1. History and Development of Phonetics

Much of phonetic structure is available to direct inspection or introspection, allowing a long tradition in phonetics (see also articles in Asher & Henderson, 1981). The first true phoneticians were the Indian grammarians of about the 8th or 7th century bce. In their works, called Pratiśãkhya, they organized the sounds of Sanskrit according to places of articulation, and they also described all the physiological gestures that were required in the articulation of each sound. Every writing system, even those not explicitly phonetic or phonological, includes elements of the phonetic systems of the languages denoted. Early Semitic writing (from Phoenician onward) primarily encoded consonants, while the Greek system added vowels explicitly. The Chinese writing system includes phonetic elements in many, if not most, characters (DeFrancis, 1989), and modern readers access phonology while reading Mandarin (Zhou & Marslen-Wilson, 1999). The Mayan orthography was based largely on syllables (Coe, 1992). All of this required some level of awareness of phonetics.

Attempts to describe phonetics universally are more recent in origin, and they fall into the two domains of transcription and measurement. For transcription, the main development was the creation of the International Phonetic Alphabet (IPA) (e.g., International Phonetic Association, 1989). Initiated in 1886 as a tool for improving language teaching and, relatedly, reading (Macmahon, 2009), the IPA was modified and extended both in terms of the languages covered and the theoretical underpinnings (Ladefoged, 1990). This system is intended to provide a symbol for every distinctive sound in the world’s languages. The first versions addressed languages familiar to the European scholars primarily responsible for its development, but new sounds were added as more languages were described. The 79 consonantal and 28 vowel characters can be modified by an array of diacritics, allowing greater or lesser detail in the transcription. There are diacritics for suprasegmentals, both prosodic and tonal, as well. Additions have been made for the description of pathological speech (Duckworth, Allen, Hardcastle, & Ball, 1990). It is often the case that transcriptions for two languages using the same symbol nonetheless have perceptible differences in realization. Although additional diacritics can be used in such cases, it is more often useful to ignore such differences for most analysis purposes. Despite some limitations, the IPA continues to be a valuable tool in the analysis of languages, language use, and language disorders throughout the world.

For measurement, there are two main signals to record, the acoustic and the articulatory. Although articulation is inherently more complex and difficult to capture completely, it was more accessible to early techniques than were the acoustics. Various ingenious devices were created by Abbé Rousselot (1897–1908) and E. W. Scripture (1902). Rousselot’s devices for measuring the velum (Figure 1) and the tongue (Figure 2) were not, unfortunately, terribly successful. Pliny Earl Goddard (1905) used more successful devices and was ambitious enough to take his equipment into the field to record dynamic air pressure and static palatographs of such languages as Hupa [ISO 639-3 code hup] and Chipewyan [ISO 639-3 code chp]. Despite these early successes, relatively little physiological work was done until the second half of the 20th century. Technological advances have made it possible to examine muscle activity, airflow, tongue-palate contact, and location and movement of the tongue and other articulators via electromagnetic articulometry, ultrasound, and real-time magnetic resonance imaging (see Huffman, 2016). These measurements have advanced our theories of speech production and have addressed both phonetic and phonological issues.

Open in new tab

Figure 1. Device to measure the height of the velum. From Rousselot (1897–1908).

Open in new tab

Figure 2. Device to measure the height of the tongue from external shape of the area under the chin. From Rousselot (1897–1908).

Acoustic recordings became possible with the Edison disks, but the ability to measure and analyze these recordings was much longer in coming. Some aspects of the signal could be somewhat reasonably rendered via flame recordings, in which photographs were taken of flames flickering in response to various frequencies (König, 1873). These records were of limited value, because of the limitations of the waveform itself and the difficulties of the recordings, including the time and expense of making them. Further, the ability to see the spectral properties in detail was greatly enhanced by the declassification (after World War II) of the spectrograph (Koenig, Dunn, & Lacy, 1946; Potter, Kopp, & Green, 1947). New methods of analysis are constantly being explored, with greater accuracy and refinement of data categories being the result.

Sound is the most obvious carrier of language (and is etymologically embedded in “phonetics”), and the recognition that vision also plays a role in understanding speech came relatively late. Not only do those with typical hearing use vision when confronted with noisy speech (Sumby & Pollack, 1954), they can even be misled by vision with speech that is clearly audible (McGurk & MacDonald, 1976). Although the lips and jaw are the most salient carriers of speech information, areas of the face outside the lip region co-vary with speech segments (Yehia, Kuratate, & Vatikiotis-Bateson, 2002). Audiovisual integration continues as an active area of research in phonetics.

