The Roman classification and nomenclature of aquatic animals: an annotated checklist (with a focus on ethnobiology)

ABSTRACT This article provides a checklist of the aquatic animals the Romans appear to have recognized, including an overview of how they classified and named them. Indeed, in addition to providing scientific identifications of the animal referents and philological and/or linguistic data related to their names as they emerge in the ancient texts, the checklist focuses on ethnobiology. In particular, it provides ethnotaxonomic and ethnobiolinguistic data such as ethnotaxonomic rank and ethnobiological name typology (Conklin 1968; Berlin 1992) by which to reconstruct the actual folk taxa, i.e. the kinds or “groups of kinds” of animals each name covered in the ancient speakers' minds (but a few unnamed folk taxa have also emerged). In this respect, tree diagrams are used to represent at a glance the folk taxonomic knowledge of an ideal Roman as to the main higher-order groupings of aquatic animals they were familiar with. The results are analyzed and statistical data are provided as to various relevant ethnobiological variables. The findings accord with Berlin's universals, except for the number of additional levels where folk taxa of the life-form rank can be found, a characteristic already observed for aquatic animals in other folk taxonomies (Pawley 2006). From a more general zooanthropological perspective, these data suggest the Romans' familiarity with the aquatic world and its inhabitants, despite their traditional self-ascribed identity as peasants and soldiers.


INTRODUCTION
This article aims to provide a checklist of the aquatic animals the Romans appear to have recognized (cf., as to snakes in ancient Greek, Bodson 2012), together with an overview of how they classified and named them. Particular emphasis will be placed on ethnotaxonomy and the related ethnobiological nomenclature. This is about how members of a cultural population categorize animals or plants according to perceived morphological, behavioural (Atran 1990) or ecological (Hunn 1999) similarities, which serve to group the recognised organisms into usually named sets (ethnotaxa or folk taxa) at several different levels of hierarchical inclusion; these levels are a maximum of six in number (Berlin 1992; see Materials and methods), each more or less corresponding to equally well recognized cross-cultural patterns of nomenclature (Berlin 1992: 34;Hunn & Brown 2011: 326). Indeed, cross-cultural ethnobiological data show how these relations of similarity tend to emerge in the morphosemantic structure of the names for living things (hereafter "ethnobionyms"). As already noted elsewhere (Guasparri 2019), the Roman nomenclature of aquatic animals provides a good example of this nomenclatural descriptiveness, which is of particular value when it comes to reconstructing ethnobiological knowledge for members of ancient cultures.
As we shall see, among the distinctive features of the biological referents under examination it is not only perceptual factors (i.e. based on the external similarities mentioned above) but also culture-dependent factors (based on using animals as food, medicine, symbols, etc.) that emerge at the linguistic level (cf. a fish called gladius, literally "sword", due to a physical characteristic, vs a fish named hallecula after hallec, a sauce made from it). This is not surprising since "the intrusion of utilitarian factors into otherwise general purpose categories" (Ellen 2006: 12) has long been acknowledged by ethnobiologists (e.g., Bulmer 1967;Hunn 1980: 13;1982;Hays 1982).
In order to analyse how a historical people such as the Romans classified and named aquatic animals (or any other set of living things), one cannot rely on participant observation or interviews of any kind (e.g., sorting experiments or free listings). Instead, one must make do with the explicit or implicit information collected from written records, at best confirmed by zooarchaeological evidence. Going into the "field of texts" is indeed, mostly, an "armchair" activity, not a "hiking boots" one. However, not only do both kinds of researchers have to tackle the limits inherent in reducing a local people's perception of their environment to the written word (e.g., Ellen 2006: S6); both also end up, as a "natural" consequence of ethnobiological research, with a list of plant or animal names whose "descriptive force" alone can provide a deep insight into "what is seen most clearly by native eyes" (Hunn 2006). If such an outcome cannot fill the unquestionable gap between face-to-face and "face-to-text" research, at least it can narrow it. Moreover, as to the ethnolinguistic population being examined, the considerable amount and quality of the ancient texts available (French 1994: 179) -with all the possible philological problems commonly related to their manuscript tradition -can further serve to reconstruct how the members of even a "dead" culture perceived, and ultimately experienced, their unique ethnobiological reality (although only limited bits of it can be disclosed).
Much like Greek aquatic animal names, the Roman ones have traditionally been a training ground for modern aquatic zoology, as represented by its "fathers", the greatest ichthyologists of the sixteenth century: Belon, Rondelet and Salviani (e.g., Gudger 1934). They had to master both the Greek and the Latin sources if they wanted to match the ancient name and description of a marine animal to the folk names and the living species known to them (e.g., Pinon 2005: 263). Nonetheless, by the early seventeenth century onwards, with the gradual extension of the biogeographical areas naturalists took into account, ancient au-The Roman classification and nomenclature of aquatic animals thors such as Aristotle or Pliny became more and more obsolete (Mayr 1982: 158;Ogilvie 2006: 138, 181, 207f.) -their study remained a matter of historical interest, as shown by the detailed commentary Cuvier devoted to Pliny's Natural History "zoological" books (8-11) (Ajasson de Grandsagne & Cuvier 1827, 1828. Subsequently, with possibly the notable exception of D'Arcy Thompson (1947), it was basically "pure" philologists who devoted themselves to translating the many aquatic animal names found in classical authors -either in the form of inquiries into single etymological issues or of all-encompassing works such as a specific dictionary (de Saint-Denis 1947;Fruyt & Lasagna 2015).
This is also what I have been doing myself for quite some time, by surveying the Roman literary sources from Plautus (third century BCE) to Ausonius (fourth century CE) and the Greek ones whenever, as has often been the case, the Latin name was a Greek loanword.
At first, the idea was to tackle the whole corpus of Latin aquatic animal names in order to refine and possibly correct the "traditional" biological identifications found within the existing literature. Afterwards, since a lot of etymological work was involved and several nomenclatural patterns began to emerge, the project took first an ethnolinguistic, then an ethnobiological turn. Indeed, in order to provide the most reliable identifications according to the latest checklists of Mediterranean marine species, it was essential to inspect the ethnobionyms' linguistic form. At the end of the analysis, the number of descriptive names, i.e. those whose morphosemantics captures a phenotypical, ecoethological or generally cultural feature perceptually related to their biological referent, proved very high (Guasparri 2019). As expected, this cross-cultural property of descriptive names, which we may call the "referential constraint" (Guasparri 2007: 74) and which is realized in Latin aquatic animal names mostly at the phenotypical level (Guasparri 2013: 350; Fig. 1) proved crucial for disclosing the biological identifications of many ethnobionyms, notably whenever any additional information from the sources was scarce or unavailable.
Among the limits of this kind of research, one may include the impossibility of focusing on a particular place and time: the ethnobionyms under examination are not only dependent on the unpredictable -often centuries distant -chronology of the works and the authors available, but the latter may in turn deal, with second-hand data collected from different places in different times (this is famously the case for Pliny, on whose accounts most identifications rely -another inevitable limit of the research). However, the common point about the sources is their constant reference to fish, mammals, molluscs and other marine organisms of the Mediterranean Sea. This "sea in the middle of the land", however noted for its rich biodiversity as a marine ecosystem (Cuttelod et al. 2009;Coll et al. 2010), is rather homogeneous as to the classes of human-salient organisms more likely to occur within the experiential range of a given Mediterranean community through time -suffice it to say that, with the notable exception of purple-producing molluscs, the most exploited species in the Roman period have virtually remained the same until recently, if not to this day, despite the many anthropogenic events which have shaped the area (Coll et al. 2010;Marzano 2013;Abulafia 2014). Moreover, the peoples of the Mediterranean Basin show a vital network of contacts and hence of shared cultural traits dating back to at least 3000 BCE (Abulafia 2014), i.e. well before the Greek and Roman periods (Coll et al. 2010;Lopes 2013;Abulafia 2014).

