Just over 40 years ago, I (Cox 1979) wrote a paper exploring symmetry and valve structure as taxonomic criteria for naviculoid diatoms, arguing that structure is more informative of relationships than valve shape and symmetry. As I pointed out then, the use of symmetry as a basis for diatom classification has historical origins and “worked tolerably well with the light microscope … it has been a useful character for species identification” (Cox 1979: 28). Although details of live structure were included in some early works (Kützing 1833, 1844; Ehrenberg 1843; Cleve 1894; Smith 1853, 1856), by the late 19^th century Schutt (1896) had adopted Smith’s (1872) work as a basis for diatom classification, which was followed by Hustedt (1927–1966, 1930) and subsequent workers (Hendey 1964; Patrick and Reimer 1966; Krammer and Lange-Bertalot 1986–1991). Yet, Smith (1872: 23) had clearly indicated that his classification was “an artificial key to facilitate diatom identification” rather than an attempt to reflect relationships.

The development of electron microscopy, particularly SEM in the latter 20^th century, revealed unsuspected diversity in diatom wall structure and facilitated the recognition of many new genera, or provided evidence for reviving some abandoned ones, particularly among naviculoid taxa, e.g. Berkeleya Grev., Brachysira Kütz., Climaconeis Grunow, Craticula Grunow, Diadesmis Kütz., Dickieia Berk. ex Grev., Encyonema Kütz., Placoneis Mereschk., Sellaphora Mereschk. (Round et al. 1990). Fourtanier and Kociolek (1999) documented the increase in new (validly published) diatom genera over time, showing the strongest increase occurring with the expansion of SEM studies (from 750 to 907 between 1979 and 1999) (Henderson and Reimer 2003). That increase has continued (Guiry and Guiry 2025; Kociolek et al. 2025) although the differences in actual numbers show that many of the validly published names (Fourtanier and Kociolek 1999) do not represent accepted genera (Guiry and Guiry 2025; Kociolek et al. 2025).

Thus, closer attention has been given to the significance of ultrastructural differences, particularly variation in raphe and areola structure, for generic delimitation, e.g. Aneumastus D.G.Mann & A.J.Stickle, Biremis D.G.Mann & E.J.Cox, Cavinula D.G.Mann & A.J.Stickle, Cosmioneis D.G.Mann & A.J.Stickle, Fallacia A.J.Stickle & D.G.Mann, Luticola D.G.Mann, Petroneis A.J.Stickle & D.G.Mann, Seminavis D.G.Mann in Round et al. (1990), Chamaepinnularia Lange-Bert. & Metzeltin in Lange-Bertalot and Metzeltin (1996), Humidophila Lowe et al. in Lowe et al. (2014). However, even with greater understanding of diatom structure, symmetry has still exerted a strong influence on how individuals have grouped taxa, particularly at generic level and above (e.g. Kulikovskiy et al. 2023; Guiry and Guiry 2025; Kociolek et al. 2025). Nevertheless, whilst advocating closer consideration of morphological characters, Kociolek et al. (2019) describe symmetry as a “feature of convenience” pointing out that “gomphonemoidness” is an example of homoplasy rather than homology, having evolved on at least six occasions.

Only more recently, with the development of molecular studies exploring phylogenetic relationships, have some conflicts in diatom classifications (Round et al. 1990), particularly at suprageneric levels (families and above) become more apparent (Medlin and Kaczmarska 2004; Kermarrec et al. 2011; Nakov et al. 2014). This paper explores how continuing SEM studies, better understanding of frustule ontogeny, and the results of molecular phylogenies have shown where reliance on shape and symmetry have misled interpretations of taxonomic relationships, with a focus on naviculoid diatoms. But it also shows that shape and symmetry are still being used to separate structurally similar taxa.

