Folk habitat sets
Csángós recognized and named at least 142–148 habitats. This number is higher than in any other landscape partitioning ever studied in the world cf.
. Though Fleck and Harder
 estimated 178 rainforest habitat types (104 primary and 74 secondary forest habitats) that the Matses in the Peruvian Amazon might distinguish, only 47 of them were named. Csángó habitat categories closely reflected the diverse habitat pattern of the valley bottoms. In the mountains, however, partitioning was less detailed. Consequently, the majority of folk habitats was concentrated in and around settlements, similar to what was found in the Alps
Our high number of folk habitat categories was not simply the result of the diverse mountainous landscape, nor did it reflect solely the deep ecological knowledge of the local Csángó people, but was also the consequence of our in-depth investigation cf.
. We argue that the method of our research and the questions we asked made it possible to elicit such a high number of habitat terms. The knowledge of habitat preference of plant species was mostly the result of personal experience, and, as such, was mainly implicit cf.
[13, 26, 52]. By asking the question „what kind of place does species X like?” we “forced” people to verbalize their knowledge related to habitats. Admittedly, this method prompted informants to answer a strange, culturally inappropriate question cf.
. We argue, however, that in this way, the rarely shared, rather implicit knowledge was verbalized more often than usual in a conversation or an interview. Even with this method there were habitat names (e.g. tepsányos – a type of fen) that were only elicited after many months of field work, although, as turned out later, they were known with the same meaning by all successive informants we asked.
Csángó habitat categories were more or less discrete units, but often had diffuse boundaries with neighboring types, because the selection and delimitation of prototypic types along continuous topological and topographical gradients is not obvious
[4, 22, 24], see e.g. “choeni ovogeshi” – transitional zone from floodplain to uplands, documented by Shepard et al. . This phenomenon is not unknown in folk biological (species) classification systems either, though classification difficulties are usually smaller in them
, and the spatial delimitation of tokens is usually not a problem. However, classification of taxa further from the so-called core taxon at the periphery of the prototype faces similar classification difficulties (cf. prototypicality effect
, prototype extension model
The great majority of the habitat names was known by all Csángós, which was thus part of their shared knowledge. The kaláka (a type of inter-familial collaboration) could be one of the key platforms of knowledge sharing. The Csángó landscape ethnoecological partitioning was highly lexicalized and had only few synonyms (all are listed in Tables
10). This is in contrast to the findings of Molnár
[5, 54] who reported a large number of synonyms (up to 17) in a steppe landscape, which he attributed to the erosion of knowledge, limited sharing and the diverse origin of the local community. Similarly high level of lexicalization was observed among South American tribes
[10, 11, 35], whereas the opposite was found in Southeast Asia
. The reason for the latter was likely the implicit nature of knowledge and not the lack of substantive knowledge.
Features used to distinguish habitats
Csángós distinguished habitats by the following features: land-use, dominant plant species, vegetation structure, successional stage, natural and anthropogenic disturbances, soil types, hydrological features, and geomorphology. In a similar high mountain environment in the Alps, Netting
 and Meilleur
[3, 4] found similar features used in the recognition and naming of habitats. Johnson also documented the importance of similar features (physiography, hydrologic features, vegetation and wildlife habitats) among Kaska Dena, Gitksan, and Witsuwit’en First Nations in Western Canada
[13, 26]. The use of abiotic and biotic features (typically vegetation-related ones, e.g. physiognomy, dominant and/or salient plant species cf.
[10, 11]) was typical among the Csángós, as was found in a number of communities in the world e.g.
[10, 11, 25, 35]. As opposed to some tropical peoples
[11, 21], and the First Nations in Canada
, Csángós usually did not consider the presence and absence of animals indicative of habitat types (but see one exception below).
Expressed distinction of primary (non human-transformed) habitats and semi-natural habitats was not typical in Csángó landscape partitioning. This is in contrast to the well-documented basic dichotomy found in the classification systems of several tropical people (see e.g. the primary forest vs. secondary forest in swidden systems)
[10, 11, 21]. It is perhaps because the entire area has been exposed to farming or forestry activities, resulting that the whole landscape has been rather significantly altered. There was only a single – rare and small in extent – habitat type known that was characterized by the absence of anthropogenic influences: it was the vadas hely (wild place – area not affected by humans; territory of wild animals).
