Boustrophedon | Hacker News Reader

Boustrophedon

en.wikipedia.org

82 pointsbreck59 comments

Fun story about this: in college we had to make a barcode reader out of a simple phototransistor. To scan the barcode, you had to slide the device along a piece of paper. The barcode scanner had to be isolated from ambient light so that we'd get a good reading.

However, our project was to read a lot of information via the barcode (a paragraph of English text encoded as Code 39 or similar, can't remember the exact details), so it wouldn't fit on one line of paper. Lots of teams decided to implement complex heuristics to determine when to ignore readings when lifting the barcode reader off of the page and to scan the next line.

But we decided to use boustrophedon to generate the page with the barcodes on it so that we never had to lift the barcode scanner off of the page. You just had to scan left to right, then down one line, and then right to left.

Scott Kim has a wonderful talent at designing "ambigrams". Check out his classic book "Inversions" and his gallery of more recent work!

http://www.scottkim.com.previewc40.carrierzone.com/inversion...

An inversion is a word or name written so it reads in more than one way. For instance, the word Inversions above is my name upside down. Douglas Hofstadter coined ambigram as the generic word for inversions. I drew my first inversion in 1975 in an art class, wrote a book called Inversions in 1981, and am now doing animated inversions.

I can see an advantage to having the beginning of the next line immediately beneath the end of the previous line -- it avoids a greater refocus adjustment distance to start new lines. This would greatly reduce mistakes where the nth or (n+2)th line is started by mistake instead of the (n+1)th.

However, I don't immediately see an advantage to flipping the letters across the vertical axis for each new line.

I found it surprisingly easy to read the mirrored examples in the article. I think it may be more comfortable -- after a while, at least -- to read the mirrored letters; because we don't read letter-by-letter anyway and the "word image" is intact (just mirrored).

Here's the example image (with mirrored letters): https://en.wikipedia.org/wiki/Boustrophedon#/media/File:Bous...

And here's a non-mirrored version:

  THIS EXAMPLE OF BOUSTROPHEDON TEXT WAS
  AIDEPIKIW EHT ROF YLLACIFICEPS NETTIRW
  ARTICLE  ON  THIS OX TURNING METHOD OF
  TNEICNA  NI TXET HTIW LLAW A GNIREVOC
  GREECE AND ELSEWHERE

You raise a good point in that we don't read letter-by-letter anyway. So, this variant, which leverages recognition of whole words at a time, would have the line tracking advantage of Boustrophedon without the problem of duplicating letters for forward and reverse directions:

  THIS EXAMPLE OF BOUSTROPHEDON TEXT WAS
  WIKIPEDIA THE FOR SPECIFICALLY WRITTEN
  ARTICLE  ON  THIS OX TURNING METHOD OF
  ANCIENT  IN TEXT WITH WALL A COVERING
  GREECE AND ELSEWHERE

While we don't read letter by letter, this does still require some left-right scanning for longer words. In texts (or languages - German?) with many long words, this would be an issue.

The initial method seems to do the best job of not triggering the part of my brain that wants to read normally - which is a benefit, as I can then use a specific reading method for boustrophedon text, rather than fighting against my initial reactions.

Problem here is that if you enter mid-page it's not necessarily obvious which direction to read.

But you can figure it out trivially easily based on which way the letters are flipped.

A sibling comment suggests that it's to help encode the direction, but I wonder if there's a less intentional reason, which is: why not mirror the letters? If I were coming up with conventions for reading and writing for the first time or possibly based on descriptions that some sailor had reported, it's not obvious to me that the direction of the letter should matter, especially if the letters you are using aren't mirror symmetric with other letters (lower case wasn't invented until ~1,000 years after the Hellenistic period, when boustrophedon fell out of favor).

Exactly. And I think the idea to encode direction is a bit anachronistic, in those days they didn't even use punctuation or spaces so it was pretty hard to read anyway, you'd have to do a lot of thinking. And of course with Semitic writing you didn't even encode the vowel sounds.

In hieroglyphic inscriptions the animals etc do face the direction of reading (which varies), but I'm not sure this was in order to help people read it easier.

I was told in my university class that the glyphs faced the direction of reading because they were thought of as talking to the reader. And, of course, it's rude to talk to someone when you're facing away from them.

> it was pretty hard to read anyway

Yes, and they did not invent the space character back then.

the wrtng looks like ths tht my couisn just writes andgets mad if i dont understn him

If you're interested in this concept, you might try BeeLine Reader, which uses line-wrapping color gradients to avoid the line transition errors you describe (and doesn't involve letter flipping). Disclosure: I am the founder!

http://www.beelinereader.com

Please add real Boustrophedon! I'm so interested in what it is like to read longer documents like this.

