I am no Kant scholar (thankfully!) but I do regularly teach ethics and so am constantly re-reading the Groundwork of the Metaphysics of Morals. I have never really known what to think about Kant’s argument in the section where he is discussing the formula of the law of nature and how it applies to suicide. Here is what he says:
…[the] maxim…is: from Self-love I make it my principle of action to shorten my life when its longer duration threatens more troubles than it promises agreeableness. The only further question is whether this principle of self-love could become a universal law of nature. It is then seen at once that a nature whose law it would be to destroy life itself by means of the same feeling whose destination is to impel toward the furtherance of life would contradict itself and would not therefore subsist as nature…
I have always interpreted this passage as expressing some kind of natural law theory rather than a deliverance of the categorical imperative. The function or role of self-love is the preservation of life. Using it in this way violates that natural function and so results in a contradiction of a sort but not the usual sort you find in the case of perfect duties. Of course the flaw in that reasoning is that it is not obvious why it must be the case that to use something contrary to its function is immoral (is it immoral to use a butter knife to screw in a screw?).
For full disclosure, I take myself to be a Kantian of a sort, though I would probably not be viewed as an orthodox Kantian (I, for instance, think we can derive Utilitarianism from the Categorical Imperative (like Hare did as well) and I think we can derive duties towards animals in much the same way). And I also take it that suicide is morally permissible in at least some cases and so I have always been happy to see something odd about Kant’s argument against suicide. Yet today in class I found myself wondering whether this natural-law-remnant theory is the most charitable reading of that passage.
In the case of the other perfect duties we seem to find a pattern where in the imagined world the very thing that we are trying to do becomes impossible. Thus in the lying promise case the very act of making a promise becomes impossible. Since we cannot be coherently described as promising in a world where everyone makes lying promises we arrive at a contradiction. So, too, in the case of lying in general. In a world where everyone lies to avoid discomfort lying itself is impossible. So too for stealing. In a world where everyone steals stealing is itself impossible (since there is no ownership in such worlds and when we steal we take ourselves to be owners). So if we are to extend this general line to what Kant says here then it must somehow be the case that in the world where the maxim in question is a universal law suicide itself is impossible.
This seems implausible but it might not be. Suicide is the act of taking one’s life out of self-love. If self-love is itself not possible then suicide is not possible. So maybe what Kant had in mind was that self-love is by definition the thing which impels us towards life and so in the suicide-maxim world there is no self-love since there is no universal drive towards life, but rather only a limited drive towards life-as-long-as-it-is-agreeable or some such. Since there is no self-love there can be no suicide (which requires that you act from self-love). This would still be a questionable argument (since the limited thing still seems like self-love enough to do its job in ‘normal’ circumstances) but might be a more fair way of reading Kant’s claim in the above passage.
Richard Brown 
I’ve always also been troubled by that passage, and like you, have thought that it indicates some version of natural law theory, which should not be that surprising, since Kant thought of nature teleologically (I think the word he used was purposive).
Still, if we want to be charitable to him, since he is asking whether the maxim could be universalized to the status of a law (of nature), the question then becomes whether nature could maintain life if living beings were unable to endure hardship, and his answer is a resolute no. Now, that may seem like a false dilemma to us, since it would not be difficult to add some nuance to that black-and-white claim, but I think that’s where he’s going.
If at least part of the function (or purpose, if you prefer) of self-love is to ensure survival, as he seems to think, then an instinct designed for self-preservation would contradict itself if it would drive itself into extinction. Again, this may strike us as dubious, since self-love may be more about a quality of life than mere life itself, but I think that’s where Kant was going.
Hi Richard. If you haven’t already, take a look at Rae Langton’s essay, “Maria von Herbert’s Challenge to Kant” (http://homepages.ed.ac.uk/rhl/maria.html). I think you’ll enjoy it; plus, it is certainly relevant to the problem you note.
Stan Grof reports that according to his patients suicide was a means to end an intolerable experience and not a means to end life. The distinction is important.
However, this is a distinction that is unavailable to cognitive neuroscientists. Their reductionism and employment of the western pathology model outlaws the experiences that Grof describes his patients as attempting to complete. Such experiences involve a shattering encounter with symbolic physical and mental death. They include “death-rebirth” experiences. A person will attempt to complete an uncompleted experience of this sort but be led astray as to how to go about it, choosing physical death and rejecting or being unaware of symbolic death and subsequent epiphany of “rebirth”.
Instead, we must look to the tribal archeologists, particularly in their investigation of the local use of entheogens, for knowledge of the human condition and its natural encounter with experiential, symbolic death. Grof provides a modern source in his use of LSD and breathing techniques; but it must be said that the birth/death experiences that arise from these methods can be had by all, with or without these triggers.
So Kant’s intuition is right, though that would be impossible for workers in the Clinic sciences to accept. My complaint is that the reductionism of the medical model has, for many years now, prevented the emergence and dissemination of knowledge of the human experiential wardrobe. Knowledge of ourselves is shrinking, leaving thinkers bewildered by Kant’s thoughts.
I think that it’s a clear case of a contradiction in the will. “The same feeling whose destination is to impel toward the furtherance of life” is roughly the terminology Kant repeatedly uses in his third Critique when describing the feeling from beauty. Think of doing the right thing as “behaving beautifully” (as Kant seems to shift towards in his third Critique) – then it’s evident that suicide, if a universal law of nature, would directly contradict the universal aesthetic law (of beauty).
“It is then seen at once that a nature whose law it would be to destroy life itself by means of the same feeling whose destination is to impel toward the furtherance of life would contradict itself and would not therefore subsist as nature.”
We are autonomous beings who give ourselves universal laws of nature. The feeling of love or the “furtherance of life” (third Critique, repeatedly) from beauty is a universal law of nature. So it would be a contradiction in the will to will life and not life at the same time out of (self)-love. The categorical imperative just tells us to do the beautiful thing (to kalon). The only immoral actions are suicide/homicide/fratricide; everything else is permissible according to Kant as he shifts to in his third Critique.