Surviving parsimonists: just tree-naive or tree-blindfolded?

A generation ago, an epic battle took place, largely unnoticed outside mathematical and biological sciences: the Phylogenetic Wars. By the mid-1990s, the war was over and probabilistic methods for tree-inference replaced traditional parsimony. But there is a small, quite dusty realm that still pretends nothing happened: palaeontology. A long read about undervalued data and stubborn old white men.

When you read an up-to-date molecular phylogenetic paper, e.g. based on phylogenomic data, you will cross two types of trees: maximum likelihood phylograms, i.e. phylogenetic tree in which the branch-lengths represents expected evolutionary change, and Bayesian majority rule consensus (MRC, usually presented as cladogram) or maximum clade credibility (MCC) trees. In any case, there will be values along the branches ('internodes') giving you the branches' support. How confident are the data about this taxon bipartition, the phylogenetic split, the clade, once we root our tree (usually post-analysis using outgroups)?

But in many palaeozoological and most palaeobotanical papers, you will see something like this as the outcome of the "phylogenetic analysis":

Diagram from a recently published Science paper (Wilf et al. 2019a) taken as "good support". The reference (only the fossil was free to move, the position of modern taxa was fixed) is to a molecular-phylogenetic paper from 2001.

A naked cladogram showing a strict consensus tree of the equally, ideally most, parsimonious trees (MPT) obtained during tree inference. The fossil(s), Castanopsis (Cs) rothwellii from the Eocene of Patagonia is resolved as part of a Castanopsis clade (the trident) and interpreted as the South American ancestor of all living, in tropical-subtropical Asia, modern Castanopsis spp. (Wilf et al. 2019a). Here based on a "DNA-scaffold" using no molecular data, seven scored morphological traits and leaving aside four Fagaceae genera that do not match the fossil (Ockham's Razor applied but not used).

Be hostile to molecular data. Fossils are to be treated the same as living species.
[quoted from Felsenstein 2001, the intellectual package of bygone cladists]

Similar graphs can been seen in palaeophylogenetic literature of the 1980s (A bit of heresy...), before the 'Molecular Revolution' changed how we establish phylogenetic relationships; the following example covers the same group (Fagaceae), involved the same second author but two fossils and more characters (which, however, remained undocumented).

The above graph deals with exactly the same group but Eocene North American fossils: Crepet & Nixon (1989), published in American Journal of Botany. The undocumented matrix seems to have got lost over time, noting the decrease of scored characters: 25 (legend) viz 26 (see in-figure info) in 1989 to 7 in 2019.

Creed of the Holy Cladistic Church – Farris' branch

James Farris (GoogleScholar-search), the unofficial demi-god of the Willi Hennig Society (WHS), postulated that evolution (of phenotypes and, subsequently, genotypes) is parsimonious, that it follows Ockham's Razor. Hence, we just have to infer a tree that fulfils the criterion of maximum parsimony (requires the smallest amount of inferred changes, called "steps"), and this tree will show us the true evolutionary tree.

Use only parsimony methods. History: William of Ockham told Popper to tell Hennig to use parsimony.
[quoted from Felsenstein 2001, the intellectual package of bygone cladists]

Since the most parsimonious tree is the true evolutionary tree (or the best-possible approximation of the true tree), all clades in the tree represent monophyletic groups per se (fide Hennig 1950; Ashlock 1971 proposed to use a less ambiguous term: "holophyletic"), i.e. groups of inclusive common origin. Inferred grades are either associated with paraphyly, exclusive common origin, or lack of resolution (whatever suits best the narrative).

The (very prominent in palaeobotany) authors of the two papers above, and many leading figures (prototypical old white males) in palaeo-"phylogenetics" are faithful followers of what I will call the HCC – the Holy Cladistic Church. The way I experienced it – wandering far beyond the beaten path and fighting off more than once the Cladistic Ghost – it is not a scientific philosophy but a science-philosophical cult (Once Upon a Time... by Dan Graur). If regular sermons would be held, the extremely well-merited Kevin Nixon co-authoring both papers, a polymath interested in "all -ics" (homepage/GoogleScholar-profile), would surely qualify for a cardinal of the HCC. He and Pope Farris claim co-authorship for The Holy Tool, the tree-inference programme TNT (which actually always was, even before the WHS cashed it in and Farris and Nixon put their names on it, a very versatile piece of software with many useful, bit underused, options). TNT has replaced the First Holy Tool: the framework programme WinClada including Hennig86 and NONA, originally distributed via www.cladistics.com (URL currently for sale) as shareware for 50 US$ by His Eminence Nixon (see e.g. Should we infer trees on tree-unlikely matrices).

Use only computer programs written by leaders in the Hennig Society, all others are fundamentally flawed
[quoted from Felsenstein 2001, the intellectual package of bygone cladists]

Believing that clades (no matter how well supported) in a parsimony tree equal monophyly, neither Crepet & Nixon (1989) nor Wilf et al. (2019a) felt the necessity to ...

... establish any form of branch (internode), clade (i.e. a subtree in a rooted tree, e.g. Felsenstein 2004 not Wikipedia) or node support. Rothwell & Nixon (2006) give a compelling argument: support values can mislead the reader to put too much trust into certain aspects of the tree.
... show the actual branch lengths or indicate the number of steps inferred along each branch. We will see below why.

Here's the common definition of what is a clade (in Anglo-saxon literature).

A clade (from Ancient Greek: κλάδος, klados, "branch"), also known as monophyletic group, is a group of organisms that consists of a common ancestor and all its lineal descendants. ... A clade is by definition monophyletic, meaning that it contains one ancestor (which can be an organism, a population, or a species) and all its descendants.
[Quoted from the English Wikipedia, the usual definition found in text books]

Just by comparison of the two graphs, probably produced by His Eminence Nixon himself, we see the belief clade = monophylum (grade = paraphylum), has some issues, especially regarding the interpretation of Castanopsis-like fossils.

Overlap of two cladistic (but not really phylogenetic) hypotheses spanning 30 years. Monophyla (confirmed) in bold font, all other labelled clades (grey outlines) are not probably not monophyletic (next section).

Only using morphological traits (Crepet & Nixon 1989), Castanopsis forms a grade within the Castaneoideae clade: Castanopsis is paraphyletic to Lithocarpus, Lithocarpus and Castanopsis have an inclusive common origin excluding Castanea and Chrysolepis. Crepet & Nixon's fossils (Castanopsoidea) may be part of that monophylum, or not. One of the two Lithocarpus in 1989 may include or represent Notholithocarpus, the tanoak, which became a genus only in 2008 (Manos et al. 2008). In Wilf et al's fixed tree, we find a Castanopsis clade/monophylum with Castanea as sister clade, the 1989 grade was an artefact camouflaging a monophylum. The other monophyletic group includes the monophyletic Lithocarpus s.str. and Chrysolepis + Notholithocarpus; the 1989 Lithocarpus-clade within a larger Castanopsis-Lithocarpus clade is not monophyletic. The Castaneoideae are inferred (Crepet & Nixon 1989) or depicted (Wilf et al. 2019a) as monophyletic, forming clades in both trees. Note that Wilf et al. (2019a) didn't include Quercus and the trigonobalanoids: Crepet & Nixon's Querceae monophylum.

