Between Subjects Design

• Between-subjects design


  A study in which individuals are assigned to only one treatment or experimental condition and each person provides only one score for data analysis
• Within-subjects design


  An experimental design in which the effects of treatments are seen through the comparison of scores of the same participant observed under all the treatment conditions
• The American Psychological Association (APA) recommends that the word participants be used when referring to human participants in research, and the word subjects when referring to animals
• Selecting and recruiting subjects

  • Accurately generalizing results depends on the representativeness of the sample
  • Random sampling procedure is highly desirable to increase external validity
  • Practical limits in recruiting participants (e.g. using university subject pools, asking friends/relatives)
• Representativeness


  The extent to which a sample mirrors a researcher's target population and reflects its characteristics
• There is no simple answer to how many subjects are enough for an experiment
• Determining number of subjects

  • More subjects are needed when individuals in the population are likely to be quite different
  • Larger samples are more likely to mirror the actual state of the population
  • The number of subjects needed depends on the size of the effect produced by the independent variable (effect size)
• Effect size


  A statistical estimate of the size or magnitude of the treatment effect
• Two-group design


  The simplest experiment with only one independent variable and two treatment conditions
• Two-independent-groups design


  Subjects are placed in each of two treatment conditions through random assignment
• Random assignment


  Every subject has an equal chance of being placed in any of the treatment conditions
• If a researcher cannot use true random selection in deciding who the subjects of an experiment will be, he or she always uses random assignment to the treatment conditions
• Even when it is not possible to select your subjects entirely at random, the two-independent groups design can still be used as long as subjects are randomly assigned to each group
• In an independent groups design, the groups are "independent" of each other. The makeup of one group has no effect on that of the other
• If subjects are not randomly assigned to treatment groups, confounding can occur
• Random assignment controls for the differences that exist between subjects before the experiment
• Random assignment guards against the possibility that subjects' characteristics will vary systematically along with the independent variable
• Random assignment

  1. Flipping a coin (for two treatment conditions)

  2. Using a random number table (for more than two conditions)
• Random selection and random assignment are two separate procedures
• Random assignment is critical to internal validity
• If the "warm" descriptions were given out first, it could have confounded the results
• Experimental condition


  Apply a particular value of the independent variable to the subjects and measure the dependent variable
• Control condition


  Carry out the same procedures as the experimental condition, except for the experimental manipulation
• No-treatment control condition

  Measure subjects' responses without trying to alter them in any way
• A no-treatment control condition tells us how subjects ordinarily perform on the dependent measure
• Researchers must be careful that the control group is not inadvertently engaging in behaviors that would affect the results of the experiment
• A two-experimental-groups design can be used to look at behavioral differences that occur when subjects are exposed to two different values or levels of the independent variable
• Forming two independent groups

  1. Randomly assign subjects to experimental and control groups (e.g. by flipping a coin)

  2. Aim to create groups that are equivalent on important subject variables
• Even if individual subjects differ, the average of the groups should be about equal if random assignment was successful
• Hypothetical weights of subjects assigned to treatment conditions

  • Experimental Group
  • Control Group
• Even though individual roaches in each group weigh different amounts, the average weight of the groups is about equal
• The difference between the groups is not significant; it is not enough to merit concern
• Assigning subjects at random, we created two groups that are equivalent on an important subject variable
• Randomization controls for differences we have not identified but that might somehow bias the study
• The more subjects we have available to assign to treatment conditions, the better the chances are that randomization will lead to equivalent groups of subjects
• The more subjects we have, the better our chances are of achieving groups that are equivalent on a number of different characteristics
• Hypothetical weights of new groups

  • Experimental Group
  • Control Group
• Randomization did not produce comparable groups, the control group turned out to be much heavier than the experimental group
• The difference in weight is a confounding variable that could contaminate the results
• Two-matched-groups design


  There are two groups of subjects, but the researcher assigns them to groups by matching or equating them on a characteristic that will probably affect the dependent variable