As emphasized throughout this text, the study of development is a science. Social scientists spend years in graduate school, studying methods and statistics. Chapter 1 touches on some of these matters (observation and experiments; correlation and statistical significance; independent and dependent variables; experimental and control groups; cross-sectional, longitudinal, and cross-sequential research), but there is much more. A few additional aspects of research are presented here to help you evaluate research wherever you find it.

The entire group of people about whom a scientist wants to learn is called a population. Generally, a research population is quite large—not usually the world’s entire population of more than 7 billion, but perhaps all the 4 million babies born in the United States last year, or all the 31 million Japanese currently over age 65.

The particular individuals who are studied in a specific research project are called the participants. They are used as a sample of the larger group. Ideally, the participants are a representative sample, that is, a sample that reflects the entire population. Every peer-reviewed, published study reports details on the sample.

Selection of the sample is crucial. People who volunteer, or people who have telephones, or people who have some particular condition are not a random sample; in a random sample, everyone in a particular population is equally likely to be selected. To avoid selection bias, some studies are prospective, beginning with an entire cluster of people (for instance, every baby born on a particular day) and then tracing the development of some particular characteristic.

A-4

For example, prospective studies find the antecedents of heart disease, or child abuse, or high school dropout rates—all of which are much harder to find if the study is retrospective, beginning with those who had heart attacks, experienced abuse, or left school. Thus, although retrospective research finds that most high school dropouts say they disliked school, prospective research finds that some who like school still decide to drop out and then later say they hated school, while others dislike school but stay to graduate. Prospective research discovers how many students are in these last two categories; retrospective research on people who have already dropped out does not.

Every researcher begins not only by formulating a hypothesis but also by learning what other scientists have discovered about the topic in question and what methods might be useful and ethical in designing research. Often they include measures to guard against inadvertently finding only the results they expect. For example, the people who actually gather the data may not know the purpose of the research. Scientists say that these data gatherers are blind to the hypothesized outcome. Participants are sometimes “blind” as well, because otherwise they might, for instance, respond the way they think they should.

Another crucial aspect of research design is to define exactly what is to be studied. Researchers establish an operational definition of whatever phenomenon they will be examining, defining each variable by describing specific, observable behavior. This is essential in quantitative research, but it is also useful in qualitative research. For example, if a researcher wants to know when babies begin to walk, does walking include steps taken while holding on? Is one unsteady step enough? Some parents say yes, but the usual operational definition of walking is “takes at least three steps without holding on.” This operational definition allows comparisons worldwide, making it possible to discover, for example, that well-fed African babies tend to walk earlier than well-fed European babies.

Operational definitions are difficult to formulate, but they are essential when personality traits are studied. How should aggression or sharing or shyness be defined? Lack of an operational definition leads to contradictory results. For instance, critics report that infant day care makes children more aggressive, but advocates report that it makes them less passive. In this case, both may be seeing the same behavior but defining it differently. For any scientist, operational definitions are crucial, and studies usually include descriptions of how they measured attitudes or behavior.

You already know that results should be reported in sufficient detail so that another scientist can analyze the conclusions and replicate the research. Various methods, populations, and research designs may produce divergent conclusions. For that reason, handbooks, some journals, and some articles are called reviews: They summarize past research. Often, when studies are similar in operational definitions and methods, the review is a meta-analysis, which combines the findings of many studies to present an overall conclusion.

A-5

Table 1.2 (p. 22) describes some statistical measures. One of them is statistical significance, which indicates whether or not a particular result could have occurred by chance.

A crucial statistic is effect size, a way of measuring how much impact one variable has on another. Effect size ranges from 0 (no effect) to 1 (total transformation, never found in actual studies). Effect size may be particularly important when the sample size is large, because a large sample often leads to highly “significant” results (results that are unlikely to have occurred by chance) that have only a tiny effect on the variable of interest.

Hundreds of statistical measures are used by developmentalists. Often the same data can be presented in many ways: Some scientists examine statistical analysis intently before they accept conclusions as valid. A specific example involved methods to improve students’ writing ability between grades 4 and 12. A meta-analysis found that many methods of writing instruction have a significant impact, but effect size is much larger for some methods (teaching strategies and summarizing) than for others (prewriting exercises and studying models). For teachers, this statistic is crucial, for they want to know what has a big effect, not merely what is better than chance (significant).

Numerous articles published in the past decade are meta-analyses that combine similar studies to search for general trends. Often effect sizes are also reported, which is especially helpful for meta-analyses since standard calculations almost always find some significance if the number of participants is in the thousands.

A-6