Naturalistic Observation

J Mark Eddy. Encyclopedia of Educational Psychology. Edited: Neil J Salkind & Kristin Rasmussen. Volume 2, Sage Publications, 2008.

Naturalistic observation refers to the scientific study of the social interactions of children and adults within their native environments, such as in school or at home, or within artificial situations designed to elicit specific behaviors of interest that occur in native environments, such as child social and problemsolving skills with peers or parents. In educational psychology, the phrase is most commonly used to describe studies in which researchers or practitioners attempt to observe subjects of interest as unobtrusively and impartially as possible. Observation techniques were developed throughout the 20th century but came into prominence with the rise of behavior therapy and behavioral assessment in the 1960s. Naturalistic observation is particularly useful for describing the social, physical, and temporal contexts within which educational activities occur. Observation data can be used to address a wide variety of basic and applied research questions in relevant areas such as learning, instruction, assessment, evaluation, diagnosis, and intervention.

An early proponent of naturalistic observation was Arnold Gesell, a student of founding American psychologist G. Stanley Hall, who developed coding systems to record the behavior of infants during the 1920s. Gesell employed a variety of techniques that would become commonplace in observational studies, including the use of one-way screens to minimize observer influence and motion picture cameras to capture observations for later study. Florence Goodenough developed a short-sample technique, whereby children were observed for brief periods of time and the presence or absence of specific behaviors were noted. Over the next decade, a variety of other psychologists observed children in nursery schools and other commonplace settings and used coding systems to record social interactions.

Henry Murray employed observational techniques with adults during the late 1930s, which he then applied to personnel selection for the U.S. Office of Strategic Services during World War II. His work included the early use of role-playing as a way to sample behavior during normally occurring social situations. During this same time period, Kurt Lewin favored observation techniques over psychological tests and played a key role in popularizing the use of observational techniques. Roger Barker worked with Lewin in studies of frustration in children, and with colleagues later coded the moment-to-moment behaviors of a child documented in One Boy’s Day. Another seminal figure in naturalistic observation was Robert Bales, who developed a method for categorizing behaviors documented in the influential book Interaction Process Analysis.

The rise of behavior therapy during the 1950s and early 1960s brought observational techniques into widespread use in psychology. Operant behavioral procedures pioneered by psychologist B. F. Skinner and colleagues had particular influence in educational psychology. In early studies, observational measures of child behaviors were employed to monitor the effectiveness of various techniques to change behaviors within the school context. Two key research groups that were among the first to employ behavioral observational techniques extensively were those of Sidney Bijou and Gerald R. Patterson. During the late 1960s, Patterson, John B. Reid, and colleagues spearheaded the development of sophisticated behavioral observation coding systems to study child aggression and other antisocial behaviors within family and classroom settings. Since this period, a wide variety of coding systems have been developed by educational, clinical, and developmental psychologists to monitor various aspects of social interactions within school, home, clinic, and laboratory settings.

The hallmark of naturalistic observation is the accurate recording of specific behaviors at the time they occur within a real-life setting or situation of interest. To accomplish this, either an observer must be present so that coding can be done “live,” or an audio or visual recording device must be active so that coding can be done at a later date. To ensure impartiality, observers should not personally know the research subjects or the details of their involvement in the research study, and to ensure accuracy, observers should be adequately trained and monitored in the reliable use of a well-developed and tested descriptive coding system.

Clearly, two important aspects of naturalistic observation are the characteristics of the coding system to be employed and the ability of observers to appropriately apply the coding system. There are two major types of coding systems, those that focus on the discrete behaviors by a given individual, or “micro” coding systems, and those that focus on larger series of a variety of presumably related behaviors by one or more individuals, or “macro” coding systems. Within each type, coding systems have been developed to code a variety of details of social interactions, including physical movement, language content, affective tone, physical location or position relative to others, and the time spent in any of the above.

Regardless of the type or purpose, a coding system must clearly specify what is to be recorded; should contain a limited number of mutually exclusive code categories; and should not require observers to make inferences about phenomena that cannot be seen, such as the internal thoughts or feelings of an individual about a given behavior. Basic training in the reliable use of a comprehensive coding system usually takes several months of intensive work by a coding team. Training in relatively simple coding systems may take less time, but it requires a significant amount of time investment by a team nonetheless.

Once data are being collected for a study, the day-to-day reliability of the observers in continuing to apply the coding system accurately should be monitored. Generally, this is done through the use of regular, random checks where two observers code the same interaction, and their agreement on specific codes is checked. Usually, from 10% to 20% of observations are coded by two or more observers to conduct such reliability checks. Observers who do not meet set reliability standards should be removed from the data collection process and provided further training.

Because of the recognition that human behavior tends to vary across situations, researchers who collect naturalistic observation data often attempt to collect multiple measures of the behaviors that are being observed. This includes multiple observations within the same setting, observations across different settings, and observations as well as impressions by several raters. For example, child physical aggression might be assessed via micro-observational coding of several home, classroom, and playground interactions, as well as by more global questionnaires completed by parents, teachers, children, and peers. Important questions within most areas where naturalistic observation is used include how many observations and what length of time is required to obtain a truly representative sample of the behavior of an individual, how study participant characteristics such as culture affect the reliability and validity of coding systems, and how the significant costs related to creating and maintaining a reliable coding team can be managed to enable the widespread use of observational techniques in research and practice settings.