The dependent variable is the major variable that will be measured in the research. For example, if you wanted to study the construct communicative competence of a group of students, then the dependent variable is the construct and it might be operationalized as your students' scores or ratings on some measure of communicative competence. The measurement would be part of the operational definition of communicative competence for your study. This variable (communicative competence) is the dependent variable. We expect performance on the dependent variable will be influenced by other variables. That is, it is 'dependent' in relation to other variables in the study. Let's consider two more examples.
Assume you wanted to know how well ESL students managed to give and accept compliments, one small part of communicative competence. Their videotaped performances during role-play were rated on a 10-point scale (10 being high) for giving compliments and a 10-point scale for receiving compliments and bridging to the new topic. The student's total score could range from 0 to 20. The rating, again, might be influenced by other variables in the research. The rating measures the dependent variable compliment performance.
In a text analysis, you hypothesized that use of English modals may be influenced by types of rhetorical organization found in texts. Each place a modal occurred in the text it was coded. Tallies were made of overall frequency of modals and also of each individual modal (e.g., can, may, might, should). The dependent variable would be modal and the levels within it might be the actual modal forms or, perhaps, the functions of modals (e.g., obligation, advisability, and so forth). The frequency of the dependent variable, modal, might be influenced by other variables included in the research.
An independent variable is a variable that the researcher suspects may relate to or influence the dependent variable. In a sense, the dependent variable "depends" on the independent variable. For example, if you wanted to know something about the communicative competence of your students, the dependent variable is the score for communicative competence. You might believe that male students and female students differ on this variable. You could, then, assign sex as the independent variable which affects the dependent variable in this study.
In the study of compliment offers/receipts, L1 membership might influence performance. L1, then, would be the independent variable which we believe will influence performance on the dependent variable in this research. There might be any number of levels in the independent variable, LI.
In text analysis, rhetorical structure may influence the use of modal auxiliaries. Rhetorical structure (which might have such levels as narrative, description, argumentation) would be the independent variable that affects the frequency of the dependent variable.
Sometimes researchers distinguish between major independent variables and moderating independent variables. For example, in the study of compliments, you might believe that sex is the most important variable to look at in explaining differences in student performance. However, you might decide that length of residence might moderate the effect of sex on compliment offers/receipts. That is, you might believe that women will be more successful than men in offering and receiving compliments in English. However, you might decide that, given more exposure to American culture, men would be as successful as women in the task. In this case, length of residence is an independent variable that functions as a moderator variable.
While some researchers make a distinction between independent and moderator variables, others call them both independent variables since they influence variability in the dependent variable. However, specifying variables as "independent" and "moderator" helps us study how moderating variables mediate or moderate the relationship between the independent and dependent variables.
A control variable is a variable that is not of central concern in a particular research project but which might affect the outcome. It is controlled by neutralizing its potential effect on the dependent variable. For example, it has been suggested that handedness can affect the ways that Ss respond in many tasks. In order not to worry about this variable, you could institute a control by including only right-handed Ss in your study. If you were doing an experiment involving Spanish, you might decide to control for language similarity and not include any speakers of non-Romance languages in your study. Whenever you control a variable in this way, you must remember that you are also limiting the generalizability of your study. For example, if you control for handedness, the results cannot be generalized to everyone. If you control for language similarity, you cannot generalize results to speakers of all languages.
If you think about this for a moment, you will see why different researchers obtain different answers to their questions depending on the control variables in the study. Comby (1987) gives a nice illustration of this. In her library research on hemispheric dominance for languages of bilinguals, she found researchers gave conflicting findings for the same research question. One could say that the research appeared inconclusive--some claimed left hemisphere dominance for both languages while others showed some right hemisphere involvement in either the first or second language. Two studies were particularly troublesome because they used the same task, examined adult, male, fluent bilinguals, and yet their answers differed. The first study showed left hemisphere dominance for each language for late bilinguals. The second showed the degree of lateralization depended on age of acquisition; early bilinguals demonstrated more right hemisphere involvement for L1 processing than late bilinguals, and late bilinguals demonstrated more right hemisphere involvement for L2 processing than early bilinguals.
The review is complicated but, to make a long story short, Comby reanalyzed the data in the second study instituting the control variables of the first study. Bilinguals now became only those of English plus Romance languages, handedness was now controlled for no family left-handed history; age was changed to one group--Ss between 20 and 35 years of age; and finally, a cutoff points for "early" vs. "late" bilingual were changed to agree with the first study. With all these changes, the findings now agreed.
It is important, then, to remember the control variables when we interpret the results of our research (not to generalize beyond the groups included in the study). The controls, in this case, allowed researchers to see patterns in the data. Without the controls, the patterns were not clear. At the same time, using controls means that we cannot generalize. Researchers who use controls often replicate their studies, gradually releasing the controls. This allows them to see which of the controls can be dropped to improve generalizability. It also allows them to discover which controls most influence performance.In the above examples, we have controlled the effect of an independent variable by eliminating it (and thus limiting generalizability). The control variables in these examples are nominal (discrete, discontinuous). For scored, continuous variables, it is possible statistically to control for the effect of a moderating variable. That is, we can adjust for preexisting differences in a variable. (This procedure is called ANCOVA, and the variable that is "controlled" is called a covariate.) As an example, imagine we wished to investigate how well male and female students from different first-language groups performed on a series of computer-assisted tasks. The focus in the research is the evaluation of the CAI lessons and the possible effect of sex and LI group membership on task performance. In collecting the data, we would undoubtedly notice that not all students read through the materials at the same speed. If we measure this pre-existing difference in reading speed, we can adjust the task performance scores taking reading speed into account. Notice that this statistical adjustment "controls" for preexisting differences in a variable which is not the focus of the re-search. Unlike the previous examples, the variable is not deleted; i.e., slow readers (or rapid readers) are not deleted from the study. While reading speed may be an important variable, it is not the focus of the research so, instead, its effect is neutralized by statistical procedures.
Other Intervening Independent Variables
We often hope to draw a direct relation between independent and dependent variables in our research. For example, we might want to look at the relationship between income and education. We would expect that with additional education, income would increase. If you collected data, you might be surprised to find that the relationship is weak. Additional education might increase the income of some people and not help others. How can we explain the lack of a direct relationship between additional education increased income? If you think about it for a moment, you can see that education is likely to increase the earning power of young people. They might earn the minimum wage at McDonald's while in high school and earn a much larger salary at IBM after college. So the increase in income is great. On the other hand, additional education is not likely to increase the income of older adults. Their salaries are already fairly high and the value of added classes may not be reflected in income. So for one age group the relation is that of additional education higher income, but for the other group this is not the case. There is an intervening variable, a variable that was not included in the study, at work.
As you can guess from this diagram, an intervening variable is the same thing as a moderating variable. The only difference is that the intervening variable has not been or cannot be identified in a precise way for inclusion in the research. In planning research, we want to be able to identify all the important variables (or control for them). However, sometimes this is impossible. For example, intervening variables may be difficult to represent since they may reflect internal mental processes. For example, when we talk about L1 L2 transfer or L1 L2 interference, we are talking about an internal mental process that we may or may not be able to measure accurately. Intelligence and test-taking talents may not be directly measurable yet play some role in changing research outcomes.
In all research, we can only account for some portion of the variability that we see in the major, dependent variable. We may look at the influence of many different independent variables to explain the variability in the dependent variable. However, there are many other factors which may or may not be important that we may fail to consider and that thus contribute to "error." However, whatever the findings, it is important to consider whether we have defined our variables well so that they reflect (as well as they possibly can) all the processes that we hope to tap.