TL;DR Paper 3 of O-Level Physics (6091) is a 40-mark practical exam split across four skill areas: Planning (P), Manipulation, Measurement and Observation (MMO), Presentation of Data and Observations (PDO), and Analysis, Conclusions and Evaluation (ACE). Most marks are lost not because of wrong physics, but because of presentation mistakes - missing units in table headers, dot-to-dot lines instead of best-fit curves, the phrase "human error", and gradient triangles that are too small. This guide explains exactly what examiners reward and penalise in each section, so you can stop losing marks on things you already know how to do.
Paper 3 structure and mark allocation
Paper 3 is a practical examination lasting around 2 hours. The 40 marks are distributed across four assessed skill areas defined by the SEAB 6091 syllabus:
MMO carries the most weight (40 % of the paper), but PDO and ACE together account for another 20 marks. Students who focus only on getting the experiment to work but neglect how they present and evaluate the data leave a large number of marks on the table.
MMO - what earns marks
MMO is assessed while you carry out the experiment. The examiner checks whether you handle apparatus competently, measure correctly, and record observations in appropriate detail.
Setting up apparatus correctly and safely
Follow the diagram in the question paper precisely. If the setup involves a circuit, check polarity and ensure no short circuits before switching on. If it involves optics or mechanics, secure the apparatus so it does not shift between readings. Examiners note whether you needed help to set up - if an invigilator has to intervene, MMO marks are deducted.
Reading instruments to the correct precision
Every measuring instrument has a resolution, and you are expected to read to the correct level:
Ruler - to the nearest mm (0.1 cm)
Protractor - to the nearest degree
Stopwatch - to 0.01 s (record the full display)
Thermometer - to 0.5 °C (read between the markings)
Ammeter / voltmeter - to half the smallest scale division
Recording a ruler measurement as "12 cm" when it should be "12.0 cm" costs marks.
Taking repeat readings without being told
If a measurement involves timing - such as oscillation period - take at least two readings and average them. If the question does not explicitly instruct you to repeat, do it anyway. Writing down a single timing measurement signals to the examiner that you do not understand the need for reliability.
Recording raw data with correct units and significant figures
Write down exactly what the instrument shows, not a rounded version. If the stopwatch reads 14.38 s, record 14.38 s, not 14.4 s. Units must accompany every reading in your raw data (or be stated in the table header - see PDO below). Maintain consistent significant figures across a set of readings.
PDO - table and graph conventions that score
PDO marks are about presentation discipline. You can get every reading correct and still lose PDO marks through formatting errors.
Table conventions
Column headers must include quantity AND unit. Write headers as "Length / cm" or "Time / s", not just "Length" or "Time". The forward-slash format (quantity / unit) is the standard convention examiners expect.
Units go in the header, not next to every reading. Once the header says "Length / cm", the data cells should contain only numbers: 12.0, 15.3, 18.6. Do not write "12.0 cm" in every row.
All readings must have consistent decimal places. If your first length reading is 12.0 cm (one decimal place), every length reading must also be to one decimal place. Mixing 12.0, 15.35, and 18 in the same column costs marks.
Processed quantities belong in a separate column. If you calculate 1/L from your raw length L, add a new column headed "1/L / cm⁻¹" with the calculated values. Do not overwrite raw data.
Graph conventions
Graphs are a high-value target in PDO. A well-drawn graph can earn up to 5 marks; a poorly drawn one can lose most of them.
Axes must be labelled with quantity and unit. The x-axis and y-axis labels follow the same convention as table headers: "Temperature / °C", "Resistance / Ω".
Scale must use more than half the grid in both directions. If your data ranges from 0 to 50 on one axis but you draw the axis from 0 to 200, you are compressing the data into a quarter of the grid. The examiner checks that plotted points span at least half the available grid area on both axes. Choose scales that are easy to read - multiples of 1, 2, 5, or 10 per large square.
