Plan investigations to ensure that they are valid and reliable, including the use of an appropriate experimental design with consideration of independent, dependent and controlled variables

What this dot point is asking

NESA wants you to identify the independent, dependent and controlled variables for any given investigation, explain the role of a control group, and design an experiment that produces valid first-hand data. This is fundamental and appears in nearly every Investigating Science paper.

The answer

A scientific investigation tests a hypothesis by changing one thing, measuring the result, and holding everything else constant. Three types of variables structure this design.

Independent variable

The variable that the researcher deliberately changes to test its effect. There is normally one independent variable per experiment, set at multiple levels (e.g. 0, 10, 20, 30 degrees Celsius).

Dependent variable

The variable that the researcher measures to see how it responds. It depends on the independent variable.

Example. In a fertiliser experiment, the independent variable is fertiliser concentration (mg/L) and the dependent variable is tomato yield (kg per plant).

Controlled variables

The variables held constant across all treatment groups to prevent them from confounding the result. In the fertiliser experiment, controlled variables include soil type, watering schedule, sunlight, plant variety and starting plant size.

Control group

A group treated identically to the experimental groups except that it receives no treatment (or a placebo). The control group establishes the baseline against which treatment effects are compared.

Example. A control group of tomato plants given water but no fertiliser. The difference in yield between treated and control plants is attributed to the fertiliser, provided controlled variables are properly held constant.

Replication and sample size

Repeating the experiment, or running multiple individuals per condition, accounts for biological variation and measurement error. NESA expects students to identify a minimum number of replicates (often 5 to 10 per condition) and to repeat the experiment at least three times.

Randomisation and blinding

In human or animal studies, randomisation assigns subjects to treatment groups by chance to reduce selection bias. Blinding prevents the researcher or the subject from knowing the treatment assignment. Double-blind designs are the gold standard in clinical trials.

Common designs

Comparative design

Two or more groups treated differently and compared. Most school experiments fit here.

Time-series design

One subject measured repeatedly over time as the independent variable changes.

Factorial design

Multiple independent variables changed simultaneously to detect interactions (e.g. temperature and pH on enzyme activity).

Examples in context

Example 1. Pittwater oyster growth at the Sydney Institute of Marine Science. Researchers investigating whether ocean acidification affects Sydney rock oyster (Saccostrea glomerata) growth manipulate the independent variable, dissolved CO2 (set to 400, 600, 800 and 1000 ppm in seawater tanks). The dependent variable is shell mass gained per oyster over 12 weeks. Controlled variables include temperature (22 degrees C), salinity (35 ppt), feeding rate (algae per day), light cycle (12:12) and oyster size at the start. A control group is held at present-day atmospheric CO2 (400 ppm). Sample size is 30 oysters per tank, with three replicate tanks per CO2 level to account for tank-effects, allowing statistical separation of the treatment effect from biological variation.

Example 2. Murray-Darling environmental flow trials. The Murray-Darling Basin Authority releases controlled water pulses into the Macquarie Marshes to test whether environmental flows improve waterbird breeding success. The independent variable is flow volume in gigalitres per release; the dependent variable is the number of nesting pairs of straw-necked ibis. Controlled variables include season of release, baseline water level, surrounding land use and observer protocol. Because true randomisation is not possible (sites are fixed in geography), researchers use a before-after-control-impact (BACI) design, comparing the same site across years with and without flow events. This compromise is common in field ecology where laboratory-style controls cannot be imposed.

Try this

Q1. A student investigates whether "music helps memory." Identify the independent and dependent variables they should specify, and list three controlled variables required for a valid investigation. [4 marks]

• Cue. IV: specified genre and decibel level. DV: number of words recalled from a 20-item list after 5 minutes. Controlled: list order, ambient noise, participant age, time of day.

Q2. A trial compared two batches of CSIRO drought-tolerant wheat at sites in Wagga Wagga and Dubbo. Yields were higher in Dubbo. Explain why this difference does not establish that the wheat variety performs better in Dubbo, and design a more valid comparison. [2+3 marks]

• Cue. Confounders: soil, rainfall, fertiliser history all differ by site. Valid design grows both varieties side-by-side at each site with randomised plot allocation and replication.

Q3. A pharmaceutical company tests a new analgesic on 200 volunteers. (a) Explain the difference between single-blind and double-blind designs. (b) Explain why double-blind is preferred. (c) Identify one situation in which double-blinding is not possible. [2+2+2 marks]

• Cue. (a) Single: participants unaware of group; double: researchers also unaware. (b) Removes observer expectancy and participant placebo. (c) Surgical interventions, or where treatment has an obvious side-effect.