There is a rich body of theory and empirical work on the evolution of selfing in flowering plants [e.g. [1, 2, 4–7, 9, 10]], but practically none of it involves an explicit role of selection involving time-limitation. The present paper is only the second to explore the implications of the time-limitation hypothesis and contribute to the maturation of this idea. According to the time-limitation hypothesis, selfing in annuals has evolved as a consequence of strong r-selection in ephemeral habitats, resulting either as an indirect consequence (trade-off) of selection for shorter time to reproductive maturity (Figure 1a), or as a direct consequence of selection for shorter pollination time, i.e., the time between flower maturation and ovule fertilization (Figure 1b), or both . Consistent with the predictions of this hypothesis, we found, using phylogenetically-independent contrasts, that (compared with outcrossing annuals) selfing annuals in general had significantly shorter plant heights, smaller flowers, shorter bud development time, shorter flower longevity and smaller seed sizes.
At the same time, these results are not inconsistent with the predictions of selection resulting from pollinator/mate-limitation associated with the traditional reproductive assurance hypothesis. Just as with many situations where two different mechanisms can potentially produce the same outcome/pattern, it is not easy here to clearly distinguish between the roles of "pollinator/mate-limitation" and "time-limitation". Nevertheless there are two important contributions from our study: First, in reporting significant life history differences between selfers and outcrossers, our multi-species study is unique in its comparison of monocarpic annual species exclusively. All previous multi-species studies of trait comparisons between selfers and outcrossers have involved variable mixes of monocarpic and longer-lived polycarpic species. Second, by comparing annuals exclusively, our results provide indirect support for the time-limitation hypothesis, not by rejecting the role of pollinator/mate-limitation, but rather by representing a system in which it is more plausible to argue for the role of time-limitation; i.e., compared with pollinator/mate-limitation, time-limitation as a selection factor favoring selfing is likely to have been much stronger, more persistent and more widespread. The strength of this argument lies in the fact that the annual life history is unequivocally a product of some type of time-limitation favoring an abbreviated life cycle, which is promoted by (among other things) selfing (as opposed to outcrossing) (Fig. 1). It is much less plausible to suspect that selection associated with pollinator/mate-limitation has been sufficiently strong and persistent to favor selfing in such a wide range of annual taxa across the many genera and families considered here. We emphasize, therefore, that for annuals the time-limitation hypothesis provides a more parsimonious explanation for the differences in traits between selfers and outcrossers. We consider each of these traits in turn below.
Plant height and time to anthesis
Taller plants may attract more pollinators and, hence, experience greater outcrossing rates [20, 21]. The pollination benefit of being relatively tall, therefore, is presumably experienced only by outcrossers. If, however, selfers have evolved from outcrossers , then why should selfers be shorter than their outcrossing ancestors? The relatively small size, including short height of selfers can be predicted as an indirect consequence of selection, from time-limitation, favoring precocious maturation time [22, 23] (Figure 1a). In the present study, however, selfers and outcrossers did not differ significantly in age at first flower. Andersson  found similar results between selfing and outcrossing populations of Crepis tectorum. Arroyo , however, reported that selfing individuals of Limnanthes floccosa flowered earlier than the outcrossing L. alba, as predicted by the time-limitation hypothesis. The results for flowering times in the present study may be confounded by the controlled greenhouse environment of constant day-length, temperature and moisture regime. In the field, flowering times may be triggered by environmental cues. L. floccosa, for example, uses soil moisture to trigger the early onset of flowering, thus escaping the detrimental effects of soil desiccation during seed development . Note also that age at first flower is only a crude estimate of time to reproductive maturity. Future studies may employ more detailed measures such as rate of mature flower production.
One of the most well established trends of predominantly self-fertilizing species is their reduced flower sizes compared with outcrossing species [1, 17]. The present results indicate that this trend is also evident even within annuals exclusively. In all but three of the 14 PICs, selfing annuals had smaller flowers than the outcrossing annuals (Figure 2b). Outcrossers and selfers had similar flower sizes in the Fabaceae and Plantaginaceae. In the Poaceae, outcrossing annuals had smaller flowers than selfers.
Under the time-limitation hypothesis, smaller flowers and selfing may be tradeoffs of selection for precocity (Figure 1a), or smaller flowers may be favored by selection because they promote selfing and hence, direct fitness benefits by abbreviating pollination time (Figure 1b). Also, if selfing evolves from outcrossing (by whatever mechanism), then selection may subsequently favour a reduction in flower size since relatively large flowers are no longer needed to attract pollinators. Hence, higher fitness may result if the resources required to construct and support these larger flowers are invested instead in other functions (e.g. seed and fruit development) .
Bud development time
Selfers had significantly shorter bud development times in all but one of the independent family contrasts (Figure 3) and all but one of the genus comparisons (Figure 5). Results from previous studies, however, are inconsistent. Shorter bud development times were found in selfing populations of Mimulus guttatus  and in Clarkia xantiana . However, no significant differences in bud growth rates were found between the selfing and outcrossing populations of C. tembloriensis . Hill, Lord and Shaw  reported that flowers from selfing populations of Arenaria uniflora develop over a longer period of time than observed in outcrossing populations. In the field, selfing populations of A. uniflora were also observed to flower at the same time or even later than outcrossing populations , suggesting that time-limitation is not currently a strong selection pressure. Self-fertilization in A. uniflora may have arisen through reproductive assurance in response to competition for pollinators . The evolution of self-fertilizing species from outcrossing progenitors has occurred repeatedly and independently in several lineages [1, 3, 14], each of which may have been associated with different contexts of natural selection vis-à-vis the fitness benefits of selfing.
