Chapter 5: Twenty Archival Sound Recordings:
Statistical Timing Analysis

Performance timings have in general been derived by performing stopwatch timings, with the identical instrument, during the initial audition periods. In some cases, timings have been provided by the issuing companies, and where these have been found by spot-checks to be serviceable they have been used. In general it should be remarked that variations from the reported timings may be considered unremarkable and within the presumed tolerance limits for this type of study. The exactitude of the timings surveyed, therefore, is within statistically acceptable limits only insofar as they serve to indicate general trends in those differences; a very fine difference will not be as significant as a dramatic one.

In three instances, irregularities in the performances have been noted which have required particular attention. These irregularities pertain to "missing data," that is, portions of the work not contained in the recording in question. As it is apparent that the omitted portions were deemed "unnecessary repeats" of already-performed material by the conductors, these variations have merely been indicated here and will be treated in greater depth in the chapter dealing with subjective comparisons.

However, for a statistical survey, the problems posed are potentially quite serious. The inclusion of this raw data would seriously affect the integrity of the analysis of the movement timings in each instance. A method of correction has therefore been applied so that the movement timings in question will indeed refer to timings (either real or estimated) for the entire movements in question. This correction is based on the assumption that the omitted portions, having been omitted as "unnecessary repeats," may have taken up a performance time essentially identical to the corresponding passages which were included in the performance. Thus the presumed timing of the missing data, derived from a comparison of the omitted and included passages, and the timings for the latter, in whole and by passage, has been estimated and added to the raw timings in order that the statistical study of these instances may be a survey of the identical material. True timings for the actual incomplete performances are given as footnotes in Table 1 as appropriate.

Table 1 displays timings, in seconds, for the individual movements of the twenty recordings, as well as for the total timings for the performances. (As indicated earlier, the timings for the Agnus Dei in AV, as well as for the Sanctus in FO and AV, have been corrected for the missing data, i.e., performance cuts.) Shown for comparison with these timings is the mean performance time for each movement (and for the entire work), and for the "score" timing, which might be termed a function of the "ideal" metronomic tempo indications in Berlioz' score. Certainly such timings would have their interest as baselines against which to compare actual performances. These "ideal" timings will be referred to as Metronomic Baseline Timings, or MBT, and have been derived as follows:

A close examination of available scores revealed identical tempo markings; since these metronome markings have their origins in editions of the score prepared and approved by Berlioz, they may be taken as representative of Berlioz' "ideal" concepts of the flow of the musical segments.1,2 A count of measures for each tempo segment was thereupon undertaken to determine the probable "ideal" timings for each of these passages. For convenience, a fermata on a beat was arbitrarily taken to double the duration of the beat; likewise, a fermata on a full measure, or a G. P., was arbitrarily taken to double the duration of the measure, In view of Temperley's caution, the passage in the Dies irae from measure 64 to the commencement at measure 141 of the Tuba mirum (i.e., the brass bands' entry), marked animez un peu, has been taken to mean a smooth accelerando from the crotchet = 96 which begins the movement (and therefore extends through measure 63) to the crotchet = 72 at measure 141, where Berlioz indicates "One bar of this movement has the time-value of two of the preceding movement"; the tempo at measure 140, then, is crotchet = 144, and the average tempo of measures 64-140 is crotchet = 120.3 The heightening dramatic tension of the passage, with each subsequent entry of voices and complications of the texture--as well as modulation up a semitone at each of these junctures--suggests a more "even" tempo modification throughout, as opposed to Temperley's suggestion of an accelerando mainly concentrated at measures 138-40. Fortunately, such problems, requiring an arbitrary aesthetic decision, were few and have had little overall effect on the MBTs reached. Rallentandi at ends of phrases and of movements were assigned arbitrary brief extensions of duration, based also on arbitrary aesthetic judgements.

Interestingly, Berlioz has provided us with a possible check for the accuracy of these presumed timings. Kindermann remarks: "On p. 148 [of the autograph full score], at the end of the Offertoire, Berlioz has penciled '8 minutes.'"4 If this is a suggestion of an acceptable or likely performance time for this movement, then the timing of it at 480 seconds seems a bit long in comparison with the MBT of 449 seconds derived for this section. However, since this is a timing written in the autograph score, it should be remembered that Berlioz later tightened the movement in order to remove a setting of some text that is redundant with that transferred (in his reordering of the text) to the Rex tremendae. This tightening takes the form of twenty-two measures from the original version of the work (i.e., the autograph, as well as the first published edition, by Schlesinger, of 1838) being replaced by five measures (not six, in Kindermann's miscounting) in the Ricordi score of 1853 and the Brandus choral parts of 1852. Thus, lightened by seventeen measures of common time at a Moderato of crotchet = 84, or about 49 seconds of music, this "check" timing comes to 431 seconds, which varies from the MBT of 449 by only 18 seconds, or about 4% of the MBT. Since it is likely that Berlioz' indicated "8 minutes" was a round figure in any event, this mere 4% variation may be taken as of little significance.

