Plant identification credibility in ethnobotany: a closer look at Polish ethnographic studies

One of the main problems ethnobotanists face when publishing their results is the possibility of a mistake in the identification of the studied taxa. Therefore securing voucher specimens is now standard procedure in ethnobotany [1‐3], required by major journals and discussed in ethnobotany method manuals [e.g. [4, 5]]. On the other hand the results of studies not documented by voucher specimens are still sometimes published, particularly in the field of historical ethnobotany, where not only is there a lack of voucher specimens, but often we have to hypothesize about the taxonomic position of certain species known only by their extinct folk/local names [6]. Ethnobotanists may include sources in their databases, which contain Latin binominals that come from reliable authors (preferably from professional botanists), but which are not confirmed by voucher specimens. This situation comes about because historical data are often too important to be discarded just on the basis of insufficient documentation [7].

It seems that no one has ever endeavored to estimate the possible percentage of mistakes in ethnobotanical publications. One of the very few authors who has dealt with the credibility of data in historical ethnobotany is Svanberg [8, 9]. He presented a few examples of some so called "ghost data" - old and erroneous information, which has been repeated by subsequent authors. The importance of identification credibility in historical ethnobotany can be clearly shown by the study of Kufer et al. [10], who compared present use of plants by the Ch'orti' Maya from Guatemala with data gathered in the same population in the 1930s by Charles Wisdom. It turned out that some mistakes occurred in the former study, where a taxon was misidentified as belonging to a different family.

The quality of ethnobotanical information is increasingly discussed in a variety of contexts [11‐13], for instance ethnobotanical databases [14, 15]. For example in a database of ethnobotanical data on the Campania region in Italy [14], levels of certainty of identification were introduced (sure, unsure, etc.). Generally, the likelihood of a mistake in identification probably increases with the age of the studied publication/information. This happens for a variety of reasons, e.g. changing folk names or uses in time.

In order to analyze the issue of mistakes in plant identification we should look at the whole process of plant identification. With ethnobotanical data a few scenarios are most likely:

Obviously the ideal situation is 1.2.2.2, particularly if voucher specimens were shown/brought by more than one informant. However, different scenarios happen for a variety of reasons, of which the major three are:

1 the ignorance of the researcher,

2 the fact that the information may be published/recorded even if securing of a voucher specimen is not possible, because of the importance of studying the use of a taxon for the researcher,

3 the use of a plant is extinct and we have only historical records without voucher specimens.

In this study I would like to consider the problem of the credibility of ethnobotanical data in one country - Poland. Poland, like a few other European countries, has a rich 19 and 20^th century ethnographic literature concerning the traditional use of plants - for a bibliography see Klepacki's review [16]. As the Polish flora is relatively poor in plant species (it has approximately three thousand species), the concept of voucher specimens was difficult to understand, not only for ethnographers studying the traditional use of plants, but also for botanists, who were relatively sure of their identifications.

The first person who tried to verify the credibility of older ethnobotanical studies in Poland was Köhler in 1996 [17], who checked the identification of plants in Udziela's herbarium from the turn of the 19^th and 20^th century. A few years earlier Radwańska-Paryska [18] re-examined the herbarium of an 18^th century monk, Brother Cyprian, containing Slovak and Polish plant names from the Pieniny and Tatra mountain ranges bordering the two countries. Later, the author of this paper (ŁŁ) published an article on the taxonomic issues concerning the quality of the data and mistakes in the identification of taxa in ethnobotanical studies in Poland [19].

The aim of this article is to extend the investigations of the previous work [19], in particular:

A sample of Polish-language ethnobotanical literature consisting of 45 published sources [20‐64] (Table 1) and four voucher specimen collections were analyzed (Table 2). The analyzed publications consisted of a large proportion of Polish-language ethnographic publications with ethnobotanical content, which contained lists of regionally used plants including at least one Latin name. All such papers available to the author were taken into account. Most of the analyzed sources deal with either wild food plants (reviewed in the Journal of Ethnobiology and Ethnomedicine in 2007 [65]) or medicinal plants. Papers without Latin binominals or monographs on the use of single species were not included. Maurizio's [66] and Moszyński's [67, 68] major works were not taken into account, as they are syntheses concerning the whole of northern Eurasia (the former author) or all Slavs (the latter). Lists of plant names and databases compiled mainly on the basis of other published sources were not included either [e.g. [69‐71]]. The analyzed publications usually concern studies from the present area of Poland and in a few cases - western Belarus [44, 45, 47], western Ukraine [37] and Lithuania [56]. The publications from these countries were included in the analyses as they were written by Polish ethnographers working close to the present area of Poland, within its former, broader territory from before World War II.

