The position of the testes in some selected mammals illustrating three character states: (a)testicles not descended (testicond), as exemplified by the rufous elephant shrew, Elephantulus rufescens; (b) descended but ascrotal as exemplified by a harbor seal, Phoca vitulina; (c) testicles descended and scrotal as exemplified by a horse, Equus caballus. T=testicle; K=kidney; P=penis; T & S=testicle in scrotum.

This is an odd thing about mammalian testicles: they're often in a peculiar and very exposed place, bouncing about uncomfortably and dangerously, hanging out of the body in a scrotal sac. Other animals don't do anything so foolish; why mammals?

A couple of hypotheses have been proposed that might explain what advantage we derive from this peculiar arrangement.

• The display hypothesis. Males are flaunting their strength and health and virility and ability to protect a delicate package. Some species do make a show of it; some baboons have bright blue scrota, for instance. Most do not, however, and the idea seems unlikely to explain why this trait is preserved in mammals with widely diverse mating strategies.
• The cold storage hypothesis. It may not be so much the testis that is maintained at a cooler temperature outside the body core, but the epididymis. The epididymis is a storage organ for sperm. It may boost the longevity of the stored sperm if they are kept cool, maximizing the quantity of sperm that can be delivered. There are some good reasons that support this idea—some mammals do have internal testicles, but keep them organized so the epididymis is superficial—but personally, this hypothesis seems unlikely, and if it is sufficient to keep just the epididymis cool, it doesn't explain how the testicles ended up exposed in a bag.
• The training hypothesis. This one seems a bit of a stretch to me: the idea is that the testicle and epididymis are like boot camp, a tough, difficult environment that will weed out the unfit and allow only the strong to go on. Cool temperatures, more variable temperature, restricted blood flow...all prepare the sperm for the hostile environment they will confront upon ejaculation. There are many untested assumptions in this hypothesis, and it seems particularly improbable in species that practice sperm competition that they would sacrifice quantity for hypothetical quality, or that they would need to do something as risky as descensus and scrotal exposure to create an unpleasant cellular environment.
• The temperature hypothesis. The most likely explanation is that there is something in the function of the testis that is optimized for a lower temperature, and that the clumsy kludge that evolved to reduce that temperature was to let the organ hang out in the breeze. This seems reasonable, since fertility can be directly affected by temperature—a difference of a few degrees can be the difference between fecundity and sterility. The specific constraint in the process of sperm development that imposes this limitation hasn't been identified, however; one contributing factor (but definitely not the sole factor) may be that mutation rates increase with temperature, and the male germ line, which undergoes many more cell division than the female line, is much more sensitive to small changes in the mutation rate.

Werdelin and Nilsonne (1999) have taken a phylogenetic approach to examine this question. It may surprise some of you to learn that not all mammals have a scrotum. Many, like us people, have fully descended testicles contained in an external sac, the scrotum. Some have descended testicles, but no scrotum; the testicles are imbedded in the body wall. And some don't bother with that descensus nonsense at all, and keep their testicles high up in the dorsal body wall, near the kidneys, a condition called testicondy. If you'd asked me before I read this paper which animals have scrota and which do not, I wouldn't have had a clue (it really does seem like a rather personal thing, don't you think?), nor would I have had much of an idea of the distribution of scrota across the Mammalia. Werdelin and Nilsonne have peeked under the kilts, though, and summarized the distribution. Here is the abstract of the paper:

The adaptive significance of the scrotum and the evolution of the descent of the testicles and epididymis have been a focus of interest among biologists for a long time. In this paper we use three anatomical character states of the scrotum and descensus: (1) testicles descended and scrotal; (2) testicles descended but ascrotal; (3) testicles not descended (testicondy). These states are then mapped on an up to date phylogeny of the Mammalia. Three main points arise out of this mapping procedure: (1) the presence of a scrotum is either primitive in extant Mammalia or primitive within eutherian mammals except Insectivora; (2) evolution has generally proceeded from a scrotal condition to progressively more ascrotal; (3) loss of testicular descensus is less common in mammalian evolution than is loss of the scrotum. In the light of these findings we discuss some current hypotheses regarding the origin and evolution of the scrotum. We find that these are all incomplete in so far as it is not the presence of the scrotum in various mammal groups that requires explaining. Instead, it is the reverse process, why the scrotum has been lost in so many groups, that should be explained. We suggest that the scrotum may have evolved before the origin of mammals, in concert with the evolution of endothermy in the mammalian lineage, and that the scrotum has been lost in many groups because descensus in many respects is a costly process that will be lost in mammal lineages as soon as an alternative solution to the problem of the temperature sensitivity of spermatogenesis is available.

What they see in the phylogenetic distribution of scrota is that our common, ancestral proto-mammal had probably evolved a scrotum as a solution to its fertility requirements, and really, probably the best answer to why we have this odd scrotal arrangement is that that is the way great-great-greatⁿ-grandpa did it. It is the primitive condition, inherited from mammal-like reptiles of the Permian, and mammalian lineages have been independently ditching the idea as fast as they can...which isn't very fast. Here's one of their summary diagrams:

Results of mapping the three character states onto a phylogeny of the Mammalia. Testicle position unordered: red, testicond; gray, descended and ascrotal; black, descended, scrotal; blue, marsupial; white, equivocal.

The black bars are us, the species that let it all hang out. The most parsimonious explanation for their distribution, the one that requires the smallest number of novel adapations, is that they represent the ancestral condition. Some lineages, such as rhinos and tapirs and bats and whales (whales and seals, by the way, still keep their gonads cool by shunting blood vessels that pass through the skin back to the testicles—they essentially have water-cooled balls), still have descended testicles but don't have scrota; these are in gray. And the rare red lineages, elephants and manatees most prominently, have dispensed with the whole silly arrangement and keep their testicles safe and deep in the body cavity. (This is very useful knowledge to have if you ever get angry at an elephant and are tempted to kick him in the nads...don't bother. He'll just laugh at you.)

Here's the rough history of the mammalian testicle, in summary:

Studies of the origin of endothermy in mammals indicate that incipient endothermy may have evolved in mammalian ancestors as early as the middle Late Permian, at least 260 million years ago, which is some 40 million years before the origin of the Mammalia and 130 million years before the origin of crown group mammals, i.e. those for which we have here demonstrated that the scrotum may be primitive. A plausible, though at present untestable, scenario is that in the course of the evolution of mammalian endothermy, core body temperatures eventually reached levels at which spermatogenesis was disrupted. At this point, the testicles had to be cooled, and descensus into a scrotum was one way of achieving this result. However, because of the various costs of descensus the evolution of the male reproductive system of mammals since then has revolved around finding alternative solutions to the original cooling problem. This means that testicondy probably evolved only once among Eutherians, strengthening the conclusions herein.

Werdelin L, Nilsonne A (1999) The Evolution of the Scrotum and Testicular Descent in Mammals: a Phylogenetic View. J. theor. Biol. 196:61-72.