Historically, frustule symmetry has been used as one of the diagnostic characters for allocation both to genus and higher taxonomic groups, and within the naviculoid diatoms, valve shape and symmetry were easy guides to the discrimination of several genera, e.g. Gyrosigma Hassall, Pleurosigma W.Sm., Amphora Ehrenb. ex Kütz., Cymbella, Gomphonema (Hustedt 1930; Patrick and Reimer 1966; Barber and Haworth 1981). Although classifications have differed in the ranks accorded to groups of these genera, if naviculoid genera were grouped at family level, heteropolar taxa (Gomphonema, Gomphoneis Cleve, and Didymosphenia Mart.Schmidt) were placed in the Gomphonemataceae, while dorsiventral ones (Cymbella, Encyonema, and sometimes Amphora) were placed in the Cymbellaceae (Patrick and Reimer 1975). Interestingly within his Naviculaceae, Hustedt (1930) created a subfamily, the Gomphocymbelloideae, for the heteropolar and dorsiventral taxa but only with Round et al. (1990) was the order Cymbellales created, containing the Cymbellaceae, Gomphonemataceae, Mastogloiaceae, and Rhoicospheniaceae. The rationale for their new order rested on the shared chloroplast structure (Round et al. 1990), although by this time SEM had also provided new evidence of shared wall features within these families. There are, however, significant differences in some characters between the families, particularly with respect to areola and raphe structure.

In my earlier paper (Cox 1979), I also discussed both the similarities between areola structure in Gyrosigma, Haslea, and Navicula and the variation in internal raphe structure in Gyrosigma and Pleurosigma, later discussing the value of morphogenetic information for taxonomy (Cox 2010). Areola structure with hymenate internal occlusions, internal raphe fissures flanked by one or two longitudinal ridges and non-porous girdle bands are all shared across Navicula sensu stricto, Gyrosigma, Haslea, and Pleurosigma, although Gyrosigma and Pleurosigma were recognised primarily by their sigmoid shape and the angle of intersection of the striae. However, the valves of some Gyrosigma and Pleurosigma species are almost straight (Sterrenburg et al. 2015) and Du et al. (2023) argue that, based on their phylogenetic analysis, sigmoidality cannot be used as a criterion to define Pleurosigma. Similarly, Poulin et al. (2004) argued that, despite its sigmoid valve, Gyrosigma nipkowii F.Meister belongs in Haslea, sharing both structural and biochemical features with the latter.

Gradations in areola structure and in raphe features are seen across Navicula, Haslea, and Gyrosigma such that allocation of individual taxa to one or other of these genera has been somewhat contentious (Sterrenburg et al. 2015). In summarizing the taxonomic history of Haslea, Sterrenburg et al. (2015) emphasised the continuous longitudinal fissures (external areola openings) as a distinctive characteristic of the genus, while questioning that all members should be lanceolate in outline. They (Sterrenburg et al. 2015) emended the generic description but also provided morphological comparisons with Gyrosigma and Pleurosigma, emphasising the similarities in the wall construction, with the formation of a basal network of ribs over which silica strips are deposited and gradually widen, leaving narrow external fissures over the areolae. But whereas the external siliceous strips in Gyrosigma (and Haslea) are longitudinally oriented and build up from the vimines, in Pleurosigma the oblique orientation of the areolae means that the external layer builds up over vertical “pillars” on the basal layer, as a mesh or in a stellate manner (Sterrenburg et al. 2005) and, unlike in most pennate diatoms, a clear virga-vimen system is not found. Unfortunately, Sterrenburg et al. (2005) do not show any very early stages of basal layer formation, which might reveal how the oblique striation is initiated.

A transition from the presence of continuous slits over the areolae in Haslea, to external slits extending over several areolae in Gyrosigma and over single areolae in Navicula was postulated (Cox 1979: figs 10–12) and supported by observations (Cox 1999a: figs 40–42). It was also suggested that the formation of additional cross connections would explain the formation of double pores in some Navicula spp. (Cox 1999a: figs 44, 45, 52) and in Hippodonta Lange-Bert., Witkowski & Metzeltin (Cox 1999a: figs 12, 47, 48, 2002: figs 24.9, 24.10), which could be interpreted as the end of that developmental trajectory. Thus, the contrasting final external appearance of the areolae in these genera should not be interpreted as representing fundamentally different (non-homologous) characters, but as different stages along the same ontogenetic trajectory forming the areola system (Cox 2010). It is of course possible that double pores could be initiated in a different way, which would constitute another character.