Points must be plotted to half-a-small-square accuracy. Use a sharp pencil. Each point should be marked with a small cross (not a dot or a blob). If the examiner cannot verify the position of a point to within half a small square, the plotting mark is lost.
Draw a best-fit line (straight or curve) - not dot-to-dot. A best-fit line is a single smooth line that follows the overall trend of the data, with roughly equal numbers of points above and below. Connecting the points sequentially with straight segments is not a best-fit line and earns zero marks.
Anomalous points should be circled and excluded from the best-fit. If one point clearly deviates from the trend, circle it and draw the best-fit through the remaining points. Do not force the line through the anomaly. For a longer treatment of when to include vs exclude an outlier and how to justify it in ACE, see our anomalous results guide.
Gradient calculation must use a large triangle - at least half the line length. When determining the gradient, draw a right-angled triangle on the graph using points on the best-fit line (not plotted data points). The horizontal and vertical sides of this triangle should span at least half the length of the line. Show the coordinates of the two points and the arithmetic clearly. A tiny gradient triangle loses the mark.
Graph plotting quick-rules checklist
This checklist answers the questions students most often ask on forums - particularly "must the best-fit line pass through the origin?"
Scale selection. Choose a scale so that plotted points fill more than half the grid on both axes. Use easy intervals: 1, 2, 5, or 10 per large square. Never use 3 or 7 per large square - they make plotting and reading off values error-prone.
Axis labels. Write "Quantity / unit" on every axis - for example, "Time / s" or "Resistance / Ω". No label or a label without a unit costs a mark.
Plotting. Mark each point with a small, neat cross (×). Do not use dots, circles, or large blobs. The examiner must be able to verify each point to within half a small square.
Best-fit line. Draw one single smooth line (straight or curved) that follows the general trend. The line does not need to pass through the origin unless the physics of the experiment requires it (for example, a proportional relationship with a theoretical zero intercept). If the data do not start at (0, 0), do not force the line there.
Best-fit line vs dot-to-dot. Never connect points with short straight segments. This is the most common graph mistake and earns zero marks for the line.
Anomalous points. If one point clearly falls off the trend, circle it and ignore it when drawing the best-fit. Do not let one outlier distort your entire line.
Gradient triangle. Use two points on the best-fit line (not your plotted data points). The triangle should span at least half the length of the line. Write out both coordinates and the full calculation: gradient = Δy / Δx.
Rate from a graph. If asked for rate at a specific point, draw a tangent to the curve at that point and calculate its gradient. If asked for average rate over a range, use total change divided by time interval. These are different - check what the question says.
ACE is where many students lose marks unnecessarily, because the answers require specific language and logical structure rather than vague statements.
Conclusions must reference the data
A conclusion is not just "the hypothesis is correct." You must quote specific values from your results - the gradient, an intercept, or key calculated values - and explain what they tell you about the relationship. For example: "The gradient of the graph of T² against L is 4.02 s² m⁻¹, which is consistent with a linear relationship between T² and L."
Sources of error must be specific
This is the single most common ACE mistake. The phrase "human error" earns zero marks every time. Examiners require you to identify a specific, physical source of inaccuracy. For examiner-credit phrasing keyed to each apparatus you might meet in Paper 3, see our O-Level Physics sources of error bank.
Replace vague statements with specific ones:
Instead of "human error" → "Parallax error when reading the meniscus of the mercury thermometer"
Instead of "timing error" → "Reaction time when starting and stopping the stopwatch manually"
Instead of "measurement error" → "The ruler was not aligned with the zero mark at the start of the object"
Instead of "the experiment was not accurate" → "Heat loss from the surface of the beaker to the surrounding air reduced the final temperature reading"
Improvements must directly address the identified error
Each improvement you suggest must correspond to the specific error you named. If you identified heat loss, the improvement should be insulation (e.g., "Wrap the beaker in cotton wool to reduce heat loss to the surroundings"). If you identified reaction time, the improvement should be a method that avoids manual timing (e.g., "Use a light gate and datalogger to measure the time interval electronically").
