Trend or Change?
We must realize that at the Bach’s time, break (sic) in boy’s voice appeared much later than today, at about 16-17 years of age. We know that Bach singers were about 16, 17 or even 18 years old. Today we have boy singers much younger, at about 14, because children grow faster and break comes earlier. So we can suspect, that Bach singers were more skilled, because of older age. (Wygnaski, 1998).
Introduction
In the previous chapter we looked at what puberty is in boys, how it is measured in the paediatric clinic and how it can be estimated to a workable degree of validity and reliability by non-intrusive means. We looked at how the speaking voice can be used alongside height and weight measurements and saw how much correspondence there is between speaking voices and the definitive clinical measures of puberty. In the next chapter, we will look at the vastly more complex issue of the singing voice, but first we need to consider the extent to which the timing of puberty varies over time and place. Small changes in when puberty begins and how long the process takes from start to finish can have quite big effects on singing, as we shall see in later chapters.
The secular trend
“Secular trend” is an epidemiological term for a slow, gradual drift in a population statistic and we live in a time when many people report belief in, or experience of, a trend towards earlier puberty. Associated with a secular trend to earlier puberty is a secular trend to greater stature. Such trends are of direct relevance to our work because if puberty is earlier it means that the voices of living boys are likely to begin the process of change sooner than in the past. Not only might the time of onset vary, but so too might the intensity or tempo of puberty. If there is a trend for this to increase, boys’ voices will change more quickly than they did in the past.
Greater stature in living boys affects how we view boys who would have been contemporary with now dead composers. We should not, however, assume any direct correspondence between puberty trends and stature trends. In this chapter we will need to consider the extent to which these matters are purely of academic interest or have real and perhaps significant implications for practice.
Cole (2000) identifies the first complication we need to consider. The term “secular trend” implies movement in one direction only but as we shall see, the onset and tempo of puberty varies in both directions over time and indeed place. The term “secular change” is probably therefore preferable to “secular trend”. As far as we are concerned, this matters most when we compare boys alive during the period of the Oxford Movement with boys alive at the time of Tallis and Byrd. A unidirectional trend would lead us to expect that boys alive when Stanford was composing would be taller and experience puberty earlier than boys alive when Byrd was composing, but the reverse is almost certainly the case. As we understand more of this we will come to see that the implications for part placement, singing technique, chorister welfare and historically informed performance are such that a good understanding of the contents of this chapter is desirable for musicians who work with boys. Recreation of the sounds of the past may be harder than is sometimes imagined. The mere fact that boys are used to perform works from previous centuries is no guarantee that the resultant sound is what the composer heard.
What do we know?
Though I later use the term “mean” to refer to the lower of the two vocal parts for boys of the fifteenth and sixteenth centuries, I use it in this chapter in its statistical sense. There is an unfortunate tendency amongst some writers on music history to draw too much inference from anecdotal examples. It is recorded that Haydn’s voice “broke” at age 16, but we cannot infer from this that 16 was the age at which most voices broke at that time. Indeed, when we look more carefully, we find that by sixteen, Haydn whose childhood voice had been exceptionally good was “singing like a crow”, almost certainly the consequence of puberty having begun for begun for him some years before age 16.
We saw in the previous chapter how wide of the mark it is to assume an arbitrary age such as 13 when a particular voice or vocal event might be expected. Either side of any central tendency there extends a dispersion marked by the extremes of range. We saw one boy whose SF0 had reached well into in-puberty values by age ten and another who did not reach the same values until nearly sixteen. Add to that the uncertainties of when choir directors decide by their own criteria that voices have “broken” and we can see that we must be very wary of statements about when boys’ voices “broke” in the past.
In 2013 I collaborated with Anne-Christine Mecke on a meta-analysis of all the historic sources we could find on puberty and ages at which “voice break” was reported in previous centuries. The title of our paper, Boyes are apt to change their voyce at about fourteene yeeres of age (taken from Plato) reflects the fact that, in spite of all that is said about changes in the timing of puberty, the age at which the voice drops most rapidly in pitch has, on average, hovered at just before or after the fourteenth birthday for over two thousand years (Ashley and Mecke, 2013). However, much of our word count was taken up by pointing out how unreliable and inconsistent our many sources were. The age of 14 is not the age of puberty, it is the age at which observers without access to modern medical knowledge thought that a voice has “broken”. We need to contend not only with a demonstrably wide range in the onset and tempo of puberty, but an even wider range in how persons alive before present day medical knowledge identified it and what counted as “voice break”.
