These planning notes support learners preparing for Cambridge International AS and A Level Marine Science 9693. They develop aims, variables, sampling and reproducible methods; consult the relevant examination-year syllabus and task for exact paper coverage and planning requirements. A complete plan turns a question into a fair, safe, ethical and analysable test rather than a list of apparatus.
Scope of these planning notes
Practise applying planning skills in unfamiliar laboratory and field contexts. Match the biological knowledge, techniques and analysis to the investigation rather than assuming that one prepared method fits every question.
A useful planning framework covers problem definition, estimates, predictions, hypotheses, techniques, controls, replication, data presentation, analysis, conclusions and evaluation. Select statistical methods only when the question and data justify them.
This note focuses on planning decisions. Companion practical notes develop apparatus and risk, raw records, graphs, uncertainty, statistics, evaluation and examination strategy in greater depth.
From a question to a testable aim
A research question identifies a possible relationship, such as whether salinity affects mussel filtration rate. The aim states what will be changed and what will be measured:
To investigate the effect of salinity on the filtration rate of mussels.
This phrasing exposes the variables immediately. Avoid vague aims such as "to investigate mussels" because they do not define a measurable problem.
In observational fieldwork, the aim may examine association rather than direct cause, such as whether limpet abundance changes with distance up a rocky shore.
Independent and dependent variables
In a manipulation experiment, the independent variable is the factor deliberately changed. In observational work, a naturally varying predictor or position may be called the independent variable, but it has not been assigned experimentally. The dependent variable is the measured response; the terminology alone does not establish causation.
For an experiment on salinity and filtration rate:
independent variable: salinity of the water
dependent variable: a defined filtration or clearance measure. Particle-concentration change per unit time can be a proxy, but is not automatically a volume of water cleared per unit time; specify the calculation, water volume and any normalisation for organism size, and account for background particle loss
For a shore transect:
Check this topic from memory
Attempt the matching topic bank before reopening the notes. Use each missed idea to decide what to review next.
independent variable: distance from a fixed shore reference or height above chart datum
dependent variable: abundance or percentage cover of the named organism
Operational definitions matter. "Growth" could mean change in mass, length, area or cell number. Name the measurement and time interval.
Prediction and scientific hypothesis
A prediction states the expected pattern. A scientific hypothesis proposes a testable relationship and can include a mechanism.
For example:
Within a specified low-light range, increasing light is predicted to increase the photosynthetic response because more light energy is available. A plateau may occur at higher light, but identifying the constraint needs further evidence; measured oxygen change is a net response influenced by respiration and gas exchange.
A predicted graph can express direction, curve shape, plateau, optimum or threshold. It should match the variables and not add unsupported precision.
A hypothesis can be supported or not supported by evidence. A single investigation does not prove it permanently.
Null hypothesis
When a statistical test is planned, state an appropriate null hypothesis. For correlation:
The population correlation coefficient for distance up the shore and limpet abundance is zero, for the specified correlation measure and target population.
For a comparison:
The population mean shell-mass loss is equal under the two specified solution conditions.
A null hypothesis states a population relationship or parameter, not whether the observed sample will be statistically significant. Specify the test, alternative and significance criterion in advance. Zero correlation for one measure is not a claim that every possible relationship is absent. The conclusion must refer back to the same relationship or difference and use the statistical outcome. Do not write "there is no difference" when the test only says there is insufficient evidence of a statistically significant difference.
Variables to standardise
A standardised variable is kept sufficiently constant so it does not provide a competing explanation for change in the dependent variable.
Choose variables with a plausible material effect. In a shell dissolution investigation, these include shell species, exposed surface area, solution volume, duration, temperature and agitation. Variation among nominally identical test tubes is unlikely to need a separate control statement.
State how each important variable will be standardised. "Control temperature" is incomplete; "place all tubes in the same thermostatically controlled water bath at the stated temperature" is operational.
Some variables cannot be held exactly constant in fieldwork. Measure or restrict them, randomise placement, block the design or acknowledge their effect.
Experimental controls
A control treatment provides a comparison that omits the tested factor or represents a reference condition. It helps identify background or procedural effects. A control treatment alone does not establish causation; treatment allocation, replication and competing explanations also matter.
Examples include:
a known reference pH when testing acidification
a chamber without the organism when measuring oxygen change
an unshaded treatment when testing shade
an identical setup without added nutrient when testing enrichment
A controlled variable and a control treatment are different. One is held constant; the other is a comparison group.
Negative and positive controls can test absence of effect or confirm that the measurement system can detect a known effect, where appropriate.
Choosing the independent-variable range
Choose enough levels over a justified range to address the question, while retaining adequate independent replication. A multi-level trend investigation may use five or more values, whereas a two-condition comparison can require two treatment groups. These are design examples, not a verified universal examination minimum. The range must be safe and feasible; follow any explicit task requirement.
