Hebrew With Points... in Hot Metal

quadibloc's picture

The topic says it all. Newspapers could be set in Frank-Rühl with a Linotype machine well enough, and in cold metal, tiny spaces and quads could handle the job of putting vowel points and Masoretic accents below letters.

But it's hard for me to believe that Hebrew, when set with points, was invariably set by hand, and only by hand, until the computer age. My web searches have not pointed me to any information on how Hebrew with points was set, either on a Monotype caster, or with classic photocomposition equipment, but I would think that somehow it was managed.

I could be too optimistic. For Monotype, something like the method used in four-line mathematics would be needed - to retype the line of text when producing accents so as to produce blank slugs of the same width as each letter when an accent was not desired. But then making the accents match their corresponding letters in width would presumably require counting units all the time. The alternative, somehow casting the accent on the same body as the letter, doesn't appear to have been something that would have been considered practical.

gohebrew's picture

Funny, I spoke today with Scott-Martin Kesofsky about this topic.

When Scott was young, he worked in this field. He has friends who also did this, and I believe that one currently still does (the last of the Mohekans). Mathew Carter, he told me, was actually a punch-cutter, a long long time ago. I'm sure that he might be able to point you in a useful direction. He's on this forum too.

It would be fascinating, if those who had access to metal Hebrew type would scan and post them. High resolution scans I would create into OpenType fonts, to preserve the designs for future generations.

I know at RIT in Rochester, NY had set up a printing press for metal type, and used it about 20 years ago. I don't know if they have the manpower anymore. You might look there.

quadibloc's picture

I've hunted around some more, and I've been able to confirm, from the Alembic Press web site, that Monotype did offer the ability to set Hebrew with points for at least one typeface - Sonzino, a face similar to Meruba (apparently a classic Siddur style).

gohebrew's picture

Sonzino, aka Soncino. was the surname of the first families to engage in Hebrew publishing. I believe David Bomberg was afterwards.

The Soncino I don't think was associated with them. I might be wrong. Rather, the Soncino name was applied to denote an old design.

In my view, Soncino is not at all like the Merubas design, which appears in many many versions.

quadibloc's picture

I found one image of Monotype Sonzino on the web, in a U.S. government document giving a classification of type styles, which it illustrated with Latin and non-Latin scripts. That short sample, visible at


(scroll down to 4.3.1, Serifs::Modern::Continental)

looks more like Vilna than the sample of Soncino given. (The image is apparently a .BMP file, and it may not display properly in all browsers.)

Also, I should mention that in my web searches earlier today, I ran across this page, in which some people might be interested:


They're selling an edition of sixty numbered copies of "The Hebrew Types of the Jericho Press".

gohebrew's picture

Here is the sample. It's misnamed.

It is a variant of the classic square letter, rarely - if ever - used by Orthodox Jews, and recently by most everyone.

One of the ways to tell that the design is "not in the fold", is by looking at the design of the shin.

quadibloc's picture

I re-visited the Alembic Press site, and noticed that Peninim was also available with points. Peninim 217 was the version of that with points, and Sonzino 218 was the version of that one with points.

And here's another sample of Sonzino, misnamed as it may be, at


The image itself is


and the image seems to answer my question as to how Hebrew with points was achieved with the Monotype caster. Apparently, they achieved it the same way as polytonic Greek was achieved, by making every combination a separate character - which was possible because only the vowel points, and not the Masoretic accents or cantillation marks were offered.

Of course, at 22 letters, times 12 possible vowel point values (7 basic vowels, sh'va, no point, 3 vowels with sh'va... oh, wait, furtive patah makes 13), times 2 for the presence or absence of the dagesh still won't quite fit in a matrix-case, but it could be that some combinations are uncommon.

gohebrew's picture

See how each is a little different.

quadibloc's picture

It is a variant of the classic square letter, rarely - if ever - used by Orthodox Jews, and recently by most everyone.

That seems quite strange to me.

From my viewpoint, shaped by typographic fashions in the Latin alphabet world, the Vilna/Romm/Siddur faces seem to have the same kind of timeless elegance as Helvetica and Times Roman, so I would expect them to be the standard.