Wait seriously? So Boustrophedon, with color gradients to pull you down in the right direction? I'll have to play with this (and vanilla Boustrophedon) a bit to see if we want to offer it. Thanks for checking us out, and for the suggestion!

My best guess there is that flipping the letters encodes the reading direction of the line, to make it easier to pick up reading at any specific place.

Hi, this is completely unrelated to this discussion but I can't find any other way to message someone on this website.

In an earlier comment you mentioned being signed up for cryonics: https://news.ycombinator.com/item?id=10209152

I'm currently considering the procedure and I'd love to talk with someone who has signed up for it. Could you send me an email please? Nagnazul at gmail dot com

Thank you very much!

With enough practice, I could see this being much easier to read. There's no jumping from the end of one line and searching for the next, plus each line is visually distinct from the ones immediately above and below.

I wonder how much text I would have to read like this before it felt natural?

> I wonder how much text I would have to read like this before it felt natural?

If you see my note above I mention that it wasn't a big deal. More to your question, frankly learning ancient Greek grammar was far more painful -- the effort to learn to parse old inscriptions was in the noise compared to the effort to understand what the hell they were writing.

http://boustro.com/app

It's surprisingly difficult to stop my eyes from snapping back to the left side after each line.

I also find that my brain wrongly pattern-matches words that form a different word when reading the letters in reverse.

So, I for example often misread "was" as "saw" or "on" as "no".

I first encountered the word in the excellent book, Expert C Programming, by Peter van der Linden, who explained that declarations in C are read boustrophedonically.

In C++, on the other hand, they're read stroustruphedonically.

I encountered "boustrophedon" a long time ago because it is one of the favorite words of the creator of the niche-but-awesome phrontistery.info

http://phrontistery.info/favourite.html

What was the reason for writing this way on stone? It can't be to keep text aligned - you can just walk to the left and keep going, and the text will stay aligned just as well as if you'd been writing right-to-left.

What is the reason for writing always in the same direction?

I don't ask this as a joke: is there seriously a cognitive advantage to having to pass your eye all the way back to the beginning of the next line (whether you read right-left, left-right or top-bottom) vs immediately dropping down and continue reading?

On the other hand I spent a lot of time in high school and university reading pictures of boustrophedon greek texts and once you are used to it you don't even think of it. Likewise R-L texts -- my difficulty with handwritten arabic is the wide variation in characters; the directionality rapidly became automatic, even in mixed-language texts.

In the case of boustrophedon texts, a dominating factor is the sheer irregularity of character shapes (nothing was machine made) and worse, a lot of the black and white photos in books are old, crummy ones which make the deciphering even more "interesting". And this was all before the development of spaces between the words too, and at a time when character shapes were evolving.

Regardless, all written languages I know of (even counting Korean, which is slightly different in-word), seem to have picked a single direction as a dominant, and even before the development of mechanised writing (which would make the characters extremely regular, which might drive a tend to optimize). So there could be some cognitive benefit. I could imagine there's an advantage in that you could train a NN in your visual system to recognize regularly-shaped characters and do look-ahead, but I have seen no studies of it.

> What is the reason for writing always in the same direction?

Only about half the glyphs to recognize. Being able to skip further down the "page" and trivially continue reading. Being able to rapidly scan a line of text.

I guess this is all predicated on a mostly literate populace with a lot of reading material. You probably don't need to rapidly scan a wall carving extolling the virtues of your God-Emperor and recording all of his battle victories, unless you're trying to win a bar-bet. ("I'm telling you, he beheaded TWELVE infidels with a single swing!" "Yeah, well, find it on the wall and I'll buy you a beer.")

> Only about half the glyphs to recognize.

This is not really a big deal. Some writing systems require you to recognize a few thousand symbols, and I read (i.e. "effortlessly recognise") three alphabets and with some concentration two more (the grammar is a bigger deal for the latter...my overall fluency in those languages is, at best, poor and I am sure will never get better. But symbol recognition is not the barrier).

This is because I put in the effort to study these languages, but many people, without much formal education, can read that number of alphabets (much better than I can) simply because they live in (typically border) regions where that is reality of daily life. My own mother grew up in a town where four writing systems were in use, plus used a different alphabet in her home.

Don't mock that older writing by the way. Most people were illiterate; Especially anything in paper was the writing of some pretty sophisticated folks.

I wonder if the standardization of directions, rather than boustrophedonic writing, is due to the technology of paper and ink. If you're carving into stone or clay, then your hand isn't as likely to smudge the already written text.

