Can species naming drive scientific attention? A perspective from plant-feeding arthropods

Abstract

How do researchers choose their study species? Some choices are based on ecological or economic importance, some on ease of study, some on tradition—but could the name of a species influence researcher decisions? We asked whether phytophagous arthropod species named after their host plants were more likely to be assayed for host-associated genetic differentiation (or ‘HAD’; the evolution of cryptic, genetically isolated host specialists within an apparently more generalist lineage). We chose 30 arthropod species (from a Google Scholar search) for which a HAD hypothesis has been tested. We traced the etymologies of species names in the 30 corresponding genera, and asked whether HAD tests were more frequent among species whose etymologies were based on host-plant names (e.g. Eurosta solidaginis, which attacks Solidago) versus those with other etymologies (e.g. Eurosta fenestrata, from Latin fenestra, ‘window’). Species with host-derived etymologies were more likely to feature in studies of HAD than those with other etymologies. We speculate that the etymology of a scientific name can draw a researcher's attention to aspects of life-history and thus influence the direction of our scientific gaze.

1. Introduction

Why do we study this species and not that one? Some species are more charismatic; some are more fundable; some are more economically important; some are more abundant; some are found closer to home. The result is ecological and taxonomic bias in scientific attention. Some such biases are well known: for instance, arthropods are dramatically understudied compared to birds and mammals. Others are likely to be more subtle. These biases shape our understanding of the natural world, but seldom receive much explicit attention (although see [1] and [2] for some discussion of how scientists should select study organisms, and how they actually do).

Could the names of species drive taxonomic bias in scientific attention? Since the mid-eighteenth century, scientists have used the Linnaean system to give each species a formal, Latinized name at the time of its scientific description. The bulk of species names follow a few etymological themes. Many species names refer to a species's morphology, behaviour, geographical occurrence or habitat, or are based on the name of a person [3–6]. However, there are few restrictions on how names can be formed or applied (even an arbitrary combination of letters can be a valid species name [7]). Naming is thus an entirely creative act. Because humans are deeply interested in names, we wondered if the etymology of a species’s name might influence the kind of scientific attention later paid to it. In particular, we asked whether a name's etymology might spur, or suggest, investigations of biological processes related to that etymology.

We explored associations between species-name etymology and scientific study in the plant-feeding (or ‘phytophagous’) arthropods (insects and mites). The phytophagous arthropods are extraordinarily diverse, many are narrowly host-specific [8] and many have species epithets based on the names of their hosts (for example, the gall making tephritid fly Eurosta solidaginis, which attacks several species of Solidago). There is strong evidence that host specialization has been important in driving insect diversification [9–11]. As a result, evolutionary ecologists have frequently asked whether an apparently oligophagous or polyphagous plant-feeding arthropod (one that feeds on a few or many species of plants, respectively) might actually represent a complex of genetically distinct host forms each specializing on a single host species (e.g. [12–14]). Such host forms are the result of ‘host-associated differentiation’, or HAD, an evolutionary process that disrupts random mating and allows independent adaptation of each resulting host form to its own host plant. Eurosta solidaginis is an excellent and well-studied example: what was once thought to be an oligophagous species attacking several goldenrod species is now understood to comprise genetically and ecologically distinct races specialized on Solidago altissima and S. gigantea [15], with further genetic structure associated with habitat in the west [16] and likely with the alternative host S. rugosa in the east [17].

While E. solidaginis is unusual in the depth and breadth of attention paid to it, the basic question—one generalist, or a complex of specialists following HAD—has been asked of many other arthropods. But with hundreds of thousands (probably millions) of phytophagous species that might be studied, which ones tend to be assayed for the occurrence of HAD? We asked whether such assays are more likely for species whose scientific names are based on the names of their host plants—as is true for E. solidaginis, for example, but not for its congener E. fenestrata. This hypothesis suggests itself because host-plant-derived names are common among phytophagous arthropods, and because of the link between that etymology and the biological mechanisms behind the HAD scenario. We speculated that host-plant based names might suggest to researchers, either consciously or subconsciously, that host specialization is an interesting part of the species's biology and therefore ought to be studied.

