Forests: the cross-linguistic perspective

. Do all humans perceive, think, and talk about tree cover (“forests”) in more or less the same way? International forestry programs frequently seem to operate on the assumption that they do. However, recent ad-vances in the language sciences show that languages vary greatly as to how the landscape domain is lexicalized and grammaticalized. Different languages segment and label the large-scale environment and its features according to astonishingly different semantic principles, often in tandem with highly culture-speciﬁc practices and ideologies. Presumed basic concepts like mountain , valley , and river cannot in fact be straightforwardly translated across languages. In this paper we describe, compare, and evaluate some of the semantic diversity observed in relation to forests. We do so on the basis of ﬁrst-hand linguistic ﬁeld data from a global sample of indigenous categorization systems as they are manifested in


Background
There is a well-documented discrepancy in how different official bodies define forest (Comber et al., 2005). Typically, however, similar sets of parameters are used for all such definitions, e.g. minimum physical requirements like "tree height" and "canopy cover" in combination with "land size". For example, the Food and Agriculture Organization of the United Nations' Global Forest Resources Assessment 2010 (FAO, 2010) defines forest as "land spanning more than 0.5 hectares with trees higher than 5 meters and a canopy cover of more than 10 percent". In spite of this official definitional variability, there seems to be an assumption still that treecovered environments are somehow a basic category that everyone can easily put a word to. Linguists too have long assumed that there is something fundamental about tree cover: the original Swadesh 200 Word List -a compilation of presumed basic, stable, and universal concepts for the purpose of comparison and historical reconstruction of languagesincludes the item woods (Swadesh, 1952). Several subsequent comparative lexical studies do the same, such as Dyen et al. (1992) and Ringe et al. (2002). This reasoning presupposes that the concept expressed in English as woods or forest is a natural one for languages to delimit and instantiate lex-456 N. Burenhult et al.: Forests: the cross-linguistic perspective ically, and that it ought to be translatable across languages with little effort.
However, the idea that some vocabulary is more "basic" or "universal" has long been under scrutiny (Hoijer, 1956). Furthermore, there is a growing appreciation within the field of lexical semantics that few concepts, if any, have identical lexical expression across all human languages (Majid, 2015;Evans, 2010;cf. Goddard, 2010), and that there are substantial challenges to cross-linguistic comparison of meaning (Haspelmath, 2010). In cross-linguistic studies, domains considered fundamental to human experience have received particular attention, such as motion (Talmy, 2000), the body (Enfield et al., 2006), and the senses (Majid and Levinson, 2011). These all point to considerable diversity in how meanings are lexicalized.
Landscape is another domain that has received increased semantic attention in recent years (Burenhult, 2008a;Mark et al., 2011;O'Meara, 2010;Rybka, 2016). This research shows that different languages segment and label the large-scale environment and its features according to astonishingly different semantic principles, often in tandem with highly culturespecific practices and ideologies. Presumed basic concepts like mountain, valley, and river cannot in fact be straightforwardly translated across languages. Yet, landscape features are a domain for which ideas of lexical basicness and universality persist, as in Chomsky's treatment of river as an elementary concept available to everyone independently of one's cultural and linguistic background (Chomsky, 2012) or in the computational semantic network analysis of Youn et al. (2016), which relies on traditional basic vocabulary, much of which is landscape-related.
Although an established topic of investigation in human geography and ecological anthropology, vegetation cover has so far taken a back seat in the linguistic inquiry into landscape. As with other landscape features, several questions come to mind. Which semantic principles and parameters do languages employ to classify tree cover? Is forest a universally recognized category? How do semantic strategies vary across languages? Existing literature does offer hints of an interesting diversity in how languages may lexicalize tree cover. For example, Lowland Chontal, a highly endangered indigenous language of Oaxaca in Mexico, is described by O'Connor and Kroefges (2008:299) as having a forest term muña, which apart from incorporating meanings similar to English "forest" or "jungle" also covers meanings like "bush", "underbrush", "overgrown wilderness" or "any type of weeds or garbage". Although muña is the closest Lowland Chontal equivalent of English forest or woods and a likely candidate for inclusion in a list of basic meanings, the term appears to invoke a wider meaning of a disorderly environment, one that is not kept in check by humans. An entity referred to as muña does not have to contain good-sized vegetation in the form of trees or in fact any vegetation at all.
An example of another semantic pattern comes from several indigenous languages in the Australia-Pacific region.
