What does it mean when a study is considered internally valid? This term refers to the design of the study, which allows you to be confident that the results are not open to alternative interpretations. No study is truly perfect, but there are ways you can maximize the integrity of an experiment.
The examples of validity given in the introduction to Section 2 are worth revisiting. For example, if you want to get an idea of whether people prefer cats or dogs, you should not just sample people from the Cat Protection League because they are already biased towards cats. If you want to examine if introducing a specific teaching technique is effective at improving academic performance, you need a comparison group because it may be that students improve over time. If you do not have a comparison group, any improvements could be a natural improvement over time, referred to as maturation, a common error in research design. Studies that examine effectiveness without a comparison or control group are considered internally invalid because you do not know if the improvements are because of the intervention or just maturation. These are all examples of threats to validity, and you will find an extensive list of others in your readings this week.
Reliability is a check of consistency. For example, if an IQ test is supposed to assess IQ, then your score on Monday should be the same as your score on Tuesday (test, re-test reliability), or that your score on half the random questions should be similar to your score for the other half (split-half reliability). When reviewing a video of children playing, if you conclude that child A played for 10 minutes, you should conclude the same each time the video is viewed (inter-rater reliability). You will see from the readings there are several different types of reliability.
Generalizability is the ability for the findings from your study to transfer to another setting. The generalizability of your findings depends on the internal validity of your study and the quality of your sampling.