Cladistics greatest foe: genes

The genes tell us indeed that modern Castanopsis may be monophyletic (although not necessarily forming a clade in all molecular trees, see ITS tree in Denk & Grimm 2010). Castanea is its sister. The subfamily/clade Castaneoideae is without a doubt paraphyletic, they are collected in the same subtree (clade) in all molecular trees but the oaks (Quercus) are nested inside it. A potential sister of oaks is Notholithocarpus. Oaks, Notholithocarpus and Castanea-Castanopsis share an inclusive common origin (are monophyletic). Thus, oaks have evolved from Castaneoideae (which, by the way, perfectly fits with the fossil record) and are not related to trigonobalanoids. The trigonobalanoid grade is monophyletic. Crepet & Nixon's Querceae clade with its trigonobalanoid grade is an inference artefact due to convergently (I would say, in parallel) evolved traits: a polyphyletic clade. Crepet & Nixon's Fagoideae clade is wrong, too. Fagus, the beech and namesake of the family, is a very distant sister of all other Fagaceae — a rooting error in Crepet & Nixon's 1989 tree. Which Wilf et al. (2019a) still accept, note their taxon selection: Fagus (outgroup) + Castaneoideae (paraphyletic focus clade), no Querceae (polyphyletic sister clade).

Diagram showing the synopsis of all molecular-phylogenetic studies so far. Likely monophyletic (holophyletic) taxa in bold font, paraphyletic taxa (incongruent plastid clades, morphology-based subfamily Castaneoideae) in normal font shaded grey.

The above scheme summarises clades, inclusive subtrees, seen in the various published molecular trees (red = plastid DNA, blue = nuclear DNA), the labels monophyletic/ paraphyletic are interpretations following Hennig's definitions of inclusive, all descendants of a common ancestor (monophyly), and exclusive common origin, some but not all descendants of a common ancestor (paraphyly). Since there can only be one common ancestor at the base of monophylum (unless we have reticulate histories: Monophyletic groups in networks), we need to make a pick between conflicting nuclear and plastid clades. My pick, nuclear outcompetes plastid, is based on my near 20-year long dealings with Fagaceae (and Fagales) data and takes into account not only all published trees but also the differentiation patterns in the data (Denk & Grimm 2010; All solved a decade ago...; Can we depict the evolution of ... ribosomal RNA genes). Note that molecular clocks and fossil evidence put the divergence between the reciprocally monophyletic (= sister lineages) core Fagaceae and Trigonobalanoideae at 80–60 Ma. The oaks, indubitably a monophyletic genus with poorly sorted plastomes (geography outcompeting phylogeny), evolved from their castaneoid ancestors about 60–55 Ma; the beeches, Fagus, were isolated from the remainder of the Fagaceae at 82–81 Ma (Grímsson et al. 2015, 2016; Hipp et al. 2019; and literature cited therein).

Looking at the genetics, we realise:

Morphological evolution is not parsimonious in Fagaceae (also applies to pretty much any other group of organisms).
Forming a clade in an inferred tree is neither a sufficient nor a necessary criterion for monophyly (also applies in general, and to some clades in non-parsimony molecular trees, a recent example). Being monophyletic helps a lot but doesn't ensure being resolved as a clade (also in molecular phylogeny, we synonymise but should not, clades with monophyla).
Castaneoid phenotypes are primitive within the (core) Fagaceae. The two main clades in the morphology-based tree seem to reflect similarity due to lack of derived traits (paraphyletic Castaneoideae clade) or accumulation of convergently evolved derived traits (artefactual Querceae clade, Fossils and network 1...)

Proper definitions for "clade" and "monophyletic":
Clade – a subtree in a rooted phylogenetic tree (e.g. Felsenstein 2004).
Monophyletic (fide Hennig = Ashlock's holophyletic) – a group of organisms that consists of a common ancestor and all its lineal descendants (Hennig 1950, 1965; Wheeler 2014 for phylogenies including reticulation).
Only when the inferred tree represents the 'true tree', i.e. when we can rule out any data or branching artefacts and the tree depicts the evolutionary pathways, clades are monophyletic (holophyletic).

The common ancestor does not need to be a single Eve, it can be a population, a species, a species aggregate or even a genus, for practical reasons — genera are typically the lowest taxon, we can trace back in time and space.

Hence, inferring just a parsimony tree, even when we fix the topology to (sort of) match genetics (and the one or other trait**), can tell us very little about the evolutionary significance of Eocene Castanopsis-like fossils found throughout the world, even when their similarity to modern species thriving in tropics and subtropics of East Asia is striking.

Early fossil record of Fagaceae and Castaneoideae mapped on Scotese's (2014) palaeoglobes (GoogleEarth overlays, © Scotese 2013; The Easter Egg 2019).

As I will elaborate in detail below any fossil that can be linked to the morphologically paraphyletic Castanopsis (Castanopsoidea, Cs rothwellii) can either share an inclusive or exclusive common origin with the modern-day species, i.e. be a direct relative of the modern-day Castanopsis, i.e. a genus-crown fossil, or a stem-fossil (of Castanopsis, Lithocarpus, or any other core Fagaceae, as we will see). And, the further we go back in time, the more likely it will be to find Castanopsis-like Fagaceae fossils. If divergence of modern genera started between 80–60 Ma, we can expect finding Castanopsis-like fossils prior to 50–40 Ma that are not direct relatives (precursors) of modern-day Castanopsis. While a Castanopsis-like fossil that is only 30 or 20 Ma old, may be a modern Castanopsis.

There's no way to tell them apart unless we can associate the fossil with a distinct intra-generic lineage, i.e. a lineage within Castanopsis, e.g. a group of closely related species, and define a suite of derived characters ('apomorphies' according Hennig) diagnostic for this intra-generic lineage (ideally "autapomorphies" – unique and derived traits). Which would require to unfold the two artificial composite taxa into species groups that are genetically and morphologically coherent. Which may be tricky (Cannon & Manos 2003); also, such an approach would have been detrimental to Wilf et al.'s narrative of a "South Route".

What about the alternatives?

Just by looking at a strict consensus cladogram of MPTs, we have no way to tell how meaningful a clade of interest is. A minimum requirement would be to test alternative placements. Being faithful followers of the HCC, Wilf et al. (2019a) didn't bother (see also: More on networks placing fossils, such as Eocene lantern fruits):

Phylogenetic analysis strongly supports our taxonomic assignment, placing the fossils either as sister to living Castanopsis or within the Castanopsis crown group (Fig. 3 and Table 1).

"Strongly supports" is a bold statement given that no support test, means, measure, data analysis are provided or applied (for the lantern fruits, Wilf et al. 2017 did an actual DNA-scaffold analysis, to "proof" the obvious, and ignore everything else). Especially in the light of the results 30 years earlier that showed Castanopsis cannot be resolved even based on 25/26 morphological traits. Using Wilf et al.'s data corrected for apparent errors, Denk et al. (2019) did test alternative placements, and showed the seven characters are not really conclusive. Here's their figure with a few extra annotations:

A – Reconstructed as ancestral within all Fagaceae; P – as primitive within core Fagaceae (potential core Fagaceae symplesiomorphies or inherited ancestral traits); D – as derived, blue: shared with other core Fagaceae (potential homoiology), green: only found in Castanopsis (potential autapomorphic variation).

