This study captured daily and weekly mood ratings using a smartphone from bipolar disorder (BD) and unipolar major depression disorder (MDD) subjects at high (HRMDD) and low risk (LRMDD) for developing Bipolar Disorder (BD) and healthy controls (HC). METHOD: 40 subjects (18 - 30 yr) (6 BD, 13 HRMDD, 16 LRMDD and 5 HC) were studied and a total of 2401 daily and 744 weekly ratings were collected. HRMDD and LRMDD subjects were naturalistically treated with antidepressants. We investigate if latent-class analyses of ratings can detect mood instability among MDD and BD groups. RESULTS: Our analyses revealed four underlying mood states correlating with clinical mood states. There was a trend for greater number of state changes in BD and HRMDD subjects compared to LRMDD and HC groups. CONCLUSION: Smartphone ratings may adequately capture mood instability in BD subjects and at risk HRMDD subjects and offers a prudent way for monitoring development of serious manic symptoms.