'Handel's Temperament' revisited 

Colin Pykett

“One must take refuge in temperament when tuning claviers and harps.  Many books make as much to do of this as if the welfare of the entire world depended on a single clavier.”

Johann Mattheson [12]

Posted: 8 August  2016

Revised: 15 September 2019

Copyright © C E Pykett 2016-2019

Update notice: this article describes an earlier realisation of Handel's Temperament which matches the tuning instructions of 1780 reasonably well.  Since then a more recent version has been developed which matches them exactly.  This version is outlined in an article published in the September 2019 issue of Organists' Review.  However for reasons of space and the anticipated audience, that article omitted some technical details of the temperament (e.g. numerical data relating to the circle of fifths) which might interest those wishing to know more.  Should you be interested in these details, they are included in a further article on this website entitled More on Handel's temperament.

Abstract.   In about 1780 a description was published in London of what has become known as 'Handel's Tuning' or 'Handel's Temperament'  Although the association with Handel has little foundation, this article suggests that it is worth revisiting the temperament attributed to him for several other reasons.  One is the unfortunate fact that many modern realisations of it are wrong because they are incompatible with the complete set of tuning instructions given.  These illuminate tuning practices at a time when beat counting was uncommon and, while therefore vague on detail, they nevertheless incorporate an explicit set of overarching constraints which place definite bounds on the possible outcomes.  In particular, all fifths must be flat rather than pure or sharpened, and all thirds must be considerably sharp.  It is therefore regrettable that the modern realisations of this temperament which are embodied in electronic tuning devices are mostly wrong because they incorporate pure fifths.  It is also impossible to conclude, as some authors have done, that the temperament was just another variation on the meantone tunings which were common at that time in Britain.  On the contrary, it is difficult to see how the temperament can be other than a mildly unequal one.  Consequently it probably lies on the haphazard path which ultimately led to equal temperament in the nineteenth century rather than being merely another example of an already-outmoded meantone approach, although the temperament is not equal because the tuning instructions also dictate that the fifths should differ in their deviations from pure. 

The frustratingly unfocused tuning instructions nevertheless allow for several subtle variations in realising a version of the temperament, and therefore I took advantage of this to produce my own which is described in the article.  It addresses the tuning instructions at a level of detail not found elsewhere, providing a reading which I therefore believe to be novel.  The outcome is a pleasing well temperament with a range of subtle key colours which can render music in any key.  Because of the imprecision of the tuning instructions it is not possible to interpret them in a single, unequivocal fashion - the best that can be done is to come up with one of several possible variations on the theme of 'mildly unequal'.   That described here is represented as a set of deviations from equal temperament for each note in the scale, a format which is well suited to most of today's electronic tuning devices and apps, thus it should be reasonably simple to evaluate in practice.

Contents

(click on the headings below to access the desired section)

Introduction

The original tuning instructions

Tuning the temperament today

Note offsets

Circle of fifths

What does it sound like?

Concluding remarks

Acknowledgement

Notes and References

Introduction

In about 1780 a description was published in London, England of what has become known as 'Handel's Tuning' or 'Handel's Temperament' [1]. It is a means of tuning (a method for setting a temperament) on instruments such as the organ and harpsichord, and it was said to have been invented, or at least used, by Handel himself. The association with Handel is probably spurious ([2], [7]) and as far as this article is concerned it is best ignored, to be pigeonholed alongside similar claims elsewhere for the temperaments allegedly employed by J S Bach. Such unfalsifiable hypotheses, with their inherent encouragement of eternal dispute, do not find a natural home with most scholars.

Nevertheless, Handel or no, the fact remains that we have an 18th century temperament here, probably English, for which full tuning instructions are given, and this is of intrinsic importance. The development of temperament in Britain followed a different path to that in continental Europe [3] and, being well documented, this one represents an interesting evolutionary waypoint. The instructions are particularly useful in an historical sense because they can help to illuminate tuning practices if nothing else. For instance it is not always clear whether tuners were counting beats at (and before) this time, though the alternative methods employed are seldom described. The detailed instructions in this case make no reference to beats as we shall see, and it is therefore useful for modern tuners to note how things were sometimes done then, accustomed as they are to having electronics do accurate beat-counting for them today.

