It is de facto a requirement these days for aspiring academic scientists, that a number of years be spent on postdoctoral appointments. I have written about the usefulness of the postdoctoral experience before. Two (closely related) questions that I am frequently asked by current or soon-to-be postdocs are: How long should a single postdoctoral appointment last ? and How many years should one spend as a postdoc ?
To a certain extent, the answer to the above two questions depends on the specific field of inquiry, but some general considerations are valid across the board.
First of all, one must be clear on what “postdoctoral appointment” means. It is a term position, that the postdoc (PD), a junior scientist who has completed his/her doctoral degree typically no more than a few years earlier, spends doing research under the supervision of a Principal Investigator (PI), normally a faculty at a university or a permanent member of the research staff at a national laboratory. Salary, benefits, as well as the degree of intellectual freedom enjoyed by the postdoc can vary (even significantly), but the most important aspect of the position is that it is non-permanent.
There are cases when the institution hosting the PD will be willing to extend the duration of a term appointment (in some cases essentially indefinitely), as long as the PDs themselves or their PIs are willing and able to obtain the funding required to prolong it. Alternatively, the position may be extended as a PD is employed part-time by the hosting institution for purposes other than those for which the person was initially hired, independently of the PI ,.
Because the specific terms of postdoctoral appointments can vary so much, often times a non-permanent, term position for which benefits are significantly better than average will be re-labeled to make it sound more “official” than it is (e.g., “research associate”, “staff scientist” or something along these lines). It is important to be clear on one thing: these are rarely more than cosmetic, “feel-good” type changes, aimed at providing some formal reassurance to the PD — the bottom line, in the eyes of those who will evaluate the curriculum vitae of a PD for hiring purposes, is that this person spent that time in a non-permanent setting.
Why do I insist on the ‘non-permanent’ aspect ?
Because most of us aspire at a permanent employment. Nobody (well, almost nobody) likes the uncertainty, stress and sense of precariousness associated with a term position; therefore, the general perception is that anyone spending what could be regarded as too long a time on a term position, likely did so out of inability to secure a permanent appointment. In turn, such perceived inability will often negatively affect the strength of one’s candidacy for a potentially permanent (e.g., tenure-track faculty) position.
Is this fair ? Of course not (since when is life fair anyway). There are many reasons, both professional as well as personal, which may induce someone to remain on a term appointment rather than seeking a permanent position . However, very few search committee members will be willing to investigate such reasons, and one will be more likely branded as “good, but maybe not excellent”, “possibly a problematic individual”, “not quite faculty material”.
Thus, extending the duration of one’s postdoctoral appointment beyond what is commonly regarded as the average in that field (I come to that in a moment), while clearly providing relief in the short term, will almost never be a good thing for one’s professional future. Not only will it make the PD less palatable at institutions other than the one where (s)he is employed, it will also make it far less likely that that very institution may consider the person for permanent employment (something that happens seldom, but is not unheard of), should an appropriate hiring opportunity arise .
Based on my experience and observation, these are my rules of thumb for Physics PDs, assuming that the goal is that of strengthening one’s CV in light of competing credibly for faculty appointments, or permanent staff positions at national laboratories:
1) No more than three years at the same place
2) No more than five years as a PD altogether .
I think that, give or take one year, the above is probably valid across all scientific fields. To be a PD for longer than that, in all likelihood means reducing one’s chances of getting that coveted academic/science job, in the process being taken advantage of, putting one’s life on hold, and being disappointed and embittered in the end.
 For example, a PD may end up taking on teaching assignments at a university, or technical tasks such as computer administration. In many cases, the person performs an important, mission-critical task at a fraction of the cost of hiring a full-time person for the same task, resulting in a significant save of money for the institution.
 A common expression in the community, to describe positions of this type is soft-money position.
 Those hoping, for instance, that since they have spent a number of years as (glorified) PDs doing research and teaching at the same university department, in the process proving themselves competent, conscientious and hard-working, will have an edge over other applicants for a tenure-track faculty position in that very department, are simply deluding themselves.
It is simple economics: Why waste a faculty or permanent position to hire someone who is already there, has been (and likely still is) willing to work on a lesser appointment, and is apparently uninterested or unable to find a permanent job anywhere else ? Much better to use the position to bring in another person.
 In the interest of full disclosure: I did five years as a PD, my first post-doctoral appointment lasted three, and as I was interviewing for faculty jobs in my fifth year I was contemplating industry jobs. I interviewed with a bank, as well as with (gasp) these people too, and was seriously considering a career change (in the financial sector), should, my job search have failed.