Who’s who in Magelona: phylogenetic hypotheses under Magelonidae Cunningham & Ramage, 1888 (Annelida: Polychaeta)

Phylogenetic hypotheses and the myth of evidential support

The present study does not assess the veracity of either the Magelonidae or less inclusive phylogenetic hypotheses in terms of evidential “support,” nor are there attempts to use popular methods such as the bootstrap or Bremer analysis to assert that hypotheses are either “strongly supported” or “robust.” Such perspectives have become all too common in systematics research, including those on annelids (e.g., Struck et al., 2011; Borda et al., 2012; Weigert et al., 2014; Aguado et al., 2015; Andrade et al., 2015; Struck et al., 2015; Weigert & Bleidorn, 2016; Gonzalez et al., 2018; Nygren et al., 2018; Shimabukuro et al., 2019; Martín et al., 2020; Stiller et al., 2020; Tilic et al., 2020; Gonzalez et al., 2021). Reasons for not speaking of support in the present study follow from the basics of reasoning described above. More detailed treatments of misconceptions regarding evidence of, and support for phylogenetic hypotheses can be found in Fitzhugh (2012), Fitzhugh (2016b), Fitzhugh (2016c).

First, to correctly speak of evidence means to refer to the premises that allow for a certain conclusion (Longino, 1979; Salmon, 1984a; Achinstein, 2001). Whilst that evidence is said to “support” a given conclusion, it is only inductive, i.e., test evidence or support (cf. [3]) that matters in the pursuit of empirical evaluation, whether in matters of everyday life or during scientific inquiry. An emphasis on test evidence certainly permeates all of evolutionary biology and by extension it is understandable that it would be of concern in systematics. But this requires being clear about (1) what is meant by the test evidence, (2) if that evidence is being used in the proper context, and (3) whether or not bootstrap and Bremer “support” values are meaningfully interpretable in relation to phylogenetic hypotheses.

Recall that the examples of ab-, de-, and induction in (1)–(3) draw clear distinctions between premises and conclusions derived from those premises. Those premises constitute the evidence for the respective conclusions [see also (4)–(6)]. Beyond basic matters of logic, we need to be cognisant of the way evidence is interpreted in the separate stages of inquiry. Evidence in relation to abduction is unremarkable in the sense that conclusions must be as they are because of the premises [cf. (1), (4), (5)] (Fitzhugh, 2010a; Fitzhugh, 2012; Fitzhugh, 2016a; Fitzhugh, 2016c; Fitzhugh, 2016d). As well, abductive conclusions cannot be changed or “defeated” with the introduction of additional effects to be explained. Adding those effects as premises only results in the inference of a new set of hypotheses that have no relevance to previously inferred hypotheses.

A related misconception is that statistical consistency applies to phylogenetic inference (e.g., Felsenstein, 1978; Felsenstein, 2004; Swofford et al., 1996; Heath, Hedtke & Hillis, 2008; Assis, 2014; Brower, 2018). When correctly interpreted, consistency is the view that as more and more test evidence is introduced in support of a hypothesis, the closer one is supposedly getting to a true proposition. The mistaken interpretation of consistency in systematics has led to the popular yet erroneous view that sequence data are beneficial for discerning “phylogenies” because of the vast increase in characters that are available (e.g., Goto, 2016; Tilic et al., 2020). For example, in relation to phylogenetic inference, Felsenstein (2004: 107, 121, respectively) states,

“An estimator is consistent if, as the amount of data gets larger and larger (approaching infinity), the estimator converges to the true value of the parameter with probability 1…”

“The inconsistency of parsimony [sic] has been the strongest challenge to its use. It becomes difficult to argue that parsimony methods have logical and philosophical priority, if one accepts that consistency is a highly desirable property.”

Contrary to claims that consistency is relevant to phylogenetic inference, it has been acknowledged since the early 20th century that continued additions of effects to abductive inferences provides no indication that a true, as opposed to plausible, set of explanatory hypotheses has been attained (Peirce, 1902; Peirce, 1932: 2.774–777; Rescher, 1978; Fitzhugh, 2012; Fitzhugh, 2016b). As new effects in need of explanation are sequentially added to abductive premises, each subsequent conclusion will only offer a new set of hypotheses that have no relevance to previously inferred hypotheses. The emphasis on consistency in systematics is yet another consequence of abduction not being recognised in lieu of a misplaced statistical (inductive) mindset.

Deduction serves to conclude predictions of potential test evidence [cf. (2)] that might be later sought during the process of testing. Much like premises-as-evidence in abduction, the premises in deduction do not warrant any special attention in terms of “support.” Where evidence is of importance is during the act of testing, i.e., induction [cf. (3)]. For instance, a conclusion that test evidence supports or confirms a hypothesis can be subsequently revised if additional support through further testing is obtained, or additional test evidence might eventually lead to a conclusion of disconfirmation. As noted by Lipton (2005), the process of testing puts hypotheses at risk of being disconfirmed since there is no guarantee predicted test evidence, via (2) and (3), will be found. Obtaining contrary evidence is always a possibility. It is test evidence that matters when speaking of support. Abductive “support” lacks that qualification, and character data are not test evidence.

Because an emphasis on causality has been largely wanting in recent decades in relation to inferences of taxa-as-explanatory-hypotheses, references to support for phylogenetic hypotheses has centered on support [sic] for “groups” or “clades” within cladograms (Fitzhugh, 2012). Such claims of support are misleading for two reasons. First, as mentioned above, “support” for any abductive conclusion is trivial since the conclusion could not be otherwise given the premises. In other words, unlike test evidence associated with induction, observations of characters do not present any risk to the hypotheses explaining those characters. To assert, for example, that characters “support” some phylogenetic hypothesis verges on circularity: the conclusion(s) is/are determined by the premises and the premises support the conclusion(s). Second, our concern is not for groups, but rather various explanatory hypotheses. As all phylogenetic hypotheses are composite (Fig. 3), an emphasis on groups is both artificial and denies consideration of the variety of explanatory accounts implied by cladograms. Support in relation to groups is epistemically meaningless.

There is a third difficulty, faced by the two popular “support” measures for phylogenetic hypotheses [sic] (Efron, 1979; Efron & Gong, 1983; Felsenstein, 1985; Felsenstein, 2004; Bremer, 1988; Bremer, 1994; Efron & Tibshirani, 1993; Davis, 1995; Efron, Halloran & Holmes, 1996; Holmes, 2003; Soltis & Soltis, 2003; Fitzhugh, 2006a; Fitzhugh, 2012; Fitzhugh, 2016c; Lemoine et al., 2018). Neither method provides epistemically meaningful values regarding evidential support, contrary to the popularity of their use, or claims that they provide a basis to say phylogenetic hypotheses are “robust.” The bootstrap was originally developed to test statistical hypotheses through a process of random resampling from an original set of random samples taken from a population. The intent is to test the hypothesis that a statistical parameter, inferred from the original samples, has the hypothesised value, without having to perform additional sampling from the population. It is important to note that the bootstrap approach is designed to test statistical, not explanatory hypotheses. Phylogenetic hypotheses are not statistical constructs, as clearly shown in (4) and (5). Like explanatory hypotheses, however, hypotheses of statistical estimates are inferred by way of abduction. Such estimates are generalisations that account for observed instances as representative of the population from which the sample was drawn. For example,

Testing the hypothesis would proceed by predicting what should be observed if additional random samples are taken. Such an inference could not be deductive, but rather what Peirce (1932: 2.268) called a “statistical deduction,”

Notice that the prediction of test evidence in (10) is different from what is shown for the explanatory hypothesis in (2). Statistical hypotheses are tested by seeking the same class of data used to originally infer the hypotheses, whereas testing explanatory hypotheses requires test evidence that is different from the effects originally used to infer the hypotheses (Fitzhugh, 2016c; Fitzhugh, 2016d).

As suggested by Felsenstein (1985) and Felsenstein (2004), the bootstrap method applied to phylogenetic hypotheses involves randomly sampling characters from an original character data matrix to create contrived data sets of the same size from which new cladograms are produced. Keep in mind that these cladograms, as “bootstrap replicates,” are not explanatory hypotheses; they are merely branching diagrams derived from artificial data sets. Yet, it is claimed that the more often “clades” (actually only branching patterns) are obtained among these “replicates” that match the original groupings indicates higher “support” for those clades. This claim is, however, entirely incorrect. The “bootstrap replicates” have no epistemic relation to the original phylogenetic hypothesis(es) under consideration, such that any measure of “support” would be impossible to ascertain. Along with the statistical, i.e., inductive, mindset that was introduced into phylogenetic inference in the 1970’s, the bootstrap method applied to phylogenetic hypotheses was largely an exercise in filling a perceived methodological void. Had the abductive nature of systematics been recognised early on, and character data not confused with test evidence, the state of the field might have avoided so many scientifically questionable schemes. The only support relevant to cladograms is valid test evidence produced in the process of testing the various classes of hypotheses within a (composite) phylogenetic hypothesis [Fig. 3; cf. (2), (3)]. But as noted earlier, such testing virtually never occurs, making the need for talk of support both gratuitous and positively deceptive.

