Reading the scientific record

Ask what is known about a species of Psilocybe and you will be handed four quite different things. A name record, giving an author, a date and a page. A type specimen, dried and in a cupboard. A sequence, deposited in a public database under an accession number. And sometimes a figure painted on a wall or drawn in a codex, with an argument attached about which fungus it shows.

All four get reported in the same register, as though each were simply a fact about the species. They are not the same kind of claim, they fail in different ways, and an account that treats them as interchangeable will be confidently wrong at exactly the points where it matters. Most of the editorial decisions behind the accounts on this site are decisions about which of the four is carrying a given sentence.

The specimen is the only thing that was there

A type specimen is the anchor of a name. Bradshaw and colleagues put the principle plainly: type specimens are the ultimate authority for applying names to other collections, which is why their study went to the trouble of sequencing 23 of them.

What that study then found is the reason to take the principle seriously. Of 71 fungarium specimens filed under Psilocybe, 52 were Psilocybe in the current sense. Fourteen belonged to Deconica, one to Kuehneromyces, three to an unplaced group within Strophariaceae, and one could not be assigned to any existing genus. Four more specimens disagreed with the taxa they clustered with and were re-examined under a microscope and given different names. The authors describe misidentification as a common issue in fungaria and in Psilocybe specimens in particular.

The label, in other words, was wrong about a quarter of the time, and the specimen was what revealed it. This is the one form of evidence that cannot be corrected by consulting something else, because everything else was ultimately derived from it.

A sequence answers a narrower question than it looks

Sequence data feels like the end of an argument. It usually is not, for two reasons that have nothing to do with the chemistry of sequencing.

The first is that a sequence is only as good as the specimen it came out of. Schoch and colleagues established the internal transcribed spacer region as the universal barcode marker for fungi, and that is cited here at abstract level only, for the selection of the marker rather than for any comparison within the paper. A barcode identifies the material in the tube. Whether that material is what the label says is a separate question, answered by the paragraph above.

The second is that the material often does not exist. Bradshaw and colleagues report that across all publicly available sequence data, 40 species of Psilocybe were represented, roughly 24 per cent of the genus, and that no species had every gene region drawn from a single voucher specimen. Their own dated estimates rest entirely on secondary calibrations, and they say so, adding that the absolute dates should be considered tentative until primary calibrations are available.

So a sequence answers: what is this particular piece of tissue related to. It does not answer: what does this name mean, or what does this species contain, or where does it grow. Those require the specimen, the literature and the occurrence record respectively, and each of those can be missing while the sequence sits there looking authoritative.

A database is a dated assertion

This is the distinction most easily lost, because taxonomic databases are the most convenient thing to cite and they present themselves as current. A database record is a statement by an organisation about what it currently accepts, read on a particular day. It is not a finding, and it is not necessarily the most recent published view.

The clearest case in this genus concerns Psilocybe serbica. Three fungi had been described from central Europe at species rank: Psilocybe arcana, Psilocybe bohemica and Psilocybe moravica. Borovička and colleagues tested them with three markers and concluded that the P. cyanescens complex in Europe consists of three well-defined species, P. cyanescens, P. azurescens and P. serbica, the last being very variable, and that the three most distinct entities previously described at species level needed a formal rank below species. They duly published the combinations: P. serbica var. arcana, var. bohemica and var. moravica, each as comb. et stat. nov.

Index Fungorum holds all three of those combinations, dated 2011. It also routes the current-name field of each one back to the species-rank name the combination was made in order to supersede. Var. arcana resolves to Psilocybe arcana, var. bohemica to Psilocybe bohemica, var. moravica to Psilocybe moravica. The database has recorded the nomenclatural act and declined the point of it, three times.

Neither source is defective. The revision is a published scientific judgement about rank; the database is a curatorial position about which name to treat as current, and curators are not obliged to follow every revision immediately. What would be defective is an account that cited whichever of the two it happened to open first and presented the result as the state of knowledge.

There is a further detail in that paper worth holding onto. The authors sequenced holotype material where they could, and sampled original localities where they could not, and then wrote:

We were not successful in obtaining DNA data from the paratype of P. serbica

The revision that reorganised the group could not obtain DNA from the type material of the species it reorganised the group around. Its circumscription is therefore open to future testing, on its own account. That sentence is the difference between a result and a settled question, and it is the kind of sentence that disappears first when a finding is summarised.

Where a depiction cannot reach

The fourth kind of claim runs from an image to a species. A considerable ethnomycological literature on this genus concerns figures in Mesoamerican codices, carvings, and rock paintings at Tassili in the Sahara. Guzmán assigns species to several of them, and states his reasoning, which is what makes the class of claim assessable. A figure in the Magliabechiano Codex is more likely P. zapotecorum, he writes, because that species is more common in Mexico. For the Tassili paintings he compares the habitat Psilocybe mairei occupies in Algeria and Morocco with reconstructed prehistoric vegetation, concludes that this supports the hypothesis that P. mairei was growing in the region and is the species represented, and notes that he had been making the suggestion in lectures since the 1990s.

Follow the direction of inference. It runs from where a species grows now, to what a person painted several thousand years ago. There is no specimen at the far end and no sequence, and the present-day distribution is doing all the work. This reference records such claims as published hypotheses, which is what their author calls them, and does not convert them into species records.

The limits are visible in the material itself. Psilocybe mairei is the one taxon of the two stirpes that Borovička and colleagues had to leave out of their molecular study, because no recent collections were available. The species attributed to paintings thought to be seven to nine thousand years old is a species that a study in 2011 could not obtain fresh material for. That is not an argument against the attribution. It is a statement of how much weight the attribution can bear.

What follows for this reference

Three habits of the accounts on this site are consequences of the above rather than house style.

Sources are named and dated at the point of use, including database records, because a database record is only a claim about a particular day. Where this reference follows one treatment over another, as it does in adopting Agrocybaceae after Niskanen and colleagues while older placements remain in circulation, the older reading is kept as source history rather than deleted.

A conflict marked resolved records an editorial choice about which treatment to follow. It does not record the end of a scientific disagreement. The serbica account says exactly that, and it is the honest description of what an editor can actually do when a revision and a database disagree.

And the gaps are stated rather than filled. Sixteen of the 25 published accounts say that the original publication behind the name has not been inspected. Ramírez-Cruz and colleagues sampled 11 of the 19 morphological sections proposed for the genus and found most of them unsupported, so this reference does not organise accounts by section. Bradshaw and colleagues put the number of accepted species at roughly 165, against which 25 accounts is a beginning.

None of that makes the record weaker. A reference that reports four kinds of claim in one voice is not more reliable than one that separates them; it has simply hidden where it could be wrong. The point of keeping them apart is that a reader can see which sentences would have to be withdrawn if a particular specimen, sequence or curatorial decision turned out differently, and can go and check.