A Primer of Scientific Terms
A concept is "an abstraction formed by generalization from particulars" (Kerlinger, 1986; p. 26). For example, geologists observed that some stones are clear, such as diamonds, and others not clear, such as onyx. Other stones are somewhere in between, such as garnet. In casual conversation we might say some stones are harder than others. The concept we are considering is "hardness." Concepts can be concrete (e.g., “cup” or “fight”) or abstract (e.g., “fear” or “aggression”).
A construct is a term that has been devised for a scientific purpose and is usually defined specifically (e.g., intelligence), although there may be debate about definition. A construct provides a definition that allows us to know if a particular instance is or is not a member of the category. For example, we might define violence as "intentional physical harm caused to a person." By this definition shouting at someone, being in an automobile accident, or butchering a chicken are not acts of violence. However, we could define violence differently in order to include or exclude other types of instances. The important point is that the construct of violence is precisely defined while the concept of violence is simply a referenced to an idea we understand in a casual sence.
Variables and Operationalization: A variable is a construct that has been defined so that instances of it can be assigned value and can be counted. Two characteristics are necessary: 1) A variable must be defined so that counting can occur. 2) The value of a variable must change either from one time to another or from one person (or unit) to another. A researcher might be interested in physical aggression on the playground. The researcher might define physical aggression as "a single instance of hitting, kicking, pushing, or biting." This is called an operational definition. It describes the "operation" conducted to measure the variable. After creating an operational definition the researcher could watch children on a playground and count the number of times one child commits an act of physical aggression toward another child. The sum would be a measure of physical aggression for any child. Another variable might be gender. The researcher could count the number of boys and girls. She then might compute the average number of acts of physical aggression committed by boys and the average for girls. Then by comparing the averages, she could determine if a) boys are more physically aggressive than girls, b) girls are more physically aggressive than boys, or c) boys and girls do not differ in physically aggressive behavior. The second important point about variables is that a variable must vary (or change). For example, over a series of weeks, one adult's height is not a variable: it does not change from one day or week to the next. However, weight may change from day to day. Thus, weight is a variable. This is important because if a construct does not vary, it can not be related to another construct. That is it cannot covary or be correlated with another variable. For example, daily caloric intake (how much one eats) might be correlated with weight: the more you eat the more you weigh. But for adults, caloric intake is not related to height on a daily or weekly basis: no matter how much you eat this week, you will be the same height next week. Notice that height is a variable when we consider more than one person. Again, consider gender. Height changes from one person to another, and on average, males tend to be taller than females.
Independent and Dependent Variables: These terms indicate that one variable is thought to cause another to change. The independent variable (X) is thought to cause a change in the dependent variable (Y). One way to remember how the two differ is to remember that a dependent variable "depends" on the independent variable for change. If the independent variable does not change, the dependent variable should not change. In experiments researchers change (or manipulate) one or more independent variables. The researchers then observe whether the dependent variable also changes. In the example above, theoretically, changes in caloric intake cause changes in weight. It is theoretically doubtful that weight causes caloric intake.
Causation: It is important to consider the issue of cause and effect. If we are to argue that X causes Y, then we must meet three criteria or answer three questions. 1) Covariation - Are X and Y related? If the two variables do not covary or are not correlated, then a change in one variable cannot cause or result from change in the other. 2) Temporal Order - Did the change in X precede the change in Y? If X did not change before Y, then the change in X could not have caused the change in Y. 3) Elimination of Alternative Explanations - This is the most difficult criterion. If one is arguing that X causes Y, then one must be able to effectively argue that one's explanation of the relationship between the two variables is the best explanation. A somewhat silly, but classic example will help illustrate this third criterion. I may argue that rain causes umbrellas to open. Rain or no rain is the independent variable and umbrellas being opened or closed is the dependent variable. The two variables covary: more umbrellas are opened when it is raining than when it is not. X precedes Y in time: umbrellas very rarely open before it rains, but they frequently open shortly after rain starts. So logically, rain causes umbrellas to open. However, if one lays ten umbrellas in a field before a rain. One will observe that when the rain starts, the umbrellas remain closed. Therefore, my argument is wrong: rain does not cause umbrellas to open. An alternative, and better, explanation is that people's desire to remain dry causes umbrellas to open. The rain activates people's desire to stay dry. Their desire to stay dry causes them to open their umbrellas. My model would look like this: Rain --> Open Umbrella. The better model would look like this: Rain --> Desire for Dry --> Open Umbrella. In the end, we can never be certain our explanation is the best one, only that it is the best we have of so far.
Research tends to look at variables as differing between groups or as
being related to other variables. Both
are usually true. The distinction is
more a function of how research is conducted than the true state of the
variables. For example, one can conceive
of tobacco and cancer as the difference between smokers and none smokers or as
the relationship between amount of smoking and the incidence of lung cancer. Group differences are measured as the
difference in means or frequency between or among groups and are tested using a
number of methods including Chi-square, t-tests, and more sophisticated
analyses of variance (ANOVA).
Relationships are measured through correlations, regressions, and path
analysis. A more thorough discussion of
all of this is available in many research texts. My favorite is the one cited below.
Kerlinger, F. N. (1986). Foundations
of Behavioral Research. Hold, Rinehart, and Winston, Inc.,