How do we know?
We saw in the previous chapter how we can now identify with precision the stage of puberty a boy has reached. J.M. Tanner’s seminal Harpenden study was begun in 1948 with publications from the 1960s onwards. The Prader orchidometer was introduced in 1966. We can conclude from this that accurate, clinical measurement of puberty is a relatively recent development in medicine. So, the short answer to the question of how do we know about secular change in the timing of puberty and voice break over the period covered by the term “dead composers” is that we know less than many people assume! Statements such as “at Bach’s time, break (sic) in boy’s voice appeared much later than today” need to be treated with greater circumspection than is often the case.
There are two broad time periods that we need to consider. The more recent past, from the middle of the previous century onwards offers us the chance to compare the progress of boys singing soprano lines in the 1960s, 70s and 80s with boys of today. This can be relevant wherever traditions of choir training have been passed on as empirical knowledge as happens when a former boy chorister becomes a choir director and treats his boys the way he himself was treated. The more distant the past, the more we must come to rely on the discipline that has become known as bioarchaeology. The two time periods require different approaches and have different messages for practice with currently living boys. I begin with a brief survey of the more recent past before looking back to the sixteenth century. I return to this more distant past in the later chapters on historically informed performance.
Our main sources of evidence for secular change in the era before accurate measurements of puberty were made are army medical records and prior to these, skeletal remains. There are anecdotal stories such as Sweet Polly Oliver (a woman who disguised herself as a man to join the army) but these have minimal scientific validity and the best we can do is infer puberty from what we know about stature and skeletal development. Skeletal remains and army medical records are of some value and if we understand the relationship between stature, growth velocity and puberty, they can be quite revealing.
The period 1850 to 1950 was the period of unbridled enthusiasm for the new voice science ushered in by Garcia’s laryngoscope. I referred to this period in How High Should Boys Sing? as the “golden century” (Ashley, 2009: 27 – 30). Not only did this period include the important scientific work of writers such as Lennox Browne (1876) it included the blossoming of the Oxford Movement in poor industrial parishes and the explosion in newer foundation public school chapels with lusty singing by upper class boys. It almost certainly saw the beginnings of the “aesthetic brand” of the mid twentieth century cathedral choir (Day, op. cit.). Against this, the nineteenth and early twentieth centuries were notorious for under-nourishment, stunted growth, and short stature in the poverty-stricken, dense urban populations of the new industrial towns in which the Oxford Movement took root. These facts are well documented by large scale population studies such as those of Köpke et al (2003) and Stekel (2008).
There are reports of the impact this had upon the military, leading amongst other things to the founding of the boy scouts. Considerable alarm was expressed by Baden-Powell and others at the short stature and feeble physique of potential army recruits in the wake of the Boer War (Jeal 1989). Comprehensive data on the heights of 18-year-old British army recruits between 1860 and 1974 are found in Rosenbaum and Crowdy (1992). These show an average of 160.7cm in 1864. There were year on year increases to 174.2cm in 1974. The equivalent in the RCPCH data used in this study is 176cm in 2012. Data on weight are even more striking. In 1825, the average weight at age 18 was only 124.3 lbs (56 kg). Today it is 67kg. Eleven-year-olds in 1825 weighed on average only 57.5lbs (26 kg, the fiftieth centile today for an eight-year-old). The average weight for young men aged between 21 and 24 was 135lbs (61.2 kg). This is reached today at a little under sixteen years of age.
One of the most dramatic illustrations of secular change in height is provided by Van Weiringen (1986) who examined the height distribution of Dutch army conscripts between 1851 and 1983. The study used changing positions on the centiles for 1863. Thus 170cm in 1863 was taller than average, falling on the 76th centile. The same height in 1983 was well below average, falling on the 5th centile. Centile positions are a useful means of comparison and I use them in later chapters for comparison of case studies.
