Experimental Research
Variables · Hypotheses · Control · Randomisation · Experimental Designs · Validity · MCQs
What Is Experimental Research?
Experimental research is a systematic approach in which a researcher deliberately manipulates an independent variable or introduces an intervention to examine its effect on an outcome, while controlling or accounting for other relevant factors.
Educational example
A researcher wants to investigate whether retrieval practice improves students’ science achievement. One group uses regular retrieval quizzes, while a comparison group follows the usual revision approach. The researcher measures learning outcomes and compares the results.
Manipulation
The researcher introduces or varies a teaching intervention.
Measurement
The researcher records a defined learning outcome.
Comparison
Outcomes are examined across conditions or over time.
Inference
Evidence is assessed against alternative explanations.
Independent, Dependent and Extraneous Variables
Independent variable (IV)
The factor deliberately manipulated or varied by the researcher in an experiment.
Example: Teaching method: retrieval practice versus usual revision.
Dependent variable (DV)
The outcome measured to assess the effect of the independent variable.
Example: Students’ scores on a science assessment.
Extraneous variable
A factor other than the IV that could influence the dependent variable.
Example: Differences in prior knowledge or access to study resources.
Confounding variable
A factor that varies with the condition being studied and provides an alternative explanation for the observed outcome.
Hypothesis and Operational Definition
Research hypothesis
A hypothesis is a testable proposition about a relationship, difference or expected outcome.
Null hypothesis (H₀)
Often states that there is no difference or relationship of the specified kind in the population or model being tested.
Alternative hypothesis (H₁ or Hₐ)
States the difference or relationship the researcher is testing for. It may be directional or non-directional.
Operational definition
Specifies how a concept or variable will be measured or implemented.
Control Group and Experimental Group
Experimental group
Receives the intervention or condition being evaluated.
Control / comparison group
Provides a benchmark, such as usual instruction, no additional intervention or another treatment.
Why use a comparison group?
- To compare outcomes under different conditions.
- To help separate intervention effects from changes that would occur anyway.
- To reduce reliance on a simple before-and-after comparison.
- To improve the credibility of causal conclusions when the design is appropriate.
Random Sampling vs Random Assignment
Random sampling
Concerns how participants are selected from a target population.
Main relevance: It can improve the representativeness of the sample and support generalisation when properly implemented.
Random assignment
Concerns how participants already in the study are allocated to experimental conditions.
Main relevance: It helps balance known and unknown confounding factors on average, strengthening causal inference.
Example
A university randomly selects 200 students from its eligible population, then randomly assigns those students to two teaching conditions. The first random procedure is sampling; the second is assignment.
A study can use random assignment without random sampling. Therefore, strong causal inference and broad generalisability are not identical.
Matching, Constancy and Counterbalancing
Matching
Participants or groups are matched on relevant characteristics, such as prior achievement, before comparisons are made.
Holding conditions constant
Keeping relevant conditions similar across groups, such as test duration, instructions or classroom resources.
Counterbalancing
Varying the order of conditions or tasks across participants to reduce order effects in repeated-measures designs.
Blinding
Concealing group allocation from participants, assessors or other relevant parties when feasible to reduce expectation-related bias.
Three Broad Experimental Design Categories
1. Pre-experimental
Usually includes an intervention but has limited control over alternative explanations. Examples include one-shot case studies and one-group pre-test/post-test designs.
2. True experimental
Uses random assignment to conditions, together with an intervention and a suitable comparison design.
3. Quasi-experimental
Evaluates an intervention without random assignment to conditions, using comparison strategies to address selection differences and other threats.
One-Shot and One-Group Designs
One-shot case study
X = intervention; O = observation or measurement after the intervention.
Example: A class receives a new teaching activity and is tested afterward. There is no pre-test and no comparison group.
One-group pre-test/post-test
The same group is measured before and after the intervention.
Example: Students take a test, complete a learning programme, then take another test.
Static-group comparison
One group receives the intervention and another group is measured without random assignment, often without a pre-test.
Pre-Test/Post-Test Control-Group Design
Participants are randomly allocated to an experimental group or a control group. Both groups are measured before and after the intervention.
Control: R O₁ — O₂
R: Random assignment. O: Observation or measurement. X: Intervention.
Example
Students are randomly assigned to retrieval practice or usual revision. Both groups complete the same pre-test and post-test under comparable conditions.
Why use a pre-test?
- It provides information about initial achievement.
- It helps describe baseline differences and change.
- It can support more precise analysis.
Post-Test-Only and Solomon Four-Group Designs
Post-test-only control-group design
Control: R — O
Randomly assigned groups receive different conditions and are measured after the intervention. A pre-test is not required for the basic design.