Similarly, the variation in internal rib development beside the raphe sternum within the Naviculales can be considered variants on the same basic pattern. In considering the morphological variation in and between Navicula, Haslea, and Gyrosigma, possible developmental pathways for both raphe and accessory rib formation in these genera have been presented (Cox 2002: figs 24.5, 24.7, 24.8). However, although when evaluating Haslea valve morphology in relation to that of Gyrosigma and Pleurosigma, Sterrenburg et al. (2015) mentioned the structure of the accessory rib in Haslea, they did not discuss internal raphe structure in detail, simply noting the presence of a similar rib in Gyrosigma. Rather they (Sterrenburg et al. 2015) focussed on the sandwich-like structure of the valve as seen in Gyrosigma and Pleurosigma discussed above. However, if the rib system beside the raphe sternum is considered (Cox 2002: fig. 24.5), degrees of development of single and double accessory ribs can be seen across Navicula, Haslea, Gyrosigma, and Pleurosigma. But, whereas the first three genera show greater accessory rib development on the primary side of their valves, in Pleurosigma there is more or less equal development of ribs on either side of the central nodule, restricted to the central part of the valve.

Cardinal et al. (1989) compared the degree of development of central bars in Donkinia, Gyrosigma, and Pleurosigma and discussed its taxonomic value at the species level for Gyrosigma and Pleurosigma. Donkinia has a relatively small central nodule flanked by short, very thick bars (Cox 1983a: figs 40–42; Cardinal et al. 1989: fig. 1). In Gyrosigma and Pleurosigma, the central nodule is usually somewhat more elliptical and the flanking bars are usually narrower (Cardinal et al. 1989: figs 1–38). In Pleurosigma, the central bars are more or less the same length, in Gyrosigma, the bar on one side is often longer or more strongly developed than on the other (Cox 1977: figs 26–29, 1979: figs 6, 7, 9; Cardinal et al. 1989: figs 4, 7–9, 11–13, 16–18). Similar asymmetry is seen in Haslea and Navicula (Cox 1999a: figs 59, 63–76, 2002: fig. 24.5) suggesting a closer relationship between these genera and Gyrosigma, less so with Pleurosigma and Donkinia (see Lobban et al. 2020 below).

One answer to this question is that all are potentially taxonomically informative, but care should be taken that their significance is not based on a priori assumptions or personal preference, perhaps influenced by the degree of familiarity with particular groups of taxa. In revising their concept of Haslea, Sterrenburg et al. (2015) suggested that earlier descriptions (Round et al. 1990; Massé et al. 2001; Talgatti et al. 2014), which all mentioned the accessory rib along one side of the raphe sternum, did not “accurately describe the essential point of the Haslea valve morphology”, stressing instead the sandwich-like structure of the valve, like that of Gyrosigma and Pleurosigma. However, despite suggesting that the valve formation process in Pleurosigma has been fully documented (Sterrenburg et al. 2015), the earliest stages in basal layer development were not illustrated (Sterrenburg et al. 2005). Thus, it remains unclear how the oblique striation of Pleurosigma is established, whereas the perpendicular intersection seen in Gyrosigma and Haslea follows from the sequential formation of virgae and vimines (as in other naviculoid diatoms, Cox 1999b).

In disregarding the significance of the presence of an accessory rib that often partially overlaps the raphe sternum, Sterrenburg et al. (2015) considered areola structure rather than raphe details a diagnostic character for Haslea, placing several taxa in that genus, although their accessory ribs do not overlap the raphe sternum (cf. Cox 1977: figs 26, 38). Two such taxa, Haslea tsukamotoi Sterrenburg & F.Hinz and H. avium M.A.Tiffany, Herwig & Sterrenburg, were transferred to Navicula by Li et al. (2017) who also pointed out that the helictoglossae are elongate and narrow in Haslea, shorter and more rounded in Navicula. This transfer is supported by phylogenetic analyses (Li et al. 2017: figs 2, 3; Lobban et al. 2020: fig.1), which clearly place these taxa within the Navicula clade. On the other hand, despite its sigmoid valve outline, the transfer of Gyrosigma nipkowii F.Meister to Haslea (Poulin et al. 2004) is supported by the accessory rib overlapping the raphe sternum as well as other morphological, biochemical and molecular evidence.

