An experiment is a procedure that uses control to determine causal relationships among variables. Specifically, one or more variables thought to influence a behavior of interest are changed, or manipulated, while other variables are held constant. Researchers can then examine how the changing variable influences the behavior under study. If the behavior changes as the variable changes, this suggests that the variable caused the change in the behavior.
For example, Gentile, Bender, and Anderson (2017) examined the effect of playing violent video games on children’s physiological stress and aggressive thoughts. Children were randomly assigned to play a violent video game (Superman) or a nonviolent video game (Finding Nemo) for 25 minutes in the researchers’ lab. The researchers measured physiological stress as indicated by heart rate and cortisol levels before and after the children played the video game. Children also completed a word completion task that the researchers used to measure the frequency of aggressive thoughts. Gentile et al. (2017) found that children who played violent video games showed higher levels of physiological stress and aggressive thoughts than did the children who played nonviolent video games. The researchers concluded that the type of video game changed children’s stress reactions and aggressive thoughts.
Let’s take a closer look at the components of an experiment. Conducting an experiment requires choosing at least one dependent variable, the behavior under study (e.g., physiological stress—heart rate and cortisol—and aggressive thoughts), and one independent variable, the factor proposed to change the behavior under study (e.g., type of video game). The independent variable is manipulated or varied systematically by the researcher during the experiment (e.g., a child plays with a violent or a nonviolent video game). The dependent variable is expected to change as a result of varying the independent variable, and how it changes is thought to depend on how the independent variable is manipulated (e.g., physiological stress and aggressive thoughts vary in response to the type of video game).
In an experiment, the independent variable is administered to one or more experimental groups, or test groups. The control group is treated just like the experimental group except that it is not exposed to the independent variable. For example, in an experiment investigating whether particular types of music influence mood, the experimental group would experience a change in music (e.g., from “easy listening” to rock), whereas the control group would hear only one type of music (e.g., “easy listening”). Random assignment, whereby each participant has an equal chance of being assigned to the experimental or control group, is essential for ensuring that the groups are as equal as possible in all preexisting characteristics (e.g., age, ethnicity, and biological sex). Random assignment makes it less likely that any observed differences in the outcomes of the experimental and control groups are due to preexisting differences between the groups. After the independent variable is manipulated, if the experimental and control groups differ on the dependent variable, it is concluded that the independent variable caused the change in the dependent variable. That is, a cause-and-effect relationship has been demonstrated.
As another example, consider a study designed to examine whether massage therapy improves outcomes in preterm infants (infants who were born well before their due date) (Abdallah, Badr, & Hawwari, 2013). Infants housed in a neonatal unit were assigned to a massage group (independent variable), who were touched and their arms and legs moved for 10-minute periods once each day, or to a control group, which received no massage. Other than the massage/no-massage periods, the two groups of infants were cared for in the same way. Infants who were massaged scored lower on the measure of infant pain and discomfort (including indicators such as heart rate, oxygen saturation, and facial responses) at discharge (dependent variable). The researchers concluded that massage therapy reduces pain responses in preterm infants.
By experimentally manipulating which infants receive massage therapy, researchers determined that massage can help preterm infants gain weight.
AP Photo / AL GOLDIS
Developmental scientists conduct studies that use both correlational and experimental research. Studying development, however, requires that scientists pay close attention to age and how people change over time, which requires the use of specialized research designs, as described in the following sections.
Developmental Research Designs
Do children outgrow shyness? Are infants’ bonds with their parents associated with their peer relationships in adolescence? These challenging questions require that developmental scientists examine relationships among variables over time. There are several approaches to examining developmental change.
Cross-Sectional Research Design
A cross-sectional research study compares groups of children of different ages at a single point in time. For example, to examine how vocabulary improves in elementary school, a researcher might measure the vocabulary size of children in first, third, fifth, and seventh grades. The resulting comparison describes how the vocabulary of first-grade children differs from older children in Grades 3, 5, and 7. However, the results do not tell us whether the observed age differences in vocabulary reflect age-related or developmental change. In other words, we don’t know whether the first graders will show the same pattern of vocabulary ability and use as the seventh graders, 6 years from now, when they are in seventh grade.
Cross-sectional research permits age comparisons, but participants differ not only in age but also in cohort, limiting the conclusions researchers can draw about development. A cohort is a group of people of the same age who are exposed to similar historical events and cultural and societal influences. Although the first-grade and seventh-grade children may attend the same school, they are different ages and different cohorts and thus may have different experiences. For example, suppose the elementary school changed the language curriculum, leading the first-grade children to be taught a new, improved curriculum, whereas the seventh graders received the old curriculum. The first graders and seventh graders therefore have different experiences because they were taught different curricula. Any differences in vocabulary may be due to age but also to different experiences. Therefore, cross-sectional research is an important source of information about age differences (how the first graders differ from seventh graders), but it cannot provide information about age change (whether the first graders will show similar development as the seventh graders).
Longitudinal Research Design
A longitudinal research study follows the same group of participants over many points in time. Returning to the previous example, to examine how vocabulary changes between Grades 1 and 7, a developmental scientist using longitudinal research would measure children’s vocabulary size in first grade, then follow up 2 years later in third grade, then 2 years later in fifth grade, and finally 2 years later in seventh grade. This longitudinal study would take 6 years to complete.
Longitudinal research provides information about age change because it follows people over time, enabling scientists to describe how the first graders’ vocabulary progressed through childhood. However, longitudinal research studies only one cohort or age-group over time. Are the findings due to developmental change or are they specific to the children studied? Is the pattern of change experienced by these children over a 6-year span similar to other cohorts or groups of children? Because only one cohort is assessed, it is not possible to determine whether the observed changes are age related or unique to the cohort examined.
Cross-Sequential Research Design
A cross-sequential research study combines the best features of cross-sectional and longitudinal research by assessing multiple cohorts over time, enabling scientists to make comparisons that disentangle the effects of cohort and age (see Table 1.5). Consider the vocabulary study of children in Grades 1 through 7 once more. A cross-sequential design would begin by measuring