Research Methods in Psychology

Descriptive Statistics

Find meaningful patterns and summaries in large sets of data Measures of Central Tendency Descriptive statistics, such as these frequency distributions of SAT mathematics scores, are one way of viewing the numeric results of a research study. These statistics give researchers a broad snapshot of the data. Typically, they're used to summarize the results. Descriptive statistics can also illustrate trends or patterns within the data. Researchers must use additional methods, however, to understand and interpret the data. Among the most frequently used descriptive statistics are measures of central tendency This gives researchers an idea of how the typical participant performs. The 3 most popular measures of central tendency are: Mean, Median, and Mode. Source: Adapted from "2009 College-Bound Seniors: Total Group Profile Report," by The College Board, 2009, retrieved from http://professionals.collegeboard.com/profdownload/cbs-2009-national-TOTAL-GROUP.pdf.

The longitudinal approach

has few logical drawbacks, but it is expensive and time-consuming to do. Participants drop out of the study due to moves or lack of incentive. Researchers then must worry about whether those who remain in the study still comprise a representative sample.

Nonexperimental Methods: Case Studies

A case study provides an in-depth analysis of the behavior of one person or a small number of people. Many fields, including medicine, law, and business, use the case study method. Psychologists often use case studies in situations where large numbers of participants are not available or when a particular participant possesses unique characteristics, like a particular disease or specific injury. Interviews, background records, observation, personality tests, cognitive tests, and brain imaging provide information necessary to evaluate the case. Consider the famous case of Henry Molaison, known for much of his life as patient H.M. to protect his identity (Scoville & Milner, 1957). In 1953, as a twenty-seven-year-old man, his epilepsy was so severe that he had brain surgery in an effort to stop his seizures. The surgery removed parts of both his left and right temporal lobes, including parts of the hippocampus on both sides. The surgery removed parts of both his left and right temporal lobes, including parts of the hippocampus on both sides. This radical treatment helped control his seizures, but it left H.M. with a dense amnestic syndrome that persisted until he died at age of eighty-two in 2008. H.M. was carefully studied for years, and the lessons learned from him prompted research on memory loss in other people after medial temporal lobe damage.

An example experiment: Does Listening to Music While Studying Affect Learning?

A good experimental design features random assignment of participants to groups, appropriate control groups, control of situational variables, and carefully selected independent and dependent variables. Let's examine a simple experiment. Suppose you notice that you seem to study better while listening to your iPod. This suggests the hypothesis that listening to music improves learning. We could test this idea by forming an experimental group that studies with music. A control group would study without music. Then we could compare their scores on a test. In this experiment, the amount learned (indicated by scores on the test) is the dependent variable. We are asking, Does the independent variable affect the dependent variable? (Does listening to music affect or influence learning?)

The Mean

A numeric average of a data set. Can be skewed by extreme values The mean represents the numeric average of a data set. You take the total scores of everyone and divide by the number of scores you have. If I want the mean score of the first exam, I would add up every student's score and divide by the number of students I have. One problem with the mean is that it can be skewed by extreme values. If the data contain a particularly high (or low) score, the mean may not represent the typical participant. Imagine that you're looking at the net worth of the faculty at this college/university. You'll probably find some professors doing well (maybe very well), while others are at the start of their careers. Now, imagine that Bill Gates, who is worth roughly $80 Billion, joins the faculty. The mean net worth will skyrocket, but it probably won't be very representative of what's typical.

Selecting Participants for a Research Study

An important part of deciding what you are going to research is deciding the group you will be studying. In a perfect world, researchers would include every person they are interested in studying. This is termed the population of interest. For example, for a developmental psychologist who specializes in infant development, all infants would be the population of interest. It is impossible to test everyone, however, so researchers select a portion, or subset, of the population of interest called a sample. Because the sample will be used to make inferences or judgments about the entire population, the sample should reflect the whole population as much as possible; that is, it should be a representative sample. Random sampling of participants ensures a representative sample. In a random sample, every member of the population has an equal chance of being selected to participate in the study; this avoids introducing sampling bias into the research. The more representative the sample is, the more the results will generalize (or apply) to the population of interest. But random sampling is not always possible. Instead, psychological research often uses samples of convenience, or groups of people who are easily accessible to the researcher. The students in your psychology course are a sample of convenience.

