---
title: Phonetics
module: Speech
moduleNumber: 8
lessonNumber: 1
order: 801
summary: >
  Before a recognizer can read speech it has to know what speech is. This first
  part covers the linguistic substrate: phones and their transcription in the IPA
  and ARPAbet; articulatory phonetics — how the vocal tract shapes airflow into
  consonants and vowels; and prosody — stress, tune, and the F0 contour. The
  acoustic side — the waveform, its spectrum, formants, and the spectrogram — is
  the second part.
topics: [Speech]
sources:
  - book: Jurafsky
    ref: "Ch. 25 — Phonetics; §25.1 Speech Sounds and Phonetic Transcription"
  - book: Jurafsky
    ref: "§25.2 Articulatory Phonetics; §25.3 Prosody"
---

A written character like `p` or `a` is already a scientific model of speech: a
claim that the continuous, gliding sound of a spoken word decomposes into a small
set of discrete, reusable units. That claim is old — the earliest writing systems
already used some symbols for sounds rather than whole words — and it is the same
claim that makes [speech recognition](/natural-language-processing/speech/automatic-speech-recognition)
and text-to-speech possible.[^jm-intro] **Phonetics** is the study of those units:
what they are, how the human vocal tract produces them, and what their acoustic
signal looks like once it reaches a microphone.

This lesson is the linguistic and physical groundwork for the recognizer that
follows. That recognizer never sees a phone directly; it sees a waveform, turns it
into a **spectrogram**, and learns to read the acoustic traces that phones leave
there. Everything here — the vocal tract as a filter, formants as resonances, F0
as the vibration of the vocal folds — explains where those traces come
from. The [log-mel front end](/natural-language-processing/speech/automatic-speech-recognition)
of a modern ASR system is a mechanized, learned version of exactly the acoustic
analysis this chapter does by hand.

$$
% caption: The speech chain. A speaker's articulators shape airflow into a
% pressure wave; a microphone records it as a waveform; acoustic analysis turns
% the waveform into a spectrogram, which is what a recognizer reads. This lesson
% walks the chain left to right.
\begin{tikzpicture}[>=stealth, font=\footnotesize,
  box/.style={draw, minimum width=23mm, minimum height=12mm, align=center, font=\scriptsize}]
  \definecolor{acc}{HTML}{2348F2}
  \node[box] (art)  at (0,0)    {articulation\\(vocal tract)};
  \node[box] (wav)  at (3.4,0)  {waveform\\(air pressure)};
  \node[box] (spec) at (6.8,0)  {spectrogram\\(time-frequency)};
  \node[box, draw=acc, text=acc, thick] (asr) at (10.2,0) {recognizer\\(words)};
  \draw[->, acc, thick] (art)  -- (wav)  node[midway, above, font=\scriptsize] {sound};
  \draw[->, acc, thick] (wav)  -- (spec) node[midway, above, font=\scriptsize] {Fourier};
  \draw[->, acc, thick] (spec) -- (asr)  node[midway, above, font=\scriptsize] {learn};
  \node[font=\scriptsize, text=black, anchor=north] at (0,-0.95)   {production};
  \node[font=\scriptsize, text=black, anchor=north] at (5.1,-0.95) {acoustics};
  \node[font=\scriptsize, text=black, anchor=north] at (10.2,-0.95){this course};
\end{tikzpicture}
$$

## Speech sounds and phonetic transcription

We represent the pronunciation of a word as a string of **phones** — speech
sounds, each written with a symbol adapted from the Roman alphabet.[^jm-intro] A
phone is a physical event, an actual sound produced by a vocal tract, and a symbol
in square brackets, like `[p]`, denotes a specific articulation.

The standard cross-linguistic system is the **International Phonetic Alphabet**
(IPA), first developed in 1888 and still evolving, which has a symbol for every
sound in every documented language. For English work it is common to use the
**ARPAbet** instead: a plain-ASCII phonetic alphabet covering the American-English
subset of the IPA, which is convenient precisely because it needs no special
glyphs. The two are in one-to-one correspondence over their shared inventory; the
ARPAbet `[sh]` is the IPA `ʃ`, `[iy]` is `i`, `[aa]` is `ɑ`.

