Hi, When very simple models are tested in step.gam, I have observed the some difference in the final output result based upon the version of the package tested. Is this an expected behavior or a bug in the latest version of the gam package? With gam 1.4: > data(gam.data) > gam.object <- gam(y~1, data=gam.data) > step.object <- step.gam(gam.object, scope=list(x=c('1','x'))) Start: y ~ 1; AIC= 232.8823 Trial: y ~ x; AIC= 126.5148 Step : y ~ x ; AIC= 126.5148 > step.object Call: gam(formula = y ~ x, data = gam.data, trace = FALSE) Degrees of Freedom: 99 total; 98 Residual Residual Deviance: 19.53955 With gam 1.9.1 or 1.12: > data(gam.data) > gam.object <- gam(y~1, data=gam.data) > step.object <- step.gam(gam.object, scope=list(x=c('1','x'))) Start: y ~ 1; AIC= 232.8823 Step:1 y ~ x ; AIC= 126.5148 > step.object NULL Thank you Sebastien