Conducting Animal Research

Can be controversial APA guidelines for what kind of research is permissible The topic of using animals in research is guaranteed to stimulate lively, and possibly heated, discussion. Some people are adamantly opposed to animal research of any kind, whereas others accept the concept of using animals as long as certain conditions are met. Currently, about 7 to 8% of published research in psychology journals involves the use of animals as subjects. Ninety percent of the animals used are rodents and birds, with 5% or fewer studies involving monkeys and other primates. According to the American Psychological Association (APA), the use of dogs and cats in psychological research is rare. Research using animals must demonstrate a clear purpose, such as benefiting the health of humans or other animals. In addition to serving a clear purpose, animal research requires excellent housing, food, and veterinary care. The most controversial ethical standards relate to minimizing the pain and suffering experienced by animal research subjects. The American Psychological Association provides guidelines for the use of pain, surgery, stress, and deprivation with animal subjects, as well as the termination of an animal's life. The standards approximate the community standards we would expect from local humane societies tasked with euthanizing animals that are not adopted.

Another Kind of Research Ethics Violation: Plagiarism

Cite your research references! To present someone else's ideas or words as your own is to commit plagiarism. Plagiarism, like fraud, is a serious breach of ethics. Reference citations (giving others credit when credit is due) must be included in your paper whenever someone else's ideas or work has influenced your thinking and writing. Whenever you use direct quotations or even paraphrase someone else's work, you need to give that person credit.

A Meta-Analysis of Video Game Violence and Aggression

Conducting a meta-analysis, or a statistical analysis of many previous experiments on the same topic, often provides a clearer picture than single experiments observed in isolation. In the example shown, combining the findings of over 300 studies representing more than 50,000 participants, Anderson and Bushman (2002) argue that a positive relationship exists between exposure to video game violence and aggression, aggressive cognitions, aggressive affect or mood, and arousal. Video game violence was negatively correlated with helping behaviors. This type of graph is known as a boxplot. The width of each box corresponds to the number of studies of each type. Twenty-five percent of the results fall below the bottom of the box and another 25% are above the top of each box. The line in the middle of a box shows the median, the point where half of the data are above and half below. Source: Adapted from Anderson and Bushman (2001).

Problems in Experimental Research

Confounding (extraneous) variables Experimenter bias When conducting experiments, researchers must be mindful of problems that can produce inaccurate results. Confounding variables (sometimes called extraneous variables) are uncontrolled factors that influence the results of an experiment. Individual differences among participants are an example of confounding variables. Random assignment to groups typically controls for confounds due to these types of individual differences, but other sources of confound exist. Situational confounds, such as time of day or noise levels in a laboratory, could also affect the interpretation of an experiment. Scientists attempt to run their experiments under the most constant circumstances possible to rule out situational confounding variables. Another hazard is experimenter bias, the unintentional effect that researchers may exert on their results. To prevent experimenter bias from influencing results, experimenters may use a double-blind design. In this arrangement, both the research participants and those giving the treatments are unaware of ("blind") who gets the placebo. Only researchers who have no direct contact with participants have this information, and they do not reveal it until the experiment is over. Double-blind testing has shown that at least 50 percent of the effectiveness of antidepressant drugs, such as the wonder drug Prozac, is due to the placebo effect (Kirsch & Sapirstein, 1998; Rihmer et al., 2012). Much of the popularity of herbal health remedies also is based on the placebo effect (Seidman, 2001).

Studying the Effects of Time

Cross-sectional design Longitudinal design Mixed longitudinal design

Not all observations are scientific. How does science differ from everyday observations, like the belief that "opposites attract"?

