Pitch, loudness and length are among the most salient of the properties of speech perceived by the listener. Linguists sometimes refer to these features as suprasegmental, suggesting that they are somehow ‘above’ the string of consonants and vowels. This connotation is misleading: rather, in speech, the string of consonants and vowels is overlaid onto a base of phonation. Phonation, or voicing, is generated by an airstream from the lungs passing through the larynx, which results in the perception of fluctuations in pitch and loudness distributed over chunks of varying durations. The term prosody, and the related adjective prosodic, commonly refer in a broad sense to features of pitch, loudness and duration in speech, encompassing their use in individual words and their component syllables, as well as the use of these features over longer stretches of speech, i.e. phrases, complete utterances, conversational turns. These longer stretches are the main focus of this book. Over such longer stretches, the meaningful patterning of pitch, along with related patterning of loudness and length features, is commonly referred to as intonation – hence the title of the book. While some authors restrict the use of this term to the patterning of pitch alone, we use it as a shorthand to refer to communicatively relevant systems, operating over stretches of speech consisting minimally of a single word but usually longer, which are realized primarily through features of pitch, also often of loudness and length, and sometimes of voice quality and articulation.
From the very beginning, long before words made up of consonants and vowels can be identified, infants produce the prosodic features that are used for intonation. Moreover, adult carers respond to them as signalling various kinds of meaning. Infants, in their turn, appear to react to the prosodic components of adults’ speech addressed to them; indeed, adults systematically exaggerate some of these prosodic features when addressing infants and young children (see Chapter 6). These observations appear to be true of all cultures that have been studied so far, although the degree of modification varies across cultures. It appears then, that from birth, prosodic features form a set of resources that the child can exploit for communicative purposes, even though in different mature adult languages, prosodic features turn out to be organized in a wide variety of ways. Some of the differences will be described later.
Some children’s intonation develops in ways that are unusual for their linguistic environment. The basis for postulating an impairment in intonation is likely to be the listener’s auditory impression that the child’s use of prosodic features is in some way different from that of the speech community and cannot be attributed to other causes, e.g. being a non-native speaker whose intonation in the second language is affected by the mother tongue. Beyond that, the identification, description and explanation of the impairment are influenced by the approach to analysis taken by the investigator, who may be a speech and language pathology practitioner or a researcher, for example. The basic tools of description are phonetic and linguistic. These complementary approaches will now be explained.
The phonetic approach
Typical and atypical prosodic development can be explored by using a range of methodologies that are based on auditory-perceptual and instrumental techniques. Instrumental techniques enable features of the speech signal to be recorded and measured reliably. Some of these techniques measure speech production directly. For example, using electrodes attached to the neck close to the larynx, the laryngograph (or electroglottograph) monitors the vibration of the vocal folds directly as they produce voicing (Abberton & Fourcin, 1997). While this can produce an accurate signal, and is not subject to interference from other noises (unlike a microphone recording), the need to wear a neckband attached by a wire to a computer militates against the recording of natural conversational speech, not least with young children.
For such reasons, the most common type of instrumental analysis used in intonation research is acoustic. An audio recording of the speakers is made via microphones. This can be done using digital audio or video tape recorders, a solid state digital audio recorder, or direct onto a PC or laptop. Conventional audio or video cassette recorders produce an analogue recording, which can subsequently be digitized using special software. The resulting digital audio files can then be analysed using one of the many computer speech analysis packages available. A much-used and freely downloadable program for acoustic analysis is Praat (Boersma & Weenink, 2014), which can be used to display and measure prosodic parameters.
However, acoustic analysis has its limitations too. The first, already mentioned, is that as soon as recording moves outside the artificial environment of the speech laboratory into children’s everyday talking environments, for example, home and school, then there will be sources of noise that may be difficult to control – traffic, clattering of play bricks, and so on. It can be difficult and time-consuming to remove such noises from the recorded signal. One particularly important and tricky source of noise is voices of speakers involved in background conversations or even in the same conversation: in a review of the occurrence of overlapping talk, Kurtić, Brown, and Wells (2013) report that, even in relatively formal meetings, for up to 10% of the time, two or more speakers will be talking at the same time and that, in spontaneous adult conversation, up to 45% of all changes of turn contain overlap between the speakers. Overlap is thus a natural part of human spoken interaction; it can only be suppressed by using an artificial speech elicitation task such as a monologue or reading aloud. Although speech research is making progress in developing techniques for separating out the voices of different talkers in overlap, at present, the only satisfactory procedure is to record talkers on separate channels, with each talker using a close-fitting microphone attached to a headset. While this has been done successfully when recording adults in meetings, it is too invasive for research with young children. A commonly adopted response to this issue in child language research has been to omit instances of overlapping talk from analysis. However, this is not a satisfactory solution when analysing intonat...