We tested the etymology-attention hypothesis by compiling a dataset of species-name etymologies for 30 arthropod species that have been assayed for the presence of HAD, and for their 2709 known congeners. We found that HAD-assayed species are, indeed, disproportionately likely to be species that were named for their food plants. This appears to be an example of taxonomic etymology directing, or at least influencing, the scientific gaze.

2. Methods

(a) . Data gathering

To find phytophagous arthropod species for which the hypothesis of host-associated differentiation had been tested, we performed a Google Scholar search on 19 December 2019 for the terms ‘HAD insect phytophagous herbivorous host associated differentiation’. In Scholar, this returns papers including all those terms in title or body, but the search is not case sensitive. We did not impose any date limits on the search, which returned about 25 900 results ordered by ‘relevance’ (which reflects Google's search algorithms). From this list, we chose the first 30 papers that met two simple criteria for inclusion. First, we retained only papers that tested, with genetic data, the HAD hypothesis for a phytophagous insect or mite using more than one host plant (disregarding whether the paper's results supported or refuted the hypothesis). Second, we retained only papers that indicated the species name of the HAD-tested species in the title or abstract. In addition, we removed the genus Euura from our list. Euura [18] has complex and uncertain genus-level taxonomy [19], making some of our subsequent data-gathering steps impossible. We replaced Euura with Nemorimyza since Nemorimyza posticata Meigen, 1830 (a leaf-mining agromyzid fly) was recently tested for HAD [14] but was not flagged in our search. Although ‘insect’ was part of our search string, we did not exclude papers that executed HAD tests on other phytophagous arthropods (and one such species, the mite Tetranychus urticae, was included in our final list). The resulting 30 HAD-tested species—our focal species—are listed in table 1.

Table 1.

Phytophagous arthropod species, tested for the occurrence of host-associated genetic differentiation, identified by our search. The genus Nemorimyza replaced Euura, whose current phylogenetic resolution is too poor for our study.

