Professional Documents
Culture Documents
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
https://about.jstor.org/terms
Galpin Society is collaborating with JSTOR to digitize, preserve and extend access to The
Galpin Society Journal
EUCHMI (4302):
A Case Study of Harpsichord Identity
nsigned historical instruments often present Collection of Historic Musical Instruments. Formerly
particular issues for modern researchers, in the collection of Raymond Russell, it was included
especially if the example is of particular amongst the gift of his collection to the University
interest because of a connection to a particular maker by his mother in 1964. It appears as number 2 in
or school, or because attribution would give a much the original catalogue (1968),1 re-numbered as
greater understanding of instruments of that type HS1-A1620.2 at the time of the initial web-based
in regard to the geographic or time spread for that data sheet lists,2 and numbered (4302) following the
form of construction or model, or for its potential amalgamation of the University's musical instrument
musical use. One field of organological study which collections into a single entity (EUCHMI) in 2004.
has received considerable interest, particularly in The harpsichord has a single keyboard, present
the past several decades, has been that of Italian compass C-d"\ with two registers at 8' pitch. In
keyboard instruments. Even with the studies by its construction it is entirely consistent with an
Italian origin, consisting of an inner instrument
many highly-respected writers it is clear that there is
no consensus about some of the methodologies used, with a cypress case and softwood (probably fir)
much less the results which have been concluded soundboard which is placed in a painted deal3 outer
from them. Rather than any attempt to provide an case, resting on a similarly-decorated stand made of
overview of the various research approaches in a lime or poplar (Figure la in the colour section).
systematic way, this paper will provide a study of an Including the casework (but not the applied
unsigned Italian harpsichord belonging to Edinburgh mouldings), the harpsichord is 1872 mm long,
University considering various approaches with the 775 mm wide across the front of the instrument and
intention of identifying which methods stand up to is 193-194 mm high. The cheek/bentside, bentside/
scrutiny with this example, and hopefully leading to tail and tail/spine corners are mitred, and the
a logical and sound attribution. decorative scrolls at either end of the keyboard are
thickened by added pieces. The wrestplank is made
A BRIEF DESCRIPTION OF THE INSTRUMENT of walnut, topped with cypress which has its grain
EUCHMI (4302) (Figure 1 in the colour section) is an parallel to the spine. The top of the soundboard
unsigned harpsichord in the Edinburgh University and wrestplank is 62-65 mm from the top of the
1 Sidney Newman and Peter Williams, The Russell Collection of Early Keyboard Instruments (Edinburgh, 1968).
2 These are individual data sheets which were mostly written by the then Curator/Director Dr Grant O'Brien. The
data sheet for this particular instrument was made public on 14 October 2000.
3 The term 'deal' is used to refer to a softwood used for parts in which its tonal qualities are not important. It might
be pine, spruce or fir.
17
y_
c'
386
517%
103%
112
368
492
79%
88
768 122 742 98
c 1011 129 973% 104
F 1412% 136 1376% 114
C 1529% 142% 1492% 119%
4 Strictly speaking there is no way to determine that the natural touchplates are original. However, the likely
originality of the accidental touchplates is due to their being entirely consistent with the evidence of the instrument's
alterations, and it is most plausible that if the accidental touchplates were re-used, then so to would be the naturals.
5 The likely original state will be discussed in detail below as it is dependent on the original nut position. The
measurements given in this table assume that the analysis given below is correct. The plucking points have also been
measured from the presumed nut position to the present rear (right plucking) quill line.
6 The initial identification of the original and second states was made by John Barnes, and his arguments can be
found in his unpublished report, Italian Harpsichord c. 1600 Russell Collection No. 2., written May 1975 (kept in the
archives of the Edinburgh University Collection). He also discusses the instrument in some detail in his article 'The
specious uniformity of Italian harpsichords', Keyboard Instruments: studies in keyboard organology, 1500 - 1800, ed.
Edwin Ripin (Edinburgh, 1971), pp. 2-3.
7 As Barnes, (1990: 2) shows, this is confirmed by an X-ray of the balance rail showing balance pin holes from the
first and second states.
8 Barnes' 1975 report was written before a second set of X-rays were taken which showed that the second (52 note)
state included re-pinning the wrestplank and, presumably therefore, re-pinning the bridge. These X-rays can no longer
be found in the Russell Collection archives, but the various balance and tuning pin positions were mapped onto
drafting film which has been preserved in the archives.
9 The other logical alternative ? than a different-feeling keyboard was desired ? can be effectively eliminated as
the balance rail position is unaltered and the octave span remained the same (as the touchplates were re-used). It is
remotely possible that the original keyboard was badly damaged and needed replacing, but against that is the fact that
the touchplates were able to be re-used and that there is no evidence of damage to the casework.
101 would like to thank Denzil Wraight for this attribution, based on a comparison of the moulding profile of the
nut with other instruments.
the EUCHMI files, written by Barnes reads 'Part instruments while also mentioning
of a date is written on the parchment backing of of a Roman origin. It should be po
the lowest part of the rose, which has evidently Wraight's discussion of the harp
come from a legal document. It reads 'Vij Octobris paper 'A construction principle
MDCX...' indicating a year between 1610 (MDCX) harpsichords'17 should not be co
and 1649 (MDCXLIX) (Figure lb in the colour attribution in favour of a Roman
section). Legal documents were probably at rather as support for the argumen
least 10 years old when used for roses giving the the main subject of his paper.
harpsichord an earliest date of about 1620'.15 ? Grant O'Brien has attributed th
? In his doctoral thesis16 Denzil Wraight describes to 'Naples, c. 1620' on the basis of
the instrument (his No. W325) as 'possibly a measurements. This appears on the
Roman harpsichord judging from the case style, added to the University of Edinbur
a Florentine origin is also possible, and probably 14 October 2000, and on his per
more likely considering that the instrument where his arguments are given in
appears to have been modified in Florence'. claviantica.com/Neap_sch_files/Ce
Wraight retains his attribution on his website and elsewhere on the same site used
listing the harpsichord amongst Florentine of a Neapolitan instrument.20 Mos
11 The present author does not necessarily agree with that view. AC- d'" c
music up to and including that of J. S. Bach. The compass was also very
decorated Hamburg-made harpsichords and clavichords by Hass, Zell an
Specken, who is believed to have worked with Gottfried Silbermann.
