To understand this issue you need to delve back into the design history
of "themed" systems.  The small accent holes for Aeolian's Themodist,
Wilcox & White's Solodant and Hupfeld's Solophonola systems predate the
88-note roll scale, being used in Aeolian and W&W's 65-note rolls, and
Hupfeld's 73-note rolls.

Both of these scales used smaller holes for accenting.  The large size
of the note perforations in these early formats forced the use of
smaller accent holes to keep the accents to a short length -- the time
taken for a large punch to cross the tracker bar would have made the
accent far too long for musical purposes.

In the case of the 65-note scale, the practical solution was to use two
small punches fitted into the width of a single one-sixth-inch note hole.
This design was carried forward almost unmodified into the 88-note scale
(which left a gap between bottom bass note and sustain pedal for the accent
holes to fit in so that Themodist rolls will play on any standard piano).

When the Themodist evolved into the Duo-Art the small accent holes were
retained, this time for backwards compatibility with Themodist rolls.
Aeolian's involvement with rolls for other systems, particularly the
Wilcox & White ones, carried the smaller perforations there as well.

With the 88-note scale the relatively smaller note perforations
substantially reduced the need for accent perforations to be smaller.
There are few musical accents that cannot be achieved by standard-sized
punches if they are correctly placed.  It is noticeable that the
numerous European makers who used accenting frequently chose to use
standard-size punches for the accents (and a couple of Italian makers
used only a single punch, not paired-up in "snakebite" form).

I suspect that if Aeolian had not already got used to the concept
of smaller punches it's possible they would not have introduced them
specifically for the Duo-Art.  However, the system did use small holes,
and the question is how critically it depends on them.

One thing that I have noticed with the software used to reconstruct the
master rolls for producing Duo-Art replica recuts is that the placing
of the accents relative to the notes is not as consistent as you might
expect.  It's not something that you'd really think about until
presented with software that can show you explicitly which punch is
on which row.

What I found with a very early recut I did was that it didn't accent
correctly, and when looked at closely it turned out that the accents
had come out one row early.  The rolls was cut at 21 rows per inch,
so the accent perforation finished before the note started; if the roll
had been one of the finer-resolution 31.5 rows per inch rolls the slots
would have just overlapped and the fault would not have been as
apparent.

Looking into this more closely, it turned out that the British roll I
had copied was cut with the accent perforations centred on the leading
edge of the punch row, and the software assumed that everything was cut
on a common leading edge as per the American-made rolls.  The advanced
accents had been 'reconstructed' a row early.

Once the situation was understood it was possible to tweak things so
that the accents came out in the right place, but it made me suspicious
of the results so I delved in a bit more, and changed my roll editor
program to highlight any accents that did not align with a note.  What
this showed was that in many cases the accents do actually lead the
notes by a row, and indeed frequently wander from being aligned to one
row in advance and back again.

If you have a look at the few Duo-Art master rolls that survive and
are accessible you can see how this might have occurred.  The accents
were put in by hand, not automatically during the recording.  The roll
editor decided which notes were to be accented by marking a pencil tick
against them.  Somebody then went though the roll and made a horizontal
pencil line at the roll margin, over which the accent perforation was
punched in _by hand._

These hand-entered accents were made using a single normal-size punch
in the original rolls that I have seen.  The originals were playable
standard rolls, not on the triple-length stencil used to control the
perforators, and as a result the accents are not rigorously aligned with
the note perforations.  Presumably accents that sounded wrongly would
be corrected during the editing process, and the others were left alone.

If you have a look at my web site www.pianorolls.co.uk you'll find
a new page I've just created briefly describing the Duo-Art recording
process and showing an original roll with hand-punched accents.

This begs the question about how carefully the accents were edited to
control the performance dynamics.  It's not just a case of whether the
accent is one or two punch rows long, but also a case of alignment with
the note being accented.  Some rolls certainly have one- and two-row
accents that look as if they have been designed that way, but the
majority don't.  I suspect it's a case of "maybe, sometimes" rather
than "always".  Up to the individual roll editor, perhaps?  I'd be
inclined to suspect happenstance rather than design unless the music
clearly shows otherwise.

As for the inaccuracy of using standard-size perforations for accents,
the majority of Duo-Art rolls that use two-row accents could be recut
perfectly accurately using standard punches, as long as the two-row
accent is substituted with one row of the larger punch.  Not that
anybody does this, of course!

That said, even the musically-fussiest folks I know have been entirely
satisfied with high-quality recuts using standard punches for accents.
What they do find unacceptable are the analogue (one-to-one) recuts
that tend to smear and elongate the accent perforations: these may
well have been recut using correct accent punches, but this is of no
use if the accent perforations are far too long and misplaced, which
has commonly been the case.

Anyhow, I hope that my own replica recuts satisfy all the requirements:
the right number of correctly-size perforations, correctly located.
It can be done, and is being done!

Julian Dyer