Military recruitment records cease to be a useful source once we wish to look back further than the nineteenth century. One of our main sources of hard evidence on past stature appears to be the work of welfare economist Richard Steckel. In pursuit of information on historical social wellbeing, Steckel and his team have evaluated skeletal remains of some 15,000 individuals who lived in Europe over the past 1,800 years. Amongst their conclusions are that a geographic gradient in health exists that favoured central and northern Europe over the Mediterranean region. They also found a general decline in stature that largely coincided with the little ice age [about 1300 – 1850 in total duration] (Steckel, 2008).
The sixteenth century was not the low point in this trend, so it is quite possible that boys singing in Byrd’s Lincoln choir were on average taller than boys singing in an a nineteenth century industrial town. A more recent paper by Galofré-Vilà et al (2018) confirms this through noting that the early years of the sixteenth century were unusually healthy. The mean adult male stature rose to reach 173-174 cm, before commencing a falling trend after the 1650s that continued into the 1800s. The comparable figure for today is 177 cm – only 3 or 4 cm taller. Another large study by Köpke et al (2003) provides further confirmation that heights declined from approximately 171 cm in the sixteenth century and expands on the fact that the low point in stature was the nineteenth and early twentieth century, not the sixteenth.
It is well established that child stature is reliably predictive of adult stature, a fact that is supported by comparative growth studies undertaken in differing geographic regions today (Myburgh et al, 2017; Godoy et al. 2008; Bogin, et al. 1999) Based upon this, it is possible to use centiles to extrapolate how tall sixteenth century boys might have been. The process is shown in the table below. Taking age 20 as adult height and 173-174 cm as the median, an adult of today’s 25th centile (172 cm) would equate approximately to the 50th centile for the sixteenth century. An “average” (if by that is meant 50th centile) twelve-year-old of the sixteenth century would therefore have been 143 -144 cm tall, resembling more closely an eleven-year-old of today (143cm at 50th centile).
The difference is not radical and it is possible to see that a taller twelve-year-old boy living in the sixteenth century (perhaps 143 cm) would have been greater in height than a shorter twelve-year-old living at the present time, for example, 139.7 cm (actual height of longitudinal study subject at age 12:05 in July 2012).
Fascinating as they are, these cross-sectional figures tell us relatively little about puberty itself, still less about voice pitch. At this point we run into a significant difficulty. Most of what we know about the progress of living boys through puberty is obtained from longitudinal studies which allow the determination of growth velocities. Any bioarchaelogical study must necessarily be cross-sectional, with the further limitation that soft tissue essential for the definitive measurement of puberty (the testes in boys) is not preserved in skeletal or dental remains, nor is laryngeal tissue for vocal analysis (but see Howard et al, 2020).
A significant recent study of considerable significance by Lewis et al (2016) takes us about as far as we can go without longitudinal or soft tissue data. Their study examined a total of 994 adolescent skeletons from three contrasting medieval burial sites that have recently been excavated in England. The headline finding was that the onset of puberty occurred between 10 and 12 years of age (mean for both sexes), much as it does today. However, the tempo of progression through puberty was slower. The authors give the present day age of 16 – 18 for the completion of the maturation process by the criteria used in their study. They suggest that the same milestone was not reached until as late as age 21 in medieval times.
The immediate consequences for singing and part placement are:
- Medieval boys could, much like today’s boys, be considered to be biologically adolescent from the age of ten or eleven, though adolescence as defined by economic independence ended much sooner. Boys were expected to support themselves by the age of fourteen, being required to submit to the poll tax at that age in the fourteenth century and at various times later.
- The period during which they could have sung parts in the mean range or higher would have been longer. By this criterion they could have remained “boys” until their late teens or early twenties.
The possibility that boys who sang during the sixteenth, seventeenth and eighteenth centuries could have been “more skilled because of older age” than boys living today becomes supported by bioarchaelogical as well as some of the documentary evidence found by musicologists. The picture that may be developing is that of a relatively protracted period of “in-puberty” and consequently longer career. The brief, final flourish of the modern boy who experiences a high tempo of puberty may be a relatively recent phenomenon. The economic status of medieval teenage choirboys is another factor to consider.
The bar chart below from Lewis eta al is for the Barton upon Humber excavation, the wealthiest of the four sites in an area of good agricultural productivity.