The MBTs which result from the above-described derivations provide, therefore, a useful tool against which to compare the interestingly varied performance timings. No attempt has been made here to further subdivide the movements back into passages (even when there might admittedly be some interest in comparing these timings segment by segment, as for example in the Dies irae or Rex tremendae), as each movement may be taken to be representative of its own peculiar "mood" in the work as a whole, and it is these grosser considerations which are being addressed here.

Chart 1, directly derived from the data in Table 1, displays the population distribution of the performance timings, against a reference point of the MBT; the chart has been drawn so as to keep the MBT in the same place throughout the chart, and the variations shown in whole seconds minus (to the left) or plus (to the right), indicating quicker and slower performances respectively. The mean has also been indicated in each instance.

Table 2 represents a further derivation from the data in Table 1. It represents these variations of performances from the MET (with the mean once again shown for comparison) in terms of percentage of the MBT. Thus a performance which lasts precisely as long as the MBT would be shown here with a zero value. a performance lasting one-tenth again as long as the MBT with a +10 value, and so forth.

Chart 2 is thus a derivation from the numerical values in Table 2. Once again a fixed location for the MBT is used as a reference point against which to display the variegated timings of the performances. Means have also been charted in an attempt to display trends of these actual performances against the presumed Berlioz timings.

In brief, Chart 1 displays the variation of the actual performances from the MBT in raw second values, while Chart 2 displays these same variations proportionally, as percent variations from the MBT.

From even the most cursory examination of the charts and tables it is patent that the population of performance times encompasses a great deal of variation. Nowhere is this more evident than in comparing WE or OR against SC in the Quid sum miser, a range from 160 seconds in the former two identically-timed performances against 277 in the latter; or, given the proportional approach, a range of 73% (from -1% to +72% of the MBT).

Some more formal observations of the variations demonstrate other trends. For example, it may be seen that virtually all performances of the Requiem et Kyrie are slower (more sustained) than Berlioz' timings (only BE and AV coming close to the MBT), some (SC) to a highly marked degree, The greatest spread, of course, is in the already-mentioned Quid sum miser, whose distribution resembles that of the Requiem et Kyrie; here it is WE and OR on one side of the MBT line, along with AV on the other side of the MBT line, which run closest to Berlioz' indications, in a field which again runs generally into the "slower" category. Note again the extremely marked divergence of SC. To a somewhat lesser extent, the range and spread of the Offertoire displays a similar shape, with CH and BU running slightly quicker than the MBT, and separated enough from the remainder of the range so as to appear anomalous, In this instance SC once again shows the same sort of marked variation, though here it is not alone, being joined in a "crowd" by MA, HO, BA and DA.

The mean variation from the MBT in these three movements is worth noting by way of comparison: +19% in the Requiem et Kyrie, +23% in Quid sum miser, and +21% in the Offertoire. These are the three greatest mean variations from the MBT, and it is interesting to note the similarity in shape of the distributions for each.

Perhaps the other most interesting distribution shapes are those in the Dies irae, Rex tremendae, Lacrimosa and Hostias. All of these distributions are reasonably tight ones, with mean variations from the MBT being +01%, +05%, -03%, and -01% respectively. (The Quaerens me, with its mean variation of -03%, might at first glance at the mean values be included in this group, but has not been so included because the shape of the distribution spread is markedly wider than in the other mentioned movements; and the Sanctus, which displays a concentration of distribution similar to those mentioned movements is to be discussed separately on account of its unusually high-value negative mean variation,) Note in particular the narrowness of range, which runs from -14% (CH) to +15% (DA) in the Dies irae, a sweep of 29%; from -06% (MH and BE) to +26% (DA) in the Rex tremendae, a sweep of 32%; from -24% (WE) to +18 (SC) in the Lacrimosa, a sweep of 42%; and from -18% (MI) to +18% (SC) in the Hostias, a sweep of 36%.

This leaves three movements with "anomalous" distribution shapes (i.e.,, not really comparable to those of other movements), namely, the Quaerens me, Sanctus, and Agnus Dei. The Quaerens me outwardly resembles the distributions of the four movements referred to in the previous paragraph; however, there are three markedly disjunct performance timings, CH at -33% and MH at -31%, and SC at +29%, for a range of 62% overall. The Quaerens me does share the distinction with the already-discussed Lacrimosa and Hostias of mean timings quicker than the MBT, with a value of -03%. The Sanctus again demonstrates a reasonably compact distribution not unlike that of the Dies irae, Rex tremendae, Lacrimosa and Hostias, but the distribution is most noticeably shifted into the range for shorter duration--mean variation is -07% from MBT, the greatest such variation in a negative direction of the mean value for any of these populations. The range here is in fact quite compact, from -18% (CH) to +07% (PR), for a range of only 25% overall. (Note also the heavy accumulation of values from -11% to -08% inclusive, to a total of eight performances in that narrow spread, a concentration matched only in the Rex tremendae with its accumulation from -01% to +03%, or 00% to +04%.)

