Analysing a speech doesn’t mean judging it, it means taking it apart in layers: occasion, structure, language, delivery, and the one moment it’s remembered for. Anyone who works through these five layers on three speeches by other people afterwards understands more about their own text than after thirty speeches merely consumed. Here’s the model, with a worked example at the end.
Layer 1: occasion and audience
Before a single sentence gets analysed, one question comes first: what did this speech need to achieve? A eulogy needs to console, a campaign speech needs to mobilise, a graduation speech needs to sum up a class that won’t exist in this form again afterwards. Note the goal of the speech in one sentence, and in a second sentence who was in the room: age, expectations, prior knowledge. These two sentences are the benchmark for everything that follows. A device that works in a policy speech to a specialist audience can fall completely flat in an emotional speech, and the other way round.
Layer 2: structure, counting the stations
Almost every good speech has a manageable number of stations, usually three. Read or listen through the speech once in full and mark only the transitions: where does the topic change, where does the tone change, where does the speech tip from analysis into a call to action? Most speeches split afterwards into three or four blocks, often with a recognisable pattern: problem, cause, way out. Or: past, present, future. Write the stations one under another in keywords, no more than one sentence per station. If you find more than five stations, the speech was probably too sprawling to stay in memory, and that’s already a finding in itself. The guide to structuring a speech shows how to plan these stations yourself, rather than just spot them.
Layer 3: language, counting the devices
Now it’s down to the sentence. Read the speech a second time, this time watching only for rhetorical devices, and count them: how many anaphoras, how many tricolons, how many images instead of abstract terms? A good speech rarely needs more than three or four strong devices in total, concentrated at the spots meant to land. What stands out usually isn’t the quantity but the placement: the strongest device in a speech sits almost always either in the first two minutes or the last third, rarely in the middle. Note the line and type for every device you find. This list is the raw material for layer five.
Count the images alongside it: concrete comparisons instead of abstract terms, a figure translated into an everyday scale, say, or an object standing in for a whole issue. Speeches with a few clear images stay in memory longer than speeches with many abstract nouns. If a speech delivers almost nothing but terms like “challenge,” “opportunity” or “change” in layer three, and barely any concrete images, that’s a finding in its own right, however skilfully the sentences are otherwise built.
Layer 4: delivery, timing pauses, pace, gaze
If a recording exists, the analysis gets concrete. Time the pauses with a watch: how long does the speaker stay silent after the core statement, how long before a turn? Two to four seconds are no rarity in strong speeches, considerably longer than it feels on first listen. Also watch for changes of pace: is a sentence deliberately spoken slower than the ones before it? And observe the gaze, if video is available: does it settle on one spot in the room, does it wander, does it search for the text on the page? These observations can’t be drawn from a transcript alone, which is exactly why this layer is worth doing with audio or video rather than text alone.
Layer 5: the one moment
Almost every great speech has a single spot it’s remembered for, rarely longer than a sentence. In Kennedy’s inaugural address, it’s “ask not what your country can do for you.” In many good campaign speeches, it’s a single image, a single figure, a single sentence that gets quoted later while the rest of the speech is forgotten. The task in this layer: find that one moment and work out what carries it. Usually it’s a combination of the previous four layers landing at exactly one spot: the strongest device, placed at the point with the longest pause, in the last third of the structure. The guide to speech opening lines shows how often this one moment is already set in the first few sentences.
Exercise: the Gettysburg Address in five layers
Lincoln’s Gettysburg Address makes a good exercise, because it’s public domain, short, and fully preserved: 272 words, spoken in around two minutes. Occasion: the dedication of a soldiers’ cemetery in the middle of a civil war, an audience of mourners and politicians mixed together. Structure: three blocks, clearly marked by tense, past (“four score and seven years ago”), present (“now we are engaged in a great civil war”), future (“that government of the people, by the people, for the people, shall not perish from the earth”). Language: a single load-bearing tricolon at the close, otherwise strikingly little ornament for a speech of this weight. Delivery: reports describe a very calm, almost understated delivery, with the audience barely registering in the moment that history was being spoken. The one moment: the final three lines, with the threefold preposition “of, by, for the people,” still the most quoted formula of American democracy today.
Anyone who works through these five layers on three different speeches, a campaign speech, a graduation speech and a eulogy, say, builds a comparison framework that never emerges from simply watching thirty speeches. Consumption without taking things apart stays entertainment. Only stopping, counting and noting turns a heard speech into a blueprint.
Transfer: the findings into your own text
The analysis is only half the work. The second half: carrying the patterns you found across into your own speech, not copying them. If the campaign speech you analysed places its strongest device in the last third, check whether your own speech even has a device at that spot, or just keeps running. If the analysed speech works with a long pause before the closing line, plan that pause deliberately into your own text, rather than leaving it to chance.
The transfer works best piece by piece, not all at once. For your first own speech, take exactly one finding from the analysis, the placement of the strongest device in the last third, say, and work in only that one point deliberately. For the second speech, the next finding joins it, the planned pause, the reduced ornament in the middle. Anyone who tries to cram all five layers into a new text at once usually loses their own voice in the process, and that’s the thing that counts for more than any single technique in the end.
eloqole suits exactly this transfer: the draft is built from the same building blocks the analysis makes visible, structure in stations, one deliberately placed device, a planned moment at the close. In the teleprompter, you can then rehearse until the pause and pace sit the way they did in the speech you used as a model.