12 Raymond Russell, The Harpsichord and Clavichord (London, Faber and
13 Sidney Newman and Peter Williams, (1968).
14 John Barnes, 'Does Restoration Destroy Evidence', Early Music, Vol. 8, N
15 The author would like to thank Denzil Wraight for bringing this to his
16 Denzil Wraight, The stringing of Italian keyboard instruments c. 1500 -
Belfast, 1997, Part II, p. 327.
17 Paper given at the Edinburgh Early Keyboard Instrument Symposium, S
available in printed version from his website <www.denzilwraight.com/CP
18 Wraight uses (4302) to show that the principle he is presenting can
geographical locations. This principle, and its application to (4302) will be
19 <http://www.music.ed.ac.uk/russell/instruments/hslal6202/datasheet
20 For example, <www.claviantica.com/Neap_sch.htm>, and <www.clavian
claviantica.com pages accessed 4 April 2009).
21 Grant O'Brien, 'The Single-Manual Italian Harpsichord in the Royal College of Music, London, Cat. No. 175: An
Organological Analysis', Galpin Society Journal LX11 (2009), pp. 72-73. In his footnote 35 O'Brien uses the former
rather than the present catalogue number. It should be pointed out that the original state of (4302) had 57 notes, and
there was a second state with 52 notes, so O'Brien's statement that the present registers (with 51 notes) are re-used
original boxslides is incorrect.
22 Eleanor Smith, A Discussion of the Use of Divided-Accidental keys in Italian Strung-Keyboard Instruments pre 1700,
MMus thesis, University of Edinburgh, 2008.
23 This will be discussed in detail below.
24 See Denzil Wraight, 'Checklist of Italian harpsichords and virginals with split sharps' <www.denzilwraight.com/
ChecklistSplitSharps.rtf>, (accessed 4 April 2009). It should be pointed out that the listed Roman instruments only
reach c'", rather than/'" as in (4302).
25 <www.claviantica.com/Characteristics.htm>
26 For example another anonymous instrument, formerly in the possession of Grant O'Brien, now owned by the
Markiezenhof Museum, the Netherlands. See Grant O'Brien 'Determination of the Centre of Construction, Anonymous
Single-manual harpsichord, Property of Grant 0'Brien',<www.claviantica.com/Neap_sch-files/Cen_GOB.htm>.
27 Although Grant O'Brien, (2009: 70) states that 'the most compelling reason ... is the use in its design and
construction of the unique workshop measurement...', he also gives similarities of stylistic features between RCM 175
(the subject of his paper) and other instruments which have been attributed to Guarracino.
28 This relies, of course, on the assumption that the characteristics used are actually representative of a single location
rather than giving a number of possibilities. Grant O'Brien lists a number of characteristics of Neapolitan harpsichords
<www.claviantica.com/Characteristics.htm>, a number of which apply to instruments from many parts of Italy. The
lack of signed (rather than attributed, thus creating a circular argument) Neapolitan harpsichords of the sixteenth- and
seventeenth-centuries means that even some of the features one would accept as Neapolitan might actually be better
defined as, perhaps 'south Italian' at best. Other characteristics might be entirely meaningless, such as (using O'Brien's
list) cypress case sides, tail angle, rose construction, panelled nameboard, two straight bridge sections and decorated
sharps. O'Brien points out, when discussing (4302) that it is missing many Neapolitan characteristics, mentioning
only the nameboard. Even that is found elsewhere: the Venetian harpsichord by Bernardinus de Trasuntinis, EUCHMI
(4471), presently displayed next to (4302), has an identical nameboard construction, as also does the 1627 Bolcioni
harpsichord (EUCHMI (4304)). In practice, study of the broad construction features alone is not enough to determine
the location.
29 Denzil Wraight, 'The Identification and Authentication of Italian String Keyboard Instruments', in The Historical
Harpsichord Volume 3, ed. Howard Schott, (Stuyvesant, NY, 1992).
30 The fact that Cristofori's tools were passed on impedes the use of this method for identifying instruments by
members of the Cristofori school. It can also be shown that a large moulding (a flaw in the profile proves it was made
using the same cutter) was used on English virginals by Philip Jones (1671, Tabley House, Cheshire) and Stephen
Keene (1675, Private Collection, Cheshire). However, these two instruments are closer in design and style to each
other than the two surviving Keene virginals, and the virginal by Jones was made four years after he completed his
apprenticeship. It can be safely assumed that Jones was working in Keene's workshop at the time and that, probably,
Keene passed the order to Jones. The possibility that the instrument bears a false signature and is actually by Keene can
be discounted as no other mouldings match (whereas they do on the two Keene virginals, which date either side of the
Jones example. This agrees with our later knowledge of Keene, who went into 'partnership' (co-signing instruments)
with both Edward Blunt and Charles Brackley. Therefore it cannot be used as evidence to support the idea of specialist
moulding makers. Further evidence that suggests mouldings were made in-house (by the maker or, as may also have
been the situation with Cristofori, a sub-contracted case-maker) is the lack of correspondence between mouldings by
makers working in the same town and period, whereas roses (and possibly jacks and other parts) do appear to have
come from a common supplier.
31 There is no reason why the design should be expected to provide tangible proof which can be used in identification.
32 This point is made in Stephen Birkett and William Jurgenson, 'Why Didn't Historical Makers Need Drawings? Part
II - Modular Dimensions and the Builder's WerkzolV, Galpin Society Journal LV (2002), p. 189, when discussing the use
of a trammel. Such a method would almost certainly be observable since, even if any arcs were subsequently erased
(by planing, for example), the centre around which the trammel was rotated would leave a comparatively deep point.
However, as they also point out in 'Geometric Methods in Stringed Keyboard Instrument Design', Galpin Society Journal
LIV (2001), p. 276; there is no need to actually make any arcs, the trammel being used simply to transfer distances. This
would not necessarily leave any visible marks, but is prone to a larger risk of inaccuracy in the transfer process.
33 This point has been made by Grant O'Brien 'The use of simple geometry and the local unit of measurement in the
design of Italian stringed keyboard instruments: an aid to attribution and to organological analysis', Galpin Society
Journal LII (1999), p. 111.
34 <www.claviantica.com/Neap_sch_files/Cen_Russ.htm>.