Frank-Rühl, on the other hand, seems dated. Perhaps like a so-so Scotch Roman, for example, Monotype's Modern No. 7, much beloved of the makers of scientific textbooks before 4-line mathematics came along. Of course, this would apply to some forms of the classic square letter too; Drugulin is clearly dated.

On the other hand, the popularity of faces like David, Hadassah, or Koren seems mystifying. They are good typefaces, and so they should be popular... up to a point. But for them to be as popular as they are now, for general purpose use, seems strange. After all, Lydian or Freehand or Zapf Chancery, praised as they may be, are not normally used for setting ordinary text.

Of course, some of the answer to that may be found in the form of the Hebrew letters themselves. The dictum "Type follows the pen" (I am not sure to whom this one is owed; I saw it come up in a discussion here, but I forget the name of the type designer: I don't think it was Gerrit Noordzij or John R. Biggs) is considerably more true for Hebrew than it is for the Latin alphabet, for which type often follows the chisel.

gohebrew's picture

Let me clarify, for many non-Jews find my earlier point illogical. It is illogical. The issue is not design related, but use related.

This typeface design, and others, have nothing 'wrong' about them. But they have been used for Christian publications. As such Orthodox Jews feel intimidated to use the same designs for their holy books.

Hence, Romm etc is acceptable, but this is not.

During the last 40 years, Frank-Rühl has become an established standard in Jewish religious publications, as Hadassah too.

Koren and David never caught on, even though the main Bible in Israel is in Koren. If we study the design attributes of Hadassah, it may well be the best Hebrew design as of yet. Frank-Rühl is more popular than you describe.

quadibloc's picture

I'm confused. I wasn't describing the popularity of Frank-Rühl; I'm aware that it's very popular, but to me, it looks a bit dated, and so, it seems to me it would be less popular if Israelis looked at Hebrew typefaces with an aesthetic based on the same fashions as generally applied to typefaces in the Latin alphabet.

I can understand what you describe, though. Since Christian editions of the Old Testament in Hebrew follow their form of practice of respect for the text of Scripture - put the best text known from current research in the body, and list any alternates in footnotes, even if the traditional text is now an alternate - a Christian edition of the Old Testament in Hebrew is likely to give offense to many observant Jews.

And so the idea - whether it is on the part of readers or publishers - may be to ensure that when someone picks up a genuine copy of the Tanakh prepared according to the Jewish tradition, they ought to be able to know at once and at sight that they're holding the real thing in their hands.

What I am puzzled by, instead, is why Vilna isn't being used for newspapers and ordinary secular literature!

gohebrew's picture

>> I'm confused ... in the Latin alphabet.

Don't be confused. In Hebrew, there is a pertinent saying which explains the unendin appeal of Frank-Rühl. The Hebrew saying is: "טעם וריח אין להתווכח". This means literally: "We don't debate taste and smell". This means that things like beauty, tastes, or smell, are subjective, and can not be objectively judged.

Frank-Rühl established itself in the early twentieth century, and by default replaced Vilna as the standard Hebrew typeface design. Perhaps, it became so popular, because it was so exposed, in newspapers, ads, books etc. Perhaps, it was so used because it has narrower widths than Vilna, because it did away with the Bodoni like thin and thick strokes, or maybe because G-d wanted us to look at Frank-Rühl a lot.

>> I can understand what you describe, ... in their hands.

No, you don't understand yet.

Since a typeface was used in a non-Jewish book, even in a phonebook, were there such a thing, Orthodox Jews would shy away from that typeface. They would prefer a Jewish-looking typeface instead used only for Jewish books.

It doesn't have to do with Christian scholarship, or the fact "their" typeface was used in an Old Testament. In fact, the early Hebrew printers, like David Bomberg, was not Jewish, yet his printed works were warmly accepted in the Jewish community.

In fact, there is an interesting discussion in the Talmud (tractate 12, Bava Metzia, the Middle Gate) about at what point can a Jewish person buy a book from a non-Jew, if the same kind of book is available from a Jew.

Jews are considered relatives to each other ("כל ישראל חברים הן"), and try to support each other financially. However, if the non-Jewish bookseller cries out, then Jewish buyers must buy from him or her.

>> What I am puzzled by, instead, is why Vilna isn't being used for newspapers and ordinary secular literature!

