Observational methods are excellent tools for the study of the complex relationships examined in developmental psychology (Bakeman & Gnisci, chap. 10, this volume). One of the key distinctions between the approach of many developmentalists and that of some other disciplines is the importance placed on context. Individuals do not exist in isolation; they actively construct their environment and simultaneously are influenced by their environment. The complexity of these interactions is often best captured through observational methods that allow the researcher to view the relationship and transactions in their entirety rather than testing very small and isolated pieces of the relationship. Moreover, an emphasis on the importance of context necessitates the investigation of variables in relation to specific contexts. Observational methods allow for the examination of contextual specificity.
Observational methods are thus a crucial element of developmental research, but along with their many advantages come some disadvantages. These methods result in extremely rich data, and this is both an advantage and a disadvantage. The data from observational studies are often able to capture "natural" phenomenon that cannot be verbalized or recognized by the participants themselves, and observations are less subject to experimenter effects and issues of social desirability bias (although individuals may act in socially desirable ways when they know that they are being filmed; Zegiob, Arnold, & Forehand, 1975). Yet these data are often difficult to codify or quantify, and frequently both extensive training and time are required to reach interrater reliability. Additionally, the complexity of observed relationships makes it difficult to test single constructs, therefore possibly lowering internal validity. Furthermore, observational methods performed in a laboratory sometimes may not generalize outside the specific situation, and because of the unfamiliar environment, they may be criticized for being artificial. Nevertheless, there is not another method presently available that is better at addressing relationships in context and interactions between individuals.
Convergent and discriminant validities of observational methods vary with the construct tested, the type of measure the observational method is tested against, and the population tested. For example, Black, Hutcheson, Dubowitz, Starr, and Benson-Howard (1996) found low to moderate convergent validity, rs = .01 to .49, of parent-child interaction observations from the Parent-Child
Early Relational Assessment (Clark, 1985) with maternal self-report on the Brief Symptom Index (Derogatis & Melisaratos, 1983) and The Parenting Stress Index (Albin, 1990), depending on the parental construct and the context (feeding or free play) of the observation. In another observational study of parent-child interaction, Crowell, Feld-man, and Ginsberg (1988) found 93% discriminate validity for predicting infants' clinical status (i.e., placement into clinical or nonclinical groups) and attachment classification using their structured play procedure for assessing mother-infant interactional behavior.
Observational methods are analyzed through a variety of statistical techniques. Because they are more open to interpretation than many other methods, multiple raters are often used to ensure validity of results. Kappa statistics or correlations, depending on the type of scale, often are used for inter-rater reliability, which tests for the congruence between raters' observations (Nussbeck, chap. 17, this volume). Hierarchical linear modeling (HLM) can be used to create a model including multiple observations of an individual (Hox & Maas, chap. 19, this volume). In addition, statistical methods such as sequential analysis are used to relate patterns in the sequencing of observations of two behaviors emitted by one reporter or potentially interrelated behaviors of two actors (as in a mother-infant interaction). In lag sequential analysis, the occurrence of one behavior preceding or following another behavior is recorded, and behavior frequencies and conditional probabilities can be computed (Farrell, 1994). For example, Feldman, Greenbaum, and Yirmiya (1999) used lag analysis to compute mother-infant mutual synchrony, in which the infant leads affectively and the mother follows as a form of emotional validation.
The use of observational methods in developmental psychology can be illustrated through the study of the parent-infant attachment relationship. Reports of parenting are especially prone to social desirability biases, as this construct is often emotionally charged because the parental role is highly valued in society. Furthermore, parenting behaviors may be difficult to recognize by those in the relationship and require objective assessment. Attachment theorists hypothesize that children form "working models" of self and attachment figures from their early relationships that serve as templates for future relationships and situations (Bowlby, 1969). Thus, the attachment relationship is an important construct that can be used to predict developmental trajectories when accurately measured.
A commonly used measure of the attachment relationship, Ainsworth, Blehar, Waters, and Wall's (1978) Strange Situation Procedure, frequently is used to activate the attachment system for observational purposes. From the attachment perspective, when an infant experiences distress, separation, fatigue, illness, fear, or other types of stress, the child's attachment system motivates him or her to seek security from a caregiver (Boris, Aoki, & Zeanah, 1999). An unfamiliar situation is believed to heighten such attachment reactions. In the Strange Situation, the infant participates in eight different episodes in an unfamiliar laboratory room where the parent and a stranger are present in varying combinations. The observations from the Strange Situation Procedure are later coded for type of attachment: securely attached, insecure-avoidant, insecure-resistant, and disorganized or disoriented (Main & Solomon, 1986). Impairments in the infant-parent attachment relationship as assessed with such methods are correlated with a variety of poor outcomes throughout the developmental literature, from psychopathologies (Fagot & Pears, 1996; Shaw, Owens, Vondra, & Keenan, 1996) to increased maltreatment risk (Cicchetti, Toth, & Maughan, 2000). Attachment classifications in the Strange Situation, when consistent and clearly observable, were also found to be useful for making predictions about the home caregiving environment (Gaensbauer et al., 1985). However, convergence between the Strange Situation and the Attachment Q-Sort (Waters & Deane, 1985), an observational rating system using a card piling technique, has been inconsistent across studies (Thompson, 1998), perhaps because the content of the two types of measures differs somewhat. Despite these inconsistencies, observational methods as illustrated by the Strange Situation Procedure can provide rich information about the nature of complex relationships in specific contexts.
Was this article helpful?