> **Definition (Phone).** A speech sound, the atomic unit of pronunciation,
> written in brackets with an IPA or ARPAbet symbol. A word's pronunciation is a
> string of phones: _parsley_ is `[p aa r s l iy]`.

The reason we transcribe at all — rather than just spelling — is that English
orthography is a poor guide to sound. The mapping from letters to phones is opaque:
one letter spells many sounds, and one sound is spelled many ways.

$$
% caption: English spelling hides the phones. The single letter c is the phone
% [k] in cougar but [s] in cell; conversely the phone [k] surfaces as c, k, ck,
% cc, and the x of fox. Transcription records the sound, not the spelling.
\begin{tikzpicture}[>=stealth, font=\footnotesize,
  lit/.style={draw, minimum width=17mm, minimum height=8mm, align=center, font=\scriptsize},
  ph/.style={draw, draw=acc, text=acc, minimum width=10mm, minimum height=8mm, align=center, font=\scriptsize}]
  \definecolor{acc}{HTML}{2348F2}
  % one letter -> two phones
  \node[lit] (c) at (0,1.6) {letter c};
  \node[ph]  (k1) at (3.6,2.4) {[k]};
  \node[ph]  (s1) at (3.6,0.8) {[s]};
  \draw[->, acc] (c) -- (k1) node[midway, above, font=\scriptsize, text=black] {cougar};
  \draw[->, acc] (c) -- (s1) node[midway, below, font=\scriptsize, text=black] {cell};
  % many letters -> one phone
  \node[ph]  (k2) at (7.6,1.6) {[k]};
  \foreach \sp/\y/\w in {c/2.9/cook, k/2.2/kite, ck/1.5/jackal, cc/0.8/raccoon, x/0.1/fox} {
    \node[lit, minimum width=13mm] (\sp) at (11.3,\y) {\sp};
    \draw[<-, acc] (k2) -- (\sp) node[midway, above, font=\scriptsize, text=black] {};
  }
  \node[font=\scriptsize, text=black, anchor=south] at (1.8,3.15) {one letter, many sounds};
  \node[font=\scriptsize, text=black, anchor=south] at (9.4,3.15) {one sound, many spellings};
\end{tikzpicture}
$$

Many languages — Spanish, for one — are far more transparent, but for English,
talking about sound requires a notation for sound. The table
below is a working subset of the ARPAbet, with the IPA equivalent and an example
word.

| ARPAbet | IPA | Example | | ARPAbet | IPA | Example |
| --- | --- | --- | --- | --- | --- | --- |
| `[p]` | p | **p**arsley | | `[iy]` | i | l**i**ly |
| `[t]` | t | **t**ea | | `[ih]` | ɪ | l**i**ly |
| `[k]` | k | **c**ook | | `[eh]` | ɛ | p**e**n |
| `[b]` | b | **b**ay | | `[ae]` | æ | **a**ster |
| `[d]` | d | **d**ill | | `[aa]` | ɑ | p**o**ppy |
| `[g]` | g | **g**arlic | | `[ao]` | ɔ | **o**rchid |
| `[m]` | m | **m**int | | `[uh]` | ʊ | w**oo**d |
| `[n]` | n | **n**utmeg | | `[uw]` | u | t**u**lip |
| `[ng]` | ŋ | baki**ng** | | `[ah]` | ʌ | b**u**tter |
| `[f]` | f | **f**lour | | `[er]` | ɝ | b**ir**d |
| `[s]` | s | **s**oup | | `[ax]` | ə | lot**u**s |
| `[sh]` | ʃ | squa**sh** | | `[ay]` | aɪ | **i**ris |
| `[ch]` | tʃ | **ch**erry | | `[aw]` | aʊ | fl**ow**er |
| `[l]` | l | **l**icorice | | `[oy]` | ɔɪ | s**oi**l |
| `[r]` | r | **r**ice | | `[ow]` | oʊ | l**o**tus |

