The general trend among seed plants is for gradually tighter and deeper enclosure of ovules within protective stuructures,including the tighter closure and fusion of carpels. Because of the adaptive advantage attached to this trend, reversals back to apocarpy are considered unlikely:
ovules on open, leaf-like structures (ancient seed ferns)----->
ovules on specialized leaf-like, cone-like, or shoot-like structures (gymnosperms) ------>
ovules within loosely-closed, leaf-like carpels (stem angiosperms) ------->
carpels apocarpous (early crown-group angiosperms and basal magnolids, eudicots, and monocots) ------->
carpels syncarpous or unicarpellate, and sometimes surrounded by tissues of the receptacle ("inferior ovary") (most advanced angiosperms)
Armbruster and colleagues did detect two probable reversals from syncarpy to apocarpy among eudicots (in the genus Crossosoma and members of the Saxifragales), and speculated that the advantage might be to extend the period of ovule fertilization so as to receive pollen from different sources. It's not clear that such an extension happens in these examples, however, as each only has a few carpels. Likewise, in archaic monocots like Butomus (in the Alismatales), the small number of carpels are receptive at the same time and for only one day (Bhardwaj & Eckert 2001). In Sagittaria, also in the Alismatales, the carpels are more numerous and physically spread out, but again are receptive at the same time for only one day. If multiple visitors fail to arrive during that window, it potentially leaves carpels unfertilized. Multiple insect visitors are possible in one day, but does the potential advantage of genetically different pollen arriving in the same flower outweigh the usual advantages of syncarpy? The idea needs fleshing out with stronger selective arguments and actual examples of it working.
From another perspective, if there is an advantage in spreading out the fertilization of ovules in either time or space, a much simpler way to effect that advantage in syncarpous flowers is to make the flowers smaller, with just one or a few ovules in each, and produce a series of them. This has in fact happened many times, as in the Saururaceae, Piperaceae, Asteraceae, palms, aroids, etc. I am not aware of any series of carpels in an apocarpous flower functioning this way, let alone a syncarpous ovary splitting apart to do so.
Despite the foregoing, Sokoloff et al. (2013) concluded that the earliest monocots were syncarpous and that apocarpous flowers evolved several times among them. In some cases, according to this scenario, syncarpy re-evolved a short time later from newly-apocarpous ancestors. What is the basis for this counter-intuitive proposal? It appears not to be based simply on the fact that the most ancient monocot lineage (Acorus) is syncarpous, but on a more extensive theoretical exercise involving cladistic analysis plus an additional step of "optimized parsimony analysis."
DNA-based cladistic analysis provides a clear, and increasingly accurate picture of the ancestral branching patterns of groups of organisms - - i.e. a phylogenetic tree. When only DNA information is used, the tree tells us nothing about when and where new adaptive traits arose. For this, we can "map" such physical characteristics onto the tree. For example, we can make a little mark on each branch containing only species known to be syncarpous. If two adjacent (sister) branches have syncarpous flowers, we assume their common ancestor had the same. It is also possible, however, that syncarpy evolved independently on each branch from an ancestor that was apocarpous. But this interpretation is less parsimonius, because it involves more independent evolutionary changes (syncarpy evolved twice instead of only once).
Optimized parsimony analysis takes into consideration a wide variety of tests, assumptions, taxa lists, character definitions, etc. for the same set of data and determines the most parsimonious interpretation of the evolution of particular characteristics. Some of the tests performed by Sokoloff et al. were ambiguous about the nature of the ancestors, but overall they suggested that syncarpous flowers were present first in the monocots. Quite possibly, the basal position of syncarpous Acorus tilted the final results in this favor, raising again the questions I raised in my previous post.
So from this cladistic perspective, multiple evolution of apocarpous flowers from a syncarpous ancestor is more parsimonious than multiple origins of syncarpy from an apocarpous ancestor. But does evolution necessarily follow the most parsimonious path? In the real world, perhaps one evolutionary trend, because of its adaptive value, is more likely to occur multiple times than the opposite. So to evaluate this proposal further, we need to consider the adaptive basis for each trend. Sokoloff and colleagues in fact raised the question: “Assessing the functional and adaptive significance of evolutionary transformations is clearly important" ( p. 75). They proceed to reiterate the advantages of syncarpy, but made no suggestions as to why reversals might occur. They made another disturbing statement: "Interestingly, in the monocot order Alismatales, congenital intercarpellary fusion was first lost and then re-appeared in three independent clades according to this scenario" (p. 64). One of those reversals would include the Butomus pictured above.
I recently outlined the principles advocated by Stebbins for evaluating alternate evolutionary scenarios (G. L.Stebbins and the process of adaptive modification). These principles result in the "other parsimony," the parsimony in which sequences of adaptive changes are assumed to proceed along the simplest paths, or "along the lines of least resistance." My example above, in which shifting to a series of small flowers, instead of splitting the ovary to make a series of separate carpels, is an example of such a simpler path. Further, once a developmentally complex and highly functional structure like syncarpy evolves, one would need a rather powerful selective pressure to undo it, something giving an advantage to apocarpy strong enough to cancel out the documented benefits of the syncarpous ovary. No one has yet offered such an adaptive scenario.
Therefore, "cladistic parsimony" must be balanced against "adaptive parsimony." Seventeen independent transformations from apocarpy to syncarpy may be more reasonable in view of selective pressures we know about than even one reversal. I think it is therefore premature to dismiss apocarpy in the ancestral monocots. Despite the fused carpels and other specializations of Acorus, apocarpy and looser forms of syncarpy (due to post-genital fusion) are widespread among the Alismatales, which are nearly as old as Acorus, and possibly more conservative with respect to the condition of their carpels. In the palms as well, separate carpels occur in more archaic groups (Nypa and the Coryphoideae), though the first branch, the lepidocaryoid palms, are syncarpous. This is a more complex situation, however, as the palms appear to have originated among clades that were already syncarpous. Still a scenario is needed for why fused carpels might become separate again in these palms. Does such a scenario make sense in the real world of adaptive pressures?
Armbruster, W. S. , E. M. Debevec & M. F. Willson. 2002. Evolution of syncarpy in angiosperms: theoretical and phylogenetic analyses of the effects of carpel fusion on offspring quantity and quality. Journal of Evolutionary Biology 15 (4): 657-672.
Bhardwaj, M., Eckert, C. G. 2001. Functional analysis of synchronous dichogamy in flowering rush, Butomus umbellatus (Butomaceae). Am. J. Bot. 88(12):2204-13.
Sokoloff, D. D., M. V. Remizowa and P. J. Rudall. 2013. Is syncarpy an ancestral condition in monocots and core eudicots? in Early Events in Monocot Evolution, Eds. P. Wilkin & S. J. Mayo. Cambridge University Press.