Here it is not uncommon to have a single term for "tree", "firewood", and "fire" (Schapper et al., 2016). For example, those are the English translations of the term rowa in Duna, a Papuan language spoken in New Guinea. The meaning of rowa also extends to instruments of fire, like "matches" and "lighter". Similar meanings are given for the term thuma in Wik-Ngathan and yugu in Guugu Yimithirr, two Pama-Nyungan languages of Cape York Peninsula, Australia (see Evans, 1992:495-496 for additional examples). The semantic connection between these concepts is obvious: fuel (firewood) and its primary material source (wood/trees) receive the same lexical expression as the burning of such fuel (fire). To our knowledge, no single language has so far been described as extending the meaning further to, say, "stand of trees" or "forest". However, cognate sets (i.e. sets of words with the same origin within a group of related languages) in some Australian languages provide evidence that such a semantic extension is not conceptually far-fetched. For example, in Kayardild -a Pama-Nyungan language spoken on the South Wellesley Islands, in the Gulf of Carpentaria -the word thungkuwa means "mangrove scrub" or "mangrove forest"; in closely related languages the cognates of thungkuwa have meanings like "wood", "firewood" (N. Evans, personal communication, cf. Evans, 1992). Indeed, familiar European languages sometimes display traces of co-lexicalization on parts of this semantic continuum, e.g. tree as land cover and tree as material, as in English wood/woods and French bois.
These examples underscore that the analyst must take maximal semantic diversity into account when approaching forest. The categories relevant to our inquiry may take us far beyond tree cover itself and its physical characteristics, and in very different directions according to the lexicalization principles of individual languages. As we explore further in the following section, it cannot be taken for granted that forest semantics can be straightforwardly defined, measured, or compared only in physical terms.

Tree cover: evidence from six speech communities
Our aim in this section is to identify and characterize semantically the closest lexical equivalent of forest in each of six languages, capturing some of the variation that may pose challenges to translations of the concept. The terms we investigate correspond to those that would be candidates for a list of basic vocabulary, such as the Swadesh 200 Word List (see Sect. 1). That is, they represent what we determine to be the best possible translational approximations of forest, and typically they are the translations offered by native speakers. We concede that this is a simplistic and selective approach to lexical semantics, and one that does not do justice to the complexities of language-specific lexical systems, nor their historical background. However, the approach does mimic and problematize the introduction of an external category and its instantiation across human representational systems, thereby serving to highlight some of the pitfalls that forest is facing when confronted with linguistic and cognitive diversity. In addition to exploring the meaning of forest terminology, we examine for each language the attendant question of whether treed environments are talked about as containers (that is, bounded entities which enclose other entities and which can be "entered" and "exited") or not. Containment is considered to be a fundamental topological relation which nevertheless displays cross-linguistic variation in terms of how it is semantically structured (Talmy, 1983;Bowerman, 1996). It has not yet been studied in depth in relation to potential landscape-scale "containers", such as forest. All languages in our sample give expression to this distinction in one way or another, offering interesting hints at crosslinguistic variation in how treed environments may be conceptualized.
Our sample is genealogically and geographically diverse, spanning six language families and four continents. Although the respective ecologies inhabited by the speech communities are also diverse, our sample does have a tropical bias. All six communities interact directly in one way or another with treed environments. All of them represent small-scale rural communities, typically follow traditional subsistence practices, and their lives are likely to be deeply affected by any forest-related policy. The accounts rest on long-term involvement in each speech community by the respective language expert and are based on a range of data types collected first-hand in the field, including stimulus-based tasks, elicitation, interviews, and natural occurrences of relevant terms in recordings. In each case, the analysis and interpretation involves a combination of such data types, resulting in the integrated accounts presented below. Methodological triangulation of this sort is recognized as best practice in language documentation and description (Lüpke, 2009).

A brief digression on the term forest
Before embarking on our global tour, we should briefly address the meaning of the term we have chosen to represent the starting point of our inquiry, namely English forest. The etymology of forest remains to be fully settled. It is likely that it derives via Old French forest from Late Latin forestis "royal wood, hunting ground protected by the king", but accounts of its earlier origins diverge. Gamillscheg (1969) proposes that the Latin term is derived from Frankish *forhist (cf. German Forst "tended forest"), in turn derived from Proto-Germanic *furho "fir", *furhísa "fir forest" (the asterisks signify that the forms are reconstructed, not attested historically). An older analysis by Wartburg (1949) holds that the Latin form instead derives from foras "outside" (suggesting protected forest which is out of bounds for members of the public) or alternatively forum "court of law", implying that forestis evokes something subject to royal jurisdiction. Either account shows that the term has gone through significant semantic change through history.
History aside, describing the present-day meaning of the term is far from uncomplicated. English is spoken as a first language by large numbers of people of diverse backgrounds and in vastly different environments, so speakers' notions of the meaning of forest are likely to vary across individuals and dialects. For example, an Australian English speaker's notion of forest may differ significantly from that of a British English speaker (cf. Bromhead, 2011).
For our present purposes we rely on several standard dictionaries in characterizing forest. What these all have in common is a definition of forest, which involves trees covering a large area. Some definitions are supplemented with references to high density and the presence of undergrowth. For example, the English Oxford Living Dictionary defines it as "a large area covered chiefly with trees and undergrowth", Wiktionary as "a dense collection of trees covering a relatively large area", and Merriam-Webster as "a dense growth of trees and underbrush covering a large tract". These definitions draw on parameters similar to the official ones (cf. the FAO, 2010 definition, Sect. 1). Trees and size of the area covered by them are clearly essential to the English meaning.