MATERIALS AND METHODS
The first step in research of this kind, based mainly on collecting ethnobiological data by going into the "field of texts", is to build a body of ethnobionyms. This has been done by recording every name of aquatic animal found not only in the Latin literature but also in the Greek one, in case of loanwords -due to ever-increasing Greek acculturation from at least the third century BCE (e.g., Lomas 1993), one can assume that Roman speakers used Greek and Latin ethnobionyms interchangeably, as Varro (Ling. 5, 77) attests (Guasparri 2013; Bodson 2014).
As a starting point, I have simply relied on pre-existing lists of aquatic animal names taken down from the main modern glossaries on the subject (de Saint-Denis 1947;Thompson 1947), only to dismiss some entries, while adding several others, in the course of the research. Subsequently, I have recorded and reviewed all the occurrences found by searching the Thesaurus Linguae Graecae (TLG) and the Packard Humanities Institute (PHI) databases (the two main digital libraries of Greek and Latin literature respectively) through the software Diogenes (ver. 3.2.0, for Windows). The retrieved passages and the contexts in which each term occurs have been thoroughly analysed in order to assess, first of all, the legitimacy of a term as an ethnobionym and hence its actual inclusion within the corpus. For this purpose the ancient texts available have been "interviewed" in order to collect as much information as possible about the biological referent(s) and, in particular, zooanthropologically relevant data such as (a) their biological identification; (b) their ethnotaxonomic status (or their rank as ethnotaxa); (c) their cultural importance; (d) their cultural construction (or the way the ancients construed and ultimately perceived them). Naturally not all these elements have emerged in every case, as this depended on the amount of material available for the ethnobionym under examination. However, even in those cases where the data were very scarce, the sheer linguistic analysis of a descriptive ethnobionym may provide valuable information, notably as regards points (a), (b), (d), but also as regards point (c) since, for example, taxa of the folk-specific level (see below), usually named with "specific binomials" (e.g., white oak, red oak), can be named with unitary lexemes (e.g., oak, woodpecker) or with "generic binomials" (e.g., mockingbird, catfish) due to their cultural importance (Berlin 1992; for unitary lexemes see Conklin 1968; for generic and specific binomials see Hunn & Brown 2011). As to, again, (a), (b) and (d), Lat. pagurus, for instance, a loanword from Greek págouros (a traditionally unidentified crab), has been identified with the warty crab (Eriphia verrucosa (Forskål, 1775)) due to the conclusion that the name is not an unanalysable monomial (e.g., oak) but actually a binomial (e.g., woodpecker) and hence an analysable compound, whose gloss is "guardian of the reef " (Guasparri 2006a).
The fact that the referential constraint is connected to the typical rocky habitat of this crab is not only relevant to its biological identification and ethnotaxonomic status as a folkgeneric (see below), but also to its cultural construction (e.g., in the ancient speakers' mental lexicon the term will have been linked to other "guardians" such as kēpourós "guardian of the garden", múkhouros "warder of the interior", nuctoûros "nightwatchman", etc.). However, besides etymology and referential constraints, the impossibility of eliciting a biological identification has been overcome diachronically, by looking for phonological similarities to ethnobionyms currently in use in the geographical area of interest. It. occhiata, for instance, is both the morphosemantic and the phonological equivalent of Lat. oculata (lit. "eyed [fish]", with reference to the black spot near the tail of the saddled seabream (Oblada melanura (Linnaeus, 1758)), more salient than the one found on similar and co-occurring seabreams (Diplodus sp.; Fig. 2); and indeed, the ancient descriptions match the biological referent of the modern name, as one may somewhat expect when inherited forms are involved.
Actually, the link between the ancient and the modern ethnobionym is not always so easy to explain. In the case, again, of Greek págouros, both Modern Greek kábouras, kaboúri and Turkish pavourya are attested as general terms for "crabs", not as more specific terms for a particular kind of crab (and specifically Eriphia verrucosa) as the reconstructed reference of the ancient name suggests. However, given the high intraspecific similarity within Mediterranean crabs, Eriphia verrucosa (the biggest non-demersal species, hence the most visible), "could have easily served as a prototypical species to designate any kind of crab" (Guasparri 2006a: 131), which might explain, diachronically, the ethnotaxonomic shift from the ancient folk-generic (e.g., Eng. mallard) to the modern intermediate taxon (e.g., Eng. duck).
But before going deeper into ethnotaxonomic status attribution, let us briefly recall the presumed universal hierarchical levels of ethnotaxonomy whereby people the world over classify the living organisms around them (Berlin et al. 1973;Berlin 1992). In this regard, the folk-generics just mentioned above are the elemental taxa, the most numerous in every folk-taxonomy, those on which people not only base their understanding, memorization and naming of the living things around them (e.g., dog, mallard, catfish) but also construct more inclusive higher-order taxa. The latter would be, in descending order of inclusiveness, folk-kingdom taxa (e.g., our "plant" or "animal" categories, usually "covert", i.e. unnamed, in most cultures); life-form taxa, which directly include most folk-generics (cf., typically, the equivalents of our "snake", "fish", "bird", more frequently named cross-culturally than "mammal" and "wug"; Brown 1984); intermediate taxa, usually covert categories, less common than life-form taxa and comprising much fewer and more morphologically similar folk-generics (e.g., duck, bird-of-prey). Finally, taxa of subgeneric rank go from folk-specifics -usually a few taxa named mostly through specific binomials and based on single contrasting features selected among referents subsumed by the same folk-generic (cf. white oak vs red oak with respect to oak) -down to folk-varietals ("nothing more than a structural replication of the folk-specific [ranks]" [Hunn & Brown 2011: 328]).
With this outline in hand, we can now give some examples of how to use the sources as "informants" when it comes to assigning a given taxon to one of the above ethnotaxonomic ranks. When Varro (Ling. 5, 77) numbers umbra "brown meagre" among the Roman "fish names" (vocabula piscium) formed metaphorically (cf. umbra, lit. "shade"), he provides significant evidence that this taxon is part of the life-form piscis "fish" (Fig. 3). Then again, based on the same passage, it is most likely that umbra itself, as a taxon, be a folk-generic, since the majority of folk-generics the world over is included in taxa of the life-form rank (Berlin 1992: 23;Hunn & Brown 2011: 327). This conclusion is confirmed by Columella (Rust. 8,16,8), who names two kinds of umbra, the indigena (lit. "native") and the punica (lit. "Carthaginian"). Indeed this pair can be said, both semantically and linguistically, to form a "contrast set", or one "comprised of just those taxa that are immediately preceded by the same taxon" (Berlin 1992: 169), i.e. oc-cupy the same ethnotaxonomic level. In other words, the two umbrae are just two specifications of the kind umbra, and do not constitute a "kind" in themselves, being distinguished not by many features simultaneously but by just those single contrasting features which modify the superordinate kind umbra. This kind of relation is typically (but not necessarily) observed between a polytypic folk-generic and its folk-specifics and is reflected, ethnobiolinguistically, by the latter being commonly named through specific binomials or, more generally, "productive" (Berlin 1992: 28) binomials, i.e. compound names whose head term denotes the superordinate category the animal referent is included in. This is also the case for the relation between the two folk-specifics umbra indigena, umbra punica and their superordinate folk-generic umbra.
As one may expect, things are not always so straightforward: cross-culturally, for instance, there are many cases of folkspecifics of great cultural importance being named through monomials (alias "simple uninomials" [Bulmer 1974: 22] or "simple primary lexemes" [Berlin 1992: xx]).
Thus, "ultimately a taxon is judged as folk-specific by virtue of the fact that it is immediately included in a folk-generic" (Hunn & Brown 2011: 328), which means at least that the ethnobiolinguistics of a taxon is not always a confirmation of an ethnotaxonomic affiliation. However, when working on "silent" informants as in our case, the ethnobiolinguistic structure and its regularities may be particularly useful for reconstructing the ethnotaxonomic rank of an ancient ethnobiomym. Sponges, for instance, are named by the Romans (and the Greeks) through what may appear simple monomials. In fact, at least in three cases out of four (i.e. tragos, manos, Achillium), a closer look at the linguistic form of these ethnobionyms reveals not only their adjectival nature as modifiers of an understood spongea "sponge" (the "head" of the compounds) but ultimately their ethnobiolinguistic status as "specific binomials". As a consequence, spongea of the ancients will very likely be a folk-generic taxon. This is not only revealed, perceptually, by the fact that sponges represent a very salient physical discontinuity with respect to other marine organisms (i.e. they have some very distinctive physical characteristics), but also by the fact that our ancient "informants" group sponges together with sea-anemones in a superordinate covert category (Pliny [HN 9,146] speaks of animals having a "third nature", somewhere between plants and animals; Fig. 4). Thus the sponge folk-specifics in question reflect a division into four kinds which, as one may also expect from the high cultural importance of sponges in the ancient world (Marzano 2013), does not include all the biological species found in the Mediterranean but just those collected and/or marketed -the others were simply labelled as aplysiae, lit. "unwashable".