Just as the bootstrap provides results that cannot be interpreted as support for phylogenetic hypotheses, the Bremer Index produces values that are equally impossible to defend. Like the bootstrap, Bremer Index is claimed to determine support for groups, as opposed to any particular hypotheses implied by cladograms. This support is based on the extent to which groups or “clades” are present in cladograms of ever-increasing “length” beyond the original minimum-length tree. The greater the tree length with a particular group still present is supposed to represent greater “support” for or resiliency by that group. The problem is that trees of greater length can only be interpreted as sets of explanatory hypotheses that have no epistemic relation to the originally inferred hypotheses. The consequence is that values provided by the Bremer Index offer no indication of support. Once again, in the absence of actually testing hypotheses, as indicated in (2)–(3), no evidential test support is possible for hypotheses.

Finally, an oblique yet equally specious approach to garnering support comes from comparisons of cladograms inferred from different sets of characters, sometimes referred to as taxonomic congruence (Fitzhugh, 2014; Fitzhugh, 2016c). The idea is that if tree topologies are identical, or nearly so, this justifies a sense of greater confidence or belief in those topologies (not the implied explanatory hypotheses). Just as was noted earlier with regard to the error of character mapping, comparisons of cladograms are an inferentially indefensible approach. Cladograms inferred to explain separate sets of characters are distinct compilations of explanatory constructs that have no epistemic relevance to each other. A popular, somewhat related approach is to speak of comparisons of past and present cladogram topologies inferred from updated compilations of observations, remarking on the degree of similarities between groups in those topologies. The same basic criticism against taxonomic congruence applies here. For example, previous abductive inferences using data set D ₁ give composite phylogenetic hypothesis H ₁. With subsequent characters added, D₁₊₂, a new composite phylogenetic hypothesis, H ₂, is inferred. It might seem reasonable to speak of topological similarities and differences between H ₁ and H ₂ but this is a mistake. Again, H ₁ and H ₂ are distinct sets of explanatory hypotheses from different sets of abductive premises. Just as there are no meaningful comparisons to be made between the actual explanatory hypotheses implied by the different cladograms (cf. Figure 2), there are no relevant conclusions to be made regarding topological similarity or difference. In the absence of recognising the nature of abduction or associated intent of those inferences, focusing on phylogenetic tree comparisons offers no scientifically meaningful evaluation.

Sequence data and explanatory hypotheses

The pursuit of sequence data was not given consideration in this study. The reasoning behind this decision will be outlined in this subsection. A more complete account is presented in Fitzhugh (2016b; see also Nogueira et al., 2017: 683–684, Lovell & Fitzhugh, 2020: 270, and Fitzhugh, 2021). Contrary to a claim made by a reviewer of an earlier draft of this paper, what is presented in this section should not be interpreted as the assertion that we “do not believe in sequence data.” Rather, our intent is to point out inherent and significant difficulties that generally preclude causally accounting for differentially shared nucleotides or amino acids via inferences of phylogenetic hypotheses.

The line of argument developed in this more inclusive section—establishing the basis for systematics—follows from the fact that the goal of scientific inquiry is to continually acquire causal understanding of phenomena, in the form of differentially shared characters of organisms, that we encounter as well as describe. The pursuit of that understanding begins with reactions to those effects, in the form of implicit or explicit why-questions, and subsequently inferring answers to those questions by way of abductive reasoning. As discussed earlier, abduction produces hypotheses that posit possible past causal events/conditions that account for observed effects, as shown in (1), (4), and (5). And, almost all actions in systematics are centered around abduction as opposed to testing hypotheses via induction (Fitzhugh, 2005a; Fitzhugh, 2006a; Fitzhugh, 2006b; Fitzhugh, 2008a; Fitzhugh, 2009; Fitzhugh, 2010a; Fitzhugh, 2012; Fitzhugh, 2013; Fitzhugh, 2014; Fitzhugh, 2015; Fitzhugh, 2016a; Fitzhugh, 2016b; Fitzhugh, 2016c; Fitzhugh, 2016d).

We pointed out earlier that composite phylogenetic hypotheses provide almost no causal specifics among the four classes of hypotheses implied by cladograms (Fig. 3). This is a consequence of the fact that phylogenetics computer algorithms emulate a vague form of abduction [e.g., (4), (5)]. Consider again what the Phylogenetic theory in (4) offers regarding character origin/fixation:

character x(1) originates by [unstated] mechanisms a, b, c…n, and becomes fixed within the population by [unstated] mechanisms d, e, f…n (= ancestral species hypothesis), followed by [unstated] event(s) g, h, i…n, wherein the population is divided into two or more reproductively isolated populations.

The cause of character fixation in an ancestral population could at a minimum be assumed to be natural selection or genetic drift. In the absence of explicit causal specifics being presented in the Phylogenetic theory, it would have to be assumed that either selection or drift are reasonable causal alternatives. This equivalence is plausible for explaining phenotypic characters but cannot be applied to sequence data. For selection to operate, whether purifying or directional, variation among heritable traits within a population must be identified that directly result in differential fitness among individuals. Nucleotides and amino acids lack emergent properties that could directly manifest fitness differences (for important nuances, cf. Linquist, Doolittle & Palazzo, 2020). Whilst there are tests that can indicate whether selection has occurred in relation to sequence data, such as the McDonald–Kreitman test (McDonald & Kreitman, 1991; Sawyer, 1994; Hey, 1999; Hurst, 2002; Biswas & Akey, 2006; Zhang & Yu, 2006; Koonin, 2012; Petrov, 2014), these tests cannot discriminate at what organisational level selection might have been directed. It is at higher organisational levels that phenotypic characters can result in fitness variation. If selection is a causal factor, it comes into play due to phenotypic effects at these higher levels. Alternatively, since there is no direct selection for sequence data, phenotypic characters have the potential to lead to differential fitness and intergenerational selection that can influence the occurrences of those sequence data that produce the selected characters. The result is a form of indirect selection, termed downward causation by Campbell (1974: 180),

“Where natural selection operates through life and death at a higher level of organisation, the laws of the higher-level selective system determine in part the distribution of lower-level events and substances.”

Since the 1970’s the importance of downward causation has become increasingly recognised (e.g., Vrba & Eldredge, 1984; Salthe, 1985; Lloyd, 1988; Auletta, Ellis & Jaeger, 2008; Ellis, 2008; Ellis, 2012; Ellis, 2013; Ellis, Noble & O’Connor, 2011; Jaeger & Calkins, 2011; Laland et al., 2011; Martínez & Moya, 2011; Davies, 2012; Okasha, 2012; Walker, Cisneros & Davies, 2012; Griffiths & Stotz, 2013; Martínez & Esposito, 2014; Walker, 2014; Mundy, 2016; Callier, 2018; Pouyet et al., 2018; Salas, 2019; Yu et al., 2020), but its relevance to how sequence data are explained via phylogenetic hypotheses has only recently been considered (Fitzhugh, 2016b).

If there can be no direct selection for sequence data, the issue then becomes one of deciding to explain individual nucleotides or amino acids either directly via drift or indirectly through downward causation. If the latter, it is first necessary to associate those sequence data with the higher-level phenotypic characters to which selection is the hypothesised causal condition. Those associated sequence data would then be excluded from phylogenetic inferences since they would already be explained in conjunction with higher-level characters (Fitzhugh, 2016b). On the other hand, it might be argued that one could avoid the issue altogether by assuming all sequence data should be explained by drift. This would be unrealistic as it necessitates that selection at higher organisational levels never occurs. In the absence of being able to discriminate between drift versus selection by downward causation, the only sensible option is to forgo altogether attempts to causally account for shared nucleotides or amino acids by way of phylogenetic inference. Otherwise, to explain shared nucleotides or amino acids with phylogenetic hypotheses inferred under an entirely agnostic perspective [i.e., the Phylogenetic theory in (4)] would be epistemically and scientifically unwarranted.

The inherent limitations for inferring explanatory hypotheses for sequence data have associated relevant consequences for mapping morphological characters on cladograms only inferred for sequence data. As discussed earlier (Phylogenetic inference=abduction; see also Fitzhugh, 2014), mapping is an epistemically unfounded approach to explaining characters. This problem is compounded by the fact that not discriminating between genetic drift and natural selection as possible causal factors when explaining sequence data yet again precludes rational acceptance of phylogenetic hypotheses accounting for mapped characters.

