How to Measure Voice Onset Time (VOT) in Praat

Stop production involves four stages providing acoustic cues: closing, closure, release, and opening. Release bursts resemble fricatives at the same place—bilabial/labiodental have diffuse spectra, alveolar lack low-frequency energy, velar show intensity peaks at 1500-2000Hz. During closure, only low-frequency voicing bars escape for voiced stops. Formant transitions reveal place: bilabial transitions decrease F2, alveolar transitions show locus convergence around 1600-1700Hz, and velar transitions exhibit F2-F3 convergence (velar pinch).

This section covers digital audio processing: Nyquist-Shannon sampling theorem (sampling frequency must be ≥2× highest frequency to avoid aliasing), sampling (temporal discretization) and quantization (amplitude discretization). Spectrogram analysis methods: cepstral analysis shows horizontal stripes (better frequency resolution for formant identification), time-frequency analysis shows vertical stripes (better temporal resolution). VOT (Voice Onset Time) distinguishes voiced from voiceless stops: 'd' has VOT ≈ -70 ms (negative = voiced). Stop consonants are identified by formant transitions from closure to vowel position.

Stop consonants produce acoustic signatures including closure periods of near silence, release transients, and aspiration noise in voiceless stops. Aspiration represents a delay between stop release and vowel onset, absent in voiced stops which may show continuous voicing through the closure. Closure duration correlates with voicing: voiced stops have shorter closures than voiceless stops. Spectrograms reveal these patterns: voiceless stops show near-silence followed by aspiration bursts, while voiced stops display continuous voice bars through closures. Combining fricative and stop analysis enables word identification. For example, a word beginning with voiceless palatal fricative /ʃ/ and ending with voiceless velar stop /k/ would be identified as 'shock', with vowel /a/ showing characteristic formant values and voice bar.

This lecture explains how acoustic analysis of consonants uses spectrograms to identify three key properties: manner of articulation (shown by release bursts and noise patterns), place of articulation (revealed by formant transitions), and voicing (indicated by voicing bars). Stops appear as gaps with release bursts, fricatives show noise regions, and nasals display weak formants with low F1. Formant transitions—particularly F2 and F3 movements—provide crucial cues for determining place of articulation, such as the distinctive 'velar pinch' for velar consonants. The source-filter theory underlies these observations, where the glottal source generates sound and the vocal tract filter shapes it into recognizable consonant sounds.

This comprehensive section covers vowel acoustics and stop consonant identification. Vowels are characterized by formant patterns: front vowels show F2 and F3 close together, back vowels show F1 and F2 merged, with high vowels having lower F1 and low vowels having higher F1. Stop consonants display three key signatures: closure (near-zero energy during articulatory closure), burst (impulse-like spectrum at release), and onset (immediate formant appearance). Voiced stops show periodic voicing bands during closure, while aspirated stops show delayed vowel onset with intervening noise. Formant transitions during closure indicate place: bilabial stops show F1 and F2 rising/falling symmetrically, dental/alveolar stops show opposing F1 and F2 movements, and retroflex stops show distinctive F3 involvement.

Praat is a widely-used software for speech and sound acoustics analysis across Linguistics, Phonetics, Hearing Sciences, Music Science, Psychology, and Computer Science. Students must download it from pr.org, selecting their operating system. Audio files for exercises are available in the course Canvas portal under the 'Pro' folder, and should be downloaded simultaneously using Shift-click selection. Sound files open through drag-and-drop or Control+O navigation. The Object List displays opened sounds as temporary copies. The View and Edit button reveals the waveform (raw pressure deviations over time) and spectrogram (frequency content over time). Time information appears with three decimal places for millisecond precision. Clicking and dragging creates selection boxes displaying start time, end time, and duration. Periodic signals like vowels show equally spaced vertical stripes in the waveform, representing repeated glottal cycles. The fundamental frequency equals the reciprocal of the period duration—for a 7.5ms period, this is approximately 133 Hz. Pitch represents the rate of repetition in periodic signals, displayed as a contour showing fundamental frequency over time. Pitch range settings should match the speaker's actual range (typically 70-200 Hz for adult males) for optimal visualization.

