I am a TA currently grading a set of 80 midterms for an engineering math course for fourth-year engineer undergrads. The test had four questions, each question had around four parts.
Of course, typically a student can make a wide variety of different types of errors on the exam. But what I have noticed in this exam is that some of these errors are not necessarily conceptual, yet we are basically taking all the points away for these "minor errors" (per instructor's grading scheme).
For a question with several parts. A student copied one of his own (correct) answers from the previous part incorrectly.
A student wrote down all the steps correctly, until the last step where he had to evaluate the expression at a number and the mistake was forgetting a negative sign.
After a long derivation, a student substitute formula incorrectly (forgot a square root in the formula, ahem, Gaussian). Everything else completely correct.
There was a litter of other minor errors. Now, these students are in their final year of engineering school, so I think I can give them the benefit of the doubt that they wouldn't make silly mistakes such as writing "0 - 1 = 1" or copy things down wrong in a non-time-constrained setting, which is the situation they will most likely face when they graduate and work in industry.
Is there a way to provide feedback for a student if he or she makes silly mistakes on the exam? I do not think that giving them a zero for writing a minus sign incorrectly is a good way of either providing feedback or preventing the same mistake from happening again in their (long) lives.