Another reason for revisiting this temperament is the unfortunate fact that many realisations of it are just plain wrong. This seems to have arisen partly because of a fairly recent and perhaps over-enthusiastic interpretation by Jorgensen of the tuning instructions in which two pure (perfectly-tuned) fifths arise [4]. Even a superficial reading of the instructions shows that all fifths must be tuned flat [5]. Jorgensen also seemed to deviate from the instructions in other ways which are difficult to explain. Yet it is his version of this temperament which surfaces in most, if not all, electronic tuning devices and apps today [6]. Clearly this is to be deprecated, and people need to be aware of it.  Aside from this and one or two other passing references, little attention seems to have been paid to the temperament.  A paper by Johnson [7] concerned itself mainly in distancing the temperament from any Handelian connection rather than examining the tuning instructions in detail.  Although she seemingly tested and evaluated this method of tuning, her discussion of it is little more than a summary of the original instructions.  She also quoted the opinion of an instrument maker and tuner whose interpretation of the instructions apparently resulted in a temperament close to one-sixth comma meantone.  This is an inexplicable conclusion because the latter temperament has a grossly sharpened 'Wolf' fifth which, like Jorgensen's realisation with its pure fifths, is explicitly disallowed by the instructions which make clear that all fifths must be flat.  One therefore has little option but to dismiss both the Jorgensen and Johnson versions of the temperament.

Disregarding major errors such as these, the somewhat vague tuning instructions nevertheless allow for many subtle variations in realising a version of the temperament, and therefore I took advantage of this to produce my own which is now described here.  It was developed independently of Johnson's research since I was unable to read her paper until after I had completed my version.  Moreover, I am unaware of any other attempt to address and interpret the tuning instructions at the level of detail to be found in this article since Jorgensen published his somewhat perplexing deductions in 1991 [4], and I believe this to be novel.

The original tuning instructions

The tuning instructions from 1780 [1] are entirely non-numerical and qualitative in the sense that beat counting is not used, indeed beats or anything remotely resembling them are not mentioned at all. Instead there are vague references to interval 'quality' or purity such as "let the Fifth be nearer perfect than the last tho' not quite". It is therefore next to impossible to produce a unique, unequivocal version of this temperament because different tuners will arrive at different results, and the same tuner will have difficulty reproducing identical results on each occasion if tuning by ear alone.

It is assumed that the intention was to produce a 'well' temperament, one in which all keys can be used without restriction. The music of the late Baroque and classical periods would have virtually demanded it. This does not mean, of course, that equal temperament must be the output of the tuning process. However an interesting statement is casually dropped into the mix more than halfway through the exercise. It says that "all Thirds must be tuned sharp more or less, as all fifths should be flat". This is important because, by disallowing any pure fifths or thirds, it makes one wonder whether the intention was indeed to create something close to equal temperament (ET). In ET all fifths are flat and the thirds are sharp. This impression is reinforced by the even more definite statement elsewhere that "NB. the Fifth will not bear to be reduced so much below its true accord, as the Third will to be raised above it". Presumably this means that the thirds must be made significantly sharper (in the sense of their fractional deviations from pure intervals) than the fifths are made flat. In ET all fifths are tuned 1.96 cents flat from pure whereas the thirds are 13.66 cents sharp, a considerable difference, so perhaps something similar was intended here.  (One cent is one hundredth of a semitone).  It should also be noted that the tuning instructions make the temperament deliberately difficult and time consuming to tune, because none of the tuned intervals is pure (it is of course easier and quicker to tune an interval pure because it exhibits no beats). Therefore in this awkward practical respect the temperament is also similar to ET, and the designer of the temperament must have had good reasons for choosing this option.  Nevertheless, it is unlikely that the strict arithmetical uniformity of ET was intended because the tuning instructions ask for some fifths to be nearer to pure than others.  One such statement (already quoted above) is "let [this] Fifth be nearer perfect than the last".  I therefore concluded that whilst the tuning process should result in a well temperament, there should be sufficient aural interest to differentiate it from an exact realisation of ET.  In other words I was looking to create a mildly unequal temperament while satisfying the constraints of the tuning instructions.

This does not merely reflect some nonchalant musing on my part because it is difficult to see how the outcome can be other than mildly unequal, a conclusion also reached by some others (e.g. [4], [11]).  Although the tuning instructions are vague at a detailed level, they nevertheless lay down a number of unequivocal overarching constraints which have been summarised above, and together these limit one's room for manoeuvre when one tries to implement the temperament for real.  Albeit somewhat disguised at first sight, the instructions simply tune the twelve fifths and fourths in a single octave in a more or less conventional manner, and the fifths must all be flat, not pure or sharp.  Therefore one has to be careful not to flatten any of them too much otherwise it will be impossible to compensate elsewhere to prevent the sum of the deviations from pure exceeding the Pythagorean Comma.  This problem does not affect strongly unequal temperaments because any number of fifths can be made pure or sharpened to compensate for the flattened ones.  Taking an extreme example, in sixth-comma meantone (the so-called 'Silbermann' temperament) this leads to the single grossly sharpened Wolf interval.  But because the quite explicit constraints of the 'Handel' temperament rule this out, it is difficult to understand how anyone could suggest that it resembles a meantone tuning, as Johnson did in reference [7] for instance.  I found that trying to tune the temperament was a salutary, if frustrating and time-consuming, exercise in ferreting out what the instructions were actually saying, rather than making baseless assumptions as to what one might like them to say.