Avatime
The Avatime are about 15 000 subsistence farmers settled in eight villages in the southern Akwapim-Togo Ranges in the Volta Region of southeastern Ghana, western Africa. Their language belongs to the Kwa branch of the Niger-Congo language family. The Avatime territory is located in the Dahomey Gap, a section of the Guinean forest-savanna mosaic ecoregion which extends all the way to the Atlantic coast and interrupts the belt of tropical moist broadleaf forest that otherwise extends along the western African coastal region. The hills are covered by mostly secondary forest interspersed with farmland, most of which is shifted regularly but some of which is permanent; the area also features savanna-like environments with high grasses and few trees. The Avatime traditionally practise shifting cultivation, clearing new land with slash-and-burn techniques every few years. The main food crops are hill rice, cassava, yam, and corn, supplemented by a number of fruits and vegetables. The farmland is often located quite far from the villages, and the Avatime make daily trips on foot to tend their crops.
The Avatime category to be investigated here is li . ŋwàfù . , the term used to refer to tree-covered areas. The history of this term is partly obscure, but it probably has three components: a prefix li . -, which specifies which class of nouns the category belongs to, and a root wàfù . , which can be analyzed as a combination of the verb ŋwà "to clear bush" (alternatively the noun aŋwà "a type of thorny vine") and a form fù . , the meaning of which is unknown. li . ŋwàfù . is the only word in the language which refers to a tree-covered area and is thus seemingly the translational equivalent of English forest. However, the presence of trees is not central to its meaning, as it can be used to refer to any undeveloped or uncultivated place where grasses, bushes, or trees grow. More appropriate English translations might therefore be "bush" or "wilderness". For the purpose of this study, eight Avatime speakers were asked in their own language "What is li . ŋwàfù . ?". Of these, only five respondents mentioned trees, three of whom also mentioned grass and weeds. Two respondents explained that li . ŋwàfù . is a place where people do not live. Three people responded that li . ŋwàfù . is the place where people go to make fields for cultivation ( 7 o any undeveloped or uncultivated place where grasses, bushes or trees grow. efore be 'bush' or 'wilderness'. For the purpose of this study, eight Avatime t is li ̣ŋwàfụ̀?'. Of these, only five respondents mentioned trees, three of whom nts explained that li ̣ŋwàfụ̀ is a place where people don't live. Three people go to make fields for cultivation (ɔ̀nyɔ). Two respondents were happy to use one of them was asked whether it is possible to say that òdzògbè ('savannah, er, a third respondent contrasted li ̣ŋwàfụ̀ with òdzògbè, and said that there is what people do in the li ̣ŋwàfụ̀, the most common answer (from seven out of nd most common was 'hunting' (five respondents) and four respondents  . Two respondents were happy to use the term li . ŋwàfù . for an area with few or no trees -one of them was asked whether it is possible to say that òdzògbè ("savannah, grassland") is li . ŋwàfù . and responded yes. However, a third respondent contrasted li . ŋwàfù . with òdzògbè, and said that there is no li . ŋwàfù . where there is òdzògbè. When asked what people do in the li . ŋwàfù . , the most common answer (from seven out of eight respondents) was "go to farm". The second most common was "hunting" (five respondents) and four respondents mentioned one or more of "clearing the bush", "cutting trees", and "setting fire", all activities done to make a plot of land suitable for cultivation. Fields for cultivation ( loped or uncultivated place where grasses, bushes or trees grow. h' or 'wilderness'. For the purpose of this study, eight Avatime . Of these, only five respondents mentioned trees, three of whom that li ̣ŋwàfụ̀ is a place where people don't live. Three people ields for cultivation (ɔ̀nyɔ). Two respondents were happy to use was asked whether it is possible to say that òdzògbè ('savannah, spondent contrasted li ̣ŋwàfụ̀ with òdzògbè, and said that there is do in the li ̣ŋwàfụ̀, the most common answer (from seven out of mon was 'hunting' (five respondents) and four respondents nd 'setting fire', all activities done to make a plot of land suitable ) are made in the li . ŋwàfù . but are themselves not li . ŋwàfù . . However, because fields are always inside the li . ŋwàfù . , it is possible to say you are in the li . ŋwàfù . when you are in your field (see below for further discussion of containment).
In reference to particularly dense forest with big trees where people do not farm or have not farmed for a long time, Avatime speakers will use the phrase li . ŋwàfù . tsìtsì "old forest". Such referents would presumably all qualify as English forest but, again, it is not a perfect semantic match, since the English term in this case probably extends to treed environments which are not considered to be li . ŋwàfù . tsìtsì to speakers of Avatime.
In discourse the term li . ŋwàfù . almost always co-occurs with a postposition mὲ "inside", e.g. in the compound li . ŋwàfù . -nε mὲ bagaa "bush animals", literally "animals of the bush's inside" (-nε is the definite article). This is a characteristic which it shares with most other terms referring to landscape features, and it suggests that referents of li . ŋwàfù . are perceived as having containing properties. You can "go into" and "go out of" a li . ŋwàfù . , and "be inside" it.

