Quiz 7 (Ch. 8a) Flashcards
List 5 reasons for performing an experiment
- Testing a theory
- Examining potential functional relations
- Developing new control techniques
- Systematic replication
- Trying out a new method or technique
Why do we need to be wary about using the presumed importance of data as a criterion for evaluating them?
Presumptions about the importance of data may lead the investigator to be biased if results are the not predicted or “wanted.” The true importance of the data may be overlooked.
What does Sidman mean when he says “good data are notoriously fickle?”
They can change their allegiance from theory to theory and may maintain their importance in the presence of no theory at all.
Why is it important to separate data from your purpose in collecting it?
If you hold onto your hypothesis too closely you may develop tunnel vision and miss accidental discoveries. The most important purpose of a study is to understand how the environment influences bx. All well controlled studies yield positive results.
What is a major advantage of studies performed to indulge the investigator’s curiosity about nature?
They will reveal observations and knowledge not formally known. These observations then lead to other questions and answers. They never produce negative results and the scientist’s life is said to be “…full of surprises”.
How do theories typically emerge from studies designed to indulge the investigator’s curiosity?
Inductively. Functional relationships b/t variables are first determined and sets of relations are linked together in theories.
What are some advantages of bringing an area of concern like depression into the laboratory?
You can better control variables and produce experimental control
What are some problems with treating all subjects alike?
- One level of the tx cuts the curve of different subjects at different points. If you know this you need to treat them differently to get a similar effect.
- You also have to wait for bx to stabilize before introducing a tx. Different subjects will take a different amount of time before bx stabilizes. If you train all subjects for 10 days each subject will all be at a different point in learning. This introduces another source of poor experimental control.
- The shape of the function between different levels of the IV and DV will likely differ for different subjects. It is better to control for this.
(For example, rather than say you will train all subjects for 10 days on a discrimination say you will train them until the discriminations reach 90% or better for 3 days. Another example would be to adjust shock intensity until you get 50% suppression in responding rather than say you will use the same level of shock for all animals.)
Why is it impossible to control variables through the use of statistics?
Statistical techniques only hide the effects of an uncontrolled variables. They only randomly distribute them. You can only control variability by controlling sources of variability.
Why is it futile to attempt to increase control by testing additional subjects? What effect will testing large numbers of subjects have on experimental control?
Using additional subjects allow you to detect a smaller effect but the additional subjects have no influence on the level of experimental control and will not identify the important variables that are causing the variability in the behavior of interest.
Why are parametric data useful in behavior analysis?
Parametric data show us the shape of the function. The function for several individuals may have the same shape, but they may peak at different points. Once you have the function you can see that what appeared to be differences in effects of the same variable on different subjects was really the result of the IV hitting those different subjects at different points on the curve.
What should be assumed when a large degree of variability is observed in the data? How does this result represent an opportunity?
If there is a large degree of variability one should assume that there is a powerful uncontrolled variable present. This represents an opportunity to identify this variable and control it. Controlling this variable will result in better experimental control and in better understanding of the variables and factors that have a large effect on the behavior you are studying.
Why does Sidman say that variables are not randomly distributed?
Bx in new environments, including social environments, have order due to previous selection. In a new environment, there may be a variable or even multiple variables one has been previously exposed to. People have learning hx that may be distributed in a non-random way and may vary within the community or b/t communities. Learning opportunities arranged by the environment are not typically randomly distributed but are instead a function of systems in place.
In what two situations is it not appropriate to report all the data? What is the problem with reporting all data?
- When the source of variability has been isolated and controlled
- When a uncontrolled variable is discovered that violates the design of the study, there is no purpose in diverting the reader from the purpose of the study by showing them data that resulted from poor control.
What are two direct behavioral methods to increase variability?
Poor experimental control results from poor experimental control techniques. If you fail to control relevant variables and they vary from day to day the bx will vary from day to day. If you haven’t tracked down important sources of variability you don’t even see these variables changing and conclude that it is “random or chance variation.”
What must be done to increase experimental control?
Control important variables