The problem with the seven characters – one of which is parsimony-uninformative, three are variable within the target genus, two are linked – is that the best alternative(s) (lush green) has many, only a teensy-weensy bit worse (one step more) alternatives (light green). Denk et al.'s analysis shows that Castanopsis includes living fossils, having the same character suite than the reconstructed ancestor of all core Fagaceae! The alleged crown-group Castanopsis fossil has only one derived trait: cupules with partially dehiscent valves in contrast to completely dehiscent ones. Which is something only recorded for both composite Castanopsis taxa, however, only as intra-generic variation; so not really the much needed autapomorphy, a uniquely derived, genus-conserved trait, but a potentially genus-restricted homoiology (if you are unfamiliar with the term, see Has homoiology been neglected...) And vague. Other Castanopsis species and Fagaceae genera have either completely dehiscent valvate cupules or "hemispheric indehiscent" ones. "Well supported"? I call this a leap of blind faith (or fraud-ish: The grey zone between obvious and less obvious scientific crankery).

As response to Denk et al., Wilf et al. (2019b) produced the following graph adding three characters, arguing the above (actual phylogenetic) analysis is "misleading" because the original 7-character set is not enough to accomodate all Fagaceae (well, cf. Crepet & Nixon's 1989, fig. 1, based on 25/26 characters, failing to resolve modern Castanopsis). Which wasn't the point (see above) but who cares? For sure not the friendly, impact-conscious editor of Science and the pal reviewers.

Ups, they did it again. No support values, no topology tests, no nothing (for sure no link to the used matrix/TNT file). All that matters, is the tree's topology (which may well be hand-drawn or an artefact*).

By adding just three relevant characters to the Denk et al. scaffold to accommodate the genera they added (Table 1), the fossil Castanopsis rothwellii is placed only with Castanopsis in the single [text says: two**] most parsimonious tree (Fig. 1). We note that even when the same morphological data are used alone, without any scaffold, the fossil resolves with the Castanopsis fissa group [true for most MPTs but Castanopsis dissolves*].

A curiously blunt expression of extreme tree-naivety exhibited by highly-published palaeobotanists (see GoogleScholar profiles of Wilf and Nixon), some of which can look back on a long history in related fields, not necessarily phylogenetics but cladistics (Cladistics vs. Phylogenetics: what's the difference?). Or, the tree-naivety they expect from the readers.

Thirty years earlier, Crepet & Nixon (1989), who had no molecular backbone topology at hand, at least tested alternative placements for their fossils by moving the fossils through the backbone tree inferred for the modern-day taxa (stippled and fat lines in their fig. 3 shown above):

Based on the [undocumented] characters used in the cladistic analyses, the fossil [Castanopsoidea] has affinities with the genera Castanopsis and Chrysolepis or the common ancestor [sic!] of these genera (Fig. 1). The addition of phenetic [continuous] characters (i.e., style and valve length) not used in the cladistic analyses suggest that the fossil is actually more similar to modern Castanopsis than to modern species of Chrysolepis.

Would be interesting to know where Wilf et al.'s 'scaffold' would have placed Castanopsoidea, away from Castanopsis? Drawn to the core Fagaceae root? Why Wilf et al. don't bother about this? A Castanopsis-like lineage connecting the Eocene of South with North America simply wouldn't have fit the narrative, so Castanopsoidea must be just some extinct Fagaceae:

Otherwise, the diverse North American reproductive fossils that are similar to Castanopsis belong to extinct genera.

* If one corrects for the acknowledged "typographic error" (Fagus scored as one-flowered, but has two flowers) and fuses the information in Wilf et al. (2019a), table 1 and Wilf et al. (2019b), table 1 adding characters 1–7 as scored by Denk et al. (2019) for the missing genera (trigonobalanoids, Quercus), one can infer 37 most parsimonious trees using PAUP*'s exhaustive branch-and-bound algorithm (tree-length 30 steps, CI = 0.80, RI = 0.81). 31 MPTs place Castanopsis rothwellii as sister to "C. fissa group", but 25 of them place the modern Castanopsis in different clades. In the logic of the HCC, clades = monophyla, the modern genus Castanopsis is either para- or polyphyletic. In the remaining six equally optimal trees, C. rothwelii is part of a (hard) Castanopsis polytomy, the inclusive but unresolved Castanopsis clade seen in Wilf et al.'s (2019b) fig. 1. The graph on the right (click to enlarge) shows the strict consensus network of all MPTs (splits seen in the comb-like strict consensus tree are highlighted by blue colour; edge labels give number of MPTs with this split: blue – fit with molecular data, red – conflicting with molecular data). One has to be truly tree-blind or data-ignorant to consider this a) "strong support" for including C. rothwellii in modern-day Castanopsis (poorly resolved), and b) interpret Cs rothwellii as the ancestor of all living Castanopsis ("South Route" hypothesis of Wilf et al. 2019a). Wilf et al. don't document the matrix used in the response to infer their 'scaffold' tree, the reconstructed matrix used here is included in my Fagaceae collection on figshare.

Useless cherry-picking

We can assume that Wilf et al. (2019b) chose the three additional characters (from a matrix of originally 25/26) carefully to make their point. However, to make a point, they would have, at least, had to repeat the very simple, parsimony-based (so, not heretic) topological test of Denk et al. (2019).

Maybe they didn't because they consider any test being heretic. The HCC teaches that only the most-parsimonious tree(s) shall be considered! Trees that are one step longer are less parsimonious, hence, irrelevant.

Or they didn't because adding three characters (two symplesiomorphies/ homoiologies of core Fagaceae, one parsimony-uninformative intra-generic variability) didn't change anything, or, rather, made things worse.

Most parsimonious placements of the Patagonian fossil (Cs rothwellii) within a molecular-informed backbone tree. White squares: ancestral states (as inferred by parsimony ancestral state reconstruction), coloured and black squares (potentially) derived states. Top: character #1, bottom character #10 (in case of character #7, the parsimony ancestral state reconstruction cannot resolve the ancestral state). The asterisk indicate the Castanopsis character Wilf et al. coded differently to avoid "imprecision" (Wilf et al. 2019b) at the cost of accuracy.

The best placement (under parsimony) of the fossil is at the branch representing the ancestral core Fagaceae, any placement within the castaneoid part of the tree requires only one step (change) more**. The (slightly?) younger North American Castanopsoidea infructescences can be directly derived from the Patagonian fossil (Cs rothwellii) and show one or two additional derived traits exclusive or nearly exclusive to Castanea-Castanopsis. If Cs rothwellii shares an inclusive common origin with (Castanea-)Castanopsis, the "extinct genus" Castanopsoidea does, too (even more).

Ten characters may be not enough for a morpho-phylogenetic placement, parsimony is the most conservative, least decisive optimality criterion (Large morphomatrices – trivial signal). Neither do they suffice for a scaffold-without-DNA as done by our expert palaeocladists. The 'scaffold' here only means that we feed a backbone topology into The Holy Tool and then let it find the most parsimonious placement for the fossil in this predefined tree — where can we put the fossil, so that additional (usually convergent) changes are minimised? Tree-blinded, Wilf et al. consider it "strong support" when it's placed next to Castanopsis in their 'scaffolds', or morpho-based tree (Wilf et al. 2019b; quote above):

The fossil Castanopsis rothwellii is placed only with Castanopsis in the most parsimonious tree ... without any scaffold, the fossil resolves with the Castanopsis fissa group.

I say it's phylo-sham when pretty different alternatives require the same number of steps or even one less.

Zero distance = sister taxon. No tree inferences needed (click to enlarge).