Finally, it might appear at first sight that the exclusion of sequence data on either the basis of downward causation or inability to distinguish drift from selection is at odds with the requirement of total evidence (RTE; cf. Fitzhugh, 2006, for a discussion of the RTE in relation to systematics). Such is not the case. The RTE is a recognised maxim for all non-deductive reasoning (the requirement is automatically satisfied in deduction), wherein rational acceptance of a conclusion is based on considering all relevant evidence that can affect support (either negative or positive) for that conclusion (Carnap, 1950; Barker, 1957; Hempel, 1962; Hempel, 1965; Hempel, 1966; Hempel, 2001; Salmon, 1967; Salmon, 1984a; Salmon, 1984b; Salmon, 1989; Salmon, 1998; Sober, 1975; Fetzer, 1993; Fetzer & Almeder, 1993): “All that the requirement of total evidence says is that one’s confidence in a hypothesis must be proportional to the support that that hypothesis receives from one’s evidence...” (Neta, 2008: 91). Nearly all of the philosophical literature on the RTE has focused on test evidence in relation to induction [cf. (3)], which is understandable given that, for instance, to exclude known test evidence that could enhance or compromise acceptance of a theory or hypothesis would be less than rational.

Whilst abductive reasoning has almost never figured into discussions of the RTE, the non-deductive nature of abduction necessitates its attention (Fitzhugh, 2006b). Consider the following example. For members of three species, a-us, b-us, and c-us, there are characters distributed among subjects 1–3 (cf. Fitzhugh, 2006c):

If the characters among members of a-us, b-us, and c-us are explained separately, there would be three sets of explanatory hypotheses (outgroup is excluded), represented here in parenthetical form:

Subject 1: (a-us (b-us, c-us))

Subject 2: (b-us (a-us, c-us))

Subject 3: (c-us (a-us, b-us)).

The three sets of hypotheses are contradictory, and while each is supported by abductive evidence [cf. (1)], choosing among them based on the information provided would be problematic. This is the sort of situation the RTE is intended to address, as well as those instances that involve other degrees of data partitioning in relation to inferring phylogenetic hypotheses and character mapping. The nature of the abductive inferences involves explaining features of semaphoronts, such that the observed characters are part of the integrated network that makes up those individuals. The various phylogenetic theories applied to those characters [cf. (1)] assume that past causes operated on individuals, not distinct characters. Thus, explaining the characters under each subject would be relevant to explaining the other characters. The plausibility for any of the phylogenetic hypotheses would be based on an abductive inference where all the characters are treated as part of the premises.

Related to the topic highlighted in this subsection, it would not be a violation of the RTE to exclude from phylogenetic inferences those sequence data for which no empirical basis is available for discriminating drift from selection-via-downward-causation for sequence data. The difficulty faced is not a matter that falls under the purview of the RTE, but rather an inability to engage in abductive reasoning for a particular class of characters.

Sequence data are not more objective than other classes of observations

Consider the following statements,

“Molecular data are more objective and subject to considerably more rigor than morphological data. DNA sequence contains four easily identified and mutually exclusive character states[….] Morphological and embryological character definitions and scoring of character states are far more subjective, and most characters have been repeatedly used without critical evaluation, calling into question the utility of morphological cladistic studies that span Metazoa[….]” (Halanych, 2004: 230).

“As we move forward with the deep-animal tree [sic], we should not employ morphology and developmental data to reconstruct the tree. There is too much historical baggage and subjectivity with these data. We should reconstruct the tree with molecular data and then use that tree to independently interpret the morphology and development in light of that tree[….]” (Halanych, 2016: 325).

“We present the view that rigorous and critical anatomical studies of fewer morphological characters, in the context of molecular phylogenies, is a more fruitful approach to integrating the strengths of morphological data with those of sequence data[....] We argue[...] that a main constraint of morphology-based phylogenetic inference concerns the limited number of unambiguous characters available for analysis in a transformational framework” (Scotland, Olmstead & Bennett, 2003: 539).

Taken at face value, each statement appears damning of all or most observations of features of organisms other than sequence data in relation to inferring phylogenetic hypotheses. We have seen throughout this section, however, that the disconnect between principles of scientific inquiry and recent views on phylogenetic inference have led to perspectives that cannot be defended. This outcome extends to the notion that sequence data are somehow “more objective,” “more fruitful,” or more reliable than other observations. Asserting that one class of observations is more effective than other classes presumes that the objective of phylogenetic inference is to obtain “phylogenies” or “trees,” which is at odds with pursuing explanations of all relevant observations. If the objective of systematics is causal explanation, thus consistent with all of scientific inquiry, then the pursuit of “phylogenies” or “trees” is not only contrary to that objective but has led to erroneous inclinations in addition to the largely contrived objective/subjective dichotomy. These erroneous pursuits include, but are not limited to (1) inferring phylogenetic hypotheses for separate classes of characters, especially sequence data versus morphology, and drawing comparisons between tree “topologies” (cf. Phylogenetic hypotheses and the myth of evidential support), and (2) inferring phylogenetic hypotheses for sequence data, then “mapping” morphological characters on the “tree” and making evolutionary claims regarding the latter characters (cf. Phylogenetic inference=abduction). To address the claim that sequence data are more objective or less ambiguous requires that we first consider what is meant by scientific objectivity. Then there are several additional issues in the above quotes that need to be addressed.

As might be expected, objectivity has multiple meanings. Useful overviews can be found in Daston & Galison (2007), Gaukroger (2012), Reiss & Sprenger (2017), and Wilson (2017). Common interpretations of objectivity range from the view that judgements should be (a) free of prejudice or bias, (b) free of assumptions, and/or (c) an accurate representation of, or “faithfulness” to facts (Gaukroger, 2012). Each meaning can be applied to everyday life as well as within fields of science. Regarding freedom from prejudice or bias, this is to speak of limiting levels of distortion beyond what is perceived. Freedom from assumptions would be impossible to accomplish given the theory-laden nature of observation statements and other types of propositions: “Every instance of scientific inquiry, every study, rests on a vast submerged set of political, moral, and ultimately metaphysical assumptions” (Wilson, 2017). Scientific objectivity has often been interpreted along the lines of “faithfulness to facts” (Daston & Galison, 2007; Reiss & Sprenger, 2017), where we attempt to minimise arbitrary judgements; we want to answer questions on the basis of the most appropriate evidence we have available. Objectivity occurs in degrees—there is no absolute objectivity, which is why objectivity cannot be equated with truth, which is absolute. Ultimately, “While we can strive for objectivity, inquiry is inherently subjective” (Spencer, 2020).

Reiss & Sprenger (2017) distinguish “product” and “process” objectivity. Product objectivity regards the products of scientific inferences, such as theories, laws, and observation statements, as accurate representations of the external world. Process objectivity refers to processes and methods that are not dependent on social or ethical values, or the individual biases of the scientist. It is product objectivity that is of concern when it comes to assessing the quotes given above, claiming sequence data are in some way more objective or less ambiguous than other classes of observations.

There is nothing in the conception of product objectivity that would support the claim that observation statements of morphological features are less objective than sequence data, much less that such a contrived distinction warrants the exclusion of an entire class of observations from the goal of systematics. Indeed, that biologists hold different interpretations of what they perceive of the properties of organisms is not tantamount to proffering less objective observation statements. Those statements are themselves products of abductive reasoning (Peirce, 1935; Hoffmann, 1995; Burton, 2000; Magnani, 2009; Anjum & Mumford, 2018), constrained by one’s background knowledge, and like all hypotheses, are open to empirical evaluation. If anything, it can be argued that sequence data are less than objective relative to other classes of characters given the fact that attempted phylogenetic inferences explaining differentially shared nucleotides or amino acids typically operate under the assumption that rates of substitution must be imposed within abductive inferences (cf. Phylogenetic inference=abduction). That assumption carries with it the implication that observation statements of shared characters must be treated as potentially false, hence the need to interject substitution rates within inferences, as indicated by the all-too-common emphasis on branch lengths. Considering the rule for valid reasoning—that we should assume the truth of premises, discussed earlier—the treatment of sequence data with “likelihood” and “Bayesian” methods pales against any criterion of objectivity.

