How to give a clear science talk
A good science talk tells one clear story: what question you asked, how you tested it, what you found and what it means. Plan the take-home message first, give each slide one point, make every figure readable from the back row, be honest about limitations, and rehearse until you finish on time.
The short version
- Decide your one-sentence take-home message before making any slides.
- Structure: title → outline → question and background → methods (brief) → results → discussion and limitations → conclusion.
- One message per slide, ideally written as the slide's headline in a full sentence.
- Figures must be readable: large labels, few lines, units on the axes, the key result highlighted.
- Rehearse aloud and time it; plan to finish a little early and leave room for questions.
Plan the story
Most weak talks have the same cause: the speaker tries to show everything they did. Your audience will remember very little, so choose what that little should be. Write down, in one sentence, what you want people to leave with ("Learning the Krebs cycle with a song improved recall after one week, but only for students who also tested themselves"). Every slide should help support that sentence; anything that does not can go into backup slides.
Next, think about who is listening. A lab meeting can handle jargon and detail; a general department seminar or a class needs more background and fewer acronyms. If in doubt, explain more, not less: experts rarely mind a clear 30-second introduction, and everyone else needs it.
Build the slides
Title slide
Give a title that states the topic or, better, the finding, plus your name, affiliation, and the names of co-authors. Keep it on screen while you introduce yourself and the question.
Outline
For talks longer than about 10 minutes, a short outline slide (three to five items) helps listeners know where you are going. In a 5-minute talk you can say the outline instead of showing it.
One message per slide
Use the slide headline to state the point as a sentence, for example "Treated cells grew twice as fast as controls", rather than a label like "Results 2". Then support it with evidence: a graph, an image or a diagram. This "assertion–evidence" style, promoted by engineering educators, means the audience can grasp the point even if they miss what you say. Avoid paragraphs of text on slides; people read faster than you speak, and then stop listening.
Figures people can read
- Axis labels and tick numbers large enough to read from the back of the room (much bigger than in a paper figure).
- Units on every axis; say what the error bars show (standard deviation, standard error or confidence interval).
- Only the data needed for this slide's message; split busy figures across slides.
- Highlight the key comparison with color or an arrow, and choose colors that also work for color-blind viewers.
- Tell the audience what the axes are before you tell them what the data show.
Methods, results and discussion
- Background and question. Why does this matter, what was unknown, what exactly did you ask? State your hypothesis.
- Methods, briefly. Give enough for the audience to judge whether the result is believable: the design, the sample size, the key measurement, the controls. Put protocols and settings in backup slides.
- Results. One finding per slide, in a logical order. Report effect sizes with uncertainty, not just "significant" (see p-values and confidence intervals).
- Discussion: limitations and assumptions. Say plainly what the study cannot show: small sample, one population, short follow-up, a test's assumptions that may not hold, a correlation that is not proof of cause. Audiences trust speakers who name the weaknesses before being asked.
- Conclusion slide. Restate the take-home message and, if useful, one or two next steps. Leave this slide up during questions instead of a slide that only says "Thank you" or "Questions?".
- Acknowledgments. Thank collaborators and funders, briefly.
Timing, rehearsing and questions
Timing. A common rule of thumb is about one slide per minute, though complex data slides take longer. Running over is the most common and most resented mistake; at conferences, chairs may cut you off. Aim to finish a minute early.
Rehearse aloud, standing up, with a timer, at least two or three times; once in front of colleagues who will give honest feedback. Recording a practice run and watching it is uncomfortable but very effective. Know your first two sentences by heart so that nerves do not derail the start.
Delivery. Face the audience, not the screen. Use a pointer sparingly. Pause after showing a new figure so people can look at it.
Questions. Listen to the whole question, repeat or rephrase it so everyone hears it, then answer briefly. If you do not know, say so and offer to follow up; guessing is worse. If a question is really a long comment, thank the person and move on. Backup slides with extra detail are useful here.
Memory aid
Use the shape of an hourglass: start broad (why it matters), narrow to your specific question and methods, stay narrow through results, then widen again in the discussion and conclusion (what it means, what it cannot show, what next).
And a one-line test for each slide: "If they read only the headline, do they get the point?"
A talk checklist rhyme
Check yourself
What should you decide before making any slides?
Your one-sentence take-home message, and who the audience is.
What is better as a slide headline: "Figure 3" or "Mice on the new diet gained 20% less weight"?
The full-sentence headline. It states the point, so the audience understands the slide even if they miss part of what you say.
Why should you discuss limitations yourself?
It shows you understand your study, helps the audience judge the result fairly, and builds trust. Expert listeners will notice the weaknesses anyway.
Which slide should be on screen during questions?
Your conclusion slide with the take-home message, not a slide that only says "Thank you".