While some of the formal notions of type theory can seem daunting, I personally have found it extraordinarily useful in being able to understand the different expressiveness of type systems in programming languages. There are languages that clearly implement say, parametric polymorphism well (ML-style languages) and safely compared to far more ad-hoc approaches, such as C++.
At the same time it's very cool to be able to grasp more advanced type systems (e.g. dependent types) that haven't been popularized yet and imagine how they could lead to radically different methods to software engineering.
I like Benjamin Pierce's Types and Programming Languages book, it's mathematically rigorous but well motivated and structured for CS-oriented readers. It covers everything from pure lambda calculus to mutable objects, Featherweight Java, and type systems with subtyping.