Rather than focusing on the false argument from objectivity as a basis to eliminate observations of morphological characters in need of explanation, systematists should embrace as much as possible all relevant observations per the goal of scientific inquiry discussed earlier and the requirement of total evidence (Fitzhugh, 2006b). Two notable exceptions to considering explaining differentially shared characters by way of phylogenetic hypotheses come from conditions discussed earlier: (1) why-questions addressing some characters might not be amenable to being answered by phylogenetic hypotheses, but instead one of the other classes of explanatory hypotheses applied in systematics (Fig. 1); (2) epistemic limitations to explaining sequence data because of an inability to discern explanations due to genetic drift versus selection by way of downward causation (cf. Sequence data and explanatory hypotheses). Pursuing causality is the intent—not producing “trees” or “phylogenies,” which are typically devoid of causal considerations in relation to sequence data.

Characters of the anterior end: prostomium (subjects 1–9)

The form of the magelonid prostomium (subject 1) is considered a synapomorphy of the group, the name “shovel head worms” having been coined due to its peculiar shape. The dorso-ventrally flattened, shovel-shaped prostomium [character 1(1), Figs. 4 and 5] is a feature not observed among members of the outgroups or any other polychaete group. Nuchal organs (subject 2) have not been recorded among adult members of any magelonid species [character 2(0)] but are present among all members of the outgroups [character 2(1)].

The distal margin of the magelonid prostomium (subject 3) can vary from smooth [character 3(0), Figs. 4A and 4F], crenulate [character 3(1), Figs. 4B and 4C] or medially indented [character 3(2), Fig. 4D]. Uebelacker & Jones (1984) stated that the presence of a crenulated prostomial margin is an excellent character for separation of species, although the degree of crenulation present shows great intraspecific variability. The crenulations may differ in number and form, for example, the distinct, triangular crenulations of Magelona sp. L of Uebelacker & Jones (1984) or M. crenulifrons (Fig. 4B), in comparison to the numerous minute crenulations observed for M. longicornis or M. wilsoni (Fig. 4C). The crenulations among members of some species are difficult to discern and this is perhaps the reason they have been previously overlooked in some descriptions and erroneously reported as absent in others. The prostomial distal margin (subject 4) may additionally vary in shape, from triangular [character 4(0), Fig. 4E], to rounded [character 4(1), Figs. 4A and 4F] or straight [character 4(2), Figs. 4G, 4H, 5C and 5E]. As subjects 3 and 4 relate directly to the shovel-shaped prostomium, characteristic of magelonids, it does not apply to any members of the outgroups.

Horns are the distal projections of the prostomium (subject 5) that occur among members of some species of magelonids [character 5(1)], such as Magelona pacifica (Fig. 5A), or M. montera (Fig. 5B). Uebelacker & Jones (1984) drew attention to this character for the initial separation of magelonid species. When present, horns can vary in size/shape (subject 6), from rudimentary [character 6(0), Figs. 4G, 4H, 5C and 5E], such as those observed among members of M. variolamellata and M. alleni (often described as a squared anterior margin), or conspicuous structures as seen among members of species such as M. posterelongata [character 6(1), Fig. 5D]. In the latter case, these are often very distinct from the distal prostomial margin. Prostomial horns do not occur in the outgroups, thus subjects 5 and 6 are not applicable.

Several authors have identified dimensions of the prostomium (subject 7) as an important feature in the separation of magelonid species (Hartman, 1944; Wilson, 1958; Wilson, 1959; Jones, 1963; Jones, 1971; Jones, 1978; Uebelacker & Jones, 1984; Bolívar & Lana, 1986; Blake, 1996c; Mortimer, 2019). According to Uebelacker & Jones (1984), the general shape and relative dimensions are fairly constant within each species, something which the current authors are in agreement. Consequently, the ratio of width to length was used to recognise three prostomial characters: prostomium longer than wide [character 7(0), Figs. 4A, 4F, 5B and 5D], as wide as long [character 7(1), Figs. 4B, 4D and 4G], and wider than long [character 7(2), Figs. 4C, 4H and 5E]. Members of three species within the outgroups (Spio filicornis, Prionospio lighti and P. ehlersi) are considered to possess prostomia which are longer than wide [character 7(0)], whilst members of Phyllochaetopterus limicolus and Laonice cirrata have prostomia which are wider than long [character 7(2)].

Magelonid prostomia carry paired “muscular”, dorsal longitudinal ridges (subject 8), which extend from the base of the prostomium towards the distal tip [character 8(1)]. Jones (1968) stated that their “inner surfaces, particularly dorsally and ventrally, are provided with longitudinal muscles,” describing them as hollow, cylindroid structures, which he presumed were fluid filled. The number of prostomial ridges (subject 9) varies from one pair [character 9(0), Fig. 4H], such as members of Magelona equilamellae or two pairs [character 9(1), Figs. 4A–4C, 4E and 5B], as observed among members of M. pacifica (Fig. 5A). The ridges may vary in size and distinctiveness, the outer pair among members of some species being more difficult to discern, e.g., M. alleni (Fig. 4G), to the distinct and transversely ridged structures seen in, for example, members of M. wilsoni (Fig. 4C). Prostomial ridges are not present among members of the outgroups [character 8(0)] and thus subject 9 is not applicable.

Characters of the palps (subjects 10–13)

Magelonids possess two long palps, considered to be peristomial, arising from the larval prototroch (Wilson, 1982). There has been much discussion about palp origin (subject 10), since they are ventrally inserted [character 10(1), Figs. 4F and 5G], rather than dorsally [character 10(0)], the latter of which is seen among members of all other polychaete groups with palps, including the outgroups. In addition, the palps do not have the characteristic feeding groove seen among members of other taxa such as spionids and Myriowenia. However, an inconspicuous ventral line devoid of papillae is generally present. According to Orrhage (1966), palps among members of the Spionidae and Magelonidae are homologous because they are innervated by nerves from the same part of the brain. More recently Beckers, Helm & Bartolomaeus (2019) stated that two nerves innervate the palps of members of the Magelonidae, Spionidae, Chaetopteridae and Oweniidae, indicating the purported plesiomorphic condition in Annelida.

Magelonid palp surfaces (subject 11) are uniquely papillated, with ampulliform to digitiform papillae arranged in rows along the longitudinal axis [character 11(1), Figs. 4H, 5B and 5F–5H]. Members of the outgroups, along with all other polychaetes possess non-papillated palps [character 11(0)]. Although the number of rows of papillae varies among members of magelonid species, it is relatively stable among members of any particular species. As papillae are involved in obtaining food (Jones, 1968; Fauchald & Jumars, 1979; Mortimer & Mackie, 2014), variations in number of papillae may reflect different feeding strategies. In general, the number of rows of papillae are fewer distally in comparison to proximally (Jones, 1963; Jones, 1971; Jones, 1978). The number of rows of papillae in the proximal region of the palp (subject 12) were coded as three characters: two rows of papillae [character 12(0)], 4–8 rows [character 12(1)], and 10–14 rows [character 12(2)]. The number of rows of papillae at the distal end (subject 13) were coded for 2 [character 13(0)], 4 [character 13(1)], or more than 4 rows [character 13(2)]. Where a range was observed among members of a species, the maximum number of rows was the value used. As with palps in other polychaete groups, those of magelonids are easily lost upon collection. Many descriptions do not include information about them, whilst others do not detail differences in number of papillae along the palp length. Many authors have undervalued the importance of palp characteristics in the descriptions of members of species. These two factors present the largest problem in coding these subjects. It should be noted that the number of rows of papillae may vary in animals with regenerating palps, particularly at the distal tips, thus affecting the ranges observed. Subjects 12 and 13 are not applicable to members of the outgroups since papillated palps are unique to magelonids.

Burrowing organ (subject 14)

Magelonids have long been described to possess a ventral eversible proboscis. A heart-shaped sac (when fully everted; oval when only partially everted), carrying longitudinal ridges (Mortimer, 2019: fig. 4.2.4). However, recent studies show that this structure plays no part in feeding (and is not connected to the buccal region) but functions in burrowing only (Mortimer, 2019). The term “burrowing organ” should be applied to this structure. The burrowing organ is present in members of all known magelonid species [character 14(1)] but not present in any members of the outgroups [character 14(0)].

Characters of body regionation (subjects 15–16)

The body of magelonids (subject 15) is divided into an anterior region comprising the head and “thorax” (the latter of which is characterised by the presence of capillary chaetae), and an “abdomen” comprising many segments with hooded hooks [character 15(1)] (Fig. 6A). As an aside, the terms thorax and abdomen are ill-defined when applied to polychaetes overall, other than to vaguely indicate anterior-posterior distinctions along the body. Unfortunately, the characters associated with these regions are not uniformly applied across all polychaetes.

The body of members of the outgroups do not have this regionation [character 15(0)]. Behind the head region (subject 16), the thorax carries an achaetous first segment, followed by either eight [character 16(0), Octomagelona; Fig. 6] or nine [character 16(1), Magelona] chaetigers. Subject 16 does not apply to members of the outgroups.