AnnotAted checklist of RomAn AquAtic AnimAls
The annotated checklist presented in Appendix 1 contains all the Roman folk taxa for aquatic animals recognized as such in the course of this study. Before presenting the list, some preliminary remarks are necessary.
Entries are ordered alphabetically according to their form in the ancient texts. The first column contains the folk taxon as named (or, in two cases, not named) by the Romans. Some entries (usually "doublets", but there are also cases of "triplets") are marked with either a superscript or a subscript number. The subscript number (e.g., the pair asellus 2 and asellus 1 ) refers to what has been called "taxonomic polysemy" (Blank 2003: 272f.;Seto 2003: 198f.), or the usage of the same term to denote "nested" sets of two related folk taxa, one unambiguously included within the other (Forth 1995: 28;1998: 325-330;Zariquiey 2014: 264f.): e.g., Eng. cats denotes not only the common house pet but also those known more technically as "felines", of which the cat is -at least for most westerners -the prototypical species, i.e. the most common and salient one (Hunn & Brown 2011: 321). Compare, in the list, the plural form aselli and its singular form asellus (entered as "asellus 2 (aselli)" and "asellus 1 " respectively), the latter denoting the prototypical cod (the hake, Merluccius merluccius (Linnaeus, 1758)), whereas the former (found as aselli, the plural form, in the sources) denotes cods in general, i.e. scientifically, all Mediterranean "cod-like" fish included in the Merlucciidae and Gadidae families. In this particular case, the taxonomic polysemy involves a folkgeneric (asellus 1 ) and an intermediate taxon (asellus 2 ), but it may occur between any two ethnotaxonomic levels. As for the names with a superscript number (e.g., perca 1 , perca 2 , perca 3 ), they refer to so-called "metaphoric polysemy" (Blank 2003: 268; see more recently Guasparri 2019), which comes into play when the same term is used for denoting two or more different organisms (or sets of organisms) which are somewhat related, typically because they resemble one another -cf., e.g., ethnobionyms such as perca 1 , the comber  whenever the identification involves more than three scientific species (e.g., apua "larva of several fish"), while the third column contains the vernacular name. It is essential to note that the identifications provided (for whose details see also Guasparri 2005) accord mostly to the traditional ones, i.e. those already proposed by renaissance authors, as well as by more recent ones. In all such cases I will refer, among others, to the rich information provided by the Thesaurus de noms de poissons et de créatures aquatiques developed by researchers from the Ichtya research program (Bisson et al. 2020). These and other references related to "traditional" identifications, will be quoted after the vernacular name in the third column. All the biological identifications have been either confirmed, corrected or proposed -in several cases with the help of specialists -by checking relevant information provided by the sources against updated checklists of Mediterranean fish and sealife (e.g., Coll et al. 2010), as well as detailed taxonomic inventories (e.g., Relini 2008Relini , 2010 2020]). The latter have been also used as a reference for the vernacular English names in the third column, together with the Multilingual dictionary of fish and fish products (OECD 2008). Non-native species have been excluded to the advantage of both endemic and native ones (i.e. respectively, those which have evolved and occur naturally in a given region and those which have arrived and established there spontaneously).
The fourth column contains information about the ethnotaxonomic ranks ("FK" stands for folk-kingdom, "LF" for "life-form", "INT" for "intermediate", "FG" for "folk-generic, "FS" for "folk-specific"). It is important to note that the acronym "LF+" indicates that the taxon in question has been assigned to a "supernumerary" life-form (see Data analysis), cf. e.g., piscis 1 or mitulus 3 . In the Figures and Tables captions, the abbreviations of these ethnotaxonomic ranks are ordered from most inclusive to least inclusive.
The fifth column contains one (sometimes two) Latin passage which provides the most relevant information for assigning a folk taxon to an ethnotaxonomic rank.
The sixth column contains information about the number of occurrences, i.e. how many times an ethnobionym occurs in the sources, which may be useful as a marker of both its actual spread in the Roman culture and, more generally, the reliability of the data. The ">" ("greater than") and the "<" ("less than") symbols are followed by a number in brackets showing the actual number of occurrences recorded in the Latin texts (not provided when occurrences exceed 30). A second value appears in case of a Greek loanword, i.e. with reference to the number of occurrences found in Greek sources. As to the definition of "occurrence", in cases of long passages where the name has been repeated many times by the author being quoted, contextual occurrences have been taken into account, rather than strict numerical ones (i.e. the whole passage has been counted as one occurrence, even though the name occurred more than once within it).
The seventh and eighth columns indicate, respectively, the superordinate and the subordinate taxon. In fact, following Conklin's (1968) key suggestions over what a dictionary entry for each lexeme which can be placed in a taxonomy should contain, these and other ethnotaxonomic relations (e.g., which other members of its folk taxon an ethnobionym contrasts with), are also found, more descriptively, in the cross-referenced tree diagrams (see Data analysis). Ethnobionyms separated by slashes are synonyms.
Finally, the ninth column is devoted to any linguistic, ethnobiolinguistic, zooanthropological (and sometimes philological) observation relevant to the analysis of the ethnobionym under examination. For example, remarks are made as to the reasons why a biological identification (ID) has been suggested, e.g., whether the ID is based on descriptions from the sources or just based on the name's descriptiveness; whether it is supported by diachronic clues (i.e. modern phonological and/or semantic equivalents of the ancient name), etc. Several inventories have been consulted for retrieving the animal names currently in use in the Mediterranean area: e.g., Froese & Pauly (2020); cf. also Bini (1967Bini ( -1970, Palombi & Santarelli (1986) and, for Italian "regional" names, Costa (1991) -implemented, for certain linguistic areas of the Italian peninsula, with personal survey or fieldwork (Russo 2018; Sessa 2019; Cheli 2020). The section ends with information about the ethnobiolinguistic type assigned to the name, i.e. "monomial", "productive binomial" and "unproductive binomial", the last two types referring to compound names whose head term is or is not a superordinate category -cf. swordfish (productive) vs sea hare (unproductive).