Duna -hi
The Duna (Yuna) language is spoken by approximately 25 000 people in Hela Province, in the highlands of Papua New Guinea, and is classified as a Trans-New Guinea language (Ross, 2005;Haley, 2002). The Duna homelands are in the central cordillera of New Guinea, and cover a broad altitude range from about 400 m (along the Strickland River) to over 3000 m in the south and northeast (Robinson, 1999). Forest in this area is broadly classified as Central Range montane rainforest, with dominant species likely to include beech, laurels, and conifers (e.g. hoop pine), and plentiful orchids, ferns, and mosses (http://wwf.panda.org/). While forest cover and type vary depending on the area and altitude, most people live in close proximity to densely treed areas. The majority of Duna people are subsistence farmers (homes and gardens usually being established between 1200 and 1600 m in altitude), with sweet potato as the main garden cultivar. A diet of gardened food and domesticated animals (especially pigs) is mixed with seasonal semi-cultivated tree crops (such as pandanus), other foraged forest foods (e.g. mushrooms, ferns), and game. Thus, people spend time in the forest as part of subsistence activities, but the intensity of this varies greatly from person to person (e.g. depending on location, gender, and skill set).
The simplest way to translate forest in Duna is likely with the word hi. However, the Duna word evokes a more general notion of outdoor, undomesticated space, as well as (potentially) the presence of trees. As an environmental term, hi prototypically refers to wild forest areas, usually those that rise above the zone of settlement and agriculture. Such areas are spiritually charged and associated, for example, with powerful spirits and rites of passage that saw boys transformed to men in the testing ground of an untamed environment. Hi seems to contrast in at least some respects to the term phou, which refers to swamps and grasslands. However, beyond this contrast, it is not clear how the concept of hi enters into the more specific environmental classifications of the Duna. As described in detail by Haley (2002), these environmental classifications center on features such as altitude, drainage, water features, soil type, and characteristic species, for example, recognizing two main altitude regions, a lower one, wapia rindi "low ground" (which favors the growth of fruit pandanus), and a higher one, karia seke rindi "mountain base ground" (that favors the growth of nut pandanus). The latter is probably more typical of hi, but the overlap is not complete.
In addition to referring to forested places, hi is also used with the sense of "outside", in contrast to internal spaces. For example, when one is inside a house one can speak of going to hi to undertake an outdoor activity such as washing or to investigate a noise one has just heard outside. However, hi does not appear to be commonly used in relation to the outside of smaller objects or containers (cf. Jahai hüp in Sect. 2.4), although this needs to be tested further. The syllable hi also occurs in the verb hiwa-"to go/come out, to leave (a dwelling)", which is very likely a related word. Hi is furthermore a signifier of a lack of domestication. For example, hi can form part of the name of older or wilder varieties of cultivars, such as hi da, roughly translatable as "wild taro". Similarly, domesticating a pig can be described as making it such that the thoughts or consciousness of the pig are no longer hi but are directed towards humans.
There is little linguistic evidence that the Duna think of a forest or hi as a container to be entered. In the available data there are no examples of hi occurring in combination with the spatial term kou "inside" (e.g. to mean something like "the inside of the forest"), nor of hi occurring with the verb uku-"go into", (as in "go into the forest"). However, hi is found in phrases with the case marker -ra, which is used to mark concealed and internal locations (such as the inside of a bag). Thus, hi can be associated with hidden places and is indeed still used by Duna people as a place of retreat, refuge, and concealment in times of strife and flight.

Jahai -hüp
The equatorial rainforests of the Malay Peninsula are home to minorities speaking Austroasiatic languages. One of these is Jahai, a community of about 1000 band-based foragers in the Malaysian states of Perak and Kelantan and adjacent parts of Thailand's Narathiwat province. The mountainous territory which they traditionally roam is covered in equatorial Dipterocarp rainforest characteristic of lower mountain regions. Much of the forest cover is still primary, but there is ongoing logging activity in the southern half of the territory. The area lies entirely within the rainforest ecosystem, although large waterways, occasional rocky outcrops, and open river flats sometimes leave gaps in the forest cover. Jahai habitation traditionally consists of temporary camps of huts or lean-tos, typically under tree cover but sometimes also in open areas. Although predominantly subsistence foragers, the Jahai occasionally make swiddens. Jahai society is deeply embedded in the forest environment, which provides the constant scene for livelihood, mobility, social and kin relations, and beliefs. The shaded environment underneath the canopy is considered cool and healthy, protecting people from the heat of the sun and from the monitoring senses of supernatural beings above.
