Britain on Hash

# part 2 

The "hash"?

In part one, we looked at the history of the number sign in the last few hundred years.  A brief overview of some very rough dates and usages (most of which probably happened a little or a lot earlier than what I've found):

1500s - musical sharp
1600s - proofreading symbol for space
1820s - Tic-Tac-Toe
1850s - number sign
1850s - pound sign
1960s - octothorpe.

But I left out the hash (or hash sign or even possibly hash mark) as a name and a usage.  This name is by far the most dominant name in England for #.  And after all the British created the English language, so they should know, right?

In a blog posting at oxforddictionaries.com about the name of the symbol, they claim that "the most common is probably hash."  I'm not sure who's posterior this "fact" was removed from, but lets hope it wasn't too painful.  It's certainly not the most common name in American English.  Perhaps they didn't realize the Internet extends outside of Britain.

Of course I'll readily accept that the British could never readily accept # as pound sign for weight, because that's already taken with their monetary unit, pounds sterling (£).  We'll just ignore the fact that the monetary unit comes from pounds of weight, which even the British have written as the symbol ℔ (which turned into the #) and which they'll read as "pounds".  Whatever.

Perhaps the real question is, what is the pedigree for "hash"?  Maybe the Brits really are right, maybe this is the true and genuine real name for this ancient symbol.

So I looked.

I've searched around in the 19th century, where I knew that "pound" and "number" were both well-established.  They certainly didn't use the word as a name for that symbol back then.  "Hash" was however a slang term for food, usually bad mass-produced food, sometimes a specific type of dish of meat and potatoes and other things all mixed together (i.e. bad mass-produced food).

In the 1910s and 1920s I find references for "hash mark" that referred to the military (or civil) navy marks on the sleeve used to indicate rank or experience.  We still call them that today.

Then, there it was, a reference to "hash mark" in 1920 that seemed like some other meaning entirely.  Granted it was still in a naval context.  But this usage was new.  Perhaps I was on the right track?

From The Log of the Columbia Naval Unit, 1920 (via googlebooks).

There it is!  A "hash mark" is also a synonym for a food stain.  That settles it.

At some point in the middle of the 20th century, Americans began referring to some of the markings on a football field as "hash marks".  That is to say, American football, not real (cough) football.  So yeah, we're not even going to explore this one as a possible origin.

In the 1960s I see hash tables appear in computer discussions.  And hashing algorithms.  This is a method of compactly storing and organizing random data in a semi-sorted and rapidly accessible form.  But as far as I can tell (and I'm a software engineer) there's never been any association between this method and the number sign.

Also in the 1960s, some old manuscripts from Mark Twain are published, including a story called "Three Thousand Years Among the Microbes", originally written sometime between 1896 and 1905.  In this story, a new monetary system is recommended, consisting of tenths of pennies called bashes.  A thousand bashes is one hash.  Of course, in the printing I've seen of this, no number sign is apparent.  Plus, Twain is American, so... no help here.

It isn't until the 1970s that I see the word "hash" used to refer to #.  A couple of things happened around this time.  One is that touch tone phones became popular, and incorporated the same keypad as in America, with this new and alien (to British) symbol.  The other is that "hash" became a popular nickname for hashish (which by the way is the origin of the word assassin).

Now don't get me wrong.  I'm not suggesting that these two new uses of the word "hash" that happened at the same time had anything to do with each other.  I'm just presenting the facts here.  But the mere possible suggestion is interesting, because of the vacuum it fills.  That vacuum is the void that is not filled by a true origin story for calling # a hash.

Several online sources claim that hash came from "hatch", as in the cross hatching artists use when drawing.  Except, the only place I find this is in these claims.  They cite no additional sources, and I have found no additional sources.  I've only seen this as a just-so story presented to fill that vacuum.

Honestly, if we're just going to make stuff up, I prefer the hashish theory.

OK, all the jest aside, I did actually find SOMETHING.  A google search in books for "hash sign" brought me to this:

Can you see that here?  I can't get to the full content.  It looks like two different printings of the same (or a similar) article.  One uses a number sign, the other uses something new:

From Research Review. Navorsingsoorsig, 1968-1971 as found in a Google search.

The source of this is from some sort of review of papers of that day.  According to Google it's a publication of the South African Council for Scientific and Industrial Research.  It seems to be a description of a Keyword-in-context (KWIC) index, or of some sort of method for marking up or automatically creating a KWIC index.

I have found many other references to KWIC software systems in that era, and I haven't found any others yet that mention a hash sign, or a number or pound sign.  Several mention a dash, and also mention it together with a slash.  "Slash-dashing" was used as a verb in a few.

What does all this mean?  I don't know yet.  This could be a printing error.  Or a symbol I'm not familiar with being imported in, and later substituted with a number sign.  It could be some combination of slash and dash, which ends up looking like an "H".  (Thus "hash"?  Oh man, do NOT quote me on that one, that is how just-so stories are created.)  But if we are making up just-so stories, it does strike me that this new symbol along with the original number sign are highly reminiscent of hash marks on the sleeve of a naval officer.

Until more information comes in, I'm convinced the British made something up off the top of their heads, to explain an alien symbol that appeared on their phones.  They couldn't call it a pound sign, because that was taken.  They couldn't call it a number sign, what with all the other phone buttons having actual numbers on them.  So they just hashed it out I guess.

So where does this leave us?

I set out to find the pedigree of "hash".  It has none.  It's a word with no long history.  In fact, at the moment it has no real history at all, since nobody knows where it came from.  Perhaps more to the point, it seems to be completely meaningless.

Compare this to "number sign", which dates back at least 160 years, and probably quite a bit more.  It's still widely used and recognized as a number sign today.  It means "number" in this usage.  So we have both history and semantics on the side of "number sign" (and of "pound sign" for that matter).

We can't defer all decisions to the British simply because that's where the language started (although not really).  And of course we have to acknowledge that language changes.  This is an inevitable process.  But that doesn't mean we can't have opinions on how it might best change or not change.  The ability to continue to understand past writers is a great one, worth preserving.  Meaningless and pointless changes to the language don't provide any benefit to anyone, and it doesn't matter if it's the British introducing these pointless changes.  If we are going to take the time to think about our language, we might as well encourage meaningful communication.

One thing is certain: we can no longer entrust the English language to the English.

The Sign of the Number

#

What is it?  How old is it?  Where did it come from?  And what should we call it?

The first time anything like this symbol appears in our history (as we've uncovered it so far) is one of the oldest known abstract marks made by any humanoid, in this case Neanderthals, about 39,000 years ago.

Neanderthal social media? (via Scientific American blog)

But that's a bit outside the focus of my blog, which is usually related to printing technology.  And so that's where we'll pick up this story.