DATA ANALYSIS
folk tAxA And ethnobionyms Since this study focuses both on classification and nomenclature, the total number of entries in the checklist refers to a "hybrid" value, in that it comprises not only the ethnobionyms, or the word-forms used by the Romans for denoting a folk-taxon, but also two unnamed folk taxa (see COVERT 1 and COVERT 2 ), which yields, as a result, a total of 386 entries. In fact, if we consider the actual The Roman classification and nomenclature of aquatic animals number of Roman folk taxa for aquatic animals emerging in this study, i.e. the kinds or "groups of kinds" of aquatic organisms recognized by our Roman "informants", the final figure is 311, a number which results from the very biological identifications being suggested. In particular, this value has been obtained by lumping together those ethnobionyms which label the same organisms (or groups of organisms). As far as the number of named folk taxa (or ethnobionyms) is concerned, their figure is much higher (384) than the number of folk taxa, precisely because of the many synonyms used in the sources for denoting the same folk taxon. This figure, however, does not correspond exactly to the number of entries (386), precisely because of the two unnamed categories mentioned above.
Let us consider now the biological referents of our Roman folk taxa as they result from our identifications (Table 1): fish constitute 52.1% of the total (n = 162), invertebrates (i.e. molluscs, crustaceans and others) 38.9% (n = 121), marine mammals 3.9% (n = 12) and reptiles (sea turtles) 1.6% (n = 5). Let us analyse each category (all results, transferred into a database, were analysed through the software SPSS 20).
The fact that fish are the most recognized grouping of aquatic animals is not unexpected, given their high perceptual and ecological salience (i.e. they are both very noticeable and spread). Although with significant differences from scientific zoology (and all the more so when suprageneric categories are taken into account), this is also reflected at the level of folk-taxonomy. In this regard, in order to represent at a glance the folk taxonomic knowledge of an ideal Roman, I have used tree diagrams, i.e. branched diagrams representing taxa and the relationship between them. Each tree diagram is dedicated to one of the major suprageneric categories the Romans appear to have recognized for aquatic animals, starting from fish, i.e. piscis 1 "fish", mollia "cephalopods" (Fig. 7), conchylium 2 "externally shelled molluscs", cancer 2 /crustata "crustaceans", beluae marinae/cete (lit. "marine monsters", e. g., marine mammals, turtles, adult tunas, etc.). Nodes represent folk taxa and are arranged hierarchically from left to right (i.e. the left-most node, corresponding to the most inclusive rank, contains all the other nodes/taxa). The acronyms at the top (FK, LF, INT, FG, FS) refer to Berlin's ethnotaxonomic ranks and identify each folk taxon in the corresponding column as belonging to the rank in question; supernumerary life-forms are indicated as "LF1+", "LF2+", "LF3+" -numbers followed by the "+" (plus) symbol indicate the number of extra life-form levels exceeding the "canonical" life-form level (LF) proposed by Berlin 1992 (see Materials and methods). As to the synonyms appearing in the tree diagrams (cf. the labels with two or more ethnobionyms separated by slashes), their relatively high number is very likely due to the long time span separating the Latin authors on which this study is based and to the wide usage of Greek loanwords alongside Latin ones -Greek loanwords make up 55.4% (n = 164) of all the 296 word-forms recorded in this study.

Roman fish
The most inclusive Roman folk taxon for denoting aquatic animals in general is the life-form piscis, which literally means "fish" (i.e. bony plus cartilaginous fish, alias "true fish"), evidently the prototypical aquatic animals for the Romans. However, piscis is also used in a second broader sense (distinguished here as piscis 2 ) for denoting other non-fish aquatic animals. This case of taxonomic polysemy involving the term for true fish is common cross-culturally, although with dif-ferences as to the higher-order biological groups denoted by the polysemically broader "fish" name (e.g., Anderson 1967, Hunn 1982, Pawley 2006. In particular, the Romans used the broader term (piscis 2 ) for referring not only to true fish (piscis 1 ) but also to all aquatic invertebrates (molluscs, crustaceans, sea urchins, jellyfish, etc.). As the tree diagram in Figure 8 shows, piscis 1 is included within piscis 2 at the sublife-form level 1 (LF1+), while several intermediate taxa are subsumed, in turn, by piscis 1 -it is relevant to note that the Symbols: *, prototypical; /, synonymy; (…), ethnotaxonomic ascription only presumed, due to lack of explicit statements in the sources; + (superscript), multiple ethnotaxonomic ascription due to different statements in the sources; ?, presumed folk taxon. See each entry in Appendix 1 for details.
The Roman classification and nomenclature of aquatic animals tree diagram displays only true fish included in supra-generic taxa of intermediate rank. All such intermediates not only involve taxa of significant utilitarian value for the Romans (cf. pelamys 2 , lacertus 2 and asellus 2 , i.e. tunas, mackerels and cods; Fig. 9), but appear to be related to special-purpose categories also in Brown's "extended" sense (Brown 1984: 10), i.e. are defined, monothetically, in terms of "a single especially criterial attribute" rather than, polythetically, on the basis of overall morphological resemblance (cf. saxatiles, lit. "rock fish", plani, lit. "flat fish" and pelagici, lit. "open sea fish"). Also worth noting here are the two prototypical taxa asellus 1 "cod" and pelamys 1 "little or juvenile tuna", which are taxonomically polysemous with the intermediates asellus 2 and pelamys 2 , i.e. respectively, all cod-like and tunalike fish. As to the latter (pelamys 2 ), this is the "trade" name for tuna and allies (tuna cuts included) and comprises the most valued fish-based food among the Romans, i.e. what they called cetus or, with a Greek loanword, thynnus "the (adult) tuna". Equally popular, but definitely characterized by a lower status, were mackerels (lacertus 2 ), found again as an intermediate taxon, although in this case the prototypical polysemous folk-generic lacertus 1 "Atlantic mackerel" is only presumed (see lacertus 1 ). It is interesting to note that lacertus 2 includes the highest number of folk-specifics among Roman fish, which is further evidence of the fish's popularity.
RomAn molluscs, cRustAceAns And otheR inveRtebRAtes Let us consider now the "non-fish" part of piscis 2 in greater detail, first, again, in terms of higher-order biological categories (Tables 1, 2)  (n = 83) -23.2% of which is represented by externally shelled molluscs (n = 72) and 3.5% by cephalopods (n = 11); crustaceans constitute 6.1% (n = 19) of the total, the same value as the remaining invertebrates (see "other invertebrates"). However, as expected, Roman higher-order taxa differ from those found in scientific taxonomy.  fig. 10. -The Roman folk-taxonomy of conchylium 2 (i.e. mostly, our externally shelled molluscs). Abbreviations: LF, life-form; LF1+, sublife-form exceeding LF level by one more level; LF2+, sublife-form exceeding LF level by two more levels; LF3+, sublife-form exceeding LF level by three more levels; INT, intermediate; FG, folkgeneric; FS, folk-specific. Symbols: *, prototypical; /, synonymy; (…), ethnotaxonomic ascription only presumed, due to lack of explicit statements in the sources; + (superscript), multiple ethnotaxonomic ascription due to different statements in the sources; ?, presumed folk taxon. See each entry in Appendix 1 for details.

