A recent Nature article, repeated in a Nautilus blog post, talks about author accountability. The article suggests that at least one author per collaborative group signs a statement with reference to Nature's publication policies. This policy would certainly help avoid honorary authorship, but it can be difficult to enforce in large research projects.
I would like to make another suggestion. The quality of a research paper should not only be judged by the number of citations it receives (which improves the Hirsch number of the author), but also by the number of authors. There are of course research projects that are only possible with large numbers of collaborators, but many biomedical papers probably only need 2-4 authors, but rather have 5-8 authors.
One good rule of thumb is a number of papers published per year. If that number is too high (e.g. more than 10), than the author has probably not contributed significantly to all those papers. Department heads often fall into this category. These rules can easily be applied when reviewing job or grant applications.
One of the most famous examples for authorship not perused is the 1922 CMAJ paper about the action of pancreatic extracts on blood sugar in diabetics. John MacLeod was not a coauthor, but rather thanked in the acknowledgements. As head of the department he provided mainly logistical support. He still went on to win the Nobel Prize for the discovery of insulin the following year.