A New Assessment of an Old Measure: Utilizing Latent Class Analysis to Examine the Strange Situation

Abstract

This work examined the Strange Situation coding system, utilizing Latent Class Analysis (LCA) to assess the number of latent categories underlying this measurement. In addition, the four behavioral dimensions of attachment behavior employed in Strange Situation coding (proximity & contact seeking (PCS), contact maintaining (CM), avoidance, and resistance) were evaluated with the LCA methodology to determine if simplifications (dichotomous vs. multi-level coding) or modifications (continuous vs. nominal/ordinal coding) were achievable. Participants were 1,191 children from the NICHD Study of Early Child Care and Youth Development (SECCYD) who were assessed using the Strange Situation procedure conducted at 15-months of age. Analysis 1 of the current work treated PCS and CM as 7-level nominal variables, and avoidance and resistance as 2-level variables (e.g., none vs. some avoidance). Analysis 2 was identical to Analysis 1, except participants classified as Unclassifiable (U) or Disorganized/Unclassifiable (DU) were excluded for comparison purposes. Analysis 3 treated PCS and CM as continuous, while avoidance and resistance remained dichotomous. Analysis 4 returned PCS and CM to their original 7-level state and freed the avoidance and resistance variables to vary by making them 3-level. None of the analyses yielded a clear model of best fit, although both Analysis 1 and 4 returned multiple potentially suitable models. Analysis 3 returned no acceptable models, calling into question the common practice of treating the four behavioral dimensions as continuous variables. No definitive answer was reached as to whether binary coding offered similar discrimination to 3-level coding. The 2-class model of each analysis contained latent classes displaying A- (insecure-avoidant) and B- (secure) type behavior, with participants showing classic C behavior seemingly included in the B class. This is counter to the typical secure vs. insecure comparison conducted in much of the attachment literature. The 3-class model of each analysis contained classes resembling the Strange Situation A and B groups, and the B2 subgroup, which diverges from the expectation of replicating the ABC coding system. Taken together, these findings begin to call into question the classic Strange Situation classification system, although further work is needed to address the many issues encountered with these analyses.