35 <http://www.music.ed.ac.uk/russell/instruments/hslal6202/datasheet.html>, (accessed 18 March 2009).
the cheek/bentside join. Arguments can be made for the present author, and by others at the author s
either of the measured points reflecting the design request;40 seven separate sets of figures were so
intentions of the original maker.39 obtained. In examples where the same figure
In addressing the first of the above issues, the occurred on at least four occasions this mode figure
baseboard was measured a number of times by both was used, and where this did not occur the average
(to the nearest quarter-millimetre) was used. In this
Table 4. The mode or average measurements of the
baseboard using the same components as measured by way the following measurements were derived.41
O'Brien Using the same six components (spine, cheek,
Spine 1860 mm baseboard width, height, and tail parallel and
Baseboard width
perpendicular to the spine) as O'Brien (to ensure
764% mm (cheek/bentside join)
the methodology is identical) gives a total number
761% mm (front)
of millimetres of 3490% using the front baseboard
Cheek 433% mm
width, or 3492% using the baseboard width at the
Tail length 173 mm
cheek/bentside join. Both results are less than
Tail component in 93 mm the figures of O'Brien and would, using the same
direction of spine nominal inches, give respective figures of 21.814
Tail component 146 mm mm and 21.830 mm for the oncia. Both of these
perpendicular figures are even closer to (different) Neapolitan
to spine oncia measurements than the figure found by
Maximum height 196 mm O'Brien (Table 3) which would, at first view, appear
36 The error distance has been calculated using a linear (Lx) method rather than a Euclidian distance (L2) which is the
sum of the squares of the errors. The linear distance is more robust'. I would like to thank Prof. Joseph Kung, Dept. of
Mathematics, University of North Texas for his advice on this issue.
37 A negative sign indicates that the measured figure is smaller than the calculated equivalent and vice versa.
38 O'Brien, (1999: 164-5) (Table 11) shows that the historical sources all point to the Neapolitan once being divided
into five rather than four. On that basis it is historically unlikely that the original maker would have divided his once
into another subdivision for the baseboard design.
39 There are two very reasonable scenarios by which a difference can be explained. The first is that the maker is
simply inaccurate with his work, reducing the width at one end of the cheek by too much. There is no reason one
particular end is more likely to be reduced that the other. The second possibility is that the baseboard shrinks across
its width after being planed to size. If the lower belly rail was glued in place early after being planed the width would
be established there (the baseboard width would be determined by the length of the lower belly rail, which would not
shrink along its length) but the front could shrink further. Alternatively, if the keyboard front batten was glued early
in the process it would be the front that was stable, whereas there could be shrinkage at the cheek/bentside corner. Both
scenarios are plausible. The risk of shrinkage is notably greater with freshly prepared (ie. thicknessed) planks.
40 I would like to thank Matthew Hill and Eleanor Smith for their assistance in taking these independent
measurements. On no occasion was the instrument measured by the same person twice in a single day. On each
occasion two people were involved in taking the measurements, the second acting to ensure that the other end of the
tape measure or ruler did not move before the figure was read.
41 Including Grant O'Brien's figures listed in Table 3 (making eight sets in total) does not alter the results presented
in Table 4.
Table 5. Measurements of (4302) using the figures taken from Table 4, and the assumed number of once as determined
by O'Brien, using the case width as measured from the front of the case
Millimetres once Error in millimetres
(using average oncia)
Spine 1860 85 + 5.8
Baseboard width 761% 35 -1.7
Cheek 433% 20 2.8
Tail component in direction of spine 93 4% + 0.3
Tail component perpendicular to spine 146 6% 1.2
Maximum height 196 -0.3
TOTAL 3490% 160
AVERAGE 21.814
TOTAL ERROR 12.1 mm
Table 6. Measurements of (4302) using the figures taken from Table 4, and the assumed number of once as determined
by O'Brien, using the case width as measured from the cheek/bentside corner
Millimetres Error in millimetres
(using average oncia)
Spine 1860 85 + 4.5
Baseboard width 764% 35 + 0.2
Cheek 433% 20 '3.1
Tail component in direction of spine 93 4% + 0.2
Tail component perpendicular to spine 146 6% -1.4
Maximum height 196 -0.5
TOTAL 3492% 160
AVERAGE 21.830
TOTAL ERROR 9.9 mm
to confirm the attribution. his instrument so that a quarter of an oncia was the
However, do they represent two different historical chosen incremental accuracy, no error should be
oncia7. That they represent different figures for the greater than ? % of the oncia, i.e. more than 2.6 mm.43
same location would seem to imply there is no issue, Put another way, 1860 mm has an error of 3.3 mm
yet there is no reason why the figures could not assuming it is intended to represent 85 once at 21.844
represent different cities.42 To proceed in a logical mm each, but only 2.2 mm if it represents 85% once.
fashion it is necessary to reverse the process and The calculated average oncia, and thus the error in the
compare the errors. Table 5 assumes the baseboard other measurements, would be different; this needs
width at the front represents the intended dimension, to be taken into account. Whereas a smaller error
Table 6 assumes it to be the cheek/bentside corner may accrue by altering the interpretation, the idea
width. that a maker would deliberately design an instrument
It can be seen that although the calculated average to be 85% once long is somewhat unattractive. An
oncia is closer to published historical figures, alternative, concentrating on the measurements
both individual and total error discrepancies have taken by and for the present author, would be that the
increased, (in both of the tables) to methodologically cheek was intended to be 19% once rather than 20.
unsound amounts. Given an oncia of 20.844 mm, Tables 7 and 8 work on that assumption.
and accepting the principle that the maker designed On the basis of the above calculations it can be
42 Related to this is the question whether various figures from the same location that are close to each other actually
represent different standards or simply slight variations of the same standard. This cannot be fully addressed here.
However, using O'Briens list the figures for the Neapolitan once include 19.363, 19.813, 19.834, 19.835 and 19.973.
Almost certainly the three middle figures represent the same actual nominal length.
