Lodewijk Muns

Tones and glides

Everyone is familiar with speech melody in practice, even if few people ever pay attention to it. Upon closer inspection, it turns out to be a challenging and complex phenomenon.

The study of speech melody or intonation dates back at least to the Greek philosopher and music theorist Aristoxenos of Tarentum, who lived in the fourth century BC and was a student of Aristotle. Aristoxenos provides what is, as far as we know, the earliest formulation of an insight that many people have gained spontaneously. It involves a difference between speaking and singing.

Aristoxenos speaks of melodic progression in terms of ‘movement’ [1]. That movement is either continuous, gliding from tone to tone, or stepwise. Stepwise movement is that of singing, continuous movement that of speaking. When singing, we hold each tone and try to make the transitions between them inaudible.

The voice does not function like an organ or piano, with a different pipe or string for each note. The vocal cords are tensed or relaxed with each change in pitch. When singing, it takes effort to distinguish the notes clearly without losing continuity. Gliding or sliding cannot be completely avoided; it is also used deliberately, but (in the European tradition) only occasionally, as an expressive device, portamento.

Calling, exclaiming, declaiming

It is true, Aristoxenos notes in passing, that when we speak with great emotion and emphasis, we raise our voice and produce more distinct pitches. This can also happen when we are not speaking emotionally, but need to make ourselves heard over a long distance. When we call out to someone, Come over here, our call may come close to singing.

The situation is similar when we speak to a large audience without amplification. We ‘raise our voice’ in both senses of the word by speaking louder and higher. A larger space requires more breath support, a slower tempo, stronger articulation, and often a higher intonation. Every stage actor (and in the pre-electrical past, orator) must develop techniques for breath and voice control. At a slower tempo, individual pitches will become more prominent.

Say loudly and with great enthusiasm: But this is beautiful! Then repeat it without the words, humming only the melodic contour m-m-m-m-m-m. The result may be an expressive little melody, a musical phrase that you could more or less imitate on a musical instrument. This is the informal experiment proposed in 1796 by music critic Johann Friedrich Rochlitz [22]. He transcribes (in a footnote, p. 154) the exclamation Wie ich mich freue! (How happy I am!) in ¾ time with this melody:

Strange behaviour

It is generally an illusion that we can capture speech in musical notation. Nevertheless, there is a long and largely unwritten history of attempts in this direction.

Outside that tradition, but relevant to it, there is the strange phenomenon known as the Speech-to-Song Illusion (henceforth STS); I will refer to it as an ‘effect’ rather than an ‘illusion’, for reasons explained below. It was discovered by chance in 1995 by music psychologist Diana Deutsch [3, 4, 5, 6, 7]. She played a recording of her own voice containing the following sentence:

The sounds as they appear to you are not only different from those that are really present, but they sometimes behave so strangely as to seem quite impossible.

It is curious that precisely the words “sometimes behave so strangely” had an unforeseen and strange effect when they were repeated in a tape loop.

Diana Deutsch: “The sounds as they appear to you…” (1995) [3]

It is very likely that during these repetitions you began to hear the words as sung rather than spoken, with a melody that was not so obvious at first. Most people experience this effect, often after only three repetitions [18]. If listeners are asked to repeat the melody, they do so singing, and with a high degree of consistency; the melody they heard is therefore the same.[6] So here we have an instance of speech melody that – after some repetitions – is so obvious that it can be adequately represented in musical notation (Ex. 2).

The effect is, of course, not limited to this particular spoken phrase; but neither does it occur with any utterance. In fact, it is rather rare [11]. You have probably noticed that in this case a text is read aloud, and the delivery is somewhat more declamatory than one would expect in conversation.

Over the past twenty years, various aspects of this effect have been extensively (though not exhaustively) studied: cognitive and neurological processing; the prosodic properties of the signal; linguistic semantics; the conditions of repetition; memory and persistence of the effect over time; individual differences; language-related differences; and the comparison with repetitive transformations of sounds other than speech.

As you might expect, the effect is stronger with an unfamiliar language [16, 19]. When you don’t understand the meaning, you are forced to guess the speaker’s intention by focusing on intonation. Predictably, also, ‘tonal’ languages like Mandarin, in which pitch has a semantic value, are less susceptible to the effect [13].

Uncertainties remain. Some are of a conceptual rather than factual nature, and different terminologies (linguistic, acoustic, musical) sometimes make it difficult to compare these studies.

Melody

What interests me most is the involvement of our linguistic capacity in making music, and how this somewhat bizarre effect can shed light on the musical qualities of speech and the speech-like qualities of music.

