Every card in CREST investigation toolkit

The independent variable

HintThe one they chose the values of before starting, and the one that runs along the bottom axis of the graph.

WhyHang masses on a spring and you choose the load: 1 N, 2 N, 3 N and so on. Those values were picked in advance, which is what makes this the factor being changed. Say it out loud as a sentence before you start — "I am changing the load and measuring the extension" — and the whole plan falls into place, because everything left over is a factor to hold still. A project that changes two things at once cannot say which of them caused the effect.

The dependent variable

HintIts values are read off an instrument, never chosen, and it goes up the side of the graph.

WhyDrag a block over carpet, then over wood, and the force on the newtonmeter is the reading you take: you did not choose it, the experiment handed it to you. Its name says it depends on what you changed. Deciding which quantity this is also decides the instrument, the units and how precisely you can read it, so it belongs in the plan rather than being sorted out at the bench.

control

HintThese are the factors deliberately held still so the comparison stays fair.

WhyA fair test changes one factor, measures one factor, and holds every other one still. Miss one and the result quietly stops meaning anything: if the beaker under the thicker jacket also started hotter, it would have cooled faster anyway. Listing these before you begin is the most useful ten minutes of the whole project — and a CREST assessor reads the plan as well as the results.

No repeat readings

HintThey rolled it once on each surface. What would tell them whether that single difference was real, and not just luck?

WhyOne run against one run cannot separate a real effect from an accident of how hard the ball was pushed. Three runs on each surface, with a mean taken, turns the same afternoon into evidence. Notice the fix is free — no better apparatus, no cleverer idea, just doing it again. This is the single most common gap in a first project, and it is also the easiest one to close.

An anomalous result

HintFour values huddle together and one stands well apart. There is a name for the stray value.

WhyThe honest thing to do is name it, leave it in the table, and say what you did about it — a thermometer read too soon, a hand in the way, a number written down wrong. It may be excluded from the mean, but only if you say so and say why. Silently deleting it is the one move that turns a good project into an untrustworthy one, and it is exactly what an assessor is trained to look for in your raw table.

To reduce random error

HintOne measurement could be a slip of the eye or a protractor knocked out of line; several averaged together even those slips out.

WhyScatter that pushes readings above and below the true value in no particular pattern is smoothed out by averaging, because the highs and lows cancel. What averaging cannot fix is a fault that pushes EVERY reading the same way — a protractor lined up wrongly, a ruler that starts at 3 mm, a balance not zeroed. That kind of fault is systematic, and taking a hundred readings would only give you a very precise wrong answer.

hypothesis

HintMore than a guess — it must be possible to show it wrong, which is what makes the project worth running.

Why"The steeper the ramp, the greater the acceleration, because more of the weight acts down the slope" does the job: it names the factor changed, the factor measured, the direction of the effect, and a reason. "Something interesting will happen" does not, because no result could ever contradict it. Writing it before the data matters — a prediction invented afterwards to fit the numbers is not one.

resolution

HintThe smallest step the scale can actually show — millimetres beat centimetres for that.

WhyMatch the instrument to the size of the change you expect. A bar that expands by 2 mm cannot be studied with a scale whose smallest division is a centimetre — every reading would come out the same and the effect would look like nothing. Choosing the instrument is part of planning, not something to sort out at the bench, and explaining that choice is worth credit in itself.

repeatable; reproducible; accurate

HintSame person again, then somebody else entirely, then closeness to the truth — three separate tests of how much a number can be trusted.

WhyThree different questions, and a set of readings can pass one and fail another. Measure a resistance five times with a badly zeroed meter and the five will agree beautifully with each other and with your own repeat runs, while every one of them is wrong — tightly clustered around the wrong place. Handing your method to somebody else and seeing whether they land where you did is the strongest check a school project can make.

best fit

HintDo not join the dots — draw one smooth line with roughly as many crosses above it as below.

WhyIt may be straight or a curve, and it need not touch a single cross. Its job is to show the pattern the readings point at, which is why joining the dots is wrong: that draws the scatter, not the trend.

Too few points

HintA line ruled between just two crosses could be hiding a curve. What is missing in between?

WhyTwo crosses always sit on a straight line, so a straight line drawn through them proves nothing at all. Springs are a good example: they stretch evenly up to a certain load and then stop doing so, and testing only the two ends would miss the bend completely. Five or six loads spread across the range, and the shape shows itself.

A risk assessment

HintWritten before anything is switched on: what could hurt someone here, and how will I make that less likely?

WhyThree columns are enough for a school project — the hazard, who could be harmed, and the step that reduces it. Hot water scalds, so use a clamp and small volumes; glass breaks, so keep the beaker back from the edge. It is not paperwork tacked on afterwards: CREST assesses whether a student made decisions that took safety and ethics into account, so the thinking has to be visible in the plan.