43 This is, of course, an issue of methodology rather than of instrument-making practice.
Table 7. Measurements of (4302) using the figures taken from Table 4, and the assumed number of once as determined
by O'Brien (with the exception of 19% rather than 20 once for the cheek), using the case width as measured from the
front of the case
Millimetres Error in millimetres
(using average oncia)
Spine 1860 85 + 2.9
Baseboard width 761% 35 2.9
Cheek 433% 19% + 2.0
Tail component in direction of spine 93 4% + 0.1
Tail component perpendicular to spine 146 6% -1.5
Maximum height 196 -0.6
TOTAL 3490% 159%
AVERAGE 21.848
TOTAL ERROR 10.0 mm
Table 8. Measurements of (4302) using the figures taken from Table 4, and the assumed number of once as determined
by O'Brien (with the exception of 19% rather than 20 once for the cheek), using the case width as measured at the cheek/
bentside corner
Millimetres Error in millimetres
(using average oncia)
Spine 1860 85 + 1.6
Baseboard width 764% 35 1.0
Cheek 433% 19% + 1.7
Tail component in direction of spine 93 4% + 0.1
Tail component perpendicular to spine 146 6% 1.6
Maximum height 196 -0.8
TOTAL 3492% 159%
AVERAGE 21.864
TOTAL ERROR 6.8 mm
seen that Table 8, in which the width is taken from This argument has been raised by previous
the cheek/bentside join, and the cheek length is authors. In particular, Denzil Wraight44 shows
assumed to be 19% once gives the best fit (even being convincingly that an unsigned clavichord ('Leipzig
a better fit than the figure found by Grant O'Brien) No. 2') can be expressed in both the Frankfurt inch
and can be said to represent (fairly closely) a quoted (as earlier proposed by Herbert Heyde45) and the
Neapolitan inch. Vicenzan inch.46 Whether or not the measurements
We wish to avoid being bogged down by further used by Heyde represent those designed by the
minutiae that would be of no advantage to the original maker is immaterial here, the point is that
questions posed in this article, so as the above tables on the basis of measurement alone the attribution
show that (4302) can be expressed in Neapolitan once of this clavichord to a particular location cannot be
using O'Brien's methodology, this approach will not substantiated.47 Wraight comments there is always
be pursued further here. Next, the second issue raised some other unit of measurement which can be found
above ?whether there is another interpretation that will fit the available millimetric dimensions
using the same methodology? will be considered. reasonably well.48
Table 9. Measurements of (4302) using the figures taken from Table 4, and assuming a spine length of 100 once. The
case width as measured from the front of the case
Millimetres Error in millimetres
(using average oncia)
Spine 1860 100 + 1.0
Baseboard width 761% 41 0.4
Cheek 433% 23% + 1.3
Tail component in direction of spine 93 + 0.1
Tail component perpendicular to spine 146 -2.7
Maximum height 196 10% + 0.8
TOTAL 3490% 187%
AVERAGE 18.590
TOTAL ERROR 6.3 mm
Table 10. Measurements of (4302) using the figures taken from Table 4, and assuming a spine length of 100 once. The
case width as measured from the cheek/bentside corner
Millimetres once Error in millimetres
(using average oncia)
Spine 1860 100 -0.3
Baseboard width 764% 41 + 1.5
Cheek 433% 23% + 1.0
Tail component in direction of spine 93
Tail component perpendicular to spine 146 -2.8
Maximum height 196 10% 0.7
TOTAL 3492% 187%
AVERAGE 18.603
TOTAL ERROR 6.3 mm
Table 11. Measurements of (4302) using the figures taken from Table 4, and assuming a spine length of 95 once. The
case width as measured from the front of the case
Millimetres Error in millimetres
(using average oncia)
Spine 1860 95 -0.2
Baseboard width 761% 39 1.9
Cheek 433% 22 2.7
Tail component in direction of spine 93 4%
Tail component perpendicular to spine 146 7% 0.9
Maximum height 196 10 + 0.2
TOTAL 3490% 178%
AVERAGE 19.581
TOTAL ERROR 5.9 mm
For the attribution of (4302) to Naples to stand This can be tested by assuming other inch values.
up, at least on the basis of the currently-discussed For the purposes of argument it has been assumed
methodology, it would need to be shown that the that the spine length will be an integral number
Neapolitan oncia is the only reasonable solution. of inches.49 Using the tables of values published by
49 Although O'Brien gives examples in which he assumes that the spine length could be half-inches (for example in
his analysis of the 1627 Bolcioni harpsichord, and his analysis of the Markiezenhof Museum harpsichord as shown in
Table 15), using integral inches is all that is required to prove the point here.
Table 12. Measurements of (4302) using the figures taken from Table 4} and assuming a spine length of 95 once. The
case width as measured from the cheek/bentside corner
Millimetres once Error in millimetres
(using average oncia)
Spine 1860 95 1.5
Baseboard width 764% 39 + 0.1
Cheek 433% 22 + 2.4
Tail component in direction of spine 93 4% 0.1
Tail component perpendicular to spine 146 7% -1.0
Maximum height 196 10 + 0.1
TOTAL 3492% 178%
AVERAGE 19.595
TOTAL ERROR 5.2 mm
Table 13. Measurements of (4302) using the figures taken from Table 4, and assuming a spine length of 90 once. The
case width as measured from the front of the case
Millimetres once Error in millimetres
(using average oncia)
Spine 1860 90 + 1.3
Baseboard width 761% 37 2.4
Cheek 433% 21 0.2
Tail component in direction of spine 93 4% + 0.1
Tail component perpendicular to spine 146 + 1.4
Maximum height 196 9% -0.2
TOTAL 3490% 169
AVERAGE 20.652
TOTAL ERROR 5.6 mm
Table 14. Measurements of (4302) using the figures taken from Table 4, and assuming a spine length of 90 once. The
case width as measured from the cheek/bentside corner
Millimetres once Error in millimetres
(using average oncia)
Spine 1860 90 0
Baseboard width 764% 37 -0.4
Cheek 433% 21 -0.5
Tail component in direction of spine 93 4%
Tail component perpendicular to spine 146 + 1.3
Maximum height 196 9% -0.3
TOTAL 3492% 169
AVERAGE 20.667
TOTAL ERROR 2.5 mm
Grant O'Brien50 the smallest Italian-inch value is using other units of measurement, with inch sizes of
approximately I8V2 mm, which would give 5" for the 18.590,18.603,19.581,19.595,20.652 and 20.667 mm.