**To ensure everything is placed where wanted, Wilf et al. coded modern Castanopsis in a way that it differs by 0–4 characters from their fossil (and in reality 1–4), the latter differing at least one character (more) from any other Castaneoideae (including the second Castanopsis). Zero distance = sister taxon, pure parsimony "scaffold" tree magic!

The 0 vs. 0.1 difference of Cs rothwellii to either Castanopsis also explains why Wilf et al. (2019a) say in the text a single optimal 'scaffold' tree was found, but show a, most probably faked, consensus tree of two MPTs in their fig. 1. The single-optimal placement of the fossil is a sister to the 'Castanopsis fissa group', i.e. embedded within the modern-day Castanopsis. Their new fake fig. 1 matches however their original figure, which could be still interpreted towards their narrative.

A morphological mess, and how to clean it up

Why is it so difficult to unambiguously place the fossil although it only differs by a single character (when coded accurately, Denk et al., but "less precise", Wilf et al. 2019b) from the consensus of Castanopsis?

Here's a graph depicting all 45 MPTs that can be inferred from the 10-character matrix, a rooted strict consensus network with averaged edge lengths.

Rooted (with Fagus) strict consensus network of all 45 MPTs that can be inferred using the branch-and-bound algorithm for the 10-character matrix. The fossil is either placed ("resolved") as sister to Quercus in a clade including Castanopsis or Lithocarpus + Notholithocarpus, or as sister to a Castaneoideae clade**. Reddish edges conflict with the current molecular framework, bold green edges agree.

When we just take the morphomatrix, we can see that phenotypic evolution in Fagaceae is chaotic, struggles hard to find a consensus, not to mention clades reflecting monophyla (green edges in the strict consensus network; the strict consensus tree would be a 2-branch comb). In addition, each MPT may be affected by branching artefacts because the signal from the matrix is not tree-like at all.

A better way to visualise the principal, usually non-trivial signal in the matrix, is a distance-based neighbour-net (as always when it comes to placing fossils unless your matrix is very gappy).

Neighbour-net based on mean morphological distances. The grey-cyan tree gives the best-fitting molecular phylogenetic synopsis (based on nuclear differentiation patterns not just inferred and published trees – I like to know the data, I work with (Why ... map trait evolution along networks, pt. 1).

From the net it's obvious that morphological similarity poorly reflects inter-generic phylogenetic relationships in core Fagaceae. It's important to realise that in contrast to HCC lore, parsimony and distance methods are very much the same (Fossils and networks 1...; see also Felsenstein 2004). The similar two taxa are, the higher the chance they will be placed in the same subtree, i.e. clade. Nevertheless, based on the graph and with the molecular-inferred phylogeny in mind (mapped onto the graph), we can say a lot about what a certain phenotype seen in a fossil inflorescence means for its phylogenetic affinities.

Lithocarpus/ Notholithocarpus-like fossils could only be tentatively interpreted based on their provenance. In lower latitude, tropical (Palaeo-)Asia, they may be Lithocarpus. In high-latitude, subtropical to temperate (Palaeo-)North America, it may be a relative/precursor of Notholithocarpus and/or the oaks (Quercus). The putative last common ancestor of oaks and Notholithocarpus likely was more similar to the latter.
We can expect that ancestors/stem fossils of both the Lithocarpus-Chrysolepis and Castanea-Castanopsis lineages where much more Lithocarpus-/ Castanopsis-like than Chrysolepis-/Castanea-like. The last common ancestor of Lithocarpus-Chrysolepis may be impossible to discern, since both genera evolved in opposite directions (see experimental graph below).
Castanopsis is placed like an ancestor (close to the centre of the graph, no terminal edge, see also A (wal)nut to crack..., A bit of heresy..., Trivial data, ..., Two papers you may want to read...). Four of the ten characters are coded as polymorphic, the intra-generic polymorphism obviously includes the ancestral states (see experimental graph below) for the core Fagaceae (the distance between e.g. "0" and "0/1" is 0 in standard implementation; likewise parsimony doesn't count a step). Thus, any primitive-suite Castanopsis-like fossil will be difficult to interpret: It could be an

ancient extinct relative of core Fagaceae,
a core Fagaceae stem fossil,
a precursor of Castanea-Castanopsis, or
a member/precursor of Castanopsis.

Potential Castanopsis crown-fossils would be those bridging between the hypothetical primitive and fully derived Castanopsis morphs, and potentially more similar to Notholithocarpus and Lithocarpus.
Cs rothwellii is either not a Castanopsis at all, or a uniquely derived Castanopsis, a descendant not an ancestor (which clashes with its age, Wilf et al.'s storyline, not its place, but see the graph below). It associates itself with Castanopsis in topologically constrained trees because the latter is the most similar and least-derived modern taxon in the data set. The lack of an according edge bundle shows that Castanopsis and Cs rothwellii lack a suite of shared derived traits (either due to random convergence, homoiologies, synapomorphies). Shared, unique traits would produce tree-like (narrow) terminal structures in graphs like this; shared, but not unique traits would build up more or less complex and prominent box-like structures connecting the fossil(s), Castanopsis and Castanea or Lithocarpus + Notholithocarpus.
Quercus-, Castanea- or Chrysolepis-like fossils are probably Quercus, Castanea or Chrysolepis. All three genera are derived with respect to Castanopsis, their sister genera (Notholithocarpus, Castanopsis, Lithocarpus), and the core Fagaceae ancestor (graph below).
A fossil placed in between Castanea and Chrysolepis represents a derived (temperate?) core Fagaceae, but its phylogenetic affinities cannot be further determined. It could be a sister of either one, or an extinct (temperate) lineage that underwent similar morphological adaptations (the less derived sister genera Castanopsis, Lithocarpus are elements of the Palaeotropics, not found beyond the hot subtropics; Castanea and Chrysolepis are fully temperate tree genera).
Assigning a fossil to trigonobalanoids would be straightforward, but not distinguishing between crown-, stem- or extinct sister-taxa. The genetic distance between the three monotypic modern genera strongly suggests that they are the surviving iceberg tips of their lineage. The arcane 1989 matrix possibly includes more traits to resolve further fossil members of the Trigonobalanoideae. Or maybe not: the monophyletic Trigonobalanoideae were resolved as a grade, "basal" to the unrelated oaks.
Fossils of the lineage leading to beeches, Fagus, should be easy to spot (indeed, they are: The challenging and puzzling ordinary beech).
Anything that increases further the complexity of the graph, is probably an extinct Fagaceae lineage that convergently (or in parallel) evolved certain traits.

A little experiment to further resolve the primary signals in the matrix; adding the reconstructed (parsimony) ancestor of the core Fagaceae to the taxon set and replacing the polymorphic (modern) Castanopsis by two artificial OTUs, one collecting all primitive, the other collecting all advanced traits scored as part of the polymorphic characters. The arrows indicate the morphological evolution (derivation) on the background of the molecular-phylogenetic synopsis. The American Eocene fossils don't fit within the framework of modern taxa, and seem to constitute a lineage, "Eocastanopsis", that evolved independently from the core Fagaceae ancestor.

Behind the clades

Traits that are beneficial will be positively selected; the probability for change cannot be the same for each trait, nor are all changes Dollo-like mutations; reversals – secondary loss of a traits – is possible. Or parsimonious: a positively selected trait may be Dollo-like and evolved/ fixed many more times than reconstructed via character mapping. Furthermore, we can expect temporal convergence (discussed in Bomfleur et al. 2017): even if a trait, or set of traits evolved in adaptation to a niche shift, is today only found in one (surviving) lineage, it (they) may have been convergently evolved in an ancient lineage. 50 million years is a lot of time for evolution to fool around.

