Doi Seal Timeline - Koitsu 'Benkei Bridge' Study

We continue our Doi seal timeline study, following on from the initial analysis undertaken by Marc Kahn and myself (Ross Walker) on Tsuchiya Koitsu's 'Zozoji Temple' print. In this most recent study we target Koitsu's 'Benkei Bridge' print - arguably one of the most popular prints of this artist. Four versions of this print are analysed:

  1. My Doi Hangaten print with in-line publisher/printer/carver seals (A)
  2. My 'Hanmoto Tokyo-Do' print (B)
  3. Marc's Doi Hangaten print with offset publisher/printer/carver seals (C)
  4. Marc's Doi Hangaten print with in-line publisher/printer/carver seals (D)

Hangaten in-line - print 'A'        Hanmoto Tokyo-Do - print 'B'
Left: Published by Doi Hangaten with in-line seals (A). Right: Published by Hanmoto Tokyo-Do (B). Hangaten in-line - print 'A'        Hanmoto Tokyo-Do - print 'B'
Left: Published by Doi Hangaten with offset seals (C). Right: Published by Doi Hangaten with in-line seals (D).

The original files used for this analysis are available for download here. Please note that some of the files are very large (over 5 megabytes). Print (A)
600 DPI
Print (B)
600 DPI
Print (C)
150 DPI
Print (D)
150 DPI

Obvious Differences

1. Margin text

Prints (A), (C), and (D) have the title "Benkei Bridge" written in the right margin. This is completely missing from print (B). Also, the dates of publication for the two versions are different: 1933 for prints (A, C, D) and 1934 for print (B).

2. Koitsu seal and signature

Koitsu signature and seal

Prints (A, C, D) have a different Koitsu seal compared to print (B). An analysis of the Koitsu signatures written above the seals suggests that the signature on print (B) is not a copy of those of prints (A, C, D), i.e., written by the same hand, but at different times. The locations of the seals and signature within the print are also different for the two print versions.

We can see some block wear in the locations indicated by the yellow arrows, which varies between prints (A), (C) and (D). Seal (A) has the two kanji characters joined in two locations, and the two characters are joined to the outer ring at the top. Print (D) has the kanjis joined at only one location, but the kanjis are still joined to the outer ring. Print (C) has the kanjis completely separated from each other as well as the outer ring. Also, it appears to have some block damage indicated by a printing loss to the ring on the left-hand side.

Is this joining of the seal components (kanjis and outer ring) the intended result of the original block carving? A result of block wear? Or perhaps the over-use of printing ink?

3. Publisher seals and paper toning

Doi Publisher SealsThe publisher, printer and carver seals for each of the four prints are as shown on the left. Not much can be said about paper toning and print age at this stage, because different scanners were used to scan the four prints. Each scanner will introduce its own colour, brightness, and contrast distortions. The scan of print (C) is particularly dark. Also, paper toning not only depends on time (aging process) but also storage conditions, paper quality, etc.

Block Analysis

1. Keyblock wear - outer border line width measurements

Print (A) Print (B) Print (C) Print (D) Here we show the widths of each prints' outer border, measured at 5 locations. The locations of measurement were decided by finding those points on the border where the line width could be measured relatively accurately. Units of measure are 'pixels'. The scans of prints (C) and (D) were resampled to 600 DPI (from 150 DPI) to allow comparison with those of prints (A) and (B). A rough calculation of the line width in inches for each print can also be determined, i.e., width in inches = width in pixels * 600.

The measurements for prints (C) and (D) are estimates only, because the original scans were made at only 150 DPI. At this resolution, the line widths are only 2 or 3 pixels wide, so a measurement error of even only one pixel represents up to 50% of the line width.

This quantitative analysis suggests that Print (D) has narrower keyblock lines, followed by print (A), and this is supported by visual inspection of the print scans. However, as can be seen in many of the following images (particularly images 1, 4, 9, 10, and 11), print (D) has particularly poor, textured, heterogeneous keyblock lines, while those of print (A), while being slightly wider, are also quite smooth and homogeneous. Perhaps the washi paper used for print (D) was quite textured, or less ink was used to make the impression(??).

11 13 15 11
15 16 15 12
12 15 18 8
9 14 11 8
11 13 12 13

2. Keyblock wear - broken keyblock lines

Here we will show a series of images which compare areas of each print containing what appear to be broken keyblock lines.

Keyblock faults 1
Image 1 - keyblock damage

The image above details possible keyblock damage on print (B) and (D), expressed as several vertical discontinuities in the horizontal black keyblock line. It is interesting to note that prints (B) and (D) appear to show damage in the same locations. The locations of damage shown by the right arrows were certainly an exact match. Damage as indicated by the left arrows appeared to be ever-so-slightly offset.

