Here is a an interesting test for some one to try.

Create a maze with a starting point and a finishing point.  Tell Group A people to solve the test by starting at the starting point and working forward to the finishing point.  Tell Group B people to solve the test by starting at the finishing point and working backward toward the starting point.  Determine if there is a statistical difference in the results.

Then create the same maze, but change the starting point to the finishing point, and the finishing point to the starting point.  Create Groups C and D in the same manner you created them for Groups A and B.  Compare the results again with each other like before, but also check them against the results in Group A and Group B.

What you are checking for is to see if there is a bias between working a project forward or working a project backward.  In a perfect world of a maze, it would seem like all the various options should be the same.  If they are different, then it may be useful to understand why they are different.  If they are not different then it seems either approach would work--although Tristan's idea of starting work at both ends may be best.

I know in my own background history of working projects that on many of them i gave a lot more attention to the end of the project before i did the beginning of a project.  I am not sure how that factors in the above, but it is just a question i have.