I would go with "How to design programs" - looks at the beginning to be siplistic, but it gives a lot of understanding about lisp like languages.
Then I would go with SCIP and then Koans.
I am treating lisps as a kind of mind game (I dont like crosswords :) so I do Lisp) since I cannot use it for everyday job so from my point of view Scheme and Lisp are not very different. There is different approach to things like mutable state and other more advanced concepts, but basics are the same.
If you want to have at least an illusion of learning Lisp that can catch up one day, learn Clojure. Or go with Scheme's ugly duckling son - JavaScript. :)
SICP [2nd edition] is flat out just a more readable book than HtDP [first print edition, second edition is not in print but looks to be significantly improved].
As textbooks, it may be a bit of a different story, the tone of SICP is certainly cockier. There's less acknowledgement that mathematics is just domain knowledge within the context of an introductory "computer science" course. Then again, HtDP had two decades of experience with SICP upon which to draw. It also has the practical advantage of assuming that students had high levels of access to computers capable of running Lisp. That wasn't the case in the late 1970's and early 80's (or even the late 80's and mid 90's).
A course (lightly)based on HtDP is also starting in Coursera in a few days- https://www.coursera.org/course/programdesign. It uses Racket and the first two weeks of videos are already up so you can preview them.
Then I would go with SCIP and then Koans.
I am treating lisps as a kind of mind game (I dont like crosswords :) so I do Lisp) since I cannot use it for everyday job so from my point of view Scheme and Lisp are not very different. There is different approach to things like mutable state and other more advanced concepts, but basics are the same.
If you want to have at least an illusion of learning Lisp that can catch up one day, learn Clojure. Or go with Scheme's ugly duckling son - JavaScript. :)