It is clear that valve shape and stria orientation alone do not allow reliable separation of Pleurosigma, Gyrosigma, Haslea, and Navicula, but greater attention must be given to raphe features, particularly accessory and central bars. Pleurosigma and Gyrosigma species can lack sigmoidality, whereas Haslea species may be straight, sigmoid or arcuate. A pseudostauros may or may not be present, while some Navicula species have Haslea-like striae, with longitudinal strips over the external valve surface (Cox 1999b). There are also differences in the shape of the helictoglossae, elongate in Haslea versus short in Navicula (Li et al. 2017). Although Sterrenburg et al. (2015) discounted the significance of an accessory rib overlapping the raphe sternum, this would seem to be a generic characteristic of Haslea, even if the degree of overlap may vary. On the other hand, its absence, the ribbon-like chloroplasts and more or less symmetrical central bars in Haslea alexanderi Lobban & C.O.Perez may indicate that this species, despite its shape and perpendicularly intersecting striae, belongs with Pleurosigma. Unfortunately, the authors (Lobban et al. 2020) were unable to obtain molecular data for this species and therefore could not include it in their phylogeny.

Phylogenetic studies using genetic data are providing another perspective on diatom relationships although taxonomic coverage remains patchy and not all taxa are amenable to culturing, which is often required to obtain sufficient material for extraction and analysis. Whereas genetic data have often been used to differentiate taxa (within and between species, genera) (Jahn et al. 2019; Majewska et al. 2022; Abarca et al. 2023), they have also provided support to confirm morphologically inferred relationships (Poulin et al. 2004; Kermarrec et al 2011; Stepanek and Kociolek 2014; Li et al. 2017). However, some rather unexpected relationships are also being revealed that should stimulate re-examination of the criteria on which taxa have been discriminated, in particular some of the more obvious features, such as frustule and valve shape and symmetry.

In exploring the relationships of raphid taxa with a stauros, Ashworth et al. (2017) created a new genus, Druehlago, distinguished from Craspedostauros by its stalked habit, cuneate frustules, heteropolar valves, and multiple lenticular chloroplasts. They (Ashworth et al. 2017) discussed the similarities between Craspedostauros and Achnanthes Bory, which their molecular phylogeny showed as closely related, also pointing out that, whereas Achnanthes (like Craspedostauros) usually had two fore and aft chloroplasts, it may also have many lenticular chloroplasts as seen in Druehlago. Despite their similarities, Majewska et al. (2022) supported the recognition of Druehlago as distinct from Craspedostauros, but a recent phylogeny from Sugawara et al. (2024) places Druehlago within Craspedostauros, in a clade with Craspedostauros and Achnanthes (Sugawara et al. 2024: appendix S2). Sugawara et al. (2024: fig. 4a) also show a mature auxospore of Druehlago in which the chloroplast configuration resembles the fore and aft arrangement of Craspedostauros and Achnanthes, in contrast to the initial and vegetative cells with many chloroplasts (Sugawara et al. 2024: figs 1c, 4b). The position of Druehlago within Craspedostauros argues against its recognition as a distinct genus and again illustrates the unreliability of symmetry as a basis for generic allocation, although not all workers would agree (Majewska et al. 2022).

This group of taxa also illustrate the potential that frustule shape can be modified while retaining the same structural features, i.e. areola, raphe and cingulum construction. Thus, Craspedostauros species are usually symmetrical about three axes, whereas Druehlago is heteropolar and Achnanthes has flexed frustules and has become functionally monoraphid. Achnanthes and Druehlago share a stalked habit, whereas Craspedostauros is epipelic or epibiontic, albeit without any evidence of stalks (Majewska et al. 2021). One species, Craspedostauros danayanus Majewska & Ashworth is reported to attach by one end of the valve, but there is no evidence of specialised apical pore fields (Majewska et al. 2021).

Shifts in frustule shape but retention of areola, raphe and cingulum structure are also shown by the phylogenetic relationships of Navicula, Pseudogomphonema, Seminavis, and Rhoikoneis (Nakov et al. 2018; Lobban et al. 2020). Whereas Navicula species are isopolar with straight girdle regions, Pseudogomphonema is heteropolar with slightly cuneate girdle regions, Seminavis has a dorsiventral valve with a slightly asymmetric girdle, and Rhoikoneis has isopolar valves with a flexed girdle region. Yet they all share the same areola structure, typical Navicula-like raphe systems and plain (non-porous) girdle bands. Structure is a better guide to their relationships than cell shape and symmetry.