Characters of thoracic parapodia (subjects 17–30)

Unique terminology introduced and later modified by Jones (1963), Jones (1971) and Jones (1978) has been applied to the family, particularly related to parapodial structures (Fig. 7). Brasil (2003: fig. 4) provided a comprehensive review of this terminology and problems related to inconsistencies in applying terms. Consequently, several authors have suggested its abandonment (Rouse, 2001; Mortimer & Mackie, 2003). Magelonid chaetigers possess biramous parapodia with lamellae in both the noto- and the neuropodia. Lamellae are generally postchaetal, flattened structures that range from filiform to foliaceous in shape. They encircle the chaetal bundle, attaching to a low, triangular or rounded prechaetal lamella, almost cuff-like (Fig. 8A). However, among members of some species, the lamellae are somewhat subchaetal in position, underneath the chaetal bundle and appearing somewhat U-shaped when viewed laterally (Figs. 8B, 8D). All members of all species in the ingroup and outgroups have postchaetal noto- and neuropodial lamellae in thoracic/anterior chaetigers (subjects 17 and 19, respectively) except for Phyllochaetopterus limicolus, in which they are absent [characters 17(0) and 19(0)].

The development of notopodial (subject 18) and neuropodial lamellae (subject 20) along the anterior chaetigers may be similar in all chaetigers [characters 18(0) and 20(0), respectively], such as among members of Magelona papillicornis, and M. minuta (Fig. 7A), or vary along the thorax [characters 18(1) and 20(1), respectively]. Subjects 18 and 20 are considered inapplicable to members of all species in the outgroups. The variation in degree of development can be easily observed among members of M. obockensis (Fig. 7C), and M. conversa (Fig. 7D), in both the noto- and neuropodia, whilst among members of Octomagelona bizkaiensis it is evident only in the neuropodia (Fig. 6E). The lamellae of chaetigers eight and nine often vary in comparison to those of the first seven chaetigers, e.g., M. crenulifrons, and M. johnstoni (Fig. 9I). Thus, characters of the lamellae of the first seven chaetigers were coded separately (subjects 21–30) to those for chaetiger eight (subjects 34–40), and chaetiger nine (subjects 41–49).

The notopodial lamellae in chaetigers 1–7 (subject 21) may be postchaetal [character 21(0); noted also for members of the outgroups, except Phyllochaetopterus limicolus, for which the subject is not applicable] or subchaetal [character 21(1)], whilst neuropodial lamellae (subject 26) may be postchaetal [character 26(0); observed for members of the outgroups, except P. limicolus, for which the subject is not applicable], subchaetal [character 26(1)] or prechaetal [character 26(2)]. Prechaetal neuropodial lamellae, such as are observed among members of Magelona conversa and M. mirabilis (Figs. 7D and 8C), are less frequent than both post- (Fig. 7A) or subchaetal (Fig. 9E). The position of the thoracic neuropodial lamellae (subject 27) may be consistent throughout the thorax [character 27(0)], or may vary [character 27(1)], e.g., being prechaetal on chaetiger 1, occurring in a ventral position in the mid-thorax, and distinctly postchaetal in the posterior thorax (Figs. 8F and 8G). This subject was not considered applicable to members of the outgroups.

Thoracic notopodial lamellar shape (subject 22) may be filiform, having a long, tapering structure [character 22(0), Figs. 9A, 9B and 9G] or more foliaceous [character 22(1), Figs. 9C–9F]. The margins of the latter (subject 23) may be smooth [character 23(0)], crenulate [character 23(1), e.g., M. montera, Figs. 9D and 9E] or bilobed [character 23(2), e.g., M. obockensis, Fig. 9F]. All members of the outgroups are considered to have foliaceous [character 22(1)], smooth-edged [character 23(0)] anterior notopodia, except for members of Phyllochaetopterus limicolus for which subjects 22 and 23 are not applicable. Above the notopodial lamellae and in a slightly prechaetal position, superior dorsal lobes (subjects 24, 25, 37, 45) (dorsal medial lobes sensu Jones) may be present (Figs. 7C, 7D, 8D, 8E and 9C–9F) or absent [characters 24(0), 37(0), 45(0)]. These structures are generally smaller than the notopodial lamellae and somewhat cirriform to digitiform in shape. However, among members of some species they may be larger and distinctly foliaceous. When present, superior dorsal lobes may be present on all thoracic chaetigers [characters 25(1), 37(1), 45(1)], on chaetigers 1–8 only [characters 25(1), 37(1), 45(0)], or only occurring on a small number of thoracic chaetigers [often starting in the mid-thoracic region, e.g., M. johnstoni or M. conversa, character 25(0), Fig. 7D]. Superior dorsal lobes are absent among all members of the outgroups on chaetigers 1–9 (thus subject 25 is not applicable), and subjects 24, 25, 37 and 45 are considered inapplicable for Phyllochaetopterus limicolus.

The shape of thoracic neuropodial lamellae (subject 28) may be filiform [character 28(0), Figs. 7C, 8F and 8G] or foliaceous [character 28(1), Fig. 9C, also seen in members of the outgroups, excluding Phyllochaetopterus limicolus, for which the subject is not applicable]. The distal ends (subject 29) of which may be pointed [character 29(0)], as among members of M. equilamellae (Figs. 8F and 8G), or distally expanded and scoop-shaped [character 29(1)] as observed among members of M. cincta (Fig. 9H). The latter subject is not applicable to members of the outgroups. The lengths of noto- and neuropodial lamellae (subject 30) may be equivalent [character 30(0), Fig. 9G], the notopodial lamellae may be longer [character 30(1), Figs. 9D–9F], or the neuropodial lamellae may be longer [character 30(2), Fig. 7D], e.g., chaetiger 8 of M. conversa. The thoracic notopodial lamellae among members of the outgroups, Prionospio lighti and P. ehlersi are longer than neuropodial lamellae [character 30(1)], whilst for members of Spio filicornis and Laonice cirrata they are equivalent [character 30(0)]. Subject 30 is not applicable to P. limicolus.

Characters of the chaetae from chaetigers 1–8 (subjects 31–33)

Capillary chaetae (subject 31) occur only in the thorax of magelonids (Figs. 6, 8, 9, 10 and 11), but many authors have undervalued the differences between them, describing them simply as limbate capillary chaetae (Hartman, 1944; Hartman, 1961; Hartman, 1965; Nateewathana & Hylleberg, 1991). Jones (1977) was one of the few authors who referred to the presence of unilimbate [character 31(0), Fig. 10C] and bilimbate capillary chaetae [character 31(1)] in descriptions of members of species. For some individuals, both unilimbate and bilimbate chaetae have been recorded, however, the predominate form in each case was coded in the current analysis. Bolívar & Lana (1986) and Blake (1996c) additionally noted some members of species possess bilimbate chaetae (subject 32) with an irregular blade [character 32(1)], such as Magelona variolamellata (Fig. 10E), and Brasil (2003) additionally noted it to occur among members of species such as M. riojai (Fig. 10D), M. pacifica and M. pitelkai. Members of the outgroups possess unilimbate capillary chaetae [character 31(0)], whilst subject 32 is not applicable. Additional to the limbations of capillary chaetae, the relative lengths of noto- and neurochaetae may vary (subject 33). Either being equivalent in length [character 33(0)] e.g., M. nonatoi, longer in the notopodia [character 33(1)], e.g., M. sacculata, or longer in the neuropodia [character 33(2)], e.g., M. variolamellata. For members of Phyllochaetopterus limicolus, Spio filicornis and Prionospio lighti, chaetae are longer in the notopodia [character 33(1)], whilst for members of Laonice cirrata and P. ehlersi they are equivalent in length [character 33(0)]. In some members of magelonid species chaetae of all thoracic chaetigers are similar, however, in others the chaetae of the ninth chaetiger are modified. For this reason, the chaetae were coded collectively for the first eight chaetigers, and separately for chaetiger nine (see below).