Sharp

The sharp is probably the oldest use of this symbol, although it isn't very clear when it really began to look like #.  Originally the sharp symbol (as well as the natural) derived from the same letter "b" that gave us the flat symbol in music (♭).  These symbols developed because in early attempts to record music (around the 12th century or earlier), the note B was the first (and for a long time the only) note that seemed to require modification.  To distinguish between the different versions of B they would add a notation using "b rotundum" (the round b), "b quadratum" (square b) and "b cancellatum" (barred b).

I haven't given this a complete study, so I don't know exactly when and how these morphed into our modern flat, sharp, and natural symbols.  But around 450 years ago, notation began to use multi-lined staffs to write the notes, often five line staffs but sometimes more.  In 1575, Queen Elizabeth granted a printing monopoly to Thomas Tallis and William Byrd as exclusive printers of music.  Movable type was applied to the printing of music, and I believe this helped to standardize developing notations.  Their early published pieces included flats which are easily recognizable but the sharps looked more like a double X (and nothing at all like any form of "b").

From Cantiones Sacrae, 1575, Thomas Tallis and William Byrd (via hathitrust.org).

Some forms even looked a bit like a flower.

From A Compendium of Practical Music in Five Parts, 1667, Christopher Simpson (via googlebooks)

Soon the sharp began to look more familar.

Sample of sharps and flats used in text rather than music, from A Preliminary Discourse to a Scheme Demonstrating the Perfection and Harmony of Sounds, 1726, William Jackson (via googlebooks)

Of course in modern music, the sharp looks noticeably different from the number sign, and in our digital world they're two separate characters, which also differ in appearance.  Here they are in this font:

Number:        #

    Sharp:       ♯.

Proofing

There's one other very old form of # which could be even older than the sharp sign (at least in recognizable form).  In the printing process, after a page of text and figures has been "locked up" in a "form", a proof is printed from the form and checked for errors, after which the form can be corrected.  There's a standard set of notations for marking these errors for correction (because the corrector is usually not the same person as the compositor).  The # symbol is used in the margin to mark lines where a space is missing (and the spot on the line is marked with a caret or slash). This same notation is still in use today.  The oldest reference I know of on printing, Joseph Moxon's 1683 text, also describes this notation:

From an 1896 reprint of Moxon's Mechanick Exercises or the Doctrine of Handy-Works Applied to the Art of Printing, Vol. 2, 1693 (via hathitrust.org).

Of course, Moxon was reporting it, not inventing it, so the usage must be quite a bit older than 1683.

In addition, volume 1 of this reference includes a standard layout for the compositor's cases.  The # is used here to mark the sort that other sources say is for spaces or large spaces (bottom row).

From an 1896 reprint of Moxon's Mechanick Exercises or the Doctrine of Handy-Works Applied to the Art of Printing, Vol. 1, 1693 (via hathitrust.org).

You might think, as I first thought, that this sort was where the number signs were kept, but this isn't so—other sources clearly mark it for spaces.  This leads to a question though, where in the case was the number sign?  It simply didn't exist as a standard printable character.  Perhaps you noticed that in my 1726 image above, the sharps appear to be hand-drawn (or more likely hand-carved characters).  Sharps were standard for music printers, but other than that, nothing like a number sign existed in the standard fonts.  Of course, many fonts had "specials", and much later on it it became  more common for specials to include a number sign, although the exact details are difficult to track down.

Tic-Tac-Toe

Another very old usage for this symbol is as a Tic-Tac-Toe board.  In printed descriptions, this is the third oldest form (based on what I've found so far), with a clear description of the game from 1823 under the name "Kit-Cat-Cannio" (Suffolk Words and Phrases; although they do not reproduce the game board).   An 1852 reference shows a game board, as shown below.  For reasons that belong in another post, I tend to suspect the game as we know it (and more to the point of this article, the game board) developed during the 1800s or at most the late 1700s.

From The Book of Children's Games, Constance Wakeford Long, 1852 (via googlebooks).

Number

In 1853 I find this symbol described in a reference as an abbreviation for "number".  This is the earliest description of this usage I've found so far.  However three years earlier (1850) I've found the symbol used in account ledgers as a pound sign.  Which usage came first?  Safe bets at this point are that the printed description of number in 1853 trumps pound usage "in the wild" in 1850, because to be described the number usage had to be quite common.  And the earliest description of the pound usage I've found is thirty years later.  Still, I've barely scratched the surface in references so who knows what might turn up next?  At any rate, my assumption now is that usage as "number" is older than "pound".

From An Elementary Treatise on Book-keeping by Single and Double Entry,  1853 (via googlebooks).

Still, this usage remains quite perplexing.  In spite of the above description, I have so far not found any actual usage of it as such prior to the introduction of the typewriter.  Many of the sources that mention this usage do so in the context of business (or commerce or retail or something similar).  And yet I have still seen no sign of this symbol on old ledgers or inventories or anything of the sort.  I don't doubt that examples exist, I just haven't found them yet.

It's also a bit odd that there's no particularly compelling theories on how the number usage originated, even though we have tons of (bad) theories on the pound usage.  In print "number" is usually written out, but if it is abbreviated, it is either done as "No.", or as a single character that was apparently not uncommon to printers: №, as in "№ 2".  It's not clear if there's any relationship between the № and # characters, but there is a clear street use versus print dichotomy which I'll cover in more detail in the next section.

Pound

There's a number of theories out there on how this came to be used as a pound sign that all seem to center on confusion between the apparently American number sign and the British pounds sterling sign (£), relating to other technology like the telephone, the typewriter, the Baudot code (the first digital encoding scheme for the telegraph), or other early telegraph codes.  The Baudot code theory is by far the most common.

The telegraph system was invented in the late 1830s, more or less simultaneously in the United States by Samuel F. Morse and in England by Charles Wheatstone and William Cooke.  By the 1850s this "Victorian Internet" was changing the world.  And then in the 1870s, a Frenchman named Émile Baudot began tinkering with automatic decoding and printing of telegraph signals.  To do this, he invented a totally new code, a five digit binary code.  The claim is that American and British versions of this code used the number sign and pounds (sterling) sign in the same place.  There's a number of problems with this, a big one being that the Baudot code itself was almost never used in the United States.  Another big problem is that (despite an unsourced claim in Wikipedia) I can find no evidence that such a code collision ever occurred.  At least not exactly.  French and British codes overlapped the numero (№) and pound sterling (£) on the same code (shift-N).  However it isn't clear if # ever had dual meaning anywhere in Europe, or if it was an entirely American symbol.  Even if codes had overlapped, an obvious issue is that American usage of 12# (weight) and British usage of £12 (currency) are easily distinguished both based on order and context.

