Intonation & Prosody

Intonation is sentence melody — the pitch contour rising and falling across a sentence. Prosody is the broader rhythmic and stress patterning of speech. Together they're the third major axis (after pitch and resonance) that the brain reads as gender, and they're often the variable that resists self-correction longest because we don't hear our own intonation patterns.

For pitch and resonance fundamentals see Pitch Fundamentals and Resonance Fundamentals. For practice material that trains these patterns see Practice Sentences.

Intonation is the variable that most often explains the "everything else is right but I still get clocked on the phone" complaint. Speakers who have moved F0 into a female-typical band and lifted resonance forward will still read as masculine if their contour is flat, their stress is even, or their sentences collapse into vocal fry at the end. Several converging perceptual studies report that listeners rate transfeminine voice as more feminine when intonation is animated, even when F0 is held constant (Wolfe et al. 1990, Hancock et al. 2014, Gelfer & Schofield 2000).

What intonation is

Sentence melody — the pitch contour

Every spoken sentence has a melodic line. Pitch rises on stressed syllables, falls into unstressed ones, and either rises, falls, or holds at the end of the phrase. This contour carries grammatical information (statement vs question), emotional information (interest, boredom, surprise), and social information (engagement, certainty, politeness). It's a continuous variable — pitch glides between targets rather than stepping between fixed levels.

Intonation differs from pitch the way melody differs from a single note. A speaker with a "feminine" F0 of 195 Hz can still sound masculine if the contour through that F0 is flat; a speaker with a lower F0 can read as more feminine if the contour is animated and the sentence-final pattern is right.

Why semitones, not Hz

The contour is described in semitones, not Hz, because the ear hears pitch logarithmically. A 50 Hz swing at a low F0 of 110 Hz is roughly seven semitones; the same 50 Hz swing at a high F0 of 220 Hz is only about four. Comparing intonation ranges in Hz across speakers with different baseline pitches gives nonsense answers — women's higher mean F0 inflates absolute Hz values for any given semitone span, which is why older Hz-based studies overstate the female-male intonation gap.

A semitone is one twelfth of an octave. Twelve semitones is a doubling of frequency. For a speaker at 180 Hz, one semitone up is about 191 Hz; for a speaker at 110 Hz, one semitone up is about 116 Hz. Quote semitones, or quote Hz only with the speaker's mean F0 attached.

Semitones in practice

Most pitch-tracking apps display semitones alongside Hz. If you can only see Hz, the rough conversion for a span is 12 × log2(top / bottom). A speaker whose voice swings between 160 Hz and 240 Hz across a sentence has about a 7-semitone range, regardless of where in their voice that swing happens to sit.

Statement vs question contours

In English, the canonical statement contour ends with a fall — the final stressed syllable peaks, then pitch drops through the remaining unstressed syllables to a low, often creaky terminus. The canonical yes/no question contour ends with a rise — pitch climbs through the final stressed syllable and continues upward through any remaining material. Wh-questions ("who," "what," "where") usually fall, like statements.

These are defaults, not rules. Real conversational speech mixes contours freely: rising statements signal continuation or invitation to respond, falling questions signal that the speaker already knows the answer or is challenging the listener. The feminine-typical pattern in conversational American English uses rises and held tones much more often than the masculine-typical pattern, in a wider variety of pragmatic contexts.

Stressed-syllable patterns

English is a stress-timed language. Stressed syllables anchor the rhythm and the pitch contour; unstressed syllables compress between them. Within a word, stress falls on a fixed syllable ("record" the verb vs "record" the noun). Within a sentence, stress falls on the words the speaker wants to emphasize, and the pitch peak lands on the stressed syllable of the stressed word.

Equally-stressed delivery — every syllable getting the same loudness, the same length, and the same pitch step — produces what voice teachers call "robot voice." It reads as masculine partly because monotone is masculine-coded and partly because it sounds emotionally flat regardless of speaker.

