Haviland, Boulifard, and Magai propose interesting points of view on the study of the self and emotions, and illustrate their positions with admirable research. I especially admire their efforts to present parts of the life trajectories of real people and illuminate their plights by analyzing underlying processes. The main plea, as I see it, is an appeal to leave the traditional methods of analysis in the field and switch to non-linear model building accompanied by computer/spreadsheetbased simulations of the model. Such a model-based simulation of the processing of fearful stimuli has been presented by Haviland et al. to demonstrate the power of non-linear models over classical approaches. The Haviland et al. model predicts the course of the intensities of seven emotions over a time period of eight seconds in the life of a single (hypothetical) individual.

Below I will try to mitigate the enthusiasm in the Haviland et al.-plea by pointing out four warnings, extending beyond the Haviland et al. model. They concern: (1) anchoring the model-time unit in real developmental time; (2) the Haviland et al. (and my own) plea that the analysis of individual developmental trajectories should be accompanied by a compatible type of (statistical) treatment of the simulated trajectories; (3) the turn toward non-conscious processing of selfrelevant information at microlevel, which is implied in many nonlinear dynamical models, should not divert attention from solving unanswered questions about the cultural construction of the self and its development; (4) classical methods of analysis have not (yet) been exhausted in advancing our understanding of the development of the self, they were just not applied to the right type of data.