Like the original publication by Deutsch et al. [7], subsequent studies were based on reading texts aloud rather than spontaneous conversation. This is a limitation which the researchers seem to have simply accepted. On the whole, they confirm what one might expect: a key requisite for the effect is the presence of relatively stable (level) pitches and regular rhythm, producing a clear melodic contour [8, 12, 16, 18, 19, 21, 26, 28, 29]. Surprisingly, the correspondence between the spoken pitches and a musical key proves to be of only marginal significance for the effect [8, 29], although there seems to be some disagreement or misunderstanding about this [9, 10, 11, 18].

‘Pitch’ is a somewhat problematic concept in relation to speech. Apart from being generally unstable, the fundamental frequency (F₀) of any speech sound, what we identify as ‘pitch’, may be less noticeable to our perception than its overtone spectrum. It is this spectrum (together with pitched and non-pitched consonantal sounds) that is responsible for articulation, and therefore contains information crucial to our understanding. We produce different vowels by adjusting the vocal tract, a capability that most musical instruments do not have: one can change the length of the vibrating string or tube, but not the shape of the resonating body. Which is what we do when we say aa, ee, oo. Pitch may therefore not be the focus of attention.

Intonation however does have a function in making speech understandable, primarily by giving audible form to the syntactic structure of the sentence: we distinguish between main clauses and subsidiary clauses, closure and nonclosure, etc. Punctuation in writing mimics much of the function of intonation.

Repetition and rhythm

In music, metre is the most basic instance of repetition: it is an accent or beat that recurs (typically) at equal intervals.

Speech rhythm, a regular distribution of accents, also is a factor in the STS effect, which is not too surprising [8, 12, 29]. Sometimes behave so strangely can without ambiguity be rendered in musical metre and rhythm. (I have chosen 2/4 rather than 4/4, because it gives a better suggestion of tempo.) The phrase could be a line from a song lyric. The regular poetic metre of traditional poetry is an adaptation of language that makes it more suitable for singing. What in poetry is called ‘metre’, the regular pattern of accents, corresponds primarily to rhythm in music. Musical metre is a regularity on a higher level.

What is surprising (and, as far as I know, unexplained), is that the regular beat that results from exact repetitions seems to be irrelevant to the effect [8, 16]. In other words, when we vary the intervals between repetitions, this does not diminish the effect, even though it destroys metric regularity.

Repetition, generally, is a way of defining a pattern. Even an arbitrary sequence of sounds (or visual elements) becomes a recognisable pattern through repetition. For my podcast, I made this compilation of mostly pitchless sounds. In this case too, in my experience, the two vaguely pitched events become more pronounced during the repetitions.

noises, looped

Most listeners, I suspect, will have the feeling that in this example a boundary is crossed between mere noise and music. In connection with similar (more formal) experiments, it has been proposed that STS is an instance of a more general repetition-to-music effect [23, 24]. However, that misses the point of STS: that we hear a clear melody where before we didn’t; and if we call it ‘music’, it is for this reason.

The STS effect could thus be generalised beyond speech to a pitch- and melody-enhancing effect of repetitive sound patterns. What counts as ‘music’, on the other hand, is a matter of aesthetic judgment, not of perception, and involves matters of intention and function. The ‘musicality’ of a sound excerpt can be rated only subjectively, not objectively measured.

Repetition occurs in music in many ways, at various levels, and with various functions. The regularity of metre should perhaps be regarded as its most basic manifestation. Some musical patterns can go on indefinitely. In Classical Music, such patterns are subjected to changes of harmony, and typically serve as a backdrop for a more eventful foreground. At these higher levels, we find repetition of melodic and harmonic patterns, motifs, phrases, and whole sections. One can choose to create music without any repetition or regularity, but the overwhelming majority of ‘musics’ is repetitive to a considerable degree (and much present day popular music excessively so).

In terms of function, the most basic and universal is probably that of maintaining movement, simply to ‘keep going’. This seems to connect music more with dance and physical movement than with language. Humans have a peculiar fondness for periodic movement that is stimulated and sustained by regular sound patterns. In that sense, almost all music is a kind of lullaby.

However, there are functions for repetition in language. Some are ritualistic in nature, from incantations to chanting slogans (perhaps also a kind of incantation). Perhaps one could say that in such uses language is closer to music; or rather, that it shares qualities and functions with music.

Other functions are rhetorical; and to the extent that music has a rhetorical or discursive character (as is often the case with Classical Music in the narrower sense), such functions can be mimicked by music. The more discursive or rhetorical it is, the more a musical phrase has the character of a ‘statement’, the less repeatable it is, and conversely, the stronger the urge to find meaning in its repetition. A typical musical statement (think of the ‘fanfare’ or ‘motto’ openings of sonatas, symphonies and overtures) arouses our expectation that something new will follow, possibly complementary or contrasting. The repetition of such a statement gives it emphasis and heightens the expectation, just as saying something twice is usually a matter of emphasis.