It was until the mid-tentieth century, but since then Frank-Rühl took over. I think the main reason that Frank-Rühl was so used, and Vilna was put out to pasture, was because the main typesetting manufactures only made Frank-Rühl.

This no longer is the case.

gohebrew's picture

Vilna Hudar - one of many Vilna "square-letter" designs

gohebrew's picture

The old version

Scott-Martin Kosofsky's picture

I haven't read through this entire thread, but I'd like to quickly shed some light on Quadibloc's question, before Shabbos begins in Boston. Mechanical typesetting with nikkud was achieved by the American Monotype Company in 1927, and was the system used to set hundreds of liturgical works throughout the world, though perhaps most notably by the Press of Maurice Jacobs, in Philadelphia. Linotype, who had offered Hebrew since about 1903, had never been able to solve the nikkud problem within the confines of their system. (One should also mention that the Ludlow Company, which made hand-set line-casting machines, mostly for display, offered a wide variety of Hebrew types, many specifically for the Yiddish newspaper market, which was a thriving and highly competitive business in early 20th-century America.)

My friend J. F. ("Chip") Coakley, ex of Harvard and now of Cambridge University, has recently written the best short short history of early mechanical Hebrew typesetting (in metal) to date, in a lovely book he printed himself (letterpress, of course) at his Jericho Press. I recommend it highly. The website is www.jericho-press.com.

More later!

quadibloc's picture

Now that the sun has set here in Edmonton this Saturday night, I thank you! I did come across the Jericho Press book in my web searches. I wish I could just buy a copy, but at the moment, sadly, that would be awkward. (I presume the book you're referring to is "The Hebrew Types of the Jericho Press"? Or is there one that I didn't see mentioned even more closely connected with the subject?)

My main interest is in finding out what technical device was used to solve the problem. As I may have noted, I am guessing that something like the scheme used for 4-line mathematics is involved, where one first types the letters (with any points appearing above them or within them precomposed) and then types another line with the points. Qoph and some final forms would overhang, so one would have to always put a space/quad in the corresponding position when doing the line of points, with their points also precomposed.

Since overhangs in the Monotype system extend upwards, and not downwards, however, it may be that the points are typed first - and that the usual reversal mechanism for setting Hebrew is not used, with the characters being upside-down in the matrix case instead. (Which would explain where the upside-down Alephs appearing in mathematics came from.)

And, if there were a need for so many precomposed characters that both the points and the letters couldn't fit into the same matrix case - there was an attachment to the keyboard to allow preparing two tapes at once (usually used for preparing the same body copy in two different point sizes or column widths)... and so one could use that to prepare a points tape and a letters tape, and have two separate runs on the caster.

In that case, one should do the points run first, so that the kerns aren't lying around to get broken.

EDIT: Seeing a video of a caster in operation, I see that doing the letters and points in separate passes is not an option, because the lines of text are stacked together by the machine.

quadibloc's picture

The fact that the machine stacks successive lines of text together also affects the utility of dispensing with the reversal mechanism. Hmm...

quadibloc's picture

Ah. Let us assume that the existing mechanism for setting type in Hebrew operated not by reversing the direction in which type left the mold beneath the matrix-case, but instead the Hebrew matrices were upside-down in the mold, and what was reversed was the direction in which lines of type were assembled in the galley.

That would seem to take care of everything: now, the overhang from characters taller than the slug would extend below the slug, and the line of characters would be typed first, followed by the line of points.

However, the problem would then be how to avoid the kerns below the slugs from being broken as each new line of type is added to the column of text upon the galley.

First, the bar that shoves the type to the operator's left, now, instead of to the operator's right, with the galley now on the other side of where the row of type exits the machine, would have to be made lower so that it definitely only makes contact with the row of type under space height, so as not to push on the kerns.

Second, instead of pushing the type just out of the way, it would have to push the row of type a little extra distance away, so that the kerns, if any, are out of the way of the next row of type.

Third, so that the type slugs containing the points, when pushed into those containing the letters, don't themselves catch on the kerns, the first small area of the galley, past which each row of type is pushed, needs to have a small incline, so that the new row of type approaches the preceding one from below.