species	year described	common name	order and family	citation for HAD test
Acrobasis vaccinii Riley	1884	cranberry fruitworm	Lepidoptera: Pyralidae	Medina et al. [20]
Acyrthosiphon pisum Harris	1776	pea aphid	Hemiptera: Aphididae	Ferrari & Godfray. [21]
Anastrepha fraterculus Wiedemann	1830	South American fruit fly	Diptera: Tephritidae	Malavasi & Morgante [22]
Aphis gossypii Glover	1877	cotton aphid	Hemitera: Aphididae	Vanlerberghe-Masutti et al. [23]
Archips argyrospila Walker	1863	fruit-tree leafroller	Lepidoptera: Tortricidae	Goyer et al. [24]
Bactericera cockerelli Šulc,	1909	potato psyllid	Hemiptera: Psyllidae	Liu et al. [25]
Caryedon serratus Olivier	1790	groundnut borer	Coleoptera: Chrysomelidae	Sembène et al. [26]
Chilo suppressalis Walker	1863	Asiatic rice borer	Lepidoptera: Crambidae	Zhong et al. [27]
Chrysomela aeneicollis Schaeffer	1928	bronze-necked leaf beetle	Coleoptera: Chrysomelidae	Rank [28]
Cydia pomonella Linnaeus	1758	coddling moth	Lepidoptera: Tortricidae	Chen & Dorn [29]
Dendroctonus ponderosae Hopkins	1902	mountain pine beetle	Coleoptera: Curculionidae	Stock & Amman [30]
Diplolepis rosae Linnaeus	1758	mossy rose-gall wasp	Hymenoptera: Cynipidae	Kohnen et al. [31])
Eurosta solidaginis Fitch	1955	goldenrod ball-gall fly	Diptera: Tephritidae	Waring et al. [32]
Gnorimoschema gallaesolidaginis Riley	1869	goldenrod spindle-gall moth	Lepidoptera: Gelechiidae	Nason et al. [33]
Heliothis virescens Fabricius	1777	tobacco budworm	Lepidoptera: Noctuidae	Groot et al. [34]
Hesperotettix viridis Thomas	1872	snakeweed grasshopper	Orthoptera: Acrididae	Sword et al. [35]
Hyalesthes obsoletus Signoret	1865	none	Hemiptera: Cixiidae	Kosovac et al. [36]
Maconellicoccus hirsutus Green	1908	pink hibiscus mealybug	Hemiptera: Pseudococcidae	Rosas-Garcia et al. [37]
Maruca vitrata Fabricius	1787	bean pod borer	Lepidoptera: Crambidae	Agunbiade et al. [38]
Nemorimyza posticata Meigen	1830	none	Diptera: Agromyzidae	Mlynarek & Heard [14]
Neoleucinodes elegantalis Guenée	1854	tomato fruit borer	Lepidoptera: Crambidae	Diaz-Montilla et al. [39]
Oxya japonica Thunberg	1815	Japanese grasshopper	Orthoptera: Acrididae	Li et al. [40]
Phylloxera notabilis Pergande	1904	pecan-leaf phylloxera	Hemiptera: Phylloxeridae	Dickey & Medina [41]
Rhagoletis pomonella Walsh	1867	apple-maggot fly	Diptera: Tephritidae	Feder et al. [42]
Rhinusa antirrhini Schoenherr	1825	toadflax seedhead weevil	Coleoptera: Curculionidae	Hernández-Vera et al. [43]
Spodoptera frugiperda Smith	1797	fall armyworm	Lepidoptera: Noctuidae	Pashley [44]
Tetranychus urticae Koch	1836	two-spotted spider mite	Trombidiformes: Tetranychidae	Gotoh et al. [45]
Thaumetopoea pityocampa Denis & Schiffermüller	1775	pine processionary	Lepidoptera: Notodontidae	Salvato et al. [46]
Thrips tabaci Lindeman	1889	tobacco thrips	Thysanoptera: Thripidae	Brunner et al. [47]
Urophora cardui Linnaeus	1758	Canada thistle gall fly	Diptera: Tephritidae	Eber & Brandl [48]

For our 30 focal species, there are two possibilities. The species could have been described and named after, and perhaps even as a result of, a study documenting HAD; or the species might have been described and named before any such study was done. Only in the latter case could the etymology of the species's name have influenced the researchers’ decision to study it, so we recorded both the year of taxonomic description and the year of HAD testing. This also allowed us to test for changes through time in naming practices.

We took our list of 30 focal HAD-tested species and compiled lists of all currently recognized species in each of their genera, using online resources and also by directly asking taxonomists with relevant expertise. There were 2739 species distributed across the 30 genera (electronic supplementary material, table S1 at doi:10.5061/dryad.fj6q573zg).

We then determined the etymology of the specific epithets. Where possible, we based our determination on the original species descriptions. When these descriptions did not include etymologies, or when they could not be located, we inferred the etymology from the linguistic formation (Latin or other root or suffix) of the name. For example, we inferred that a name ending in -ensis refers to a place of origin or distribution, whether or not this was explicitly indicated in the species description. We classified etymologies into seven categories: host, behaviour, habitat, morphology, place, person (eponymy) or other. The ‘other’ category was assigned either when we could not determine a specific epithet's etymology, or when the etymology didn't fit any of the other six categories. We included in the ‘other’ category two names based on host common names or cultural references: Aphis sumire, where ‘sumire’ is a girl's name meaning violet, and violet is the host; and Phylloxera kunugi, where ‘kunugi’ is a Japanese common name for the oak species Quercus acutissima. We reasoned that although these etymologies are ultimately based on hosts, that connection would be inapparent to most of the global research community.