tail component parallel to the spine. Comparison of these figures with Table 11 in O'Brien
As the above tables show, the methodology that (1999) it suggests that Rome (18.563 or 18.619) and
led O'Brien to conclude the instrument was made Genoa (20.674) are other locations which, strictly
in Naples can also be used to show it was made on the basis of the measurements and total error
50 Grant O'Brien, (1999). The tables in question appear on pages 164 - 171.
Table 15. Calculation of the local unit of measurement for the Markiezenhof Museum harpsichord, using the same six
component measurements as for the previous tables and O'Brien's assumed number of once
Millimetres once Error in millimetres
(using average oncia)
Spine 1989 90% -1.7
Baseboard width 747 34 0.9
Cheek 478 21% -0.4
Tail component in direction of spine 67 + 1.0
Tail component perpendicular to spine 186 8% -1.0
Maximum height 190 8% + 3.0
TOTAL 3657 166%
AVERAGE 21.997
TOTAL ERROR 8.0 mm
Table 16. The same figures as Table IS, but assuming a spine length of 96 rather than 90% once
Millimetres Error in millimetres
(using average oncia)
Spine 1989 96 -0.1
Baseboard width 747 36 + 1.1
Cheek 478 23 +1.5
Tail component in direction of spine 67 3% 0.3
Tail component perpendicular to spine 186 0.5
Maximum height 190 9% -1.7
TOTAL 3657 176%
AVERAGE 20.720
TOTAL ERROR 5.2 mm
are more likely.51 The net result of these calculations IS THE UNIT OF MEASUREMENT NECESSARILY
is that the methodology does not allow that Naples THE LOCAL INCH?
is the only possible place of manufacture of (4302). That (4302) can be described as closely matching
That Rome has been suggested on stylistic grounds the Genovese oncia figure (with a much smaller
by other researchers including Denzil Wraight,52 and error than any other result) presents another issue
that its oncia fits the measured dimensions better which must be considered, particularly in light of
than any of the Neapolitan figures, implies that it is another unsigned instrument which also (by the
the most likely place of manufacture. same methodology) has a Neapolitan attribution.
However, it would be unwise to make a Roman This harpsichord, formerly in the possession
attribution merely because it fits better than Naples. of Grant O'Brien (now in the collection of the
It can be shown that all the potential locations Markiezenhof Museum, the Netherlands) has a
give results which are within the accuracy that the calculated construction oncia of 21.995 mm,53
original maker achieved himself, since the difference essentially identical to a recorded Genovese oncia
in baseboard width between the front and cheek/ of 21.994 mm. If one is restricted to the same six
bentside join shows that anything less than 2Vi mm component measurements as used above a similar
produces, effectively, no error. figure results, but, like (4302) an alternative inch
51 The 19.595 mm figure does not closely relate to any unit of measurement, the nearest being Naples (19.363). Of
course, that the instrument was made in Naples using that published figure cannot be excluded from the possibilities.
52 Denzil Wraight, (1997: Part 11:327).
53 <www.claviantica.com/Neap_sch-files/Cen_GOB.htm>, accessed 26 March 2009. The determination of this unit
of measurement used more than just the six components described above and should, therefore, be more accurate,
assuming all of the added measurements actually represent the original design intentions.
Figure 4. 7b/? view o/?/ze octave spinet labelled Petrus Michael Orlandus anno 1710 (EUCHMI (4311)).
Photo - Copyright the Friends of St Cecilia's Hall
unit can also be determined.54 of the instrument' is more striking when compared
Table 16 gives a calculated figure which is closest with (4302) than any signed instrument. That both
to Rome (20.749), or a slightly large Genovese oncia harpsichords can be shown to fit the Genovese
(20.674). It should be noted that this measurement oncia better than any other unit might be used
for the Genovese oncia is different from the one as evidence to support a relationship between the
calcuated in Table 14.55 O'Brien argues against two instruments, but does not, as has been shown,
Genoa as a plausible location of manufacture, guarantee a Neapolitan origin.
stating 'Although Genoa also used a unit near to It is possible that the methodology discussed
that found here the fact that no seventeenth-century above accurately reflects the design intentions
harpsichord makers are known to have worked there of the makers, but that the unit of measurement
and the similarity of the instrument to those made was not the local inch. If it can be shown that the
in the Neapolitan tradition excludes Genoa as the measurement is totally unrelated to the place of
centre in which this instrument was made.'56 manufacture then the whole approach becomes
There are three important points here. First, questionable. This argument has been presented by
although there may be no known makers of the Birkett and Jurgenson58 who argue that the builder
period working in Genoa it cannot in itself rule out used a convenient workshop inch (or werkzoll) which
that location (a point made by Grant O'Brien (1999: might have no relationship whatsoever to the local
150) where an unsigned instrument is suggested unit of measurement. This will be discussed in
to have been made in Urbino), especially since it further detail below, but it is germane to give here an
is known that non-keyboard instruments (guitars example of a related unit of measurement being used
and violins) were made in Genoa, albeit perhaps in an instrument.
later. Second, as O'Brien argues was the case with An Italian octave spinet which also forms part of
the Raymond Russell Collection (EUCHMI (4311))
Giovanni Natale Boccalari,57 the unit used by the
maker might be associated to him by something has a modern inscription59 reading 'Petrus Michael
other that his place of work. Third, the 'similarity Orlandus anno 1710' on the back of the nameboard.
54 The raw measurements have been taken from O'Brien, ibid. As seen in his figures, the baseboard width varies
from 747 mm at the front to 742 mm at the bellyrail. Although both figures were tried it was found that the 747 mm
baseboard width gave a smaller error. Another inch of 19.122 mm (assuming a spine length of 104') also fits very well,
but has not been shown because it does not further the argument.
55 It is not entirely obvious whether this is purely coincidental or not. It is more likely to be coincidental that both
this instrument and (4302) can be described in units from Naples, Rome and Genoa.