But of course the biggest problem is the timeline.  The # symbol already had dual meanings by 1853, and almost certainly much earlier.  This precludes any technology interference except possibly the original telegraph codes.  But again, I find no evidence of collisions (and in fact, no evidence that the number sign had any encoding at all) in any of the common telegraph codes of that era.

This all makes it seem more likely that if there was any confusion or overlap, it went in the opposite direction—documentation of a variety of different national code sets were confused by the pre-existing double meaning of #.  Or perhaps, if # was used or known in Europe, the double meaning even encouraged the use of the same code slot for these meanings.

The most popular theory and almost certainly the right one (in my opinion) is that the pounds meaning derived from the common usage of lb as an abbreviation for pounds, or more specifically for the latin "libra pondo".  This abbreviation of "lb" or "lbs" was common both then and now, but in addition to the abbreviation, it was also used as a symbol, by crossing a line through the letters like this: ℔.  (Some sources distinguish ℔ and lb. as pounds Troy and pounds avoirdupois respectively, however I haven't seen enough consistency across the long span of these usages to say that this is an accurate or useful distinction).  This notation goes back at least to Isaac Newton's time, as noted in online sources.  Here's an enlarged version of one of his ℔ symbols, together with a sample showing his use in context.

Isaac Newton's pound symbol, 1692 (via chemheritage.org).

As you can see, in its usual cursive form, it already looks a bit like a pound sign.  Note how he has used ℔ as a superscript.  I've found this it be a highly consistent usage in handwriting from his time through the 19th century, and one that helps convince me of the # usage when it turns up.  It's quite easy to find ℔ in old handwritten receipts and ledgers, as well as "lbs."  However use of an actual # is not that common.  I first found an 1887 receipt that includes 302 pounds of cheese (shown as 312-10=302) sold at 12 cents per pound for $36.24:

Unmistakable pound sign, 1887 receipt (via 19th Century Wellington blog).

Later I found this example from 1850.  It includes two pound signs, although the second one is smudged.  It's from a ledger in Hustonville Kentucky from 1850 and includes a line with 7-1/4 pounds of nails, then later 6 pounds of nails.

Two pound signs (one clear, one smudged), 1850 (via Kentucky Digital Library).

Early on it's even harder to find it described in a text than it is to see it in use.  The first mention in print I have found so far is this 1880 book, where its dual usage as both number and pound is mentioned:

From Book-Keeping by Single and Double Entry, 1880 (via archive.org).

This mention comes thirty years after I see it "in the wild".  Of course it has to be fairly common to be mentioned in a book at all.  My assumption is that there was common usage among the people, in markets and in handwriting, where the symbol was long used to mean both number and pound.  And then there was the printing business, dominated primarily by academic concerns, and where these notations would be written out fully, or at least abbreviated in the more traditional printers' way as ℔ and №.  These two worlds rarely collided which is why it was so long before this usage was acknowledged.  This is supported by the fact that the character used in the above sample seems to be a sharp symbol, not a pound or number sign—they couldn't actually print the literal character in common usage on the streets, because even by 1880 it was still very unusual in typography.

Two technologies seem to confirm this notion: the Linotype, and the typewriter.  The Linotype was an extension of printing technologies.  The typewriter on the other hand was a replacement for people's handwriting.  The first typewriter in 1873 had no number sign (and no lower-case letters for that matter), but by 1886 at the latest, there it was, above the three, right where it is today.  Contrast this with the first Linotype that came out in 1886.  It had a rack of 72 matrices that could be used to mold characters.  None of them was a number sign.  But then, by 1906, Linotype had expanded to a 90-character rack of matrices.  But still no number sign.  It did however have the ℔ symbol.  Even by World War II, there was no standard number sign included in most Linotype fonts.  (And for the record, they existed, but were extras that had to be added in by hand, rather than from the Linotype keyboard.)

Diagram of Remington typewriter keys, from catalog ca. 1887 (via Harvard Library).

There's another clue hidden here that's related to the notion of printing versus handwriting.  If you go back and look at the full manuscript I linked to for that Isaac Newton reference, he seems to use a form of the ampersand that many still use today—a sort of a cursive plus sign.  It's a perfect example of common usage that we can all recognize, and yet it is very rarely mentioned in most printed documents.  This is possibly for the simple reason that its representation in typography is very rare.  In fact, I'm not even sure that version of the ampersand exists in Unicode.  If that form of ampersand can exist for so long without significant discussion, it gives me hope that there are some very old number signs and pound signs out there somewhere, represented as #.

Official name?

So you may have noticed that I have called this symbol a number sign a few times in this post as a more generic name.  Today, it's the most common usage (outside of social media of course).  It's seldom used to mean pounds anymore.  In my opinion, because of the widespread use of the symbol in this context, together with a long history as the same, "number sign" is our best bet as a name for this symbol.

But you may be wondering about a couple of other common names.  One of them, the octothorpe, has at least two origin stories, both linked to different sets of employees at Bell Labs around 1963.  One of them is almost certainly close to the truth.  In 2014, one of the claimants wrote a pretty detailed piece on the history.  But even the authors didn't seem to want it to catch on, and for the most part it hasn't.  Despite a few sources that call it the "official" name, this seems to have been little more than a joke.

Which brings me to the other common name, the "hash", or "hash mark" or "hash sign".  I think that one deserves a posting all its own, so you'll have to wait to see how I demonstrate that the British are ruining the English language in part two.

Revenge of Telescope Identification

One of the pitfalls of being a dabbler is being optimistic about your research.  Or maybe that's one of the pitfalls of being a member of Homo sapiens.  At any rate, in my rush to turn a few hours of useless Google Books research into something fleetingly useful, I quickly summarized what I'd found about that telescope in my previous posting.

I chose the 26-inch Naval Observatory refractor as the likely model for the boys' drawing.  But thanks to twitterer @jackdoerner I now know that this telescope had a twin.  The McCormick telescope at the University of Virginia.  They had been ordered at about the same time, but financial issues with both Leader McCormick and UVa meant that their telescope wasn't delivered and installed for another fifteen years.  Ultimately, UVa had to rely on philanthropists to complete the funding of their project.  Because of this delay the McCormick was never the biggest in the world, and was far less famous.

The Naval Observatory 26-inch on the left.
It's twin, University of Virginia's McCormick 26-inch on the right.

This story of twin telescopes and delays shouldn't have been at all surprising to me.

The telescope I'm most familiar with, Harvard's Great Refractor, also had a twin.  And like these two 26 inch telescopes, the Harvard 15 inch telescope and its twin at the Pulkovo observatory near St. Petersburg Russia, were the largest in the world in their day too.  Like the 26 inchers, there was a winner (Pulkovo), and an also-ran (Harvard).  While Harvard wasn't ever really in contention to be the first, financial reasons nevertheless added delays in getting Harvard's 15 inch telescope pointed towards the sky.  Harvard actually raised funds from astronomy enthusiasts to pay for their telescope.  Donors got their names on a plaque that is still there today.  (And they did it without kickstarter.)