Morphology-informed clades are, on their own or as part of a total-evidence (TE) approach, poor indicators for monophyly or close relationship. Morphological similarity or (near-)identity, the latter will result in artefactual sister placements in scaffolds and TE trees, can be severely misleading.

Thus, the critical question is not (challenging not a few palaeo-phylogenetic papers and "phylogenetically placed" fossils):

Is my fossil placed as sister to X in a MPT (or TE-tree)?

But:

What characters support this placement?

What quality do they have from an evolutionary point of view?

More Hennig less Farris. More testing and thinking, less tree-belief.

When one has molecular data (gene banks include 250,000+ accessions of Fagaceae), there is no reason to not make use of it. E.g. to put up hypothesis for morphological evolution and estimate the probability of change within lineages, along the branches of preferred tree (or trees). Easy to do, hence, hardly seen in publications of the HCC reviewed by brothers of faith:

How many potential synapomorphies (unique, shared, lineage-conserved, derived traits), homoiologies (shared, derived, and restricted to but not found in all members of a clade), symplesiomorphies (unique, shared but primitive within a clade) support each clade?

Trivial under parsimony.

Better done in a probabilistic framework, e.g. using branch-length-aware maximum likelihood as the simplest of all methods (also fun, visual mapping along networks: Why ... map trait evolution using networks, pt. 2).

Ancestral state reconstruction using maximum likelihood. Done with Mesquite, nothing fancy such as BAMM, RAxML-NG or otherwise R-based, e.g. Potts & Grimm 2017.

For the above, I read in the topology used for the parsimony mapping and optimised branch-lengths and the substitution model(s) with RAxML using the consensus sequence matrix in the linked Why... map trait evolution GWoN post (available via figshare). Then read in the branch-length optimised tree into Mesquite and chose "trace character history" with standard Mk model (Lewis 2001). Polymorphic states (ML requires unambiguous tip states, hence, polymorphisms are coded as additional states) are counted as such, i.e. the probability for the polymorphic state is added to the monomorphic states' probability. The main difference between MP and ML is that the roots are much too long, and the characters much too homoplastic, to infer ancestral states for the all-Fagaceae ancestor. 6/10 unambiguously resolved under MP, only 1–2 under ML. Furthermore, ML tells us to expect more flexibility in gaining or losing a potentially derived (advanced) trait. Which is important to keep in mind since (most) modern genera were probably established by the early Eocene, 50 myrs ago, and that is a lot of time to evolve or modify one's trait set, to try out the same than your distant relatives.

Gold tree (as above), branch-lengths indicating (expected) character change (~ steps under MP) deduced from the ML ancestral state reconstruction above. Note that the core Fagaceae can be characterised by a symplesiomorphic character suite, three of which are found in the American Castanopsis-like fossils. The North American Castanopsoidea show 2–3 traits derived within the Castanea-Castanopsis lineage, the South American Cs rothwellii 0–1 (partially dehiscent cupule valves).

Irrespective whether we used branch-length (change over time) ignorant MP or branch-length aware ML, the data lacks traits of high (unambiguous) diagnostic value but some (likely) monophyla (and paraphyla) can be diagnosed by distinct character suites. For instance, primitive core Fagaceae – a paraphylum fide Hennig which would include the modern-day castaneoid genera Castanopsis, Lithocarpus and Notholithocarpus – are characterised by unisexual and mixed inflorescences with one female flower per hemispheric-indehiscent (ML) or completely dehiscent, valvate (MP) cupules with scaly appendages, and an unexpanded stigmatic area. Pretty fit with the Patagonian fossil with its intermediate dehiscence (partially, valvate). Castanea, Chrysolepis and oaks modified this, within core Fagaceae primitive, character suite.

But there is no stable derived character suite that defines modern-day Castanopsis. Derived traits (convergences, homoiologies, reversals) are shared either with the sister genus Castanea or other Fagaceae. Even if we would find an inflorescence of a modern Castanopsis sp., it might be impossible to judge whether it is a Castanopsis, or a primitive Castanea or part of the primitive core Fagaceae paraphylum. At least based on Wilf et al.'s 'enhanced' 10-character set.

Consequently, a Castanopsis-only clade including the fossil(s) but excluding other castaneoid core Fagaceae has very little support: it would be solely defined by the potential to produce a valvate, partially dehiscent and asymmetrical cupules in contrast to symmetrical cupules that are either 'hemispheric indehiscent' (ML: ancestral within core Fagaceae, a symplesiomorphy; or MP: derived, a lineage-specific homoiology) or with valves that open completely (primitive or derived within Fagaceae, a plesiomorphy or convergence). Any other trait set is either ancestral within core Fagaceae or Fagaceae at large, or highly instable along the Fagaceae phylogeny. The hypotheses made just based on the distance-based neighbour-net hold (and, hence, Denk et al.'s "misleading" critique):

Any ancient member of the core Fagaceae would likely have had "Castanopsis-like" inflorescences/ -infructescences.

Be it a precursor or early sister of any of the modern genera or morphologically disparate sibling lineages (Chrysolepis + Lithocarpus; Castanea + Castanopsis + Notholithocarpus + Quercus).

And we would expect castaneoid pollen (Denk & Grimm 2009) and foliage, reported for Castanopsis rothwelli but also Castanopsoidea sites, and widespread in the Northern Hemisphere since the Cainozoic (an overview about fossil record of Fagaceae can be found in Grímsson et al. 2016a; see also Grímsson et al. 2015, 2016b).

A phylogenetic plot of some Fagaceae fossils (modified after Grímsson et al. 2016, fig. 2). The colours give indicate the geographic distribution: green – western Eurasia (W. Asia + Europe, incl. Iceland); yellow – East Asia (east of Indus Valley/Ural); red – North America (Americas north of Isthmus of Panama); pink – South America. Stippled lining: modern genera without a fossil record (or, not recognised as being part of these lineages). Abbrev.: C. = Castaneoideae, Cs = Castanopsis.

Interpretation of Eocene Castanopsis-like fossils

Interpreting Castanopsis rothwellii as the ancestor of all modern-day Castanopsis, the only indication Wilf et al.'s provide for their spectacular "South Route" hypothesis, can be rejected based on their own data. Their "phylogenetic trees" are HCC-lore doodles. Palaeo-cladistics too often ignore something Darwin and Wallace came up with: give time and space, you'll have evolution, and, given enough time, space and selection pressure, convergence. And something Hennig pondered: similarity can have two reasons, uniquely shared derived traits (his 'synapomorphies') and shared primitive, ancestral within a lineage, traits (his 'symplesiomorphies'). Tree inferences are ignorant of these issues (parsimony utterly – distances pretty much – Bayesian inference and maximum likelihood least).

A data(matrix)-based hypothesis about morphological evolution in core Fagaceae. The American fossils can be easily interpreted as an extinct (tropical) Eocene lineage. All their shared (potentially) derived traits have been evolved (fixed) at least twice in near or distantly related modern-day lineages. Green: lineage-diagnostic traits, unique within core Fagaceae; red: traits shared by more than one lineage.