So it may come as a surprise that the Jahai language does not have a term which corresponds straightforwardly to forest. The closest equivalent is the term hüp. This is the term which Jahai consultants will respond with if asked, for example, to translate the Malay term hutan "forest, jungle" into their own language (Malay is the unrelated majority language and lingua franca of Malaysia, spoken as a second language by all Jahai). In fact, they provide this translation with such ease that one will be forgiven for thinking that its semantics are unproblematic and that the case is closed. However, closer examination will show that hüp is a relational noun, that is, its meaning expresses a spatial relation rather than an entity. Its closest lexical equivalent in English is the nominal reading of outside but with the pronounced notion of an outside region or realm (and not a surface, as in an outside wall). It represents the ordinary way of expressing a spatial relationship of non-containment, such as location outside a container like a quiver or a hut. It is also used for reference to regions outside a perceived realm of human communication, activities, and social relations, for example, a location outside the zone of a conversation (Burenhult, 2008b). A third context is reference to regions and locations outside bodies of wa-ter. In all of these readings, hüp is conceptually and lexically contrasted with klεŋ "inside", an opposite and equally relational noun used for reference to the insides of containers, dwellings, camps, conversations, water, and so on.
Given the densely forested nature of the area, any region referred to as "outside" at a larger scale is highly likely to have tree cover. This may well contribute to the confusion about the translational equivalents of hüp. However, none of the readings explicated above requires tree cover in order to apply and, in fact, none of them encode any vegetation at all. Whether permanent or not, treelessness does not have an influence on meaning: a large expanse of barren rock is still hüp, as are treeless river flats, and deforested areas. Saying, for example, of a logged area that "the hüp is gone" is completely nonsensical in Jahai, as is "let's plant new hüp". Any approximation of the English concept of forest will have to involve paraphrasing by means of the terms jhũ ʔ "tree" and hali ʔ "leaf, foliage". The latter term in particular is used frequently to express whether an area is forested or not, as in means of the terms jhũʔ 'tree' and haliʔ 'leaf, foliage'. The latter term in particular is used frequen area is forested or not, as in sɔc ka=haliʔ 'the foliage is gone', in reference to a logged area.
Unlike some of the other languages discussed in this paper, Jahai does not offer any evidence t perceived as being contained by them. The language does not have expressions like 'go into', 5 environment. This is in agreement with the notion of hɨp as 'outside realm', and with the fact that t the Jahai ecosphere. It is in striking contrast with Malay, the neighbouring majority language, wh masuk hutan 'to enter the forest' and dalam hutan 'inside the forest'. It is worth noting in this conte policy conveyed to the Jahai will be communicated in Malay and thus conjures an understandin fundamentally at odds with the Jahai notion.

25
The Lokono divide their environment into two main categories: konoko and karhow. We will here b konoko. This term refers to tall vegetation whose canopy provides shade and obscures vision. It is and coolness. At its densest and darkest, the konoko is called konoko êbera 'forest's giant', a phr ka=hali ʔ "the foliage is gone", in reference to a logged area.
Unlike some of the other languages discussed in this paper, Jahai does not offer any evidence that being among trees is perceived as being contained by them. The language does not have expressions like "go into" or "be in/inside" a treed environment. This is in agreement with the notion of hüp as "outside realm", and with the fact that the forest is tantamount to the Jahai ecosphere. It is in striking contrast with Malay, the neighboring majority language, which has expressions like masuk hutan "to enter the forest" and dalam hutan "inside the forest". It is worth noting in this context that any official forest policy conveyed to the Jahai will be communicated in Malay and thus conjures an understanding of forest which seems fundamentally at odds with the Jahai notion.

Lokono -konoko
Our next example comes from Lokono, a moribund language of the Arawakan language family, spoken in the pericoastal areas of the South American nations Suriname, Guyana, and French Guiana. The Lokono people are settled agriculturalists, who also practise hunting, fishing, and gathering. The areas inhabited by the Lokono are typified by dry plateaus dissected by numerous creeks and rivers. The depressions formed by these watercourses are dominated by rainforest, large parts of which are seasonally flooded. In northern Suriname, where the data were collected, four general vegetation types prevail -forest, wood, scrub, and grass -each of which is found across three different ecotopes -land, marsh, and swamp -resulting in a kaleidoscope of environments (van der Eyk, 1957:37). Primary forest on land is highly diversified and reaches up to 30 m. Marsh swamp forests are lower and less variegated. Woods are lower than the forest and make up dense walls of thin trees, overshadowed here and there by larger specimens. Scrubs and grasses form closed and open savannas on both land and marsh. These areas, populated with grasses, sedges, and scrubs, are often surrounded by "savanna forest" -a transitory form between the savanna and the forest. The Lokono villages are located on the border between the rainforest and savanna, usually in the vicinity of creeks.
The Lokono divide their environment into two main categories: konoko and karhow. We will here be mainly concerned with konoko. This term refers to tall vegetation with a canopy that provides shade and obscures vision. It is associated with humidity and coolness. At its densest and darkest, the konoko is called konoko êbera "forest's giant", a phrase with the noun êbera "giant", or konoko anakhubo "forest's midpoint", with the spatial noun anakhubo "midpoint". Konoko is associated with a specific type of spirit called konokokoya, venerated and avoided through a system of cultural practices (de Goeje, 1943;Roth, 1915). Places such as konoko êbera and konoko anakhubo are described as far removed from the village and suggest a continuum from the primary forest, into which the Lokono venture only rarely, to the more "tame" forest around the village, where trees are felled to make swiddens.