Characters of parapodia of chaetiger 8 (subjects 34–40)

As described above, development of lamellae in the magelonid thorax may be similar on all chaetigers or vary between them. For members of species in which they vary, the parapodia of chaetigers 8 and 9 in particular may show larger differences in comparison to preceding chaetigers. However, the same characters that were coded for the first seven chaetigers can be applied to the posterior-most thoracic chaetigers. The notopodial lamellae (subjects 34–36) of chaetiger 8 may be postchaetal [character 34(0)] or subchaetal [character 34(1)] in position relative to the chaetal bundle, filiform [character 35(0)] or foliaceous [character 35(1)] in shape, with smooth [character 36(0)], crenulate [character 36(1)] or bilobed [character 36(2)] margins. Subjects 34–36 are considered not applicable to members of Phyllochaetopterus limicolus. However, remaining members of the outgroups possess postchaetal [character 34(0)], foliaceous [character 35(1)] parapodia with smooth margins [character 36(0)]. Superior dorsal lobes (subject 37) may be absent [character 37(0)] or present [character 37(1)]. The neuropodial lamellae (subjects 38–40) of chaetiger 8 may be postchaetal [character 38(0)], subchaetal [character 38(1)] or prechaetal [character 38(2)] in position relative to the chaetal bundle, and may be filiform [character 40(0)] or foliaceous [character 40(1)] in shape. However, an additional postchaetal expansion (subject 39), often triangular in shape, as is seen among members of Magelona montera (Fig. 9D) is sometimes observed in the neuropodia of chaetiger 8 (and 9, see below) [character 39(1)]. Subjects 38–40 are considered not applicable to members of Phyllochaetopterus limicolus. Members of Spio filicornis, Prionospio lighti, P. ehlersi and Laonice cirrata have postchaetal [character 38(0)], foliaceous [character 40(1)] neuropodia in chaetiger 8, without an additional postchaetal expansion [character 39(0)].

Characters of the parapodia of chaetiger 9 (subjects 41–49)

For chaetiger 9, observed characters were coded in the same way as for chaetiger 8. The notopodial lamellae (subjects 41, 43–44) of chaetiger 9 may be postchaetal [character 41(0)], or subchaetal [character 41(1)] in position in comparison to the notochaetae, and may be filiform [character 43(0)] or foliaceous [character 43(1)] in shape, with smooth [character 44(0)], crenulate [character 44(1)] or bilobed [character 44(2)] margins. Members of the outgroups possess postchaetal [character 41(0)] e.g., notopodia of chaetiger 99 which are foliaceous [character 43(1)], with smooth margins [character 44(0)]. However, subjects 41, 43 and 44 are not applicable to Phyllochaetopterus limicolus. Superior dorsal lobes (subject 45) may be present [character 45(1)] or absent [character 45(0)].

The neuropodial lamellae (subjects 46, 48) of chaetiger 9 may be postchaetal [character 46(0)], subchaetal [character 46(1)] or prechaetal [character 46(2)] in position in comparison to the chaetal bundle, and filiform [character 48(0)] or foliaceous [character 48(1)] in shape. As with chaetiger 8, an additional postchaetal expansion (subject 49) may be absent [character 49(0)] or present [character 49(1)]. Members of the outgroups possess foliaceous [character 48(1)], postchaetal [character 46(0)] lamellae without an additional postchaetal expansion [character 49(0)], except for P. limicolus for which these subjects are not applicable.

The parapodia in chaetiger 9 among members of some magelonid species may be markedly different in comparison to preceding chaetigers. For example, the lamellae in chaetiger 9 among members of Magelona johnstoni (Fig. 9I) are much broader and lower [characters 42(0) and 47(0)] in comparison to the elongate lamellae of chaetigers 1–8 in both the noto- and neuropodia, and the elongate lamellae of chaetiger 9 amongst members of other species, e.g., M. equilamellae [characters 42(1) and 47(1), Fig. 7B]. For this reason, additional subjects were added to accommodate the height of lamellae in both rami of chaetiger 9 (subjects 42 and 47). As members of Octomagelona species only possess eight thoracic chaetigers these subjects were coded as inapplicable for these taxa and additionally for members of Phyllochaetopterus limicolus. Members of Spio filicornis possess broad and low parapodia in chaetiger 9 [characters 42(0) and 47(0)], whilst for members of Prionospio lighti, P. ehlersi and Laonice cirrata they are more elongate [characters 42(1) and 47(1)].

Characters of the chaetae of chaetiger 9 (subjects 50–51)

The chaetae of chaetiger 9 in magelonids may be similar to or vary from those of chaetigers 1–8 (subjects 50 and 51). They may be the same length [character 50(0), as in the outgroups], shorter than [character 50 (1)], or longer than those of chaetigers 1–8 [character 50(2)]. The distal ends of chaetiger 9 chaetae may be gently tapered as is seen in the preceding chaetigers [character 51(0), Figs. 8F and 8G] (observed also for members of the outgroups), or have distinctly mammiform, expanded, mucronate tips [character 51(1), Figs. 7C, 7D, 9I and 10F–10H] such as among members of Magelona johnstoni and M. mirabilis. In members of other species, the distal ends are pennoned [character 51(2), Fig. 10I], such as M. pitelkai, M. hobsonae, and M. hartmanae.

Characters of the thoracic region (subjects 52–53)

Whilst many authors have noted the often-marked difference between thoracic and abdominal regions in magelonids (Fig. 8B), other characteristics of the thoracic region have been largely overlooked. Mortimer (2019) noted that among members of certain magelonid species, the thoracic interparapodial margins (subject 52) may be characteristically bulbous and rounded [character 52 (1), Fig. 11A], as seen among members of Magelona alleni, whilst among members of others, such as M. johnstoni, they are straight [character 52(0), Fig. 11B]. Members of the outgroups possess straight thoracic interparapodial margins [character 52(0)].

Uebelacker & Jones (1984) first coined the phrase “oblique lateral slits” (subject 53) in relation to members of their undescribed species Magelona sp. I from the Gulf of Mexico. Later, Bolívar & Lana (1986) described “sulcos latero-dorsais” (dorso-lateral slits) and “sulcos transversais” (transverse slits) for members of M. variolamellata (Fig. 11C). Brasil (2003) used the term dorsal furrows/grooves, highlighting them to be present also among members of M. polydentata [character 53(1)], whilst herein, the term dorso-lateral grooves is utilised. Brasil (2003) stated that they vary in number, position, and location, and they were noted to be absent in the majority of members of magelonid species [character 53(0)] and are additionally absent among members of the outgroups.

Dimensions of post-chaetiger 9 segments (subject 54)

Uebelacker & Jones (1984) described posterior segments that are much longer than wide [character 54(1), Fig. 12A] among members of an undescribed species, Magelona sp. H, from the Gulf of Mexico. Later, M. posterelongata was the first species with individuals formally described with this feature. Many authors have ignored this characteristic, a problem compounded by the difficulty in collecting entire specimens. A further problem in coding this character is that abdominal chaetigers may vary in proportion along the length of the posterior region. It is believed, however, that the majority of magelonid species possess abdominal chaetigers equal in length and width, or only marginally longer than their width; a characteristic shared with members of the outgroups [character 54(0)].

Characters of the parapodia of post-chaetiger nine segments (subjects 55–59)

The abdominal region of magelonids consists of many chaetigers carrying biramous parapodia. In contrast to the thorax, the abdominal lamellae of the notopodia and neuropodia are generally symmetrical and of similar size and shape in both rami. The abdominal region starts at chaetiger 9 for members of Octomagelona (Fig. 6) and chaetiger 10 for members of Magelona (Fig. 11A) and is marked by a change in chaetal type, from capillary chaetae to hooded hooks. Abdominal lamellar shape (subject 55) varies among members of species, from triangular to rounded. They may be basally constricted [those with rounded lamellae, character 55(1), Figs. 12C, 12H], or without a basal constriction, being broad based [character 55(0), Fig. 12B]. The latter situation being generally observed among members of species possessing more triangular lamellae, tapering to pointed tips (Fig. 12D). Subject 55 is not applicable to members of Phyllochaetopterus limicolus, however all other members of the outgroups have lamellae without a basal constriction [character 55(0)].

At the inner margins of chaetal rows, triangular to digitiform processes (subject 56) may be present [character 56(1), Figs. 12C, 12F, 12H, 12I]. Jones termed these structures dorsal medial lobes (DML) for notopodia and ventral medial lobes (VML) for neuropodia. Uebelacker & Jones (1984) believed that they may occur universally throughout the family, but this is a viewpoint not shared by the current authors. They are absent [character 56(0)] in members of the outgroups, however, this subject is considered not applicable to Phyllochaetopterus limicolus. In members of magelonid species, the size of the medial lobes in the abdomen varies widely, from minute and difficult to discern, to long and conspicuous structures, e.g., among members of Magelona montera. For comparison, the medial lobes were coded depending on their height relative to abdominal hooded hooks (subject 57): character 57(0) for members of species in which they are smaller than the hooks and character 57(1) when they are larger than the hooks. The latter subject is not applicable to the outgroups.

The abdominal lamellae may extend behind chaetal rows (subject 58). This postchaetal expansion (previously termed the interlamella) is often more conspicuous in the anterior abdomen and when present [character 58(1), Figs. 12C, 12E and 12H] it may be rounded to triangular. Among members of species in which it is absent [character 58(0), Figs. 12B, 13D and 14B], the abdominal hooded hooks arise from a distinct ridge (Figs. 12B and 13E). Postchaetal expansions are absent among members of the outgroups, whilst in members of Phyllochaetopterus limicolus the subject is considered inapplicable.