Csángós do their best to minimize disturbance during farming. Natural and anthropogenic disturbances usually reduce the amount and/or quality of biomass, which Csángós try to avoid. This might be a reason why disturbance was a salient feature in their landscape ethnoecological partitioning. Disturbance may have been regarded as the main feature in 29 habitats. To counteract the effect of disturbance, Csángós tended to facilitate regeneration in several ways. They, for example, sowed seeds of Onobrychis viciifolia cf.
 in the dry, south-facing slopes to restore the continuity of the grass mat. Certain gaps in the grass were restored by sowing hayseeds collected in barns cf.
, Babai and Molnár unpubl.
The overwhelming majority of ethnic groups of which landscape partitioning has been studied so far live either in the tropics or the boreal zone. As a consequence, their habitat vocabulary focuses on the dichotomy of forested and cleared, cultivated areas
[10, 11, 15, 26], but see
[21, 25]. In Gyimes, where 99% of the landscape is under ‘cultivation’ (arable, pasture, meadow, or managed forest), habitat categories identified by their land-use formed the basic-level habitat partition. The economic importance of grasslands is greater than that of forests and arable fields due to the greater share of animal husbandry in the Csángós’ farming economy. Hay meadows are the most significant habitats in terms of survival, because they provide fodder for winter that should last up until May. The Csángós’ management of hay meadows is meticulously detailed. Thus, the habitat set of hay meadows also was much more detailed than that of other sets (e.g. pastures, forests). Csángós distinguished altogether 48 meadow types based on their soil, exposure and dominant species. On the other hand, their partitioning of pastures was coarser than that of the herdsmen in the Hortobágy and the Fulani in Burkina Faso
Some Csángó habitat terms are locative, especially in the geomorphological, hydrological sets (e.g. among trees, along streams), and are literally the same as those used in Amazonia by the Matsigenka
, or by the Gitksan and Witsuwit’en in British Columbia
[13, 26]. Similar locative expressions were also found by Molnár in the Hortobágy steppe
. In many Csángó habitat terms used for abiotically salient habitats, the term “place” was used. Martin found among the Chinantec and Mixe people a similar phrase (kam), but for habitats named after the dominant species cf.
. “Locative-like verbal expressions” were also used by Csángós (e.g. areas trampled (or not) by animals or humans, Table
In some cases, different people described the habitat of a particular plant species by using different habitat terms. In these cases, the exact meaning of the terms used was often not the same, as they referred to a different (though often overlapping) section of the same environmental gradient, or described the habitat by using another environmental feature. Shepard et al.
 also documents such examples in the Matsigenka folk classification (“apamankera nia – place of inundation, flooding” and “otsegoa – seasonally flooded island, branch of river”), though without a detailed discussion. A typical example from Gyimes was the terms porond (gravel deposit) and vízmeghajtotta hely (an area often flooded, with erosion and accummulation of gravel and sand). The former was the prototype, emphasizing that the area is elevated, whereas the latter mostly referred also to such places, but emphasized disturbance. The latter and its variants were used 14 times, of which 8 described habitats of Myricaria germanica and bushy Salix spp., species typical of gravel deposits (poronds).
Habitats that the studied community hardly knows may also represent important data in terms of folk landscape partitioning
. Csángós did not care much about rocky habitats and habitats of the extremely rare aquatic vegetation, and did not consider the differences caused by the heterogenous spatial pattern of co-dominant tall grasses. The former two were insignificant in their economy, whereas the dominant grasses formed a functionally uniform group (broad-leaved, tall grasses) whose division into any further subgroups might have seemed meaningless.
While some geomorphological features were also used as a habitat term (e.g. völgy (valley), tető (mountain top), oldal (mountain slope)), others were never mentioned as a habitat (e.g. bitkó (a small peak), nyak (neck), hegyláb (foot hill)). It seems that the former three have a special combination of environmental features, while the others have not.
Compared to the scientific botanical classification
, the Csángó system was only less detailed in the case of fen communities (communities dominated by Eriophorum sp., Glyceria plicata, Carex spp.) and rocky vegetation. In all other cases, the traditional landscape partitioning was more detailed than the scientific classification [cf. 39].