Gendered intonation differences

Feminine sentence-final patterns

The single most identifiable prosodic difference between female-typical and male-typical conversational English is what happens at the end of a sentence. Feminine speech often holds the final pitch — the last stressed syllable peaks and the contour stays elevated rather than falling — or rises slightly through the final word, even on a declarative statement. Masculine speech defaults to a hard fall, often terminating in vocal fry below the modal range.

This is not the same as upspeak. The held or slightly-rising final is subtle: a few semitones, not the dramatic question-rise of "you're going to the store?" Listeners parse it as engagement, openness, or willingness to keep talking, rather than as a question.

The masculine fry-terminus pattern is particularly costly for trans women. Pitch may have been correctly placed for the body of the sentence, but the final drop into fry exposes the masculine baseline and gets the speaker clocked. Many voice trainers consider eliminating sentence-final fry the single highest-yield prosodic intervention (Hancock et al. 2014, Davies, Papp, & Antoni 2015).

Wider pitch range in conversational speech

Female-typical conversational English uses a wider pitch range than male-typical speech. Estimates vary by study and by speaking task — read speech is narrower than spontaneous, lab speech narrower than naturalistic — but the direction is consistent.

Speech type	Typical female range	Typical male range
Conversational English	~7–9 semitones	~4–6 semitones
Read prose	~5–7 semitones	~3–5 semitones
Animated / expressive	10+ semitones	6–8 semitones

These numbers overlap heavily and shift with affect and context. What matters for training is the direction, not the exact figure: feminine-typical speech is more melodically animated, with bigger excursions on stressed syllables and more frequent contour shifts within an utterance. Masculine-typical speech is flatter, with smaller excursions and longer stretches at a held pitch (Gelfer & Schofield 2000, Hancock, Colton, & Douglas — Intonation and gender perception in transgender speakers, 2014).

The "upspeak trap"

A predictable failure mode in early intonation training is to caricature every statement as a yes/no question — "I went to the store? And got some milk? And then I came home?" This is the upspeak trap. It sounds performative to listeners, it doesn't match how cis women actually speak, and it frequently makes the speaker read as more obviously trans, not less.

Cis women use rising and held finals, but only on a fraction of statements and in pragmatically appropriate contexts (continuation, engagement, listing, softening). They don't use them on every sentence, and they vary the contour. The trap is mistaking the most salient feature of feminine prosody for the only feature.

A useful diagnostic: record yourself reading a paragraph naturally, then listen for whether the rises feel motivated or mechanical. If every sentence ends with the same rise, that's upspeak caricature, not feminine prosody.

Variety beats rises

The fix for a flat contour is not "rise on everything." It's a mix of rises, held finals, gentle falls, and occasional emphatic falls — distributed by what you actually mean, not by mechanical pattern. A delivery with three contour shapes used appropriately reads as feminine; a delivery with one contour shape used everywhere reads as performed.

Common errors

Monotone delivery (the masculine default)

Trans women coming out of years of masculine-presenting speech often default to a narrow pitch range — the male-typical 4–6 semitone span — even after moving their average F0 up. The body of the sentence sounds correctly placed; the contour through it is flat. This reads as masculine even if F0 and resonance are otherwise feminine.

Monotone is the hardest error for the speaker to hear in their own voice. The brain perceives intonation relative to its own production; if the speaker's internal model of "expressive speech" is calibrated to a 5-semitone range, a 5-semitone delivery will sound expressive to them. Recording and listening back is the only reliable way to break this loop — see practice section.

Over-corrected upspeak (the caricature trap)

Already covered above. Worth repeating because it's the most common over-correction. The fix is not less rise but more variety: some falls, some holds, some rises, distributed by pragmatic intent rather than by pattern.

Sentence-final drop into vocal fry

Vocal fry — the creaky, low-frequency register below modal phonation — is gender-neutral in moderation but masculine-coded when used as a default sentence-terminus. Many trans women carry this pattern over from pre-transition speech without noticing it, because pre-transition the fry sat near the bottom of a much lower modal range and was less audible as a register shift. Post-transition, with modal pitch lifted, the fry stands out.