Given this, one can go further.  Although variants of meantone tuning were common in England during the 18th century and they persisted into the 19th, there was also some pressure to adopt less aggressive tunings which might have been imported from continental Europe, and eventually they led to the comparatively late adoption of equal temperament in Britain.  It is therefore my contention that the so-called 'Handel' temperament was an evolutionary step along this newer road, rather than yet another example of an already-outmoded meantone approach.

Tuning the temperament today

The temperament was set up using a digital organ for convenience, using a mild-toned 8 foot Principal stop not unlike a Dulciana.  Such stops would have been common in England in the 18th century, and I considered it important to use one whose harmonic spectrum would have been similar to those used to tune an organ to this temperament.  Tuning, and therefore setting any temperament, is a matter of detecting beats.  Whether one counts or times them or not is largely immaterial - the fact remains that they exist except when an interval reaches purity and this is what a tuner listens for, consciously or otherwise, when tuning pure or setting a deviation from pure.  It follows that the harmonic spectrum of the stop used for setting a temperament is important because the beats arise exclusively from harmonics above the fundamental frequency except when tuning unisons and octaves.  For example, the strongest and lowest-frequency beat of a fifth arises between the third harmonic of the lower note and the second harmonic of the upper.  Other beats at integer multiples of this frequency also exist if there are sufficient audible harmonics, but they are of lower amplitudes because the generating harmonics themselves become weaker in flue pipe spectra as their frequencies increase.  In the case of a fourth the predominant beat arises from the fourth harmonic of the lower note and the third of the upper, and for a major third it involves the fifth harmonic of the lower note and the fourth of the upper.  Therefore the stop chosen for tuning purposes must allow these harmonics to be heard, yet there should not be too many others which might confuse the ear if too many higher-order beats are generated.  For similar reasons one should desirably choose a stop whose timbre (harmonic spectrum) relates to the epoch in which the temperament of interest arose because the temperament designer would have had in mind the tone colour, and thus the beat-generating properties, of the pipes which would be used to tune the instrument.  Hence my choice of a mild Principal tone.

The tuning instructions were presented in the original document as a set of eight major triads written as broken chords together with various instances of octaves (which are of course tuned pure). Each triad has its own set of instructions and we shall now consider them in turn.  They were printed using the Georgian 'long s' typeface which looked similar to an 'f' but here the modern letterform has been substituted.  The tuning for each note in the octave which is derived from the instructions is represented here as a deviation in cents from its ET counterpart. This was thought to be the most convenient format for today's tuners, who will most likely wish to use the offsets in some form of electronic tuning device. Because of the vagueness of the instructions however, the data are represented to a precision of only one cent with no decimal places, as a greater precision was not warranted.

First chord (C major): 

The original instructions state:

"In this Chord tune the Fifth pretty flat and the Third considerably too sharp. NB. the Fifth will not bear to be reduced so much below its true accord, as the Third will to be raised above it"

One interpretation of these instructions is as follows:

The fundamental frequency (pitch) of C was chosen as 261.63 Hz. This arises in ET if the A above is tuned to today's usual standard of 440.00 Hz and it is therefore a choice having practical advantages. However any other figure can be chosen. Proceeding with this choice, the offset of C from the usual ET scale is therefore defined as 0 cents.

The term "pretty flat" relating to the interval C-G is to all intents and purposes meaningless. In view of the probable inclination of the whole temperament towards one which is well-tempered, or even something akin to ET as discussed above, G was given its ET frequency value, resulting in an offset from ET of 0 cents.

For the same reason E was also given the value it would assume in ET. This choice certainly satisfies the requirement for E to be "considerably too sharp". Thus the offset of E from the usual ET scale is also 0 cents.

Summarising, the frequency offsets in cents from ET are C (0), E (0), G (0).

Second chord (G major):

"Let the Fifth be nearer perfect than the last tho' not quite, tune the Third a Fifth to E, make it good but just bearing flat".