Pilot work or supplied information can prevent values that all produce no response or immediate mortality. Include the biologically relevant range and, when useful, a reference condition.
Intervals can be equal when a general trend is expected. Smaller intervals near a suspected threshold or optimum may resolve the pattern better, provided the overall range remains adequate.
List actual values rather than saying "use several concentrations".
Changing the independent variable accurately
Describe how values will be prepared or located. Salinity may be set by measured dilution or salt addition, temperature by water baths, light by a measured light quantity at the sample, using distance, filters or source output to adjust it, and shore position by a tape placed along a defined transect.
Verify the achieved value where possible. A labelled treatment is not evidence that pH, salinity or temperature remained at its target.
Avoid changing a second factor inadvertently. Moving a lamp changes light but may also change temperature, so use a heat shield, LED source or water bath and monitor temperature.
Measuring the dependent variable
Choose a method that is valid, sufficiently sensitive and repeatable. State the instrument, unit, measurement interval and calculation.
If the response is rate, collect a change over a known time and calculate change divided by time. If the response is abundance, define the sampling unit and counting rule. If it is colour, use a quantitative scale or sensor rather than an unexplained judgement.
The measurement resolution should detect differences expected across the chosen range.
Replicates and sample size
Independent replication repeats a treatment on independently allocated experimental units. It reveals variation among those units and supports uncertainty estimates. Define the unit that receives the treatment before counting the sample size; an unusual observation should be investigated, not automatically deleted.
Technical repeats of the same prepared sample mainly test measurement repeatability. Different organisms can reveal biological variation, but individuals sharing one treatment tank are not automatically independent treatment replicates. For example, if salinity is assigned to four separate tanks containing three mussels each, there are four treatment units and twelve organism observations, not twelve independent tanks. State the sampling hierarchy.
More genuinely independent units can improve precision and power, but do not remove systematic bias, confounding or dependence. They also cost time and may use more organisms. Choose a feasible number and keep it the same across treatments unless a justified design requires otherwise.
Repeating one reading without resetting the experimental unit is not always independent replication.
Randomisation and order effects
Random assignment of organisms or samples reduces systematic allocation bias. Randomising treatment order reduces effects of time, warming, fatigue or instrument drift.
If conditions change predictably, use blocks. For example, test one replicate of every treatment in each time block rather than completing all replicates of one treatment first.
On a shore, random coordinates within defined zones can prevent the observer from choosing unusually dense patches.
Planning ecological sampling
Field investigations must define the target population, sampling frame, sampling unit and placement method. Use a defined sampling frame and appropriate random or stratified sampling to estimate the target mean; random placement does not guarantee that one realised sample is representative. Use systematic sampling along a transect to investigate a gradient.
Quadrat size and number affect precision and feasibility. Small quadrats may miss patchy or rare organisms; very large quadrats reduce the number possible. Pilot sampling can guide a suitable design.
Define abundance as count, density, frequency or percentage cover and keep the rule consistent.
Avoid pseudoreplication. Many quadrats along one transect do not automatically represent multiple independent shores.
Ethics and organism welfare
Use the minimum number of organisms consistent with useful evidence. Reduce handling, exposure time and stress; maintain appropriate oxygen, temperature and salinity; and return organisms safely where lawful and ecologically appropriate.
Do not release non-native, captive or exposed organisms without a justified protocol. Empty shells or non-living models may replace live organisms for some questions.
Build ethical treatment into the method and follow applicable permissions, welfare rules and task requirements; do not leave it as a generic closing sentence.
Risk within the plan
Identify the hazard, who or what may be harmed, the route of harm and a specific precaution. Acids, glass, hot water, electricity near water, slippery shores, tides and living organisms require different controls.
Risk detail belongs primarily to Practical Skills 2, but every complete plan must show that its method is safe enough to perform.
Planning the results table
Construct the table before data collection. Include the independent variable first, raw replicate columns, calculated values and deductions if required. Put units in headings, not repeatedly in cells.
Record raw readings at the precision supported by the instrument and method, retaining the actual readings. Resolution helps set reporting precision but is not the same as accuracy or total uncertainty; derived values also depend on propagation and calculation. A plan that says "record results" without defining table structure and precision is incomplete.
The raw-data plan must preserve enough information to check subsequent calculations.
Planning analysis and conclusion
State how the data will answer the aim. This may include calculating a mean and spread, plotting paired measurements, reporting confidence intervals, or applying a justified correlation or difference test. Overlap of separate confidence intervals is not a substitute for a test or interval for the difference.
Choose a graph for the question and data: a scatter plot for paired quantitative measurements, a line when an ordered sequence or justified connection is meaningful, an appropriate categorical comparison plot for groups, and a histogram for a numerical frequency distribution. A continuous predictor alone does not require joining every point.