The left half of the table is consonants, the right half vowels; the rest of the
lesson explains what that division means physically. A companion notion is the
**phoneme**, the abstract sound category a language treats as one unit, versus the
**allophone**, the concrete variant that actually surfaces in context. English
treats the aspirated `[p]` of _pin_ and the unaspirated `[p]` of _spin_ as the
same phoneme even though they are physically distinct phones; the two are
allophones of one phoneme. Phonetics studies the phones; phonology studies which
distinctions a language treats as meaningful.

## Articulatory phonetics

**Articulatory phonetics** studies how phones are produced as the organs of the
mouth, throat, and nose modify airflow from the lungs.[^jm-artic] Sound starts as
air expelled from the lungs up the **trachea** (windpipe). At the top of the
trachea sits the **larynx** — the Adam's apple — which holds two folds of muscle,
the **vocal folds**. The gap between them is the **glottis**.

$$
% caption: The vocal organs in side view. Air from the lungs passes the vocal
% folds at the glottis, then through the oral and nasal tracts, which act as the
% resonating cavities that give each phone its acoustic shape.
\begin{tikzpicture}[>=stealth, font=\footnotesize]
  \definecolor{acc}{HTML}{2348F2}
  % head/tract outline (schematic)
  \draw[black, thick] (0,0) .. controls (0,3.2) and (2.2,3.6) .. (4.4,3.6)
    .. controls (5.6,3.6) and (6.0,3.0) .. (6.0,2.4);
  \draw[black, thick] (1.2,0) .. controls (1.2,1.6) and (1.8,2.0) .. (2.4,2.1)
    .. controls (3.4,2.3) and (4.2,2.2) .. (5.0,2.2)
    .. controls (5.6,2.2) and (5.7,2.4) .. (5.7,2.4);
  % lips at right
  \draw[black, thick] (6.0,2.4) -- (6.0,2.9);
  % nasal tract (upper cavity)
  \draw[black] (4.4,3.6) .. controls (5.4,3.4) and (5.8,3.2) .. (6.0,2.9);
  % tongue
  \fill[acc!12] (2.7,2.15) .. controls (3.4,2.9) and (4.2,2.9) .. (4.9,2.25)
    .. controls (4.2,2.25) and (3.4,2.25) .. (2.7,2.15) -- cycle;
  \draw[acc] (2.7,2.15) .. controls (3.4,2.9) and (4.2,2.9) .. (4.9,2.25);
  % trachea to lungs
  \draw[black, thick] (0.6,0) -- (0.6,1.2);
  \draw[black, thick] (1.2,0) -- (1.2,1.2);
  % glottis / vocal folds marker
  \fill[acc] (0.9,1.35) circle (1.6pt);
  % labels (kept off the strokes)
  \node[anchor=east, font=\scriptsize, text=acc] at (0.4,1.35) {glottis /};
  \node[anchor=east, font=\scriptsize, text=acc] at (0.4,1.05) {vocal folds};
  \node[anchor=north, font=\scriptsize] at (0.9,-0.1) {from lungs (trachea)};
  \node[font=\scriptsize] at (3.8,2.62) {tongue};
  \node[anchor=west, font=\scriptsize] at (6.15,2.65) {lips};
  \node[anchor=west, font=\scriptsize] at (6.15,3.25) {nasal tract};
  \node[font=\scriptsize] at (3.6,1.75) {oral tract};
\end{tikzpicture}
$$

If the folds are held close together, airflow makes them vibrate, and the sound is
**voiced**; if they are apart, no vibration occurs and the sound is **unvoiced**
(voiceless). All English vowels are voiced, as are `[b] [d] [g] [v] [z]`; `[p] [t]
[k] [f] [s]` are their voiceless counterparts. Above the larynx is the **vocal
tract**, split into the **oral tract** (out the mouth) and the **nasal tract**
(out the nose). Sounds routed partly through the nose — `[m] [n] [ng]` — are
**nasal**.