As stated above, abdominal lamellae (subject 59) in magelonids are generally symmetrical in size and shape [character 59(0), Figs. 12D and 12H] as in members of the outgroups. However, in some magelonids, such as members of Magelona alleni, the lamellae are sub-equal between the two rami [character 59(1), Fig. 12G]. This subject is inapplicable to members of Phyllochaetopterus limicolus.

Characters of lateral abdominal pouches (subjects 60–65)

Lateral pouches (subject 60) are present among members of some species of magelonids, located laterally between the parapodia of certain abdominal chaetigers [character 60(1)]. Their presence [character 60(1)] or absence [character 60(0)] is often a key diagnostic feature, although their function is currently unknown (Jones, 1968; Mortimer & Mackie, 2014). Pouches vary in morphology, direction of opening, number, and location. However, descriptions of these structures have generally been limited, often only indicated as being present or absent. Fiege, Licher & Mackie (2000) described two types of lateral pouch: Σ-shaped occurring in the anterior abdomen, generally paired, and opening anteriorly, whilst C-shaped pouches were noted to occur on median and posterior abdominal chaetigers, opening posteriorly and occurring in pairs or singly. Mortimer (2010) suggested these terms should be abandoned to enable variations to be better understood, and subsequently, Mortimer & Mackie (2014) and Mortimer (2019) detailed further pouch morphologies. Following the latter authors, lateral pouches when present, were coded based on six subjects: arrangement of pouches (subject 61), direction of pouch openings (subject 62), posteriorly open lateral pouch arrangement (subject 63), lateral pouch distribution (subject 64), and margins of posteriorly open pouches (subject 65). Firstly, pouches are either all paired [character 61(0), Fig. 13C] or unpaired [character 61(1), Fig. 13B], or both types are present [character 61(2), Fig. 13A]. Secondly, both anteriorly and posteriorly opening pouches are present [character 62(1), Figs. 13A, 13E and 13G], e.g., members of Magelona johnstoni, or only posteriorly open pouches are present [character 62(0), Figs. 13B and 13C]. Posteriorly open pouches may occur on alternate segments [character 63(1), Fig. 13B], or on consecutive segments [character 63(0), Fig. 13C], or alternatively, they may be present on both consecutive and alternating chaetigers at varying points along the abdomen [character 63(2)], e.g., members of M. pulchella. Fiege, Licher & Mackie (2000) described the first occurrence of lateral pouches on the body as an important distinguishing character between members of species and is relatively easy to observe. In members of some species, lateral pouches are present throughout most of the abdomen [character 64(0)], whilst in others they are present only on median and posterior chaetigers [character 64(1)], and in others restricted to the posterior abdomen only [character 64(2)] (Mortimer & Mackie, 2014). Whilst the margins of most posteriorly open pouches are smooth [character 65(0)], in members of some species, such as M. pacifica, they are medially split [character 65(1), Fig. 13H]. One of the biggest problems in coding these characters is the lack of information known about the posterior regions of many magelonids, whilst some pouches are present throughout most of the abdomen and are therefore likely to be recorded even in posteriorly incomplete specimens. For members of species in which pouches only occur in the extreme posterior region, pouches are likely to be recorded as absent if individuals are described from anterior fragments only. Lateral pouches are present [character 60(1)] in members of outgroups, Laonice cirrata and Prionospio ehlersi. These are paired [character 61(0)], posteriorly open [character 62(0)] and with smooth margins [character 65(0)]. They occur on consecutive segments [character 63(0)] on median and posterior chaetigers [character 64(1)]. Lateral pouches are absent [character 60(0)] in the remaining members of the outgroups.

Characters of post-chaetiger 9 chaetae (subjects 66–74)

Magelonids possess hooded hooks in the abdominal region (subject 66), located between the lamellae and medial lobes (when present). They are present [character 66(1)] in members of all magelonid taxa and members of the outgroup except Phyllochaetopterus limicolus [character 66(0)]. In magelonids, hooded hooks generally occur in one row per ramus, but recent studies suggest that they can occur in two rows, at least in the medial part of the rami (Mortimer et al., 2020) (Fig. 14B for members of M. equilamellae). Although, this may prove to be an important characteristic in future studies, due to lack of information for the majority of species this was not considered in this study.

The dentition of abdominal hooks (subject 67) may be bidentate [character 67(0), e.g., members of Magelona minuta and M. papillicornis, Figs. 14C and 14D] with one secondary tooth above the man fang, tridentate [character 67(1), Fig. 14E], with two secondary teeth above, or polydentate [character 67(2)], having three or more secondary teeth above the main fang. Quadridentate, pentadentate, and hexodont hooks (Figs. 14F–14H) have been recorded among Magelonidae, with some polydentate individuals carrying more than one type of hook. Where members of species have been recorded with more than one, the dentition which predominates was coded. For example, members of M. minuta are recorded to possess the odd sporadic tridentate hook (Mills & Mortimer, 2018), but carry predominately bidentate hooks. Members of the outgroup Spio filicornis possess bidentate [character 67(0)] hooded hooks, whilst members of Prionospio lighti, P. ehlersi and Laonice cirrata possess polydentate hooded hooks [character 67(2)]. This subject is not applicable to members of Phyllochaetopterus limicolus, which have uncini.

Jones (1971) was the first author to draw attention to the fact that in members of some species the hook at the base of the lamellae (subject 68) may be smaller than the others [character 68(1), Figs. 12E and 14I]. Later it has been suggested that not only their size but also their appearance may indicate groups of closely related species (Magelona pitelkai, M. hobsonae, M. hartmanae, M. dakini, M. filiformis, M. capensis). However, Brasil (2003), upon re-examination of the holotypes of the first four species, observed no differences in shape, only differences in size. Among members of other magelonid species, the hook at the base of the lamellae is of equivalent size to the rest [character 68(0), Fig. 13D]. The subject is not applicable to members of the outgroups.

Uebelacker & Jones (1984) first reported the presence of enlarged chaetae (subject 69) in the abdominal region of members of four undescribed species from the Gulf of Mexico; Magelona sp. C, possessing an enlarged recurved hook with a minute apical tooth in each ramus; Magelona sp. D, with enlarged hooded apical spines (Fig. 12F); Magelona sp. E, possessing large bidentate hooks akin to “ordinary” hooks; and lastly Magelona sp. H, with large unidentate recurved spines. Hernández-Alcántara & Solís-Weiss (2000) formally described members of two of these species (Magelona spp. D and H) as Meredithia spinifera and M. uebelackerae, respectively, erecting the new genus based on this distinctive feature. The authors stated that “although this character is the only one that differentiates this group from the other species described in this monogeneric group, it is sufficiently distinctive to make it a valid generic level character.” However, Mortimer & Mackie (2003) felt that stronger evidence for recognising other genera within the family was needed, preferring to follow Uebelacker & Jones (1984) by including species with enlarged abdominal chaetae in Magelona, and synonymising Meredithia with Magelona. Two additional species with enlarged abdominal chaetae have since been described: Magelona magnahamata and Magelona falcifera. For the current analysis, the presence [character 69(1)] and absence [character 69(0)] of enlarged chaetae in the abdomen was noted as well as the morphology of the distal ends (subject 70), i.e., recurved [character 70(0), Fig. 14K], spine-like [character 70(1)], or an enlarged normal hook [character 70(2)]. Enlarged hooks are not present in members of the outgroups [character 69(0)], therefore subject 70 is not applicable (together with subject 69 for Phyllochaetopterus limicolus).

The number of hooded hooks per ramus in abdominal chaetigers (subject 71) is variable and characters distinguished as eight or more hooks per ramus [character 71(0), e.g., members of Magelona equilamellae, Fig. 14B, and members of outgroups Prionospio lighti, P. ehlersi and Laonice cirrata ], and less than eight [character 71(1), Fig. 14J], such as among members of M. papillicornis and members of outgroup Spio filicornis. This separation in number was based on the examinations of specimens from a number of different species and assessing several chaetigers of the same specimen. As variations in the number of hooks have been observed along the length of the abdomen (e.g., members of M. papillicornis have a higher number of hooks in the mid-abdominal region), the number of hooks was counted in the anterior abdomen. Subject 71 is not applicable for members of Phyllochaetopterus limicolus. Hooks may be arranged in two formations (subject 72): occurring in two groups per ramus with main fangs arranged face-to face (vis-à-vis) [character 72(0), Figs. 12C, 12G, 12H, 14B and 14J], or in unidirectional rows (vis-à-dos) with main fangs pointing in one direction [character 72(1), Fig. 13D]. Vis-à-vis orientation is more common among members of the Magelonidae, whilst vis-à-dos is present for all members of the outgroups (subject 72 is not applicable to P. limicolus).