(b) . Statistical analysis

We began by asking whether the distribution of etymologies differed among genera. This involves a G-test of independence applied to a 7 × 30 contingency table (7 etymological categories, 30 genera). We then used logistic regression to test the hypothesis that the proportion of species named for their host plant changed with year of description. In this analysis, the response variable was a binary value indicating whether or not the arthropod species was named after its host.

To test our hypothesis that species-name etymology influenced likelihood of study, we asked whether those species that have been assayed for HAD were disproportionately named after their host plants, compared with all members of their genera. This is achieved using a G-test of independence to compare two proportions (in a 2 × 2 contingency table): first, the proportion of the 30 HAD-tested species with host-plant-based names, and second, the proportion of the 2709 congeners with host-plant-based names. While a few of these congeners may themselves have been subjected to HAD testing, we did not attempt to separate such species out. This makes our G-test somewhat conservative. We repeated this G-test on two subsets of our data: first, we omitted Aphis (the largest genus, and something of an outlier in naming practice with by far the most host-plant-derived names); and second, we omitted Tetranychus (the only non-insect arthropod genus in our dataset).

All analyses were performed in R using the packages DescTools [49] and vegan 2.5-7 [50]. Our R scripts are available as Supplementary Material.

3. Results

In our dataset, for all species with published HAD studies, the naming preceded the HAD test (table 1, compare 2nd and last columns).

There was substantial variation among genera in the breakdown of species-name etymologies (G = 1052, d.f. = 174, p < 2.2 × 10⁻¹⁶; electronic supplementary material, table S2 at doi:10.5061/dryad.fj6q573zg). Morphology (895 species), host (646 species), person (423 species) and place (300 species) were the most popular origins of names given to species. Names referring to host plants are most common (57%) in Aphis, which is the only genus to exceed the proportion of host-plant names for our HAD species; figure 1). Three genera (Hesperotettix, Hyalesthes and Maruca) have no species named after host plants, and for Eurosta, only the focal E. solidaginis has such a name. For the remaining genera, between 4 and 37% of species were named after plants. Species were significantly more likely to be named after their hosts if described earlier (z = 3.93, p < 0.001. However, the effect size was only moderate: between 1760 and 2020, the modelled fraction of insects named after host plants declined from about 36% to 21% (figure 2).

Figure 1.

Proportion of species named for host plants across the 30 focal genera, compared with the proportion for the 30 focal HAD-tested species (black bar at right). The plotted proportions for each genus exclude the focal species.

Figure 2.

Arthropod species that were named for their plant host, or otherwise, organized by their year of description. Blue line represents the probability curve generated from the logistic regression.

Of our 30 focal HAD-tested species, 14 (47%) were named after a host plant. Just less than half this proportion of the 2709 congeners (632 species, or 23%) had a host-plant etymology, and the difference was highly significant (G = 7.67, d.f. = 1, p = 0.0056). This pattern remains robust if we omit from the analysis either Aphis (G = 15.8, d.f. = 1, p = 0.00007) or Tetranychus (G = 6.49, d.f. = 1, p = 0.01).

4. Discussion

There are far more species on Earth than can possibly be studied; and even among species that do attract the scientific gaze, some are studied far more than others. Using data for phytophagous arthropods, we asked whether the etymology of a species name—in particular, whether it is named after a host plant—might influence the kind of scientific attention that species receives.

We found striking and significant variation among genera in the use of different etymological categories in constructing species names. This included strong variation in the proportion of names based on plant hosts (from just one species in Eurosta to over half of the speciose genus Aphis). Poulin et al. [6] documented a similar pattern (but at a coarser phylogenetic scale) among parasitic worms, with strong variation among classes and phyla in naming etymologies. This kind of variation is not unexpected [6], given that different taxonomists are involved with naming in different clades. We also found a tendency, significant but weak, for names based on host plants to have become less common over time. Mammola et al. [5] found a similar but stronger effect for ecologically informative names among spiders, although Poulin et al. [6] found no such trend for parasitic worms. These patterns underscore the fundamentally creative nature of the naming act. The assignment of a specific epithet to a newly described species is ultimately an arbitrary and creative decision [4] that might reflect the biology of the named species, but can also (or instead) reflect the culture and personalities of the taxonomists assigning the name. And yet naming is potentially important, because humans are extraordinarily interested in names and may perceive information in names even if that perception isn't intended by the namer.