A special dimension in Csángó landscape partitioning is scale. Scale has two aspects: (1) topographical (spatial scale): some habitat categories (e.g. high mountains, settlements) occupy larger areas often comprising a habitat mosaic of finer-scaled habitats, while others extend only to smaller areas (e.g. stream side, birch woodland), and (2) topological scale: some more inclusive habitat categories (e.g. forest) are subdivided into subcategories (e.g. into spruce and beech forest, dense or sparse forest etc.), while others are not (e.g. gravel deposit, Nardus grassland). The two aspects are not independent. More inclusive categories often cover larger territories, while less inclusive ones usually cover smaller areas (cf. to different ranks in folk biological species classifications [cf. 1], and in landscape ethnoecological partitionings e.g.
[11, 18, 54]). Gilmore et al.
 for example identified three main categories in his general group of habitats: (1) “cuadu” (lit. soft earth) for the swamp habitats, (2) “aqui” (ugly forest) and (3) habitats without soft earth and ugly forest. Shepard et al.
 also lists inclusive categories. The Matsigenka nigankipatsa (lit. ‘middle earth’) is a broad category for habitats that are not flooded, but nigankipatsa is not included into any specific biotic/abiotic habitat type (other inclusive habitats: ‘flat area’, ‘many hills’ (= montane vegetation, above 600 m)
Habitats that Csángós distinguished could be arranged into characteristic categories at three spatial scales: macro-, meso-, and microhabitats. Similarly to Shepard et al.
, Abraão et al.
, and Gilmore et al.
, we also found that abiotic features (e.g. geomorphology, hydrology, edaphic conditions) often defined larger, broader habitat categories, while biotic feature-defined habitats used in finer partitioning. Habitats in the land-use set were often intermediate between macro and meso scales. Habitats elicited by Shepard et al.
 are mostly mesohabitats. Microhabitats are not mentioned. Martin
 found that the Chinantec have a keen sense of microclimate. They distinguish sites with a special environment (i.e. microhabitats) that can be used for cultivation of certain plants outside of their zonal range.
In conclusion, macrohabitats occupy usually large areas and comprise many habitat types, mesohabitats are usually smaller in extension and homogenous and are often dominated by one vegetation type, while microhabitats are embedded in mesohabitats and provide special niches for particular species. One reason for the use of different scales in landscape partitioning may be ecological, since species occupy somewhat different places (niches) in a landscape, some species are specialist, others generalist. For a precise description of these species-specific habitats a multi-scaled landscape partitioning might be better suited.
Csángós never used species composition (list of characteristic and/or dominant species) as one of their features, although species composition is the key feature in recognizing and classifying habitats for scientific purposes see e.g.
[34, 39]. Even when they were asked to list the typical species of a particular habitat, they listed only a few species (Babai & Molnár unpubl.). We emphasize that the many habitat terms derived from a species name (Table
3) did not reflect a scientific plant community used by phytosociologists and as suggested by Rab
 in the nearby Gyergyó-basin, but only pointed to the most salient species of that spot in case of the Csángós.
Indicator species, however, were often used by Csángós in describing plant species’ habitats. These indicator species were not necessarily the commonest ones in a given habitat, but were morphologically or ecologically salient see also
[18, 20, 25]. Whereas the majority of indicator species are woody in the communities studied elsewhere
[18, 35], the indicator species were primarily herbaceous plants in Gyimes.
Species regarded as indicators were mainly very well known in the community, and their habitat preferences were widely understood (Babai & Molnár unpubl.). These features made them well suited to characterize the habitat preferences of other species. The majority of indicator species were ecological specialists in this landscape, found mostly in a single habitat type (e.g. Onobrychis viciifolia, Nardus stricta, Leontopodium alpinum, Tragopogon orientalis). Surprisingly, the list also included some generalist species (e.g. Picea abies) that, however, were used to describe the habitats of other generalist species (e.g. Leucanthemum vulgare, Abies alba).