Neuropodia with sabre chaetae (subject 73) are present [character 73(1)] among members of the spionid outgroups. Such chaetae are absent [character 73(0)] among members of Phyllochaetopterus limicolus and all ingroup species.

Internal support chaetae, or aciculae (subject 74), have been reported among members of several magelonid species. These curved and slender chaetae support the bases of the lamellae, and it is possible that they are more prevalent among members of species with larger abdominal lamellae. The proximal ends of the aciculae may overlap in the junction between the noto- and neuropodia (Figs. 12C and 12H). Notes on their presence [character 74(1)] or absence [character 74(0)], as well as their description, are poorly reported in the literature. Jones (1978) stated that all magelonid abdominal chaetigers are supported by these structures, although this is not an observation supported by the current authors. Aciculae are generally more conspicuous in larger specimens, whilst slide preparations of parapodia may be necessary for smaller individuals. The relative obscurity of aciculae may be a factor in the limited inclusion in descriptions of specimens. Aciculae are absent in all members of the outgroups [character 74(0)].

Characters of body pigmentation (subjects 75–76)

Among members of some species pigmentation in the posterior thorax (subject 75) is evident even in fixed material [character 75(1)], often deep reddish to brown in colour. Very little is known about this character beyond presence or absence [character 75(0)]. When pigment is present, there are two patterns of distribution (subject 76). Among members of some species, such as Magelona alleni and M. equilamellae, the pigment forms a distinct band [character 76(0), Figs. 15A and 15C], often from chaetigers 4/5 to 8; among members of other species only light, dispersed pigmentation in posterior chaetigers is noted [character 76(1)], e.g., members of M. symmetrica. Moreover, members of Magelona fasciata Mortimer, Kongsrud & Willassen (2021) from West Africa are noted to have distinct stripy pigmentation over much of the body (Fig. 15B). Pigmentation on the palps of members of some species, such as M. mirabilis, has been noted (see Mortimer 2019: fig. 4.2.6). However, insufficient information is currently available to enable coding of this character. Relevant pigmentation is absent in members of the outgroups [character 75(0)].

Characters of granular bodies (subjects 77–78)

The surfaces of segments can have granular bodies (subject 77) appearing as spots, distributed in a variety of patterns on both the thorax and abdomen, and remain upon fixation and preservation of material [character 77(1), Figs. 6A, 9H and 15C]. Jones (1971) felt that this characteristic had no systematic significance. However, patterns appear to be species specific, particularly those of the thoracic region. The latter may appear as transverse stripes (Fig. 16A) or distinct smaller circular regions adjacent to the parapodia (Fig. 16B). Abdominal interparapodial areas commonly have granular bodies occurring in patches (subject 78) [character 78(1), Figs. 16C and 16D] among members of Magelonidae and may cover a large part of the lateral region between parapodia of adjacent chaetigers. Both thoracic and abdominal bodies may stain with dyes such as Rose Bengal, or be highlighted by methyl green, in which the granules appear white against the contrasting stain (Figs. 16B and 16C). Whilst staining patterns can be extremely useful in separating members of species (Figs. 11D, 11E and 16A), it was not included in the current study. However, it may prove useful in the future, as more methyl green patterns are described. Granular bodies and abdominal interparapodial areas are absent in members of the outgroups [character 77(0), character 78(0)].

Tube construction (subject 79)

In general, magelonids are relatively motile, burrowing more or less continually through sediments (Jumars, Kelly & Lindsay, 2015), without constructing tubes [character 79(0)]. However, members of several species are known to build distinct multi-layered tubes covered in sand [character 79(1), Fig. 16E]. Mills & Mortimer (2019) investigated the permanency of these tube-lined burrows among members of the European species, Magelona alleni. Members of some species, such as M. filiformis, have been described living in fragile tubes of a secretion to which sand grains adhere. These are not considered as tubes per se but are believed to be the worm’s response to removal from its habitat. Members of the outgroups, Spio filicornis and Phyllochaetopterus limicolus, are known to build distinct tubes [character 79(1)], whilst the situation is unknown for members of Prionospio lighti, P. ehlersi and Laonice cirrata.

Prostomial features

Magelonid prostomia may carry distinct markings on either side of the dorsal muscular ridges (Figs. 4A, 4C–4E, 5A, 5B and 5E), the pattern of which is generally species specific, whilst members of other species show no obvious markings (Figs. 4G and 4H). Although this can be an important diagnostic feature, this was not included in the current analysis due to difficulties in adequately describing this character.

Thoracic chaetae

Mills & Mortimer (2018) highlighted differences in the number of thoracic chaetae among members of different species, and additionally observed variations in number of chaetae on different chaetigers of the same animal for members of several magelonid species. They noted that members of some species, such as Magelona cincta, have distinctly splayed chaetae in the posterior thorax when compared to members of other species. Whilst these three characteristics may prove to be useful, limited information is currently available, prohibiting their use in the current study.

Features of the thoracic region

More recently, several additional characteristics of the thoracic region have been noted. Mortimer & Mackie (2009) and Mortimer (2019) described distinct V- and X-shaped ventral markings in the mid-thoracic region of members of species such as Magelona crenulifrons and M. pulchella (Figs. 11D and 11E). Among members of other magelonid species, thoracic ventral pads/swellings have been recorded, such as those observed among members of M. mirabilis (Mortimer & Mackie, 2014) or M. obockensis (Mortimer, 2010). These swellings may be oval to reniform in shape (Figs. 11F and 11G). However, occurrence of these two features in members of the majority of species is unknown.

Although, the constriction at the thorax/abdomen junction can be marked and may vary between species, the extent to which it is constricted can be greatly influenced by preservation and fixation and was not considered herein. Despite this, however, it is a character worthy of further study, and should be detailed more fully in future descriptions. Brasil (2003) indicated that members of Octomagelona present a constriction despite lacking a 9^th “thoracic” chaetiger, suggesting that the character may not simply indicate a transition between the thorax and the abdomen.

Whilst average length and the average number of chaetigers appear to be species specific, the lack of information for members of many species prohibits this as a character in the analysis at the present time. Many individuals of magelonid species are described from anterior fragments only. The breadth of individuals, however, may prove an important character. Members of species like Magelona minuta, true to their name, rarely attain widths greater than 0.5 mm, being somewhat long and slender. Members of species such as M. alleni are distinctly broad and stout, often measuring over one mm in width (Mortimer, Kongsrud & Willassen, 2021). Additionally, Mills & Mortimer (2019) indicate that stouter magelonid species generally are shorter in comparison to width and have a fewer number of chaetigers. This is something which warrants further investigation.

Abdominal hooded hooks

Mills & Mortimer (2018) highlighted differences in the angle between the main fang and secondary teeth in abdominal hooded hooks of members of several magelonid species and further suggested characteristics such as the width and roundedness of the main fang, and the angle between it and the axis of the hook shaft, as noted for example by Jones (1963) and Jones (1977), may prove useful in future studies.

Pygidium

Magelonids possess a pair of digitiform pygidial cirri on either side of the pygidium (Figs. 16F and 16G). Uebelacker & Jones (1984) noted the presence of three pygidial cirri on individuals of an undescribed species, Magelona sp. B, from the Gulf of Mexico, however, it is believed that the third cirrus may actually represent the elongated, rounded tip of the pygidium itself. As this is the only magelonid that has been described with this feature, the number of pygidial cirri was not included in this study. Mills & Mortimer (2019) noted that pygidial cirri of members of M. alleni were more truncate and triangular (Fig. 16H) in comparison to other magelonids, however, this is something that warrants further investigation.

Whilst Rouse (2001) stated that the magelonid anus is terminal, early illustrations from McIntosh (1878) of members of Magelona mirabilis (possibly M. johnstoni; see Fiege, Licher & Mackie, 2000) clearly show the anus in a distinctly ventral position. Unfortunately, for members of many species the pygidium is unknown, or the amount of information provided by authors has been relatively limited. When included in descriptions, the pygidium is often drawn from a dorsal perspective, thus conclusions about the position of the anus are often difficult to make. Mills & Mortimer (2019) investigated the situation among members of five European species, concluding that the anus was ventral in members of four species, whilst in the latter, M. alleni, it was distinctly terminal in position (Fig. 16H). Subsequently, Mortimer, Kongsrud & Willassen (2021) have noted that members of Magelona fasciata from West Africa also have a terminal anus. Whilst this may prove to be a valuable character in subsequent studies, the lack of information among individuals of many species precludes inclusion in the current study.