The etymological basis of a species name was significantly associated with the likelihood of that species being studied as a potential example of host-associated differentiation (HAD). In particular, species with names based on those of their host plants were strongly overrepresented—over twice as frequent—in HAD studies. While our analysis cannot break this pattern down to the level of individual genera, we were able to rule out the possibility of its being driven by two potential outliers: Aphis, a diverse genus for which naming practices have been somewhat unusual (with far more host-species-derived names than any other genus), and Tetranychus, the only non-insect arthropod in our dataset. Our results suggest that decisions made by the taxonomists describing species, often decades or centuries earlier, may be shaping the attention evolutionary ecologists pay to taxa now. Because the frequency of host-plant names varies across insect taxa, this effect could be driving a taxonomic bias in HAD studies—and as a result, in the detection of HAD—across arthropod lineages.

Why might HAD researchers disproportionately study host-plant-named arthropods? Experimental answers to this question would be challenging at best, and without them we must be cautious about inferring causality. Nevertheless, we offer a speculation: perhaps researchers notice a host-plant derived name and are subconsciously inspired to ask questions about host specialization. One of us (SBH) first came to the study of HAD by noticing moth galls on stems of Solidago altissma and Solidago gigantea, and wondering if they represented a pair of host specialists. It's certainly possible that learning the gallmaker's name—Gnorimoschema gallaesolidaginis—played some role in feeding this wonder. Humans have well-documented cognitive biases that make such a scenario plausible—for example, the frequency illusion (or ‘Baader–Meinhof phenomenon’ [51]), under which noticing something once primes one's mind to notice it repeatedly. It would be surprising, we suspect, if humans were able to avoid this kind of ‘inspiration by connection’.

Of course, species-name etymology isn't the only thing that might influence the likelihood of a scientist studying a particular species (to test a HAD hypothesis or for any other reason). For example, a species may be more likely to be studied if it is large, colourful, charismatic, economically important, abundant, widespread or found close to where scientists live and work [1]. A species may also be chosen for study because it has lesser permitting, handling or training requirements, more previous knowledge or proven techniques (for instance, a published genome or SNP library), or—in contrast to the last factor—because its study would be novel [2]. Unsurprisingly, we were unable to assess all these possible influences on the likelihood of species study.

In multicausal systems (that is to say, in all systems), there are two ways in which an unstudied factor might affect inference about the impact of a studied one on some outcome (in our case, how something else might affect our inference about the effect of species-name etymology on the likelihood of HAD testing). First, the two factors might have independent effects. In this case, the unstudied factor will create noise, potentially obscuring a real pattern. Second, the unstudied factor might be associated with the studied one—and in this case the two are confounded and the unstudied factor might cause an apparent pattern we might ascribe instead to the studied one. Which scenario might we expect for the species attributes reviewed in the last paragraph? While a full answer to that question is a matter for a decadal-scale research program, for most attributes we are aware of no data or models predicting confounding that could drive our host-name/attention pattern. We would not expect, for example, species named for their host plants to be larger or more attractive than their congeners (if anything, we would hypothesize that large or conspicuously patterned species might be more likely to be named with reference to morphology, and if so our detection of a host-name/attention pattern is conservative). Species named for their host plants are unlikely to be easier to find or collect, or to require fewer permits to study. If they are actually harder to find and collect (because of their specialist habits) this again makes our results conservative. It is possible that species with broader geographical ranges might be noticed and described first, and thus be more likely to be named after a host (figure 2). However, the temporal trend in naming was far too modest to explain our result even if these broadly distributed species are also more likely to be chosen for study of HAD.