It's almost like the 26 inch twins were twins of the 15 inch twins.

One of my biggest mistakes though was to assume that each telescope of this size was unique.  I had believed that even our own Great Refractor at Harvard was unique, and had (for reasons I can't now pin down) assumed that Merz and Mahler had only made the lenses, and the telescopes at Harvard and Pulkovo were completely different.  This assumption was completely wrong.  Merz and Mahler made both 15 inch telescopes, and they look nearly identical.  And Alvan Clark and Sons made both of the twin 26 inch telescopes, including the mounting.

This also shouldn't have been surprising, given that the telescope I get to actually use is an Alvan Clark nine-inch refractor made in 1912. (Sadly, the 15-inch Great Refractor at Harvard is non-functioning, and there are no funds available to restore it.)  And I was perfectly aware of the fact that Clark made the entire telescope, as they did with countless others.

Incidentally, Avlan Clark and Sons' ascendancy as the preeminent telescope makers of their day was very rapid.  in the mid 1840s Alvan and his sons began making a few basic reflector and refractor telescopes, more or less as a hobby at first, while Alvan continued his career as a painter and engraver.  After a few small successes, Alvan Clark Sr. had an opportunity to look through none other than Harvard's Great Refractor, shortly after it was installed.  Based on what he had learned to that point, he saw minor problems with the telescope and decided he could do a better job.  Plus, he found out how much these lenses cost —the Harvard lens was $12,000 in 1844, which would be about $300,000 today).  He turned his hobby into a business, and in just over ten years was given the opportunity to build the largest telescope in the world, an 18.5 inch telescope commissioned by the University of Mississippi (but eventually sold to the University of Chicago in 1862, as the Civil War made the original deal impossible).  His firm continued to dominate the field right up to the end of the refractor era, making four of the final six record-breakers, including the last two at 36 and 40 inches.

So the two 26 inch telescopes are very much twins.  Even the distinctive mount, which was also designed and built by George Basset Clark.  This is particularly notable given the fact that telescopes in different locations might need different mounting hardware.  The angle of the mount has to parallel the axis of the Earth, and if latitudes are very different, different hardware might be needed.  But in this case, both telescopes were designed for nearly the same latitude (less than one degree apart), which makes it almost trivial for them to share the same mounting hardware.

So which one is the inspiration?  Or was it another telescope entirely?  Again, I have no firm answers.  But this is where it gets really interesting.  @jackdoerner also pointed out that the Viriginia telescope was operational in Cambridge (Massachusetts) prior to its installation.  When I assumed that Clark only made the lens, this would have been unimaginable.  But the evidence is there to confirm what twitter was telling me.  In 1877, the Naval Observatory discovered the moons of Mars, Deimos and Phobos, using their 26-inch telescope.  They quickly got word out to other important sites to confirm their finding.  And one of the sites they contacted was the Cambridge telescope factory, where the still undelivered McCormick 26-inch telescope was used to confirm the moons.

This is notable, because in the last article, I wondered if the boys had ever visited the 26-inch in Washington D.C.  But this new information means that its twin was only 100 miles away instead of 500 miles.  Did they ever visit Alvan Clark and Sons' factory?  Their ages probably make this impossible, but I can't say for sure because I don't know their ages.  The books were written roughly from 1892 to 1895.  The telescope was delivered to Virginia in 1885.  If the boys were teens when they wrote the books, then they might have been too young to visit the telescope in Cambridge and draw it with so much accuracy.

I'm also still guessing that Alvan Clark and Sons didn't provide the pier itself (the brick structure).  But their clock drive design would likely have mandated some sort of a pier where the mechanism could be inside the pier, and would need to be accessed.  So while the piers are clearly not identical, they are clearly similar, and either could have been the inspiration for that drawing.

It still is most likely one of these two telescopes.  Alvan Clark and Sons did make smaller telescopes that used a similar mount, but I haven't found any where the counterweights looked anything like these two telescopes.  It's unlikely the boys saw the McCormick telescope in Cambridge because of the timing.  And it's even more unlikely they'd ever see a picture of the McCormick—I haven't found a single image from the 19th century yet.  I'm sticking with one of these being correct, and the Naval telescope remains the strong favorite.

Unless of course the Internet tells me otherwise.

Telescope Identification

(I seem to be going off-topic more and more, but any research I end up doing on history is basically an off-shoot (or sidetrack or zig-zag) of my typography research, so it might as well go here.)

Slate published a great story about some brothers from the late 19th century in New Hampshire who wrote a series of books set in a fictional universe (they had no TV or Internet back then so they were forced to do something creative I suppose).  Someone discovered their books years later and there is now an online archive at Amherst College.  At any rate, one of the things they created for their fictional universe was this telescope.

Nelson brothers drawing, likely from early 1890s.

"The Great Monarch telescope at Picnic City B.C. [a town in their fictional world] is mounted already for shipment to N.P. to be set up under a solid steel roof on top of the great tower at Allenstown the great foundation upon which the telescope rests is made of the best B.C. Granite and Marble"

As it happens I work at the Harvard College Observatory (for the Smithsonian, not Harvard), where we have the Harvard Great Refractor from 1847, the largest(-ish) telescope in its day (bad timing between lens grinding and telescope construction meant we didn't quite get to be the largest) at 15 inches diameter (it's about 19 feet long I believe; all sizes given hereafter will be diameter in inches so assume the length is a similar size in feet).  I have seen this telescope many times and even acted as tour guide in a pinch.  I also have the privilege of using a 1912 Clark nine-inch telescope here from time to time.  So I knew from my exposure to these telescopes that this imaginary telescope seemed very detailed and realistic, and wondered what their inspiration was.  Even their text was very accurate, the appropriate use of the word "Great", the description of the granite and marble tower on which to mount the telescope, and (to some degree) the steel dome all rang true.  So I set out to figure out if there was any telescope back then that looked just like this.

After looking at dozens of telescopes, and pulling my hair out a bit, the answer is that the inspiration is almost certainly the U.S. Naval Observatory's 26-inch telescope, installed in 1873, and itself a record holder for the largest telescope in the world for about seven years.

U.S. Naval Observatory 26-inch Equatorial Refractor.

The brick pier (the permanent "stand" for the telescope), is somewhat distinctive, as most top quality telescopes used either solid marble or steel piers.  The ladder on the side is quite distinctive too, as others have stairs or spiral stairs or nothing fixed there at all.  The shape of the mounting bracket on top of the pier is also very similar.  Even the wheels that the dome sits on are somewhat distinctive in their spacing (and in being there at all, and visible).  But the real clincher is the split arched door panels that access a space in the pier.  I haven't found any other telescope anywhere with this feature. I can only assume that space was for the clockwork mechanism used to keep the telescope rotating in synch with the earth so the stars wouldn't drift out of view.  Such mechanisms were completely standard equipment, and sometimes inside the pier, other times on the outside.