Indications of these statistical observations offer interesting clues to the performing approaches of the represented recordings. The observed conductors allow themselves as a group a great liberty in determining the tempos in the Quid sum miser, the briefest movement of the entire work. It may be speculated that the conductors' individual approaches to the Quid sum miser may depend on their choice to view it as a postlude to the Dies irae (with which it shares thematic material), or as prelude to the Rex tremendae (the opening of which is quite jubilant and dramatic, and therefore in perfect opposition to the "fearful" Quid sum miser), or even as an interlude of little cumulative consequence to the entire composition. Coming as it does after the remarkable Dies irae (with the Tuba mirum and its brass bands), it may be that the greater proportion of the conductors view the Quid sum miser as a moment of repose dramatically, allowing for a musical-aesthetic opportunity to demonstrate a more reflective attitude with the quieter, sustained singing of the male chorus against the darker of English horns, eight bassoons and lower strings.

The Requiem et Kyrie and Offertoire similarly display the conductors' prevalent interest in tempos slower than the MBT. Note again that only BE and AV offer a cumulative timing in the Requiem et Kyrie which approaches Berlioz' markings, while all other performances are noticeably to markedly slower. The Offertoire displays a similar predilection for slower tempos; however, its nature as (essentially) a symphonic slow movement (c.f. the convoi funèbre de Juliette in Roméo et Juliette, which has a very similar choral ostinato treatment) may offer the conductors opportunity to display the orchestra as performing ensemble over the chorus. (It is interesting to note that the swiftest performance, CH, is from a conductor who specializes in choral performances, and the next swiftest, BU, has what largely amounts to an amateur ensemble, In fairness it must also be noted that the two patently amateur orchestras, those of AV and WE, contribute to performances not too much quicker than the mean value in this movement.

The closer distribution shapes of the Dies irae, Rex tremendae, Lacrimosa and Hostias--the second, fourth, sixth and eighth movements of the work--suggests a pattern of traditional response to Berlioz' markings. (Notice how the three French-raised conductors--FO, FR, MX and MZ, including both of the Alsatian Munch's two efforts--hew fairly close to the MBT in these movements specifically, though they may diverge more widely in the other movements.) The first three of these four "even-numbered" movements may be considered the dramatic centers of the work, as opposed to the more devotional movements, and as such it is interesting to note the relative similarity of approach among the conductors in general (and the French ones in specific). Then too, Berlioz specifies that the Dies irae (with Tuba mirum) and the Lacrimosa are the movements suitable for the deployment of a larger chorus (if available), and if the Rex tremendae is also used as an occasion for a large singing ensemble (as it was on the instance of the performance which produced WE), then such a consistency of approach is hardly surprising. It may be that the conductors have in these parts of the GMdm more of a sense of a consistently "right" approach than they find anywhere else in the work. The Hostias might be considered a dramatic movement, but not in the same sense as the others; that is, the unusual sonorities of the chords of eight trombones and three flutes, and for that matter the remarkable range of the male chorus writing, inspire a greater consistency of approach because the novelty of those combinations would wear too thin at a very slow tempo, and not impress itself suitably upon the mind at a quicker one.

Of the movements with "anomalous" distribution shapes, the Quaerens me is notable chiefly for the three widely separated performances, two at the end of the quicker tempo side and one at the slower tempo side. Otherwise, the performances range from -17% (AV) to +08% (FR) for a range of 25%, a more nearly "normal" form of distribution. Obviously this movement is one on which most of the sampled conductors diverge little from score tempos, where three of them (arguably the most erratic in terms of tempos in this survey) take strongly disjunct views. SC, as it will be noted in greater detail later in this chapter, is characterized by slow tempos throughout; and it is worth of note that CH and MH are among the "swiftest" conductors of the entire work.

The Agnus Dei as well would have a more "normal" distribution shape were it not for the far-outlying SC; aside from that anomaly the range would be from -10% (MH) to +21% (BA and HO), a range of 31%. Perhaps this distribution is relatively wider than some in other movements because of the conductors' individual approaches to this music as the closing statement of the work; hence there is a tendency toward broader tempos to underline the devotional and consolatory functions of the movement.