Let's for the sake of the argument assume that partially dehiscent, asymmetrical cupules really evolved only once in Fagaceae, and only within the Castanopsis lineage (Wilf et al. 2019b provide no reference or information at which point cupules can be considered to be "asymmetrical" in opposite to "symmetrical"). Then, the fossil cannot be ancestral within this lineage, since there are modern species that are more primitive with symmetrical 'hemispheric indehiscent' or valvate, completely dehiscent cupules while being more derived in other aspects. Those more primitive modern-day Castanopsis spp. must be sister lineages of the "Gondwanan" partially-dehiscent-asymmetrical-cupules-lineage. Mother Nature dices a lot, derived traits may get lost, and Bauplans may get simplified. However, Nature rarely turns back time. Organisms rarely devolve (regress) and not without a very good reason (e.g. massive changes in their niche/environment, which can be ruled out for Castanopsis). The Eocene fossils (North and South American) hence would not only represent crown-fossils but should be deeply embedded among modern-day Castanopsis.

Why Wilf et al. don't discuss Castanopsoidea: if their new fossil is a crown Castanopsis, Castanopsoidea is one, too. Even more so.

On the other hand the data and signal exploration revealed additional derived traits diagnostic for the Castanea-Castanopsis lineage such as three flowers per cupule, a lineage-restricted homoiology, and spinose cupule appendages, a core Fagaceae homoiology only shared with Chrysolepis (the "Nothocastanea" of North America). Interestingly, features possibly shared with the "extinct" genus Castanopsoidea of the early-middle Eocene of North America but not in the "crown-fossil" Cs rothwellii.***

It all adds up. From molecular clock estimates (discussed in Grímsson et al. 2016b; see Hipp et al. 2019 for oaks) and the fossil record (see the map-through-time-and-space-plot above), we know core Fagaceae must have existed in the Upper Cretaceous or around the KT boundary, ~ 80–60 Ma (has to be older than the oldest oak, but younger than the first Castaneoideae pollen). 8–28 myrs are more than enough to radiate and diverge, including an (extinct or precursor) American lineage that migrated till Patagonia during the Paleocene-Eocene greenhouse phase. And they were probably not the only Northern Hemispheric elements jumping on the opportunity.

Whether this lineage represent an extinct American(/ early European; Sadowski et al. 2018) sister lineage of the (sub)tropical Asian Castanopsis, or the Eurasian Castanea + Castanopsis, or even a precursor/sister of extratropical Notholithocarpus + Quercus (possible) or Chrysolepis (least likely) cannot be judged based on inflorescences. Modern-day Notholithocarpus and Chrysolepis are obviously the product of parallel morphological evolution within two old (~60+ Ma) lineages of core Fagaceae:

Chrysolepis, originally considered congeneric with Castanopsis, paralleling Castanea (adaption to a fully temperate niche?).
Notholithocarpus, lit. the 'false Lithocarpus' and only recognised as its own genus in 2008 based on molecular data (Manos et al. 2018), parallels Lithocarpus although the two genera have no ecological or climatic niche overlap and are genetically as distant as one can be in the core Fagaceae.

Based on modern-day phenotypes (the very boxy morpho-based net) and their molecular-based phylogenetic relationship (mapped trait evolution; see also Why ... on networks, pt. 2), one has to consider the strong possibility that "Eocastanopsis" of the Americas, core Fagaceae with striking similarity to Castanopsis, are not Castanopsis-relatives at all. They could be "Nothocastanopsis": a lineage of extinct core Fagaceae mimicking the Eurasian modern-day Castanopsis (and Castanea to some degree) because they occupied comparable niches back in the day. And while these ancient phenotypes survived in some species of Castanopsis in the Palaeotropics, a region largely unaffected by the Oligocene cooling, they went extinct in the Americas (and Europe; Sadowski et al. 2018).

*** I see no point in naming something that is Castanopsis-like something else, e.g. Castanopsoidea, just because the authors felt it's not related to modern-day Castanopsis (or where unsure), while naming a more primitive Castanopsis-like fossil Castanopsis (because the authors needed a hook for their narrative). Such naming would be even illogic in the HCC philosophical framework (such illogic is however common in phytocladistic literature). Both are resolved as sister to one of the two taxa representing Castanopsis — if one uses Wilf et al.'s "more precise" coding (inflorescence sexuality unisexual, and not unisexual and mixed) even identical to their Castanopsis consensi in all scored traits. When you claim clades are all one needs to name a fossil, you shouldn't use two different ones when they are part of the same clade, especially not when being nested inside it! Also, it is a bad but enforced during review tradition in palaeobotanical naming to come up with new names for fossils that are indistinguishable from modern taxa and, in the process, diagnoses that are non-exclusive regarding one or several modern-day genera. The cleanest (but impossible, because 'uncladistic' = heretic) taxonomic solution would be accepting Castanopsis as a increasingly paraphyletic genus the further we go back in time and away from the modern-day distribution area, and consequently name all Castanopsis-like fossils Castanopsis (in analogy to e.g. our, Bomfleur et al. 2017, definition of Claytosmunda).

Fear the inquisition, beware palaeo-lore

You may say what a fuzz about so few data. There are probably worse, more misleading papers out there. Sure, but: the first and second author are extremely influential people within palaeobotany, who can publish sham pseudo-phylogenies in the best of all possible journals, Science. Similar pseudo-phylogenies are usually published in specialised journals ensuring little or pal review (like e.g. Rothwell et al., 2009, published in a special issue edited by faithful followers). Them and like-minded, equally merited (U.S.) colleagues decide as (anonymous when hostile) "expert" reviewers which palaeobotanical papers are published and how, they control the market and classical journals in the field (such as Int. J. Plant Sci.), dragging down since now 30 years an entire scientific field to their level. And there's no resistance: Wilf et al. could publish their pseudo-phylogenetic tree and their confusing, misrepresenting and misleading response because they don't need to fear critical reviews. No one can afford to mess with them. They can easily enforce their believes — better to have a prominent co-author than a hostile reviewer. Ensuring that every young palaeobotanist is forced into their 1980's philosophical corset (being a geologist-geneticist, I escaped with ease). Including propagating this compelling logic for why they never show support values (e.g. cited in Rothwell et al. 2009, providing a "cladistic test" of Friis et al.'s 2007 conclusions; What I was not allowed to show #1...):

... support values, whether low or high for particular groups, would only mislead the reader into believing we [Rothwell and Nixon] are presenting a proposed phylogeny for the groups in question. Differences among most-parsimonious trees are sufficient to illuminate the points we wish to make here, and support values only provide what we consider to be a false sense of accuracy in these assessments. ... [Our phylogenetic inferences should] not [be] view[ed] as reasonable estimates of vascular plant phylogeny.
[Rothwell & Nixon 2006, p. 739]

Any of Wilf's, Nixon's, Crepet's, Rothwell's etc. papers that include a cladogram demonstrate that this group of U.S. based palaeo-nigh-experts are not only tree-naive but can be tree-stubborn and, when no available (molecular) tree complies their views, tree-ignorant. "Cladistic analyses" or "phylogenetic trees" in their papers just represent their convictions, "illuminate the points [they] wish to make", and are not the result of any properly done, open-minded analyses. Wilf et al. (2019b) make this explicitly clear in their response to Denk et al. (2019).

Because the fossils are castaneoid in all features, we did not include all Fagaceae in our original analysis [Wilf et al. 2019a]... Unfortunately, Denk et al. do not advance the discussion. Only time and many more fossils [not if Castanopsis-like, see above], not negative evidence and misleading assertions [Denk et al. 2019], will tell where else the Fagaceae occurred.