The total number of identified plant taxa was recorded for each publication, as well as the number of taxa which were presumably identified erroneously. A reference to a species from one publication and each herbarium specimen were later referred to as a use-report, a term, which, although mainly applied to indicate a plant-use mentioned by a given informant [72], in this case can be used with a publication as a unit. This way of treating a literature citation as one use-report is used in ethnobotanical studies, which review earlier publications, where the number of informants and informant consensus is not given. For example this approach was used by Leonti et al. [73] to analyze the influence of the 16^th century herbal of Matthioli on present day ethnobotanical knowledge in Campania (Italy), and in reviews of edible plants of Spain [74].

The following methods of identifying errors were used:

The second part of the study dealt with the re-examination of voucher specimens (Table 2). The voucher specimen collections for ethnobotanical data are extremely rare in Poland and so far only four such herbariums have been found:

Both in publications and voucher specimen collections, only taxonomic errors were taken into account. Spelling mistakes were not included, nor were cases where the author was cautious and identified only the genus (for example Equisetum sp. instead of Equisetum hyemale). However the cases when only one species was reported in the literature as used in the area, though we have firm evidence that a larger number of closely related species was/is utilized were also treated as errors (inaccuracies), for example, a passage like: "blackberries (Rubus caesius) are used as food", as "Rubus caesius" should be replaced by "Rubus subgenus Rubus" or "Rubus spp."

The author set up a code of credibility for presentation of historical ethnobotanical data in tables:

H - confirmed by (a) voucher specimen(s),

A - confirmed by authority (expert botanist),

O - obvious common name universally used in a large area,

L - highly probable Latin name or a binominal scientific name used in the language of a given country corresponding to a Latin name, given by non-botanist,

N - identified using comparative analysis of folk names,

M - identified using data on the species' mode of use (in case of unusual species/uses),

D - identified using physical description of species,

R - identified with the help of the data of a species range or/and habitat,

U - highly uncertain (should be combined with another code),

P - identified using pictures (photographs or drawings).

The usefulness of such a code was tested by compiling an up-to-date list of wild food plants used in Poland from the 19^th to 21^st century (within the present territory, excluding the German population pre-1939). The list was based on the review of edible plants of Poland [65] and amended by recent publications by Łuczaj [79‐81] and Pirożnikow [82, 83] bringing more data on the subject.

Forty-six identification mistakes were detected both in the published material using comparative methods (Tables 1 and 3) and in the voucher specimens (Tables 2, 4 and 5). This constitutes 2.3% of the analyzed use-reports for the former set of data and 10.0% of voucher specimens. The mean mistakes rates per publication differ significantly between the two sets of data (Mann-Whitney U test, U = 35.5, P (exact version) = 0.032, P (Monte Carlo version) = 0.022), they were 6.2% and 9.2% respectively.

The comparative method revealed a relatively large number of mistakes in a few publications, both older [26, 32, 52, 53] and new ones [49, 54, 60], however no or single mistakes were found in most sources.

There was no correlation between the year of publication and the percentage of errors in the species list (Pearson correlation coefficient, r = -0.004, P = 0.98, Kolberg's postmortem publications were assigned to his death date of 1890). Longer lists of plants had slightly lower error rates (the correlation between the number of Latin binominals in a list and the percentage of errors in it was r = -0.28, P = 0.060).

The mistakes concerned a variety of taxa but only a few taxa were mistaken more than twice: Thymus, ten times (e.g Thymus serpyllum confused with Thymus pulegioides or T. vulgaris), Rubus (six), Rumex (six), Cirsium, Trifolium (both four), Chenopodium/Atriplex, Malva and Mentha (three each). When the taxa from two families were confused this usually happened because of two similar folk/scientific names (e.g. Chenopodium - 'lebioda', Origanum vulgare - 'lebiodka'; Hippophae rhamnoides -' rokitnik', narrow leaved Salix spp. - 'rokita', etc.), which suggests that the author looked up Latin names in a scientific key without illustrations. This kind of error was the commonest type of mistake (eighteen out of thirty-six errors where a possible reason for the error was identified). The second commonest type (twelve cases) were simplifications and inaccuracies - such as reporting the use of only one species when more species from the same genus were used at least as frequently (Table 3).