This implies that the interval G-D should be given the merest nudge flat from pure. Tuning D one cent sharp relative to its ET value results in a beat frequency of 0.66 Hz flat from pure for this interval, or just over one beat every two seconds. In ET the beat would be twice as fast at 1.34 Hz. Therefore the instructions appear to have been satisfied in a qualitative sense. (Note that these numerical values for beat frequencies assume that middle C was first tuned to 261.63 Hz as stated above. Other choices will give different results).

B now has to be tuned as a flattened fifth against middle E which was set for the first chord above. The instructions seem more or less the same as for G-D, so the same procedure was adopted by sharpening B by one cent from its ET value.

Summarising, the frequency offsets in cents from ET are G (0), B (+1), D (+1).

Tune the octaves:

"Tune all Octaves perfect".

This instruction does not call for further discussion beyond observing that it tunes three notes which are used in subsequent steps.

Third chord (D major):

"Tune A, a good fifth to D, trying it at the same time with E, above already tuned. Tune the Third a fifth to B, and let it be near as flat as the fifth in the 1st Chord, this will in some measure bring down the sharpness of the Third".

Middle D was tuned from its octave in the previous step, and using it the interval D-A should then be tuned as a "good fifth". As with so much else in these instructions it is not at all clear what is meant by the adjective "good". For example, all fifths in ET are tuned flat by almost two cents - would these have been considered "good" in the 18th century? We know that the interval cannot be pure because we have noted several times already that perfectly tuned fifths are disallowed in this temperament. Therefore, as with beauty, 'goodness' must depend as much on the ears of the beholder as anything else. Here I sharpened A by one cent above its ET value to bring its interval with D closer to a pure fifth. (Note that this means the frequency of A will no longer be exactly 440 Hz given the frequency already chosen for C. However it can be easily adjusted back to this value when the entire temperament has been set up, and this will be described later). This choice endows the interval D-A with a beat frequency of 1 Hz, hopefully making it indeed something which could be considered a "good fifth".

The instructions then tell us that middle A just tuned now has to be tested as a fifth against the (treble) E which was set in the previous step. Since no criteria of goodness are given it is again impossible to be sure what is intended here. E retains its ET frequency value as a result of earlier decisions, causing it to beat at 2.24 Hz with the newly-tuned A. This is more than twice as fast as the beat with middle D (see previous paragraph). Maybe the instructions are implying at this point that one should aim to bring these beat frequencies (i.e. of the intervals middle D-A and A-treble E) closer to equality. Therefore this is an opportunity to experiment with my interpretation of the original tuning instructions given in this article. Attempting to equalise the beat frequencies requires that A be flattened from the +1 cent offset from ET just given to it. Removing the offset results in the intervals D-A and A-E having almost identical beat frequencies of 1.50 and 1.48 Hz respectively. So the instruction to try A against treble E, having first tuned it against middle D, boils down to a choice between leaving A unchanged or sharpening it slightly to make a better fifth with middle D. I chose the latter course, partly to give my reading of this temperament a little more aural interest rather than allowing it to drift yet closer to ET.

It is stated next that the interval B (below middle C) to F# should be tuned much as C-G was previously in terms of being "near as flat", though the meaning of this is obscure. In this reading of the temperament, B has already been given an offset of +1 cent from ET. Therefore F# is tuned with no offset, which has the effect of slightly narrowing the fifth B-F# compared both to ET and to C-G.

Summarising, the frequency offsets in cents from ET are D (+1), F# (0), A (+1).

Fourth chord (A Major):

"Tune the Third a fifth to F, sharp [sic - presumably this means 'F sharp'] already tuned, let its bearing be the same as the Third in the last Chord".

On the face of it this implies that the interval F#-C# should be narrowed by one cent from ET in the same way that B-F# was narrowed in the previous step. Therefore C# is tuned with an offset of -1 cent from ET.

Middle C# is next derived from its octave just tuned as above.

Summarising, the frequency offsets in cents from ET are A (+1), C# (-1), E (0).


Fifth chord (E major): 

"Tune the Third very fine rather bearing sharp, for in fact all Thirds must be tuned sharp more or less, as all fifths should be flat".

This instruction, relating to G# because E and B have been tuned already, is perhaps the most obscure of all. It certainly gives the most leeway for tuners to impose their signature on their version of the temperament. At one extreme it could be taken to imply that the interval E-G# should be tuned just sharp of pure, which implies an offset for G# from ET of anything up to a gross -13 cents or so. After some experiments I concluded that anything beyond an offset of -6 cents made little sense for a well temperament, being too extreme for my ears. Eventually I settled on a range of acceptable offsets lying between -1 and -4 cents. The value adopted here is -1 cent, in keeping with the presumed well-tempered ethos of the temperament as a whole.