(From Lewis et al, 2016)
Comparison with modern boys through Tanner staging is clearly impossible. Comparable dental and skeletal data on living boys can, of course be obtained, but Tanner staging, soft tissue measurement and endocrine assay have become the preferred methods for the measurement of currently living boys. Lewis et al identified five dental/skeletal stages, but these are not directly comparable with current Tanner and associated Cooksey stages, so care needs to be taken in interpreting the above graph.
Onset of puberty was identified with some confidence through observation of the mandibular canine, the hamate hook and the cervical vertebrae. Through the dental staging scheme established by Demirjian et al (1973), 54% of of ten-year-olds, 22% of eleven-year-olds, and just 6% of twelve-year-olds were prepubertal. This is quite similar to data for currently living boys established by SF0 and singing range (see below).
Acceleration of growth was harder to identify, resulting in wider age spans and greater levels of uncertainty, though as the graph shows, all remains at the Barton site indicated some acceleration of growth by age 14. Probably the most significant period then, as now, was the period of PHV (peak height velocity). Across the whole sample bone evidence of PHV spanned the entire 11 – 19 period. 22.6% of individuals appeared to have reached PHV between 11 and 13 years of age, the majority (39%) reaching it at 15. A further 24% reached PHV between 16 and 17½. There is no certain way of telling that PHV coincided then as it does now with the period of most rapid voice deepening, but there is no reason to suppose it did not. PHV is the most significant stage for voice and in later chapter I compare these data with musicological data. The deceleration of the growth spurt, which coincides with the developing and settling stages of new baritone extended in some of the medieval skeletons to as late as 21 years with complete fusion of the iliac crest (the final stage of skeletal maturity in the Risser system) being seen between the ages of 16-19 in boys.
Trends of the modern era
Cross sectional studies such as that of Lewis and colleagues can provide a fascinating snapshot of the past but unless the same data are collected longitudinally over time periods of suitable length, trends cannot be shown. The improbably high treble parts imagined by David Wulstan are perhaps a spectacular example of what can happen when imagination is uninformed by the necessary data. Data points from the end of the medieval period (1550) and the first quarter of the twenty-first century are too far apart to show a trend, although the extent to which they show little change in the time of pubertal onset is nevertheless remarkable. In theory at least, any number of accelerations or reverses could have taken place in the intervening centuries. We have already looked at the nineteenth century as a nadir in population heights though do not have complementary data on pubertal timing.
It was not until the 1960s that we had a common language of puberty in the Tanner stages, or until the late 1970s that John Cooksey showed replicable relationships between singing and speaking voices that have withstood subsequent probing by endocrinologists. However, comprehensive measurements were being made before then. Reich produced in 1929 detailed reference measurements of the testes of living children. Individual velocities were not shown, but mean values for successive age groups suggest that maturation was beginning in 1929 at twelve years of age. Gundobin published in 1921 average girth measurements of the penis for the age groups 1 – 6 months; 7 – 12 months; 2 – 5 years; 6 – 8 years; 9 – 13 years; 14 – 15 years and 16 years. These show quite clearly that rapid growth and maturation took place between the ages of thirteen and fourteen in the year 1921 (Gundobin, 1921).
Greulich et al published in 1938 a comprehensive handbook of “methods for the study of adolescent children”. A wide range of detailed and accurate measurements was given but “serial photography” was presented as the “most valuable single technique for recording the externally discernible bodily changes associated with the growth and development of children” (Greulich et al, 1938: 19). The photographic records of these authors allow us to allocate Tanner stages retrospectively to boys living during the 1930s. They provide an exemplar set of photographs of one boy, selected as typical at the time. He is aged 11:03 in the first set and 14:03 in the last. He was photographed without clothing from front, back and side at approximately 6-month intervals. At age 14:03, his height was 159cm and weight 49kg. This would place him at just above the 25th centile of boys that age in 2012.
The table below has been constructed from the serial photographs of that boy. Importantly, it is possible to calculate growth velocity as these are longitudinal data.
(Derived from Greulich et al, 1938)
PHV is achieved here at 13:08 with accelerated growth continuing until 14:03 when there is the increased weight gain that commonly follows PHV. We can also see a “pre-adolescent dip” at age 12. Height velocity slows to less than the normal pre-pubertal rate and no weight is gained between ages 12:03 and 12:09. This is very much the kind of typical growth pattern as described in the previous chapter and further demonstrated in case studies in later chapters.
