Research design

Research Design

Once a sample has been recruited for a study, the participants must then be organised into different groups according to the research design of the experiment.  The research design essentially refers to how participants in the study are grouped into experimental and control groups. It essentially determines which participants are exposed to which condition(s) of the independent variable, and in which order. Usually, the experimental group is exposed to the manipulated independent variable, while the independent variable is held constant for the control group.

While experiments, in general, may have multiple conditions, simple experiments have only two:

The research design is sometimes referred to as the experimental design, because it is unique to experimental research. In the simplest terms, the research design assigns (also referred to as allocates) participants to different conditions of the experiment. The three most common experimental designs in psychological research are:

  • Independent measures - different groups of participants are exposed to only one condition of the experiment.

  • Repeated measures - participants are exposed to all conditions of the experiment.

  • Matched pairs - participants are paired with another participant sharing similar characteristics, but each is exposed to only one condition of the experiment.

Not surprisingly, the choice of research design depends on the topic under study, and each design has its own strengths and limitations. 

Independent measures

Participants are only exposed to one condition of the independent variable, so a simple experiment is characterised by two (or more) distinct groups: the experimental group and the control group.

The main advantage of independent measures is that the design limits several of the ways participants might bias an experiment. Sometimes, if participants figure out the aim or hypothesis of the study they may change their behaviour to deliberately please, or to deliberately confound, the researcher or hypothesis. In psychological terms, such effects are known as demand characteristics, and the underlying idea is that an experiment's design might 'demand' certain behaviours. This then becomes an artefact in the collected data, where what appears to be an effect of behaving or thinking is really an effect of the experimental design. The point is, independent measures avoids all that.

The trade-off for independent measures is that it may be affected by participant variables, which means that each group of participants may have different personal characteristics, abilities, and so on. These differences between participants may affect the data, as participants in one group may behave differently than the other group not because of the independent variable, but because of participant variables.

Participant variables are a more specific type of extraneous variable, which is basically anything other than the independent variable(s) which might be affecting the dependent variable and how it is measured. For example, the results of an experiment on short-term memory capacity could be affected by any number of participant variables in the non-equivalent groups. This could include individual differences in short-term memory capacity to begin with, how much sleep each participant got the night before, what they had for breakfast and so on. Any of these factors could affect short-term memory capacity in the results, not because of the manipulated independent variable, but because of participant variables.

In order to try to minimise the effects of participant variables, random allocation/assignment to experimental conditions is used.  Allocation - or assignment - refers to the organising of participants into the different groups for the experiment, according to the research design.  With random allocation, participants have the same chance of being assigned to the experimental or the control condition. This hopefully results in a somewhat ‘balancing out’ of participant variables which then lessens the potential for any differences to influence the results.  

Repeated measures

In contrast to independent measures, the repeated measures design uses the same participants in all conditions of an experiment. 

Repeated measures are often chosen as the design when researchers want to limit the influence of participant variables on experimental results. For some research topics, factors like intelligence, gender, or socioeconomic status can have a considerable impact on experimental findings. Since repeated measures use the same participants in all conditions, this means participant variables are not a major factor.  Repeated measures control participant variables more so than any other design.  If an experiment's results are likely to be affected by participant variables, then repeated measures is a logical choice for design.

The biggest disadvantage of the repeated measures design is that it may result in order effects, which means that the effect of participation in one condition of the experiment then affects participant behaviour in later conditions of the experiment. Order effects are also known as carryover effects, which include:

  • Practise (or learning) - participants perform experimental tasks better in later conditions, because they have practised or learned the tasks in the first condition.

  • Fatigue (or boredom) - participants perform experimental tasks worse in later conditions, because of tiredness or boredom.

  • Context - testing in the first condition may influence how participants interpret experimental tasks in the second condition, especially when participants begin to guess the aim of the experiment.

Controlling for order effects

Counterbalancing is a way of controlling for order effects that often arise when using a repeated measures design. If the conditions of an experiment were labelled A and B, counterbalancing means that half of the participants complete the experiment in order AB, while the other half does it in order BA. While this doesn't actually remove order effects, it counterbalances them between the experimental and control conditions, so that any order effects from learning, practice, fatigue, or whatever, affect both conditions.

It is important to use random assignment to allocate participants to orders, so that every participant has an equal chance of completing the experiment in order AB as BA, for example.  Order effects can be further controlled through several other strategies, for example:

  • including a long enough time lag between conditions that any order effects like learning or fatigue have worn off

  • randomising stimulus materials in the experiment

  • randomising the order in which experimental/control conditions are presented to each participant.

Matched pair designs

This is the 'same but different' of experimental designs: each condition uses different participants, but they are matched to each other according to characteristics that might shape the outcome of the experiment. These characteristics could include age, gender, intelligence, culture, socioeconomic status, musical training, or any other personal characteristic relevant to the study.

In some ways, the matched pair design attempts to strike a 'happy medium' between independent measures and repeated measures. The design accounts for the impact of participant variables by matching participants in each condition according to whatever personal characteristics are relevant to the study. It also accounts for order effects by exposing participants to only one condition of the independent variable.  However, it takes a lot of work to match participants, especially when the matching involves multiple criteria. Furthermore, it is almost impossible to match each pair identically, except when identical twins are sampled.