That is a rhetorical strategy, and even if repetition is more pervasive in music than in language, I see no reason why repetition in language should generally make it more like music, or why repetition “should be most essentially connected with the functions and practice of music” [15, p. 180]. There is repetition in almost anything we do, from breakfast routines to memorisation techniques.

Making sense of nonsense

In spoken communication, it generally makes no sense to repeat a sentence or phrase many times. Speech is normally disposable: once the message is understood, its carrier can be discarded. The repetitive use of speech in the STS setup is an unusual, one might say unnatural use of language, closest maybe to chanting slogans, but very much unlike it in other respects. It thus challenges our human urge to make sense of what we hear. We can’t make linguistic-pragmatic sense of a sentence fragment spoken identically ten times. It is not what we do with language.

We might therefore explain the effect in terms of a sense-making strategy, in which we shift our attention from semantics to sound patterns.

It has been pointed out [5, 24] that the effect was already musically exploited before 1995 in tape loop compositions of the Minimalist school, such as Steve Reich’s It’s gonna rain (1965). It is indeed hard to believe that no one noticed the increasing tunefulness of the (declamatory) speech excerpts, although this seems not to have been a compositional strategy in the montage. It remains remarkable that this was not explicitly pointed out as a specific effect.

For a speech fragment such as sometimes… to make sense musically, we (unconsciously) try to interpret it in accordance with the idiom we are most familiar with. For the vast majority of listeners, this is major-minor tonality and rhythm with regular subdivisions. We involuntarily reshape what we hear – to some extent – according to musical patterns of expectation, the patterns for which musical notation was developed. (It is regrettable that the studies following Deutsch’s initial presentations do not make use of musical notation; it could be helpful in studying the preferred patterns of this ‘musicalisation’ of speech.)

Such reshaping also takes place when we listen to actual music. The integrity of the pattern can tolerate small, unnoticed deviations, or, under certain circumstances, larger ones, as when a rhythm appears to be constant during tempo changes. It also allows us to recognise ‘what it should be’ when a piece is played badly.

This is an area where musical expertise seems to have contradictory effects: musically educated or experienced listeners are more critical of intonation, and therefore less susceptible to the STS effect [8, 11, 21], while on the other hand their larger mental store of patterns enables them to find a musical match more easily [21].

Sometimes behave so strangely clearly implies such a pattern, a familiar cadential turn with its implied harmony (Ex. 3A). Deutsch herself relates it to two patterns: the pitch contour corresponds to the Westminster chimes, the rhythm to that of Rudolph the Red-Nosed Reindeer [5]. It seems that the continuation of the sentence (as to seem quite impossible) does not evoke a pattern strong enough to cancel its intonational ambiguity.

However, I find the unanimity about the tonal interpretation of sometimes… somewhat surprising, since for me the last pitch (which in the pitch tracing appears unstable [D11]) is E# rather than F#, thus harmonically implying the key of F# major (Ex. 3 B) instead of B major (Ex. 3 A). Looping the syllable -ly in fact produces a clear E#/F.

Ex. 3 A

Ex. 3 B

Is it singing? – an illusion?

What is most interesting about the STS effect is the extent to which our perception involuntarily imposes a musical pattern, making a spoken phrase sound very different – according to the general consensus, like singing. This is what makes it an illusion.

However, one might have some doubts about this characterisation as ‘song’. Part of what is strange about the effect is that although it sounds melodious, it hardly sounds like singing; it is still speech, albeit a weirdly tuneful kind of speech that is difficult to imitate.

Singing usually implies more than just vocal melody; it involves certain physiological conditions relating to breath, the vocal apparatus, and vowel production [12] that differ from those of speaking. Some or perhaps most of these are involved in the task of producing clear and distinct intonation when we sing.

Admittedly, this view or intuition stems from a cultural tradition that – since Aristoxenos – strictly distinguishes between speaking and singing. Intermediate forms have been attempted in the field of declamation (ranging from tonal declamation [17] to Schoenbergian Sprechgesang), but without much success.

Whether we should call this tuneful speech ‘song’ may be a matter of choice. But as long as we detect a significant difference from regular singing, it is perhaps not an illusion. Rather, it is an uncommon distortion of perception; but all perception involves some distortion of physical reality, without making it illusory.

---

---

Lodewijk Muns (piano), Impromptu on ‘sometimes behave so strangely’.