How an IA is actually marked

Written for the sciences, but the shape holds across subjects: a small number of criteria, each with published band descriptors, marked by your teacher and then moderated by a stranger.

Start from the stranger

The single most useful mental move is to stop writing for your teacher. Your teacher watched you do the practical. They know which variable you controlled because they saw you control it. The moderator sees a PDF.

Almost every mark lost in a competent IA is lost this way. The work was done, and the evidence that it was done never made it onto the page. Before you submit, read your own IA as though you have never met yourself.

The research question does more work than the experiment

A question that names a dependent variable, an independent variable and a range does three jobs at once. It tells the reader what you measured. It justifies your method. And it gives your conclusion something specific to answer. A vague question makes every later criterion harder, because nothing can be judged against it.

"How does temperature affect enzyme activity" cannot be assessed. "How does temperature between 20 and 60 °C affect the rate of catalase-catalysed decomposition of hydrogen peroxide, measured as oxygen volume per unit time" is answerable, and the method now writes itself.

Controlled variables are a written artefact, not an intention

Listing three controlled variables earns very little. Saying how each was controlled, and what would have happened had it not been, is a different level of the descriptor. The examiner is looking for evidence that you understood the causal structure of your own experiment.

Uncertainty is where the marks actually are

This is the most reliably under-written section in the IAs we read. Recording an instrument's uncertainty is the easy half. Propagating it through your calculation, and then saying whether the resulting error bars are large enough to threaten your conclusion, is the half that separates bands.

If your error bars overlap across your entire range, that is a finding, and saying so plainly earns more than pretending the trend is clean.

Evaluation means weighing, not listing

A list of weaknesses is a common way to lose marks while appearing thorough. "Human reaction time introduced error" is a sentence anyone could write about any experiment. The question is which weakness mattered most, how much it plausibly shifted your result, and what specifically you would change, naming the actual apparatus or protocol.

The conclusion answers the question you asked

Go back to your research question and answer it in one sentence, with a number. Then say how well your data supports that answer, and how it compares to a published value. A conclusion that drifts into general biology is a conclusion that stopped being assessed.

Want this applied to your draft? A criterion-by-criterion pass is $42, or $80 with a recheck after you revise. Comments in the margin of a copy.