Phones split into two classes. **Consonants** restrict or block the airflow;
**vowels** leave it comparatively open, are usually voiced, and are louder and
longer. Semivowels like `[y]` and `[w]` sit between: voiced like vowels but short
and non-syllabic like consonants.

### Consonants: place of articulation

Because a consonant is made by pinching the airflow somewhere, we can sort
consonants by _where_ that pinch happens — the **place of articulation**.[^jm-artic]

$$
% caption: The major English places of articulation, ordered front (lips) to back
% (glottis) along the vocal tract. The place is the point of maximum constriction.
\begin{tikzpicture}[>=stealth, font=\footnotesize,
  pl/.style={draw, minimum width=18mm, minimum height=8mm, align=center, font=\scriptsize}]
  \definecolor{acc}{HTML}{2348F2}
  % a schematic mouth axis, front (left) to back (right)
  \draw[->, black, thick] (-0.2,0) -- (11.6,0) node[anchor=west, font=\scriptsize, text=black] {back};
  \node[anchor=east, font=\scriptsize] at (-0.25,0) {front};
  \foreach \x/\name/\ex in {0.4/bilabial/{p b m}, 2.4/labiodental/{f v}, 4.4/dental/{th dh}, 6.4/alveolar/{s z t d}, 8.4/palatal/{sh ch y}, 10.4/velar/{k g ng}} {
    \node[pl] (\name) at (\x,0.9) {\name};
    \draw[acc] (\name.south) -- (\x,0.05);
    \node[font=\scriptsize, text=acc, anchor=north] at (\x,-0.1) {\ex};
  }
\end{tikzpicture}
$$

Working front to back: **labial** sounds close the two lips (**bilabial** `[p] [b]
[m]`) or press the lower lip to the upper teeth (**labiodental** `[f] [v]`).
**Dental** sounds `[th] [dh]` place the tongue at or between the teeth. **Alveolar**
sounds `[s] [z] [t] [d]` press the tongue tip to the alveolar ridge, the bump just
behind the upper teeth. **Palatal** (strictly, palato-alveolar) sounds `[sh] [ch]
[zh] [jh]` and the palatal `[y]` raise the tongue toward the hard palate. **Velar**
sounds `[k] [g] [ng]` press the back of the tongue against the velum, the soft
palate at the very back. A **glottal** stop closes the glottis itself.

### Consonants: manner of articulation

The place says where; the **manner of articulation** says how the constriction is
made — a full stop, a narrow hiss, or a gentle approach.[^jm-artic] Place and
manner together, plus voicing, almost always identify a consonant uniquely.