First, science relies on objectivity rather than subjectivity. Scientists strive to be objective, but any observation by a human being is, by definition, subjective. Recognizing when you are being subjective can be difficult, so scientists cannot rely on their own introspections to maintain objectivity. Scientific methods promote objectivity and help prevent biased, subjective observations from distorting a scientist's work. The second important difference between science and everyday observations is the use of systematic observation as opposed to hit-or-miss observation. By "hit or miss," we mean making conclusions based only on whatever is happening around us. Finally, science relies on observable, repeatable evidence, whereas everyday observation often ignores evidence, especially when it runs counter to strongly held beliefs. Scientific knowledge is both stable and changing. The fact that we may learn something new tomorrow should not convince you that today's knowledge is flawed. Most change occurs very slowly on the "cutting edges" of science, not quickly or at the main core of its knowledge base. An important feature of scientific literacy is to learn to be comfortable with the idea that scientific knowledge is always open to improvement and will never be considered absolutely certain. The skilled critical thinker has learned to follow logical arguments, identify mistakes in reasoning, prioritize ideas according to their importance, and apply logic to personal attitudes, beliefs, and values. Critical thinking is grounded in a skeptical viewpoint. You can begin by using five critical thinking questions (as illustrated above) to evaluate new information you come across in your everyday life, including in this class. It is also helpful to recognize the signs that you are not thinking critically: Instead of figuring out the answer to a problem yourself, you prefer just being given the answer. You prefer to use "gut feelings" about decisions instead of using reason to choose a solution. You do not review your mistakes or change your mind if evidence contradicts your original position. You resent criticism of your ideas

Nonexperimental Methods: Naturalistic Observation

If you are interested in learning about larger groups of people than are possible with the case study method, you might pursue a naturalistic observation, or an in-depth study of a phenomenon in its natural setting. The researcher looks at a much larger group of people than in a case study (next slide), which strengthens the ability to apply results to the population in general. We also have the advantage of observing individuals in their natural, everyday circumstances. With the naturalistic observation method, researchers in the field can examine behavior as it unfolds, but they run the risk of influencing the behavior they are observing.

Designing an Experimental Study

Manipulate one variable and observe changes in others Independent variable: the cause Dependent variable(s): the effect The scientist's most powerful tool for drawing conclusions about research questions is the formal experiment. Unlike cases in which descriptive methods are used, the researcher conducting an experiment has a great deal of control over the situation. Unlike correlational methods, the use of the formal experiment allows us to talk about "cause" A researcher begins designing an experiment with a hypothesis, which can be viewed as a highly educated guess based on systematic observations, a review of previous research, or a scientific theory. To test the hypothesis, the researcher manipulates or modifies one or more variables and observes changes in others. The variable controlled and manipulated by an experimenter ("If I do this . . . .) is known as the independent variable. We need some way to evaluate the effects of this manipulation. We use a dependent variable, defined as the observed result of the manipulation of the independent variable, to tell us "that will happen" as a result of the independent variable. Like the independent variable, our choice of dependent variable is based on our original hypothesis.

Designing a Correlational Study

Measure the direction and strength of the relationship between two variables, or factors Correlational research measure the direction and strength of the relationship between two variables, or factors that have values that can "vary," (have a range of values) like a person's height and weight. We begin our analysis of correlations by measuring our variables. A measure answers the simple question of "how much" of a variable you have observed. After we obtain measures of each variable, we compare the values of one variable to those of the other and conduct a statistical analysis of the results using a correlation coefficient. Three possible outcomes from the comparison between our two variables can occur: positive, negative, or zero correlations. In a positive correlation, high levels of one variable are associated with high levels of the other variable. Height and weight usually show this type of relationship. In most cases, people who are taller weigh more than people who are shorter. Two variables can also show a negative correlation, in which high values of one variable are associated with low values of another. For example, high levels of alcohol consumption among college students are usually associated with low grade point averages. The third possible outcome is a zero correlation, in which the two variables do not have any systematic relationship with each other at all. When variables have a zero correlation, knowing the value of one variable does not tell you anything about the value of the other variable

Standard Deviation

Measures how tightly clustered a group of scores is around the mean We might be interested in identifying the "average" score on our measures, or the central tendency of our data set. There are three types of measures for central tendency: the mean, median, and mode for each group of scores: the mode, the median, and the mean. The mean is the numerical average of a set of scores, computed by adding all scores together and dividing by the number of scores. The traditional way to look at the variability of scores is to use a measure known as the standard deviation. This measure tells you how tightly clustered a group of scores is around the mean. Many measures of interest to psychologists, such as scores on intelligence tests appear to form a normal distribution. The ideal normal curve in this illustration has several important features. One, it is symmetrical. Equal numbers of scores should occur above and below the mean. Second, its shape indicates that most of the scores occur near the mean, which is where our measure of variability plays a role. In the standard normal curve, shown in (a), 68% of the population falls within one standard deviation of the mean, 95% falls within two standard deviations, and 99% of the population falls within three standard deviations.

