The relationship between adolescence and the voice remains an important topic of investigation. Contrary to the opinions of J.M Tanner, it is now believed that the voice may be the most reliable non-intrusive measure of a boy’s progress through puberty.
Non-intrusive measurement of male puberty is something of a Holy Grail. The reference clinical standard is testicular volume. The ethical difficulties of making such a measurement, to say nothing of the embarrassment experienced by testee and quite possibly tester, mean that such measurements are made only by qualified paediatricians when absolutely necessary. Sample sizes in research are consequently small. Tanner developed a more frequently used set of reference photographs illustrative of five genital and pubic hair stages (G and PH stages respectively). These still present ethical difficulty and potential embarrassment, and are somewhat less reliable than actual measurements of the testis.
A consistent relationship between voice pitch (speaking fundamental frequency or SF0) has been demonstrated in a number of small-scale studies. That of M Harries and colleagues is the most frequently cited.
The Speech Test app has been developed in collaboration with Professor David Howard with the aim of providing an easy-to-use non-invasive instrument for the measurement of male puberty. It uses the vocal task of counting backwards from twenty as a substitute for potentially embarrassing photographs. We were motivated by the potential health benefits.
- Speech Test can be used in large-scale medical screening and research, or simply the GP surgery to indicate whether the timing of a boy’s puberty is falling within healthy norms (precocious or delayed puberty is a not uncommon worry).
- In choral settings and the teaching of singing, a significant threat to vocal health arises through allocating boys to the wrong vocal part (most commonly singing too high for too long).
There are two main components to the App. First, an analytical algorithm detects and averages out the pitch of the voice over the task of counting from twenty, with an additional focus on the numbers twelve to five (when the voice is most settled). Second, a processor compares the result to the pitch ranges of each of the five pubertal stages as defined by the one thousand voices database. Tanner’s five puberty stages are represented by the different shirt colours of the cartoon characters.
Proving the validity, reliability and operating range of Speech Test
Validity asks the question does the instrument measure what it claims to? In other words, is the app really measuring pubertal stage? We can only know the answer to this by comparing the predictions made by the app with the actual clinical reference standard. So we are back to testicular volume! We have recently completed an important validation study in collaboration with Professor Gary Butler, a leading expert on adolescent endocrinology at London’s University College Hospital Paediatric Clinic. The results of the study have now been published as a poster presentation which you can see here. As with so much in research, answering one question simply leads to asking another. Our work on this topic broadly agrees with that of Mette Pedersen who is working in a similar area. Agreement seems to be coalescing around the importance of the 200 Hz speaking voice frequency as a key milestone in pubertal progression. Boys whose voices are at or below this pitch can safely be considered to have reached the “in-puberty” phase of life. Concerning singing technique and voice parts, important further work needs to be undertaken on the prepubertal and peripubertal phases.
Reliability asks the question will the instrument give the same result over a series of repeat tests? The reliability of the algorithm has already been tested by comparing Speech Test readings with other methods of pitch test analysis, principally the Praat voice analysis software and the reference standard of Laryngograph electroglottal measurement of vocal fold collision frequency. As a quick indication that the app is working correctly, the two readings (20-1 and 12-5) should not differ by more than 10 – 15 Hz.
The issue with any hand held device is distance from microphone and interference from extraneous noise. The app has an icon that flashes blue when the signal is sufficient, but the test still needs to be done in a quiet room. Obviously if another voice is picked up (noisy boys’ environments!) it will distort the result. We have recently been approached by a firm that is developing a telephone-based system for another medical application. We remained to be convinced about the quality of telephone signals but may have some exciting announcements in due course.
Tests of the app under strictly controlled conditions, with back-up recordings analysed by Praat, have revealed a matter of considerable interest. The boys themselves are not reliable! Although some singing teachers regularly pitch match counting by ear, accurate measurement by the app has shown that there is a considerable variation in voice-pitch associated with matters such as the boy’s mood and state of general health etc. Time of day also affects results, possibly a result of tiredness/excitement levels, but also perhaps a consequence of diurnal cycles of testosterone. We cannot yet say whether these variations are sufficient to invalidate tests done by singing teachers (or indeed other research). We will need to analyse the degree of variation and publish our results.
Operating Range. In theory the app is designed work across the five stages of puberty recognized for boys. Broadly, we might expect it to produce valid results when working across the age range 10 – 15 (in fact, it will correctly identify that any mature adult male has “completed puberty”!). It is possible to use the app simply as a tool for measuring the fundamental frequency of the speaking voice. Advanced users can interpret the output in Hz according to their own understanding.
Here are the current standards for adolescent growth that the app is aiming to enhance.
Also of interest
(Page updated October 2017)