Summarising, the frequency offsets in cents from ET are E (0), G# (-1), B (+1).


Sixth chord (F major):

(This chord was thrown onto a C-clef in the original document but the F-clef has been used here as it is probably more familiar).

"Tune the A, an Octave to A above, when you have drawn the F, to a perfect fifth with C, give it a little inclination higher, that being the same as if you had tuned your C, a fifth to F, giving the C a little flatness".

Thus tenor A (the A below middle C) is first tuned to its octave which has been tuned already, as has middle C. Then, this admittedly rather confusing instruction is simply saying that the interval tenor F-middle C should be narrowed slightly from pure in the same way as all the others. I gave F an offset of -1 cent relative to ET, making the interval slightly wider and closer to pure than in ET.

Middle F is then tuned pure from tenor F.

Summarising, the frequency offsets in cents from ET are F (-1), A (+1), C (0).


Seventh chord (B flat major): 

"Tune the fifth as in the 6th Chord".

B flat was given an offset of -2 cents from ET. This makes the interval with F the same as in the sixth chord (slightly wider and closer to pure than in ET), which is what the instructions call for.

This note is then used to tune the B flat above.

Summarising, the frequency offsets in cents from ET are B flat (-2), D (+1), F (-1).


Eighth chord (E flat major):

"Tune the fifth as in the last chord".

The only note to be tuned here is E flat, to which I gave an offset of -3 cents. Therefore the interval with B flat became the same as in the previous case for the seventh chord (slightly wider and closer to pure than in ET), which is what the instructions call for. 

Summarising, the frequency offsets in cents from ET are E flat (-3), G (0), B flat (-2).

Note offsets

The frequency offsets in cents from equal temperament for each note which resulted from the tuning procedure above are summarised in the tables below for convenience.  These can be inserted directly into most electronic tuning devices or apps.  Two sets of data are given.  The first (Table 1) simply pulls together the numbers from the foregoing section.  It will be recalled that these were referenced to middle C with a frequency of 261.63 Hz.  Since A was given an offset of +1 cent, its frequency is 440.25 Hz.

Table 1 . Note offsets from equal temperament in cents (middle C = 261.63 Hz; A = 440.25 Hz)

However it might be preferred to reference the data to A = 440.00 Hz.  As this note has an offset of +1 cent in Table 1 this now has to be removed, and this means that all other offsets have to be reduced by the same amount.  This is done in Table 2.  The frequency of middle C is reduced accordingly.

Table 2 . Note offsets from equal temperament in cents (middle C = 261.48 Hz; A = 440.00 Hz)

Circle of fifths

The note offsets above are all that are needed to tune using most electronic devices.  However tuning by ear in the traditional way requires the information to be presented in a different form.  Specifically, temperaments are set by 'laying the bearings' in a single octave of the key compass using a sequence of twelve fifths, and those which stray beyond the octave boundaries are inverted to become fourths.  Each interval is tuned by adjusting its beat frequency to match a prescribed value.  It is conventional to think of the twelve fifths which have to be tuned in setting a temperament as forming a circle, a figure which closes on itself.  This is because the sum of all twelve deviations from pure intervals must equal -23.46 cents, a value known as the Pythagorean Comma.  If this does not happen the circle does not close because it is either smaller or larger than the Comma, and in either case the octaves will not be pure.  These matters are explained in almost any treatise on temperament, such as Padgham's 'The Well-Tempered Organ' [8].  (This well known book, while useful in some respects, should be used with caution as it contains many numerical errors [9]).

Having said this, it is not really necessary to represent the relative tunings of the fifths in circular form on the page as a tabular representation is just as convenient if not more so, and this is given below in Table 3.  Intervals are expressed here using the tuner's convention of referring to notes either as naturals or sharps, which means that not all of them make sense musically.  In particular the interval A# to F is not, musically speaking, a fifth even though it brackets the same number of physical semitones on a modern keyboard as all the others (it should really be written as A# - E#).  However it is presumed readers will understand this.