56 O'Brien, ibid.
57 Grant O'Brien, (2009: 68).
58 Stephen Birkett and William Jurgenson, (2001) and (2002).
59 Probably by Mabel Dolmetsch at the time the instrument was restored in the Dolmetsch workshop.
With the exception of one other instrument (in for an attribution (at least of an approximate date
the Horniman Museum, London, also formerly in and location since identical roses can be found in
the possession of Arnold Dolmetsch), Orlandus is instruments from different maker's workshops that
otherwise unknown. As the instrument s inscription are contemporary in date and of similar location).
is not historical and does not give a location, the There is no Florentine soldo (= oncia) which is
baseboard measurement methodology might near the calculated 20.463 mm measurement, and
be a convenient starting point in an attempt to likewise no Florentine unit comfortably fits the
understand its design. baseboard measurements on the instrument.63
The resulting figure in Table 17 for the oncia Although this realization would seem to support
suggests, using O'Briens table, that the instrument the viewpoint of Birkett and Jurgenson that the
could be made in either Palermo (20.232) or Genoa builder could use a measurement that is totally
(20.674). That Genoa is a possible location with this unrelated to the local inch, a possible related-inch
instrument is perhaps less surprising in the light of explanation can be found. John K?ster, writing in
the two instruments mentioned above. this Journal,64 shows that in some parts of present
However, a recent re-examination of the day Spain there were two related measurements
mouldings by Denzil Wraight60 carried out at (i.e. the pulgado and dedo) which are both, in effect,
the request of the present author, shows that the inches. In the case of Valladolid, for example, 12
EUCHMI and Horniman Museum instruments with dedos is the same as 9 pulgadas, so that 1 dedo is
Orlandus labels can be attributed to the same maker. three-quarters of a pulgado. By using a similar
In the case of the Horniman Museum instrument approach to the Orlandus/Poggi (4311) octave spinet,
he has previously been identified as Poggi,61 a maker and assuming the calculated figure of 20.463 mm to
who can be shown to have worked in Florence. It be the Florentine equivalent of the dedo, it would
has also been shown by Sheridan Germann62 that give a figure of 27.284 mm for a 'pulgado*, which is
the rose in octave spinet is almost identical to one very close to the quoted 27.341 mm for a Florentine
in a virginal by Poggi, providing further support soldo.65
66 This assumes, of course, that the attribution to Poggi is actually correct. There has never been, to the present
authors knowledge, any question over the attribution of the Horniman Museum instrument to Poggi, and there is
nothing about the Edinburgh octave-spinet which is inconsistent with the attribution to Florence. It might well be the
case that Dolmetsch also made a connection between the two instruments, which would explain the false signature
on (4311).
67 Although the majority of keyboard-instrument catalogues quote the 'scale' as the string length of the note c", this is
not the note that was always historically used. In particular, Italian instruments with a C/E -/"'compass appear to have
laid-out the string band from notes rather than 'c'. This allows the top and bottom of the range to be set, rather than
finishing a half-octave from the top. Some instruments, for example the 1627 Stefano Bolcioni harpsichord (EUCHMI
(4304)), have both the f and c notes marked on the soundboard and register. There is no clear evidence on (4302) itself
about which notes were used, but the compass and the evidence presented in this article suggests it was from 'f. For
further information about the laying-out of Italian harpsichords see Denzil Wraight, (1997: Part 1:121-2 and 186-7).
68 The divided meantone accidentals in the/' -/" octave (but not the upper octave) means that it is impossible for all
of the notes within that range to be Pythagorean-scaled.
69 If the top two figures and the lowest (A) are excluded as possible outliers the average is essentially unchanged at
97.4 mm.
rear of wrestplank
I spine cheek|
front of wrestplank
front of baseboard
- 7x -M
Figure 5. Schematic of the front end of the instrument, illust
Table 20. Register-slot spaces of the 1627 Bolcioni The final action part to be designed is the layout
harpsichord, arranged in groups of four for the strings, both laterally (their position in
relation to each other and a datum line such as the
Notes 1-5 (bass) 50.6 29-33 53.3
spine), and lengthwise (the speaking lengths of the
5-9 51.4 33-37 53.5
strings and the position of the bridge in relation
9-13 52.3 37-41 53.0
to a datum line). These two datum lines provide
13-17 52.3 41-45 52.9 and y coordinates respectively for the string-band
17-21 52.9 45-49 53.2 arrangement.
21-25 52.1 49 - 53 (treble) 52.3 The spacing of the original bridge pins from
25-29 52.9 TOTAL 1 - 53 682.7 the spine is not very even, which suggests that
the register itself was used to mark the bridge-pin
Table 20 shows that the spacing in the bass is positions. In the first instance this would only be
closer than towards the treble, and is fully consistent done for the layout notes (probably all the 'f ' strings).
with Bolcioni starting in the bass and working in There is no clear internal evidence about how the
groups of four86 (probably marking check' points on register was used. Presumably, it was set so that one
the register with a pair of dividers), using a gradual of the register holes was a particular distance to a
increase of block thickness until the middle of the virtual spine, although one can only speculate which
register. There he found the distance between his one was used.88 It is, however, evident that the strings
desired and actual position was closing, and he then are not placed centrally between the spine and cheek
used more-or-less the same block width to the top but instead, the outer (long) bass string is closer to
of the register.87 For the 57-note register of (4302), the spine than the outer (short) treble string is to the
the maker would be required to divide his proposed cheek. The initial layout process must involve both
25 soldi register width into 56 spaces, which can be the register (if it was that) and a long marking-out
very simply done with check points every four, seven stick to determine the bridge position in relation to
or eight notes. the front datum line. This could be either a standard
Assuming the outer edges of the lowest and ruler or a graduated stick.
highest keylevers were intended to be parallel, simple In determining the original string lengths and
geometry shows that the keyboard width must be layout, and in the absence of any evidence of the
greater than the register width. Returning to the original nut, there must be a starting hypothesis. In
dimensions of (4302), to a register width of c. 685 this case, it is that if, assuming a straight nut (in the
mm needs to be added the thickness of the jackslot treble at least), the layout notes can be Pythagorean
(c. 4.5 mm) and some extra clearance at each end (to scaled, it is likely to reflect the original intentions.
ensure the registers can safely slide between their As discussed above, the notes f, f" and can be
86 Although the 52 spaces can be divided into groups of four evenly, and the measurements as recorded by O'Brien
are consistent with Bolcioni using groups of four, there is no reason why it may not have been a different grouping
(perhaps five, six or eight).
87 Aside from the evidence shown in the table, it is very much simpler to assume Bolcioni intended the whole register
width to be 25 soldi rather than to assume he desired 50 spaces to be 24 soldi. Although 50 spaces (51 notes) corresponds
to a C/E - c'" compass, with divided D/F#,E/G# and four higher enharmonically divided notes (see Denzil Wraight,
'Checklist of Italian harpsichords and virginals with split sharps' <www.denzilwraight.com/ChecklistSplitSharps.rtf>),
the additional two bass notes would, particularly if they were side-by-side as proposed by O'Brien, create layout issues
which would make it more sensible to design from scratch, as implied which the numerous scribed lined and points
elsewhere on the instrument.