FS LF
The Roman classification and nomenclature of aquatic animals To begin with, the Romans do not seem to have conceived of aquatic invertebrates as a distinct, single life-form (unless they had used a covert one for whose detection we have no sufficient data) but, by and large, distinguished shelled molluscs (conchylium 2 ), cephalopods (mollia), crustaceans (cancer 2 / crustata). As to other invertebrates, sea-urchins (echini) were considered both shelled molluscs and crustaceans, while sea nettles (urtica/cnide) and sponges (spongea) on the one hand, and starfish (stella), sea cucumbers (cucumis) and jellyfish (pulmo) on the other, were probably grouped within two different covert categories, distinguished here as COVERT 1 and COVERT 2 , both denoting "plant-like" organisms (respectively attached and unattached). Finally, sea squirts (tethea) were classified both as shelled molluscs and as COVERT 1 .
But let us now consider these invertebrate categories in their ethnotaxonomic context as it appears from the single tree diagrams. Let us start from the life-form denoting all "externally shelled molluscs", i.e. conchylium 2 , the most numerous invertebrate folk taxon.
The tree diagram shown in Figure 10 illustrates the folktaxonomy of externally shelled molluscs, among which the Romans included also sea urchins (echini). A first significant characteristic of Figure 10 is the high number of what we have called "supernumerary" life-forms, i.e. life-form "sublevels" exceeding the universal number of taxonomic ranks proposed by Berlin (1992). Postponing the discussion as to the possible reasons for this (see Conclusions), it must suffice for now to note that conchylium 2 includes at least two of the most culturally important marine animals for the Romans, i.e. oysters (most valued for food) and purple-dye gastropods (famously the most prized sources of dye in the ancient world -e.g., Pliny, HN 9, 60ff.). Indeed it is within the folkgeneric ostrea/ostreum 1 ("oyster") and the intermediate murex 2 / purpura 2 /pelagiae ("dye-murices") that one finds the highest number of folk-specifics (12 for oysters and six for murices), a classical instance of "overdifferentiation" (Berlin 1973: 268; Fig. 11), or the split of a scientific species in two or more folk taxa. This is another marker of high cultural relevance, also confirmed, e.g., by such toponymic modifiers as Lucrina (lit. "of the Lucrine lake", renowned for its aquaculture production of shellfish), which is found both as a kind of oyster (ostrea Lucrina) and a kind of venus shell (peloris Lucrina) -the latter too was highly esteemed by the Romans. The fact that a Roman might have thought of oysters as the prototypical shelled molluscs seems proven, among other things, by the taxonomic polysemy between the folk-generic ostreum 1 and the (level +2) life-form ostreum 2 , which widens the meaning of the lexeme ostreum to include all bivalves and gastropods, i.e. all shelled molluscs, sea-urchins excluded (the latter are included in conchylium 2 ). Similar evidence exists for mussels. Indeed the ethnobionym mitulus, lit. "mussel", emerges as the prototypical mussel (mitulus 1 ) at the rank of folk-species, due to its taxonomic polysemy not only with the superordinate folk-generic mitulus 2 ("all marine mussels"), but also with mitulus 3 at the (level +3) life-form rank, where the lexeme widens its meaning even further to denote most bivalves. In the case of conchylium, this loanword imported from Greek into Latin retains the original wide biological reference (i.e. "all externally shelled molluscs plus sea-urchins") when used as the (level +2) life-form conchylium 2 , otherwise narrowing its meaning to denote both a particular mollusc-derived purple dye (the "blue" purple dye) and the mollusc which produces it, i.e. the dye-murex (see conchylium 1 ). The renown of the dye murex (Hexaplex trunculus (Linnaeus, 1758)) will have made it the Roman externally shelled mollusc par excellence, which explains the taxonomic polysemy between the prototypical folk-generic conchylium 1 and the life-form conchylium 2 .
A second invertebrate higher-order category subsumed by the life-form piscis 2 is cancer 2 /crustata, which is represented in Figure 12. As the tree diagram shows, the Romans included among cancer 2 /crustata both crustaceans and sea-urchins (echinus), which, alternatively, some Romans classified also as a kind of conchylium 2 or shelled molluscs, as just seen. Three intermediate folk taxa emerge, and in particular: l ocusta 2 /carabus 2 (used A B C D fig. 11. -Some of the most popular ostrea (externally shelled molluscs). A, a date mussel (balanus); B, a mussel (musculus -also myax or mitulus); C, a scallop (pectunculus); D, a spiny dye murex (murex or purpura). Detail from a Roman mosaic from Pompeii, 1 st century CE (Museo Archeologico Nazionale, Napoli; photo credit: A. Guasparri).
for lobster-like crustaceans; Fig. 13), squilla 2 /caris (denoting shrimp-like crustaceans) and cancer 1 (used for all crab-like crustaceans), each corresponding to three evident physical discontinuities. In other words, these groupings of crustaceans appear to be relatively predictable, to the extent that they can be explained in terms of human perception, i.e. in terms of a universal human characteristic rather than a culture-dependent one. It is interesting to note that the distinction between locusta 2 /carabus 2 and squilla 2 /caris, i.e. lobster-like vs shrimp-like crustaceans, is based on size more than overall body shape. The latter is at play when it comes to the third intermediate, cancer 1 , which includes, e.g., pinoteres 1 /pinophylax, the pea crab, a very small crab living inside a pen shell's mantle cavity. As to cancer 1 , this appears to be the prototypical crab-like crustacean for the Romans, as attested by its polysemy with the (level +1) life-form cancer 2 , which shares its ethnotaxomic role with the descriptive synonym crustata, lit. "crusted" (sc. animals). Contrary to squilla 1 and locusta 1 , which both represent  , with some externally shelled molluscs: A, a mussel (musculus); B, clams (chemae); C, spiny dye murices (murices or purpurae). Fresco from Herculaneum, 1 s t century CE (Museo Archeologico Nazionale, Napoli; photo credits: A. Guaspari).
the prototypical folk-generics polysemous with the respective superordinate intermediates, cancer 1 does not appear to be a folk-generic. Indeed, if this were the case, then its subordinate folk-specific taxa would be all designated through mononomials, rather than binomial names, which does not accord with cross-cultural data for folk-specifics.
cephAlopods And otheR inveRtebRAtes As to the remaining invertebrates ( Fig. 4), the Romans grouped together, rather predictably in terms of overall resemblance, such folk-generics as those corresponding to our octopuses, cuttlefish, squids and paper nautilus, i.e. what zoologists call cephalopods, another salient physical discontinuity within marine animals. It is remarkable, however, that the Romans distinguished two octopus folk-specifics, i.e. the "terrestrial octopus" (polypus terrenus) and the "open sea" one (polypus pelagius), with reference to the octopus's ability to crawl on land as opposed to those found at sea. As Figure 4 reveals, most other invertebrates other than cephalopods, crustaceans and shelled molluscs (the last two we have seen earlier), are represented by folk-generics which are directly included in two covert categories at the (level +1) life-form rank -we have marked them as COVERT 1 and COVERT 2 . Despite being included in the all-encompassing piscis 2 life-form, both these covert categories are described as "plant-like", either because their referents share a "third nature" between animal and plant (COVERT 1 ) or because they have merely "the same nature as plants" (COVERT 2 ) -e.g., the sea anemone, which is called urtica, lit. "nettle", by the Romans.
As to scolopendra marina or marine bristle worm, this is the only marine invertebrate not affiliated to piscis 2 but, directly, to the insects (insecta) life-form, which can be easily justified in terms of overall resemblance, inasmuch as scolopendra marina looks like a terrestrial animal found also undersea.

