Most lab-type courses grade on a "points-off" basis. That is, the exercise starts off with a maximum number of points, and then points are deducted for each omission or mistake. Thus a student ends up with some number of points out of total number available.
There are several problems with this kind of grading. First, it presumes the existence of a "perfect" paper, one which has zero points deducted. Good students are thus prompted to try to figure out precisely what the instructor wants, both in content and in format, and hit that target as closely as possible. In addition, "points-off" grading discourages all kinds of creative or out-of-the-box thinking. Since one can only lose points by venturing into unknown territory, it seems best to stick as closely to the precise assignment.
For Astro 255, I want to encourage students to think about the assignments as broadly as possible. So I'm going to modify the "points-off" model in the following way. Each assignment will start with a particular number of points (typically 10), and mistakes and omissions will result in point reductions, just as in the standard model. BUT I will also add points when a student does something interesting, clever, or creative. Thus there is no maximum number of points, and students can gain points by unusually good work, as well as lose points for blunders of various kinds.
The likelihood of gaining points depends on the nature of the assignment. Some kinds of work lend themselves to creative thinking more than others. So the typical point range will vary from one assignment to another. It is therefore not easy for students to figure out where they stand in terms of their final course grade - this is a drawback to grading with both "points-off" and "points-added". To alleviate this problem, I plan to give periodic updates on how the total number of points earned so far translates into a letter grade - this translation will depend on how I feel the overall class is doing. If everyone is doing well, there will be a lot of As - under no circumstances will a student gain from another student's poor performance.
So the question "do you grade on a curve" has two answers. I do grade on a curve in the sense that I do not expect that 90%=A, 80%=B and so forth. But I don't have a fixed overall grade distribution or a set fraction of As, Bs and Cs in mind, so in that sense I don't grade on a curve - students are not competing with each other, but will be evaluated on their mastery of the material, regardless of the performance of others in the class.