$$
% caption: Manner of articulation ordered by how completely the airflow is
% obstructed, from a total stop through turbulent fricatives to open approximants.
% Each manner is illustrated by the airstream through the constriction.
\begin{tikzpicture}[>=stealth, font=\footnotesize,
  mbox/.style={draw, minimum width=22mm, minimum height=20mm, align=center, font=\scriptsize}]
  \definecolor{acc}{HTML}{2348F2}
  % stop: blocked then burst
  \node[mbox] (stopb) at (0,0) {};
  \draw[acc, thick] (-0.9,0.4) -- (0.0,0.4);
  \draw[black, very thick] (0.1,0.7) -- (0.1,0.1);
  \draw[acc, thick, dashed] (0.25,0.4) -- (0.9,0.4);
  \node[font=\scriptsize, anchor=south] at (0,1.05) {stop};
  \node[font=\scriptsize, text=acc, anchor=north] at (0,-1.05) {p b t d k g};
  \node[font=\scriptsize, text=black, anchor=north] at (0,-1.5) {full closure + burst};
  % fricative: narrow, turbulent
  \node[mbox] (fric) at (3.4,0) {};
  \draw[acc, thick] (2.5,0.55) -- (3.1,0.55);
  \draw[acc, thick] (2.5,0.25) -- (3.1,0.25);
  \draw[acc, thick] (3.15,0.4) -- (3.35,0.55) -- (3.55,0.25) -- (3.75,0.55) -- (3.95,0.25) -- (4.15,0.55) -- (4.3,0.4);
  \node[font=\scriptsize, anchor=south] at (3.4,1.05) {fricative};
  \node[font=\scriptsize, text=acc, anchor=north] at (3.4,-1.05) {f v s z sh};
  \node[font=\scriptsize, text=black, anchor=north] at (3.4,-1.5) {narrow, hissing};
  % affricate: stop + fricative
  \node[mbox] (aff) at (6.8,0) {};
  \draw[black, very thick] (6.5,0.7) -- (6.5,0.1);
  \draw[acc, thick] (6.6,0.4) -- (6.8,0.55) -- (7.0,0.25) -- (7.2,0.55) -- (7.4,0.25) -- (7.6,0.55) -- (7.7,0.4);
  \node[font=\scriptsize, anchor=south] at (6.8,1.05) {af\/fricate};
  \node[font=\scriptsize, text=acc, anchor=north] at (6.8,-1.05) {ch jh};
  \node[font=\scriptsize, text=black, anchor=north] at (6.8,-1.5) {stop then fricative};
  % approximant: open
  \node[mbox] (app) at (10.2,0) {};
  \draw[acc, thick] (9.3,0.6) .. controls (9.9,0.6) and (10.1,0.4) .. (11.1,0.4);
  \node[font=\scriptsize, anchor=south] at (10.2,1.05) {approximant};
  \node[font=\scriptsize, text=acc, anchor=north] at (10.2,-1.05) {y w r l};
  \node[font=\scriptsize, text=black, anchor=north] at (10.2,-1.5) {open, no turbulence};
\end{tikzpicture}
$$

A **stop** (plosive) blocks the air completely, then releases it in a burst; the
silent block is the **closure** and the burst is the **release**. English has
voiced stops `[b] [d] [g]` and voiceless `[p] [t] [k]`. **Nasals** `[n] [m] [ng]`
lower the velum so air escapes through the nose. In a **fricative** the channel is
narrowed but not closed, and the turbulent air hisses: `[f] [v] [th] [dh] [s] [z]
[sh] [zh]`, of which the loud, high ones `[s] [z] [sh] [zh]` are **sibilants**. A
stop released straight into a fricative is an **affricate**, `[ch]` and `[jh]`. In
an **approximant** the articulators come close but not close enough for turbulence:
`[y] [w] [r] [l]`, with `[l]` a **lateral** because air flows over the lowered
sides of the tongue. A **tap** or **flap** `[dx]` is a single quick brush of the
tongue against the alveolar ridge — the middle consonant of _lotus_ `[l ow dx ax s]`
in most American dialects.

### Vowels and the vowel space

Vowels have no real constriction, so they are described not by place and manner but
by the _position_ of the tongue body and the shape of the lips.[^jm-artic] Three
parameters do the work: **height** (how high the tongue's high point sits),
**frontness/backness** (whether that high point is toward the front or back of the
mouth), and **rounding** (whether the lips are pursed). For `[iy]` (_beet_) the
tongue is high and front; for `[uw]` (_boot_) high and back; for `[ae]` (_bat_) low
and front.