Examiner-approved phrases that score
The following phrases are consistently awarded marks in Paper 3 mark schemes. Memorise them and use the appropriate ones:
"Parallax error when reading the scale" - for any measurement involving reading a scale at an angle
"Repeat the experiment and calculate the mean to improve reliability" - the standard reliability statement
"Use a light gate and datalogger to reduce the effect of reaction time" - for any timing measurement
"Allow the system to reach thermal equilibrium before taking the temperature reading" - for heat experiments
"Use a set square to ensure the ruler is perpendicular to the surface" - for length measurements from a baseline
"The oscillation was not in a single plane, causing the length of the pendulum to vary" - for pendulum experiments
"Switch off the current between readings to prevent heating of the resistor" - for resistance experiments
"Use a fiducial marker at the centre of the oscillation to reduce timing uncertainty" - for period measurements
Planning - the 4-mark structure
The Planning question asks you to design an experiment. The 4 marks follow a predictable structure:
1 mark - Identify independent and dependent variables. State clearly what you will change (independent variable) and what you will measure (dependent variable). For example: "The independent variable is the length of the wire. The dependent variable is the resistance of the wire."
1 mark - Describe how to control key variables. Name at least one variable that must be kept constant and explain how you will keep it constant. For example: "The diameter of the wire is kept constant by using the same reel of wire throughout."
1 mark - Describe the method with specific apparatus and quantities. Name the instruments you will use and give realistic values. Do not write "measure the length" - write "measure the length of the wire using a metre rule, starting from 10.0 cm and increasing in 10.0 cm intervals to 60.0 cm." The mark requires both named apparatus and a plausible range of values.
1 mark - State how to make results reliable. The expected answer is: "Repeat the measurement of [dependent variable] at each value of [independent variable] at least twice and calculate the mean." You may also mention controlling environmental conditions or specifying the number of data points (at least six sets of readings).
Common reasons marks are lost
These are the ten most frequent mark-losing mistakes in Paper 3, based on recurring examiner feedback:
Writing "human error" as a source of error. This is too vague and never earns the mark. Always name a specific physical cause.
Graph scale not using enough of the grid. If the plotted points cluster in one small corner, you have chosen a poor scale. The data should span more than half the grid on both axes.
Missing units in table headers. Every column header must include the unit, formatted as "Quantity / unit". No unit in the header means a lost mark.
Not taking repeat readings. If the experiment involves timing or any measurement with variability, you must repeat and average - even when the question does not say so.
Drawing dot-to-dot instead of a best-fit line. Connecting plotted points with straight segments is the most common graph error. Draw a single smooth line through the trend.
Not showing gradient calculation working. You must draw the triangle on the graph, label the coordinates of the two points used, and write out the calculation. Just stating a gradient value without working earns nothing.
Inconsistent decimal places in data columns. All values in a column must have the same number of decimal places, reflecting the precision of the instrument.
Writing "accurate" when meaning "precise" (or vice versa). Accuracy refers to how close a measurement is to the true value. Precision refers to the spread of repeated measurements. Using the wrong term in an evaluation costs the mark.
Not specifying which instrument for which measurement. In Planning, writing "measure the temperature" without naming a thermometer loses the apparatus mark. Always pair the measurement with the instrument.
Forgetting to state controlled variables by name. Saying "keep everything else the same" is not specific enough. Name the variable (e.g., "the cross-sectional area of the wire") and explain how it is controlled.
Putting it together - a practical checklist
Before you hand in your Paper 3 script, run through this mental checklist:
Every table column has a header with quantity and unit
All data in each column has consistent decimal places
Repeat readings are recorded and averaged
Graph axes are labelled with quantity and unit
Plotted points span more than half the grid on both axes
Best-fit line is drawn (not dot-to-dot)
Gradient triangle is large and the calculation is shown
Sources of error are specific (no "human error")
Each improvement matches its corresponding error
Planning answer names variables, apparatus, method, and reliability
