Acoustic Evidence for Phonetic Analysis
Chapter Objectives
By the end of this chapter, you will be able to:
- Orient a waveform and a spectrogram in time and frequency.
- Identify periodicity, silence or closure, release, and frication in controlled examples.
- Interpret duration and fundamental frequency from a labeled display.
- Interpret first and second formants at an introductory level.
- Recognize basic voice onset timing evidence.
- Connect acoustic patterns to articulatory hypotheses rather than treating them as free-standing facts.
- State clearly what a given acoustic display can and cannot establish.
An Observable Problem
Chapter 27 used waveforms and spectrograms the way a working transcriber actually uses them: as a check on the ear, consulted at step seven of a nine-step workflow, never as a substitute for listening. That chapter deliberately left the displays themselves unexplained. It asked you to trust that a closure, a burst, or a patch of frication energy could be located "in the display" without telling you what you were looking at or why.
This chapter fills that gap. It does not introduce a new transcription task. It introduces the small, defined set of acoustic concepts that let a waveform or spectrogram be read with precision instead of impression: what the axes represent, what periodic and aperiodic energy look like, how duration and pitch appear visually, and what a formant is.
This chapter asks three questions:
- What do the shapes on a waveform and a spectrogram actually represent, and how do they relate to articulation?
- Which phonetic properties, such as voicing, closure, frication, duration, pitch, and vowel quality, leave a visible trace that a trained reader can interpret?
- What must an analyst still decide using articulatory and auditory evidence, because no acoustic display decides it automatically?
Orienting a Waveform
Purpose: teach the waveform as a time-and-amplitude display, and connect its major visible patterns to articulatory events already familiar from earlier chapters.
A waveform plots amplitude, the relative loudness of the acoustic signal at each moment, against time. Time runs along the horizontal axis; amplitude, shown as the height of the wiggling line above and below a center resting position, runs along the vertical axis. A waveform contains no frequency information by itself: it shows only how loud the signal is at each instant, not which frequencies make it up.
Three patterns recur often enough to learn immediately. A flat or near-flat stretch, close to the center line, usually reflects silence or a very quiet interval, consistent with a stop closure or a pause. A tall, evenly repeating wave shape, in which one cycle looks much like the next, is periodic energy, typically produced by regular vocal-fold vibration during voicing. A tall but irregular, densely jagged stretch, with no clearly repeating cycle, is aperiodic energy, typically produced by turbulent airflow such as frication.
A common early confusion treats a tall waveform as high-pitched and a short waveform as low-pitched. Amplitude reflects loudness, not pitch. A loud, low-pitched vowel can produce tall waves that repeat slowly; a quiet, high-pitched vowel can produce short waves that repeat quickly. Pitch is read from how often a pattern repeats per second, not from how tall the waves are. This distinction matters directly for Section 28.4, where fundamental frequency is introduced.
Orienting a Spectrogram
Purpose: teach the spectrogram as a time-frequency-intensity display and connect its major visible patterns to articulation, building directly on the waveform concepts just introduced.
A spectrogram adds the dimension a waveform leaves out. Time still runs along the horizontal axis. Frequency, corresponding roughly to the auditory sensation of pitch height and to specific resonances in the vocal tract, runs along the vertical axis. Intensity, how much acoustic energy is present at a given time and frequency, is shown as darkness or color: a darker or more saturated region has more energy at that time and frequency than a lighter one.
A spectrogram is best read as a picture built from many overlapping frequency measurements taken across time, not as a single, simple curve. Regions with almost no visible darkening usually indicate silence or a very quiet interval, the spectrogram's version of the waveform's flat stretch. A dense, unstructured smear of darkening, especially concentrated at higher frequencies, usually indicates turbulent, aperiodic noise, consistent with frication. Dark, horizontally oriented bands that persist across time, changing gradually rather than abruptly, are formants, the resonance pattern most useful for vowel and approximant analysis.
Error: Describing a spectrogram as showing "the sounds themselves," as if the darkened shapes were pictures of individual IPA symbols waiting to be read off.