It was a bit harder than it should have been to find this telescope, because in 1896 the Naval Observatory moved the telescope to a more remote location away from the foggy river, and replaced this pier with one that was inspired by the record-setting 36 inch James Lick telescope from 1888.  That telescope featured a movable floor that could be raised or lowered to match the height of the bottom of the telescope regardless of the angle.  The Navy's new dome copied this movable floor, and the new pier looks much like the Lick telescope's pier.

Most images of the Navy 26-inch show the new pier from 1896, very different from the original.
Image from Scientific American,  July 1896.

It is notable that the new location was commissioned in 1891 and a "steel" dome was ordered to house the telescope, although this may have only meant a steel frame, which would be more typical.  Depending on when they wrote this description, the Nelson brothers may have heard of these plans to move to a new steel dome, and that may have been part of their inspiration.  (By the way, the Lick Telescope was almost the end of the line for refractors.  The Yerkes telescope is the only one to beat it, in 1897, at 40 inches, a record that stands to this day.  The Yerkes also featured the same movable floor design, and the pier looks like a brother of the Lick and Naval telescope piers.  The difficulty of making lenses this large, together with improvements in stabilizing large mirror designs, meant the astronomy community switched to reflector telescopes for their ever larger designs.  The current record for reflectors is now 10.4 meters, or for comparison purposes, 409 inches in diameter.  Reflectors also can be extremely short relative to their diameter, hardly longer than they are wide, whereas a 19th century refractor this diameter would have to be 300-500 feet long.)

I don't know if this telescope was mounted on a tower.  But the Naval Observatory's best telescope from 1844 was mounted as shown below.  Likewise, the Great Refractor here at Harvard is similarly mounted on a tower built inside a building (although it isn't as tall, and from the outside you'd never think of calling it a tower).  The tower provides a stable base all the way to the bedrock, to prevent telescope movement.  It also provides additional height to make sure the telescope clears any nearby trees or other obstructions.

"Tower" used to support the Navy's best telescope in 1944.  Harper's New Monthly Magazine, 1874.

With the identification made, I wondered what their inspiration might be.  It is possible that they were able to visit the telescope in Washington D.C. but I don't know how much, or even if, they travelled.  I didn't find any satisfactory answer, but I did find several different drawings of this telescope, which is basically why I'm writing this blog.  So, preamble over, here are some popular depictions I found of just this one telescope.

I should first note that I can only tell you the date where I found these images.  This is not necessarily the date when the image was created, as later publications often borrowed or stole images from earlier publications.  This image from "American Observatories" in Frank Leslie's Popular Monthly, volume 30 (1890) is clearly a reproduction of the photograph near the top of this article, which I don't have a date on, but I believe was taken in 1873 or 1874.

From Frank Leslie's Sunday Magazine, 1888

This image, from a different Frank Leslie publication, Frank Leslie's Sunday Magazine from 1888, is frankly horrible.  The scale is off, and the details don't quite match the real telescope.  I can only wonder if it was reproduced from older pre-construction drawings of the telescope.  The Nelson brothers' drawing, while less tidy, is much more accurate than this one so I tend to believe this was not their inspiration.  On the other hand, it is from the same angle, and the telescope position on the mount is the same.

From Johnson's New Universal Cyclopaedia, 1880

This image from Johnson's New Universal Cyclopaedia: A Scientific and Populary Treasury of Useful Knowledge in 1880 is a better depiction of the telescope.  It also matches the telescope position and point of view the Nelson Brothers drew, but I think it differs too much in details to really be their inspiration.

From Harper's New Monthly Magazine, 1874

This image from an article called "Observatories in the United States" from Harper's New Monthly Magazine, volume 48, May 1874 is also very similar to the photograph I have.  Not quite as similar as the previous drawing found in 1890, but similar enough that I still think it is likely linked to the photograph, and it is the reason why I think the photo dates back this far.

It's hard to judge exactly where they were exposed to information on this telescope.  It's possible they had multiple sources available.  It's entirely likely that I haven't found all the images that were floating around in their day.  It is clear though that there were plenty of popular depictions available.  This is possibly related to one of my favorite topics, the Linotype.  As much as I love to talk about its typographic shortcomings, the revolution brought about by this machine made newspapers four times longer, and made cheap magazines and books available to everyone.  From what I've read, the Nelson brothers did their work within the first decade of the Linotype's existence.  Was this too soon after the invention to be related?  Or did our culture change that quickly?

One other interesting thing I noticed in all of this is that the illustrators never seem to be credited.  Some of their depictions are awful, but others are amazingly detailed and beautiful.  There's a fantastic drawing of the Lick Telescope that puts any modern steampunk artwork to shame, and which I finally managed to find a high-resoution scan.  It's signed "E. Steinmetz", and yet I can find no sign anywhere of who this person is.  I'd love to see more of his or her work.

In the meantime, look for future blogs on QWERTY keyboard layout, and more on the Tic-Tac-Toe machine.  And someday I'll get back to the subject of sentence spacing.

The Mystery of the Victorian Tic-Tac-Toe Computer

"The Turk", 1769 chess machine.

In 1769, Woffgang de Kempelen debuted a mechanical device known as "The Turk".   The Turk was a chess-playing mechanical computer.  It consisted of a large console with gears inside, and a chessboard on top.  Seated at the console was a robot, with a mechanical, human-like, arm that would pick up real chess pieces on a real board, and move them to a new square.  The Turk won the vast majority of games it ever played.  For almost seventy years, until 1837, The Turk dazzled crowds with it's amazing (and secret) engineering.

Until it was revealed to be a massive hoax.

There was a master chess player hiding inside the desk (he could move out of the way while the cabinet was opened and shown to be empty to the audience).  The Turk was a puppet, controlled by a device similar to a pantograph.  It was only because one of the puppeteer chess masters eventually spilled the beans that the hoax came to an end.  Of course, a machine that played chess was an impossibility in that era.  Many at that time believed it would never be possible.

Charles Babbage was probably not one of those skeptics.  He actually played The Turk (and lost) in 1820, and while he concluded it was a human player, it still inspired him to invent mechanical computers.  Babbage created early designs for calculating machines called difference engines in the 1820s and '30s.  He also created designs for more complicated machines that were very capable computers.  Or could have been, if they had ever been built.  Among the devices that Babbage designed was a machine that could play Tic-Tac-Toe.  Again, though, such a machine was not built in his lifetime.

