{[ promptMessage ]}

Bookmark it

{[ promptMessage ]}

martin - 24 Chapter One I Summary As scientists of human...

Info iconThis preview shows pages 1–9. Sign up to view the full content.

View Full Document Right Arrow Icon
Background image of page 1

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full Document Right Arrow Icon
Background image of page 2
Background image of page 3

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full Document Right Arrow Icon
Background image of page 4
Background image of page 5

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full Document Right Arrow Icon
Background image of page 6
Background image of page 7

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full Document Right Arrow Icon
Background image of page 8
Background image of page 9
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: 24 Chapter One I Summary As scientists of human behavior, psychologists have many research designs available to them, all of which aim to establish relationships between events and to fit these relationships into an orderly body of knowledge. Among the quantitative designs is the experimental method, which is the primary focus of this book. This method requires that a particular circumstance be manipu- lated and some aspect of behavior measured. From an experiment it is possi- ble to say whether the manipulation of the circumstance caused any change found in the behavior. Sometimes when an experimental approach cannot be used, it is neces- sary to use correlational observations, in which variables are observed and their relationships evaluated. The results of such a study cannot be used to establish causal relationships, because none of the variables is under the control of the investigator. Correlational observations are often carried out using a survey in the form of a questionnaire or interview. Correlational data can also be obtained by doing archival research with data contained in public or private records, such as census data or court records. Some investigators are now doing research that employs qualitative designs. Qualitative researchers use descriptive data: written descriptions of people, including opinions and attitudes, and of events and environments. In ethnography the investigators use interviews and sometimes participatory observations to gather descriptive data. In one form of qualitative research they use naturalistic observation, in which data are gathered in realistic settings. A final qualitative design used when the potential number of obser- vations is limited is the case history, in which detailed accounts of the events in a person’s life or in a historical incident are described and analyzed. If _.__. xy‘ ' ' . ' . .411"! I [How to Do EXPe" ments During its long history down to the middle of the nineteenth century, psychology was cultivated by able thinkers who did not realize their need of carefully observed facts. . . . Finally psychologists decided that they must follow the lead of physics, chemistry and physiology and transform psychology into an experimental science. R. s. WOODWORTH (1940) We must guard against . . . the drawing of a preconceived conclus'on from experiments or observations which are so vaguely conditio d that a variety of inferences are as a matter of fact possible. K. DUNLAP (1920) In the first chapter we briefly discussed the experimental method. You will recall that the major advantage of doing this type of research is that it allows you to make causal statements—that a circumstance caused a change in behavior. Because this type of statement is precise, the rules required to sup- port the statement are quite stringent. Most of these rules involve being able to account for all the circumstances that could vary. By way of example, suppose that we were interested in the time it takes a person to press a button in response to a light of a particular intensity. At this point we have chosen a circumstance to manipulate—the intensity of a light—and a behavior to measure—the time to press a button. These two variables have formal names. I Variables INDEPENDENT VARIABLES The circumstance of major interest in an experiment, light intensity in our example, is called an independent variable. The best way to remember this name is to recall that the variable is independent of the participant’s behavior. As experimenters, we manipulate this variable—that is, choose two or more levels to present to the participant—and nothing the participant does can 25 26 Chapter Two change the levels we have chosen. For example, if our independent variable is light intensity, we might select a high-intensity light and a low-intensity light as our two levels and observe behavior under both circumstances. Without at least two levels, we are not doing an experiment, but we are free to choose many more levels or to have more than one independent variable. In the later chapters I discuss ways of designing these more complex experiments. DEPENDENT VARIABLES Once we have chosen the independent variable, we will want to measure a participant’s behavior in response to manipulations of that variable. We call the behavior we choose to measure the dependent variable because it is dependent on what the participant does.1 In thégeaction-time experiment, for example, our aim is to find out whether a relationship exists between light intensity and time to respond. Thus, our dependent variable is the time from when the light is turned on until the participant presses a button. Making a statement about the expected nature of the relationship is some- times useful; such a statement is called a hypothesis. In the example, we might hypothesize that the more intense the light, the quicker the response will be. The outcome of the experiment will determine whether the hypothesis is supported and becomes part of the scientific body of knowledge or whether it is refuted. I have discussed hypotheses in several different places in this book. In the next chapter we will consider how hypotheses can be deduced from theories and how they must be true if the theory is true. In Chapter 12 we will discuss the concept of a null hypothesis. As we will see, the null hypothesis is just a statistical statement saying that the independent variable has no effect on the dependent variable for a population. However, if you really believed there would be no effect in your experiment, you would probably not carry out that experiment. In actuality, you are usually predicting that the change in levels of the independent variable will cause a change in the dependent variable. This prediction is your real hypothesis. In fact, experimenters often go beyond this nondirectional hypothesis of simply predicting some change and state a directional hypothesis predicting the direction the dependent variable will change as the independent variable is manipulated. In some cases hypotheses are not even based on theories, particularly when you simply wonder what would happen to a behavior if the indepen- dent variable were manipulated. In this case, the hypothesis is simply the answer to a question. How does crowding affect aggression? Does marking your first guess or thinking longer lead to better grades on multiple-choice tests? Are politicians who smile in their campaign posters more or less likely to win elections than those who don’t? Hypothesized answers to questions such as these can also add to the scientific body of knowledge. 