Correction: Describe a spectrogram as showing acoustic energy distributed across time and frequency. That distribution is evidence for a phonetic interpretation, produced by articulation and checked against listening, not a direct printout of the transcription.
Periodicity, Closure, Release, and Frication
Purpose: connect the visual patterns from Sections 28.2 and 28.3 to the consonant-manner categories established in Chapters 8 through 10, giving learners a concrete acoustic correlate for each.
Periodic energy, the evenly repeating pattern described above, is the acoustic signature of voicing: regular vocal-fold vibration produces a regularly repeating pattern in both the waveform and, as vertical striations, in the spectrogram. A stretch of speech with no periodic pattern and little or no amplitude is consistent with a voiceless interval or true silence, though only articulatory and distributional evidence can confirm which.
| Articulatory event | Waveform correlate | Spectrogram correlate |
|---|---|---|
| Voicing (Chapter 6) | Regular, repeating wave shape | Vertical striations and, for vowels, formant bands |
| Stop closure (Chapter 9) | Flat or near-silent stretch | Blank or near-blank region |
| Stop release / burst (Chapter 9) | Brief, sharp spike | Brief vertical burst of energy, often broadband |
| Frication (Chapter 10) | Dense, irregular, non-repeating noise | Diffuse, often high-frequency noise band |
| Nasal resonance (Chapter 11) | Lower-amplitude periodic wave | Weaker, more closely spaced formant bands |
An affricate, taught in Chapter 9 as a coordinated stop-plus-frication contour, shows this coordination directly: a silent or near-silent closure interval, a brief release, and an immediately following stretch of frication noise, all produced by one continuous articulatory gesture rather than by two unrelated sounds placed side by side.
The patterns in the table above are reliable tendencies, not exceptionless rules. A very short or lightly articulated stop may show barely any visible closure at conversational rate; background noise can partly obscure a frication band; a nasal's resonance pattern can resemble a vowel's on a casual glance. Acoustic correlates narrow the range of plausible interpretations; they do not remove the need for a trained, careful reading supported by listening.
Duration and Fundamental Frequency
Purpose: teach duration and fundamental frequency as measurable quantities read directly from a display, and connect them to the stress, length, and tone systems introduced in Chapters 18, 20, and 21.
Duration is simply the length of a labeled interval along the time axis, usually reported in milliseconds. Because a waveform and a spectrogram share the same horizontal time axis, duration can be measured on either display by marking the start and end of the interval in question and reading the elapsed time between them. Chapter 18's distinction between phonetic duration and phonological length, and Chapter 20's distinction between stress and quantity, both depend on exactly this kind of measurement, applied carefully and compared against other tokens in the same recording rather than against an assumed universal value.
Fundamental frequency, abbreviated F0, is the rate at which the vocal folds vibrate, measured in cycles per second, or hertz. F0 is the primary acoustic correlate of perceived pitch: a higher F0 corresponds to a higher perceived pitch, and a lower F0 to a lower one. F0 can be estimated from a waveform by measuring how much time one full vibratory cycle takes and calculating its rate, or read more directly from a specialized F0 track that some spectrogram displays include as an overlaid line.
| Interval | Approximate duration | Approximate F0 | Interpretation |
|---|---|---|---|
| Stop closure | 80 ms | not applicable | Silence; F0 cannot be measured where there is no vocal-fold vibration. |
| Vowel, first half | 60 ms | 110 Hz, roughly level | Consistent with a level tone or non-tonal stressed vowel onset. |
| Vowel, second half | 70 ms | falling from 110 Hz to 85 Hz | Consistent with a falling contour, to be checked against the language's tone or intonation system before final analysis. |
Notice that the interpretation column in the table above does not assert a phonological conclusion. A falling F0 contour is compatible with a falling lexical tone, a falling pitch accent, or ordinary phrase-final intonation, and only evidence from the language's prosodic system, of the kind gathered in Chapters 21 and 22, can decide among these possibilities. The acoustic measurement narrows the question; it does not answer it alone.
Formants and Vowel Resonance
Purpose: introduce formants as the acoustic correlate of vowel quality at an introductory level, connecting directly to the height and backness dimensions taught in Chapters 14 through 16.
