Written analysis
Finished figures, your sub-hypotheses, and a statistical result if you chose to run one. Do not start writing while the graphs are still changing.
- Each sub-hypothesis examined in turn, and answered
- Every pattern explained, not just described
- Anomalies identified and explained
- Data quoted in the prose as evidence
- A written analysis that answers the fieldwork question
Eight marks, nearly a third of the IA, for explaining, not describing.
This is the largest criterion in the investigation and the one where reports separate. Almost everybody can describe what a graph does. The band above asks why it does it, and the band above that asks you to account for the parts that do not fit.
State what the geographical theory predicts for this variable and give the mechanism. Not "Bradshaw predicts velocity increases" but why it does.
Describe the pattern with figures quoted from your data: values, the range, the correlation coefficient, whether it is significant.
The marks. Where your river follows the theory, explain the mechanism operating here. Where it doesn't, explain what about this river caused that.
Run those three moves for each sub-hypothesis in turn, then answer the fieldwork question across all of them. Most reports do move 2 well, move 1 briefly, and skip move 3, which is precisely the difference between the middle bands and the top one.
Everything below is how we suggest you actually do it.
The difference between describing and explaining
Everyone is told to explain rather than describe, and almost nobody is shown the difference on their own data. Here it is.
A description says what the graph does: the direction, the size, where it changes. It is necessary; the mark scheme wants your patterns described and linked to your hypotheses, but on its own it caps you in the middle.
An explanation says what caused it, through a mechanism, in this river. It answers "why?" one more time than feels necessary. If your sentence would be equally true of any river anywhere, you are still describing.
Every model sentence in this guide is here to show you a structure: what a good sentence does, and in what order. None of them is a sentence to lift. Where they quote figures, those figures are one group’s and they are not yours.
Borrowing one would not work anyway. Your analysis has to be built out of your own dataset, your own sites and your own hypotheses, and the IB is explicit that the writing, the justification of methods, the analysis and the conclusion must be entirely your own work. A borrowed sentence sits visibly on top of a report rather than in it, and your teacher has read the guide too.
“Discharge increased with distance from the source. This supports the Bradshaw model and the hypothesis.”
True, linked to the hypothesis, and it stops exactly where the marks start. It would also be true of almost any river.
“Discharge rose from 0.001 m³ s⁻¹ at site 1 to 0.234 m³ s⁻¹ at site 8, in the direction the hypothesis predicted, then collapsed by 96% to 0.009 m³ s⁻¹ at site 9, only 200 m further down the Asse at the second of the two ‘Near the flats’ sites. The rise is what a larger, smoother channel predicts: as hydraulic radius increases, proportionally less water is in contact with the bed and banks, so less of the river’s energy is spent on friction. The collapse is not consistent with that, and 200 m is far too short for the downstream trend itself to reverse. The cause therefore has to be something about site 9 itself rather than about distance from the source.”
Same data. Now the figures do the evidence work, a mechanism is named, and the anomaly is tied to a hypothesis and a place, which is the linking to theory and location the top band asks for. One thing is deliberately missing: which feature of site 9 is responsible. That is your work.
Bedload is the variable students most often describe and least often explain. Two mechanisms are already in your vocabulary: attrition and abrasion, which wear stones smaller and rounder the further they travel.
The third is the one that turns velocity and bedload into a single argument. The Hjulström curve relates a river’s velocity to the particle sizes it can erode, transport and deposit, so a fall in velocity has a predictable consequence for what the bed is made of. Your own velocity readings can be traced onto it. The flowmeter manual reproduces the curve in section 3.3, and step 3 links the manual.
Naming the curve, and using your measured velocity with it, is exactly the move that separates a described pattern from an explained one.
Take any sentence in your analysis and ask: could I have written this before the fieldwork?
If yes, it is theory or description. If it needed your numbers, your sites and your river to be written at all, it is analysis.
A strong analysis has a high proportion of sentences that could only have been written by someone who was standing in that river.
