Aim 2: To examine and compare how negative prediction error reward-learning sensitivity at six months predicts ad libitum intake of milkshake and weight status in postpartum women 1 y postpartum. Data Analyses for Aim 2: The model will include five regressors of interest, two regressors modeling the cue and three modeling the taste delivery after a cue. The cue regressors will include 1) hyper-palatable logo 2) sub- palatable logo. The taste delivery regressors will include 1) paired hyper-palatable logo and taste 2) paired sub-palatable logo and tasty 3) negative prediction error hyper-palatable logo and sub-palatable taste. Negative prediction error will be compared to the paired hyper-palatable and sub-palatable delivery as well as to baseline to assess sensitivity to negative reward learning. The hyper-palatable cue will be compared to the sub-palatable cue to assess sensitivity to reward related cues. Parameter estimates from the contrasts will extracted and regressed against weight change at year postpartum to examine how sensitivity to negative rewards and sensitivity to reward related cues is associated with weight change, and regressed against hyper- palatable milkshake intake at the ad libitum milkshake challenge. To assess free-living negative reward learning, intake of the sub-palatable milkshake will be treated as a binary variable (tasted the sub-palatable, or 0.05, and will be familywise error corrected using the Theory of Gaussian Random Fields47. Voxelwise Thresholded Z statistic images will prepared with a cluster-forming threshold of Z > 2.3 and a corrected extent threshold of p < nonparametric permutation testing will be preformed using FSL’s randomize, with 5000 permutations. All statistical maps additionally will be based on the threshold-free cluster enhancement (TCFE) generated from Randomise48. Non-imaging data will be analyzed using R (version 3.3.1, R Foundation for Statistical Computing Platform, Vienna, Austria). All analysis will be available on RPubs (http://rpubs.com/) and analytic code will be available on github (https://github.com/niblunc). did not taste). We will then preform a binary logistic regression comparing the parameter estimates to the binary taste variable to assess odds of tasting the sub-palatable milkshake. As menstrual phase may impact reward-related neural function, we will record females menstrual phase, defined as the number of days between the scan and the first day of last period and we will ask about type of birth control pill used (if relevant) so that we can covary for the effects of these factors along with hunger, thirst, and breastfeeding status.