The history of diatom classification and its relationship to phylogeny is as complex as that of any other taxon [30]. It merits study [32] and [33]. Some interpretations understand it to harbour a rich history of varied intellectual development (see [32] and [33]), while others see it of less or even no significance: “History suggests that, in many cases, diatomists might just as well have been classifying scraps of wallpaper as diatoms” because “[D]iatom taxonomy has developed largely without a conceptual basis, using a restricted range of characters drawn from just one part of the phenotype (the valve)” ([13] (p. 482)). The latter idea is at best unfounded ([32], and below). Yet there has been a divorce, so to speak, of representations of phylogeny and its accompanying classification, rather than a lack of any “conceptual basis” to the investigations [32].

To begin, an early explicit phylogenetic diagram for diatoms was offered by Hippolyte Peragallo in 1897 [21] (Fig. 3). He presented a scheme depicting a series of diatom genera, their names enclosed in separate boxes, each abutting and inter-linking with other such boxes. For example, the genus Fragilaria was ‘connected’ to Plagiogramma, Synedra and Diatoma. The impression gained (and, perhaps, meant to be conveyed) is that Fragilaria ‘gave rise’ to Synedra, which in turn ‘gave rise’ to Eunotia (Fig. 3). ‘Gave rise’ being interpreted as one taxon being ancestral in some way to another, such that, in the example above, Fragilaria is ancestral to Synedra.

Another phylogenetic diagram, presented by Merezhkowsky in 1903, included hypothetical ancestral taxa, those based not on specimens but supposition: Archaideae is the ancestor of the raphid diatoms (“mobilées”), Protonées is the ancestor of the Archaideae, Copuloneis is the ancestor of Protonées and Tabellarioideae, with Urococcus as ancestor of all diatoms (Fig. 4 after [17]). Merezhkowsky also included some previously described taxa in ancestral positions: Melosireae leading to (ancestral to) the Anaraphideae, Fragilarioideae leading to (ancestral to) Copuloneis, Auricula leading to (ancestral to) the Carinatae and Libellus leading to (ancestral to) Polyplacatae (Fig. 3; Merezhkowsky revised his classification with the Archaideae added to the “mobilées” and the Anaraphideae re-named Centrales, a version of the ‘centric’ diatoms referred to above, ([18] (pp. 203–204)); some years later, Peragallo proposed a classification based on a combination of his earlier diagram and that of Merezhkowsky [22], see [32]).

Like Peragallo, the Italian biologist Achille Forti developed Merezhkowsky's ideas. Forti chose to represent genealogical relationships as a series of networks, some with reticulating lines (Fig. 5, from [5]). It might be reasonable to assume that in Forti's diagram if a genus is ‘connected’ by lines to another genus that is to be interpreted as the one having ‘given rise’ to the other, as in the above examples of Peragallo and Merezhkowsky. In comparison to Peragallo's diagram, Forti's is upside down. Nevertheless, it still implies that Fragilaria ‘gave rise’ to Synedra as well as Himantidium and Eunotia (Fig. 5; Himantidium became a synonym for Eunotia).

In more recent times, diatom phylogenetic diagrams continued attempting to represent certain aspects of the notion that some taxa ‘give rise’ to others, with the ancestral taxon often left unnamed, as in Simonsen's ‘pedigree’ ([26] ( “uncertain - not common - ancestors”)) or else diagrams with a mixture of unnamed (and unknown) ancestors and fossil taxa (Fig. 6, after [19]). Of significance is that most of the early schemes (Peragallo, Merezhkowsky, and Forti) are based on a whole suite of data (the valves, the plastids, reproductive behaviour, etc.) not “just one part of the phenotype (the valve)” ([13] (p. 482)). If anything, the “conceptual basis” was an attempt to ‘find’ and represent direct ancestry via supposed relationships, a view traceable to Ernst Haeckel, the creator of such phylogenetic trees [34].

Thus, or so it seemed, phylogeny was understood to be complex, requiring the identification of (real or imaginary) ancestors and some taxa could reasonably be assumed to have ‘given rise’ to other taxa – this is a “conceptual basis” of sorts.

