CONCEPT I: Gathering Data

The initial step of any scientific research involves the researcher in an imaginative or speculative period of incubation of ideas. During this period, the researcher formulates a hypothesis, a tentative statement about why something happens, and determines the appropriate method to be used.

Many research methods can be used, depending on the question asked. In most, subjects are selected through a process called sampling, the technique of selecting a few individuals to represent an entire group. A sample must adequately and fairly represent its population, and this can be accomplished in two ways. In random sampling, each member of a population has an equal chance of being selected in the sample. In representative sampling, a chosen group exactly matches the population on some characteristic(s) such as sex or age in terms of their percentages in the sample and in the population overall. Samples, even when they are very large, can misrepresent a population if they are drawn from a subgroup of the population.

Scientific research is intended to accomplish one or more of three basic goals: describe behavior, explain its causes, and predict when it might occur again.

The method called the experiment has the advantages of maintaining control over conditions and enabling inferences to be made about a cause-and-effect relationship between variables, which are factors capable of change. Its major disadvantage is that so much control may lead to an unnatural situation. In the experiment, after formulating a hypothesis, the next step is to specify the variables. The independent variable is the condition manipulated by the experimenter. It affects the dependent variable, which is the variable the experimenter measures. Subjects are randomly divided into an experimental group, which experiences the experimental treatment, and a control group, which does not. The control group is needed so that comparisons can be made. If a change in the independent variable can be shown to produce a change in the dependent variable, a cause-and-effect relationship has been demonstrated.

Sometimes a true experiment cannot be used to study a problem because the independent variable cannot be manipulated (for example, age or sex) or because subjects cannot be randomly assigned to experimental and control groups. In these cases, a quasi-experimental design is used. In one quasi-experimental design, called a time-series design, the experimenter measures the dependent variable before and after the introduction of the independent variable to determine if the independent variable had an effect. The major problem with time-series designs is that changes in the dependent variable may be caused by factors other than the independent variable.

Sometimes, for practical or ethical reasons researchers cannot conduct an experiment to test cause and effect. They may choose to study the degree of relationship between two variables by using correlational research. A positive correlation means that as one variable increases, so does the other. A negative correlation means that as one variable increases, the other decreases. Correlations can be represented numerically by a correlation coefficient. Correlational studies do not allow the researcher to draw conclusions about cause and effect.

A survey is an attempt to estimate opinions, characteristics, or behaviors of a population by polling a representative sample. Surveys can be conducted through interviews, questionnaires, or public records. Problems with surveys often include use of inadequate sampling techniques, or bias in questions and/or responses. Surveys can be the basis for correlation or provide descriptions of behavior that are interesting and useful in their own right.

Naturalistic observation involves watching and recording behavior as it naturally occurs. The major problem with this research method is that people behave differently when they believe they are being observed. To avoid this difficulty psychologists may use participant observation, in which they join the group they want to study. The major problem with observational studies is that researchers have no control over what happens and therefore cannot draw conclusions about cause and effect

A case study is an intensive investigation of one person, rather than a representative sample, usually because the observed behavior is rare or unique. Careful measurement of variables can be as important in a research project as the selection of an appropriate research method. A measure is valid if it measures what it intends to measure. It is reliable if it yields consistent results.

All research methods involve measurement of variables using one or all of three kinds of measures: self-report, behavioral, and physiological. Self-reports are verbal statements by subjects about how they feel, what they think, or how they behave. The usefulness of self-reports is limited because subjects may lie or may not know their true thoughts or feelings. Behavioral measures are quantified observations of how subjects actually behave which is sometimes quite different from how they think they will behave. Behavioral measures are inappropriate for studying mental events (such as fantasies) or private behaviors (such as sexual practices), and they can be more trouble to collect. Physiological assessments involve recording physical changes. Electroencephalograms (EEGs) measure brain activity and are used to study states of consciousness.

Psychologists may use all three types of measurement to gain a more accurate understanding of a research problem.

