This study describes a comparison between retrospective questions and daily diaries inquiring about positive and negative support in spousal interactions. The design was a multitrait-multimethod matrix with trait factors of positive and negative support, and method factors of retrospective questions and daily asked questions. Five questions were used for positive support and 2 questions were used for negative support. The data were multilevel, with daily measurement occasions nested within subjects. In addition, the data were ordered categorically. The negative support events proved to be so rare that the original 4-point response scale had to be dichotomized. The resulting model could be estimated using M plus, but the model and data complexities set some limits to the analysis. The results showed that at the subject level both positive and negative support could be assessed with sufficient reliability and validity. At the daily measurements level, positive support showed significant but low reliability and validity, but negative support could not be assessed reliably. It was concluded that at the daily level both positive and negative support should be viewed as transient events that do not indicate an underlying latent variable, but that could be modeled as a formative construct.