In the list of edible plants of Poland (Table 6) 39% of 192 use-reports are confirmed by voucher specimens (code H), 30% by scholars with reliable botanical expertise (code A), 13% using folk names known widely throughout the country and 11% by scientific names with unknown reliability (L). Only ten out of 192 were identified using folk names (N) and four by comparing species ranges (R; with help of other data, e.g. folk names). None of the species were identified by only using a physical description from literature (D), pictures (P) or mode of use (M). In ten cases the code U (uncertain) was used.

The lack of voucher specimens is obviously a problem in determining the real level of mistakes in older European ethnobotanical studies. A considerable number of errors was detected in the studied herbariums. The comparative analysis of species ranges and names did not reveal these mistakes. One may wonder to what extent other works can be trusted? Some ethnographers probably avoided making taxonomic mistakes by writing only about more common and widely known taxa and identifying taxa only to the genus level. Some authors mentioned in their publications that their study was documented by voucher specimens identified by a professional biologist (e.g. Orzeszkowa [77] and Wysłouchowa [64]) or that living/dried specimens were at least shown to professional botanists [30, 46] or that voucher specimens from the Polish Ethnographic Atlas were used [46]. Thus in the above mentioned cases the possibility of mistakes is much lower.

Some taxa are more likely to be confused than others. Unsurprisingly, errors often occur in genera with more than one species, in which the species are similar to each other and are poorly recognized in folk taxonomy, e.g. in the genera Mentha, Thymus, Tilia, Crataegus, Rosa, Rumex and in the subgenus Rubus. The identification may be particularly difficult in apomictic taxa (like Rubus - [84]). Intergeneric mistakes occur either between two closely related genera not distinguished by folk taxonomy (Carduus and Cirsium) or less related (or unrelated) genera, if one of them has a folk name identical or similar to the scientific or folk name of another genus (in Poland, Origanum and Chenopodium, Sonchus and Taraxacum).

Nearly all of the quoted studies were performed by ethnographers, not botanists, so it is impossible to quantitatively compare the quality of their work with that of the few people with a biological background who have contributed to ethnobotany in Poland (e.g. Pirożnikow, Rostafiński, Maurizio, Moszyński, Szulczewski, though the latter two were known mainly as ethnographers). This comparison is particularly difficult given that most of these expert botanists supplied us with relatively large synthetic studies [66‐69], and only Szulczewski and Pirożnikow contributed local monographs with ethnobotanical data [57, 82]. Each of these studies [57, 66‐69, 82] contains well over a hundred species. In five out of six of these works I have not encountered any identification mistakes. On the other hand in the work of Maurizio two mistakes can be suspected, which probably arose from the misidentification of folk taxa. Both concern famine plants used in Poland, quoted by the author. One of them is Cichorium, supposedly used as famine food in the Tatra mountains. Maurizio got this information from an ethnographic paper [25]. However the original source does not mention Cichorium but only a folk name - szczerbak. This folk name was used both for Cichorium intybus, Cirsium rivulare, as well as other Cirsium species [85]. Cirsium rivulare was a very important famine plant in the Tatras [85], whereas Cichorium intybus was never mentioned as famine food by any other source listed in this article. Another possible mistake concerns the use of Mellitis melisophyllum. Maurizio mentioned that this plant (he also cited the folk name miodnik) was used during famine in Poland. However this is a relatively rare plant. On the other hand there are a few species of plants, which were used as famine food under similar names (miodunka, medunka, miodownik), i.e. Lamium spp., Symphytum officinale and Pulonaria obscura [65, 79‐81].

Even if these two mistakes were confirmed, the total ratio of mistakes in the works of the five professional botanists would remain well below half a percent. However, due to the different character of these studies, I restrained from deeper statistical comparisons.

It is worth pointing out that the quality of the ethnographers' work is highly variable. More than half of the publications contained no detectable mistakes, in contrast to a few authors who frequently misnamed the plants they had studied.

It must be noted that there is a significant difference between the mean percentage of mistakes detected in the studied works without voucher specimens (6.2%) and the level of errors found in the works documented by voucher specimens (9.2%). This also illustrates that even in works documented by a herbarium, gross mistakes can occur when the specimens are not verified by a good taxonomist. Single voucher specimens impose yet another threat: although the plant was correctly identified in the field or due to a widely known folk name a specimen of the wrong plant may be collected. This may have happened to Orzeszkowa. Her herbarium contains a specimen of Hepatica nobilis identified as 'kopytnik' Asarum europaeum. The name kopytnik is universally used throughout Poland to name Asarum, so Orzeszkowa may have collected a wrong specimen as both species have evergreen leaves of similar size and grow in the same habitat (identification scenario 1.1 or 1.2.2.1.2 in the Background chapter).