PHV, as well as being a non-intrusive measurement is the one that is most likely to be consistent or reliable across disparate studies representative of diverse time periods. We are also confident that the relationship between PHV and rapid voice deepening is robust and consistent, though this fact appeared to be unknown to Greulich and his colleagues. They wrote that
“According to Barth (1911) and Scheidt (1923) the voice change occurs in boys, on the average, at about the fifteenth year – at a time, therefore, when the pubic and axillary hair are already well developed. . . .” (Greulich et al, 1938: 68).
It is remarkable that, for all the careful attention to almost every anthropometric measurement possible, not even simple perceptual categorisations of voice were made. Tanner himself did not attach much significance to voice. With modern measuring equipment and the normative data supplied by workers such as Cooksey, we can be confident of the PHV/SF0 relationship and state on that basis that “voice break” occurred for this boy in 1938 between ages 13:08 and 14:03, in other words consistent with the long-standing empirical principle that “boys are apt to change their voice at fourteen years of age.”
How much Barth or Scheidt might be trusted is a moot point. The means by which they arrived at age fifteen are not stated. We know now that in most cases (including the Greulich one cited above) pubic hair is not fully developed by the time of voice change so I am inclined to dismiss the voice estimates of Barth and Scheidt. It is frustrating that their estimates were made at a time when average stature was near its nadir, but we simply do not have the data to make any reliable connection.
Moving forwards a decade from Gruelich, a detailed paper with longitudinal data was published by Ellis in 1948. Ellis reported data gathered during 1947 from two boarding schools in Scotland. Boarding schools were chosen in order that living conditions, particularly diet, could be as constant as possible. Height velocity and weight velocity, albeit with only two data points, were recorded. The particular concern was to investigate the relationship between childhood stature and rate of growth during puberty. Ellis did find a relationship, with boys of greater childhood stature reaching puberty sooner and progressing with more rapid tempo. Confusingly he uses different terminology, although it relates well to the three-phase system described in the previous chapter. His use of the term “adolescent” for “completing puberty” is not untypical of medical literature of the era, as indeed is the term “pubescent” for the “in-puberty” Tanner stages 2 and 3. The table below is presented to avoid confusion.
I have constructed the two tables below from Ellis’s data. The boys were measured twice. Thus, for example, non-pubescent – pubescent would mean non-pubescent at the first examination and pubescent at the second. The light green shading in each table (my interpretation) indicates that a majority of boys will have voices that are beginning to change under the influence of puberty. The pale-yellow shading indicates that the majority of boys have reached the point at which voices begin to deepen rapidly (i.e “break” as PHV is achieved).
Once again, we see the same pattern of a dip to below pre-pubertal growth rate in the 9 – 13 age range with PHV beginning before peak growth as measured by weight. PHV appears to be occurring across the age range 13 – 14 and we may therefore assume that there is little difference in comparison to Greulich’s data of ten years previously. Most boys aged fourteen in 1948 will no longer be living, although it is of course possible that at age 90 a small number might be. We can, by this criterion, reasonably set the age of just turning fourteen as the time of “voice break” in the middle of the twentieth century, a time when the recently retired generation of cathedral musicians would themselves have been boys. We now need to consider whether there has been any significant change since then.
The present era
Rather conveniently, the beginning of the twenty-first century is marked by an authoritative study of secular trends published by Timothy Cole. Fourteen years later, Cole and two colleagues published detailed data that can be considered a baseline for our present era and compared with the decades leading up to 1950. I have extracted the most relevant information from the 2014 data for the table below.
(From Cole et al, 2014: 79)
We can see here that mean PHV has been reached by the age of 14.1 years. This means that there is little change from the 1930s and 40s when, as much as we are able to tell, boys were on average reaching PHV at around the time of their fourteenth birthday. We have to reconcile this with the anecdotal reports that are regularly received from cathedral musicians that they are losing boys sooner.