A formant is a frequency region where the vocal tract's shape causes acoustic energy to resonate more strongly than at neighboring frequencies. Formants appear on a spectrogram as the dark horizontal bands introduced in Section 28.3. They are numbered from lowest to highest frequency: the first formant, F1, sits lowest; the second, F2, sits above it; higher formants exist but matter less for the introductory vowel analysis this chapter supports.
F1 and F2 relate directly to the articulatory vowel dimensions from Chapter 14. F1 correlates inversely with vowel height: a higher, more closed vowel typically shows a lower F1, while a lower, more open vowel typically shows a higher F1. F2 correlates with backness and, to some extent, rounding: a fronter, unrounded vowel typically shows a higher F2, while a backer or more rounded vowel typically shows a lower F2. These are reliable tendencies for introductory analysis, not exact formulas, and they must always be checked against a specific speaker's own vowel space rather than a fixed universal chart, since vocal-tract size and shape affect exact formant values considerably.
Two speakers producing what a listener judges to be the same vowel quality can show different absolute formant values, because vocal-tract length and shape differ from speaker to speaker. Introductory formant interpretation therefore compares a vowel's F1 and F2 with other vowels produced by the same speaker in the same recording, tracking relative height and backness, rather than treating any single fixed hertz value as a universal target for a given vowel quality.
Voice Onset Time
Purpose: introduce voice onset time as a specific, measurable interval that supports the aspiration and stop-voicing analysis begun in Chapter 9, without requiring advanced instrumentation.
Voice onset time, abbreviated VOT, is the interval between the release of a stop closure and the beginning of regular voicing in the following segment. VOT is measured directly on a waveform or spectrogram: locate the release burst, then locate the first clearly periodic vibration that follows it, and measure the time between the two.
A short or negative VOT, in which voicing begins at or even slightly before the release, is consistent with a fully voiced stop. A longer positive VOT, in which a noticeable voiceless interval separates the release from the onset of voicing, is consistent with a voiceless, and especially an aspirated, stop, connecting directly to the aspiration diacritic ʰ introduced in Chapter 9. VOT is a genuinely useful, learnable measurement at the introductory level precisely because it targets one clearly locatable interval rather than requiring a full acoustic analysis of the entire utterance.
What a Display Can and Cannot Establish
Purpose: state explicitly, as a stand-alone principle, the boundary this chapter and Chapter 27 have both assumed throughout: acoustic evidence supports phonetic analysis and does not replace it.
A waveform or spectrogram can reliably support claims about timing, relative amplitude, periodicity versus aperiodicity, approximate pitch movement, and approximate vowel quality once formants are correctly identified. These are genuine, checkable observations, not guesses, and a trained reader can defend them from the display alone.
A display cannot, by itself, establish a phonological analysis, a speaker's intended target, the correct IPA symbol, or a claim that goes beyond what the specific recording's context supports. Two different articulations can sometimes produce very similar acoustic patterns; noise, overlapping speech, or a low-quality recording can distort a pattern in ways an inexperienced reader may misread; and a measurement taken from one token says nothing certain about an entire language's sound system without further, appropriately sourced evidence.
| Supported | Plausible, needs more evidence | Not established by the display alone |
|---|---|---|
| Duration of a measured interval | Whether a duration difference is phonemic length or ordinary variation | The speaker's phonological intention |
| Presence of periodic vs. aperiodic energy | Manner classification consistent with that energy type | The correct IPA symbol without any auditory check |
| Approximate F0 and its direction of movement | A tone or intonation category, pending the language's system | A universal pitch meaning true of every language |
| Relative F1 and F2 position for one speaker | Approximate vowel height and backness for that speaker | An exact, speaker-independent vowel quality |
Treat every acoustic claim in this chapter, and in your own work, the way Chapter 27 treated a transcription draft: as a hypothesis, tested against the recording and against listening, and reported with a level of confidence that matches the strength of the evidence actually available.