As long as there is tree cover that provides shade, the area is called konoko, even if it is seasonally or permanently flooded. The noun konoko does not receive any modifiers of the type "swamp forest" or "marsh forest" when referring to such waterlogged areas. The presence of water is fully compatible with the term konoko. However, konoko is not felicitous when talking about open, flooded savannas without trees or with dispersed trees that do not form a continuous canopy (e.g. stands of moriche palm, Mauritia flexuosa). Such areas are called onêbera "rain's giant", a combination of the noun oni "rain" and êbera "giant". Similarly, on dry land, konoko is limited to vegetation that has an umbral canopy. The forests, woods, and shrub on land can all be tall and dense, but if there is no canopy providing shade konoko does not apply.
For canopyless environments, the term karhow is more appropriate. It applies to grass and scrub savanna and is associated with higher lying areas, heat, dryness, and lack of shade. However, karhow can also apply to areas containing trees, some relatively tall, if the trees are dispersed or not tall enough to provide continuous shade. This is the case with the aforementioned "savanna forest". Such areas are sometimes referred to in Lokono with the compound konoko karhow "forest savanna".
Lokono offers first-rate linguistic evidence that its speakers conceptualize treed environments as containers. When talking about motion into konoko, the Lokono use the verb kodonon "enter containment", restricted to contexts in which an entity moves into a contained space (e.g. into a house or from a big river into a small creek). Moreover, konoko frequently appears with nouns encoding containment such as loko "inside" or koboroko "inside (of multiple objects, e.g. trees)". By contrast, when describing motion into the karhow, the opposite verb fotikidin "enter non-containment" is used. This verb applies when an entity moves out to an open space (e.g. out of a house or from a small creek into a big river).

Makalero -alah
Makalero is spoken by some 7000 people on the south coast of easternmost Timor, in the nation of East Timor. It belongs to the Timor-Alor-Pantar family, a small group of Papuan languages thought to originate in New Guinea (Ross, 2005). The Makalero territory extends from the coast to the mountainous interior of Timor, and the area is variably covered with tropical dry forest and moist deciduous forest. Most Makalero speakers are subsistence farmers who own different types of agricultural land: swiddens, orchards with fruitbearing trees, small vegetable plots, and rice paddies. Main crops include rice, maize, vegetables, and coconut.
The Makalero word corresponding to the English forest most closely is alah. There appear to be three aspects central to the meaning of this term: firstly, it denotes areas covered in trees and other woody plants; as one Makalero speaker puts it: "The alah is a place where trees live". This is a prominent aspect of its meaning and is the first thing mentioned by Makalero speakers when asked to define the term. In this dimension, alah can be said to contrast with fereh "grassland" or with mu ʔa namar "land that has been cleared of trees" and mu ʔa hare-hare ʔ "clean land, land with no vegetation". Secondly, the vegetation in alah grows wild, as opposed to that planted by humans. In this respect, it can be arranged on a continuum of human intervention along with ama "field, swidden" or ado "orchard", dana "abandoned field where grasses and trees are growing again", and alah kirin or just kirin "old forest". It thus becomes clear that alah functions on two levels: on the one hand, it is a generic term referring to any type of wooded area as explained above, and on the other hand, it contrasts with (alah) kirin. On this lower level, alah refers to a not very dense forest cover with relatively young trees which is exploited and tended (but not created) by humans. Kirin is distinguished by having tall trees and by being untended and located far away from people. Thus, in effect, alah has two meanings -"forest" in the generic sense and "tended forest" -and only the context can disambiguate the two. Lastly, alah contrasts with mu ʔa hopan "flat or treeless land". This not only shows again that tree cover is central to the definition of alah but also that it is associated in particular with mountainous terrain.
The term kirin "old forest" calls for some elaboration. It also translates as "old", "valuable", "heirloom", "taboo", and may be used to describe any old and prized object. Elsewhere in the landscape domain, kirin is used with ira "water" in the expression ira kirin "permanent spring". On its own, however, kirin can only be understood to mean "old forest". In Makalero culture, old forests and springs (and some other landscape features) are associated with potent spirits (called mu ʔa ki-ouar "the land's master"), which must be treated with the proper respect. They require regular sacrifices and trespassing in their domain is prohibited; visiting such places is safe only for prominent members of the clan which owns the land.
In Makalero, descriptions of location usually require that the location and located entity are accompanied by what is referred to as a locative verb, such as isi ʔ "to be at" or mutu ʔ "to be inside" (Huber, 2017). However, with some nouns which refer to natural, large-scale entities which are prototypical reference objects for locations, such verbs are optional in certain syntactic contexts (not unlike Australian English to go bush vs. to go into the bush). This class of nouns includes place names as well as landscape terms such as larin "mountain", meti "sea", and ama "garden". Alah also belongs to this class, which shows that treed environments are conceptualized as natural locations at the landscape scale. Interestingly, in those contexts when such nouns do occur with the locative verbs, alah always co-occurs with mutu ʔ "to be inside", unlike the other nouns, which co-occur with the semantically more general isi ʔ "to be at". Thus, like Lokono (Sect. 2.5), Makalero provides compelling grammatical evidence that location in treed environments, here represented by alah, is conceived of as containment.