Sign language, a modality largely devoid of sound, has also adopted the term “phonetics” to describe the system of realization of the message (Goldin-Meadow & Brentari, 2017; Goldstein, Whalen, & Best, 2006). Similarities between reduction of speech articulators and American Sign Language (ASL) indicate that both systems allow for (indeed, may require) reduction in articulation when content is relatively predictable (Tyrone & Mauk, 2010). There is evidence that unrelated sign languages use the same realization of telicity, that is, whether an action has an inherent (“telic”) endpoint (e.g., “decide”) or not (“atelic”, e.g., “think”) (Strickland et al., 2015). Phonetic constraints, such as maximum contrastiveness of hand shapes, has been explored in an emerging sign language, Al-Sayyid Bedouin Sign Language (Sandler, Aronoff, Meir, & Padden, 2011). As further studies are completed, we can expect to see more insights into the aspects of language realization that are shared across modalities, and to be challenged by those that differ.

2. Phonetics in Relation to Phonology

Just as phonetics describes the realization of words in a language, so phonology describes the patterns of elements that make meaningful distinctions in a language. The relation between the two has been, and continues to be, a topic for theoretical debate (e.g., Gouskova, Zsiga, & Boyer, 2011; Keating, 1988; Romero & Riera, 2015). Positions range from a strict separation in which the phonology completes its operations before the phonetics becomes involved (e.g., Chomsky & Halle, 1968) to a complete dissolution of the distinction (e.g., Flemming, 2001; Ohala, 1990). Many intermediate positions are proposed as well.

Regardless of the degree of separation, the early assumption that phonetic implementation was merely physical and universal (e.g., Chomsky & Halle, 1968; Halliday, 1961), which may invoke the mind-body problem (e.g., Fodor, 1981), has proven to be inadequate. Keating (1985) examined three phonetic effects—intrinsic vowel duration, extrinsic vowel duration, and voicing timing—and found that they were neither universal nor physiologically necessary. Further examples of language- and dialect-specific effects have been found in the fine-grained detail in Voice Onset Time (Cho & Ladefoged, 1999), realization of focus (Peters, Hanssen, & Gussenhoven, 2014), and even the positions of speech articulators before speaking (Gick, Wilson, Koch, & Cook, 2004). Whatever the interface between phonetics and phonology may be, there exist language-specific phonetic patterns, thus ensuring the place of phonetics within linguistics proper.

The overwhelming evidence for “language-specific phonetics” has prompted a reconsideration of the second traditionally assumed distinction between phonology and phonetics: phonetics is continuous, phonology is discrete. This issue has been raised relatively less often in discussions of the phonetics-phonology interface. An approach to phonology that has addressed the need to reconsider the purely representational phonological elements in a logical compliance with their physical realization is Articulatory Phonology, whose elements have been claimed to be available for “public” (phonetic) use and yet categorical for making linguistic distinctions (Goldstein & Fowler, 2003). Iskarous (2017) shows how dynamical systems analysis unites discrete phonological contrast and continuous phonetic movement into one non-dualistic description. Gafos and his colleagues provide the formal mechanisms that use such a system to address a range of phonological processes (Gafos, 2002; Gafos & Beňuš, 2006; Gafos, Roeser, Sotiropoulou, Hoole, & Zeroual, 2019). Relating categorical, meaningful distinctions to continuous physical realizations will continue to be developed and, one hopes, ultimately be resolved.

3. Phonetics in Relation to Other Aspects of Language

Phonetic research has had far-reaching effects, many of which are outlined in individual articles in this encyclopedia. Here are four issues of particular interest.

Perception: Until the advent of an easily manipulated acoustic signal, it was very difficult to determine which aspects of the speech signal are taken into account perceptually. The Pattern Playback (Cooper, 1953) was an early machine that allowed the synthesis of speech from acoustic parameters. The resulting acoustic patterns did not sound completely natural, but they elicited speech percepts that allowed discoveries to be made, ones that have been replicated in many other studies (cf. Shankweiler & Fowler, 2015). These findings have led to a sizable range of research in linguistics and psychology (see Beddor, 2017). Many studies of brain function also take these results as a starting point.

Acquisition: Learning to speak is natural for neurologically typical infants, with no formal instruction necessary for the process. Just how this process takes place depends on phonetic findings, so that the acoustic output of early productions can be compared with the target values in the adult language. Whether or not linguistic categories are “innate,” the development of links between what the learner hears and what s/he speaks is a matter of ongoing debate that would not be possible without phonetic analysis.

