How to analyze a speech instead of quoting its lines

Most guides to examining a speech assume you have the text in front of you and can quote it freely. For historic speeches that assumption is often wrong: much of the material people most want to work on is still owned, sometimes by estates that license it actively.

The good news is that quoting was never where the analysis lived. What follows is a set of questions people actually ask, answered in a way that works when reproduction is off the table.

What to take away

Three objects: the text, the performance, the event

The text, the performance and the event are three different objects, and

The text is a sequence of words, in some particular version. The performance is what a specific person did with their voice and body on one occasion. The event is the whole thing: a room, an audience, a moment in time, a speaker with a history, a reason for everyone being there.

Claims about one do not transfer to another: a brilliant text can have a flat performance, and a performance can rescue weak text.

An event can matter enormously while its speech is ordinary. Decide which you are writing about and say so in the first paragraph; most confused speech analysis silently switches between the three.

Working without reproduction

Yes, and the results are often better, because the constraint pushes you away from

the two weakest moves in the field: retyping the famous passage, and pointing at individual sentences to admire them.

What remains available without reproducing a word:

A brief phrase can sometimes be discussed as the object under examination when there is genuinely no other way to make the point. A paragraph cannot, and a page built out of extracts is not analysis at all.

Choosing a version, and naming it

Whichever one you can name. The problem is not choosing wrong; it is not saying

A single speech can leave behind a prepared draft, a reading copy with the speaker's marks, an advance text given to the press, a stenographic record, a broadcast recording, an official published version, and an anthology extract. These disagree, sometimes substantially. All of them are real documents. None of them is "the speech".

Practical handling: name the document, name where it is held, and say what it is a record of. Where two versions differ in a way that affects your argument, say so rather than picking the convenient one. And treat "as prepared for delivery" as the warning label it is.

The move as the working unit

The sentence is too small and the speech is too big. The useful unit is the

A move might be establishing common ground, conceding a point, introducing the central claim, addressing an objection, naming what is being asked, or signaling the close. Moves are usually between fifteen seconds and two minutes. They have audible boundaries, because speakers change pace and pitch when the job changes.

Working in moves makes speeches comparable. Two addresses on unrelated subjects can be laid side by side if you have described both as sequences of functions. Classical teaching divided an oration into parts with distinct jobs, and the summary of dispositio is a serviceable starting grid, provided the labels are treated as your map rather than as the speaker's plan.

Mapping the structure in three passes

A three-pass method that works on a recording alone.

First pass: listen without taking notes. You are establishing the overall shape and, crucially, your own untutored reaction, which you should write down immediately afterwards in a couple of sentences. That reaction is data you cannot recover later.

Second pass: mark the boundaries. Note the timecode every time the job changes. Do not name the moves yet, just mark where they start. Pace and pitch changes will tell you.

Third pass: name each move by its function and record its duration.

You now have a table (timecode, function, length), that describes the speech without reproducing it, can be compared to any other speech treated the same way, and will show you proportions that are invisible when reading.

Separating the effect from your reaction

You cannot, entirely, and pretending otherwise is the field's biggest source of

Separate the reaction from the mechanism. Write down what you felt first, then go looking for what in the speech might have produced it. Doing this in the other order produces confirmations rather than findings.

Check against the audience track. If a passage moves you and the room was silent, something interesting is happening: usually that you know things the audience did not, or that a later reputation is doing the work.

Correct for outcome. Knowing how the story ended changes how the speech sounds. A passage that seems prophetic may simply be one of several predictions, the others forgotten.

Ask whether the effect survives paraphrase. If you can describe what the passage does and the description is still interesting, it is probably about the speech. If it collapses into "it was beautifully put", it may be about you.

Reading the audience track

More than most people use, and less than it appears.

Useful signals: whether applause starts at once or must be led; whether laughter is general or scattered; whether the room quiets at certain moments.

Whether people applaud at a natural break or interrupt; how long applause runs before the speaker resumes.

Serious cautions. Broadcast audio is mixed, and audience level is a production decision. Recordings are sometimes made in near-empty halls, or re-recorded later. Applause in a hostile or a compulsory setting means nothing about persuasion. Studio recordings and audience recordings of the same speech sound completely different. Establish what kind of recording you have before you interpret a single reaction.

Analyzing delivery

Concentrate on change rather than absolute values.

Pace, and where it changes. Most speakers speed up under pressure and slow before something they consider important. A speaker who slows into a passage is telling you what they think matters.

Pause placement. Pauses before a phrase and pauses after it do different jobs. The first creates expectation; the second gives the audience room to react.

Pitch at sentence ends. A drop signals completion. Held or rising pitch signals more coming. Speakers control the audience's applause almost entirely with this.

Volume floor, not ceiling. How quiet a speaker is willing to go is a better measure of control than how loud they get.

Breath. Audible in most recordings. Where a speaker breathes tells you how the sentences were built and whether they are in trouble.

Where they look up. In a scripted speech, the moments of eye contact are chosen, and they cluster around what the speaker wants to land.

Writing it up

• State what you could not check. Versions you could not reach, a recording you could not hear, a translation you had to rely on. This is the part that makes an analysis usable by someone else.

Related: cultural speeches for material that reaches you through a translator before reaching you at all.

political speeches for reconstructing the occasion, award speeches for a genre whose primary document is a recording, not a text.

commencement speeches for one where published text and delivered speech routinely differ.

Common questions

Who actually wrote it? Usually more than one person, and often a committee. The speechwriter is a normal fixture of political, corporate and institutional settings, and has been for a long time. Drafts circulate, departments object, and the final text is negotiated.

This has real consequences for analysis. Claims about "the speaker's style" based on a single speech are weak. Inconsistencies of register within one address are frequently the visible joins between contributors. And the material a speaker adds live, departing from the script, is disproportionately revealing, because it is the part that is unambiguously theirs.

Where drafts survive, comparing them to the delivered version is one of the most productive exercises available. What was cut is often the most interesting document in the file.

What about a speech in translation? Say so, and say whose translation. Then be careful about what you claim.

Analysis of structure, proportion, timing, argument, and audience response survives translation intact. Analysis of sound, rhythm, register, and word choice does not. If your argument depends on the second kind and you are working from a translation, you are analyzing the translator.

What makes an analysis wrong? Four recurring errors.

Hindsight. Reading a speech through what happened afterwards, and treating the speaker as having known it.

Quote-centring. Building the whole reading around the famous passage. That passage is famous because it was extractable, which is a different property from being central.

Assuming the audience heard what you hear. They had different knowledge, different expectations, and one pass at it.

Outcome bias. Judging the speech by what followed. Events have many causes and speeches are rarely the largest one.

Do I need a recording to write anything useful?

No, but you need to say that you do not have one. A text supports claims about argument, arrangement and visible technique. It supports nothing about pace, pause or emphasis.

How long should an analysis be?

As long as the evidence supports. A well-sourced page with three observations and a stated limit is worth more than a chapter of description, and it is what most readers actually want.