Solomon four-group design
Uses four randomly assigned groups: two receive a pre-test and two do not. Within each pre-test condition, one group receives the intervention and the other serves as a control.
Within-Subjects and Between-Subjects Designs
Between-subjects
Different participants receive different conditions.
Example: Group A uses one teaching method; Group B uses another.
Within-subjects / repeated measures
The same participants experience multiple conditions or are measured repeatedly.
Example: Students complete assessments under two counterbalanced learning conditions.
Advantages and limitations
- Between-subjects designs avoid some carryover effects but may require more participants.
- Within-subjects designs can control stable individual differences.
- Within-subjects designs may be affected by order, practice, fatigue or carryover effects.
- Counterbalancing can reduce some order effects when suitable.
Non-Equivalent Control-Group Design
In educational settings, random assignment may be impractical because schools already have intact classes. A quasi-experimental design can compare an intervention group with a non-randomly assigned comparison group.
Comparison group: O₁ — O₂
Example
School A introduces a new reading programme, while School B continues its usual teaching. Both schools assess students before and after the programme.
Key issue: selection bias
The schools may differ in teacher experience, student background, resources or leadership before the intervention. These differences can affect the outcome.
Interrupted Time Series and Regression Discontinuity
Interrupted time-series
Repeatedly measures an outcome before and after an intervention to examine changes in its level or trend.
Example: Monthly school attendance rates are recorded for a year before and after a new attendance policy.
Strength: Multiple pre-intervention observations help establish the prior trend.
Limitation: Other events occurring around the intervention may explain the change.
Regression discontinuity
Uses a defined eligibility threshold to compare units just above and below the cut-off.
Example: Students scoring below a set baseline threshold qualify for a tutoring programme; outcomes are compared near that threshold.
Key requirement: Credible inference depends on assumptions about the assignment rule, continuity and the absence of manipulation around the cut-off.
Internal Validity and Threats
Internal validity concerns how confidently a study supports the conclusion that the intervention caused the observed outcome in the study setting.
History
An external event occurs during the study and affects outcomes.
Maturation
Participants change naturally over time, independent of the intervention.
Testing effect
Taking a pre-test influences performance on a later test.
Instrumentation
The measurement tool or scoring procedure changes over time.
Selection bias
Groups differ systematically before the intervention.
Attrition
Participants drop out, potentially changing the composition of the groups.
Internal vs External Validity
Internal validity
How credible is the causal explanation within this study?
External validity
How far can the findings reasonably generalise to other people, settings, interventions or times?
Other important quality concepts
- Construct validity: Do the measures and manipulations represent the intended concepts?
- Statistical conclusion validity: Are the statistical conclusions justified by the data and analysis?
- Reliability: Are measurements or procedures sufficiently consistent?
- Ecological validity: How well do the study conditions reflect the real-world contexts of interest?
Steps in Conducting an Experiment
- Identify a research problem and formulate a research question.
- Review relevant theory and previous evidence.
- Define the IV, DV and operational measures.
- State hypotheses where appropriate.
- Choose the design and determine the sample.
- Plan assignment, comparison conditions and procedures.
- Obtain ethical approval and informed consent where required.
- Conduct the intervention and collect data consistently.
- Analyse the results using appropriate methods.
- Interpret findings, assess limitations and report transparently.
Interpreting Experimental Results
Statistical significance
Indicates how incompatible the observed data are with a specified null model under its assumptions. A p-value is not the probability that the null hypothesis is true.
Effect size
Describes the magnitude of a difference or relationship, such as a mean difference or standardised mean difference.
Confidence interval
Communicates the precision of an estimate under the assumptions of the statistical procedure.
Practical significance
Considers whether the size and consequences of the effect matter in a real educational context.
Experimental Design Comparison Table
| Design | Intervention | Random assignment | Comparison / measurements |
|---|---|---|---|
| One-shot case study | Yes | No | Post-test only; no comparison group |
| One-group pre/post | Yes | No | One group measured before and after |
| True experimental | Yes | Yes | Randomised conditions with suitable outcome measurement |
| Quasi-experimental | Yes | No | Comparison or time-series design without random assignment |
| Ex post facto | No researcher manipulation of suspected cause | No | Existing groups or conditions are examined |
Memory Map for UGC NET
Variables
IV = manipulated factor; DV = measured outcome; confounder = alternative explanation for an observed relationship or difference.
Assignment
Random sampling selects participants; random assignment allocates participants to conditions.
Designs
Pre-experimental = limited control; true experimental = random assignment; quasi-experimental = intervention without random assignment.
Validity
Internal = credibility of causal inference; external = generalisability.
Results
Statistical significance, effect size and practical importance answer different questions.
15 Experimental Research MCQs
Select one answer for each question. Feedback and explanations appear immediately.
Score: 0 / 15 answered