The next great events in the world of computing and strategy games seemed to wait another century.  In the late 1930s, Alan Turing and Claude Elwood Shannon laid down the theoretical basis for modern computing.  Alan Turing built an actual electromechanical computer to break the german ENIGMA cypher machine.  Perhaps based on this reaserch, or perhaps simply because we had reached the right time in history, clever engineers gave us two computerized strategy games.  In 1940 and 1941, two different machines were created that could play the game Nim—the Nimatron, and the Nim Machine.  One of these is generally credited with being the first machine that could play a game of strategy.

The Nimatron, 1941 World's Fair.

So imagine my surprise when I stumbled across an article from the January 1879 Issue of the Journal of the Franklin Institute, in which "An Automatic Tit-Tat-To Machine" is described by Frank T. Freeland.  The article is an explanation of how the machine is to be built and how it will function.  But more than a mere design, the article states that the machine was built, and was able to successfully play the game.  Allegedly it never lost.

How is this possible?  Could such a machine exist, and yet escape the observations of decades of computer historians?  Perhaps, like "The Turk", this machine itself was a hoax?  At this point, I have no corroborating evidence, but I'll lay out the case for the machine's existence.

Francis Theodore Freeland was born in Philadelphia in 1859.  He attended the University of Pennsylvania where apparently his game machine was an undergraduate project in his senior year.  He graduated in 1879.  After graduation he stayed on for two years as an Instructor of Mechanics.  He was a member of the Engineer's Club of Philadelphia.  He wrote at least two more papers (Linkages for x^m, American Journal of Mathematics, volume III, 1880 describing a machine for finding roots; and A Machine for the Solution of the Equation of the Nth Degree, Engineers and Engineering Volume 2, 1880) and gave several talks (such as calculating the weekday from a date, and space-filling calculations for sphere stacking).  In 1880, Engineers and Engineering mentioned that he exhibited a machine for solving quadratic equations, although I haven't been able to find any signs of a paper on this, or even if the machine worked on digital or analog principles.  But it's also an indication that his mechanical and mathematical skills were well-demonstrated.

According to a footnote of the Freeland's paper, the machine and his paper ended up at the Franklin Institute by the recommendation of William D. Marks.  Dr. Marks was a professor of mechanics at the University of Pennsylvania.  Eventually he would be an honorary lifetime member of the Franklin Institute, and the Whitney Professor of Dynamical Engineering at the University, a very prestigious position.  These facts, more than anything else, lend a great amount of credibility to the device's proper functioning.

Still, I can't find any corroborating evidence.  While I have found a few articles in other sources about this device (Gentlemen's Magazine, American Stationer, and a newspaper in Illinois), they all seem to be rehashes of the Franklin Institute article, rather than independent stories based on seeing the machine.  This is the most troubling issue.  Wouldn't such a machine be major news, and create a big stir?  Perhaps.  Or perhaps with the folklore of "The Turk" still looming in the culture people were skeptical and unimpressed.

During his time as an instructor, Mr. Freeland took a post-graduate course (which I assume means an entire program, not merely one class) in mining engineering.  He moved to Colorado, where he became a well-known figure in the mining world, publishing papers, writing, filing at least two patents, and making a fairly large pile of cash.  He retired young and travelled the world.  Unfortunately he also died fairly young, at the age of 48, during a return visit to Philadelphia.  His obituaries talk at length about his mining contributions, but no mention at all is made of his Tic-Tac-Toe machine or other mechanical contributions.

Mr. Freeland apparently had a most promising career ahead of him as a mathematician and mechanical engineer.  Why did he move to Colorado and take up mining?  Was it simply a desire for fortune?  Unfortunately we're not likely to find answers on these questions.

The next example I can find of a Tic-Tac-Toe machine is a 1949 "Noughts and Crosses" machine built by Donald Watts Davies in Great Britain.  It included some sort of lighted display grid, and a plugboard for making moves (see it in action).  This was followed in 1952 with a Tic-Tac-Toe machine made from phone relays by the Bell Telephone company to demonstrate the power of their new relays.  If his device really worked, Mr. Freeland was seventy years ahead of his time.

Davies' 1949 Noughts and Crosses machine.

As for Freeland's device, I've only given a brief look at the design.  One positive is that the "user interface" looks like something that anyone could use.  It has a Tic-Tac-Toe board with mechancial Xs and Os that slide up and down in front of openings.  To play, you would move the sliding X/O into place yourself.  It seems that no help was needed by the player.  Without a human operator between the player and the machine, it would seem fairly difficult to pull off a hoax.

Cutaway front view of Freeland's machine, showing the X/O sliders.

Another positive is that the machine wasn't that showy.  "The Turk" was all about the visual performance, and (of course) the inner workings were kept secret.  Here the inner workings for this machine were published and reviewed by the best mechanical minds of the day (decades after difference engines were designed and built).  (And as an aside, the Nimatron pictured earlier does not survive this standard very well—it was very showy, large enough to hold a person, and supposedly had a built in delay to simulate the time it would take a person to play.  Hmm.)

So far, I'm fairly optimistic that this was a real working machine, but I have to remain open to the possibility that this machine was simply another "Turk".  The obvious next step is to analyze the device in more detail and see if it could actually function.  Stay tuned.

The Magic of the Modern Font

One of the oft-repeated myths around sentence spacing is that modern fonts don't need an extra space because the fonts know how to fix sentence spacing.

What does that mean, exactly?  The statement is so vague as to be meaningless.  What needs to be fixed?  How is the fixing accomplished?  How does the font magically determine what is and is not a sentence?

If you feel that sentences don't need any extra space at all, I suppose you could argue that fonts do nothing, and that this “fixes” the problem.  But I really don't think this is what anyone means.  Everyone seems to believe that the font is doing something.  What could it be doing?

The answer is, not much.  Fonts have almost no magic built into them (except for the scaling and low-resolution rendering magic, which is impressively magical).  But they do have pair-kerning, and this may be the magic that some people are putting their faith in.

Kerning is the adjustment of the overlap of, or space between letters (or other characters).  Historically, a kern used to be the piece of metal that stuck out on letters that needed to overlap.  For example the upper-case T should overlap with most letters.  For letters where this wouldn't work, typesetters had to insert an extra space to avoid a collision with the kern.  And “kerning” itself didn't really exist, other than as a word for manufacturing type to a given (and permanent) width.  Once a font was created, kerning was never adjusted.

But these days, you can't shake a stick without hitting some hipster designer babbling on about how he tuned the eff out of the kerning in his new t-shirt design.  Of course, this is always about a designer making a mostly graphical design with some text.  Nobody actually goes through and tinkers with the kerning of the text in the middle of their term paper.  For plain text, people use the kerning that is built into a font, just like they did with handset type.

