Latent variables as dependent or independent variables?

I'm a little confused concerning whether or not factors are "causal" or if they can be both causal (as in independent variables), or an outcome (as in dependent variables), or both! For example, figure 14.2 in TF shows the factor "grad school success" as causal of observed variables and an outcome gender and undergrad success.

One of the strengths of SEM is that it allows the specification of latent variables as both an outcome (dependent) variable and a predictor (independent) variable. When we study SEM we will distinguish variables as either "exogenous" (serving as independent variables only) or "endogenous" (serving as dependent variable for at least one other variable). Using this nomenclature, in the rather complex model represented in TF figure 14.13 (on page 762 of TF 5ed):

• AGE, STRESS, and SELF are exogenous (b/c they serve as dependent for no other variables)
• PERCHEAL is endogenous (even though it is an IV for USEHEAL, it serves as DV for AGE, STRESS, and SELF)
• USEHEAL is endogenous