Table 3.  Tuning data for the twelve intervals making up the 'circle of fifths'

In this temperament all fifths are tuned flat from pure because the instructions specify that this must be so.  Therefore there are no pure or sharpened fifths as there are in Jorgensen's version [4] or as implied by Johnson [7].  In equal temperament all the deviations would be the same and equal to one twelfth of the Pythagorean Comma or -1.955 cents [10].  Because the deviations in the second column of the table are not the same, this indicates that the temperament is an unequal one.  However three of the intervals nevertheless take the ET value whereas five are smaller (-0.955 cents), which means these intervals are nearer to pure than in ET.  Three have larger deviations from pure than in ET at -2.955 cents and one is larger still at -3.955.  These variations are reflected in the beat frequencies one hears when the two notes bracketing the interval are sounded together, and they are tabulated in the right hand column [13].  The three ET intervals beat at or close to 1 Hz, those with smaller deviations beat more slowly and those with larger ones more quickly.  (Note that the beat frequencies depend on the pitch of the stop and the octave in which the intervals are played whereas the cent values are independent of frequency.  Here the numbers relate to an 8 foot stop in the octave starting at middle C).  The slowest beat at 0.5 Hz occurs for the interval D# - A#, which is one beat every two seconds, making this interval nearly perfect.  The other intervals with the smallest deviation values are also very close to pure (within one cent).  The fastest beat at nearly 3 Hz, three beats per second, occurs with G# - D#, and this is the nearest one gets to the 'Wolf'  intervals which occur in some other unequal temperaments. Even so, this interval is far removed from the unpleasantness of any interval meriting the Wolf soubriquet.  The variations in beat frequencies of these and other intervals are responsible for the impression of key colour which one receives from this temperament about which more will be said presently.

Another column of the table represents the deviations from pure as fractions of the Pythagorean Comma.  For the three ET intervals these values are all 1/12, whereas for the others the fractions are larger or smaller.  It is noteworthy that the smallest fraction is 1/24.  A divisor of 24 does not appear frequently in unequal temperaments from this era, and particularly so when one goes further back in time to the 17th and early 18th centuries.  It only seems to make an appearance later in the 18th century in temperaments such as certain versions of  the French Ordinaire in which the fraction 5/24 occurs.  Two other examples, mentioned by Padgham [8], occur in the Finchcocks and Oakes Park English organs where fractions of 1/24, 7/24 and 11/24 arise (Finchcocks) and 1/24, 5/24, 7/24 and 9/24 (Oakes Park).  The English examples might be no more than interesting coincidences because the temperaments could only be reconstructed from fragmentary data, and Padgham's book cannot always be relied upon in any case in terms of the numbers it contains [9].  Nevertheless the coincidence is worthy of note all the same, particularly in the context of the evolution of English temperament.   The minute deviation of only 1/24th of the Pythagorean Comma in some of these temperaments (including the one derived here) implies that tuners were honing their art towards greater precision by the mid to late 18th century because it represents a deviation of only one cent from the pure interval.  It might also reflect an intermediate phase of the haphazard drift towards equal temperament which subsequently became better established in Britain as the 19th century advanced.  Statements in the tuning instructions above such as "let the Fifth be nearer perfect than the last tho' not quite" and  "tune the ... F, to a perfect fifth with C, give it a little inclination higher" give the impression that some exactness is required to nudge intervals only a smidgeon away from pure, even though the numerical precision of specified beat frequencies did not appear in the tuning instructions.  This is some distance removed from many earlier temperaments which were strongly unequal with multiple perfect fifths on the one hand, much larger fractions of the Comma (such as 1/4th and 1/6th) for the detuned ones on the other, the inclusion of Wolf intervals and the appearance of several unusable keys.  Therefore the picture of this temperament seems to be painted on a canvas which confirms my reading of it as a mildly unequal one containing hazy pre-echoes of ET.

What does it sound like?

It is next to impossible to describe the sound of something as complex as a musical temperament in words or numbers.  One needs to try it for oneself.  However a few remarks might be of interest. Firstly all keys are useable and it is therefore a well temperament.  Secondly a range of subtle key colours is noticeable.

Thirdly it is worth remembering that temperaments produce different effects on different instruments, such as the organ compared with the harpsichord.  There are several reasons for this but an important one is seldom mentioned.  This reason is that the organ is unique among keyboard instruments in that  its sound while it is being tuned is usually very different to that when it is used to render music.  This is because tuning is done using a single stop as discussed above, whereas several quite different stops and many combinations thereof are available for performing music.  Although music will sometimes be played on the stop used for tuning alone, most of the time it will not.  No other keyboard instrument has this attribute.  A single Principal-toned stop is usually used for tuning in which harmonics up to at least the fifth must be clearly audible, otherwise the necessary beats will not exist which are needed to inform the ear that the interval of a major third is pure or not.  As explained earlier this interval only generates beats if the fourth and fifth harmonics of the two notes are audible.  Yet a Flute stop either does not have such an extended harmonic retinue or, if it does, the harmonic amplitudes fall off much faster than for a Principal.  If they did not, the stop would not sound like a Flute at all.  Therefore one either does not perceive the fast beats of the sharpened thirds which characterise equal temperament when playing on a Flute stop, or they are insignificant.  Consequently this often-deprecated property of ET is suppressed for Flutes, which can therefore sound pretty much as attractive in ET as they do in (certain keys of) some unequal temperaments where the thirds are better in tune.  Hence one cannot be too prescriptive in condemning or approving temperaments for the organ, because their subjective effects vary dramatically for different stops and combinations of stops.  This needs to be borne in mind because the version of 'Handel's' temperament which has emerged here, although an unequal temperament, shares with ET its rather rough major thirds - as the tuning instructions suggest that it should.  The point is that one does not necessarily perceive them as 'rough' if the stop(s) being used only have weak fourth, fifth and subsequent harmonics.