Diatom phylogeny, as revealed by molecular data, is certainly progressing from its humble beginnings. Its most comprehensive ‘estimate’ can be found in Medlin and Kaczmaska [15] (see also [1]) based on ca. 9000 sequences. While these might be useful places to examine progress (see below), it is worth noting that there are numerous diatom phylogenies (most, if not all, based on 18S RNA) in various papers and book chapters [2], [9], [10], [11], [15], [23], [28] and [29]. Still, even with such vast amounts of data depicting relationships precisely, curiously accounts of phylogeny continue to have as its “conceptual basis” the idea that some taxa gave rise to others, even though the phylogenetic diagrams specify only sister-group relationships. Thus such statements appear, derived from representations of diatom phylogeny:

“Radial centric diatoms begot multipolar centrics, multipolar centrics begot pennates and the araphid pennates begot the raphid pennates” ([9] (p. 92)).

However that sentence is supposed to be interpreted, more puzzling is the straight narrative accounts, stories focusing on this or that ancestor, tales of what actually happened:

“It is of interest that the molecular phylogeny of diatoms also favours the centric forms as ancestral … with some molecular evidence agreeing [data] that the earliest diatoms could have been neritic … Calibrating the molecular clock of diatoms and other Heterokontophyta from the fossil record, and applying the same nucleotide substitution rate to parts of the molecular phylogeny before the occurrence of fossil diatoms, indicate that diatoms originated close to the Permian–Triassic boundary 250 Mya, some 130 million yr earlier than the first silicified fossil diatoms … This suggests that silicification evolved late in the evolutionary history of diatoms. It is possible that any earlier silicified diatoms have failed to be preserved for some environmental or taphonomic reasons; however, silicified organisms (sponges) have been found in phosphoritic marine sediments 580 million yr old … and both radiolarian and sponges were common from the Cambrian onwards …” ([24] (p. 44)).

One wonders what – scientifically – might be made of these kinds of statements and what might be made of their “conceptual basis”. Perhaps one should simply be amazed at the precision and move on.

Taxon relationships, as expressed in phylogenetic diagrams (regardless of the source of data), are expressed as sister-groups: Taxon A is more closely related to Taxon B than it is to C. This is different to stating that Taxon A ‘gave rise to’ (begot) Taxon B. The contrast being, the former is a statement of relative ancestry, the latter of direct ancestry.

Given the number of different diatom phylogenies available and the number of differences between the various trees, it is difficult, if not impossible, to summarise relationships with any accuracy, notwithstanding efforts to create a new classification that includes possibly (or possibly not) paraphyletic groups [1] and [16]; for commentary, see [2] and [33].

Given the significance, in terms of diversity, of pennate diatoms (those with bilaterally symmetrical valves, Bacillariophyceae, as emended in [25] plus Fragilariophyceae), their sister-group relationship is of some import. The results in Sorhannus [29] supported a sister-group relationship of the pennate diatoms with a group that included Thalassiosirales (a radially symmetrical group) plus Cymatosirales (a bilaterally symmetrical ‘centric’ diatom) (Fig. 7a); a similar relationship was discovered recently in [23], although the Thalassiosirales and Cymatosirales were less well defined (Fig. 7b); in that tree the Cymatosirales include Hyalosira sp. (usually part of Fragilariophyceae) and the Thalassiosirales are sister to Lithodesmiales (which includes genera such as Mediopyxis, Lithodesmium, Ditylum, Bellerochea, Streptotheca, Lithodesmioides and Neostreptotheca; for further discussion see [7]).

Medlin and Kazsmarska [15] (and reflected in the classification of Mann in [1]) found only the Cymatosirales as sister to the pennate diatoms (Fig. 8), while Sinninghe Damsté et al. [28] had only the genus Gonioceros as sister to the pennate diatoms (Fig. 9). Gonioceros is in the order Chaetocerotales, part of the Chaetocerotophycidae, the latter being divided into the Chaetocerotales and Leptocylindrales ([25] (p. 127)). The Chaetocerotales have three families, the Acanthocertaceae (with only one species, A. magdeburgense Honigmann), Attheyaceae (with only one genus Attheya) and the Chaetocerotaceae with three genera, Chaetoceros, Gonioceros and Bacteriastrum. Gonioceros may be more closely related to Attheya than any other member of the Chaetocerotaceae and all species recognised in either Attheya or Gonioceros should be in “one genus in the family Attheyaceae …” [4]. Interestingly enough, Sinninghe Damsté et al. [28] found Chaetoceros to be most closely related to the Cymatosirales, whereas in Medlin & Kazsmarska, Chaetoceros is basal to the Cymatosirales [15].

In short, basal pennate relationships remain obscure and uncertain – and the groups so implicated as closest relatives are primarily composed of fossil taxa.