Keyblock faults 2

Image 2 - keyblock damage in lantern frame

More possible keyblock damage is shown in Image 2 above. Part of the lantern frame on prints (B), (C), and (D) is missing (red arrow) -- the frame appears to be intact on prints (A).

There also appears to be damage to a keyblock line on prints (A) and (C) [on the left-hand side of the lantern cover/top]. However, upon careful inspection of prints (B) and (D) it was found that, rather than there being missing black ink on prints (A, C), this difference has been caused by extra brown ink from the brown colour block filling this area on prints (B, D).

Another difference between these two prints is the size and shape of the red lantern flame. I may be wrong, but it appears to me the colour block used to produce this element on print (B) is different from the other prints. Prints (A) and (D) appear to have the same flame impression, and the lights are out in print C!

Take a look at the right-most vertical lantern strut that supports the lantern cover (blue arrow). This strut appears to be straight on print (A), but is clearly curved on prints (B), (C), and (D).

Finally, take a look at the knob on the top of the lantern cover (green arrow). On print (B), the black and brown colour blocks are well aligned, but on prints (A), (C) and (D) they are slightly horizontally offset.

Image 3 - keyblock difference in tree branch

Here in Image 3 we show another example of keyblock line differences which separates print (B) from the other three prints.

Image 4 - keyblock damage on bridge outline

More keyblock damage common across prints (B), (C) and (D) (red arrow). Or is it damage?? Print (A) appears to have a continuous black line in this area but it is very hard to tell due to the brown ink.

The blue arrow shows keyblock damage which appears to be common to both print (B) and (D).

Also, note the poorer brown colour block alignment in (B).

Here we have more discontinuities in the tree branch outlines of prints (B), (C) and (D). This break does not appear to exist in print (A).
Image 5 - more keyblock damage on tree branch outline

2. Keyblock wear - Analysis of leaf area

Here we show two image tile overlays, taken from the same area of prints (A) and (B) -- too hard to do for all 4 prints. We've adjusted the transparencies of the two images so that they are both visible. Also, the image tile from print (B) has been moved slightly to the right, allowing us to see and compare the leaf sizes in each print.

For each leaf pair, I've marked the leaf that appears to be the largest -- either the leaf from print (A), or from print (B). I haven't had time to do an actual quantitative comparison (measuring the area of each leaf).

Well, I don't know about you, but I can't conclude much from this particular analysis. Neither print exhibited consistently larger or smaller leaf area. I guess we may be seeing the 'opposing' effects of block damage (smaller leaves) vs. block wear (larger leaves).
Image 6 - Leaf size analysis.

Image 7 - Leaf size analysis.

Image 8 - More leaf size analysis

Here we have a clear case of a leaf size being smaller in print (B) than in the other three prints. One of the leaves on the right of print B (red arrow) is almost non-existent, and appears to have it's left portion missing. Is this difference a result of a later printing of (B) with extra damage? Or was print (B) made from a different set of blocks? We know from the direction of keyblock damage in the previous images that the direction of the keyblock's wood grain appears to be in the vertical direction. This would explain the missing left-portion of the leaf mentioned above.

Image 9 -

Here we show yet another example of prints (B) and (D) showing similar characteristics (longer branches) that are different from prints (A) and (C).

Image 10 -

Too many morphological differences in all of the prints to make any comments here. Print (D) is particularly 'different'.

2. Colour block wear

Finally, we detail what appears to be the effects of wear on two colour blocks. Prints (A), (C) and (D) have negligible wear in the blue colour block and slight wear in the brown colour block. Wear in both colour blocks appears to have increased significantly in print (B). This once again suggests that print (B) is the later printing or impressed from different woodblocks.
Image 11 - Colour block wear

Image 12 - Further colour block differences in (B)

Notice the quite significant morphological differences in this brown coloured area of print (B).

3. Printing Errors?

Print (D) has double printing in a small area (two copies of each leaf). I assume this was simply due to some sloppiness on the part of the printer, rather than there being any differences in the woodblocks.


Well, the jury is still out. The body of evidence suggests that print (B) by Hanmoto Tokyo-Do has been printed from recarved blocks, but I feel that we have not yet proven conclusively whether or not prints A, C, or D were printed using any recarved keyblock or colour blocks. While differences in keyline and coloured areas do exist between each print, we do not yet have enough examples to determine whether the differences are due block recarving or due to natural block wear and ageing. This is where I need your help! If you happen to have examples of this print in your collection, it would be a great help to us if you could email us high-resolution scans of your print(s). Scans at 600 DPI saved as jpg files would be ideal, but even scans at 150 DPI are still somewhat usable. We thank you in advance.

Ross Walker

PS: Special thanks to our other research collaborators: Dr Andreas Grund, Tosh Doi, Thomas Crossland, and David Bull.

Back to Research & Articles Back to Home Back to Ohmi Gallery Home