End grain glue joints

Please give a look at this video. I remember harpsichords (maybe Delin?) whose cheek and spine were just glued to the wrestplank, without inletting and without wooden pegs. Just a rub glue joint, of course with hide glue.

The video explains why the endgrain-to-endgrain glue joint is strong enough and actually stronger than sidegrain joint. Makers, what do you think?


1 Like

I think this analysis is fundamentally flawed. The tests showed that
with side-to-side and end-to-side joints, the glue joint was always
stronger than the wood: the samples always broke in the wood and the
glue joints did not break (despite the fact, incidentally, that he was
gluing machine-planed surfaces which had not been improved by hand
planing them, and were therefore rough with the marks of the planer

He was breaking the wood, not the glue joint: all you can say about
the glue joint is that it was stronger than the wood. He has not tested
the strength of the glue joint

By contrast, the end-to-end joints always broke at the glue joint. In
that case, the glue joint was weaker than the wood.

It seems to me the tests do not prove the assertion that
endgrain-to-endgrain joints are stronger than side-to-side joints. Quite
the reverse. The traditional wisdom is therefore correct.


1 Like

There is some discussion on this here:

I agree with @Peterclav.

I would note that not all historical harpsichord makers were necessarily completely consummate woodworkers.

They are not the only profession with problems holding things together!

See here


1 Like

Sure, but if I understood correctly the video, its author’s point is that the wood is so much stronger in an end-to-end joint, due to the grain direction.
However, yes, he should have tested the strength of the glue joint directly.

1 Like

This note is in response to Domenico’s message of 1/31/23.

A first observation: Your writing “…whose cheek and spine were just glued to the wrest plank, without inletting and without wooden peg.” seems to be concerned about using glue only, without any mechanical interlocking. I have a Frank Hubbard kit (kit # 625) that uses neither “inlet” nor “wood pegs”. It does have, however, a 7.5”x1”x1.75“ batten supporting the wrest plank (7.5”X1.625” in cross section ), which battens are glued each to the spine and the cheek, and supported by three screws each, besides being bonded with Titebond glue (probably the “Original” version) that came with the kit. Its front end is also bonded to the name board (which is also simply bonded to spine and cheek) as is the west plank along its total length in addition to resting on the batten(s) with a glue bond, as well as having is end sections glued to the spine and cheek, respectively. While this is not a “simple” end-bond of the wrest plank against the spine or cheek, this combination, which has more than doubled the glue-surface for the wrest plank, has not given me any (apparent) trouble in nearly 50 years.

Incidentally, Claudio has, I think, the same kit from very nearly the same time (mine was purchased in 1973). He may recall whether I remember that situation correctly.

As a second point — I think Peter Bavington is right. This “mythical endeavor” is flawed in that it mixes qualitative results in particular test geometries with generalized notions. Mr. Sullivan uses a very special test geometry that stresses the glue to its extreme (mostly) over a very small (strip) region along the top surface of the bending test sample: that situation is not even a close model for the glue line which the wrest plank ends experience against the spine and the cheek.

Here is essentially Sullivan’s bend test.

The tension across the bond decreases from the top layer to zero half way down through the thickness, then turns into compression at the midplane, and increasing to the maximum compression across the bond line at the bottom surface (not considering the compression against the tubular support). Moreover, he never even offers an estimated value for the maximum tension stress at the failure line along the top surface, so that there is not even any type of estimate for the strength of a bond for a comparison with a different loading arrangement.

In contrast, the wrest plank bond experiences forces (stresses) that is, for present purposes, described by shear, in which two adherends are stressed so as to slide relative to each other parallel to the bond line, not across it: For the purpose of this simple discussion the stress in the glue across the “glue layer” is uniformly distributed over the entire bonding area. That is why, typically, bonds in a shear mode of deformation can be very strong. In fact, most engineering issues involving safety concerns make use of bonding under shear.


Another comparison, though somewhat weak in details, is the difference of tape adhesion: When one peels the tape, only the small region ahead of the separation point (line) is involved, similar to the line in Mr. Sullivan’s bending test. However, when one pulls on the tape along its length, the (shear) forces are distributed along the length of the tape, so that for a (sufficiently large length) of tape, the tape will break off ahead of where the adhesive starts .

So much for why the mystical treatment of adhesion is mystical.

PS: You may wonder why I respond to this issue: I have researched adhesion mechanics for decades for the Navy, the Air Force and NASA.