$$
% caption: The English vowel space, a schematic quadrilateral with tongue height
% on the vertical axis and frontness on the horizontal. Point vowels hold a fixed
% tongue position; diphthongs (drawn as arrows) glide from one position to another.
\begin{tikzpicture}[>=stealth, font=\footnotesize]
  \definecolor{acc}{HTML}{2348F2}
  \definecolor{red}{HTML}{C0392B}
  % quadrilateral frame
  \draw[black] (0,0) -- (6.4,0) -- (5.6,4.0) -- (0.8,4.0) -- cycle;
  % axis labels
  \node[anchor=east, font=\scriptsize] at (-0.15,3.9) {high};
  \node[anchor=east, font=\scriptsize] at (-0.15,0.1) {low};
  \node[anchor=south, font=\scriptsize] at (0.8,4.05) {front};
  \node[anchor=south, font=\scriptsize] at (5.6,4.05) {back};
  % point vowels (x, y in space)
  \foreach \name/\x/\y in {ih/1.5/2.9, eh/1.7/1.9, ae/1.9/0.4, aa/4.3/0.3, ao/5.2/1.3, uh/4.6/2.9, ah/3.6/1.5, ax/3.3/2.3} {
    \fill[acc] (\x,\y) circle (2.0pt);
    \node[font=\scriptsize, text=acc, anchor=south] at (\x,\y+0.06) {\name};
  }
  % high corner vowels, labels placed to the side to clear the top frame
  \fill[acc] (1.2,3.5) circle (2.0pt); \node[font=\scriptsize, text=acc, anchor=west] at (1.3,3.5) {iy};
  \fill[acc] (5.0,3.5) circle (2.0pt); \node[font=\scriptsize, text=acc, anchor=east] at (4.9,3.5) {uw};
  % diphthongs as glides
  \draw[->, red, thick] (2.6,0.6) -- (2.2,2.7); \node[red, font=\scriptsize, anchor=east] at (2.55,0.55) {ay};
  \draw[->, red, thick] (3.7,0.7) -- (4.5,2.8); \node[red, font=\scriptsize, anchor=east] at (3.6,0.6) {aw};
  \draw[->, red, thick] (4.8,2.6) -- (5.5,1.6);
  \node[red, font=\scriptsize, anchor=east] at (4.55,2.65) {oy};
\end{tikzpicture}
$$

The space is drawn as a quadrilateral because that roughly matches both tongue
positions and, more accurately, the acoustics we get to below. A vowel with a
fixed tongue position is a plain vowel; one in which the tongue moves markedly
_during_ the vowel is a **diphthong**, drawn here as an arrow from start to end —
English is rich in them (`[ay] [aw] [oy]`). Certain back vowels (`[uw] [ao] [ow]`)
are pronounced with the lips **rounded**.

### Syllables

Consonants and vowels group into **syllables**. The vowel at the core is the
**nucleus**; the consonants before it are the **onset**; the consonants after it
are the **coda**; and the nucleus plus coda together form the **rime**. _Dog_
`[d aa g]` has onset `[d]`, nucleus `[aa]`, coda `[g]`.

$$
% caption: Syllable structure for ham, drawn as a tree. The onset precedes the
% nucleus; the nucleus plus coda form the rime; sigma marks the whole syllable.
\begin{tikzpicture}[>=stealth, font=\footnotesize,
  n/.style={font=\scriptsize}]
  \definecolor{acc}{HTML}{2348F2}
  \node[n, text=acc] (sig) at (2.6,3.2) {(syllable)};
  \node[n] (on)  at (0.8,2.2) {Onset};
  \node[n] (ri)  at (3.6,2.2) {Rime};
  \node[n] (nu)  at (2.8,1.2) {Nucleus};
  \node[n] (co)  at (4.6,1.2) {Coda};
  \node[n, text=acc] (h)  at (0.8,0.2) {h};
  \node[n, text=acc] (ae) at (2.8,0.2) {ae};
  \node[n, text=acc] (m)  at (4.6,0.2) {m};
  \draw[black] (sig) -- (on);
  \draw[black] (sig) -- (ri);
  \draw[black] (ri) -- (nu);
  \draw[black] (ri) -- (co);
  \draw[black] (on) -- (h);
  \draw[black] (nu) -- (ae);
  \draw[black] (co) -- (m);
\end{tikzpicture}
$$

Which phone sequences a language permits in an onset or coda is its
**phonotactics** — English allows the onset `[str]` of _strike_ but forbids
`[zdr]`. These constraints can be captured by a finite-state or language model over
phone strings, and breaking a word into syllables is **syllabification**.

