Welcome To The EmoTe Lab

Welcome to the Emotion and Temporal Dynamics (EmoTe) Lab, directed by Dr. Sarah Sperry, in the Department of Psychiatry at the University of Michigan and affiliated with the Heinz C. Prechter Bipolar Research Program. The EmoTe Lab has a broad mission to improve early detection, predict illness trajectory, and develop personalized interventions for bipolar spectrum disorders (BSDs). Within this broader mission we are working to better characterize and understand intraindividual variability in emotion and behavior in real-world contexts. We use digital phenotyping methods (smartphones and wearables) and advanced idiographic statistical methods to model dynamics over both micro (e.g., momentary) and macro (e.g., years) timescales.

Research in the EmoTe Lab is grounded in clinical, personality, and affective science. This integrative approach allows us to interrogate emotions and behavior through multiple lenses. Our work is highly influenced by Dynamical Systems Theory, viewing processes of interest as the result of interactions of multiple subsystems within the person and environment. As such, emotion, cognition, and behavior are viewed as contextualized dynamic processes rather than static "symptoms" of psychopathology.


We aim to identify affective and cognitive mechanisms underlying bipolar spectrum psychopathology in order to:

IMPROVE early detection of illness onset and symptom re-occurrence

PREDICT​ individual illness trajectories, course, and treatment response

DEVELOP scalable and equitable personalized interventions




Using smartphones and wearable devices, we can actively and passively sample emotions, cognition, and behavior in real-time. These methods allow for the assessment of ecological valid, context-dependent, and dynamic processes that unfold within the individual.


​In order to understand heterogeneity, longitudinal trajectories, intraindividual variability, and treatment response, we use intensive longitudinal modeling (ILM) and idiographic approaches. This includes time series analysis, dynamic structural equation modeling, and machine learning approaches for deep phenotyping


Quantitative modeling approaches (hierarchical modeling) can be used to understand, characterize, and develop new measures of dimensional phenomenon. No more carving nature at its joints!