Reviews
Part of How to analyze a speech instead of quoting its lines
How to catch weak speech analysis in about a minute
Best speech analysis names its object, version and evidence early. Five failure shapes, what a good one contains, and four checks that take about a minute.
Published analyses of famous speeches fall into a few recognizable shapes, and most of them fail in the same way. They say what the speech contains without saying what the speech was trying to do or whether it worked.
This page is about telling a good analysis from a bad one, whether you are reading someone else's or checking your own. It reproduces no passages and names no speaker.
What to take away
- A good analysis names its object, its version and its evidence in the first paragraph.
- The common failure shapes are recognizable in a minute, and each one has a characteristic missing ingredient.
- An analysis that could have been written without hearing the speech usually was.
Five failure shapes
- The device-spotting essay. Identifies figures, labels them, and stops. Missing ingredient: what problem each device solved and how the writer knows
Five failure shapes
Failure shape
- Device-spotting
- Problem solved
- Hindsight
- What was uncertain
- Biography
- Occasion
- Paraphrase
- Any claim
- Greatness
- Falsifiable criterion
Missing ingredient
- Device-spotting
- How writer knows
- Hindsight
- Biography
- Audience knowledge
- Paraphrase
- Greatness
What a good one contains
Element / What it looks like
- A named object
- Text, performance or event, chosen and stated
- A named version
- Which document, held where, a record of what
- A stated question
- Something that could have come out the other way
- Structural description
- Movements, proportions, order, in the writer's own words
- Occasion evidence
- What the audience knew, what preceded it, what was pending
- A delivery observation
- Only where a recording was consulted
- A response record
- Where it exists, cited; where it does not, said
- A boundary
- What the analysis cannot settle and why
The last row is the strongest single indicator. Writers who state their limits
What a good analysis contains
- A named object
- A named version
- A stated question
- Structural description
- Occasion evidence
- A delivery observation
- A response record
- A boundary
Reading a published analysis quickly
Four checks take a minute and settle most of it.
Four one-minute checks
- First page names the version used?
- Pace claims say what was listened to?
- Speech order preserved, not reordered?
- Piece says what would count as failing?
- Does the first page say which version the writer used? If not, treat every claim about wording as unsupported.
- Are there claims about pace, pause or emphasis? If so, does the piece say what was listened to?
- Is the speech's own order preserved in the description, or reordered to suit the argument?
- Does the piece say what would have counted as the speech failing?
A piece that fails all four is a summary with adjectives. The discipline behind each of the four is set out in speech analysis. Where a writer does count something, whether devices, sentence lengths or response points, the methodological cautions collected under content analysis apply, and most published counting in this field ignores them.
The problem with academic and popular versions alike
Popular analyses tend to be hindsight essays, because the outcome is what makes the speech interesting to a general reader. Academic ones tend to be device-spotting essays with better sourcing, because the field's teaching tradition rewards identifying figures.
The scholarly tradition has its own name for the enterprise, and the survey of rhetorical criticism is worth reading for how many competing methods sit under it.
Both problems share a root: describing a text is easier than reconstructing a situation. Situation work is archival, slow, often inconclusive, and it turns an essay about a speech into an account of an event.
The kind of evidence involved is set out in political speeches. The comparable reconstruction problem in wartime material, where the record was controlled rather than merely lost, is in war and peace speeches.
Checking your own
Print your draft and mark every sentence that makes a claim. For each mark, ask
Checking your own draft
Does an artifact support this claim?
keep, cite the artifact
label inference or cut assertion
Then ask what evidence would have changed the conclusion. If nothing would have, the piece is not an analysis, and the honest fix is to restate it as a description, which is a legitimate thing to publish.
Where the material is older and freely reproducible, the temptation is to fill the gaps with quotation instead, and public domain speeches explains why availability is not an argument for using it. Ceremonial material tempts the opposite error, filling the gaps with the outcome, which award speeches addresses directly.
Common questions
Is an analysis without a recording worth anything?
Yes, if it says so. Text-based analysis supports claims about argument, arrangement and visible technique. It cannot support claims about delivery, and the good ones do not try.
Should an analysis reach a verdict?
It should reach a conclusion about the question it set. A verdict on greatness is a different genre, and the two are usually mixed.
How long should one be?
As long as the evidence supports. A well-sourced page beats a chapter of description, and the length of published analyses is mostly a function of where they are published.
What is the single best improvement to most analyses?
Naming the version in the first paragraph. It disciplines everything that follows, because half the loose claims cannot survive being tied to a document.