A recently described new biraphid genus, Yuzaoea Chenhong Li, Honghan Liu, Yahui Gao & Changping Chen, is compared to Rhoicosphenia (Liu et al. 2024), from which it differs in being isopolar rather than heteropolar and having fully developed raphe systems on both valves but sharing a flexed girdle view. Based on rbcL and SSU rRNA gene datasets the genera form a clade sister to several monoraphid genera, supporting the hypothesis that progressive reduction in the raphe system, from biraphid to monoraphid has occurred on more than one occasion (Kulikovskiy et al. 2016), ending in the araphid state (Kociolek et al. 2013). To what extent the loss of a raphe system on one valve is also linked to the development of valve flexure seen in many monoraphid genera remains unexplored.

As mentioned above, molecular studies (Poulin et al. 2004; Li et al. 2017; Lobban et al. 2020) are helping to clarify the affinities of taxa within the Pleurosigma-Gyrosigma-Haslea-Navicula complex. Lobban et al. (2020) and Li et al. (2017) showed that some taxa (H. tsukamotoi, H. avium) (accessory rib overlapping raphe sternum and shorter rather than longer helictoglossae) placed in Haslea by Sterrenburg et al. (2015) fell within the Navicula clade. They are also confirming the suggestion that genera such as Pseudogomphonema (Medlin 1991) and Seminavis (Round et al. 1990), with different valve and frustule symmetries but shared areola and raphe characters, belong with Navicula in the Naviculaceae (Lobban et al. 2020) rather than with dorsiventral or heteropolar taxa in the Cymbellales (Round et al. 1990). Other molecular studies (Kermarrec et al. 2011; Nakov et al. 2018; Jahn et al. 2019) are confirming that dorsiventrality or heteropolarity should not be relied upon to allocate to families in the latter order, confirming my earlier suggestions based on morphological and morphogenetic evidence (Cox 2002). Details of raphe and isolated pore structure (stigmata or stigmoids, Cox and Van de Vijver [2024]) are better guides to relationships within the Cymbellales than simple shape or symmetry.

In their work on “stigmata”, Cox and Van de Vijver (2024) discussed the distribution of different types of isolated pores across raphid diatoms, including the occurrence of fistulae, first described from the small, lightly-silicified cells of Fistulifera Lange-Bert. (Lange-Bertalot 1997). Phylogenetic analyses (Gastineau et al. 2018; Majewska et al. 2019; Kim et al. 2020; Tseplik et al. 2022) placed this genus in the same clade as Proschkinia Karayeva, a relationship that would not have been expected based on valve morphology, apart from the shared presence of a fistula. The relationships of taxa with a buciniportula (Olifantiella Riaux-Gob. & Compère, Labellicula Van de Vijver & Lange-Bert., and Luticola) have not been investigated using molecular data although, based on partial 18S rRNA sequencies, a putative Olifantiella species grouped with Luticola (Han et al. 2018; Kezlya et al. 2021). However, both cases (fistulae and buciniportulae) would benefit from closer morphological comparisons.

Whereas shape and symmetry are useful, easily described characters to aid identification, as recognised by Smith (1872), it is clear they are not reliable characters for revealing systematic relationships, being homoplasious rather than homologous. Closer morphological investigations and molecular phylogenetic studies show that the same underlying structure is found in taxa exhibiting different frustule shapes and symmetries (Poulin et al. 2004; Lobban et al. 2010; Kermarrec et al. 2011; Nakov et al. 2014, 2018; Sugawara et al. 2024). Within the unique diatom reproductive cycle involving gradual size reduction followed by size restitution via auxosporulation, shifts in valve shape can occur and have resulted in different names being applied to larger and smaller representatives of the same taxon (Rose and Cox 2014). On the other hand, the flexibility of the auxospore, perizonium and initial cell provide a potential opportunity to introduce more striking shape changes, which would then persist through vegetative reproduction (Mann 1982b; Cohn et al. 1989b; Toyoda et al. 2006; Idei et al. 2013).