#### A worked transcription

Put the pieces together on one word. Take _greasy_, from the TIMIT sentence _she
had your dark suit in greasy wash water_. Its ARPAbet transcription is `[g r iy s
iy]` — the IPA `[g r i s i]`. Read left to right against the machinery above: `[g]`
is a voiced velar stop (back of tongue on the velum, folds vibrating); `[r]` is a
voiced alveolar approximant; `[iy]` is a high front tense vowel; `[s]` is a
voiceless alveolar sibilant fricative; `[iy]` again. Every phone is now a triple of
articulatory facts, not a letter. Syllabify it and the two vowels each anchor a
syllable: the first is `[g r]` onset $+$ `[iy]` nucleus, the second `[s]` onset $+$
`[iy]` nucleus, with no coda on either. The onset `[g r]` is legal English
phonotactics; `[r g]` would not be.

$$
% caption: The word greasy, [g r iy s iy], split into two syllables. Each vowel is
% a nucleus; the consonants before it are the onset. Sigma marks a syllable. The
% word has no coda, so each rime is just its nucleus.
\begin{tikzpicture}[>=stealth, font=\footnotesize,
  n/.style={font=\scriptsize}]
  \definecolor{acc}{HTML}{2348F2}
  % syllable 1
  \node[n, text=acc] (s1) at (1.5,2.6) {(syllable 1)};
  \node[n] (on1) at (0.6,1.6) {Onset};
  \node[n] (nu1) at (2.4,1.6) {Nucleus};
  \node[n, text=acc] (g) at (0.2,0.4) {g};
  \node[n, text=acc] (r) at (1.0,0.4) {r};
  \node[n, text=acc] (iy1) at (2.4,0.4) {iy};
  \draw[black] (s1) -- (on1);
  \draw[black] (s1) -- (nu1);
  \draw[black] (on1) -- (g);
  \draw[black] (on1) -- (r);
  \draw[black] (nu1) -- (iy1);
  % syllable 2
  \node[n, text=acc] (s2) at (5.5,2.6) {(syllable 2)};
  \node[n] (on2) at (4.6,1.6) {Onset};
  \node[n] (nu2) at (6.4,1.6) {Nucleus};
  \node[n, text=acc] (s) at (4.6,0.4) {s};
  \node[n, text=acc] (iy2) at (6.4,0.4) {iy};
  \draw[black] (s2) -- (on2);
  \draw[black] (s2) -- (nu2);
  \draw[black] (on2) -- (s);
  \draw[black] (nu2) -- (iy2);
\end{tikzpicture}
$$

## Prosody

Everything so far concerns the _identity_ of sounds. **Prosody** concerns their
_melody and rhythm_: the use of F0, energy, and duration to carry meaning that the
phone string alone does not.[^jm-prosody] Prosody marks the difference between a
statement and a question, signals which word matters, conveys affect (happiness,
surprise, anger), and manages turn-taking in conversation.

### Prominence, stress, and schwa

Some words, and some syllables within them, sound more **prominent** — perceptually
salient — than others. A speaker makes a syllable prominent by making it louder,
longer, or by varying its F0. We mark prominence with a **pitch accent**; a word
bearing an accent is accented. But not every syllable can host the accent: it must
land on the syllable that carries **lexical stress**, a fixed property of the
word's dictionary pronunciation. _Surprised_ is stressed on its second syllable, so
an accent on _surprised_ strengthens that syllable, never the first.

Lexical stress is marked in dictionaries — the CMU dictionary tags each vowel `0`
(unstressed) or `1` (stressed): _table_ is `[T EY1 B AH0 L]`. Stress can even
distinguish words: the noun _content_ is `[K AA1 N T EH0 N T]`, the adjective
`[K AA0 N T EH1 N T]`. An unstressed vowel is often **reduced** to a weaker,
centralized quality, most commonly the **schwa** `[ax]` (the IPA `ə`) — the second
vowel of _parakeet_ `[p ae r ax k iy t]`. The result is a continuum of prominence:
accented, then stressed, then full-vowel, then reduced.