However if one plays on stops with greater harmonic development such as Principals the effect of the sharpened thirds emerges, and together with the beats between some other intervals this temperament is found to have a variety of distinct key colours.  The colours are not as strongly differentiated as in a more obviously unequal temperament but their presence is noticeable and subtly attractive nonetheless.  Key colour  is largely a property of how much the consonant intervals deviate from pure.  These intervals are mainly the major and minor thirds, the fifths and their inversions (minor and major sixths and fourths respectively).  The remaining intervals (semitone, whole tone, etc) are more dissonant because there are far fewer harmonics of the two notes whose frequencies nearly or completely coincide, therefore they sound harsher to the ear regardless of temperament.  Consequently the subjective effect of a temperament is influenced strongly by the harmonic relationships between just two intervals - the fifths and the major thirds - because these and their inversions involve the pairs of notes which form most of the consonant intervals.  In turn these manifest themselves to the ear by a complex and ever-shifting web of beats between their harmonics when music is played.  The shimmer of the beats differs in different keys, creating an aural tapestry which is strongly dependent on temperament.  Thus arises key colour.

Therefore we need to discuss how the fifths and thirds beat to arrive at a primitive understanding of why this temperament sounds as it does.  The fifths have been covered already in the previous section and Table 3, so here we dwell briefly on the thirds.

Table 4.  Major thirds: beat frequencies for equal temperament and the present realisation of 'Handel's' temperament (8' pitch; middle octave)

Table 4 shows the beat frequencies of the thirds both for equal temperament and this realisation of 'Handel's' temperament in the middle octave of an 8 foot stop.  There are several differences.  For ET, there is a gradual increase of beat frequency as one ascends the scale whereas for the temperament realised here the values jump up and down from one interval to the next in most cases.  This behaviour is most noticeable, both in terms of the numbers in the table and when listening to the intervals, for the progression A - C#, B♭ - D and B - D# over which the beat frequency suddenly increases to over 20 Hz and then drops back again.  These effects contribute to a more interesting aural experience when listening to this temperament compared with the relative blandness of ET.  The effects arise because B♭ is offset from ET by -3 cents whereas its major third (D) is not (see Table 2), consequently its detuning contributes to the impression of a distinct key colour in the keys which include it.  E flat major, F major, F sharp major and B flat major are examples.

The irregular scatter among the beat frequencies of the major thirds compared to that in ET also endows this temperament with another aspect of key colour which emerges when mutation stops are included in an ensemble, particularly those at 1 3/5 foot pitch such as Tierces or Seventeenths and the equivalent ranks in mixtures.  These are tuned pure relative to the unison stops (i.e. so that there is no beat between their fundamental frequencies and the fifth harmonics of the unisons).  This results in tuning clashes with the corresponding tempered notes of the scale, and although this also occurs with ET, the effects here depend more strongly on key because of the scattered tuning offsets of the notes shown in Table 2.  This is another contribution to key colour which ET does not possess.  Nor does any other instrument because none of them has mutation stops, which is another reason why the choice of temperament affects the organ in a unique way as noted earlier.

Together with the slower beats of the fifths tabulated earlier in Table 3, those of the thirds are largely responsible for the colours of this temperament.  The slow-beating fifths produce some 'throb' and the faster thirds add 'purr' to an ensemble, at least for 8 and 4 foot stops in the central region of the key compass.  These modulations are the underlying phenomena which in fact generate the various key colours.  The spectrum of colour within this temperament is not great, and there is certainly nothing comparable to the extreme effects in some other unequal temperaments which render certain keys poor or unuseable.  All keys are useable here.  It is a mildly unequal temperament which nevertheless possesses greater aural interest than ET itself.  However, whether it was designed to be so assumes that my reading of the tuning instructions is correct or at least reasonable.  

