BACKGROUND: Structural equation modeling examined the relationship between change in negative affect (NA) and change in heart rate variability (HRV) among 121 Latinos with type 2 diabetes. METHODS: This study leveraged data from the Community Health Workers Assisting Latinos Manage Stress and Diabetes (CALMSD) study which compared diabetes education vs diabetes education plus stress management. Participants completed surveys of NA at baseline and again 8-10 weeks later. They also wore 7‑lead, 3-channel ambulatory ECG monitors for 24 h at both time points. The latent variable NA was modeled by observed scores on symptoms of depression, anxiety, diabetes distress, and wellbeing (reversed). The latent variable HRV was modeled by observed scores in the time domain (the standard deviation of the R-R interval [SDNN] and the root mean square of the successive differences [RMSSD]) and in the frequency domain, i.e., log-transformed ultra-low frequency, very-low frequency, low frequency, and high frequency. RESULTS: At baseline, there were strong, negative cross-sectional associations between NA and HRV. Baseline NA predicted change in HRV, whereas baseline HRV did not predict change in NA. Controlling for fasting glucose and treatment assignment did not meaningfully alter the findings. Stress management improved NA but not HRV. At followup, a greater reduction (improvement) in NA was associated with a larger increase (improvement) in HRV, with a small-to-medium negative association that approached statistical significance. CONCLUSIONS: Findings indicate a longitudinal relationship between NA and HRV, and suggest that improvement in one may be associated with improvement in the other.