1 I believe it is easier to remember the term this way, although the word dependent really refers to the behavior’s being potentially dependent on the levels of independent variable. How to Do Experiments 27 CONTROL VARIABLES So far, we have chosen one circumstance to manipulate—the independent variable. However, in some way, we will need to account for other circum- stances in an experiment. One possibility is to control the other circumstances, thus making them control variables. We can control such circumstances by making sure that they do not vary from a single level. For example, in our reaction-time experiment, we might require constant lighting conditions in the room, only right-handed participants, a constant temperature, and so on. Ideally, all circumstances other than the independent variable would stay con- stant throughout an experiment. We would then know that any change in the dependent variable must be due to the changes we had brought about in the independent variable. The concept of control is vital for experimentation and makes the exper- imental method distinct from the other forms of research that I discussed in the previous chapter. In your experiments, many of the variables will be set as control variables. As an experimenter, you will want to be sure that you have indeed achieved complete command of the control variables in your experi- ment. This is why psychologists go to considerable expense to build special environments in which sound, light, and temperature are controlled and to use special equipment to ensure that stimulus characteristics are consistent and that responses are carefully measured. However, even though many variables in your experiments will be control variables, you should realize that, especially in psychology, not all variables will be assigned as control variables. First, the experimenter cannot control all the variables. It is impossible not only to control many genetic and environ- mental conditions but also to force cooperative attitudes, attentional states, metabolic rates, and many other situational factors on human participants. Second, we really do not want to control all the variables in an experi- ment otherwise we would create a unique set of circumstances. If we could control all variables while manipulating the independent variable, the rela- tionship established by the experiment would hold in only one case—when all variables were set at exactly the levels established for control. In other words, we could not generalize the experimental result to any other situation. As a rule of thumb, the more highly controlled the experiment, the less generally applicable the results. Suppose, for example, that General Nosedive from the US. Air Force came to you and said: ”Say, I understand you ran an experiment on reaction time. Tell me how intense I should make the fire-warning light in my fighter planes so that my pilots will respond within half a second.” Having conducted a well-controlled experiment, you reply, ”Sir, if you can guarantee that the pilot is a 19-year-old college sophomore with an IQ of 115, sitting in an air- conditioned, 10-foot-by-15-foot room, with no distracting sounds and nothing else to do, and if you always give a warning signal 1 second before the light comes on, then I might be able to give you an answer.” You can probably imagine the general’s reply. The moral of the story: if you want to generalize the results of your experiment do not control all the variables. 28 Chapter Two The generalizability of an experimental finding is also referred to as external validity—how well a causal relationship can be generalized across people, settings, and times. Cook and Campbell (1979) have defined several types of validity. The way they use this term, validity refers to whether draw- ing expergiental conclusions about cause is justifiable. I will introduce other terms for validity at appropriate places in the book. Threats to external valid- ity might occur if you use a limited sample, such as college sophomores, when you want to generalize to all humans of any age or intelligence (includ- ing, as in our example, Air Force pilots). Or you might have done a highly controlled laboratory experiment when you want to generalize to real-world work settings where it is noisy, hot, and crowded, and the workers are tired and unmotivated but have lots of practice. In general, the more tightly con- trolled your experiment—that is, the more circumstances you choose to make into control variables—the more likely it is to suffer from threats to external validity. RANDOM VARIABLES Having established that we do not want to control all the circumstances, what can we do with the remaining circumstances in our experiment? One possi- bility is to let them vary. In what way can we allow the circumstances to vary and still be sure that they will not bias our experiment? One alternative is to permit some of the circumstances to vary randomly. These variables are termed random variables. The term random or randomization is used in several different ways in sci- ence. One use of the term is in the context of random selection of items from a population to form a representative sample. In this case, a population of items is available and some random process is used that makes the selection of any one item from that population as likely as the selection of any other item. Random selection is used to ensure external validity, that is, to ensure that the sample of items randomly selected from the population is generaliz- able to that population. So, if you wanted to generalize the results from an experiment to all people in the United States, ideally you would use a means of selection that was equivalent to putting the name of everybody in the coun- try into an enormous hat and drawing out a sample of names. You could then say that you have randomly selected your sample and you could claim good external validity of your findings. However, in this context the word random in the term random variable usually refers to random assignment of circumstances to the levels of the independent variable. Many of the circumstances in an experiment concern individual differences in the participants. Obviously, if we use the same par- ticipants for the various levels of the independent variable, we do not have to worry about individual differences. However, if we use different partici- pants for each level of