The fix is to terminate sentences in modal voice, even at the cost of holding the pitch slightly higher than feels natural. If the sentence-end drop is reflexive, practice sentences that intentionally end on a held tone or a slight rise (see Practice Sentences) until the modal terminus becomes the new default.

Cis women do use vocal fry, particularly younger American English speakers. The issue is not fry itself but fry as a reflexive sentence-end default that exposes a masculine pitch floor.

Equally-stressed syllables ("robot voice")

When learners focus too hard on pitch placement they often flatten stress — every syllable gets the same loudness and length. This produces a mechanical, recitation quality that reads as unnatural regardless of gender, and as masculine-coded specifically because it lacks the rhythmic animation of conversational feminine speech.

The fix is to let the unstressed syllables get short and quiet. Reduction is half of English prosody; trying to give every syllable equal weight breaks the rhythm. The unstressed "the," "of," "a," "to," "for" should compress almost to schwa — getting them out of the way is what lets the stressed syllables carry the contour.

Stress and emphasis

Word-level stress

Each multisyllabic English word has a fixed primary stress. Misplacing it ("PHOtograph" vs "phoTOGraphy") sounds like a non-native speaker, not like a masculine or feminine speaker. Word stress isn't a gendered variable — it just has to be correct.

Sentence-level emphasis

Within a sentence, the speaker chooses which word to stress, and the choice changes the meaning:

• "I didn't say she stole the money" — implies someone else did.
• "I didn't say she stole the money" — denial of the saying.
• "I didn't say she stole the money — I implied it."
• "I didn't say she stole the money" — implies someone else stole it.
• "I didn't say she stole the money" — maybe she borrowed it.
• "I didn't say she stole the money" — maybe she stole something else.

Same words, six different meanings, distinguished only by which word gets the pitch peak and the duration lengthening. The unstressed words compress around the stressed one.

Feminine-typical speech makes more frequent use of contrastive sentence stress and uses bigger pitch excursions on the stressed word. Masculine-typical speech is more even, with smaller excursions and less contrast.

Emotional color

Pitch animation tracks emotional engagement. Genuine interest, surprise, sympathy, and amusement all show up as larger pitch excursions and faster contour shifts. Feigned animation — forcing rises that don't match the emotional content — reads as performative.

The training goal is not "use more rises" but "let the pitch reflect what you actually feel about what you're saying." A speaker who is bored will sound bored regardless of pitch placement; a speaker who is engaged will sound engaged, and the prosody will track. This is why intonation training tends to plateau if practice is restricted to neutral read sentences — the contour needs emotional content to anchor onto.

Practicing intonation

Mirror sentences

The Davies/Papp tradition uses mirror sentences — short utterances where the trainee listens to a recorded model, then reproduces the contour as closely as possible. The goal isn't memorization but pattern absorption: the ear learns what feminine contours sound like, and production gradually follows.

Drill structure: model says one sentence; trainee repeats; trainee compares recordings side-by-side; trainee adjusts. Sentences are graded from short declaratives ("I went to the store.") to longer multi-clause utterances with internal phrasing ("Well, I went to the store, but they didn't have what I wanted, so I came home empty-handed.") (Davies, Papp, & Antoni 2015).

Shadowing recorded voices

Shadowing is real-time imitation of a continuous speech sample — audiobook, podcast, YouTube video — where the trainee speaks along with the model at a roughly half-second lag. This is harder than mirror sentences because it forces the trainee to produce contour without time to plan, which is closer to conversational speech.

Shadowing teaches pitch and prosody simultaneously rather than as separate variables, which matters because the two are coupled in natural production — trying to drill them in isolation tends to produce the "robot voice" failure mode. Use cis female speakers as models, ideally ones whose speaking style you'd like to inhabit. See TransVoiceLessons for shadowing demonstrations and model selection guidance.

Recording and listening back

The single highest-yield practice habit is recording yourself and listening back the next day. The brain reliably under-perceives the flatness of its own intonation in real time — the speaker hears the contour they intended, not the contour they produced. Delayed playback breaks this loop because the playback is heard as someone else's voice.