In modern fonts, each character is created with a default width, much like the physical width of the sorts back in the day.  But modern fonts have this concept of pair kerning too.  Pair kerning makes adjustments to the kerning for specific letter pairs.  For example with the letter T, you want letters that could tuck under it to actually do so, otherwise they will appear much too spacey.  All good modern fonts do this, and they do it all the time for everything.  This means I can type in a sequence of letters like “LYAVATJ”, and the pair kerning built into the font should know that every single one of these letter combinations is special, and requires a bit of extra adjustment.  This makes modern fonts far superior in their unaltered spacing to old handset fonts.

A good modern font might specify special tuning for hundreds of pairs of letters.  So this can be used to fix sentence spacing, right?  Wrong.  Modern fonts almost never define any special pair kern for the period and space.  And the very few that do, define a negative pair kern, that is, they reduce the size of the visual space after the period, making it even smaller than a word space.

To understand why they don't do this, we only have to observe that periods are not simply sentence enders.  They are also used in abbreviations and (perhaps most importantly for this context), in initials.  So for example, J. K. Rowling might well need to be tightened up in the eyes of some, and in this case a negative kern makes sense.  Of course there is no way for the font itself to distinguish this usage with a sentence boundary usage.

[And as much as I love grammatical and typographic holy wars, I'm not going to get into discussions at this time as to whether one should use J.K. Rowling, J. K. Rowling, J K Rowling, or even JK Rowling.  Suffice it to say that on this one, the historical record is all over the map.  Also, let me add that neither J-period-space-K nor J-period-K did a very good job of kerning on my screen just now.]

There are even some fonts that build in extra space to the default width of the period, and then apply pair kerning exceptions for that.  This minority of fonts de facto adds extra space between sentences.  But again, it also adds extra space between initials and after abbreviations, where it isn't needed.  Hopefully fonts that use this method will then add in pair kerning for numbers so that 3.14159265 does not end up with extra space in the middle.  And hopefully letters too so that e.g. the text “e.g.” won't be split open.  This seems like a bad idea, complicating the font, while not actually fixing anything.

When you get down to it, fonts are just data, which is why they really can't do anything.  Modern software though is a different story.  Whether we are talking about page layout or word processing or even the text tools in image software, the editing of text and conversion of unformatted text into formatted text can be pretty spiffy smart.  So maybe when people say that fonts magically fix things, they actually mean that the software is fixing things for you?

Sadly this is not the case.  You can't just “trust the software”.  You have to know what your software is going to do, because there is almost no consistency in handling of sentence spacing in software.  Unix tech-heads that use vi and emacs can benefit from sentence detection based on two spaces between sentences (hooray).  But troff originally could only identify sentences when the terminal punctuation was followed by a carriage return or newline.  More modern gtroff does this, but also uses the two-space habit as a means of sentence detection.  TeX takes the view that EVERY period followed by a space is a sentence boundary, and you have to explicitly find each occurrence that is not a sentence boundary and label it by hand.  Yuck.

But those are all old-school plain text processors that format from plain text to print.  More modern WYSISYG software, like Word for example usually has much worse support.  Word dutifully trusts that all spaces you type are spaces that you mean, and keeps them.  Other editors have taken their cue from the monospacers and only let you type one space.

In fact, the only consistent software out there is the majority of software that chooses to do absolutely nothing with sentence spacing.  In other words, the majority of software does not fix sentence spacing at all.

If there is any software out there that takes into account the period-space pair kern and attempts intelligent sentence boundary detection and tries to fix spacing of initials, abbreviations, decimal points, and sentence boundaries all in different ways based on all of these factors, I'm not aware of it.

In order for software to “fix” sentence spacing, it has to know what a sentences is.  The best way for that to happen is for users to tell the software what a sentence is.  And as luck has it, we already have a way to communicate this to the software, with two spaces between sentences.  Sadly fonts don't, and can't, have any magic that will do this.  Software can have this magic, but modern software developers generally don't have a clue about real typography.

One or two spaces after a period? How about three?

The number one myth I see repeated about sentence spacing is that wide spacing came from the typewriter, to make up for the monospaced fonts.  I've talked about it before.  But it seems like it really needs its own dedicated article, and I think I have a more complete argument anyway.

There's basically two parts to this myth.  First, that monospaced fonts demand wider sentence spacing.  Second the statement or implication that wide sentence spacing was created just for the typewriter.

Let's start with the idea that monospaced fonts demand wider sentence spacing.  This is the hardest argument to make, being largely an aesthetic argument.  Still, there's always been a certain logic to visual design.

Growth (shrinkage) of letters from Times New Roman to Courier.

Most of the letters in a proportional font would have to grow significantly to match the size in a monospaced font.  And typically the space character must grow the most.  If we look at the New Times Roman font, we see that the two narrowest characters are the space and the period.  These two characters combined as they would between sentences take up less space than even a pair of "i"s or a pair of "l"s.  (And let me be clear that, contrary to another myth, modern fonts offer no "magic" that fixes sentence spacing; this is probably a topic for future blogging).  In New Times Roman it turns out the letter "n" is exactly the same size as the space and period combined.  The word "no" is twice as wide as the space between sentences.  The word "me" is even wider than that.

Comparison of sizes of various characters, Times New Roman on the left vs. Courier on the right.

But in a monospaced font, every character is the same size as the next.  If you used one period and one space between sentences in Courier (or any monospaced font), together they would be the twice the size of the letter "n" and the same size as the word "no".  In effect, with one space between sentences in both fonts, the monospaced font doubles the visual gap between sentences, relative to the other letters and words.  Or, it doubles the space relative to the letter "n".  Of course even the letter "n" grows going from New Times Roman to Courier, so in effect the sentence spacing in a monospaced font (with only one space) is even more than doubled compared to the proportional font.  The only logical argument you could make is that with those giant spaces and punctuation characters, you'd be better off reducing space between sentences, not increasing it.

But as I said, this is an aesthetic argument, so perhaps applying logic or math is just not relevant.  Fine.  The claim in this myth is that people looked at monospaced fonts, and said "this is unsightly, we must use more spaces".  If so, shouldn't there be a historical record of this?  If this myth is true then we would find narrow sentence spacing, and then the invention of the typewriter, followed by wider sentence spacing (two spaces) only on the typewriter.

And yet this is what printed text looked like before the typewriter:

From The Innocents Abroad, Mark Twain, printed in 1869.