the independent variable, then we have to make sure that the characteristics of the participants assigned to each level do not bias our conclusions. For example, suppose that you want to determine the effects of TV Violence on aggression in children. After you have randomly selected How to Do Experiments 29 two hundred 6-year-old children as a sample from some larger population, you might then randomly assign them to two levels of the independent vari- able: Viewing violent TV shows and viewing nonviolent TV shows. Perhaps you could flip a coin for each child and assign the child to the first group if a head occurred and to the second group if a tail occurred. Is it possible that most of the children in first group attend violent schools or eat lots of sugar or come from abusive homes, while few of those in the second group do? Yes, but if the selection was done in a truly random manner, it is statistically unlikely for such large samples to be biased. Suppose that you let the children watch the violent or nonviolent TV shows at home. Is it possible that most of the children in one group have large-screen theater-system TVs at home, while most of those in the second group have small portable TVs? Again, it is possible but not probable; ran- domness makes this possibility highly unlikely. There is no particular trick to random assignment or random selection. You can use any device that allows each item an equal chance of assignment or selection. As in the example, if you want to form two groups, you can flip a coin to form them.2 If there are six groups, you can throw a die. If there are 33 groups, you can use 33 equal-sized slips of paper. Most mathematical handbooks and many statistics texts have random-number tables based on a process equivalent to drawing from 10,000 slips of paper. I have included one such table in the back of this book as Appendix C. Using any column or columns in a table of random numbers, you can assign each of your items a number and select the item when that number occurs. Just ignore the extra columns or numbers that are not on your list. If you happen to be a computer buff, you can use the computer to generate random numbers or events.3 If you have chosen to make a circumstance into a random variable, you must be sure that it varies in a truly random way, because not all events that appear random are really so. For instance, if you try to randomize conditions in an experiment by assigning events yourself, you have not randomized! Humans are notoriously bad at producing random events. If you assume that participants will show up for an experiment throughout the day or through- out the semester in a random order, you are wrong! People who are morning or afternoon volunteers or early-semester or late-semester volunteers have different characteristics. New experimenters commonly make mistakes in randomization. Don’t you make them! Perhaps most of the circumstances that become random variables in your experiment will be associated with participants and can be randomized by randomly assigning participants. However, other circumstances that are not associated with participants can sometimes be treated as random variables. Suppose in our TV violence experiment that the room in which the children 2 Actually, most coins are slightly biased in favor of heads, but, unless your experiment has over 10,000 trials, don’t worry about it. 3 Computers are also less than perfect at generating random events, but they’re much better than coins. For assigning events in an experiment, it doesn’t make much difference which method you use. 30 Chapter Two watch TV is available either in the morning or in the afternoon. If you think there is a reason why watching in the morning versus the afternoon may cause differences in how aggressive children become independent of the amount of violence in the show, then you might want to randomly assign the violent-TV and the nonviolent-TV groups to morning and afternoon times. You certainly would not want one group to watch exclusively in the morning and the other in the afternoon. There are other circumtances that also might affect the aggressiveness of children that you may consider as random variables even though you have not been able to make truly random assignments; for example, how stormy the weather is or how much violence there is in the news on a particular day. If you have multiple sessions of your experiment, you would probably not be too far off if you assume that these circumstances are randomly distributed across the levels of your independent variable and that they will not system- atically bias your results. As mentioned earlier, the major advantage of random selection is the generalizability of the results. Every time you choose to make a circumstance into a control variable, you can generalize the results to only that level of the variable. However, if you allow many levels of the circumstance to exist in a population and then randomly choose a sample, you can generalize to the entire population. The major advantage of random assignment is the elimination of bias from the results. Thus, randomization can be a powerful experimental tool. RAN DOMIZATION WITHIN CONSTRAINTS In some cases you may not want to make a circumstance into either a random or a control variable. Actually, randomization and control define opposite ends of a continuum. Falling between these two extremes are vari- ous degrees of randomization within constraints. In this case, you control one part of the event assignments and randomize the other. Suppose that in our reaction-time experiment we knew that practice could be an important variable. If we presented all the low-intensity trials first, followed by all the high- intensity trials, we could be accused of biasing the experiment; any difference between response times to low- versus high-intensity light might, in fact, be due to short versus long practice. To avoid this problem, we could decide to control the practice variable and give only one trial to each individual. Or we could assign the low- and high-intensity trials randomly over, say, 12 trials by flipping a coin and presenting a high-intensity light whenever a head occurred and a low-intensity light whenever a tail occurred. However, this alternative might not be the most attractive one, because it could result in an inadequate representation of high and low intensities. (For example, the flip- ping of the coin might result in only three high-intensity trials and nine low- intensity trials.) To avoid this possibility, we could decide to have an equal number of high- and low-intensity trials. How to Do Experiments 31 Thus, as a solution we establish a constraint on the assignment of trials (an equal number of each type of trial) and make a random assignment within this constraint. We might write the word high on six slips of paper and the word low on six and...
View Full Document

{[ snackBarMessage ]}