Concluding remarks

Although the association with Handel has little foundation, this article has suggested that it is worth revisiting the temperament attributed to him for several reasons.  One is the unfortunate fact that many modern realisations of it are wrong because they are incompatible with the complete set of tuning instructions given.  These illuminate tuning practices at a time when beat counting was uncommon and therefore, while vague on detail, they nevertheless incorporate an explicit set of overarching constraints which place limits on the possible outcomes.  In particular, all fifths must be flat rather than pure or sharpened, and all thirds must be considerably sharp.  It is therefore regrettable that the modern realisations of this temperament which are embodied in electronic tuning devices are mostly wrong because they incorporate pure fifths.  It is also impossible to conclude, as some authors have done, that the temperament was just another variation on the meantone tunings which were common at that time in Britain.  On the contrary, it is difficult to see how the temperament can be other than a mildly unequal one.  Consequently it probably lies on the haphazard path which ultimately led to equal temperament in the nineteenth century rather than being merely another example of an already-outmoded meantone approach, although the temperament itself is not equal because the tuning instructions also dictate that the fifths should differ in their deviations from pure.

The frustratingly unfocused tuning instructions nevertheless allow for several subtle variations in realising a version of the temperament, and therefore I took advantage of this to produce my own which has been described here.  It addresses the tuning instructions at a level of detail not found elsewhere, providing a reading which I therefore believe to be novel.  The outcome is a pleasing well temperament with a range of subtle key colours.  Because of the imprecision of the tuning instructions it is not possible to interpret them in a single, unequivocal fashion - the best that can be done is to come up with one of several possible variations on the theme of 'mildly unequal'.   That described here has been represented as a set of deviations from equal temperament for each note in the scale, a format which is well suited to most of today's electronic tuning devices and apps, thus it should be reasonably simple to evaluate in practice.

Acknowledgement

I should like to thank my friend David Hitchin, mathematician, musician and instrument maker, for his knowledgeable perspectives on tuning and temperament during many discussions of the subject.

Notes and References 

1. "Twelve Voluntaries and Fugues for the Organ or Harpsichord with rules for tuning by the celebrated Mr Handel", Book IV, Longman & Broderip, London, c. 1780.

2. "Muzio Clementi as an Original Advocate, Collector and Performer, in particular of J S Bach and D Scarlatti", Stephen Daw. In "Bach, Handel, Scarlatti - Tercentenary Essays", ed. Peter Williams, Cambridge University Press 1985.

3. "A Clear and Practical Introduction to Temperament History", Fred Sturm, Piano Technicians' Journal, May 2010.

4. "Tuning, Containing the Perfection of Eighteenth-Century Temperament, the Lost Art of Nineteenth-Century Temperament, and The Science of Equal Temperament Complete with Instructions for Aural and Electronic Tuning", Owen Jorgensen, Michigan State University Press 1991. 

5. The tuning instructions include the simple phrase "all fifths should be flat" [1]. Nothing could be clearer. With its pure fifths, Jorgensen's realisation of the temperament [4] must therefore be wrong.

6. The 'TuneLab' digital tuning app is an example of those which apparently use Jorgensen's tuning data to realise the 'Handel' temperament. Tuning offsets for each note in the octave appear to be those derived by Jorgensen [4], and they (incorrectly) generate two pure fifths. See:

https://tunelab-world.com/tempers/George%20Frederick%20Handel%20well.tem (accessed 28 July 2016)

7. "The rules for 'Through Bass' and for tuning attributed to Handel", Jane Troy Johnson, p. 70, Early Music, February 1989.

8. "The Well-Tempered Organ", Charles A Padgham, Positif Press, Oxford 1986.

9. Padgham's 'Well-Tempered Organ', an article on this website, C E Pykett 2013.

10.  Expressing quantities in cents to a precision of three decimal places might seem excessive but in fact it is necessary if substantial rounding or truncation errors are not to accumulate when several arithmetical steps are cascaded.  For example, the Pythagorean Comma (-23.46 cents) cannot be usefully represented to a precision of less than two decimals, and this means that intermediate calculations must be performed to a precision of at least three decimals.  Spreadsheet calculations are particularly vulnerable to this insidious problem because it is not always obvious that it is happening.  It is possible that many of the numerical errors in Padgham's treatise on organ temperaments [9] arose in this way, though others were clearly a by-product of sloppy typesetting and proof reading.

11.  "Unequal Temperaments: Theory, History and Practice", C Di Veroli, Bray Baroque, Ireland, 3rd edition, p. 447, 2013.

12.  Quoted by Fred Sturm in [3].

13.  The beat frequencies in the right hand column of Table 3 are approximate values intended only to facilitate discussion rather than for tuning this temperament by ear.  One should derive more accurate values expressed (for example) as the number of beats per minute for the latter purpose.