The problem lies in a technique called fuzzy matching. Scientific names are notoriously variable: alternative spellings proliferate, typographical errors creep in during digitisation, and legitimate variants accumulate over centuries of taxonomic work. To cope, automated tools estimate the ‘edit distance’ between two names — the minimum number of single-character insertions, deletions, substitutions or transpositions needed to turn one string into the other. If two names differ by only a letter or two, the algorithm assumes they refer to the same organism, and the records are merged. The approach implicitly assumes that genuinely distinct species names are far enough apart orthographically that a small edit distance can only signal a misspelling. A team writing in the Biodiversity Data Journal decided to test that assumption rigorously, and the results are sobering.

Working with the World Register of Marine Species, the researchers computed Damerau-Levenshtein distances across approximately 230,000 unique, accepted binomial names spanning 85 phyla. For every name, they identified its closest orthographic neighbour in the register and measured how many edits separated them. The findings overturn the comforting assumption that valid species names are safely distinct. Some 1,026 names — 0.45 percent of the total — sit within a single edit of another accepted species name. Another 8,063 names, or 3.51 percent, are within two edits. Most strikingly, more than ten percent of all marine species names, 28,564 in absolute terms, differ from another valid name by three edits or fewer. The peak of name similarity occurs at distances of four to five edits, after which the number of near-identical pairs begins to decline.

Why does this matter? Because when two distinct species have genuinely similar names, a fuzzy-matching algorithm can mistake the difference for a typographical error and wrongly merge records from two different organisms — precisely the kind of confusion long associated with homonymy, where identical names are applied to different species. Homonyms are a recognised hazard in biodiversity informatics and are typically mitigated by consulting higher taxonomic ranks or the authority who originally described the name. Near-homonyms created by orthographic coincidence have received far less attention, yet they generate the same failure mode: biological data from separate species combined under a single name, silently degrading downstream analyses of distribution, abundance and extinction risk.

The researchers also examined how name similarity behaves across the two components of a binomial. In a striking pattern, 99.20 percent of the closest matches had all their edits concentrated in a single component of the name. In 81.17 percent of cases, the two names shared the same genus and differed only in the species epithet; in the remaining 18.03 percent, they shared the epithet and differed in the genus. The pattern held after filtering, with edits confined to one component in 98.57 percent of cases. This means the most dangerous collisions occur not between distantly related organisms but between close relatives — sister species within the same genus whose epithets happen to be near-identical strings.

The reason, it turns out, is that taxonomists have reused a remarkably small pool of descriptive words. Of 85,863 unique epithets in the register, 24,655 are shared by multiple species, dragging 168,300 species into shared-epithet relationships — a figure broadly comparable to the number of species sharing a genus. Descriptors such as ‘gracilis’, meaning slender, ‘australis’, meaning southern, ‘elegans’, ‘japonica’ and ‘pacifica’ have each been applied to more than 400 different species across phyla as distant as arthropods, molluscs and chordates. Geographical and morphological adjectives make intuitive names, but when the same adjective is recycled across hundreds of organisms, and closely related species receive near-identical versions of it, the resulting name space becomes densely crowded with lookalikes.

Quantifying this crowding, the study found that within shared genera, 1.45 percent of species have an epithet within one edit of a congener’s, 7.58 percent within two edits, and 19.34 percent within three. The comparable figures for genera among species sharing an epithet were lower — 0.27, 1.76 and 6.05 percent respectively — confirming that the second component of the binomial is the primary site of dangerous similarity. Whether this reflects genuine evolutionary pattern, with sister species sharing similar traits that inspire similar names, or simply chance — genera contain more species on average, raising the odds of a small edit distance — the practical implication is the same: permitting fuzzy matches up to three edits within a single name component dramatically raises the probability of incorrectly fusing two accepted species.

Can existing safeguards rescue the situation? The field’s standard defence is a set of non-phonetic post-matching filters, embodied in the Taxamatch algorithm and deployed by major platforms including the Taxonomic Name Resolution Service and WoRMS itself. These filters exploit the internal structure of names, checking that shared prefixes and suffixes are preserved and that errors fall in plausible positions. The new analysis shows they work — up to a point. Filtering had no effect at a single edit, where all 1,026 candidate matches were retained, but progressively pruned false positives at larger distances: the pool of two-edit matches shrank by roughly half to 4,070, three-edit matches fell to 9,060, a 68 percent reduction, and four-edit matches collapsed to 2,622, a 95 percent reduction. Encouraging, yet the residual false positives at two and three edits remain substantial, and the study’s authors conclude that orthographic filtering alone cannot eliminate the risk.