Acoustic Evidence Across Sound Systems
The acoustic concepts introduced in this chapter apply regardless of language, but which measurement matters most shifts with the sound system under analysis. VOT is especially useful for languages with multiple contrastive stop series distinguished mainly by voicing or aspiration timing. F0 tracking is especially useful for languages with lexical tone or pitch accent, previewed in Chapters 21 and 22. Formant tracking matters most for languages with dense, closely spaced vowel inventories, where small quality differences carry the phonemic contrast. The measurements themselves do not change from language to language; which one earns the most attention in a given analysis does.
Analysis Exercises
These exercises work from fully labeled, described displays rather than unavailable audio, so that display-reading skills can be practiced honestly. Treat each item as a worked case: the relevant visual evidence is described in prose, and your task is to interpret it using this chapter's concepts.
Exercise A — Reading a Waveform
A waveform shows, in order: 90 milliseconds of a flat, near-silent line; a brief sharp spike; then 140 milliseconds of tall, evenly repeating waves. Identify the most likely articulatory event for each of the three intervals and state the evidence for each identification.
Exercise B — Formants and Vowel Quality
Two vowels from the same speaker are measured. Vowel one shows a low F1 and a high F2. Vowel two shows a higher F1 and a lower F2, closer to the middle of the speaker's range. Using the correlates from Section 28.5, state which vowel is likely higher and fronter, and explain why the comparison must stay within this one speaker's data.
Exercise C — Stating the Limits
A learner claims that a spectrogram "proves" a speaker produced a particular IPA symbol, based only on the shape of the formant bands. Using Section 28.6, explain what additional evidence would be needed before that claim could be responsibly accepted.
Common Mistakes
Error: Confusing waveform amplitude, the height of the wave, with pitch.
Correction: Read loudness from amplitude and pitch from how often a periodic pattern repeats per second, or, more directly, from F0.
Error: Comparing formant values across two different speakers as if a fixed hertz value always corresponds to the same vowel quality.
Correction: Compare formant values within one speaker's own vowel space, since vocal-tract size and shape shift the absolute values considerably.
Error: Reporting an acoustic measurement as if it settled a phonological question, such as whether a duration difference is contrastive.
Correction: Report the measurement itself with confidence, then explicitly separate it from the phonological interpretation, which requires additional distributional or contrastive evidence.
Chapter Review
Part A — Displays
1. Explain what the horizontal and vertical axes represent on a waveform, and how they differ from the axes on a spectrogram.
2. Describe the visible difference between periodic and aperiodic energy on both a waveform and a spectrogram.
Part B — Measurements
1. Define voice onset time and explain how it is measured.
2. Explain how F1 and F2 relate to vowel height and backness, and why the relationship must be checked against a specific speaker.
Part C — Limits
1. List two claims an acoustic display can reliably support and two claims it cannot establish alone.
2. Explain why a falling F0 contour alone cannot determine whether a language has lexical tone, pitch accent, or ordinary intonation.
Cumulative Review
1. From Chapter 9: Explain how voice onset time connects to the aspiration diacritic and the stop-voicing distinctions introduced there.
2. From Chapter 27: Explain how this chapter's concepts make step seven of the nine-step transcription workflow, comparing a draft against the recording, more precise than relying on an unexplained display.
3. Preview for Chapter 29: This chapter treated every acoustic claim as evidence requiring a stated source, speaker, and level of confidence. Predict how that same evidentiary discipline will apply to evaluating sources during the research planning taught in Chapter 29.
Self-Check
Before moving to Chapter 29, confirm that you can do the following without looking at your notes:
- Orient a waveform and a spectrogram in time, and, for the spectrogram, in frequency.
- Identify silence or closure, a release burst, and frication on a labeled display.
- Measure a duration and describe how F0 is estimated.
- Explain how F1 and F2 relate to vowel height and backness for one speaker.
- Define voice onset time and explain what a longer positive VOT suggests.
- State at least two things an acoustic display can support and two it cannot establish alone.
If any of these tasks are difficult, review the relevant section before reading Chapter 29. Chapter 29 shifts from acoustic evidence for individual utterances to the broader research and source-evaluation skills needed to plan a complete language sound-system profile.