In 2011 I received funding for a research assistant to undertake a full literature review of secular trends in puberty. My conclusion from her work was that if there has been a trend to earlier puberty over the last fifty years, it has only been a slight one and is often exaggerated. Two large scale meta-analyses were reviewed and the expert panels in both cases reached a similar conclusion to the meta-analysis of “voice break” I undertook with Anne-Christine Mecke. Euling et al (2008) concluded that it was unsafe to identify a clear trend on account of the different methods used on different populations at different times and locations across between 1940 and 1994. They stated that:
… on the basis of few studies and reliability issues of some male puberty markers, the current data for boys are insufficient to evaluate secular trends in male pubertal development. (Euling et al, 2008: S172).
Amselem et al (2007) similarly concluded that the sum of available evidence demands further research and does not justify a confident conclusion that puberty is invariably coming sooner for all boys. Their review gave particular attention to the wide variations in pubertal timing across ethnic groupings, social classes and geographical regions. Smaller studies of specific populations did indeed reveal quite wide variation. In our own review we found results that varied by almost two years. Liu et al (2000) were concerned that the reference standards set by Tanner (1978) might be out of date and set out to establish new reference points through a study of changes in height velocity in 2432 Swedish children. Their conclusion was that there had been secular trend to earlier puberty which did justify new standards. Mean age of onset in boys was found to be as young as nine, with peak height velocity occurring at 13.5 years.
Although I have suggested that height velocity is a good non-intrusive measurement, it is unwise to assume a direct relationship between trends in stature and pubertal timing. Quite dramatic trends in height have been identified, in the Netherlands currently and in Japan between 1950 and 1990. Takaishi (1995) demonstrated that nearly all of this trend was explained by an increase in leg length only, concluding that “the secular trend in adult measurements, both in men and women, is chiefly or wholly a trend in length of legs”. Karlberg (2002) found that the two trends of stature increase and pubertal onset are not strongly correlated. Juul et al similarly published a study that concluded that Danish boys were significantly taller in 2000 than in 1964, but that their pubertal timing had not altered (Juul et al, 2006). Cole (2000) explains that the pubertal growth spurt occurs first in the long bones, the trunkal spurt following later. The timing of voice change may be more closely associated with trunkal growth than leg growth, but this proposition needs investigating. Such an investigation might help to explain the lack of association between low stature and vocal part ranges in the nineteenth and early twentieth century.
Kahl et al (2007) stated that the puberty is not significantly earlier, but gave the age of 15.1 as the median for “voice low”, which is late by any standards. Their data were gathered through children’s self reporting, so “voice low” may be telling us more about adolescent perceptions than either voice pitch or puberty timing. They found that 35.7% of boys had reported the appearance of pubic hair by age ten and gave the ages for Tanner P stages as stage 2, 10.9; stage 3, 12.6; stage 4, 13.4; and stage 5 14.1. “Low voice” coming later that P stage 5 seems highly improbable and can only be explained by a combination of the unreliability of self-reporting and a failure to quantify “low voice”. With discrepancies such as these, the safest position to take is indeed that suggested by the Euling and the Amselem reviews.
The most recently published study published study located at the time of writing is that of Brix et al (2018). These workers also concluded that a secular trend is uncertain in boys, though were more confident that there has been one in girls. Most studies of girls employ the age of menarche as a key marker, which is much easier to identify than the equivalent event in boys (spermarche). Brix et al analysed a total of 73 160 questionnaires returned from 14 759 Danish children born between the years 2000 and 2003, aged between fourteen and seventeen at the end of the study. Mothers completed the questionnaires for children aged under eleven and a half. 7104 boys aged eleven and over answered questions that required them to estimate Tanner P and G stage, first ejaculation and “voice break”. The authors claimed to have 77% consistent information on G stages and 87% on P stages. These figures seem high when compared with Butler’s study of medical practitioner error in judging Tanner stages (see previous chapter).