Some meadow types were named after their typical indicator species (Table
12). These species were generally not dominant, but were rather associated with certain characteristics of the habitat, most often with the treatment with manure. Species indicating manured sites were Tragopogon orientalis and Salvia pratensis. Nardus stricta was an indicator species of nutrient poor, acidic habitats at higher elevations. Onobrychis indicated nutrient poor, south-facing slopes. Pteridium aquilinum may become abundant in south-facing and rather dry slopes, whereas Alchemilla spp. reached high abundances at higher elevations, primarily in grazed grasslands. Among the Chinantec and Mixe, Martin
 also documented indicator species that are used to indicate soil types suitable or not for the cultivation of certain plants.
Multidimensionality of the Csángó landscape partitioning
Csángó landscape ethnoecological partitioning was multidimensional, as it incorporated several sets of features. Multidimensional landscape partitioning was suggested as a general phenomenon by Ellen
 and Hunn and Meilleur
, and was found among Matsigenka in Peru by Shepard et al.
, various First Nations groups in British Columbia
, and Fulani pastoralists in Burkina Faso
. The number of gradients are usually large in mountainous landscapes, but Csángós have also contributed to an increased number of gradients (or an increased expression of the existing gradients, e.g. on south facing slopes, or along the valley bottoms) by turning the forest-dominated landscape into a more diverse mosaic of grasslands, forests, and arable fields cf.
[3, 4]. This resulted in an increased number of combinations of environmental features that plant species can inhabit. Csángó people used these combinations to describe, and name habitats of plant species they use as a resource.
Where there is a single dimension of morphological discontinuity, a salient prototype can generate a one dimensional classification of related forms
, but the multidimensional nature of habitat dictates that classification too will likely be multidimensional
. In Gyimes, several gradients could be recognized in the study area, which served as the basis of multidimensional landscape partitioning (i.e. mountain top / mountain side-valley, forest / bush / grassland, rocky / poor / lush, wet / moist / dry). These gradients were often formed from simple dichotomous pairs cf.
, but more stages may have been included between the extremes in the classification of a rather complicated habitat mosaic. Besides land-use types, these antagonistic pairs formed the basis for the basic-level landscape partitioning, and helped recognize the most salient habitats in the landscape cf.
. Environmental features were combined flexibly in habitat descriptions see also
[4, 11, 13, 15, 35].
In South and Central America, South East Asia, and Canada, features used in multidimensional landscape partitioning are almost the same as those in Gyimes e.g.
[10, 11, 13, 20, 26]. Most of them also appear in Europe (e.g. hydrological, geomorphological, vegetational features)
[3, 4, 54]. The most striking exceptions are the successional stages of the swidden agricultural systems, as these agro-forestry systems were banned in Europe by the German-type forest administration during the 19th and early 20th centuries
On the contrary, Molnár
 found that herders’ habitat classification in a salt steppe environment (Hortobágy, Hungary) is not multidimensional. He assumed that in the flat, open, and on the medium- and long-term stable landscape, the various abiotic and biotic factors that determine different habitats are arranged along a single (key) gradient (depth of ground-water table), while many possible gradients (e.g. woody / non-woody, mountain / valley, rock / sand, successional, naturalness) are missing. Among the Baniwa people in Amazonia, Abraão et al.
 also documented a landscape partitioning with only one gradient. The authors could structure forests along this gradient into a single hierarchical system (with 5 ranks) analogous to Berlin’s
 folk biological classification system.
Based on these experiences, we propose the following hypothesis for testing: multidimensionality of landscape partitionings and the number of dimensions applied depend on the number of key environmental gradients in a given landscape. Hunn and Meilleur
 argue that folk landscape ethnoecological partitionings are organized into shallow classifications only. Ellen
 goes further by arguing that multidimensionality and the emic multidimensional continuum is what might prevent the development of a single well developed hierarchy. Gilmore et al.
 found that folk habitat types are grouped into several, separate, overlapping subsystems. Further detailed studies are needed that document emic classifications of habitats to deepen our knowledge on the relations of multidimensionality and hierarchy.
In conclusion we argue that multidimensional partitioning of the landscape in Gyimes made the nuanced characterization of plant species’ habitats possible. Multidimensional partitioning provided innumerable possibilities for the Csángós to combine the most characteristic and salient features when explaining the habitat preference of a particular wild plant species in their diverse, semi-natural, and intensively managed landscape. The number of dimensions applied by them seemed to depend on the number of key environmental gradients in their mountainous landscape.