Praat is a free speech analysis software downloadable from Google, available for Windows and Mac. After installation, users access two main windows: a display window (which can be closed) and a 'Praat objects' window for file management. Files are opened via 'Open' > 'Read from file'. Spectrograms are viewed by selecting sounds and choosing 'View & Edit', displaying time horizontally and frequency vertically with intensity shown through color gradients. Zooming is achieved by selecting regions and pressing Ctrl+N. Plosive releases appear as vertical bars in spectrograms, with white spaces indicating complete closures before release.

This tutorial demonstrates how to zoom in and out of audio waveforms and spectrograms in Praat software, including how to highlight specific audio segments, play portions before or after highlights, and navigate between different zoom levels using the select, back, and all buttons.

This tutorial demonstrates how to measure Voice Onset Time (VOT) in Praat by opening a .wav file and a .TextGrid file together, viewing the waveform and spectrogram, adding boundaries on Tier 1 to mark the start and end of the VOT interval, and reading the measurement from the time display at the top of the waveform window.

To open audio files in Praat for acoustic analysis, click the 'Read' menu and select 'Read from file' (or use Command+O/Ctrl+O), navigate to your desired file, and select it; the file will appear in the Objects window where you can highlight it to play or analyze it, and you can select multiple files by holding the Shift key while clicking.

Phonation involves four interconnected components: the power source (diaphragm, abdominal muscles, and intercostal muscles that generate expiratory airflow), the voice generator (vocal folds that vibrate when air passes through them), resonators (pharynx, nasopharynx, oropharynx, nasal cavity, and oral cavity that amplify and modify sound), and articulators (lips, tongue, and soft palate that shape sound into recognizable speech); vocal fold vibration follows a myoelastic aerodynamic mechanism where Bernoulli's principle causes mucosal waves traveling from inferior to superior at 0.52-1 m/s, with pitch determined by vocal fold mass, tension, and length, and voice quality modulated by three registers (falsetto, modal, and pulse) controlled by neural pathways involving the central and peripheral nervous systems.

Phonation is a three-component process involving a vibrator (vocal folds), an actuator (expiratory airflow), and a resonator (supraglottic tract including throat, oral cavity, and nasal cavity). The fundamental frequency is determined by vocal fold properties: mass (inversely proportional to pitch), length (longer folds produce lower pitch), and tension (increased tension raises pitch). Vocal fold vibration follows a cyclic pattern of opening (when subglottal pressure overcomes threshold), closing (due to Bernoulli effect and elastic recoil), and recoil. The myoelastic aerodynamic theory explains that vibration results from aerodynamic forces interacting with vocal fold elasticity, not from rhythmic nerve impulses as previously thought. Formants are the selectively amplified overtones that give voice its characteristic quality, shaped by the resonating chambers and controlled by articulatory structures like the tongue.

Phonation generates the vibrating glottal pulse source for voiced speech. Vocal folds are paired muscular structures that tense and vibrate when air passes through, creating the Bernoulli effect-driven opening/closing cycles. Tension is controlled by thyroarytenoid (adduction) and cricothyroid (stretching) muscles, with higher tension producing higher pitch. During phonation, vocal folds create periodic pulses serving as the fundamental source signal. Whispering lacks vocal fold vibration entirely. This vibration mechanism is essential for all voiced speech sounds and can be artificially induced using external vibration sources for those unable to produce natural vocal fold vibration.

Vocal fold vibration during phonation is produced by the mutual resistance between subglottic air pressure and the elastic force of the vocal fold mucosa, creating a seesaw-like cycle where air pressure pushes the folds apart and elastic recoil brings them back together; this fundamental frequency is determined by vocal fold mass, length, and tension, with the thyroarytenoid muscle controlling low pitch and the cricothyroid muscle controlling high pitch through their opposing actions on vocal fold length and tension.

Voicing is the third crucial feature for describing human speech sounds, referring to vocal cord vibration during production. Speech begins at the lungs where the diaphragm and ribcage pump air upward through the trachea to the larynx. The larynx houses the vocal cords, which can be positioned spread apart (wide glottis for breathing) or brought together with tension. When vocal cords are held together, air pressure pushes them apart, and their elasticity causes them to snap back together, creating vibration cycles. High-frequency cycles produce voicing. Lip buzzing provides an analogy: lips pressed together with tension and air pressure vibrate, similar to how vocal cords function as the sound source while the vocal tract modifies the sound.