Reliability refers to the consistency of assessment results; the degree to which students' assessment results are the same when they complete the task on two different occasions, they complete different but equivalent tasks on the same or different occasions or two more more graders mark the assessment similarly.
 

Nitko, A.J. (1996). Educational Assessment of Students. 2nd ed. Englewood, NJ: Merrill.

 
Nitko (p. 75) offers some very general rules of improving the reliability of assessment results:

1. lengthen the assessment procedure
2. broaden the scope of the procedure -- include all of the essential and important aspects of the target learning performance
3. improve objectivity by using a systematic, more formal procedure for scoring such as a rubric
4. use multiple graders
5. combine results from several assessments
6. provide sufficient time to students
7. teach students how to perform their best
8. match the assessment difficulty to students' ability levels
9. select assessment tasks that do a good job of differentiating among the best and worst students