EXPERIMENTS AND CAUSAL CLAIMS

Remember that it takes all of these things to demonstrate causality:

• Covariance

• Temporal precedence
• Internal Validity

Covariance

exists because we have demonstrated that the cause (independent variable) is related to the effect (dependent variable). In our distributed practice example, the way the study period is split up is related to memory accuracy. Specifically, a larger number of shorter study periods is associated with better memory, and a single longer study period is associated with worse memory.

Temporal precedence exists because the cause comes before the effect. The experimenter presents the words before giving the test.

Internal validity exists because a well-designed experiment

rules out alternative explanations for the data. The groups were randomly assigned, so we can't say that the distributed practice (DP) group was just smarter from the start. The two groups received the same words on the list, so we can't say that the DP group had an easier task. The two groups received the same instructions, so we can''t say that the DP group understood the task better. The only explanation for the difference between the two groups is the manipulation of the independent variable.

WHAT IF IT DOESN'T WORK?But what if you conduct your experiment , and there is no difference between the groups? Does this mean that your hypothesis is incorrect? Possibly, but there might be other explanations.

It is possible that your manipulation of the independent variable was not powerful enough to make a difference. Perhaps, for example, 8 seconds isn't enough study time to show an effect of distributing practice. 8 minutes might be more appropriate. Or perhaps studying single words isn't a realistic enough situation to show an effect; it might be better to give participants a chapter of a textbook to study in several short periods or one long period.

Maybe the task was too easy, and everyone's score was too close to perfect. If the massed practice group got all of the words right, then it would be impossible for the distributed practice group to better them. When all scores are near the maximum, we say that a CEILING EFFECT has occurred. The dependent variable should be made more difficult.

What could you do to eliminate the ceiling effect?

Notes

Think back to the example of the robbery video. What if the lighting were so poor and the crime so brief that almost no one selected the correct photo of the robber? When the task is too difficult, we may see FLOOR EFFECTSeveryone's score is near chance, and we cannot see any effect of the dependent variable. Here, the task should be made easier.

Example 1:VIDEO NOTES Two groups of participants watch a video of a convenience store robbery. In each video, the same white male actor bursts into the store and yells "give me all of your money!" to the clerk. The clerk opens the cash register, gives the robber the cash, and the robber runs out. Both videos run for exactly 10 seconds. The only difference between the videos is that in one, the actor playing the robber wears a baseball cap, and in the other, he wears no hat. Both groups of participants are given a distracting taska Sudoku puzzleto work on for 15 minutes, and then both groups are shown a lineup of 6 photographs of white males. Both groups are instructed to pick the robber out of the lineup. 40% of the "hat" group correctly choose the third photograph as the robber; 55% of the "no hat" group correctly choose the third photograph as the robber.

How could we re-do the robbery study to eliminate the floor effect?