RomAn tuRtles
The biological referents of our Roman folk taxa include also "reptiles" (Table 1), which in our corpus comprise aquatic turtles only. Actually the Romans did not conceive of such a major category of scientific zoology (the term reptilis, lit. "creeping", appears in late Roman authors with reference to any creeping animal, including, e.g., snails). It is the ethnobionym testudo that appears to be used in the Latin sources for denoting, indifferently, both aquatic and land turtles. Indeed, instead of being characterized as kinds of piscis 2 , aquatic turtles seem to have normally been categorized simply as "animals", i.e. as folk taxa not affiliated to any life-form but directly subsumed by the folk-kingdom animalia (Fig. 14).
In particular, the Romans appear to have distinguished land, sea, river and "mud" turtles as testudo folk-generics, without assigning a "prototypical" priority to any of them. The same unaffiliation to any life-form apparently goes also for mus 2 , an alternative name for the freshwater turtle (Emys orbicularis (Linnaeus, 1758)) ( Fig. 15), whose folk-specific mus marinus (lit. "sea mouse") denotes the leatherback sea turtle (Dermochelys coriacea (Vandelli, 1861)). These turtles display certain physical characteristics which differentiate them from, respectively, tortoises (Testudo hermanni Gmelin, 1789) and the loggerhead sea turtle (Caretta caretta (Linnaeus, 1758)). In particular, traits like overall colour plus drop-shaped body in the marine and long pointed tail in the freshwater species (Fig. 16) may explain both the metaphorical polysemy with mus (lit. "mouse") and the usage of an ethnobionym other than testudo for denoting what, ultimately, is a kind of turtle (Guasparri 2015) -it may be noted that the affiliation of mus 2 to the intermediate testudo is only conjectural, hence the brackets. However, as to the Roman categorization of sea turtles, there is something more to be said, as the next paragraph will show.  cete oR "mARine beAsts" (A cAtegoRy in between) Coming to marine mammals (seals and cetaceans), another category shown in Table 1, once again the Romans did not use anything like the modern scientific category but, instead, seem to have grouped them together with several fish and even some strange, imaginary, animals. Indeed such a folk category stands out when compared to all the others in that its members cross-cut higher-order biological categories (roughly coinciding with life-forms at the ethnotaxonomic level): I am here referring to the variable called "various" in Table 1. This is the life-form the Romans called both beluae marinae (lit. "marine beasts") and cete (see cetus 2 ). As the label "various" indicates, the animal referents comprised in this group are "variously" (and somewhat unpredictably) aggregated with respect to "proper" taxonomy, i.e. that based primarily on the perception of similar morphological or behavioural characteristics shared among the recognized taxa ("general-purpose classification" in Berlin's terms). For instance, as Figure 17 shows, tunas (thynni) and catfish (siluri), which also appear in the "fish" (piscis) life-form (Pliny,HN 9,(43)(44), are grouped here with turtles (testudines), which Pliny contrasts with fish elsewhere (Pliny,HN 11,180). Moreover, Pliny (HN 32, 144) includes Nereids and Tritons, i.e. mythological sea deities with half-human, half-fish body, which appears bizarre to us. This inconsistent mutual exclusivity, whereby animals may be grouped as much with non-animals as with other animals, is a characteristic of "artificial" (Bulmer 1974: 95) or special-purpose classifications (i.e. utilitarian and symbolic ones; e.g., Forth 2016: 35) and is not found in folk-taxonomies where, for instance, a blackbird is a bird and never simultaneously also a fish. Despite this, the taxon cete 2 apparently retains two features of folk-taxonomies: it involves more than two ethnotaxonomic levels and all the members of a group may be assigned to it by virtue of external perceivable characteristics (size in our case). This is not the place to discuss such "mixed" pattern in detail; suffice it to say that it seems to stand as a further confirmation of the impossibility of separating artificial from natural classifications or, more generally, utilitarian factors from perceptual ones when it comes to the animal (or plant) knowledge of a given people. We should not fail to notice that the choice of regarding the folk-taxon cetus 2 and its synonym belua marina as a life-form is based on the wide range and internal diversity of the folk-generics it subsumes (Wierzbicka 1985: 189 ff.), whereas an intermediate would be a narrower, often covert taxon, usually comprised of closely related folk-generics (Berlin 1992: 149) and typically included in a recognised life-form. However, as Figure 17 shows, cetus also appears as cetus 1 within the folk-generics immediately included in the life-form as the prototypical (hence the asterisk) "extended" taxon. This is the most perceptually and/or culturally salient member of a subordinate category (usually a folk-generic, as is the case here) whose meaning has been widened by speakers so as to denote the whole superordinate class -a widespread ethnotaxonomic phenomenon called "genericisation" (Pawley 2006) or "generic name extension" (Berlin 1992: 29). It is not surprising, then, that cetus 1 denoted the large tuna, i.e. the fish par excellence according to the Romans, at least as far as the "special-purpose classification" of foodstuffs is concerned -indeed fishmongers were called cetarii, lit. "cetus dealers", by the Romans (see cetus 1 ).

MORE ON ETHNOTAXONOMIC DATA
Let us now analyse our data by considering ethnotaxonomic ranks in greater detail (Table 2).

folk-geneRics And life-foRms
Folk-generics represent, as expected, the most numerous taxa, making up 70.1% (n = 218) of all the 311 Roman aquatic folk taxa recorded in this study. Fish represent the largest group of folk-generics, and in particular 40.8% of the total (n = 127), a number which decreases slightly when considering the aquatic animals actually included in the Roman life-form piscis 1 (n = 121), due to the inclusion of some fish folk-generics within the life-form cetus 2 (e.g., cornuta "devilfish", gladius "swordfish", glaucus 2 "blue shark"; Fig. 17). The remaining folk-generics are represented mostly by shelled molluscs (10.6%, n = 33), which approximately correspond to the Roman life-form conchylium 2 , a category where the Romans included also sea-urchins (echini) and sea squirts The Roman classification and nomenclature of aquatic animals (tethea) but, seemingly, not sea hares (lepus marinus), which adds up to 32 folk-generics in all (Table 3). The other Roman major categories which subsume folk-generics (Table 3) are, in descending order, cetus 2 (5.8%, n = 18), i.e. marine animals equal to or greater than human size, including some "sea monsters"; cancer 2 (3.9%, n = 12), a life-form that the Romans applied to crustaceans, including sea urchins; mollia (1.9%, n = 6), an intermediate taxon that the Romans used for cephalopods (octopuses and the likes); COVERT 2 (1%, n = 3), i.e. "plant-like" but unattached marine invertebrates, such as cotton spinners (holothurium), barrel jellyfish (pulmo marinus) and starfish (stella). In general ethnobiological terms, the folk-generics found among Roman aquatic animals confirm Berlin's prediction on prototypicality (Berlin 1992: 24), i.e. their internal structure shows a recurrent pattern by which some taxa emerge as prototypical (i.e. better examples) of the whole superordinate taxon. This property is found at other levels of ethnotaxonomy and, as seen in Results and discussions, is indicated by taxonomic polysemy occurring between, e.g., a folk-generic and its superordinate intermediate taxon, whereby the latter is named after the former. Linguistically, the prototypical taxon, usually named with a monomial term, is named with an additional modifier meaning "genuine", "real", etc. when it needs to be unambiguously distinguished from other taxa of the same rank. This is the case for the folk-generic monomial umbra "meagre", which appears as the binomial folk-specific umbra indigena (lit. "native", i.e. "Roman") when contrasted with another folk-specific such as umbra punica (lit. "Carthaginian meagre", likely Umbrina ronchus or U. canariensis) -cf. also lupus germanus "lit. genuine seabass", with respect to other folk-specifics such as lupus lanatus, lupus maculatus, etc. (see the respective entries in the checklist).
As Table 2 shows, the life-forms recorded in our body of data are 10, which accords with Berlin's prediction that folk taxonomies the world over have "probably no more than ten or fifteen" taxa of life-form rank (Berlin 1992: 33). However, their most relevant characteristic in the Roman ethnotaxonomy of aquatic animals seems the fact that this relatively high number is connected to a proliferation in the "depth" of belua (marina)/ cetus 2 (large and hence salient marine animals equal or beyond the size of a human, e.g., sea-mammals, large tuna and sharks, "sea-monsters", etc.)   higher-order taxa. In particular, the number of levels at which life-forms can be found is more than one, which raises the total number of ethnobiological ranks above the "canonical" limit of six ranks established by Berlin (1992). For instance, the Roman classification of true fish (piscis 1 ), crustaceans (crustata) and cephalopods (mollia) exhibits one life-form "sublevel" (i.e. two life-form levels -see, respectively, Table 1; Figs 10; 12). This is a consequence of a taxonomic polysemy involving the Roman life-form term for "fish" (piscis), which, as seen earlier, has both a broader usage for denoting true fish plus most aquatic invertebrates (i.e. piscis 2 ) and a narrower one for denoting true fish only (i.e. piscis 1 ). The ethnotaxonomy of externally shelled molluscs (conchylium 2 ) is even deeper, with up to three sublife-form levels (Fig. 10). This proliferation of levels is a characteristic that Andrew Pawley had already noticed for aquatic animals in Oceanic folk taxonomies of aquatic animals. For instance, in Wayan, a dialect of the Western Fijian language, the life-form term for "fish" (ika) is used both for typical fish only and, more broadly, for fish, marine mammals and turtles, thus bringing about two different folk taxa, one "nested" inside the other (Pawley 2006). Pawley puts it down to the fact that life-forms as well may be influenced by cultural factors (e.g., turtles, much like fish, are valued food sources for the Fijians) and records several instances of even two or more exceeding levels (e.g., among shelled molluscs and crustaceans, both equally valued as food sources). This seems definitely the case for Roman shelled molluscs as well, whose importance as sources of food for the Romans is confirmed by the high number of taxa of folk-specific rank, which is typical of taxa of great cultural significance (Berlin 1992; see Conclusions). In this regard, as Table 2 reveals, externally shelled molluscs present a number of folk-specifics even higher than that found among fish (i.e. respectively, 29 vs 20 out of a total of 57 folk-specifics). Moreover, as another marker of high cultural significance, shelled molluscs exhibits evident cases of overdifferentiation (i.e. rather than corresponding to single biological species, folk taxa cover several of them), in particular among oysters and purple-dye murices, which were both extremely "popular" species. However, the fact that ethnotaxonomic levels proliferate in the Roman folk taxonomy of aquatic animals may be also a consequence of the intrinsic textual nature of our data. In particular, as already noted in Data analysis and as graphically attested in the tree diagrams, the number of synonyms, i.e. of two or more ethnobionyms used for denoting the same organism or group of organisms, appears to be high, a characteristic likely due to the long time span separating the Latin authors on whose works this study is based. Also, the high number of Greek ethnobionyms used mostly as synonyms for the correspondent Latin ones (see Data analysis) might have further contributed to the juxtaposition of possibly different ethnotaxonomic categories and, as a likely consequence, to the growth of ethnotaxonomic levels. For instance, the (level +3) life-form cochlos "marine gastropods" is only found once in Latin, and precisely in Pliny (HN 32, 147), when translating a passage from the Greek physician Xenocrates (Xenocrates, 23). Pliny seemingly replaces the Greek term kokhlías, which he normally translates with Lat. coc(h)lea "terrestrial snail", with the Aristotelian term kókhlos, a wider ethnobionym used for denoting all marine gastropods (e.g., Aristotle, PA 678b 24). Indeed Xenocrates's usage of kokhlías in this latter wider sense would have been misinterpreted by Pliny's readers, had he translated it with cochlea: hence Pliny's substitution and the resulting increase in the number of ethnotaxonomic levels. So, although it is not to be excluded that Roman speakers actually used the loanword kókhlos, it is a fact that this database is largely founded on Pliny's Natural History, whose debt to the Greek sources (especially Aristotle) and to Greek names used as a "supercode" to disambiguate the many "provincial" Roman names is explicitly acknowledged by Pliny himself (Pliny, HN 9, 52; see Guasparri 2013: 351). This is a further suggestion that the intrinsic textual nature of our data may be involved in the proliferation of ethnotaxonomic levels and, consequently, in the proliferation of ethnobionyms.