What makes scientific thinking different than everyday observation?

Objectivity rather than subjectivity Systematic observation and repeatable evidence

The third approach, the mixed longitudinal design, combines the cross-sectional and longitudinal methods.

Participants from a range of ages are observed for a limited period of time, usually about five years. This approach is faster and less expensive than the longitudinal method and avoids some of the cohort effects of the pure cross-sectional method. Longitudinal designs control for the cohort effects that are often seen in cross-sectional designs. The longitudinal study above shows that verbal ability and verbal memory are fairly stable over the lifetime, but that perceptual speed gradually worsens with age. Source: Adapted from Schaie (1996).

Ethically Questionable Research: The Tuskegee Syphilis Experiments

Prisoners, soldiers, and mental patients were deliberately exposed to syphilis and gonorrhea to test the effectiveness of penicillin One of the most egregious examples of unethical research was the Tuskegee syphilis experiment, which lasted from 1932 until 1972. Researchers from the U.S. Public Health Service recruited about 400 impoverished African American men who had contracted syphilis to study the progression of the disease. None of the men was told he had syphilis, and none was treated, even though penicillin became the standard treatment for syphilis in 1947. While examining the papers of Dr. John Cutler, who led the Tuskegee syphilis study, Wellesley historian Susan Reverby discovered that during the 1940s, U.S. and Guatemalan health officials had deliberately exposed prisoners, soldiers, and mental patients to syphilis and gonorrhea to test the effectiveness of penicillin.

Assigning Participant Groups in a Study

Randomly assign participants to: Experimental groups, who experience the independent variable Control groups, who do NOT experience the independent variable After determining our independent and dependent variables, we still have quite a bit of work to do. In most experiments, we want to know how simply going through the procedures of being in an experiment influences our dependent variable. Perhaps the hassle of going to a laboratory and filling out paperwork changes our behavior. To evaluate these effects, we assign some of our participants to a control group, or a group that experiences all experimental procedures with the exception of exposure to the independent variable. The experience of the control group should be as similar as possible to that of the experimental groups, who do experience the independent variable. A special type of control group is a placebo control group. A placebo is a treatment that contains nothing known to be helpful but that still produces benefits because the person receiving the treatment believes it will be beneficial (the placebo effect) We want to ensure that our dependent variables reflect the outcomes of our independent variables, instead of individual differences among the participants' personalities, abilities, motivations, and other similar factors. To prevent these individual differences from masking or distorting the effects of our independent variable, we randomly assign participants to experimental or control groups. Random assignment means that each participant has an equal chance of being assigned to any group in an experiment. With random assignment, any differences we see between the behavior of one group and that of another is unlikely to be the result of the individual differences among the participants, which tend to cancel each other out. Individual differences among participants are an example of confounding variables, or variables that are irrelevant to the hypothesis being tested that can alter our conclusions. Random assignment to groups typically controls for confounds due to these types of individual differences, but other sources of confound exist. Situational confounds, such as time of day or noise levels in a laboratory, could also affect the interpretation of an experiment. Scientists attempt to run their experiments under the most constant circumstances possible to rule out situational confounding variables. Another potential confound is experimenter bias, the unintentional effect that researchers may exert on their results. To prevent experimenter bias from influencing results, experimenters may use a double-blind design. In this arrangement, both the research participants and those giving the treatments are unaware of ("blind") who gets the placebo. Only researchers who have no direct contact with participants have this information, and they do not reveal it until the experiment is over. Double-blind testing has shown that at least 50 percent of the effectiveness of antidepressant drugs, such as the wonder drug Prozac, is due to the placebo effect (Kirsch & Sapirstein, 1998; Rihmer et al., 2012). Much of the popularity of herbal health remedies also is based on the placebo effect (Seidman, 2001).

