Conducting¶
Syllabus mapping
Working Scientifically outcome — Conducting. Covers safely and accurately following (or adapting) a method, systematically recording data and observations, and identifying anomalous results as they occur. Like every Working Scientifically outcome, it's assessed through application inside a focus area's practicals and depth study, not as standalone content.
Exact NESA outcome code TODO — confirm against the official syllabus PDF (Resources) before treating any wording on this page as verbatim NESA text.
You need to know: how to record raw data properly the first time (so you're not reconstructing it from memory afterwards), and how to recognise a genuine anomaly without quietly deleting inconvenient results.
Core¶
Recording data as you go¶
A raw data table should be built so that someone who wasn't in the room can read it correctly:
- Units go in the column header, once — "Length (cm)", not a unit repeated in every cell.
- Decimal places should match your instrument's precision, consistently — if a ruler reads to the nearest mm (0.1 cm), every length in that column should be recorded to 0.1 cm, not a mix of 12 cm and 12.3 cm.
- Record what the instrument actually showed, not a rounded or "tidied" version of it. If a stopwatch reads 4.37 s, write down 4.37 s — round later, during Processing, not now.
- Record repeats as separate columns/rows, not just a running average in your head — you need the individual trial values later to check consistency and calculate a proper mean.
Worked example — good vs poor raw data table
Timing a pendulum's period at a fixed length, three trials.
Poor:
| Trial | Time |
|---|---|
| 1 | 1.4s |
| 2 | about 1.5 |
| 3 | 1.38 |
Problems: no unit in the header, inconsistent decimal places, "about 1.5" isn't a measurement.
Good:
| Trial | Time for 10 oscillations (s) |
|---|---|
| 1 | 14.22 |
| 2 | 14.35 |
| 3 | 14.19 |
Every value has the same number of decimal places, the header carries the unit once, and timing 10 oscillations rather than 1 (then dividing) is itself a deliberate method choice that reduces the relative effect of reaction-time error — a technique worth reusing whenever you're timing something short and repetitive.
Why repeats matter¶
A single measurement can't tell you whether it was typical or a fluke. Repeating each trial (see Planning) and looking at the spread of results tells you how much random variation exists in your setup — small, tightly-clustered repeats suggest a precise setup; widely scattered repeats suggest something (technique, equipment, environment) is introducing significant random error. How you turn a set of repeats into a single reported value — and how you quantify that spread — is covered properly once Processing is built; for now, the conducting-stage job is simply making sure every individual repeat gets recorded, not just a mental average.
Recognising an anomaly¶
An anomaly (or outlier) is a result that doesn't fit the pattern of the other repeats at the same IV value. Spotting one while you're still at the bench — not three days later while writing up — is what lets you investigate and, if justified, redo that specific trial.
A result is reasonable to flag or exclude only if you can point to a specific, identifiable cause observed at the time — for example, you noticed the trolley was nudged during release, or a light gate clearly misfired and the reading was wildly outside physical possibility. "It didn't match what I expected" is never a valid reason to discard a data point. If you can't identify a specific cause, the honest move is to keep the value, flag it clearly as an outlier in the table, and deal with it openly during analysis rather than quietly removing it.
Worked example — flagging, not deleting
| Trial | Time for 10 oscillations (s) |
|---|---|
| 1 | 14.22 |
| 2 | 14.35 |
| 3 | 21.90 (anomaly — pendulum was visibly bumped mid-swing; excluded from mean) |
| 4 | 14.28 |
The anomalous trial is still visible in the table with the reason stated, and a fourth trial was run to replace it rather than reporting a mean of only two values.
Working safely and honestly¶
Two things that belong together at this stage, even though only one of them is usually framed as "safety":
- Follow the controls from your risk assessment (Planning) as you go — a risk assessment that gets written and then ignored during the actual prac hasn't done its job.
- Report what actually happened. Data doesn't get adjusted, rounded in a hypothesis-friendly direction, or invented for a trial you ran out of time to do. A result that contradicts your hypothesis is not a failed experiment — a hypothesis being wrong is a legitimate, useful scientific outcome, and Analysing is where you're expected to discuss why a result didn't match a prediction, not paper over it here.
Advanced
Noticing error type while you're still collecting data. You don't need the full mathematical treatment (that's in Processing) to start noticing, at the bench, whether a setup seems to have mostly random error — repeats scattered fairly symmetrically around some central value, likely from reaction time or reading precision — or a systematic error — every repeat consistently high or low in the same direction, which usually points to something wrong with the equipment or method itself (an uncalibrated instrument, a consistent parallax error in reading a scale, a light gate positioned slightly off). Spotting a likely systematic error during conducting, rather than after, is valuable because it's often fixable mid-experiment (recalibrate, reposition) in a way it isn't once you've packed up.
Video/visual resources¶
- 🎥 Khan Academy — TODO: source a data collection / recording good practice explainer. Nice-to-have rather than essential — the worked good-vs-poor table example above covers the core skill reasonably well without video support.
- 🎥 Physics High — TODO: unlikely to have generic data-recording content — more realistic to skip this line for this page.
Check yourself¶
-
A student records a table of times as: 3.2 s, 3.15 s, 3s, 3.24 s. Identify what's wrong with this table and rewrite it correctly.
Answer
The decimal places are inconsistent (3s should not appear alongside values given to two decimal places) and there's no indication of what's being timed or its unit in a header. Corrected, assuming the instrument reads to 0.01 s: a column headed "Time (s)" with values 3.20, 3.15, 3.00, 3.24 — each to the same number of decimal places, matching the instrument's precision.
-
A student gets these five repeats for the extension of a spring under a fixed load: 4.1 cm, 4.2 cm, 4.0 cm, 7.8 cm, 4.1 cm. They discard the 7.8 cm reading because "it's obviously wrong." Is this acceptable practice? What should they do differently?
Answer
Not acceptable as described — "obviously wrong" isn't a stated cause. It's acceptable to discard if the student can point to something specific observed at the time (e.g. the mass was seen swinging when the reading was taken, or the ruler was misread against the wrong point). Without an identified cause, the correct approach is to keep the value in the table, flag it clearly as an outlier with the reason left blank/uncertain, and address it explicitly during analysis — or repeat that specific trial and record both the original and the repeat.
-
Explain why timing 20 pendulum oscillations and dividing by 20 gives a more reliable period than timing a single oscillation directly.
Answer
Reaction-time error when starting/stopping a stopwatch is roughly constant in absolute terms (around 0.2–0.3 s), regardless of how long the timed interval is. For a single oscillation, that fixed error is a large fraction of the total time, so it dominates the result. Spread across 20 oscillations, the same fixed reaction-time error is divided by 20 once the total is converted back to a per-oscillation period, making its relative effect on the reported period much smaller.