Umpila/Kuuku Ya'u -maalatha and thungkuyu
The Umpila/Kuuku Ya'u dialect group is a Paman language (Hale, 1964) spoken on the eastern coast of Cape York Peninsula, Australia. It is severely endangered, having only a small number of elderly fluent speakers. Cape York Peninsula, a large remote peninsula located in the far northeast of Australia, was traditionally home to 40 or so distinct languages of the Pama-Nyungan persuasion -thus, languages of this region group with the main phylic classification of Australian languages. Cape York Peninsula is notable today as an important wilderness area, the largest in Australia and of global environmental significance for the diversity of habitats and endemic species found there. The coastal and inland ecology and topography of the Umpila/Kuuku Ya'u people's regionlocated around Lloyd Bay -is particularly diverse. It consists of a patchwork of ecosystems: coastline with rocky outcrops, islands and extensive reef; mangrove everglades; large estuary systems; plains with salt pans; tropical savannah; rainforest, and so on. Notable for the discussion of forests is the presence of the Great Dividing Range which runs along the length of the northeastern coast of the Peninsula. The eastern face of the range, particularly in the northerly part of Umpila/Kuuku Ya'u territory, is covered in thick rainforest and vine forests. This extends coastwards along the edges of the waterways. The territory in the south and inland of the range features areas of dry sclerophyll forest and tropical savannah.
While attempting to find the closest equivalent of forest in Umpila/Kuuku Ya'u, we run into a different kind of challenge. The language distinguishes two major types of tree cover, expressed lexically as maalatha and thungkuyu. Maalatha is tree cover predominantly composed of one tree species, e.g. eucalyptus forest, sandalwood forest (Santalaceae), and so on. One speaker explains in the local Englishbased lingua franca: only one kind (of tree), not different different one. These tend to be dry forests, but speakers are not attuned to this aspect of maalatha environments, since the core semantic element is uniformity of tree type. Speakers closely associate this term with hillside environments. The following Umpila speaker's commentary on the meaning of maalatha is typical and makes this association explicit: maalatha mean when you go inside hill in proper scrub now, ngampula waathaka kani maalathaku nha'atha ampuyku ("we all will go to the forest and will enter to search for possum"). This example also suggests that speakers of Umpila/Kuuku Ya'u perceive maalatha as environments which you "go into" (lexicalized with the verb nha'ana) and are contained by.
By contrast, thungkuyu refers to rainforest, typified by a diversity of plant life including large fruit-bearing trees, strangler figs (Ficus spp.), palms and ferns, woody vines, and so on. When talking of thungkuyu, Umpila/Kuuku Ya'u speakers repeatedly highlight the impenetrability and darkness of this environment. They focus on the consequences this has for mobility and access: you can't go through; like the dark in the night inside lo thungkuyu. Thungkuyu are places where people move carefully in groups through fear of becoming disoriented or lost. Indeed, the term itself appears to be originally derived from the term thungku, which denotes dark hues.
Maalatha and thungkuyu both require tree cover that is extensive, rather dense, and contains trees of considerable size (as opposed to a category iicha "thickets", "shrubs"). Thus, in a sense, both appear to qualify as equivalents of forest and they are basically translatable into English as "monodominant forest" and "rainforest", respectively. However, neither assumes a generic meaning which covers both types of treed environments, and there is no additional term which covers both categories.
Both maalatha and thungkuyu are closely associated with the spatial term pakay "inside, under, down". Pakay and its opposing term kani "outside, top, over" are conventionally used to refer to certain environmental zones. Pakay is associated with inland densely vegetated environments, because its meaning concerns containment.

Discussion
A comparison of our translational equivalents of forest offers some interesting observations. First of all, none of the investigated terms can be conclusively analysed as evoking the same meaning as forest, although two come close: Lokono konoko and Makalero alah, which both require tree cover and which both have a generic reading applying to any densely treed environment. The Umpila/Kuuku Ya'u terms maalatha and thungkuyu also require tree cover, but they represent a more fine-grained distinction than in English. Together they refer to entities which would probably all qualify as forest, as defined here, but the language lacks an overarching forest-type term which ignores the species characteristics of the treed environment. The Avatime candidate li . ŋwàfù . takes a step away from these meanings, because it does not require the presence of trees but refers to "bush" or "vegetated wilderness" in a more general sense. Duna hi and Jahai hüp, finally, abstract away from trees and vegetation entirely, both carrying a more general spatial meaning of "outdoors", "outside", or "outside realm". Our sample languages thus appear to spread out along a continuum, with highly concrete treeencoding meanings at one end and more general, abstract spatial meanings at the other. 1 This semantic continuum is congruent with whether the languages treat treed environments grammatically as containers or not. The languages with terms that are more concrete and encode vegetation -Umpila/Kuuku Ya'u, Lokono, Makalero, and Avatime -offer grammatical evidence that referents of the terms are conceived of as containers. By contrast, the Duna and Jahai terms can encode abstract spatial meanings and do not present such evidence but instead seem to conceptualize the referents as the opposites of insides, namely outsides. Following from this, we might hypothesize that Duna and Jahai do not construe treed environments as bounded entities, whereas the other languages do. This would be in line with the reading of the Duna and Jahai terms as denoting an unbounded, infinite "outside", while the human sphere inversely represents the bounded entity and container.