Inferential Statistics

Reach conclusions about data Inferential statistics allow us to decide whether the observed differences between the performance of males and females on the math SAT represents a real gender difference or is just due to chance. Although we can learn a great deal from our descriptive statistics, most of research features the use of inferential statistics, so called because they permit us to draw inferences or conclusions from data. The descriptive statistics described earlier allow us to talk about our sample data, but do not allow us to decide what our sample data might mean more generally. To reach conclusions about how our observations fit the big picture, we use inferential statistics. How do we know when a hypothesis should be rejected? Like most sciences, psychology has accepted odds of 5 out of 100 that an observed result is due to chance as an acceptable standard for statistical significance. We can assess the likelihood of observing a result due to chance by repeating a study, like throwing dice multiple times.

Issues in Measurement

Reliability: consistency Validity: accuracy Reliability is an important aspect of measurement quality. In its everyday sense, reliability is the consistency or stability of an observation. While reliability is necessary for ensuring the quality of any measurement, it alone is not sufficient. Validity is the other important dimension that is crucial for any kind of meaningful measurement, in particular, and research, in general. Validity deals with accuracy or precision of measurement. We often think of reliability and validity as separate ideas but, in fact, they're intimately interconnected. A favorite metaphor for the relationship between reliability and validity is that of a target. Think of the center of the target as the concept or construct you are trying to measure. Imagine that for each person you are measuring, you are taking a shot at the target. If you measure the concept perfectly for a person, you are hitting the center of the target. If you don't, you are missing the center. The more "off" you are for that person, the further you are from the center. The figure shows three possible situations. In the first one, you are hitting the target consistently, but you are missing the center of the target. That is, you are consistently and systematically measuring the wrong value for all respondents. This measure is reliable, but not valid. (It's consistent but wrong.) The second scenario shows a case where your hits are spread across the target and you are consistently missing the center. Your measure in this case is neither reliable nor valid. The last figure shows the Robin Hood or William Tell scenario; you consistently hit the center of the target. Your measure is both reliable and valid.

Conducting Ethical Research

Researchers working in universities and other agencies receiving federal funding must receive the approval of institutional review boards (IRBs) for human participant research and institutional animal care and use committees (IACUCs) before conducting research. The IRBs and IACUCs are guided by federal regulations and research ethics endorsed by professional societies such as the American Psychological Association, the Association for Psychological Science, and the Society for Neuroscience. At the core of ethical standards for human research is the idea that participation is voluntary. No participant should be coerced into participating. Although psychologists are well aware that people who volunteer to participate in research are probably quite different in important ways from those who don't volunteer, we have chosen to give research ethics a higher priority than our ability to generalize research results. Scientists have a duty to obtain informed consent from all participants. In doing so, researchers must briefly describe the goals of the project, the potential risks and/or benefits to the participants, the procedures for maintaining confidentiality, and the incentives or payments offered for participation. Only after these items have been communicated, can participants agree to involvement in the study. Research should also be conducted in a manner that does no irreversible harm to participants. Most cases of deception are quite mild, as when participants are told that a study is about memory when it is actually a study of some social behavior. When researchers must deceive their participants, extra care must be taken to debrief participants and answer all their questions following the experiment. Research using human participants should be rigorously private and confidential. Privacy refers to the participants' control over the sharing of their personal information with others, and methods for ensuring privacy are usually stated in the informed consent. For example, some studies involve the use of medical records, which participants' agree to share with the researchers for the purpose of the experiment. Confidentiality refers to the participants' rights to not have their data revealed to others without their permission. Confidentiality is usually maintained by such practices as substituting codes for names and storing data in locked cabinets.

Nonexperimental Methods: Surveys

Surveys or questionnaires allow you to ask large numbers of people questions about attitudes and behavior. Surveys provide a great deal of useful information very quickly at relatively little expense. Surveys can be conducted in several different ways (telephone, internet, paper-and-pencil forms) It's important to be aware of limitations that arise from the method used. For instance, if you conduct a telephone survey, are you excluding people that don't have landlines? One of the primary requirements for a good survey is the use of an appropriate sample, or subset of a population being studied. Good results require large samples that are typical, or representative, of the population you wish to describe. Surveys use self-report, so results can be influenced by people's natural tendency to want to appear socially appropriate.