First, lets get past one other part of this myth that sometimes crops up: the font is proportional.  There are those who are so confused on this issue that they believe that proportional fonts themselves are a modern, post-typewriter invention, but this is not true.  Proportional fonts have been the standard since movable type was invented in the 1450s.  And now look at the sentence spacing: it's very wide.  It's about three times as wide as the space between words, although you'll note that the word spacing varies a bit in order to make each line justify.  Nevertheless, wide sentence spacing existed before the typewriter.  But this was no isolated incident.  Almost everything printed in English used this same standardized spacing.  Here are more examples of sentence spacing, mostly taken from books about printing and typography:

Mechanik Exercises: Or the Doctrine of Handy-Works Applied to the Art of Printing, Joseph Moxon, 1683

New England Courant, December 11 1721, published by James and Ben Franklin (this was the second newspaper published in America, founded in August 1721).

Typographia, John Johnson, 1824

The Practice of Typography: Modern Methods of Book Composition, Theodore Low DeVinne, 1904

Modern Printing volume 2, John Southward, 1915

Note that the last two are actual descriptions of sentence spacing.  The 1904 quote is instructions for the Linotype, while the later 1915 book shows more traditional hand-compositing instructions (although this is a later edition of an 1898 volume, which I was unable to find).

Almost everything printed in English from the late 1500s until around 1950 used more space between sentences than between words in a sentence.  This wasn't some haphazard practice, it was very well-defined: every typography manual I've ever read up until the invention of the Linotype that describes sentence spacing says to use an em quad.  This is a spacing element in handset type that is as wide as the typeface is tall (including the minimal built-in line spacing).  As I mentioned with the illustrations above, the em quad is about three times as large as the spaces used between words in that era (modern fonts tend to use even narrower word spaces, closer to one quarter the size of the em quad).

One of the examples above is from Johnson's Typographia, an early seminal work in typography.  In addition the excerpt shown, it also says the "m-quadrat is the proper space after a full point when it terminates a sentence in a paragraph."  This exact same wording can be found in literally dozens of other books on typography in the nineteenth century, largely because copyright laws were less well-defined back then, and Johnson's Typographia was heavily plagiarized.  Of course Johnson didn't invent this standard.  Here's what a few other books on typography (that did not plagiarize Johnson) say on the subject:

"M-quadrats mostly begin paragraphs, by an indention of the first line ; one of them is likewise the proper space after a full-point, when it terminates a sentence in a paragraph." – Typographia, J. Southward, 1824

"The m-quadrat ... is used ... after full stops, to designate that the sentence is ended." – Universal Technological Dictionary, Volume 1, George Crabb, 1833

"An em quadrat [...] separates one sentence from another when the end is indicated by use of a period" – American Dictionary of Printing and Bookmaking, W. W. Pasko, 1894

On the other hand, there are some texts that don't describe the practice at all.  For example, Joseph Moxon's "Mechanick Exercises: Or, The Doctrine of Handy-works Applied to the Art of Printing" is one of the earliest known works about typography (1673).  He makes no mention at all of sentence spacing (although he uses wide spacing, as shown in one of the illustrations above.)  Several other books on typography gloss over this detail.  How could any comprehensive book on printing or typography skip over something so critical?

Take a look at this:

Enhanced image of the original copy of the Declaration of Independence, 1776.

That's a piece of the Declaration of Independence.  And it uses wide sentence spacing throughout.  But that's got nothing to do with the printing business, right?  Remember that in this era, every single document that was not printed on a printing press was handwritten.  And the standard practice in that era for handwriting was wide sentence spacing, just like the printing business.  (Note that it's unclear to me which came first, as I haven't researched handwriting practices in the 15th or 16th centuries.)  It was what everyone learned to do in school (if they learned to write).  Wide spacing was so ubiquitous in English, some authors felt it literally was not worth mentioning.

It was into this environment that the typewriter was born.  It began production in 1873, and by the mid 1880s began to become a fixture in modernized offices.  It was only natural that the same wide sentence spacing used everywhere else was also used on the typewriter, and that's exactly what was done.  And how much space did they use between sentences?  As I said, the em quad was about three times as big as typical word spaces back then, so naturally most people recommended three spaces between sentences, to replicate professional print quality.

Here's what some sources instructed in those early years.

"Properly divided sentences add considerably to the neatness of a document, and this may be done by spacing three times between each." – The Shorthand Review, Volume 3 No. 4, April 1891 

"... it is well to strike the space-key once after a comma, twice after a colon and semi-colon, and three times after the question mark and full point." – The Shorthand Review, Volume 5 No. 10, October 1893

"A full stop at the end of a sentence should be followed by three spaces, as also the note of interrogation and the note of admiration." – A Manual of the Typewriter: A Practical Guide to Commercial, Literary, Legal, Dramatic and All Classes of Typewriting Work, I. Pitman, 1893

"...and remember to space three times following each sentence." – Typewriting Instructor and Stenographer's Hand-book, Practical Textbook Company, 1892

And here's a typewriter sample from the Remington typewriter catalog in 1886 (those circles are a typewriter keyboard diagram, bleeding through from the reverse side of the page):

Remington Standard Typewriter, typewriter sample, 1886

Not all sources agreed though.  Both "Scott Browne's Typewriting Instructor" from 1882 (ish) and "How to Become Expert in Typing" (1890) stated that two spaces should be used.  There's hardly any mention of a disagreement.  In 1899, "Pitman's Journal of Commercial Education" (also known as the "Phonetic Journal") described an article in "The Gem" and notes its disagreement on spacing with "Pitman's Typewriter Instructor".  And a year earlier, a letter was written to "The Phonographic Magazine" asking the following:

"Are there any stated rules for spacing after punctuation marks in typewriting?  I notice that some writers put a space after a comma, and two spaces after a period, while others do not put any extra space after a comma and only one space after a period.  Which is correct?"

The answer provided was the latter, one after a comma and two after a period.  But it's interesting that the writer had seen such diminishing space on the typewriter.  At any rate, over the next couple of decades, sources all converged on two spaces between sentences.  I've found no explicit statement on why this happened.  But it seems obvious that given the extra wide size of spaces and punctuation on these monospaced typewriters, three spaces just looked too big compared to printed text, and some minority even felt that two were too big.

This is the complete opposite of the modern mythology of sentence spacing.  Remember that the claim is typewriters demanded more space because of the wide and uneven letter spacing.  The myth is false—history shows that typewriters demanded less space than people were used to.

So to review what we've demonstrated here, wide spacing did not come from the typewriter, and the large spaces on the typewriter apparently demanded a change towards less, rather than more space between sentences.

In short, there is not one single shred of this typewriter myth that's true.  In my opinion, it's a great litmus test to see if anyone blabbing on about typography actually has any clue of what they're talking about.

So where did this myth come from?  I covered my hypothesis in a recent blog.  But that's just a bit of the history.  This myth is really about marketing.  It's about people (generally lazy editors) who want to make everyone change their spacing habits by pigeon-holing two-spacers as being old fashioned and clueless.  It seems to me we should be fighting fire with fire.  But that's a topic for another day.