Discuss the river, not nine sites in a row
The structural mistake here is organising by site: a paragraph on site 1, a paragraph on site 2, and so on. It feels thorough and it produces nine descriptions and no argument, because the question is about a change along the river and no single site can show one.
Organise by sub-hypothesis instead. Each gets its three moves, using whichever sites are relevant. Then a short section drawing them together to answer the fieldwork question as a whole.
That structure also stops the analysis fragmenting. Your variables are not independent of each other, cross-sectional area and velocity together are discharge; hydraulic radius links channel shape to velocity; bedload size is evidence about the energy environment the other variables describe. The most convincing paragraphs use one variable to explain another, and you cannot do that if each is quarantined in its own section.
Criterion D (Written analysis) asks whether the analysis allows the fieldwork question to be answered. Criterion E (Conclusion) then asks whether the conclusion is supported by that analysis.
So the answer has to be visible here, in the argument, before the conclusion states it. A conclusion that introduces a judgement the analysis never reached is unsupported by definition, and it costs marks in both criteria at once.
Explaining the site that does not fit
This is where the top band is won, and it is worth being methodical rather than hopeful. When a site sits off the trend, work through four possibilities in order and say which one you think it is.
The tape lifted, the rule sank into silt, the float caught an eddy. Real, and it belongs in the evaluation rather than here, but say so, don't leave the reader guessing.
A pool sampled as a riffle, a bridge fixing the width, a bend concentrating flow on one bank. Local hydraulics rather than downstream position.
An engineered, straightened or restored reach, where width, depth and gradient are partly a design decision. The strongest kind of explanation available on this river.
A tributary junction stepping discharge up, or a change in geology or land use altering what the river carries and how much water reaches it.
Then name it specifically. "Site 8 lies in a reach straightened and confined decades ago, where the channel is held between engineered banks" is an explanation. "Human activity may have affected this site" is a shrug with vocabulary.
A river straightened in places, confined in others, under a scheme to widen and reprofile about 4.5 km inside Nyon since 2019, is a river where a simple distance-from-source model should break down in identifiable places.
That is a gift. On an unmodified river an anomaly is usually measurement error and there is little to say. Here, a site that behaves oddly can often be traced to something a reader can go and look at.
Check every anomalous site against what has been done to the channel there before you write it off as error. An explained anomaly moves you from the 5–6 band to the 7–8 band; an unexplained one leaves you listing it, which is the band below.
Quote your data in the prose
The top band asks for an analysis with no or only minor gaps in its supporting evidence. In practice that means numbers in your sentences, not only in your figures.
"Discharge rose steadily" is an assertion. "Discharge rose from 0.001 m³ s⁻¹ at site 1 to 0.234 m³ s⁻¹ at site 8" is evidence. The second version costs a dozen more words and closes a gap the first one leaves open.
Quote sparingly and deliberately: the endpoints, the range, the correlation coefficient and its significance, the value at an anomalous site. Not every number; that is what the tables are for, but enough that the argument stands up when read on its own.
Refer to each figure by its number as you use it, better still, put it in brackets at the end of the claim it supports. That is what integration looks like in practice, and it is how the examiner knows which graph is carrying which claim. A figure never cited here is a figure doing no work.
Watch the significant figures. A tape measure reading to the nearest centimetre does not support a cross-sectional area quoted to four decimal places, and a spreadsheet will happily give you one. Round to what the measurement can actually justify, and be consistent.
If you have put a line of best fit on a scattergraph, add r² beside it. In Excel and Sheets it is one tick box on the trendline options, and it turns a line that merely looks convincing into a measured claim about how strong the relationship is.
Read it as a percentage: r² = 0.68 means 68% of the variation in your y variable is accounted for by x, and the remaining 32% is not. That leftover is the useful half. It is the room in which your anomalies live, and saying “distance from the source accounts for about two thirds of the variation in discharge, so something site-specific accounts for the rest” is exactly the kind of sentence Criterion D (Written analysis) rewards.