Statistical choice follows the question and data, not personal preference. The analysis must lead back to the prediction or null hypothesis.
Writing a reproducible method
A plan must give enough detail for another trained student to perform the same investigation. A reliable order is:
identify and prepare experimental units
set and verify independent-variable values
standardise important variables and apply control treatments
measure the dependent variable with stated apparatus, unit and timing
reset or prevent carry-over between measurements
repeat independently at every value
record raw data in the planned table
analyse data and decide how the evidence tests the hypothesis
Quantities, times, distances and criteria should be explicit. Do not hide essential steps behind "repeat as usual".
Worked application: planning the shell pH investigation
Investigate pH effects on empty mollusc-shell mass loss. For this illustrative trend design, prepare five or more verified pH values spanning a justified supplied range, with equal solution volumes in labelled vessels. Record solution composition and how pH is maintained: equal pH does not guarantee equal carbonate saturation, and acid addition is not chemically identical to carbon-dioxide addition. Cut or select shell pieces from the same species with similar exposed area, rinse, dry to a consistent endpoint and record initial mass. Randomly assign pieces to independently prepared treatment vessels. Define the vessel as the treatment unit when pieces share its solution, and use independently replicated vessels per pH; multiple pieces in one vessel are subsamples. Record donor shells or organisms so material from one source is not mistaken for independent biological replication. Incubate for the same time in one temperature-controlled water bath with equal agitation. Recheck pH, then rinse, dry and reweigh. Calculate mass and percentage mass loss, plot mean loss against pH and display spread. Standardise shell source, area, volume, duration and temperature. The inference concerns empty-shell dissolution under these solution conditions, not live-organism growth or physiological regulation. Empty-shell use reduces welfare concerns; precautions must match the actual solution, preparation method and other hazards, rather than assuming goggles alone make the method safe.
Common misconceptions and corrections
Writing an aim with no measurable response. Name both variables operationally.
Calling the measured response the independent variable. It is the dependent variable.
Assuming a field gradient is deliberately manipulated. It may be an observational independent variable.
Leaving a prediction too vague to test. Specify a justified expected pattern; do not invent a direction when evidence does not support one.
Calling a supported hypothesis proven. Evidence supports it within limits.
Writing a null hypothesis with no named variables. State the exact relationship or difference.
Standardising every trivial feature. Prioritise variables likely to affect the response.
Saying only "control temperature". Explain how it will be held constant and checked.
Confusing a control treatment with a controlled variable. They have different roles.
Treating one number of treatment levels as universal. Justify levels and independent replication for the question, and follow any explicit task requirement.
Choosing values outside safe biological limits. Use pilot or supplied evidence.
Using an overly narrow range. It may hide the pattern.
Changing lamp distance without controlling heat. A second variable changes.
Trusting a treatment label without verification. Measure achieved pH, salinity or temperature.
Measuring rate without a time interval. Rate requires change per time.
Calling repeated readings of one sample biological replicates. They do not capture between-organism variation.
Giving every treatment a different replicate count without reason. Comparability is weakened.
Testing all controls first and treatments later. Time drift may bias groups.
Choosing attractive field patches. Use random or systematic placement.
Treating many quadrats on one shore as many independent shores. That is pseudoreplication.
Changing abundance definition during sampling. Use one stated counting rule.
Ignoring organism welfare because the species is common. Ethical treatment still applies.
Listing "wear goggles" without a hazard. Link precaution to a specific risk.
Planning analysis only after seeing results. Predefine the evidence pathway.
Choosing a line graph solely because salinity is continuous. Start with paired quantitative data and justify any connecting line or fitted relationship.
Selecting a statistical test because it is familiar. Match it to the question and data.
Omitting raw replicates from the table. Means alone hide variation.
Writing "repeat" without saying what or how many times. Define independent replication.
Omitting quantities and timing. The method is not reproducible.
Concluding about causation from uncontrolled observational data. Alternative variables remain.
Assessment guidance
Start a planning answer by extracting the independent and dependent variables from the prompt. State a justified prediction and, when a statistical test is planned, a precise null hypothesis and appropriate alternative. Give justified treatment levels, independent experimental units, a valid measurement method, operational standardisation and an informative control. Include randomisation or a defensible field-sampling rule, ethical treatment and hazard-specific precautions. Predefine a table, graph or statistical analysis and explain how it answers the aim. For an extended planning response, connect the method into a reproducible sequence and match the required detail to the prompt. Useful detail is selective: each decision should prevent bias, reveal variation or make the conclusion possible.
Retrieval practice
Take one laboratory and one shore-field question. For each, state the aim, variables, prediction, possible null hypothesis, justified treatment range, replicate type, control, standardisation, randomisation, ethical limit, raw table and analysis. Then identify one confounder and redesign the method to isolate the proposed relationship without changing its intended dependent variable.