Measuring the Correlation

The correlation coefficient, How is the degree of correlation expressed? The strength and direction of a relationship can be expressed as a coefficient of correlation. This can be calculated as a number falling somewhere between +1.00 and −1.00. Drawing graphs of relationships can also help clarify their nature. If the number is zero or close to zero, the association between two measures is weak or nonexistent. If the correlation is +1.00, a perfect positive relationship exists. If it is −1.00, a perfect negative relationship has been discovered. Note that the strength of the correlation is NOT influenced by its sign. A correlation of −.67 is stronger than a correlation of +.32 These graphs show a range of relationships between two measures, X and Y. If a correlation is negative (a), increases in one measure are associated with decreases in the other. (As Y gets larger, X gets smaller.) In a positive correlation (e), increases in one measure are associated with increases in the other. (As Y gets larger, X gets larger.) The center-left graph (b, moderate negative relationship) might result from comparing time spent playing computer games (Y) with grades (X): More time spent playing computer games is associated with lower grades. The center graph (c, no relationship) would result from plotting a person's shoe size (Y) and his or her IQ (X). The center-right graph (d, moderate positive relationship) could be a plot of grades in high school (Y) and grades in college (X) for a group of students: Higher grades in high school are associated with higher grades in college.

The Median

The halfway mark in a set of data, with half of the scores above and half below The median represents a halfway mark in the data set, with half of the scores above and half below. The median is far less affected by extreme scores, or outliers, than the mean. In many cases, like the SAT data, means and medians are very close together. However, in other cases, like reported lifetime sex partners, these two measures of central tendency provide very different pictures. The average number of sex partners for males is 20, but half of all men report having had 8 or fewer partners. This suggests that the upper half of males have a very large number of partners indeed. Source: Adapted from ABC News: Primetime (2001).

The Mode

The most frequently occurring score in a set of data The mode in a data set refers to the score that occurs most frequently and is easy to determine from looking at a histogram. The average age of onset for the eating disorder anorexia nervosa is 17 years, but this measure masks the important fact that age of onset shows two modes— one at 14 years and the second at 18 years. For public health officials wishing to target vulnerable groups for preventive education, the modes provide better information than the mean. Source: Adapted from Halmi (1979).

The Scientific method

The purpose of psychological research is to test ideas about behavior. Researchers use the scientific method when testing ideas about behavior. The scientific method is a set of rules for gathering and analyzing information that enables you to test an idea or hypothesis. The decisions that scientists make at each step of the scientific method will ultimately affect the types of conclusions they can draw about behavior. Step 1: Observe Behavior or Other Phenomena The scientific method often begins with casual or informal observations. Simply observe the world around you until some behavior or event catches your attention. Step 2: Formulate a Research Question This step in the process usually begins by identifying other factors, or variables, that are associated with your observation. Use those other variables to generate explanations for the behavior that is of interest to you. Step 3: Generate a Testable Prediction (Hypothesis) Once you have a research question, you must develop a way to answer it. Notice that a single hypothesis can lead to several different predictions and that each prediction refers to a specific situation or an event that can be observed and measured. Also notice that the predictions generated from a hypothesis must be testable—that is, it must be possible to demonstrate that the prediction is either correct or incorrect by direct observation. Either the observations will provide support for the hypothesis or they will refute the hypothesis. For a prediction to be truly testable, both outcomes must be possible. Step 4: Collect and Analyze Data After a specific, testable prediction has been made, the next step is to evaluate the prediction using direct observation. The goal is to provide a fair and unbiased test of the research hypothesis by observing whether the prediction is correct. The researcher must be careful to observe and record exactly what happens, free of any subjective interpretation or personal expectations. A good research study will have safeguards in place to minimize the likelihood of bias. Step 5: Draw Conclusions and Create Theories The final step of the scientific method is to compare the actual observations with the predictions that were made from the hypothesis. To what extent do the observations agree with the predictions? Some agreement indicates support for the original hypothesis, and suggests that you consider making new predictions and testing them. Lack of agreement indicates that the original hypothesis was wrong or that the hypothesis was used incorrectly, producing faulty predictions. In this case, you might want to revise the hypothesis or reconsider how it was used to generate predictions. In either case, this information is used in creating new theories or refining existing ones. Researchers attempt to place their findings within a larger context of knowledge. Once completed, researchers often circle back to step 2; that is, they form a new hypothesis that further explores their theory.