88 One possibility is that the left (bass) side of the F register slot was set at 2 soldi from the bass end, and that the
register had prick-marks on it (like the 1627 Bolcioni harpsichord) to guide the long string position. Other possibilities
include that the register was set to position the long/string SV2 soldi from the bass end, or set to position the long/"
string 21 soldi from the bass end.
89 Those figures give a calculated maximum error of less than 0.6 mm, which is clearly much more accurate than the
original maker could have worked to. In practice the actual 'Pythagorean-scaled' range might be larger.
90 Grant O'Brien, (2000:177).
91 The possibility that (on both instruments) a straight nut was used in the layout process, but a two-section nut
actually used in the finished instrument cannot be excluded. This would only have a minute consequence for the string
lengths in the bass, and this difference would be tonally insignificant.
92 The distance between the front of the baseboard and the proximal side of the wrestplank is VA mm greater in the
bass than the treble.
93 Strictly speaking it is possible that the front edge of the baseboard was the datum line as, in theory it is 4 soldi
from the proximal side of the wrestplank. The inaccuracies of construction may deceive the interpretation's validity.
position of / pin in relation to the spine position of f" pin in relation to the spine
Figure 6. Method of laying-out the nut position from the datum line (the proximal side of the wrestplank for (4302).
described, half the length of/". Florentine soldi) from the straight-nut line, which
The generation of the bridge position for the itself was derived by the datum line at the proximal
F string must be somewhat conjectural, but a side of the wrestplank.
plausible solution can be proposed. Denzil Wraight
has suggested94 that some Venetian instruments THE BASEBOARD GEOMETRY AND LAYOUT
had their lengths derived along the F string, rather That the baseboard dimensions as a whole do
that the spine, this length (from the front of the not conform to the 27.341 mm measurement97
case or proximal side of the wrestplank to the tail (regardless of fitting the 'small soldo' of the 27.560
along the line of the / string) being eight times the Florentine measurement) suggests that it was laid-out
length of the f" string length. Although this method using another method, possibly as a consequence of
would ?if the maker started his design from this needing to be built around the action and stringband.
length which was then used to generate the other For this to be viable it should fit a series of 'rules'
string lengths ? probably be within acceptable limits as suggested by John Koster,98 in particular that it
(as he demonstrates), it would not be plausible if the should be repeatable on other instruments. One of
design method was as described above. However, it is the problems with examining instruments that are
possible that a variation on this method was used,95 in the sole-known example of a particular maker's
which eight times the length off" (8x7% = 62 soldi) work is being able to show that the details are due
is, to within a millimetre or so, the distance from to design rather than chance. Despite the apparent
the straight-nut line to the tail along the F string.96 excellent fit of many of the figures presented earlier
In this way, all of the layout notes can be shown in this paper there can only be one (if any) which
to have been generated (and can be described in can, by definition, be correct. It would, however,
94 Denzil Wraight, Construction Principle in Venetian harpsichords', paper given at the Edinburgh Early Keyboard
Instrument Symposium, St Cecilia's Hall, Edinburgh, 25 October 2008, available in printed version from his website
<www.denzilwraight.com/CPinVH002.pdf>.
95 Although there is precedence in Venetian instruments, the present author is not aware of any other Florentine
instrument which follows any variation of the same method.
96 This is not, of course, the actual string length, which must be shorter (even assuming a straight nut-line was used)
by the distance from the tail to the bridge along this line. This is discussed below.
97 Or, alternatively, that the internal dimensions do not coincide well with any of the prospective baseboard inch
unit measurements.
98 John K?ster, (1998: 22).
99 It is also only a single error in practice, assuming that the distance was set on a pair of dividers rather than being
marked-off on a ruler.
100 This might also, in fact, be an argument in favour of the baseboard being subject to shrinkage across the grain as
discussed above, rather than representing any error.
101 The length from the datum line is 1860 - 108% = 1751% mm, which, divided by 27.341, gives 64.06 soldi.
102 The two possibilities are not entirely mutually exclusive as two points are needed to construct a tail line.
103 Stephen Birkett and William Jurgenson, (2001).
104 John K?ster, (2008).
105 Stephen Birkett and William Jurgenson, (2001).
106 And, by extension, if he reduced the keyboard width to 25 soldi he might well have made a similar reduction (by
a ha\?-sold?) to the case-front width - in this example to 27% soldi.
107 The letter designation used in this paper will be the same for all instruments as a convenience. It does mean
that not all letters will be used in each example. This alternative seems better than having, for example, the letter 'E'
signifying the perpendicular tail component point on the spine on one example, and the actual tail point on another.
108 Two possibilities - an 'F-line' length being 8 xf", which would give one point on the tail triangle 'hypotenuse', or a
front datum line to tail point of 64 soldi, to give point {G} have already been mentioned, and it would even be possible
to join those lines together so that the tail was defined by this line until it reached a line perpendicular to the spine
from point {E}.
109 The consistency (in all the instruments discussed here) with which the length of EF (F being the bentside/tail
corner) or GF can be described as a simple fraction of either the case-front (AB) or cheek length (BC) suggests the
measurement was deliberately designed.
110 The phrase 'swinging an arc' is almost certainly a misrepresentation of the maker's actual working methods.
Although it is possible builders did, on occasion, actually rotate a trammel or graduated stick, the actual method
G K Gr^
C /^1C
-^ ^) I -1 2s 4s
A- A1-1
Figure 7. Method layout drawings for (4302). The illustration on the left shows the derivation of the straight nut line, and
the string lengths of the layout notes (which gives point {!<}). The illustration on the right shows the baseboard layout.
no (continued) wou\? probably have been simply linear. For example, once the length HC had been marked on a stick (be
it a trammel, graduated ruler or a mark on a piece of wood - the principle is identical), one end would be placed at {H}
from which point {D} was noted, and then the stick was moved up so that this point {D} became the new end point to find
point {E}. No physical rotation' is actually required.
111 The position of point {E} would actually be 12 mm closer to the front of the instrument using AC, a large enough
difference to make it impossible to argue that the method cannot allow a differentiation between the two choices. The
overall case length would be longer as point {K} is in the same place, and the line {FK} is a more acute angle to the spine.