One scenario that deserves testing is based on the connection between pest status and ecological study. Species that are pests of agricultural crops and timber trees receive an outsize share of scientific attention, in part because their economic impact makes the work more fundable. This pattern certainly extends to the study of HAD: about two-thirds of our focal species (table 1) are agricultural or forestry pests. If pest species are more likely than their non-pest relatives to be named after the host plant they impact, then our HAD-naming pattern could arise for economic (fundability) reasons. A test of this hypothesis should be a priority for future work. For now, we would simply note that among our 30 focal genera, by far the highest proportion of host-plant names occurred in Aphis. As a genus, Aphis includes many agricultural pests; however, removing Aphis from our dataset strengthened the HAD-naming pattern rather than weakening it. This observation does not support a fundability explanation for our result.

One might wonder whether species names might bias not just the occurrence of HAD studies, but also their outcome. This could be true if insects named for their host plants are more likely to be strict specialists. Under this scenario, the same species researchers are most likely to assay for HAD are the ones most likely to reveal that HAD has occurred. The result would be exaggerated estimates of the frequency of HAD among plant-feeding insects. This would be harmless (perhaps even efficient) if our only goal were only to study how HAD occurs when it does, but it's a disquieting possibility if we are also interested in how often HAD occurs. We are not, unfortunately, in a position to shed much more light on this. It seems plausible that oligophagous species are more likely to be named for their hosts than broad generalists, but it does not necessarily follow that these oligophages are more likely to in fact represent complexes of cryptic host specialists. Ideally, we would compare the frequency with which assays for HAD actually reveal its occurrence between taxa named for plant hosts and those named in other ways. However, given the file drawer effect (studies that don't find HAD may be less likely to be published) and given the relatively small number of insects for which powerful tests of HAD are available, this will have to remain a goal for the (distant) future.

Might there be other ways in which naming influences later scientific attention to species? We are unaware of data bearing on this, but we suspect so. At the most obvious level, taxa without formal names are unlikely to be studied. This might well be true even for taxa that have been identified as distinct and thus nameable lineages but have not yet been formally described and named (e.g. the divergent habitat- and host-associated clades within Nemorimyza posticata identified by [14]). Thinking further along these lines, we wonder whether species epithets that are very long, difficult to spell or difficult to pronounce might reduce the scientific attention paid to their bearers. This hypothesis should be testable with current data, or we can wait to assess the future literature corpus devoted to the recently described myxobacterium Myxococcus llanfairpwllgwyngyllgogerychwyrndrobwllllantysiliogogogochensis [52].

Our study provides yet more evidence (if it was needed) that the common stereotype of science as objective and fully rational is ill-founded. Scientists study a non-random subset of the world, and ask a non-random subset of questions. We have provided evidence that these two dimensions of non-randomness may be interrelated. Scientists, like novelists and songwriters, are often asked where they get their ideas. It would appear that sometimes the answer may lie in names.

Data accessibility

Data are available via Dryad, https://doi.org/10.5061/dryad.fj6q573zg [53].

The data are provided in electronic supplementary material [54].

Authors' contributions

J.J.M.: conceptualization, data curation, formal analysis, methodology, writing—original draft, writing—review and editing; C.C.: data curation, investigation, writing—review and editing; A.L.P.: formal analysis, software, visualization, writing—review and editing; J.L.V.: formal analysis, software, visualization, writing—review and editing; S.B.H.: conceptualization, data curation, formal analysis, funding acquisition, investigation, methodology, project administration, supervision, writing—original draft, writing—review and editing.

All authors gave final approval for publication and agreed to be held accountable for the work performed therein.

Conflict of interest declaration

We declare we have no competing interests.

Funding

This work was funded in part by grant support from the Natural Sciences and Engineering Research Council (Canada; Discovery Grant RGPIN-2015-04418 to S.B.H.).