The remedy, they argue, lies in enriching automated name resolution with two additional streams of information. The first is systematic context: restricting searches to names within a known taxonomic group, or checking that a proposed match respects known relationships, can prevent a fish record from being attached to a similarly named mollusc. The second is authorship. In botanical and zoological nomenclature, species names are frequently accompanied by the name of the describing authority and the year of description, and these strings can disambiguate two closely related species whose binomials are nearly identical. Because the study shows that most perilous matches occur within genera, between close relatives, authorship information becomes not a bibliographic nicety but an essential computational safeguard.

The stakes extend far beyond marine registers. Aggregated biodiversity databases underpin global assessments of species extinction risk, invasive-species tracking, biogeographic modelling and the monitoring of ecosystem change, and all of them inherit whatever errors originate at the name-matching stage. A misassigned record does not merely sit quietly in a spreadsheet; it can propagate through models, skew range maps and distort conservation priorities. As biodiversity science enters an era of machine learning and AI-assisted data aggregation, the new results are a timely warning that string similarity is not species similarity. The humble binomial, devised by Linnaeus as a universal label for life, turns out to be a far more crowded and collision-prone space than anyone had measured — and the algorithms that navigate it need more than spelling to find their way.

Subject of Research: Orthographic similarity among marine species names and its impact on taxonomic standardisation in biodiversity data

Article Title: How similar are species names and why does this matter for biodiversity data

Article References: Menegotto, A., Ronquillo, C., Hortal, J., & Webb, T. (2026). How similar are species names and why does this matter for biodiversity data. Biodiversity Data Journal, 14, Article e196932. https://doi.org/10.3897/bdj.14.e196932

Image Credits: AI Generated

DOI: 10.3897/BDJ.14.e196932

Keywords: biodiversity informatics, taxonomy, fuzzy matching, species names, WoRMS, Damerau-Levenshtein distance, homonymy, taxonomic standardisation, marine species, data aggregation, Taxamatch, scientific nomenclature

Chironomus melanotus is no obscure curiosity. First described from Germany in 1961, it is one of the most widespread and well-studied members of a genus of non-biting midges whose aquatic larvae play crucial roles in freshwater ecosystems. The species is in demand in classical hydrobiology and in toxicology, where larval responses to pollution serve as indicators of environmental health. But identifying these larvae is notoriously difficult. Their external morphology offers few reliable distinguishing features, which is why researchers have long turned to an unusual tool: the giant chromosomes packed inside the salivary glands of the larvae.

These polytene chromosomes, so large they can be examined under a light microscope, carry distinctive banding patterns that act like barcodes of a different kind. Each species of Chironomus has a characteristic set of banding sequences in its chromosome arms, and deviations from the norm can even signal environmental stress. In the new study, the researchers collected fourth-instar larvae from three sites in the Yaroslavl region: thirty-one larvae from a puddle on the Shumarovka river shore, seven from the Sunoga river, and four from a small stream in the shore zone of the Kotorosl’ river. Using the ethanol-orcein staining technique, they prepared chromosome spreads and compared the banding patterns against established cytomaps.

The karyotype analysis confirmed what earlier studies had suggested: Ch. melanotus is a chromosomally conservative species. The larvae carried the expected chromosome set of 2n = 8, with the arm combination AB, CD, EF and G that places the species in the so-called “thummi” cytocomplex. A distinctive feature is a heterochromatinized centromeric region forming an unstable chromocenter, a trait shared with only a handful of other Chironomus species. Across all the sampled larvae, the team found just eight banding sequences, and only one of them, melD2, deviated from the standard set, appearing in a heterozygous state in two larvae from the Sunoga river. No trisomy, rearrangements, insertions or deletions were detected, the kinds of chromosomal damage more typical of larvae living in polluted water bodies.

With the species identity of their Russian material securely established by both morphology and cytogenetics, the researchers then extracted DNA from one larva and amplified a 658-base-pair segment of the mitochondrial cytochrome oxidase subunit I gene, the standard marker for animal DNA barcoding. The sequence was deposited in GenBank under accession number OL546775. The real drama began when the team compared this verified sequence with the six COI sequences already labeled as Ch. melanotus in GenBank and the Barcode of Life Data Systems, or BOLD, five from Finland and one from Sweden.