folk-inteRmediAtes And folk-specifics
Coming to folk-intermediates and folk-specifics, folk taxa of both these ethnotaxonomic ranks have been interpreted as reflecting culturally specific values (Atran & Medin 2008: 33) and the Roman classification of aquatic animals accords with this view. The frequencies shown in Table 2 reveal that both categories are more common among marine invertebrates and fish. In particular, out of a total of 57 folk-specifics, 20 of them are used for fish and 29 for shelled molluscs; while, out of a total of 19 folk-intermediates, seven refer to fish, six to shelled molluscs and five to crustaceans. This might be somewhat expected since marine invertebrates and fish represent not only the most numerous human-salient marine animals but also the most exploited ones as sources of food the world over. However, the single tree diagrams enable us to illustrate the correlation between folk taxa of both intermediate and specific rank and culturally important species in greater detail. Figure 10, for instance, reveals that most "polytypic" folk-generics, i.e. those subsuming one or more folk-specifics 1 , are related to culturally important species such as purple-dye producing molluscs and oysters: both species significantly involve the highest number of folk-specifics recorded for the folk-taxonomy of Roman aquatic animals. This, as already mentioned in Data analysis, is a typical case of overdifferentiation, i.e. the split of the scientific species to which a folk-generic may correspond into distinct categories, due to the (usually cultural) importance of that species for a local people (which, as a consequence, will mark its internal differences as folk-specifics). As to folk-intermediates, they equally appear to involve those aquatic animals which were most exploited by the Romans, as clearly seen, e.g., in the cases of pelamys 2 ("tunas, bonitos and the likes") and lacertus 2 ("mackerels") among piscis 1 (Fig. 8); or in the case of murex 2 / purpura 2 /pelagiae ("dye-murices") among conchylium 2 (Fig. 10).
1. In fact, Berlin defines folk-specifics as "two to several named subgroups" into which a folk-generic can be divided (Berlin 1992: 102). Hence for all the folkgenerics appearing in the tree diagrams as divided into one folk-specific only, I have assumed that the second folk-specific is in fact the prototypical, linguistically unmarked, one, coinciding [i.e. being polysemous] with the superordinate folk-generic, e.g., murena (i.e. the "standard" murena) and murena fluta are the two constrasting folk-specifics into which the folk-generic murena is divided.

NOMENCLATURE
In terms of ethnobiological nomenclature, the data shown in Table 4 accord with the generalizations on the structure of ethnobiological names suggested by Berlin (1992: 29f.). For example, most folk-specific taxa -44 out of 57 (77%) 2are named with composites, alias productive binomials, i.e. compound names with the superordinate folk taxon as the head term, plus a modifier (e.g., umbra punica, where umbra is the folk-generic); on the contrary, this name type is the least spread among named folk-generics -5 out of 283 (2%) -where, as expected, unproductive binomials are not rare -33 out of 283 (12%) -while monomials are the norm -230 out of 283 (81%). It is worth noting that the twofold categories called "monomial/unproductive binomial" and "monomial/productive binomial" correspond, respectively, to ethnobionyms like canis (marinus) and cochlea (terrestris), which can optionally incorporate the modifier in brackets -in which case they can be classified as binomials (either productive or unproductive). respectively, to ethnobionyms like canis (marinus) and cochlea (terrestris), which can optionally incorporate the modifier in brackets -in which case they can be classified as binomials (either productive or unproductive). Table 4 refer to the total number of ethnobionyms (384) and are therefore different from the ones given here.

