N” (which is normally left out by default) at the end of their ID number, and contractors have a “C”.

    I see who “C.F.” is (ID 2812V), but I will leave their name out of this due to the privacy concerns mentioned above.

    As for Jason’s post above, employee 1569V is H.C., if that means anything to anyone. (Again, I don’t want to use names due to privacy concerns.)

    There are no regular employees with these numbers (e.g. there is no 1586N). There are some contractors with duplicate numbers, but I don’t think they have anything to do with this puzzle.

    Good luck!

  17. I wouldn’t want to rain on anyone’s parade here, but I’m sure that if this was done by a professional,
    they would proof-read. I wouldn’t be so quick to call “basse” a spelling error. Pretty big assumption there.
    Bye for now.

  18. An article was just released on slashdot.org (/.) today mentioning a breakthrough in quantum computing, showing equations to calculate the capacity of quantum channels in respect to the 1948 theory of communication as published by Claude Shannon while he worked at Bell Labs.

    Here’s the abstract: http://arxiv.org/pdf/0807.4935v1

    I’m re-visiting the idea that the middle stanza is a mathematical equation that is significant. Most likely an explicit formula that answers an age-old question. That this entire paper is a “teaser” for a soon-to-be released miraculous scientific find.

  19. Has anyone considered the idea that base sixteen may not refer to a “base” as a location but to a mathematical definition of base? Of course, no other clues to tell us how to manipulate the number 16. Logs or what…hm.

Leave a Reply

Your email address will not be published. Required fields are marked *