Sound Spectrography

This chapter aims at providing the reader with a short and comprehensive introduction to the spectrographic analysis of speech and its applications in clinical linguistics. Since the development of the first sound spectrograph, also called sonograph, in the 1940s (Koening et al., 1946), the rapid change in both instruments for data collection and sound analysis have greatly improved the potential for acoustic analysis of speech signals, enhancing our understanding of speech production and perception also in relation to sound change (Ohala, 1993). A radical innovation in spectrography has been introduced with digital signal processing (Farmer, 1997:, p. 22) and the development of different software for digital audio analysis. Fujimura & and Erickson (1997) recall that Potter et al. (1966) developed the first visual representation of speech: Visible Speech displayed sound as a two-dimensional time-frequency function called a spectrogram. Nowadays, there are many software available to perform spectrographic or digital audio analysis (e.g. Matlab, Python, C Speech, Speech Analyzer, Voice Vista, just to mention a few). In this chapter we will refer in particular to PRAAT (Boersma & Weenink, 2020) [AQ4]which is an open resource available for Windows, Macintosh, Chromebook, Raspberry Pi and Linux, with a large community of users online composed by both linguists and engineers. All the examples presented in the following sections were recorded by the author with a TASCAM DR-20 recorder, at a frequency of 44.1 kHz with a sampling rate of 16 bit (see Vogel & Reece, chapter Chapter 18, this volume); visualization and annotation have be performed in PRAAT 6.0.37.