The fact that molecular relationships cannot as yet be succinctly summarised implies that molecular data are either insufficient to yield robust relationships or the samples taken so far have little useful phylogenetic signal. For morphology, there have been no good morphological synapomorphies identified or proposed for any higher taxon (beyond the presence of a raphe in the raphid diatoms, a group nested within the pennate diatoms). Nevertheless, for relationships within the pennate diatoms, it is possible to reduce options to taxa (Class, sub-class, orders) whose valves are bilaterally symmetrical in a strict sense. Comparison of Orders in the (probably) paraphyletic Fragilariophycideae, Biddulphiopycidae and Mediophyceae is presented in Table 1 (to document overlap in their composition). The table can be understood in two parts: (1) among the first nine orders listed (Fragilariales to Protoraphidales) is the probable sister-taxon to the raphid diatoms; (2) among the remaining 13 orders (Biddulphiales to Triceratiales) is the probable sister-taxon to the pennate diatoms. Of the total, only four Orders lack fossil representatives. Yet, conversely, a glance at the families and genera listed in the Biddulphiopycidae (sensu [20]), for example, the vast majority appear extinct. Thus, the Biddulphiales (sensu [20]) has three families, Biddulphiaceae, Gryocylindraceae, and Parodontellaceae. Ignoring the Parodontellaceae for the moment (see below on this family [8]), there are 15 genera in the Biddulphiaceae and Gryocylindraceae is monotypic [20]. Of these 16, nearly half are extinct.

Table 1 does not include details of all known extinct groups, whose classification varies, such as artificial monotypic groups harbouring single species, such as Bilingua rossii Gersonde et Harwood [6] placed, redundantly, in its own Order (Bilinguales) and family (Bilinguaceae) [20] (there may be other species in the genus, see [31]). There are also larger groups such as the Cymatosirales, its significance noted above. The Cymatosirales presently has two families, the Cymatosiraceae and the Rutiliariaceae, the former with 14 genera, of which four are composed entirely of extinct species, the latter with five genera, of which four are composed entirely of extinct species. In terms of numbers of species, around 50% of the Cymatosiraceae are extinct and nearly all (57 of 59) species in the Rutiliariaceae are extinct. There is the possible addition of a further family, the Paradontellaceae, of which all of its four genera are composed entirely of extinct species (see [20] and [32], placing this family in the Biddulphiopycidae; for further comment, see [32]).

It is more than possible that Gonioceros, Chaetoceros and the Cymatosirales, referred to above, share putative synapomorphies, such as the possible homology (relationship) between the seta of Gonioceros and Chaetoceros (seta are long spines, extending from the poles of the valves) and the pili of the Cymatosirales (pili are hair like extensions emerging from the valve surface, at their poles), and in turn these may be homologous to related structures in the Biddulphiopycidae (sensu [20]). These notions require investigation.

In spite of much (molecular) effort, phylogenetic relationships of diatoms remain poorly understood and poorly represented in classifications [33]. Few, if any, higher taxa have associated, or even hypothesised, morphological synapomorphies. Mann offered the following comment: “Molecular data … show the centrics as a whole to be paraphyletic, but relationships between the principal groups, and whether particular groups are monophyletic or paraphyletic, is currently unclear. Several major molecular clades are cryptic, with no or few morphological or life history traits that can be convincingly argued to be synapomorphies” ([1] (p. 431, note 18)). For characters to be “convincingly argued as synapomorphies”, analysis is required rather than judgement – and no relevant morphological data has yet been analysed in such a way to allow any degree of conviction. Such analyses are possible and, in the near future, hopefully, will be published. If the results conflict with current molecular understanding, so be it – that outcome would identify a problem worthy of investigation: why do morphology and molecules differ? However, the central purpose of this contribution is to point out that without consideration of fossil taxa (extinct morphology) much of the diversity that existed between, as it were, the pennate diatoms and all the remaining diatoms, and the diversity that existed between the raphid diatoms and the remaining pennate diatoms is extinct and, of course, inaccessible to direct investigation with molecular methods. Of course, one might entertain the following thoughts:

“Thus, by the end of the Lower Cretaceous, cylindrical, circular, bipolar and possibly triangular forms were present, possessing linking spines and six types of processes including the rimoportula. Thus, the ancestors of both Clades 1 and 2 almost certainly were present. Sulci, costae, a marginal ridge, a pseudoloculate and perforate valve structure also have been identified, together with signs of vela” ([27] (p. 374)).

Without identifying which characters are derived and which are primitive (discovering synapomorphies), relative to extant forms, the above quotation simply presents a further excursion into the narrative phase of phylogeny, an episode abandoned by systematists (including molecular systematists and palaeontologists) decades ago.