Another measurement decision involves change across time. Alongitudinal study is one in which a group of people is measured again and again over a number of years. Its major advantage is consistency. A cross-sectional study also measures change across time, but it does so by comparing measures taken at the same time of subgroups of people of various ages. Although it is quicker and less expensive, cross-sectional research fails to consider time-related differences between groups. Often longitudinal and cross-sectional studies are used in combination.

Research into the effects of television violence on behavior demonstrates the need to use more than one technique to gain an understanding of an important or complex relationship. For example, laboratory research may not always mirror accurately real-life settings, a concern referred to as lack of ecological validity. To address this question, field experiments are performed in a natural setting.

CONCEPT II: Analyzing Data

Statistics are mathematical methods for assessing and presenting data in summary form. Psychologists use two general types of statistics: descriptive and inferential.

Descriptive statistics are a form of shorthand for representing a large amount of data. Research studies often begin by calculating the range, or maximum interval, over which the behavior of interest takes place. The next step often involves calculating from the data its central tendency, which is an indication of the center, or middle, of a group of responses or scores. Three measures of central tendency are commonly used: the mean (or arithmetic average of the scores), the median (or middle score), and the mode (or most frequent score).

When a distribution of data is symmetrical, the mean, median, and mode are identical. Such a symmetrical distribution is called a normal distribution, or normal curve. If a distribution is not normal but skewed, the mean, median, and mode take on different values and the median becomes the least biased indicator of central tendency.

In addition to knowing the central tendency, psychologists also need to know how scattered the scores are, a characteristic called the dispersion, or variability, of the scores. The range is one measure of variability, and it is simply the difference between the highest and lowest score. The standard deviation is generally the more preferred measure because it indicates the average extent to which all scores in a particular distribution vary from the mean.

Correlation coefficients express the degree of relationship between two variables. They range from -1 (a perfect negative correlation) to a +1 (a perfect positive correlation). A 0 correlation indicates no relationship between the variables; the closer a correlation coefficient approaches either -1 or +1, the stronger the relationship between the two variables.

Inferential statistics are used to provide the rules for determining what conclusions can legitimately be drawn from data. They use probability estimates to determine the likelihood that a given set of scores would occur by chance alone. Since chance can never be entirely eliminated, psychologists use tests of statistical significance for this determination. Generally, results are considered statistically significant if the probability of the results occurring by chance is .05 (5 times in 100) or less.

Meta-analysis is a method of synthesizing, or combining and summarizing, the results of different studies by using a set of statistical procedures to compute, select, and combine data.

CONCEPT III: Some Pitfalls in Psychological Research

Many problems can invalidate a study. Self-fulfilling prophecy is the notion that expectations of investigators can influence findings. One method used to alleviate this problem is a double-blind technique, in which neither experimenter nor subjects know who is in the experimental and control groups. In a variation, the single-blind technique, the experimenter knows who is in which group, but the subjects do not.

Demand characteristics are conditions that cause subjects to respond in ways they think will please the experimenter. These characteristics can be reduced by avoiding face-to-face encounters with subjects and by interviewing subjects after the experiment to detect any demand response patterns. Demand characteristics can also be reduced by deceiving the subjects about the real intent of the experiment. In this case, subjects must be debriefed afterward and informed of its real purpose.

The Hawthorne effect is the phenomenon of subjects' awareness of being participants in an experiment affecting their performance.

Replication - that is, repeating an experiment - serves as an essential check in making sure that the results of the experiment are reliable and dependable. Multiple studies help avoid premature conclusions.

CONCEPT IV: The Ethics of Research in Psychology

The American Psychological Association (APA) has issued principles to guide researchers in the ethical conduct of experiments using human subjects. These principles state that harm to a subject should be avoided. If some possibility of harm exists, a subject should be informed prior to the experiment. Subjects' participation must always be voluntary, and if they request, subjects should be allowed to withdraw at any time during an experiment. If deception is used, the experimenter must later explain why it was necessary. Finally, subjects' privacy must be protected.

Other organizations as well insist on ethical treatment; these include the U.S. government and most universities. Also, most public and private institutions protect the rights of animal subjects. Very few complaints have been made against psychologists regarding the ethical treatment of human and animal subjects, indicating the ethics procedures in place are working quite well.