112 Some of the possibilities include that the distance EG is 2% soldi, that EG is VA times the register gap width, that
it is in an 11:4 ratio with EF, or (most likely of these options) a 3:1 ratio of FG:EG. Other possibilities derived from
the internal layout (as proposed above for (4302) might also exist but have not been considered and are not needed to
prove the point.
113 Other than the relationship between the two instruments suggested by O'Brien (see above), no evidence is
presented here to show that the Markiezenhof harpsichord is Florentine. Although the instrument has been seen by
the present author it has not been measured or otherwise examined in detail, and all the measurements have been
taken from the published work of Grant O'Brien.
114 The measurements from which this analysis of the 1631 Bolcioni has been derived have been taken from Grant
O'Brien, (1999). The measurements of the 1631 Bolcioni have been taken by the author (in collaboration with others)
in a similar manner to that reported for (4302) above. It should be noted that the present author agrees with O'Brien's
arguments for the compass and his expectation that the keywell was the same depth as the 1631 instrument at Yale.
Likewise, the author agrees with O'Brien's evidence for the bridge line. However, the present author's examination of
the keywell shows that the only line found on the instrument itself which could indicate the front of the wrestplank is
122 mm from the front edge of the present baseboard, suggesting the instrument has not been shortened in length as
suggested by O'Brien. The analysis presented here has been tested on the original instrument.
115 It is, of course, quite possible that this step was also taken in the 1631 Bolcioni at Yale, but it has not been included
in the analysis as it has no bearing on the later processes.
116 It is possible that rather than thinking in terms of 8 soldi, Bolcioni intended the distance to be 2/7AB. The actual
dimensions (EF = 221 mm, AB = 759 mm at the case front) are well within an acceptable accuracy, particularly if
dividers were used. This would be consistent with all of the other instruments described in this article which have
either EF or GF as a simple fraction of AB or BC.
117 Both of the Bolcioni harpsichords could, in principle (and with a different tail-triangle derivation), have used the
distance EG rather than EF designed as a specific number of soldi. However, doing so would not take into account the
string layout. By designating EF as a particular distance, Bolcioni would be using a variation on the same approach as
used by the makers of (4302) and the Markiezenhof Museum harpsichord.
118 A set-square can be positioned perpendicular to the spine on which the distance 34 A has been marked.
CONCLUSIONS
The ABOU
comparat
METHOD DESCRIBED
Florentine A
vir
Four instruments
of are
harpsichorn
Florentine Museum
harpsichords inst
fr
century ? then
but it is a reason
enough o
conclusions. The
the number
surviving
can bedemonstrated
described toas h
ha
during the The
first method
half of t
is quite nothing
small; mor
combining
of Boalch119 with
compass, updated
wher
Wraight120 more thanfewe
produces a gr
It is reasonable
on a to assum
number o
instruments mostwill also be
importan
of whethersame(4302) and
maker. th
A
harpsichords are
(exceptamongst
{D}121 w
that the numbers
exist on would
the ins
APPENDIX
BIBLIOGRAPHY
Barnes (1971): John Barnes, 'The specious uniformity of Italian harpsichords', Keyboard Instruments:
studies in keyboard organology, 1500-1800, ed. Edwin Ripin (Edinburgh, 1971).
Barnes (1975): John Barnes, Italian Harpsichord c. 1600 Russell Collection No. 2., unpublished
restoration report, written 1975 (EUCHMI archives).
Barnes (1980): John Barnes, 'Does Restoration Destroy Evidence', Early Music, 8/2 (April 1980).
Birkett and Jurgenson (2001): Stephen Birkett and William Jurgenson, 'Geometric Methods in Stringed Keyboard
Instrument Design', GS/LIV (2001).
Birkett and Jurgenson (2002): Stephen Birkett and William Jurgenson, 'Why Didn't Historical Makers Need
Drawings? Part II - Modular Dimensions and the Builder's WerkzolV, GSJ, LV (2002).
Boaich (1995): Donald Boalch, Makers of the Harpsichord and Clavichord 1440-1840, third edition,
ed. Charles Mould (Oxford, 1995).
Germann (2002): Sheridan Germann, 'Harpsichord Decoration - a Conspectus', The Historical
Harpsichord, Volume 4, ed. Howard Schott (2002).
Heyde (1986): Herbert Heyde, Musikinstrumentenbau (Wiesbaden, 1986).
Koster (1998): John K?ster, 'Towards the Reconstruction of the Ruckers Geometrical Methods',
Kielinstrumente aus der Werkstatt Ruckers, ed. Christiane Rieche (Halle, 1998).
Koster (2008): John Koster, 'Traditional Iberian Harpsichord Making in its European Context', GSJ
LXI (2008).
Newman and Williams (1968): Sidney Newman and Peter Williams, The Russell Collection of early Keyboard
instruments (Edinburgh, 1968).
O'Brien (1999): Grant O'Brien 'The use of simple geometry and the local unit of measurement in
the design of Italian stringed keyboard instruments: an aid to attribution and to
organological analysis', GSJI?I (1999).
O'Brien (2000): Grant O'Brien, 'Towards establishing the original state of the three-manual
harpsichord by Stefano Bolcioni, Florence, 1627, in the Russell Collection of Early
Keyboard Instruments, Edinburgh', GS/LUI (2000)
O'Brien (2009): Grant O'Brien, "The Single-Manual Italian Harpsichord in the Royal College of Music,
London, Cat. No. 175: An Organological Analysis', GS/LX11 (2009).
Russell (1959): Raymond Russell, The Harpsichord and Clavichord (London, Faber and Faber, 1959).
Smith (2008): Eleanor Smith, A Discussion of the Use of Divided-Accidental keys in Italian Strung
Keyboard Instruments pre 1700, MMus thesis, University of Edinburgh, 2008.
Wraight (1992): Denzil Wraight, "The Identification and Authentication of Italian String Keyboard
Instruments', in The Historical Harpsichord Volume 3, ed. Howard Schott,
(Stuyvesant, NY, 1992).
Wraight (1997): Denzil Wraight, The stringing of Italian keyboard instruments c. 1500 - c. 1650, PhD
thesis, Queen's University of Belfast, 1997.
134 As with the 1627 Bolcioni harpsichord, the point {E} is not actually found on the instrument, and does not need
to be. The triangle can be positioned perpendicular to the spine using a graduated square on which the distance %AB
has been marked.
DARRYL MARTIN