Specific protocol:

1. Record a 60-second free-speech sample (describe your day, read a paragraph, narrate a photo).
2. Don't listen immediately.
3. Listen the following day with attention to: (a) sentence-final patterns, (b) overall pitch range in semitones, (c) presence of fry termini, (d) syllable stress contrast.
4. Note two things to change. Re-record.
5. Repeat.

For pitch-tracking apps and longitudinal logging see Recording & Self-Monitoring.

Practice material

The Practice Sentences page has graded material organized by speech move and phonetic load. For intonation work specifically:

• Daily script sentences train sentence-final patterns in functional context (greetings, requests, declines).
• Tone toggles train contour variation across emotional registers (warm vs neutral vs firm) on the same words.
• Reading passage trains contour stability over sustained connected speech.

For drill technique — sirens, glides, contour-matching exercises — see Exercises.

When intonation breaks under stress

Intonation collapses first under stress, often before pitch. A speaker who can produce a credible feminine voice in calm settings will frequently flatten their contour under emotional load — anger, fear, embarrassment, fatigue — even if they hold pitch and resonance reasonably well.

The mechanism is attentional. Pitch and resonance are easier to monitor as single targets ("stay above 165 Hz," "keep the resonance forward") because they're slow-moving variables. Intonation requires moment-to-moment contour shaping across every syllable, and that bandwidth disappears first when cognitive load increases.

Practical implications

• Intonation must be over-trained to survive stress. Comfortable production in practice isn't enough; the contour has to be automatic before it will hold under load. Plan for several months of consistent shadowing and recording before expecting stress-resistance.
• The phone is a stress test. Phone calls strip resonance cues (no facial cues, narrowband audio compression cuts upper formants) and force the listener to rely more heavily on F0 and prosody. A flat contour that passes in person will often clock the speaker on the phone. See Voice Across Contexts for phone-specific strategies.
• Recovery after a flatten-out: drop pitch slightly, restore the contour first, then rebuild pitch. Trying to hold pitch while reanimating the contour usually fails because both demand foreground attention and only one can be foregrounded at a time.

When to escalate

If recording reveals persistent flat contour after several weeks of deliberate intonation practice, or if the contour is fine in solo practice but collapses every time in real conversation, the issue is usually beyond self-correction range. See Troubleshooting & Plateaus for the diagnostic flow, and Working with an SLP for when to bring in professional help — intonation is one of the variables a trans-experienced speech-language pathologist can hear in your voice that you can't.

Sources

• Davies, S., Papp, V. G., & Antoni, C. (2015). Voice and communication change for gender nonconforming individuals: Giving voice to the person inside. International Journal of Transgenderism, 16(3), 117–159. https://www.tandfonline.com/doi/abs/10.1080/15532739.2015.1075931
• Hancock, A. B., Colton, L., & Douglas, F. (2014). Intonation and gender perception: Applications for transgender speakers. Journal of Voice, 28(2), 203–209. https://pubmed.ncbi.nlm.nih.gov/24315656/
• Gelfer, M. P., & Schofield, K. J. (2000). Comparison of acoustic and perceptual measures of voice in male-to-female transsexuals perceived as female versus those perceived as male. Journal of Voice, 14(1), 22–33. https://pubmed.ncbi.nlm.nih.gov/10764114/
• Wolfe, V. I., Ratusnik, D. L., Smith, F. H., & Northrop, G. (1990). Intonation and fundamental frequency in male-to-female transsexuals. Journal of Speech and Hearing Disorders, 55(1), 43–50. https://pubmed.ncbi.nlm.nih.gov/2329796/
• Hancock, A. B., & Garabedian, L. M. — Transgender voice and communication treatment: A retrospective chart review of 25 cases (2013).
• Hirsch, S., & Boone, D. — The Modification of Speech and Voice in the Transsexual Patient (clinical text, multiple editions).
• Mills, M., & Stoneham, G. — The Voice Book for Trans and Non-Binary People (2017, clinical practitioner reference).
• WPATH Standards of Care, version 8 — voice and communication chapter.
• TransVoiceLessons (YouTube) — practical drill demonstrations for intonation, pitch, and resonance.