Topological tests, ancestral state reconstructions and ignored data (Castanopsis-like fossils occur in the Eocene of North America and Europe; Castaneoideae pollen, foliage and seeds all across the Northern Hemisphere but not in Antarctica or Australia) are "negative evidence and misleading assertions" but "many more" fossils placed using unsupported, biased tree-doodles, which should not be viewed as "reasonable estimates" at all (Rothwell & Nixon 2006), are positive evidence and for sure not misleading? Palaeobotany, thou may rest in peace!

Instead of providing something more up to date like this ...

How to assess the phylogenetic information content of a character suite of a fossil. Also just 11 traits, mapped on the molecular tree (based on a complete gene bank harvest; figure from Friis et al. 2015).

... this ...

Morpho-based neighbour-net (above) and result of evolutionary placement algorithm (below), which finds the optimal placement of a query (here: fossils) within a given tree (here: fully and unambiguously resolved molecular tree). This (figs 8, 9 of 13) and more can be found in Bomfleur et al. (2015; open access and open data [ZIP-archive]).

or this:

Yes, I doodle, too. But this one is based on a probabilistic trait evolution analysis using categorical and continuous characters (Potts & Grimm 2017; Grímsson et al. 2018). The fossil, Pseudowinterapollis (Ps.) agatdalensis. is "just" a Winteraceae pollen tetrad from Paleocene of Greenland (plants did get around a lot in that time, but not from North to South America?).

May look like magic but everyone could do it.****

What I dissected here at the example of a small but not trivial matrix holds for any other and larger matrix. What is obvious for Wilf et al. (2019a,b) – limited data, no support, misinterpretation and misrepresentation of results, severe systematic bias – may be less obvious in other cases.

World-leading palaeobotanists, and with them any dependent researcher (money-, publication-wise or married as in the case of Wilf and Nixon's co-authors), are not only the last surviving true followers of the HCC ...

The cladists of that [bygone] era had accepted a number of points as an intellectual package. At one point in the mid-1980s I tried to summarize the package and came up with these points ...
[quoted from Felsenstein 2001, full list at the end of this post]

... they also seem to have no idea, and, for sure, no interest in getting an idea, about the signal in their (heavily cherry-picked) matrices, on which they b(i)ase their meaningless stick graphs. After all, it's not about placing a fossil in a phylogenetic framework but (one can't repeat it often enough) "illuminate the points" their majesties and eminences "wish to make" (The 10 Holy Commandments...)

So, when you cross a palaeo-paper just showing a naked strict consensus cladogram with no support values and/or some assessment/ discussion of alternative placements, there is only one thing to do.

Cherish the description and treat everything else as palaeo-lore.

And daters should check twice before using a "phylogenetically placed" (palaeo-speak for not obvious) plant fossil as age prior for their node dating (Most common errors ...) Better to go for the 'apomorphy-based' (palaeo-speak for obvious), just to make sure. Or do a quick re-analysis of the data. When available.

**** In case you wonder, why there are not dozens of papers like this, co-authored by me. The main work lies in getting together a morphological matrix to start with, which I'd be incapable of. You need taxonomists familiar with the organisms, the situation in the field (especially when studying fossils of modern lineages) and the collections, you can't just go and blindly re-use what people wrote in papers (I did, for my very first phylogenetic and very last cladistic analysis: Grimm 1999; compare the results with those of Hermsen et al. 2006, the accredited experts). The matrices have to be as comprehensive and unbiased as possible, they should capture all visible aspect of a fossil, ideally the modern diversity, and not be filtered towards a preferred tree or to ease the parsimony analysis. Hence, many published matrices are useless — if documented at all. Dropping taxa and characters messing with the tree inference has long been an industrial standard. And it precludes co-operation with heretics like me, who could probably get much out of the data assembled by a student or Ph.D. but insist on novel methods and full transparency. When presenting my heretics on (palaeo-)botanical conferences, some were interested but nothing happened once they talked to their boss.

Once the matrix is put up, it's a piece of cake. Relatively speaking. Harvesting and cleaning out gene bank data to get a molecular backbone tree can be tedious, too (all of my papers based on harvests include in the supplement a file detailing how I did it, and what I had to throw away, and why – usually because of taxonomist-unaccompanied barcoding studies). An increasing amount of neontologists make their matrices available, and more and more journals require them to, but it's not yet a common standard (Why want to publish our phylogenetic data...) Finally, publishing "unusual", "phenetic", "not standard", "too much for our readers", unpleasant to reviewers ("I don't feel...") analyses can take a while (The peer review should be transparent...)

Some manuscripts take longer to review than others (One date that is missing ...)

Data
All matrices used for this post (including the putative Wilf et al. 2019b matrix), have been included in version 2 of my Fagaceae figshare file collection.

Grimm, Guido (2020): Fagaceae collection. figshare. Dataset. https://doi.org/10.6084/m9.figshare.11603547.v2