Two cautions. r² describes a straight-line fit, so it is not interchangeable with Spearman’s rank, which works on ranks and does not assume a straight line; report whichever you actually ran, and do not present r² as though it carried a significance test with it. And a high r² is still only association. The mechanism paragraph is what turns it into geography.
Everything in this beat applies only if you ran a test. Step 4 makes the case: nothing in the subject guide requires one, and Criterion D (Written analysis) puts any statistical techniques used that are irrelevant or contain errors in its 3–4 band, so a test you cannot justify costs you rather than earns you.
If you have not run one, do not fake the vocabulary. Describe your patterns with real figures, explain them with mechanisms, and spend the words you saved on the explaining, which is where Criterion D (Written analysis) puts most of its marks anyway.
A question we get every year. You do not need to reproduce the arithmetic, the formula, or the rank-by-rank calculation. Nobody is marking your ability to add up, and every line of it costs you words you need for the analysis.
What has to be visible is the decision trail: which test, why that test for that data, the value of the statistic, n or the degrees of freedom, the critical value or p, whether your test was one- or two-tailed, the significance level you used, and the verdict on the null hypothesis. Then what that means for the hypothesis it was testing. Justifying the choice of test is where the credit is, because that is a judgement; the arithmetic is not.
The tables themselves are free. Tables of numerical data do not count towards the word limit, so a rank table or a contingency table can sit in the body at no cost. Put them there rather than in an appendix, which is not read as evidence.
They answer different questions, so the choice is part of the argument. Spearman’s rank tests whether two variables rise or fall together, which is what most of your hypotheses are about: velocity against distance, discharge against distance.
Chi-squared tests whether observed frequencies differ from what you would expect by chance, so it needs counts in categories. Bedload roundness at an upstream site against a downstream site is the right shape for it, and it can turn “the stones look rounder” into a tested claim.
But check it fits before you run it. Chi-squared needs an expected frequency of at least 5 in nearly every cell, and thirty stones spread across six Powers classes will not give you that: the end classes come out near-empty, sometimes with nothing in them at all, and half the cells fall below the threshold. Run it that way and the result is invalid, which puts you in the 3–4 band rather than above it.
The fix is to merge classes into three, angular / sub-rounded / rounded, or into two, before testing. Merging is a judgement you can justify in a sentence, and it is the difference between a test that works and a test that looks like one.
Two more traps. Chi-squared must be run on raw counts, never percentages or means; convert for a graph if you like, but feed the test the tallies. And state your degrees of freedom, which is (rows − 1) × (columns − 1).
Three clauses, in this order: the coefficient, whether it is significant, and what that means for the null hypothesis.
"Spearman's rank on velocity against distance from the source gave rs = 0.20 with n = 9. The critical value at the 95% level for a one-tailed test is 0.60, so the result is not significant and the null hypothesis of no relationship cannot be rejected."
That example is deliberately a non-significant one, because with nine sites it is a result you may well get. Reported like that it costs you nothing: a non-significant result named plainly and then discussed is worth more than a significant one asserted without evidence. What loses marks is quietly not mentioning it.
It can do nothing here. If a concept has not landed, how hydraulic radius affects velocity, what a correlation coefficient means, go and understand it before you open this section, at the levels set out in steps 1 and 4.
It cannot interpret your results. Explaining the pattern in your own data is not a step towards the eight marks; it is the eight marks. There is nothing left to assess once a tool has done it.
Ready for step 7?
0 of 11This is the section to write first and redraft last. If you run short of time anywhere in the IA, run short somewhere else.
Step 7 takes the last five marks: a conclusion that answers the question and an evaluation that is specific enough to be about your investigation and no one else's.
Anything marked “our river” is specific to the fieldwork we do together on the River Asse at Nyon: the sites, the equipment and the way we collect the data. Site coordinates and altitudes are from our own fieldwork records and the exact list can change from year to year, so check them against the sheet you are given on the day. Photographs are the author’s own, taken at the river.