The families and genera in which selfers had shorter bud development times also had significantly shorter flower longevities (Figure 4). In fact, all of the selfers had flowers that remained open for less than four days (except in Trifolium hirtum; Fabaceae), with a large proportion of flowers open for only one day, which is common amongst self-fertilizing species . The present data again indicate that this generalization apparently holds true even within annuals exclusively. By having flowers that remain open longer, outcrossers increase the probability of visitation by pollinators and successful cross-pollination . This fitness benefit is realized, however, only if there is sufficient time remaining after cross-pollination to complete seed and fruit development before the maternal plant succumbs to density-independent mortality in strongly r-selecting habitats . If time is limiting in this context, selection should favor selfing (Figure 1b) with no advantage in having long-lived flowers.
It is important to note that our data measure maximum flower longevity, since there were no pollinators in the greenhouse, nor was hand pollination conducted. Pollination has been shown to induce floral senescence in numerous species . This effect was not tested on any of the study species, which means that our observed flower longevities in outcrossing species may be longer than would normally be seen in the wild. Nevertheless, since selfing may have evolved as a method of shortening pollination time, and flower longevity was used as a measure of pollination time, the maximum floral longevity gives an indication of how long outcrossers can delay flower abscission or self-pollination (i.e. through delayed selfing).
Strong r-selection associated with the annual life form presumably favors wide dispersal mechanisms (for colonizing new and distant sites) which may be conferred by small seed sizes . The reproductive assurance hypothesis would predict, therefore, that most selfers are annuals because annuals are more likely than perennials to disperse further, or colonize new habitats where conditions are unsuitable for successful outcrossing (because of a shortage of mates or pollinators) and where selfing, therefore, provides reproductive assurance. The present study indicates that even among annuals only, selfers have smaller seeds than outcrossers (Figure 2c). Future studies are required to test whether smaller-seeded selfing annuals are more likely than their outcrossing annual relatives to disperse further or colonize new habitats and thereby incur potential reproductive assurance benefits of selfing.
An alternative explanation, however, is offered by an extension of the time-limitation hypothesis: strong r-selection favors an acceleration of all stages of the life cycle (Figure 1), including not only earlier reproductive maturity (Figure 1a) and a shorter pollination time (facilitated through selfing) (Figure 1b), but also a shorter seed and fruit maturation time, which, on a per-seed basis, is facilitated in turn through the production of smaller seeds. Andersson  found that self-fertilizing individuals of Crepis tectorum took an average of 16 days for fruit maturation, whereas outcrossing individuals of the same species required 43.3 days. Small seed size may also be simply a trade-off of selection for high fecundity, also favored by strong r-selection .
Habitat selection and time-limitation
While most selfers are annuals, it is not the case that most annuals are selfers. An unbiased literature survey  suggests that roughly half of all annual species are selfers and half are outcrossers. If, however, selfing annuals evolved in habitats with a short window of time for completing the life cycle (Figure 1), then selfing annuals should be significantly more common than expected (i.e. comprising greater than 50% of resident annuals) within habitats associated with historically regular, early-season disturbances (e.g. cultivated fields, gardens), or in habitats where severe droughts follow quickly after a wet season (i.e. deserts, Mediterranean climates, vernal pools). Hence, we should expect to find more selfers than outcrossers among annual weeds of cultivated habitats and among desert annuals in particular. Similarly, for annuals with both selfing and outcrossing ecotypes or races, we should expect selfers (or a higher selfing rate) to be more commonly associated with these severely time-limited habitats .
While rigorous tests of these predictions have yet to be explored, some preliminary support is available from published surveys. From a representative sample of Mediterranean annuals , we find a much greater representation of selfers: i.e. 34 selfers versus 11 outcrossers. Selfing and outcrossing desert annuals have been shown to be distributed along a moisture gradient. Outcrossing annuals are found generally in the wetter areas and selfers in the more arid zones, as seen in Clarkia xantiana  and between outcrossing populations of Limnanthes alba and its selfing relative L. floccosa . Since the length of the growing season is limited by the amount of moisture in the soil, selfers have a much narrower window of time to complete their life cycle before desiccation. During a severe drought, seed production in L. alba was reduced by one sixth, whereas the seed set of L. floccosa found in the same area was virtually unaffected by the identical drought .
The association between 'weediness' and self-fertilization has also been noted [2, 30]. An extensive survey of colonizing herbaceous plants of Canada showed that agricultural weeds of row crops and grain fields are almost exclusively annuals, and most of these are self-compatible . A published list of the world's worst weeds of agricultural crops  includes 76 species, 41 of which are annuals. Based on previous literature, we were able to identify the breeding system for 24 of these annuals, and, as predicted, the majority (20 out of 24) are selfers.