Comparing ecologies, we find that such an interpretation makes some sense. Duna and Jahai are arguably those languages in our sample with speakers who live in closest proximity to forested areas, spend the most time in them, and are most dependent on them for their livelihood. Both communities are surrounded by tree cover, which dominates the physical world that is accessible to their senses. Avatime, Lokono, Makalero, and Umpila Kuuku Ya'u speakers also interact rather closely with treed environments but do so on a more intermittent basis, approaching and penetrating them from afar for particular purposes. For these communities, tree cover is but one of several ecotopes and it does not dominate 1 Although this proposed semantic continuum helps to arrange the meanings of terms in our particular sample, we do not wish to suggest that it is inevitably relevant to basic forest terminology in other languages. We acknowledge the likelihood of forest terms with meanings that invoke rather different semantic parameters which may not be amenable to plotting on the concrete-to-abstract continuum proposed here. Further research and detailed semantic analyses from more languages are needed to determine whether the continuum is more widely applicable or needs to be adapted, and whether there are terms to which it is not applicable. their perceived landscape. The Lokono and Umpila/Kuuku Ya'u ecospheres are particularly heterogeneous in this respect. It is perhaps no coincidence that such communities conceptualize forests as bounded entities, give them distinct labels, and express them linguistically as containers which can be "entered" and "exited". It appears as if our sample communities divide the same domain into their own specific categories that are optimal for their cultural and environmental profiles.
It is also interesting to note how the communities diverge in terms of their ideology and beliefs surrounding "pristine" treed environments. For the Duna, the Lokono, and the Makalero such environments are spiritually charged; among the Makalero they are also highly respected and subject to restrictions. The Umpila/Kuuku Ya'u consider them dark, confusing, and fearful. Conversely, the Jahai consider them a safe and healthy home, whereas treeless spaces are uncomfortably hot and ritually dangerous.
We should also briefly note that our sample offers no evidence that suggests that similar subsistence modes result in similar concepts of forest. For example, those two communities in our sample which are traditionally hunter-gatherers -Jahai and Umpila/Kuuku Ya'u -represent the two extremes on our semantic continuum. They are also ideological opposites in their approach to rainforest environments.

Conclusions
It should be evident by now that forest does not lend itself to clear-cut translation across languages. Our small but diverse sample shows that terms that may be assumed to be close equivalents of forest in fact vary rather fundamentally as to their semantic encoding. While some do closely approximate the English meaning of a densely treed area of some size, others refer to untamed vegetation in a more general sense (akin to bush); yet others do not encode vegetation at all but instead evoke more abstract spatial meanings of outdoors or outside. Thus, crucially, there are languages for which the referents of the term do not need to contain that component which is so essential to both official and English definitions of "forest", namely trees. Effective translation of forest into such languages requires innovative paraphrasing, such as "outside area with trees". Conversely, our sample also demonstrates that there are languages with basic terms that partition treed environments into more fine-grained categories than forest, for example on the basis of species composition. This would be tantamount to English having, say, two completely different terms for coniferous and deciduous forest but no generic term for forest subsuming the two. Consequently, translation of "forest" into such languages requires more than one term. Previous literature indicates that the semantic diversity does not stop there: as noted in Sect. 1, languages may have forest terms which incorporate the meaning of garbage, and per-haps even fire. We leave it to the reader to ponder translational solutions to such semantic phenomena.
In this paper we have explored some of the cross-linguistic diversity applying to forest terminology. This diversity is not only of academic interest -it may also have a bearing on the practical aspects of forest policies and programs which rely on intercultural communication for their implementation. In particular, it may have consequences for current efforts aimed at standardizing forest definitions and measurements. It also suggests that great attention needs to be paid to categorial variation in designing and implementing forest agendas, and that receptiveness to and understanding of local indigenous classification systems are crucial to successful communication of those agendas on the ground (cf. Wartmann and Purves, 2014). Most importantly of all, we hope to have shown that linguistic diversity, although sometimes an obstacle to comprehension, can be a rich source of information and inspiration for scientists and policymakers alike.
Author contributions. All authors took part in the conception and design of the study, as well as the general analysis and interpretation of the data. Acquisition, analysis, and interpretation of datasets pertaining to the individual languages was carried out as follows: Avatime by SvP, Duna by LSR, Jahai by NB, Lokono by KR, Makalero by JH, and Umpila/Kuuku Ya'u by CH. NB coordinated the development of the manuscript, incorporating drafts of the language-specific descriptions by each author and drafting the introductory and concluding sections. All authors took part in the critical revision of the intellectual content of the manuscript, and all authors approve of this version to be published.