I used to think that, but a bit of reflection puts that to bed:

1 - The earliest ink-and-paper writing in the Middle East was in Egypt, where the pen writing mostly followed inscription, being a little free form in ordering. AFAIK the old language didn't acquire an alphabetic system until Greek colonization (cf Coptic, which is (evolved) pharaonic Egyptian written with essentially Greek characters). That latter is indeed a L-R system, but has been largely replaced with the R-L Arabic.

2 - Your hand is likely to smudge R-L systems unless you're left handed which most people are not.

3 - The other in-and-paper system, Chinese, until recently (couple of hundred years ago) used only brushes, not pens, so smudging was a non-issue. Top-bottom though could be attested in this way; it's better to draw the brush rather than push it, so starting at the top of the page makes sense, but why R-L vs L-R...?

I don't know about the the third independent site of the development of writing, the Americas, as the jesuits burned almost 100% of the books, and so mostly we only have monumental writing, which doesn't address this question.

AFAIK all the indigenous Alphabets of India are L-R (Urdu is R-L due to its use of an arabic alphabet)

> Your hand is likely to smudge R-L systems unless you're left handed which most people are not

Arabic was originally read Right-to-Left, but written Top-to-Bottom, leftwards, on the scroll so smudging by right-handed writers wasn't an issue. A completed scroll was then hung sideways on the wall for people to read. In fact, by writing Arabic T-B(leftwards) instead of L-R(downwards in columns) like Greek was at the time, smudging became less of an issue. The Arabs could roll up the left side of their scroll sooner without smudging or pausing, whereas the Greeks had to leave the entire column they'd just written exposed while the ink on the bottom rows was drying.

But there is also the disavantage of going back to the beginning of the page and losing your line. Happens to me too frequently to be annoying (while reading left-tp-right and top-to-bottom languages).

Boustrephodon works very well for _words_ -- but badly for numbers. Even Roman-style ones. I suspect that put it at a disadvantage.

IN the days when people used boustrophedon there were not separate glyphs for numbers and even positional number orientation wasn't especially common.

As it happens the big-endian notation of hindu (aka "arabic" in English) numbers is kind of a botch in L-R languages since you have to parse the entire goddamned thing to learn anything about it (and when it's long you have to parse it r-l anyway y divided sections). At least in R-L writing you can get evenness immediately.

Double the characters to recognize; double the heuristics to internalize for rapid reading.

No idea what the relative cost of those are but I'd wager it exceeds the mechanical cost of scanning back to the "front" of the line...

A mirrored letter still looks very similar. Most people learn the letter topologies first, and only later pin down the chirality. Just think of all the mirrored N's you probably used to write!

So it's definitely not twice as much work, I'd say it's not more work at all, you basically just don't worry about the chirality and that's it.

>So there could be some cognitive benefit

This is only really the case for phonetic writing where you learn to recognise the shape of groups of letters (this is how you are reading this text). For logographic/pictographic writing it is much less of a benefit.

And as I wrote above, even for alphabetic writing like Greek it doesn't really matter if all you read is the occasional monumental inscription.

I assume you meant to write "alphabetic" rather than "phonetic" since many languages that use alphabets use them in a way only loosely tied to sound (which you allude to by talking about reading in chunks of grapheme clusters). Korean is a notable exception but I'd bet most readers don't parse the sound either.

No I really mean phonetic. Any system where you combine graphemes more or less freely to form words will be parsed groupwise. Though much less so for Greek monumental majuscle writing than modern books.

Take Japanese writing for example, you definitely don't parse individual kana. Same with hangul, as you mention.

Even for Chinese writing it is true but to a very small extent, but my point was that it is much more important for phonetic writing, and most important indeed for alphabetic. I have even heard that it is less pronounced in Cyrillic, because all letters basically look like majuscles so the outline of words is quite uniform.

I don't quite get the comment about alphabetic script not being very phonetic, it's true that this would promote parsing words as units, but the main reason we do it is because it's much faster, not because some spellings are opaque. Kana spelling is almost completely one-to-one but it still makes sense to read e.g. -ma.shi.ta as a unit.

Writing involves a couple of different metaphors; the main one is the correspondence of a written symbol with an oralsound. Another one is that a linear sequence of glyphs matches a linear sequence of sounds. Obviously there are mechanical problems with fitting a long sequence of letters on a limited space. Boustephedron maintains the linear sequence of the letters by going back and forth. People quickly learned that it's much easier to read if you instead cut each line into segments and start each segment on one side of a page, but this breaks the visual metaphor of a continuous line of written symbols matching a continuous line of sounds.

Inventing writing is really, really, really hard. Just as, given that we learned to read so long ago, it's hard to remember how hard it is to learn to read. Boustephedron is like training wheels for writing.