References

Ashlock PD. 1971. Monophyly and associated terms. Systematic Zoology 20:63–69. — a simple solution but largely ignored until this day (except e.g. in Bomfleur et al. 2017, kudos to the first author for digging out the paper).
Bomfleur B, Grimm GW, McLoughlin S. 2015. Osmunda pulchella sp. nov. from the Jurassic of Sweden—reconciling molecular and fossil evidence in the phylogeny of modern royal ferns (Osmundaceae). BMC Evolutionary Biology 15:126. — extraordinary fossil, unique set of analyses (still, I suppose). Triggered recognition of subgenera in Osmunda as genera for PPG I, why we had to change the generic epithet soon after (Bomfleur et al. 2017).
Bomfleur B, Grimm GW, McLoughlin S. 2017. The fossil Osmundales (Royal Ferns)—a phylogenetic network analysis, revised taxonomy, and evolutionary classification of anatomically preserved trunks and rhizomes. PeerJ 5:e3433. — it says "evolutionary classification" for a reason (we formalised holophyla, paraphyla and monophyla, i.e. groups where we can't say whether they are holo- or paraphyletic), extremely long read, worth every bit, sort of a textbook (includes e.g. a glossary of important terms to describe Osmundales rhizomes).
Cannon CH, Manos PS. 2003. Phylogeography of the Southeast Asian stone oaks (Lithocarpus). Journal of Biogeography 30:211–226. — old but still a fine paper, the genus is understudied (but maybe not for long).
Crepet WL, Nixon KC. 1989. Earliest megafossil evidence of Fagaceae: phylogenetic and biogeographic implications. American Journal of Botany 76:842–855. — as typical for U.S.-experts, the second part of the title is misleading.
Denk T, Grimm GW. 2009. Significance of pollen characteristics for infrageneric classification and phylogeny in Quercus (Fagaceae). International Journal of Plant Sciences 170:926–940. — published against fierce resistance during review (the newly appointed editor apologised for having being unaware about how fierce reviewes can be in palaeobotany) and basis for the current oak classification (with much help from phylogenomics, of course).
Denk T, Grimm GW. 2010. The oaks of western Eurasia: traditional classifications and evidence from two nuclear markers. Taxon 59:351–366. — mixed to hostile reviews, because we only show a tree to demonstrate that trees mislead, but a helping editor.
Denk T, Hill RS, Simeone MC, Cannon C, Dettmann ME, Manos PS. 2019. Comment on “Eocene Fagaceae from Patagonia and Gondwanan legacy in Asian rainforests”. Science 366:eaaz2189 — to-the-point critique of Wilf et al. (2019a).
Felsenstein J. 2001. The troubled growth of statistical phylogenetics. Systematic Biology 50:465–467. — a must-read.
Felsenstein J. 2004. Inferring phylogenies. Sunderland, MA, U.S.A.: Sinauer Associates Inc. — a must-buy (still, the basics haven't changed).
Friis EM, Crane PR, Pedersen KR, Bengtson S, Donoghue PCJ, Grimm GW, Stampanoni M. 2007. Phase-contrast X-ray microtomography links Cretaceous seeds with Gnetales and Bennettitales. Nature 450:549-552. — became a co-author during review by accident; my best-cited, career-wise most important and, from an artistic viewpoint, most embarrasing paper.
Friis EM, Grimm GW, Mendes MM, Pedersen KR. 2015. Canrightiopsis, a new Early Cretaceous fossil with Clavatipollenites-type pollen bridge the gap between extinct Canrightia and extant Chloranthaceae. Grana 54:184–212. — there is a reason one frequent co-author is missing, but in contrast to papers by accredited palaeo-"phylogeneticists" (Crepet & Nixon 1989, Rothwell et al. 2009, Wilf et al. 2019a) the second part of the title holds.
Grimm GW. 1999. Phylogenie der Cycadales. Diploma thesis thesis. Eberhard Karls Universität. — I never attended a single phylogenetics course but was a big fan of Hennig (still are, even though I'm not a cladist and can never be).
Grímsson F, Zetter R, Grimm GW, Krarup Pedersen G, Pedersen AK, Denk T. 2015. Fagaceae pollen from the early Cenozoic of West Greenland: revisiting Engler's and Chaney's Arcto-Tertiary hypotheses. Plant Systematics and Evolution 301:809–832. — the prequel to Grímsson et al. (2016b).
Grímsson F, Grimm GW, Meller B, Bouchal JM, Zetter R. 2016a. Combined LM and SEM study of the Middle Miocene (Sarmatian) palynoflora from the Lavanttal Basin, Austria: Part IV. Magnoliophyta 2 – Fagales to Rosales. Grana 55:101–163. — Part of an invaluable (for palynologists) series of papers by my Vienna colleagues.
Grímsson F, Grimm GW, Zetter R, Denk T. 2016b. Cretaceous and Paleogene Fagaceae from North America and Greenland: evidence for a Late Cretaceous split between Fagus and the remaining Fagaceae. Acta Palaeobotanica 56:247–305. — Everyone who muses about Fagaceae evolution should read this (open-access) paper (Wilf et al. obviously haven't).
Grímsson F, Grimm GW, Potts AJ, Zetter R, Renner SS. 2018. A Winteraceae pollen tetrad from the early Paleocene of western Greenland, and the fossil record of Winteraceae in Laurasia and Gondwana. Journal of Biogeography 45:567–581. — just a single pollen tetrad in the wrong place, providing a template for how to trace and map character evolution of stand-alone fossils.
Hennig W. 1950. Grundzüge einer Theorie der phylogenetischen Systematik. Berlin: Dt. Zentralverlag. — the classic work, more cited than read, I guess (I haven't read it since my Diploma thesis: Grimm 1999).
Hennig W. 1965. Phylogenetic systematics. Annual Review of Entomology 10:97–116. — my feeling is, it lost a lot in translation.
Hermsen EJ, Taylor TN, Taylor EL, Stevenson DM. 2006. Cataphylls of the Middle Triassic cycad Antarcticycas schopfii and new insights into cycad evolution. American Journal of Botany 93:724–738. — a paper showing a tree reminiscent of my 1999 tree but based on a much better, bit of a Swiss cheese matrix lacking tree-like signal; for those interested in actual Cycadales palaeo-phylogenetics involving fossils, see e.g. Coiro & Potts, BMC Evol. Biol. (2017).
Lewis PO. 2001. A likelihood approach to estimating phylogeny from discrete morphological character data. Systematic Biology 50:913-925. — still the model for multistate data.
Manos PS, Cannon CH, Oh S-H. 2008. Phylogenetic relationships and taxonomic status of the paleoendemic Fagaceae of Western North America: recognition of a new genus, Notholithocarpus. Madroño 55:181–190. — A striking example for genetics recognising a good and valid genus, morphologists overlooked; also a cautionary tale regarding (very) similar morphologies and assumption of monophyly.
Potts AJ, Grimm GW. 2017. Ancestral state reconstruction of seven continuous and 20 categorical pollen traits scored for extant Winteraceae. Supplement to Grímsson et al. "A Winteraceae pollen tetrad from the early Paleocene of western Greenland, and the fossil record of Winteraceae in Laurasia and Gondwana". http://rpubs.com/AlastairPotts/WinterIsComing — a fully documented R-script by A. Potts for Grímsson et al. (2018).
Rothwell GW, Nixon K. 2006. How does the inclusion of fossil data change our conclusions about the phylogenetic history of the euphyllophytes? International Journal of Plant Sciences 167:737–749. — a paper that gives a unique insight into the believes of world-leading palaeobotanists (the first author knows his stones very well) claiming, however, to be experts in phylogenetics, too; not surprisingly published in a journal of blind faith.
Rothwell GW, Crepet WL, Stockey RA. 2009. Is the anthophyte hypothesis alive and well? New evidence from the reproductive structures of Bennettitales. American Journal of Botany 96:296–322. — A critique of Friis et al. (2007) using a matrix stripped from most discriminative signal (see also this draft-PDF, I put together for an application; Should we try to infer trees..., Age of angiosperms...); the "cladistic tests" in this obviously not reviewed point-of-view would fail for any other tip-taxon in the matrix, including the extant ones. Hard to add more characters, though...
Sadowski EM, Hammel JU, Denk T. 2018. Synchrotron X‐ray imaging of a dichasium cupule of Castanopsis from Eocene Baltic amber. American Journal of Botany 105:2025–2036. [PDF] — always better to not include any phylogenetic analysis than posing a poorly made one (as in Wilf et al. 2019a).
Scotese C. 2014. Atlas of Paleogene Paleogeographic Maps (Mollweide Projection), Maps 8-15, Volume 1, The Cenozoic. PALEOMAP Atlas for ArcGIS. Evanston, IL.: PALEOMAP Project/the author. — Just awesome.
Wheeler WC. 2014. Phyletic groups on networks. Cladistics 40:447–451. — sometimes you find good papers in the Holy Letters (#ParsimonyGate).
Wilf P, Carvalho MR, Gandolfo MA, Cúneo NR. 2017. Eocene lantern fruits from Gondwanan Patagonia and the early origins of Solanaceae. Science 355:71–75. — second part of title far-catching.
Wilf P, Nixon KC, Gandolfo MA, Cúneo NR. 2019a. Eocene Fagaceae from Patagonia and Gondwanan legacy in Asian rainforests. Science 364:eaaw5139. — second part of title pure fiction.
Wilf P, Nixon KC, Gandolfo MA, Cúneo NR. 2019b. Response to Comment on “Eocene Fagaceae from Patagonia and Gondwanan legacy in Asian rainforests”. Science 366: aaz2297. — not-a-response to Denk et al. (2019), but worthy of a point-per-point Res.I.P. response regarding who misleads, misrepresents, etc. The pseudo-phylogenetic not-an-analyses (fig. 1) is just the tip of the iceberg.