It must be emphasized that the main source of errors in the analyzed literature was the confusion of local names with Polish official names (scenario 2.1.3 in the Background chapter). This probably occurred by the researcher looking up a particular local name in a plant guide and then automatically assigning to it the Latin binominal of a different genus whose official name was identical to the local name of the studied species.

The studied papers usually contained little or no methodological information, so in most cases we cannot be sure if mistakes arose with or without seeing the actual studied plants in the field.

Most errors in the identification of voucher specimens occurred within the same genus, and only 24% of genera were misidentified. In contrast with this, 48% of mistakes detected in the publications without voucher specimens concern incorrectly identified genera. As the total number of detectable mistakes in the studies without voucher specimens is roughly four times lower than that of the studies with voucher specimens (2.3% versus 10%), we can assume that three quarters of all the errors and half of the misidentified genera remained undetected in the studies without voucher specimens.

It is a matter of dispute whether there should be separate codes for voucher specimens identified by an expert in the field and for those identified by a non-specialist (e.g. a separate code E could be used for a taxonomic expert). This could be useful, but on the other hand it is very difficult to draw a line between these two categories. As a rule, all voucher specimens should be identified/verified by a specialist - in case of easily identified taxa - a botanist, and in the case of critical taxa (in the Polish flora: Rubus, Rosa, Hieracium, Oenothera, Alchemilla) - a specialized taxonomist, or the specimens should be identified only to the genus, subgenus or section level [2‐5, 19].

Fortunately the errors made in the presented publications were rarely repeated later. The only case of erroneous "ghost information" in Polish studies is a list of plant names compiled in two ethnographic works [70, 71]. This situation contrasts with Swedish publications, which according to Svanberg [8, 9] contain numerous ghost data.

In all the studied cases the ethnobotanical herbaria contained species which had been reported in the given region or cultivated there, which supports the idea of using detailed atlases of plant distribution to verify ethnobotanical data. Such biogeographical data could be coupled with estimates of species abundance and distribution in local habitat spectra.

Not all the codes presented in the methodology section were used in the real-life list of edible plants of Poland. Most taxa were identified using the codes H, A and L, more rarely O, N or R. The codes D, M and P were not used. However this list was compiled using numerous voucher specimens (hence code H predominates) and data from 'reliable' researchers (like Rostafiński and Pirożnikow, hence code A). If a similar list were to be compiled for countries where voucher specimens were not collected, or for earlier periods, the proportion of codes in the list may have been reversed.

An interesting issue is the use of photography to document ethnobotanical studies [86]. Although photographic images cannot replace voucher specimens, they can help to document the use of plants, as many (but not all) taxa can be easily identified to a genus level from photographs [87]. Currently, as many electronic journals allow for the attachment of additional files to an article, authors could be encouraged to provide photographs of voucher specimens. Or perhaps we could start thinking about a service of online depositories of photographs of voucher specimens? Yet another option for plant identification, almost exclusively concerning historical ethnobotany, is the identification of plants from drawings (e.g. in old herbals). This is not always easy, but is sometimes possible, particularly when coupled with plant descriptions and folk names [73, 88, 89].

At the end of this paper the author must confess his own error. While preparing a table for the article on the taxonomic issues in Polish ethnobotanical studies [19], a mistaken name for Veronica chamaedrys was published, i.e. wole oczy instead of żabie oczka. ('ox-eyes' instead of 'frogs' eyes'). This mistake happened when transferring hand-written records to the computer. That is another example of the way errors and "ghost information" can enter ethnobotany, even via a botanist.

Ethnographic papers without herbarium documentation contain on average at least 6.2% of mistakes. The verification of voucher specimens can increase this ratio to 9.2%. These mistakes most often arise by looking up plants using a local name in a botanical guide, and by the lack of cooperation between ethnographers and botanists. The large difference between the ratio of mistakes in the voucher specimen collections and the ratio of detectable mistakes in the literature is an argument for the rigorous use of voucher specimens, which are identified by a specialist, and for creating a service of online depositories of photographs of voucher specimens.

The presented code of credibility may become a useful tool for historical ethnobotany.