EEG

The scientific study of dreaming was made possible by use of the EEG, a device that records the tiny electrical signals the brain generates as a person sleeps. The EEG converts these electrical signals into a written record of brain activity. Certain shifts in brain activity, coupled with the presence of rapid eye movements, are strongly related to dreaming.

Hypothesis Proposed explanation for a situation: "if A happens then B will be the result" Theory A set of facts and relationships between facts that can explain and predict related phenomena

Theory building begins with generating hypotheses that are then systematically tested. Hypotheses that are not rejected contribute to the theory and help generate new hypotheses. Science seeks to develop theories, which are sets of facts and relationships between facts that can be used to explain and predict phenomena. In other words, scientists construct the best possible models of reality based on the facts known to date. Confusion over the multiple meanings of the word theory have led people mistakenly to view truly scientific theories, like the theory of evolution, as nothing more than casual guesses or hunches rather than the thoroughly investigated and massively supported principles that they are. The best scientific theories not only explain and organize known facts, but they also generate new predictions. A scientific prediction is much more than a guess or hunch. Before attempting to generate your own scientific questions, it pays to become very familiar with relevant theories and previous discoveries. As Sir Isaac Newton noted, scholars stand on the shoulders of giants—we build on the work of those who came before us. Scientific progress often takes a giant leap forward when a gifted observer recognizes a deeper meaningfulness in an everyday occurrence. Based on your understanding of past work and theoretical foundations in your area of interest, coupled with your own observations, you can now generate a hypothesis, which is a type of inference, or, in other words, an educated guess, based on prior evidence and logical possibilities. Scientific hypotheses must be both falsifiable and testable. Falsifiable does not mean "false." Instead, falsifiable means that you can imagine situations that demonstrate your hypothesis to be false. For example, a hypothesis claiming that all planets outside our solar system are uninhabited is falsifiable, because finding an inhabited planet outside our solar system would show that your hypothesis was wrong. Hypotheses can be falsifiable but not necessarily testable, which means you can evaluate the hypothesis using known scientific methods. Scientists can never "prove" that their hypotheses are true, because some future experiment, possibly using new technology not currently available, might show the hypothesis to be false after all. All we can do is show when a hypothesis is false. A false hypothesis must always be modified or discarded.

Understanding Causation-The third variable problem

Third variables can be responsible for the correlation we observe in two other variables. In an example of video game violence and aggression, being bullied could be a third variable that predicts both choice of violent games and a tendency to be aggressive at school. The possibility of third variables is one reason we must be very careful when we reach conclusions based on correlational data. Correlational research results are frequently misunderstood. Correlations permit us to discuss the relationships between two variables but tell us nothing about whether one variable causes changes in the other. Let us say that we discover a positive correlation between violent video games and aggression: Youth who play the most hours of violent video games have the most reports of physical aggression at school. However, we still cannot say that playing violent video games causes physical aggression at school. This conclusion may seem very reasonable to you, and it may actually be true, so why must we abandon it? First, the two variables in a correlation can influence each other simultaneously. Although it may be true that playing violent video games leads to physical aggression at school, youth who experience physical aggression at school may be more attracted to violent video games as an outlet for their frustration. Second, we might be observing a situation in which a third variable is responsible for the correlation we see between our two variables of interest. Consider the observation that many school shootings have been perpetrated by people who had been bullied relentlessly by others. Perhaps the experience of having been bullied (the third variable in this case) predisposes both a choice of violent recreation and a tendency to engage in aggressive behavior at school

Behavioral Statistics

This section covers: Issues in measurement Descriptive statistics Inferential statistics Some psychologists specialize in administering, scoring, and interpreting psychological tests, such as tests of intelligence, creativity, personality, or aptitude. This specialty, which is called psychometrics, is an example of using psychology to predict future behavior. How do psychologists draw conclusions from data?

To do a cross-sectional study,

we might gather groups of people of varying ages and assess both their exposure to violent video games and their levels of physical aggression. We might be able to plot a developmental course for age-related differences in both video game exposure and aggressive behavior. However, the cross-sectional method introduces what we refer to as cohort effects, or the generational effects of having been born at a particular point in history. Being 20 years old in 1950 was very different from being 20 years old in 1980 or in 2010, due to a variety of cultural influences. A method that lessens this dilemma is the longitudinal study, in which a group of individuals is observed for a long period.