Much if not most of the world’s population is bi- or multilingual, and the phonetic effects of second language learning have received a great deal of attention (Flege, 2003). The phonetic character of a first language (L1) usually has a great influence on the production and perception of a second one (L2). The effects are smaller when L2 is acquired earlier in life than later, and there is a great deal of individual variability. Degree of L2 accent has been shown to be amenable to improvement via biofeedback (d’Apolito, Sisinni, Grimaldi, & Gili Fivela, 2017; Suemitsu, Dang, Ito, & Tiede, 2015).

Sociolinguistics: Phonetic variation within a language is a strong indicator of community membership (Campbell-Kibler, 2010; Foulkes & Docherty, 2006). From the biblical story of the shibboleth to modern everyday experience, speech indicates origin. Perception of an accent can thus lead to stereotypical judgements based on origin, such as assigning less intelligence to speakers who use “-in” rather than “-ing” (Campbell-Kibler, 2007). Accents can work in two directions at once, as when an African American dialect is simultaneously recognized as disfavored by mainstream society yet valued as a marker of social identity (Wolfram & Schilling, 2015, p. 238). The level of detail that is available to and used by speakers and listeners is massive, requiring large studies with many variables. This makes sociophonetics both exciting and challenging (Hay & Drager, 2007).

Speech therapy: Not every instance of language acquisition is a smooth one, and some individuals face challenges in speaking their language. The tools that are developed in phonetics help with assessment of the differences and, in some cases, provide a means of remediation. One particularly exciting development that depends on articulation rather than acoustics is the use of ultrasound biofeedback (using images of the speaker’s tongue) to improve production (e.g., Bernhardt, Gick, Bacsfalvi, & Ashdown, 2003; Preston et al., 2017).

Speech technology: Speech synthesis was an early goal of phonetic research (e.g., Holmes, Mattingly, & Shearme, 1964), and research continues to the present. Automatic speech recognition made use of phonetic results, though modern systems rely on more global treatments of the acoustic signal (e.g., Furui, Deng, Gales, Ney, & Tokuda, 2012). Man-machine interactions have benefited greatly from phonetic findings, helping to shape the modern world. Further advances may begin, once again, to make less use of machine learning and more of phonetic knowledge.

4. Future Directions

Phonetics as a field of study began with the exceptional discriminative power of the human ear, but recent developments have been increasingly tied to technology. As our ability to record and analyze speech increases, our use of larger and larger data sets increases as well. Many of those data sets consist of acoustic recordings, which are excellent but incomplete records for phonetic analysis. Greater attention is being paid to variability in the signal, both in terms of covariation (Chodroff & Wilson, 2017; Kawahara, 2017) and intrinsic lack of consistency (Tilsen, 2015; Whalen, Chen, Tiede, & Nam, 2018). Assessing variability depends on the accuracy of individual measurements, and our current automatic formant analyses are known to be inaccurate (Shadle, Nam, & Whalen, 2016). Future developments in this domain are needed, but current studies that use current techniques must be appropriately limited in their interpretation.

Large data sets are rare for physiological data, though there are some exceptions (Narayanan et al., 2014; Tiede, 2017; Westbury, 1994). Quantification of articulator movement is easier than in the past, but it remains challenging in both collection and analysis. Mathematical tools for image processing and pattern detection are being adapted to the problem, and the future understanding of speech production should be enhanced. Although many techniques are too demanding for some populations, ultrasound has been found to allow investigations of young children (Noiray, Abakarova, Rubertus, Krüger, & Tiede, 2018), speakers in remote areas (Gick, Bird, & Wilson, 2005), and disordered populations (Preston et al., 2017). Thus the amount of data and the range of populations that can be measured can be expected to increase significantly in the coming years.

Understanding the brain mechanisms that underlie the phonetic effects being studied by other means will continue to expand. Improvements in the specificity of brain imaging techniques will allow narrower questions to be addressed. Techniques such as electrocorticographic (ECoG) signals (Hill et al., 2012), functional near-infrared spectroscopy (Yücel, Selb, Huppert, Franceschini, & Boas, 2017), and the combination of multiple modalities will allow more direct assessments of phonetic control in production and effects in perception. As with other levels of linguistic structure, theories will be both challenged and enhanced by evidence of brain activation in response to language.

Addressing more data allows a deeper investigation of the speech process, and technological advances will continue to play a major role. The study does, ultimately, return to the human perception and production ability, as each newly born speaker/hearer begins to acquire speech and the language it makes possible.