(Even if I have correctly anticipated the mechanical improvements required in every particular, I still wonder as to just what, precisely, would be the contents of the matrix-case for such typesetting. I suspect that it would be difficult to find room for everything that was needed, but the 15 x 17 matrix-case may have been available then, although this was long before the days of the 16 x 17 matrix-case and the unit-shift that went with it.)

quadibloc's picture

What I proposed would have been possible if only an extremely limited set of accents appeared above the letter. Perhaps Rafe, `Ole Weyored, Pazer Magnum, and Azla, and no others. Combine them with the presence or absence of Dagesh, and one's matrix-case is already almost full.

Of course, there's Holem to take care of as well.

If that isn't possible, it seems like more radical steps, outside of what is usually acceptable as extensions to the Monotype system, would be required. It seems that, at the least, either the height of the mold would have to be subject to variation, or slugs would have to be allowed to have projecting kerns at both top and bottom.

Thus, in one pass, one types the accents that go above the letter - and the letter Lamed, with its accents precomposed.

In the second pass, one types the letters other than Lamed; only Dagesh and Mappiq are precomposed.

In the third pass, one types the accents that go below the letter.

It seems that a pass in which the line is quadded out would be required between the second and third passes, or blank spaces between each line of text plus excessively large projections below the line, or a way to vary mold height would be required if this is needed. So I may be missing some ingenious step that was used in practice.

EDIT: Examining closely a typical example of Hebrew with points, while the accents above the line are too varied to allow precomposed characters, they are also infrequent, so perhaps they could be set by hand. However, the two examples I've seen of Hebrew set by this method didn't include any accents above the line, except for Sin dot, Shin dot, and Cholem, including Cholem with Vau as a mater lectionis.

Also, the set sizes for the Hebrew faces with points are significantly smaller than their point sizes, so apparently the wide characters aren't included; thus, it seems that even with points, and even allowing for some rare cases to be handled by hand, this system was not intended to allow fully annotated Biblical texts to be set.

gohebrew's picture

New Hebrew

quadibloc's picture

That's definitely a nice classic Hebrew face. Incidentally, though, I can't seem to find the currently-active web site of the concern where your typefaces are made available. The site http://www.philidor.com appears to be forthcoming, http://www.gotalmud.com/ appears to be related to Jewish faith and life in a more general way or to a new edition of the Talmud, and another site I located that I thought might be associated with your typographic endeavors appears to mostly sell software for Arabic-language word processing.

quadibloc's picture

I have received a personal communication, which I have permission to share, from David Bolton of the Alembic Press, conveying information from Chip Coakley.

From a document available on the www.letterpress.ch site, I found that Peninim with points comes in 7, 7 1/2, 8, and 9 points, and these sizes are cast in 5 1/2 set, and 7, 7, and 7 set respectively, and that Sonzino with points comes in 9 and 10 points, both at 7 1/2 set.

I had wondered why they were set with such a narrow size of em. But one of the items of information from Mr. Coakley explained this.

As I surmised, somewhat like the superscripts and subscripts of 4-line mathematics, one first typed a line of Hebrew, and then under it a line of points. But to simplify typing the line of points, the Hebrew letters were cast in only two widths: 9 units and 18 units.

Given the narrow set, my assumption is that the letter Shin was designed to be the same width as most Hebrew letters, and the narrow letters (that is, regular Ghimel, Vau, Zain, Yod, Nun and final Nun) were half-width. Placing a vowel under the vertical stroke of Daleth and Resh was probably accomplished by using a half-width vowel, although certainly it would have been possible to find room in the matrix-case for a Chireq and a Sh'wa that were moved over even further to the right.

Although it was noted that overhangs were used, nothing specific was said to confirm my surmise that reversing the direction of the line of text also meant that the overhang was now on the bottom of the slug instead of on the top, as is the usual case on the Monotype system.

It was also mentioned that Monotype did have available a way of typesetting accents in addition to vowels with Hebrew, but there was some awkwardness with that.

My guess is that it would be necessary to type the accents with the letter Lamed in one line, then the other letters, then the vowels, and that would be awkward enough. But what was mentioned was that the height of the mold had to be changed between phases, perhaps with a special attachment to make that easier.

Syndicate content Syndicate content