The Sanctus, as previously noted, displays a most compact distribution, which is also noticeable shifted into the quicker-than-MBT range. It should be remembered that this movement contains the one vocal solo of the work; moreover, it is a taxing solo with a difficult tessitura, particularly if the tenor line is to be sung softly in the higher register (see the discussion of the tenor part in Chapter 6). Consequently the conductor must give thought to the breathing and stamina of the tenor for this section, which may be the reason why most performances are noticeably less sustained than Berlioz' indications require. Though a more extended survey of the tenor soloists will be undertaken in Chapter 6, it is worth noting here that the only two performances of the Sanctus slower than Berlioz' markings, FR and PR, are by the same tenor, Robert Tear; and that even the typically slow SC has a quicker-than-MBT pace in this movement, probably in consideration of the soloist's breathing. The quickest of all performances (CH) is oddly the one in which the tenor part is sung by a group of tenors (see Chapter 6).

Now that some traits of the performances of each movement have been noted as a whole, it may prove suitable to examine traits of the performances as a whole. Firstly it will be noted that the most disjunct of the performances in the selection of tempos is SC. The conductor has in the main chosen the most sustained, even static, tempos (except, as has been noted, in the Sanctus). At first glance it might be thought that special advantage was being made of the acoustic properties of La Chapelle de l'Hôtel des Invalides (the site of the work's premiere), though BS, from the same locale, does not exhibit nearly such slow tempos; and other performances from large, reverberant halls (such as DA, in Westminster Cathedral, and MZ, in the Benedictine Abbey at Ottobeuren) do not exhibit quite such traits of slowness. Benvenga has pointed out how the properties of reverberation are "built in" to the pauses in Berlioz' musical fabric, but the extremely slow tempos of SC must be regarded, at least in comparison with the MBT and other performances, as erratic and wilful.5

At the other end of the scale, AV displays relatively rapid tempos throughout, with only the Offertoire being significantly slower than MBT (though still markedly swifter than the mean in this sample.) One suspects from the recording a rather small, unreverberent auditorium; furthermore, the performing forces are entirely non-professional (students), and the conductor may not have been able to expect miracles in intonation at any sustained tempos, particularly from the singers. Note the rapidity of the a cappella Quaerens me, in contrast to the one markedly slow movement, the chiefly orchestral Offertoire.

CH similarly displays generally swift tempos, the exceptions in this case being the markedly slower, but near to mean, Requiem et Kyrie, and the exceedingly sustained Quaerens me, not surprising from a choral specialist. BA, DA, and PR offer relatively slow performances of the work as a whole, with particular slowness again in the Offertoire, perhaps reflecting the conductors' specialty as orchestral conductors.

Some observation too might be made of those instances of conductors emphasizing the dramatic canvasses of the work (i.e., such movements as the Dies irae, Rex tremendae, and Lacrimosa) over the more devotional Quid sum miser and Quaerens me, if one may take the second through sixth movements--the ones based on the text of the Sequence--as greatly representative of the alternating mood-aspects. FR, for example, sustains the devotional movements more than the dramatic ones, providing the key to his reading of the work. OR is broader than the MBT in these movements only in the Lacrimosa, perhaps underlining the concept of that movement as the dramatic center of the composition. And WE, in comparison, allows his swift conception of the Lacrimosa (swiftest of all in the sampled performances, in fact) provide a sort of scherzo in contrast to his near-MBT timings in the preceding four movements.

Any further judgement or speculation regarding the variant tempos of these observed performances must come under the heading of subjective judgement, and some comments in this regard (in terms of overall selection of tempos) will be found in Chapter 7.

Lastly, two observations may be noted regarding the significance of these timing comparisons. In the one instance in which a conductor has recorded the GMdm twice (MX and MZ), the performance timings are quite close in all but three of the movements, those being the Quid sum miser (slower in Boston, with its student chorus, than in the later Bavarian performance), Quaerens me (with here the Bavarians presumably better able to sustain intonation in this a cappella movement than the Bostonians), and Sanctus (Simoneau in Boston taking the near-MBT tempo, and Schreier in Bavaria requiring a less sustained mood). One may presume that a conductor may change his mind about performing tempos in this work (and indeed in any other) in the course of eight years; the proficiency of one's orchestra and chorus, as well as the particular technique of the solo tenor, may also alter portions of one's total conception of the GMdm.

The final observation is that it may not be useful to view conductors as in any way similar to their teachers, certainly not in the selection of tempos for performances. The timings of CH and FO would scarcely bear witness to the student-teacher relationship of Chekijian and Fournet. Similarly Fritz Mahler worked under Felix Weingartner, but the swift approach of MH must offer us little idea of what must have been the grandness of the older conductor's way with this work, sadly never recorded by him.

Copyright © 1983, 1995 by Matthew B. Tepper

Table of Contents

Chapter 4

Chapter 6


Back to The Hector Berlioz Page

Back to Matthew B. Tepper's Home Page