Answer Key (click to expand)
Analysis Exercises — Exercise A
The 90-millisecond flat, near-silent interval is consistent with a stop closure, since a full oral closure blocks the airflow that would otherwise produce visible amplitude. The brief sharp spike is consistent with the release burst that follows a stop closure. The 140 milliseconds of tall, evenly repeating waves are consistent with a voiced vowel, since regular vocal-fold vibration produces exactly this kind of periodic pattern.
Analysis Exercises — Exercise B
Vowel one, with a low F1 and a high F2, is likely higher and fronter than vowel two, following the inverse relationship between F1 and height and the direct relationship between F2 and frontness described in Section 28.5. The comparison must stay within one speaker's data because vocal-tract size and shape shift absolute formant values, so a value that indicates a high front vowel for one speaker could indicate a different vowel quality for a speaker with a differently sized vocal tract.
Analysis Exercises — Exercise C
Additional evidence should include auditory confirmation from repeated listening, comparison with other vowels from the same speaker to establish a relative vowel space, and, where relevant, distributional or contrastive evidence establishing that the observed quality functions as a distinct category in the language rather than a contextual variant. A spectrogram alone shows a resonance pattern; a specific IPA symbol is an analytical claim that requires this additional support before it can be treated as established.
Chapter Review — Part A
1. A waveform's horizontal axis is time and its vertical axis is amplitude; a spectrogram keeps time on the horizontal axis but replaces amplitude with frequency on the vertical axis, showing intensity as darkness or color instead. 2. Periodic energy appears as an evenly repeating wave shape on a waveform and as vertical striations or formant bands on a spectrogram; aperiodic energy appears as an irregular, non-repeating, densely jagged wave shape on a waveform and as a diffuse, unstructured noise band, often at higher frequencies, on a spectrogram.
Chapter Review — Part B
1. Voice onset time is the interval between the release of a stop closure and the onset of regular voicing in the following segment; it is measured by locating the release burst and the first clearly periodic vibration after it, then measuring the elapsed time between the two. 2. F1 correlates inversely with vowel height, so a lower F1 suggests a higher vowel, and F2 correlates with frontness, so a higher F2 suggests a fronter vowel; the relationship must be checked against a specific speaker because vocal-tract size and shape shift the absolute formant values that correspond to a given vowel quality.
Chapter Review — Part C
1. Reliably supported: the duration of a measured interval, and whether a stretch of speech shows periodic or aperiodic energy. Not established alone: the speaker's intended phonological category, and a claim about an entire language's sound system based on one token. 2. A falling F0 contour is an acoustic pattern compatible with several different phonological analyses, including a falling lexical tone, a falling pitch accent, or ordinary phrase-final intonation; deciding among these requires evidence about the language's prosodic system, such as whether pitch contrasts word meanings, gathered through the methods taught in Chapters 21 and 22, not the F0 measurement alone.
Cumulative Review
1. Chapter 9 distinguished stop voicing from aspiration and introduced the aspiration diacritic for a voiceless stop followed by a noticeable delay before voicing begins; voice onset time gives that distinction a precise, measurable acoustic basis, since a longer positive VOT is the acoustic signature of the delay that aspiration notation represents. 2. Chapter 27's step seven asked a transcriber to compare a draft transcription against the recording, using slowed playback and visual evidence as support; this chapter's concepts let that comparison target specific, locatable acoustic events, such as a closure boundary, a burst, or a formant transition, rather than relying on an impression of the display as a whole. 3. Chapter 29 will likely require the same evidentiary discipline applied to sources rather than displays: stating what a given source can establish, what it can only suggest, and what it cannot establish at all, and reporting confidence in a claim only to the degree the available evidence actually supports it.
Looking Ahead
What You Can Now Do
You can now orient a waveform and a spectrogram in time and frequency; identify periodicity, silence or closure, release, and frication in controlled examples; interpret duration and fundamental frequency; interpret first and second formants at an introductory level; recognize basic voice onset timing evidence; connect acoustic patterns to articulatory hypotheses; and state clearly what a given acoustic display can and cannot establish. Together with Chapters 24 through 27, you now have a complete introductory toolkit for deciding what a transcription needs, notating it correctly, reading someone else's transcription, producing your own from a recording, and supporting every step with acoustic evidence read responsibly.