Most students approach the IB Physics HL internal assessment expecting it to reward thoroughness: a complete method, a clean data table, a conclusion that matches the expected result. The current Scientific Investigation format rewards something different. Four criteria structure every mark—Research Design, Data Analysis, Conclusion, and Evaluation, each worth 6 of the 24 total—and at the top band, what’s being assessed is interpretive reasoning, explicit uncertainty treatment, and the quality of evaluative thinking, not how comprehensively the procedure was documented. Students working from pre-2025 rubrics risk directing most of their effort toward the parts that matter least.
- Frame your research question and state the physics model you expect it to test—one or two sentences.
- Map your variables: independent, dependent, and the single biggest uncertainty source, with a brief note on how you’ll measure each.
- Draft the method by explaining the reason behind each control variable and measurement choice—what it prevents, what it assumes, and what uncertainty it introduces—rather than listing steps.
- Run a pilot: produce one graph in the format you intend to use and confirm you can compute or propagate uncertainty on at least one derived quantity.
- Collect full data and process it; keep raw tables and extra calculations in appendices, and use the main body to explain what the processed results mean in relation to the model.
- Write Conclusion and Evaluation as separate jobs—Conclusion states what the data show about the model; Evaluation addresses how design choices and uncertainty limit what you can claim.
- Package the report within 3,000 words, with core analysis in the main body rather than in appendices, and treat collaboration as data-collection logistics only unless your teacher explicitly approves more.
Choosing a Research Question
Strong Scientific Investigation topics satisfy three conditions simultaneously: feasibility, clear quantifiability, and genuine grounding in physics theory. All three matter, and all three have to hold at once. A question that’s easy to implement but loosely connected to any real model leaves little to say in Data Analysis or Evaluation—no matter how tidy the apparatus is. To earn marks across all four criteria, the research question has to let you define measurable variables precisely, link them to a specific theoretical relationship, and treat uncertainty as a first-class feature of the investigation rather than something to acknowledge in the final paragraph.
A common failure mode is the single-variable linear relationship that produces one straight-line graph and not much else. When data only support “it is proportional” as a conclusion, there’s limited scope for deeper interpretation, uncertainty propagation on derived quantities, or any serious discussion of where the model breaks down. A stronger research question builds in structural complexity—a non-linear dependence, competing effects, or a clear theoretical prediction that the data either confirm or challenge—so there are multiple substantive claims to make, not just one. A useful stress test: if the planned data can’t support at least three distinct analytical statements about the relationship or model, the question is probably too narrow for top-band work—and even a question that clears that bar will lose marks if the method logic behind it is invisible to the reader.
Designing the Method and Processing Data
Under the Research Design criterion, marks are most commonly lost not because a method is wrong, but because its logic is invisible. A high-scoring design makes its reasoning explicit: why each control variable matters, how measurement choices limit random and systematic uncertainty, and which assumptions the setup relies on. Rather than a bare step-by-step list, annotating the key decisions lets examiners follow what each choice prevents, what it assumes about the system, and how it shapes the quality of the data that follow. The steps tell them what you did; the annotations tell them you understood why.
Data Analysis demands more than plotted points. The current format expects graphical and statistical treatment that makes the model visible—linearizing relationships where that clarifies the physics, propagating uncertainty through to derived quantities, and reading processed results against the theoretical relationship you set out to test. Appendices are for supporting material only; raw tables, additional graphs, and working calculations belong there so the 3,000-word main body stays focused on what the data actually say. Getting the processing right is the precondition for everything that follows—but turning clean results into a genuinely rigorous evaluation is a different and harder skill.
Writing a Distinctive Evaluation
Top-band Evaluation is sharply different from a superficial acknowledgment of error. High-scoring work identifies specific strengths and weaknesses in the experimental design, links each limitation to its concrete effect on the results and on the conclusion drawn, and proposes realistic improvements or extensions. A generic list of “human error” or “equipment limitations” doesn’t meet this standard. Naming a systematic error without explaining how it produced a bias or pattern in the data is surface acknowledgment—the examiner needs the mechanism, not just the label.
A practical structure is to build every evaluative point in three moves: name the design feature or limitation, explain how it could distort measurements or restrict how far the conclusion applies, then propose a specific, feasible modification that would reduce or quantify the effect. This discipline also keeps Conclusion and Evaluation distinct—Conclusion addresses what the data indicate about the model; Evaluation addresses how the design constrained the quality of that evidence. Both depend on choices the student personally owns, which is why the line between sharing practical logistics and genuinely independent analysis matters more in these sections than anywhere else.
Collaboration Boundaries and Self-Audit
The individual-authorship requirement is the Scientific Investigation’s clearest structural rule and, in practice, one of its most commonly misread ones. IB examiner instructions for the sciences are direct on this point: each student submits one investigation, marked out of 24 and externally moderated, with collusion explicitly listed as a form of academic misconduct alongside plagiarism. In practice, this means your research question, processing choices, and interpretation need to be demonstrably your own. Any collaboration beyond basic practical logistics should be treated as a risk unless your IA coordinator explicitly approves it.
- One-pass-per-criterion: Read the draft four times, each time highlighting only text that directly earns that criterion—anything unhighlighted is a word-count risk.
- Criterion log: After each pass, write one line noting which parts of the draft are well-supported and which are thin; thin sections show you exactly where to revise.
- Fix-first rule: If your Evaluation lists errors without showing mechanism → effect on result → specific improvement, rewrite each point into that three-step structure.
- Word-budget rule: Move any reasoning essential to following the result out of appendices and into the main body; keep appendices for supporting material only.
- Cadence: Run this loop after the pilot graph, after full data processing, and the day before submission; stop when each criterion log line has matching highlighted evidence in the draft.
Twenty-four marks is a significant fixed allocation in a course where written-paper performance can vary considerably. A well-executed investigation doesn’t soften a bad exam day—it moves your scoring baseline before you sit the paper. Working back through the IA after submission with the four-criterion framework in view also reveals something that matters beyond the mark: the investigation isn’t a standalone task but a compressed version of how the entire specification thinks about physics, and understanding it at that level is more useful than treating it as a checklist of skills to patch before the exam.
Engaging With the Specification
The IB Physics IA gets filed under internal assessment logistics before it gets understood as one of the more intellectually revealing tasks in the course—and that order of operations is worth reversing. The four-criterion, 24-mark structure rewards the same skills that carry students through the written papers: precise variable definition, uncertainty-aware analysis, and conclusions proportional to the evidence. A research question chosen for conceptual depth, a method whose reasoning is visible on the page, an evaluation that traces each limitation to a real effect rather than listing them by category—these aren’t exam-technique adjustments. They’re what doing the physics correctly looks like. Students who approach the investigation that way aren’t just securing 24 marks; they’re working with the specification rather than around it.