Two examples of species that can produce very different shapes are Phaeodactylum tricornutum Bohlin and Centronella reicheltii Max Voigt. Whereas P. tricornutum is unusual in not producing a typical siliceous frustule, C. reicheltii has a triradiate frustule that can be reproduced through a series of mitotic divisions (Krieger 1927; Schmid 1997), although it can revert to a fragilarioid form by gradual reduction of one “arm” of the triradiate cell (Schmid 1997). Schmid (1997) suggested that the triradiate morph is a teratological form induced in the auxospore and initial cell, nevertheless it has been recorded from a range of lakes and can produce sizeable populations (Echenique and Guerrero 2004). However, it is unlikely that raphid diatoms could produce tri-radiate frustules because this would require the development of a third raphe fissure from a bi-polar pattern centre (initiating the raphe system), whereas dorsiventrality, heteropolarity or flexure of the apical axis are not precluded. Nevertheless, irregularities in the structure of initial cells (Mann 1989; Nagumo 2003; Toyoda et al. 2006; Levkov 2009; Mann and Pouličková 2010) indicate that establishing the integrity of the cytoskeleton may take one or two mitoses and could offer an opportunity for a shift in shape.

How readily valve and frustule shape can change in natural populations is unclear, but presumably such changes are mediated by cytoplasmic components, nuclear movement, actin and filaments and microtubules of the cytoskeleton (Schmid 1986, 1994). Once a shape change is initiated it can then be perpetuated through subsequent mitoses thanks to the constraints of the siliceous frustule, at least until the next auxosporulation. Work with cytoskeleton inhibitors has shown that disrupting their activity results in abnormal valve morphology, particularly raphe path and stria arrangement, although areola structure seems to be maintained (Blank and Sullivan 1983a, 1983b; Cohn et al. 1989a). Studies on diatom teratologies similarly show that shape, raphe and stria patterns are often disrupted but effects on areola structure have not been reported (Lavoie et al. 2017).

In considering which characteristics are taxonomically informative but not susceptible to change over the life cycle, the following should be considered: areola structure, including the type of occlusion, e.g. hymenate or cribrate; raphe sternum structure and any associated thickenings; type of girdle bands, e.g. with or without rows of pores, type of pore occlusion. Despite still often being ignored, chloroplast structure can be informative, particularly at genus and family level, although the same chloroplast form may occur with different areola, raphe and cingulum characters. For example, within the Cymbellales sensu Round et al. (1990) (all with a single plastid with a central pyrenoid), there is a distinction between the Anomoeoneidaceae + Rhoicospheniaceae and the Gomphonemataceae + Cymbellaceae, the former families having hymenate areola occlusions that are lacking in the latter. On the other hand, variation in chloroplasts has been shown to be indicative of the different clades within Amphora sensu lato (Stepanek and Kociolek 2014).

Members of the Naviculaceae share the same type of areola, internally occluded by hymenes with an external slit, simple girdle bands lacking any pores, two chloroplasts lying along each side of the girdle, and the raphe fissure opening laterally in a distinct raphe sternum, often with some form of accessory bar(s) present. But the shape and symmetry of the genera vary from bilaterally symmetrical and isopolar valves with straight girdle regions (Haslea, Navicula), to bilaterally symmetrical and isopolar valves that are flexed in girdle view (Rhoikoneis), to heteropolar valves with slightly cuneate girdles (Pseudogomphonema), to dorsiventral valves with biconvex girdles (Seminavis), to sigmoid valves with straight girdle regions (Haslea, Gyrosigma). Yet all could be derived from an ancestor that was symmetrical about all three axes. Similarly, vaulted and keeled frustules could be derived from the same ancestral type.

In using structural characters to explore systematic relationships it is important to distinguish between homologous and homoplasic characters, often requiring a knowledge of their ontogeny rather than relying simply on visual comparison of mature valves. It is also important to describe characters accurately using appropriate terminology (Cox 2010; Cox and Van de Vijver 2024), e.g. the distinction between a stauros and a pseudostauros, which in turn often requires SEM data. LM observations alone may be inadequate.