Early writing wasn't as standardised as it is now, and most reading would be done aloud in more or less formal circumstances, so the maximum efficiency we strive for today was probably not a consideration.

And since writing on stone is a pretty cumbersome activity it was probably natural to continue writing next to where you just finished a symbol.

Note that cuneiform (which of course predates Greek by a wide margin), written on clay with a reed, was not written boustrophedon.

On the other hand, Egyptian hieroglyphics, though carved in stone, weren't either - instead they were very flexible on writing direction.

I suppose my only real point is that the Ancients didn't see a great need for a fixed writing direction.

In our system, it becomes challenging to find the start of the next line when reading long lines of text. This is pretty much the main reason why sheets of paper or modern webpages are only as wide as they are.

Here's a tangentially related article about a crossword-solving program called Dr. Fill, which got tripped up by boustrophedon in its debut at the American Crossword Puzzle Tournament in 2012.

http://www.crosswordtournament.com/articles/iw032212.htm

My son would write like this when he was first learning to write. I was constantly having to prompt him to carriage return, and write the letters the correct way round.

See also: the Boustrophedon cellular decomposition for robotic 2D coverage path planning, e.g. for mowing, painting, vacuuming, etc. The method works by decomposing an arbitrary environment into a union of smaller, simplified shapes that are easy to cover with a boustrophedon motion - moving back and forth like an ox plowing.

http://planning.cs.uiuc.edu/node352.html

Older NVIDIA GPUs (and maybe current ones) used an alternating-direction scheme for coarse rasterization which was called (somewhat tongue in cheek) boustrophedonic rasterization. The idea was to have more spatial locality for the coverage samples issued by the rasterizer compared to a conventional same-direction scheme. Spatial locality in the two-dimensional sense correlates to spatial locality in the texture cache, for example.

More or less the ancient Greeks - in their simplicity - chose to use ASCII 10 (newline/line feed) instead of 13+10 (carriage return+line feed). ;-)

https://blog.codinghorror.com/the-great-newline-schism/

>Even if you rule out unicode and stick to old-school ASCII, like most Facebook relationships â€¦ it's complicated.

Brilliant, encoding problems in an article about encoding!

The Floyd Steinberg error diffusion dithering algorithm can use a boustrophedonous scan order to eliminate the diagonal geometric artifacts you get by scanning each row the same direction.

https://en.wikipedia.org/wiki/Floyd%E2%80%93Steinberg_dither...

"In some implementations, the horizontal direction of scan alternates between lines; this is called "serpentine scanning" or boustrophedon transform dithering."

I implemented some eight bit cellular automata heat diffusion rules with error diffusion, which accumulated an unfortunate drift up and to the right because of the scan order.

Rudy Rucker pointed out the problem:

http://donhopkins.com/mediawiki/index.php/CAM6_Simulator

"Rudy Rucker: I feel like you might have some kind of bug in your update code, an off-by-one thing or a problem with the buffer flipping. My reason is that I see persistent upward drift in the action, like if I mouse drag a blob it generally moves up. Also the patterns appearing in the blob aren't uniform. I mean...this IS supposed to be the 2D Rug rule, isn't it?"

So instead of scanning back and forth boustrophedoniously (which wouldn't eliminate the vertical drift, just the horizontal drift), I rotated the direction of scanning 90 degrees each frame ("spinning scan") to spread the drift out evenly in all directions over time.

https://github.com/SimHacker/CAM6/blob/master/javascript/CAM...

    // Rotate the direction of scanning 90 degrees every step,
    // to cancel out the dithering artifacts that would cause the
    // heat to drift up and to the right.

That totally canceled out the unwanted drifting and geometric dithering artifacts! That made it possible to cultivate much more subtle (or not-so-subtle) effects, like dynamically switching per-cell between different anisotropic convolution kernels (see the "Twistier Marble" rule for an extreme example).

http://donhopkins.com/home/CAM6

And down the Wikipedia rabbit hole...

IBM 1360 Photo-Digital Storage System: https://en.wikipedia.org/wiki/IBM_1360

My Latin 101 prof used to go through the rows of seats in boustrophedon order, his words, to read passages from Wheelock. Don't miss that class very much.

>> boustrophedonically

Such a pitty I don't play Scrabble...

I used to write some of my college notes this way, when I was trying to make myself thing more about what I was writing.

I use this as my handle online in a few places (github, irc). Hi!

My first reaction to seeing this headline was recognizing it as a username. I thought that was here but given you are newly registered it must have been the place with the ducks.

I am not familiar with "the place with the ducks" as far as I know, so you probably have the wrong person.

that could have been useful to read text on a glass wall from both sides... too bad that they had not invented it yet.

I was sure it will be a very illegal substance by that name.

Reply on news.ycombinator.com