The method of first ejaculation is seldom used in puberty studies since it relies on boys reporting an event, often from memory, that many do not fully understand. The collection of semen samples through induced orgasm is a highly impractical method, so urine sampling has been used instead. The relative concentration of spermatoza in urine has been shown to be heavily influenced by sexual behaviour and the socially variable age at which boys begin masturbation. For this reason, an early study by Baldwin (1928) was abandoned and the method subsequently saw little use. However, some recent workers have attempted to revive the method. Laron et al (1980) examined and questioned 135 boys. Eighty of these reported first ejaculations and from these a mean age of 13.5 years was determined, with a range from 12.5 to 15.5. No correlation between first ejaculation and TV, pubic hair or penis length was reported, so given the other drawbacks of this method, the authors’ claims of a “milestone in male puberty” seem optimistic. More recently, Chad (2018) reopened the question of spermarche (or “thorarche”) as a male equivalent of menarche but was forced to report that the event occurs involuntarily for some boys but is induced by others, results being heavily influenced by sexual behaviour.
This leaves us with just “voice break”. The approach to this by Brix et al seems flimsy. Boys were asked simply to report from memory whether they experienced this as “yes—sometimes”, “yes—definitive changes” or “no, don’t know”. No further explanation is given in the paper. A later paper by Brix et al that examined the timing of puberty in relation to TV and semen characteristics reported that “the voice break was chosen as the primary exposure as it has been the most frequently used self-reported marker of the timing of puberty and has been successfully used in previous studies” (Brix et al 2023: 2). Examination of the three studies cited confirms that the authors do not appear to have examined the question of what “voice break” means in detail. They report a mean age of 13.1 years, which they noted was “considerably earlier than former studies reporting between 15.5 and 14.0 years in the period 1968 to 2005.” Could their method be relied upon, this would evidence a significant advance in the timing of puberty, but “voice break” as a phenomenon self reported from memory and unsupported by actual measurements of SF0 or voice deepening velocity does not, in my view, qualify as a valid method.
The table below summarises their data. If we refer back to the Harries et al study (see previous chapter) we find the event that Brix et al may be calling “voice break” occurs during the transition from Tanner G stage 3 to 4. The mean age of 13.8 for stage G4 may be closer to the time of “voice break” than the Brix estimate of 13.1, but without any means of linking subjective recall of “voice break” with measurements of voices perceived as “full broken” we are left once again to conclude that “the current data for boys are insufficient to evaluate secular trends in male pubertal development” (see above).
From Brix et al, 2018: 77
We in the singing community may not like the term “voice break”, but it is unfortunately the nearest we have to a common language across disciplines and over time. If we are prepared to accept that “voice break” and choir dispensation (the time a boy is dismissed from the treble line of a choir) are largely contemporaneous events, one of the most useful landmark studies of recent times was that undertaken by Anders Juul and colleagues in Copenhagen.
Juul is professor of growth and reproduction at Copenhagen’s Rigshospitalet and lists amongst his specialisms the timing and regulation of normal puberty and its disorders. His aim, like that of Butler et al (see previous chapter) was to find acceptable non-intrusive measures of male puberty that can be used in larger population studies than are possible when admission to a paediatric clinic is required. He worked closely with the Copenhagen Royal choir and usefully defined “voice break” as when “unintentional falsetto notes” were detected by the choir director. Since falsetto, according to Cooksey, first emerges at stage 3 and Harries et al have shown Cooksey and and Tanner stage 3s to be closely associated, we at last almost have a common language!
The table below is derived from their 2006 paper:
It shows the age for unintentional falsetto notes and PHV as identified by Cole at the beginning of the present millennium to be much the same, 14.1 years. However a small, though statistically significant advance of four months between boys enrolling in 1990-1992 and boys enrolling in 1997-1999 occurred. Of possible significance is the fact that the Juul data also show the heights and weights of the four cohorts of boys on enrolment at age 8½, and there is no similar advance in these.
Though the advance across this period was statistically significant, data from just one choir over a ten year period are not sufficient to identify any trend over the last half century. Ideally, the study would be repeated at perhaps ten year intervals but it has not been. Neither do we have comparable data for decades before 2000.
The need for measurement and a clear base line: the 2012-13 cross-sectional study
With this need in mind as well as the need to quantify the term “voice breaking” I undertook during 2102 and 2013 a cross-sectional baseline study in which boys’ speaking voice pitches were recorded as a function of their chronological ages. At the very least, I thought, researchers who come after me will have a reference point in the beginning of the twenty-first century. Over a thousand boys were seen, drawn from secondary schools across the country, youth choirs and seven cathedral choirs. The method was simple. Boys were released from classes one at a time to make short recordings which could later be analysed for pitch parameters. All boys were asked to count backwards from twenty and to read the first portion of the Arthur the Rat phonetic passage. If the boy had singing experience (as in the choir schools visited) he was also asked to complete vocal glides and pitch matching exercises. Analysis was by the Praat speech software, cross-checked against the Voce Vista system and in one sub-study EGG measurement by Laryngograph. The pitch derived was recorded against the boys’ ages in years and months.