### Tune and intonation

Two utterances with identical words and stress can still differ in **tune** — the
rise and fall of F0 over the utterance.[^jm-prosody] The classic example is the
contrast between a statement and a yes-no question on the same words: a final F0
rise (a **question rise**) versus a final drop (a **final fall**).

$$
% caption: The same words carry two tunes. A final F0 rise (blue) signals a yes-no
% question; a final fall (red) signals a declarative statement. Only the pitch
% contour over the last words differs.
\begin{tikzpicture}[>=stealth, font=\footnotesize]
  \definecolor{acc}{HTML}{2348F2}
  \definecolor{red}{HTML}{C0392B}
  % axes
  \draw[->, black] (0,0) -- (0,3.2) node[anchor=south east, font=\scriptsize, text=black] {F0};
  \draw[->, black] (0,0) -- (9.2,0) node[anchor=north east, font=\scriptsize, text=black] {time};
  % word ticks
  \foreach \x/\w in {0.9/you, 2.9/know, 5.0/what I, 7.6/mean} {
    \node[font=\scriptsize, anchor=north] at (\x,-0.1) {\w};
    \draw[black] (\x,0) -- (\x,-0.08);
  }
  % question rise (blue): flat then up
  \draw[acc, very thick]
    (0.5,1.3) .. controls (2.0,1.2) and (4.0,1.4) .. (6.0,1.3)
    .. controls (7.2,1.3) and (8.0,2.2) .. (8.7,2.8);
  \node[acc, font=\scriptsize, anchor=west] at (8.75,2.8) {question rise};
  % final fall (red): flat then down
  \draw[red, very thick]
    (0.5,1.7) .. controls (2.0,1.7) and (4.0,1.9) .. (6.0,1.7)
    .. controls (7.2,1.7) and (8.0,0.9) .. (8.7,0.5);
  \node[red, font=\scriptsize, anchor=west] at (8.75,0.5) {f\/inal fall};
\end{tikzpicture}
$$

English uses tune widely: besides the question rise, a list of comma-separated
nouns often carries a small **continuation rise** after each item, and there are
characteristic contours for contradiction and surprise. Utterances also have
**prosodic structure** — words group into **intonation phrases** with audible
breaks between them, often at commas and often aligned with syntactic constituents.
One influential labeling scheme, **ToBI** (Tone and Break Indices), tags each
accented word with one of a handful of pitch-accent types (`H*`, `L*`, `L*+H`, …)
and each phrase with a boundary tone (`L-L%` for the declarative fall, `H-H%` for
the question rise). Predicting these boundaries automatically matters for
text-to-speech, and models are trained on corpora such as the Boston University
Radio News Corpus.

## From articulation to acoustics

Everything so far concerns how a phone is _made_ and what a listener perceives — the articulatory and prosodic side of speech. But a microphone never sees a tongue or a glottis; it sees only the pressure wave those movements produce. The second part follows the sound out of the mouth: the waveform, its decomposition into a spectrum, the source-filter model that explains why each vowel has its own formants, and the spectrogram that a recognizer actually reads. This continues in [Acoustic Phonetics](/natural-language-processing/speech/acoustic-phonetics).

[^jm-intro]: **Jurafsky & Martin**, _Speech and Language Processing_ (3rd ed.), Ch. 25 — Phonetics; §25.1 Speech Sounds and Phonetic Transcription: phones as the units of pronunciation, the IPA and the ASCII ARPAbet, and the opaque letter-to-sound mapping of English.
[^jm-artic]: **Jurafsky & Martin**, §25.2 — Articulatory Phonetics: the vocal organs, voicing at the glottis, place and manner of articulation for consonants, the height/backness/rounding parameters for vowels, and syllable structure.
[^jm-prosody]: **Jurafsky & Martin**, §25.3 — Prosody: prominence, pitch accent, lexical stress and vowel reduction to schwa, prosodic phrasing, tune (question rise vs. final fall), and the ToBI labeling system.