Percentages in
As already noted earlier, the frequency of descriptive (or analysable) names is particularly high in the Roman nomenclature for aquatic animals -comparative data show much lower ratios (e.g., Forth 2016: 254). As Table 5 indicates, over 80% of ethnobionyms (i.e. 314 out of a total of 384) are descriptive. And, although this ratio is not evenly distributed in every higher-order biological category, even in those groups of aquatic animals which are less represented, such as aquatic reptiles or mammals, descriptive names are at least twice as many as non-descriptive (i.e. unanalysable) ones.
Considering now the distribution of descriptive names by ethnobiological name type (Table 6), the number of monomials -which one would expect to be less often descriptive than binomials are -is particularly significant (210 out of the total of 314). Descriptive monomials are ethnobionyms in which, as Berlin (1992: 27) put it, "salient morphological and behavioural features of plant and animal species are often encoded". Such property is made possible through a type of polysemy usually called metaphoric -but, more appropriately, "analogical polysemy" -of the kind at work in Eng. sole (a fish) as named after a sole (the underside of footwear) because of the referents' analogical similarity (Guasparri 2019). Table 7 shows the distribution of descriptive names across the semantic dimension at play in the analogical similarity selected for the output referent (i.e. the aquatic animal) to be named. In particular, "perceptual" refers to the typical ethnobiologi- - cal naming criterion -and our data confirm this typicalityi.e. the one based on the perception of external characteristics of the referent (e.g., gladius "swordfish", lit. "sword", refers to the fish's sword-like bill); "cultural" refers to culture-specific traits such as, e.g., the food-related ones (cf. a kind of tuna called cybium, lit. "little cube", from the shape of the cut fish sold in markets); "perceptual-cultural" refers to productive binomials, typically used for naming folk-specifics, like, e.g., lacertus Sexitanus, lit. "mackerel (imported) from the city of Sex", which combines a perceptual piece of information -lacertus, lit. "lizard", with reference to mottled colour -and a cultural one -the toponymic Sexitanus. Lastly, "perceptual-onomatopoeic" is used for ethnobionyms based on analogical similarity involving the sound emitted by the animal being named. The single occurrence of this name type in Table 7 refers to the fish called zaeus (John Dory, Zeus faber Linnaeus, 1758), an onomatopoeic name (cf. Gr. záēmi "to breathe hard") which captures the air blowing sound emitted by this salient marine animal (cf. the Latin synonym faber, lit. "blacksmith", with reference to the sound of a blacksmith's bellows).
The large majority of polysemous names are related to dryland referents such as, for instance, animals or objects. This matches a statement by the Roman linguist Varro that Latin aquatic animal names are "transferred from terrestrial things similar in some part" (Varro, Ling. 5, 77), which is very accurate with respect to both polysemy and referential constraints (see Materials and methods). Significantly, from a zooanthropological perspective, Pliny (HN 9, 2ff.) backs up Varro's remark by extending it to "the common opinion that anything born in any domain of nature exists also in the sea", which "contains likenesses of things and not of animals only" (translation by Rackham 1940, slightly modified). Both statements are utterly consistent with our data. Table 8 shows the number of dryland-related polysemous names (i.e. 212, or about 68% of the total number of descriptives) and their distribution across the same semantic dimensions seen in Table 7: the large majority of names are perception-based (81.1%, n = 172) and most of them are related to terrestrial animals (38.3%, n = 66). In particular, as to the single categories, the highest frequency is recorded for polysemous names related to terrestrial quadrupeds (21.7%, n = 46), followed by those constructed on similarity to objects (19.8%, n = 42) -both these sets of referents were possibly the most likely to be encountered by the Romans in their everyday life. The "others" category is represented by names related to body-parts (cf. unguis, lit. "nail"), deities (cf. Adonis, Veneria, etc.), professions (cf. cinaedus, faber, etc.), and other minor semantic domains.

CONCLUSIONS
The Roman folk-taxonomy of aquatic animals as reconstructed from our "textual" informants comprises 311 folk taxa. This number is smaller than the total number of named folk taxa (ethnobionyms) recorded in the ancient sources (384), largely because of synonyms denoting the same folk taxon.
Both from an ethnotaxonomic and an ethnobiolinguistic point of view, this folk-taxonomy reflects the universals suggested by Berlin (1992). Most folk taxa, for instance, appear to be categorized and named according to typical folk-taxonomic criteria based on perceivable characteristics of the referents such as their morphology and behaviour. In addition, taxa of folk-generic rank outnumber all other taxa, and (a characteristic they share with folk-specifics) some members emerge as more typical of the category than others; also, they are named mostly with monomials or at least unproductive binomials, while, on the other hand, productive binomials are the norm for folk-specifics. As to life-forms, taxa of this rank are ten in number, which, again, accords with Berlin's universals (Berlin 1992: 33). However, an ethnotaxonomic characteristic of this folk-taxonomy involves the number of additional "sublevels" where folk taxa of the life-form rank  can be found. Externally shelled molluscs, for instance, present three sublevels beyond the one life-form level established by Berlin (1992). This proliferation can be ascribed, on the one hand, to the great cultural significance of these animals as sources of food and, on the other, to the textual nature of our information and the long time span occurring among the authors of the texts being consulted. A folk-taxon, the life-form cete/beluae marinae (lit. "marine beasts"), has been dubbed an "in-between category" inasmuch as it presents characteristics which are typical of both specialpurpose classifications and folk-taxonomies.
As far as cultural significance is concerned, ethnotaxonomic data provide good evidence of the "popularity" of some species. The frequency of folk-specifics, for instance, confirms that purple-dye murices and oysters were the most esteemed molluscs, and, more generally, that externally shelled molluscs were very popular. Folk-specifics, together with folk-intermediates, reveal utilitarian associations also among fish: e.g., mackerels (lacerti), cods (aselli) and tuna-like fish (pelamydes) emerge as very exploited species. In particular, with regard to tunas (ceti), they appear to be the fish par excellence in terms of food consumption also for nomenclatural reasons: not only were fish-mongers named after them (cetarii), but tunas were overdifferentiated to the point that some "tuna-cuts" (cf. apolectus, cybium, tritomum) were actually categorized as tuna species. This is also consistent with the nomenclatural observation that several Roman fish (and tunas among them) were named on the basis of cultural-specific criteria like the food prepared from them or visible characteristics connected to their use as food rather than, much more typically, on the basis of pure perceptual criteria like external morphology or behaviour (Guasparri 2019: 23f.).
However, from a more general zooanthropological perspective, there seems to be some evidence that the Romans did not construe the sea and its inhabitants just as an inexhaustible larder (Guasparri 2019) -which, incidentally, would have been much more justifiable then than it is today.
From both Varro and Pliny, for instance, the sea emerges as some kind of double, a weird version of the terrestrial world, whose generative power brings about organisms which closely resemble terrestrial animals and objects (Pliny) and hence the names for the majority of them (Varro). This, however, rather than promoting the image of the aquatic world and its inhabitants as something derived and hence remote, in line with the traditional self-ascribed identity of the Romans as peasants and soldiers (de Saint-Denis 1947;Fruyt & Lasagna 2015), seems to stress the Romans' familiarity with it, as also suggested by our data in various respects. In this regard, one may once again refer to Pliny and in particular to his remark that marine animals, contrary to land and sky animals, are known insofar as they can all be given a name and thus, since they are less numerous than mammals (ferae) or birds (volucres), one can establish 144 as their exact number (Pliny,HN 32,142). In sum, the Mediterranean or mare nostrum (lit. "our sea") will have been "the Romans' sea" not only for its relatively circumscribed borders but also for the limited number of both its organisms and, ultimately, the ways of naming them.

Acknowledgements
This article is the result of at least 20 years of enquiry and reflection, and the number of people to thank would be very high. There are, however, some whose contribution cannot be overlooked: Romano Khan, who has carefully proofread and amended this text; Angela Giardino, who has given me valuable advice on the design and structure of the tree diagrams; Andrew Pawley, who has provided me with the text of a paper given at the 2006 Austronesian conference in the Philippines; Eugene Anderson, who has made his 1967 thesis on the fish classification of the Hong Kong boat people available to me. I would also like to thank (in random order): Giuseppe Manganelli, Stefano Morelli, Bianca Picciulin, Maurizio Bettini, Athanasios Giakazes, Benedetta Cheli, Enrico Maria Russo and Marina Sessa. Last but not least, I owe a very real debt of gratitude to Tarek Oueslati and an anonymous reviewer for their insightful suggestions and comments, and to Emmanuelle Rocklin for her editing work.  (Fig. 10) -Lit. "river snail". Productive binomial.  (Fig. 4) -A kind of stone which is "nothing but coral". Pliny also explains the name ("it petrifies"; cf. Gorgo, another name for petrifying-sighted Medusa -Either from Gr. (h)ellόs-όps "fawn-faced" or en-lopόs "scaly". Both etymologies are likely in referential terms.