SOCI2000 6.0 The Experiment: Introduction

Lecture Slides

Week 6 Lecture Slides

Video Lecture (from 2019)


Classical Experimental design

Independent variable

Dependent variable

Random assignment



Experimental group

Control group

Key Concepts:

Double-blind experiment





Types of experiments:

Classical experimental design

True experiment


One-shot case study

Natural experiment

Field experiment


Internal validity

External validity




Introduction: The power of experiments

What is an experiment?

Experiments in social science use the same logic that guides experiments in biology or engineering.

Experimental research is the strongest method for testing causal relationships.

Experiments are the method that most clearly satisfies the three conditions needed to show causality:

  • temporal order (A happened before B),
  • association (when A goes up, B goes up), and
  • no alternative explanations (there is no third cause - a source of spurious correlation - that gives rise to this correlation).

Why are true experiments so powerful?

  • Hardest part of proving causality is ‘ruling out other explanations’ (i.e. ruling out other theories). * Randomisation into control or treatment makes two groups IDENTICAL.
  • Any differences after treatment MUST be due to treatment. It rules out ALL other explanations*

(*) so long as you didn’t screw up the experiment.

What types of questions are experiments suited to answering?

Research questions most appropriate for an experiment have:

  • A clear and simple logic. e.g. effect of one variable on another
  • Units of analysis that can be manipulated. e.g. individuals NOT countries, economies, planets, stars
  • A cause (independent variable) that the experimenter can manipulate. e.g. image on a computer screen, NOT growing up in poverty, or fearing for your life, or taking a dangerous drug
Last updated on 13 April, 2020 by Dr Nicholas Harrigan (