Factors in Reader Responses to Negative Letters: Experimental Evidence for Changing What We Teach
http://jbt.sagepub.com/cgi/reprint/13/1/5
  access restricted (by the publisher) to members/subscribers/customers only
  peer-reviewed

Locker, Kitty O.
Journal of Business and Technical Communication
1999

Abstract:

This article summarizes the scholarly discussion about negative messages and reports the results of two pretests and two experiments using negative letters. The results show that buffers did not significantly affect college students' responses to simulated letters refusing credit and denying admission to graduate school, and strong resale was counterproductive. Students responded least favorably to rejection when they were surprised by it and when their other options were limited. On the basis of these experiments and the published literature, the author recommends that negative letters normally begin with the reason for the refusal, using a buffer only if one of several exceptions apply. If the reason makes the company look good, it should be spelled out in as much detail as possible. If an alternative or compromise exists, the writer should suggest it. Although a positive ending is not necessary, if one is used, a bland positive is better than a strong one, especially in letters to clients or customers.