Marceau, Fisler and Krishnamurthi, "On novices' interactions with error messages"

https://cs.brown.edu/~sk/Publications/Papers/Published/mfk-mind-lang-novice-inter-error-msg

A rare empirical study that I found extremely useful.

Outside of a univ course, my lessons are slightly different from the paper's:

• Maintain a map of error messages for a project, use this to minimize the vocabulary of error messages.
• Be enigmatic rather than risk misleading.
• Make affordances beyond error messages optional. Color-coding has a cost. Not highlighting may be better.

Comments gratefully appreciated. Please send them to me by any method of your choice and I'll include them here.