The results were startling. The average genetic distance among sequences carrying the Ch. melanotus label was 9.1 percent, far outside the roughly 3 percent threshold conventionally used to distinguish chironomid species. Only one Finnish sequence, MZ659620, proved to be a genuine match, sitting just 0.4 percent away from the Russian specimen. The remaining sequences diverged from the verified Russian barcode by 9.5 to 12.4 percent, distances comparable to those separating entirely different species. In other words, the databases were not storing minor variants of one species; they were storing several species masquerading under one name.

Bayesian phylogenetic analysis painted the same picture. On the tree inferred from the COI sequences, the Russian specimen and the single matching Finnish sequence clustered together, while the other supposedly conspecific sequences scattered onto separate branches. Two Finnish sequences grouped with Ch. annularius, a species confirmed karyologically in earlier work, at a genetic distance of 2.9 percent, just under the 3 percent threshold. The Swedish sequence clustered with Ch. cingulatus at 2.1 percent. Most intriguing of all, two Finnish sequences formed their own isolated cluster, matching nothing in either database. The researchers suggest these may belong to a known species simply absent from the reference libraries, or to a distinct, undescribed species awaiting formal recognition.

The study’s authors place their findings in a broader context of crisis in molecular taxonomy. Genetic databases are riddled with so-called “dark taxa,” sequences from unidentified or incorrectly identified specimens. By one estimate cited in the paper, roughly 65 percent of Chironomidae sequences from Germany lack species-level assignments. The authors argue that a related problem, the “superficial taxonomic impediment,” applies even more directly here: species are named so poorly and unreliably that they will need to be redescribed before they can be used with confidence. The Finnish sequences in question came from the Finnish Barcode of Life project, whose probabilistic taxonomic assignment tool achieves about 88.5 percent accuracy at the species level, a figure the authors consider insufficient in a group rich in sibling species.

The complications do not end with mislabeling. The standard barcoding threshold itself is unreliable for Chironomus. Interspecific COI divergence in the genus typically ranges from 9 to 20 percent, but in rare cases among karyologically verified species it falls between 1 and 4 percent, overlapping with intraspecific variation. A cautionary tale comes from Japan, where the “highland” and “lowland” populations of Ch. nipponensis, initially separated by morphology and later shown to differ by 9.1 percent in COI sequence, were eventually confirmed as two distinct species. Numbers alone, the study suggests, cannot settle questions of species identity.

The conclusion the researchers draw is unambiguous: identifying Chironomus species demands a comprehensive approach combining morphology, cytogenetics and molecular genetics, drawing on both mitochondrial and nuclear genes. DNA-based analyses remain extremely useful as a first step, narrowing the search before precise identification, but it is impossible at the present stage to trust a computer algorithm alone. The stakes extend beyond taxonomy. Chironomus larvae are a vital part of freshwater ecosystems, and reliable species identification enriches our understanding of how those ecosystems function. As DNA metabarcoding becomes the backbone of biomonitoring programs worldwide, the case of Ch. melanotus stands as a warning that the reference libraries underpinning this revolution must themselves be built on verified, karyotype-checked foundations.

Subject of Research: Species identification of the midge Chironomus melanotus using karyotype analysis and COI DNA barcoding

Article Title: Karyotype and COI gene sequences of Chironomus melanotus Keyl, 1961 from the Yaroslavl region, Russia, and the difficulties with its identification using GenBank and BOLD systems

Article References: Bolshakov, V. V., & Movergoz, E. A. (2022). Karyotype and COI gene sequences of Chironomus melanotus Keyl, 1961 from the Yaroslavl region, Russia, and the difficulties with its identification using GenBank and BOLD systems. Comparative Cytogenetics, 16(3), 161-172. https://doi.org/10.3897/compcytogen.v16.i3.90336

Image Credits: AI Generated

DOI: 10.3897/compcytogen.v16.i3.90336

Keywords: Chironomus melanotus, DNA barcoding, karyotype, polytene chromosomes, COI gene, GenBank, BOLD, dark taxa, chironomids, species identification, Yaroslavl region, biomonitoring