Anderson, D. T. 1992. Structure, function and phylogeny of coral-inhabiting barnacles (Cirripedia, Balanoidea). Zoological Journal of the Linnean Society 106: 277-339.
Barnacles are among the oddest animals on the planet. Genealogically speaking, they're crustaceans, but with such a highly derived morphology that, except for the jointed cirri (actually derived legs), you'd be hard pressed to find an obvious character marking them as such. Much is made of how barnacles spend their lives functionally upside-down (the legs are protruded out to filter food particles from the water, after which they are funneled downwards towards the mouth), their enviable reproductive organs, and how much pain a patch of them can cause while walking below the high-tide mark on a rocky coast. The most familiar barnacles are the rock-inhabiting pyramidal forms, but others seek out different habitats.
Barnacles living on corals have been assigned to three families, the entirely coral-living Pyrgomatidae, the genus Armatobalanus in the Archaeobalanidae and the genus Megabalanus in the Balanidae, but Anderson (1992) agrees with previous authors that the Pyrgomatidae and Armatobalanus form a single clade, with Armatobalanus representing the more ancestral form from which the more specialised Pyrgomatidae originated (the photo at the top of the page, from here, shows the opening of the pyrgomatid Nobia grandis with the cirri extended). Armatobalanus has a wall constructed of six plates as in other families of barnacle, while Pyrgomatidae show a trend towards fusion of the wall plates, with at most four and sometimes a single plate in the wall. The range of variation from more generalised to more specialised forms seen by Charles Darwin during his major revision of the world's barnacles was a significant factor in confirming Darwin's acceptance of the concept of transmutation of species, and the Pyrgomatidae are no exception. Coral-inhabiting barnacles run the gamut, from taxa that are merely resident on the coral and have a fairly typical barnacle morphology such as Armatobalanus, to the derived Hoekia monticulariae with fused opercular plates, vestigial cirri and enlarged mouthparts that feeds directly on the coral overgrowing it.
The greatest threat to any coral-living animal is being overgrown by the coral itself. In Armatobalanus, this is prevented using purely mechanistic means - as the cirri are extended from the aperture, they actively scrape away any overgrowing coral, and enlarged maxillipeds that also protrude from the aperture flick away the resulting debris. In the Pyrgomatidae, a frill has developed that protrudes from the aperture when open on either side of the cirri, and probably secretes a growth inhibitor that excludes the coral (the presence of some sort of chemical defense is indicated by the fact that dead barnacles are rapidly overgrown). Some species of the less derived pyrgomatid genus Cantellius, while possessing the apertural frill, also retain the teeth on the cirri and enlarged maxillipeds of Armatobalanus species. More derived pyrgomatid species show a trend towards reduction in size of the aperture, which Anderson (1992) suggests may be because that reduces the size of the perimeter the barnacle needs to keep clear of coral. The downside of aperture reduction is that it requires some degree of reduction in the size of the cirral fan, and hence reduces feeding efficiency. It has long been suggested that at least some pyrgomatids may compensate for the reduced feeding ability through some degree of parasitism from the host coral, either through tissue feeding or absorption of dissolved nutrients. While many species do show a trend towards weakening of or the development of pores in the basal shell or membrane separating the barnacle from its host (and this basis is completely lost in Hoekia), direct evidence for parasitism in genera other than Hoekia is slight. No evidence of nutrient transfer was found in a study of Newmania milleporum, but Anderson (1992) points out that Newmania is one of the more actively-feeding species, without a reduced basis, so is not one of the most likely candidates for parasitism anyway.
On the basis of morphology, Anderson (1992) suggested a phylogeny for the pyrgomatid subfamily Pyrgomatinae that placed the derived genera in three groups arising independently from the basal Cantellius, which was itself derived from Armatobalanus or an Armatobalanus-like ancestor. However, this phylogeny was not supported by the more recent molecular study by Simon-Blecher et al. (2007). In their phylogeny (shown above, from the paper - the drawings to the right of each taxon represent the arrangement of wall plates and the shape of the opercular plates), Armatobalanus is actually nested within the pyrgomatids (and one "pyrgomatid" genus, Wanella, seems to actually be a convergent member of the Balanidae). If the phylogeny of Simon-Blecher et al. (2007) is correct, then there appears to have been a fair degree of homoplasy in the fusion of the wall plates from the ancestral six retained in Armatobalanus. The most interesting possibility suggested to me by the molecular phylogeny, however, is that the mechanistic method of coral exclusion of Armatobalanus, rather than being ancestral, may actually be derived relative to the chemical inhibition method. Cantellius, the genus Anderson (1992) suggested retained relictual features of the mechanistic method, is sister in Simon-Blecher et al.'s (2007) tree to Armatobalanus, adding more credility to the idea that we should reverse our ideas of ancestral vs. derived.