The results for the whole sample of 1068 boys aged between eleven and fourteen are presented in the table:
These values can be compared with the quantifications of perceptual judgements described in the previous chapter. It will be recalled that the SF0 value of 164Hz represented the threshold between voices perceived as “slightly deepened” and voices perceived as “fully broken”.154Hz was the first value that all participants in the perceptual test confidently agreed to be a “full broken” voice. This value was achieved as a mean between the ages of thirteen and fourteen, so there is a measurement of agreement with the Juul study. Voices are “breaking” at or just before a boy’s fourteenth birthday. The range, however is large. There were still some thirteen year olds who had not even begun puberty, having SF0 as high as 237Hz. Equally, a small number of eleven year olds and a larger number of twelve year olds had “broken voices”. It will also be recalled from the previous chapter that we equated 200Hz with the achievement of 4ml TV, i.e. the medical onset of puberty. Boys with voices between 195 and 205Hz will, by this criterion, be at the onset of puberty and we can see that this similarly occurs between the ages of eleven and twelve, again with a large range.
Cross-sectional studies are, as we have said, useful in some circumstances but limited. If we read across the top row of the table, we can gain some idea of velocities. Mean SF0 falls by 9.5Hz between ages eleven and twelve, then by 25Hz between twelve and thirteen and finally by 35.8Hz by age fourteen. From this we can deduce that PHV must be occurring on average as boys turn fourteen. This agrees well with both Cole’s and Juul’s data (see above) but also suggests a similar pattern to that found by Lewis et al in medieval skeletal and dental records. A more precise picture from a somewhat smaller sample is obtained when aggregate data for the thirty two boys I have studied studied longitudinally are produced. Significantly more data on these boys, including heights and weights and a broad range of singing parameters are available. These data allow the pubertal milestones to be more confidently identified.
Here we see a threshold for completing puberty of 13:07 years. This means that PHV was, on average beginning at this age and reaching its apotheosis at just around fourteen years of age. Just turning fourteen does seem to have been the age at which boys are “apt to change their voices” over two thousand years. Our present age looks to be one of those when events occur a few months sooner than the two thousand year average. That being the case, we will need to look at other explanations for the reports of earlier loss of boys from choirs and the demise of the once familiar teen boy soprano.
Summary and key learning points
To understand and manage voice change, we need to take account both of the age of pubertal onset and the intensity or tempo of puberty.
Both of these qualities are influenced by environmental factors such as nutrition, disease and stress. Trends can be positive or negative.
The intensity of puberty and the tempo of growth are closely related. Pubertal progression in living boys can be established non-intrusively by longitudinal measurements of height. Progression in boys alive in previous centuries can be established by the examination of skeletal and dental remains.
There is a consistent and close relationship between the tempo of growth and the tempo of voice deepening. At pubertal onset, the tempo of growth increases and voices begin to deepen slowly until growth accelerates further to peak height velocity (PHV) when voices deepen rapidly, or “break”.
The medical community has long recognised the deepening of the speaking voice as a secondary sex characteristic but has been relatively slow to exploit the potential of voice pitch as a valid and reliable surrogate measure of puberty stages.
Twenty-first century technology can measure voice pitch very accurately and these measurements can be related to equally accurate measurement of puberty through ultrasound scans and other technologies previously not available.
The relatively recent development of the necessary technology means that estimates of the time of “voice break” made in previous centuries may not be very reliable. We know less about trends in the age of “voice break” than is often reported.
The best available evidence suggests that pubertal onset occurs today at largely the same age as it did in medieval times, but the intensity of puberty has increased. Boys living today tend to pass more quickly through the stages of puberty than boys contemporary with “dead composers” of historical times.
A high tempo of puberty may be associated with currently living boys leaving choirs at younger ages than during the twentieth century.
