Tapping into non-English-language science for the conservation of global biodiversity

The widely held assumption that any important scientific information would be available in English underlies the underuse of non-English-language science across disciplines. However, non-English-language science is expected to bring unique and valuable scientific information, especially in disciplines where the evidence is patchy, and for emergent issues where synthesising available evidence is an urgent challenge. Yet such contribution of non-English-language science to scientific communities and the application of science is rarely quantified. Here, we show that non-English-language studies provide crucial evidence for informing global biodiversity conservation. By screening 419,679 peer-reviewed papers in 16 languages, we identified 1,234 non-English-language studies providing evidence on the effectiveness of biodiversity conservation interventions, compared to 4,412 English-language studies identified with the same criteria. Relevant non-English-language studies are being published at an increasing rate in 6 out of the 12 languages where there were a sufficient number of relevant studies. Incorporating non-English-language studies can expand the geographical coverage (i.e., the number of 2° × 2° grid cells with relevant studies) of English-language evidence by 12% to 25%, especially in biodiverse regions, and taxonomic coverage (i.e., the number of species covered by the relevant studies) by 5% to 32%, although they do tend to be based on less robust study designs. Our results show that synthesising non-English-language studies is key to overcoming the widespread lack of local, context-dependent evidence and facilitating evidence-based conservation globally. We urge wider disciplines to rigorously reassess the untapped potential of non-English-language science in informing decisions to address other global challenges. Please see the Supporting information files for Alternative Language Abstracts.

making better use of existing knowledge that has yet to be fully utilised due to the language of 3 3 5 publication should be a cost-and time-efficient approach for filling gaps and rectifying biases 3 3 6 in the evidence base for tackling urgent global challenges. In 1922 the philosopher Ludwig 3 3 7 Wittgenstein stated "Die Grenzen meiner Sprache bedeuten die Grenzen meiner Welt" (the 3 3 8 limits of my language mean the limits of my world) [30]. Hundred years on, his quote still 3 3 9 seems applicable to science today. Scientific communities should stretch the limits of our 3 4 0 shared knowledge, and its benefits, by uncovering knowledge that has long been accumulating 3 4 1 and continues to be produced in languages other than English. The searches aimed to identify peer-reviewed scientific studies (a study is defined as a paper 3 4 8 published in a peer-reviewed journal) written in a language other than English that tested the 3 4 9 effectiveness of one or more conservation interventions for any species group or habitat. Our and manually scanning titles and abstracts (or summaries) of every document in those sources.

5 5
We adopted discipline-wide literature searching, rather than systematic mapping/reviewing, as predetermined criteria for study inclusion [32]. For more details on the Conservation Evidence database, see Section English-language studies on the effectiveness of conservation only on studies published in peer-reviewed journals, so as to enable a comparison between 3 6 3 eligible non-English-language studies and peer-reviewed English-language studies stored in We originally aimed to cover the top 15 non-English languages on the basis of the number of 3 6 7 conservation-related publications, provided in Table S1 of [16]. However, we could not find 3 6 8 native speakers of Swedish and Dutch who were willing to collaborate, and thus both 3 6 9 languages were excluded from our searches. Instead, we were able to cover three additional 3 7 0 languages (Arabic, Hungarian, and Ukrainian). In total, our searches covered 16 languages 3 7 1 (Table S2). (hereafter referred to as searchers). The number of searchers for each language ranged from 3 7 5 one to six (see Table S2 for more detail). We used a range of approaches (e.g., known networks, Before starting the searches, every searcher was trained through the following four steps.

8 3
First, searchers were directed to read through a guidance document detailing the objectives and 3 8 4 processes of the searches. Second, searchers were also requested to read and understand the full criteria for selecting eligible studies during the searches, which were described in detail, Evidence project, who provided searchers with feedback.

9 9
Identifying and selecting journals for each language 4 0 0 We first identified and listed peer-reviewed academic journals published in each language 4 0 1 which were likely to contain eligible studies. This process involved one to four researchers for 4 0 2 each language (all native speakers of the target language, with at least a bachelor's, and often 4 0 3 higher, research degree; this often included the searchers) from relevant disciplines (see Table   4 0 4 S2 for more detail), who used a range of approaches (e.g., personal knowledge, opinions from "relevant" (mostly journals in relevant disciplines, such as agricultural/forest sciences and 4 1 0 general zoology), and "maybe relevant" (all others). Subsequent searches aimed to at least 4 1 1 cover all journals categorised as "very relevant" and, when possible, those in the other two 4 1 2 categories (see Data S1 for the list of all journals searched). Searches for eligible studies in each journal were conducted by manually scanning the title and journal, and by reading the main text of all papers for which the title and/or abstract were 4 1 7 suggestive of fulfilling the eligibility criteria (fully described below). All papers that appeared Study validity assessment). The journals were searched backwards from the latest volume, Data S1 for publication years covered for each journal). We also recorded the total number of 4 2 3 papers screened in each journal.

2 4
The following eligibility criteria, which were developed and published by the Conservation 4 2 5 Evidence project (https://osf.io/mz5rx/), were used. Criteria B. If no, the study will be excluded. 3. Could the intervention be put in place by a conservationist/decision maker to protect, 4 3 5 manage, restore or reduce impacts of threats to wild taxa or habitats, or control or mitigate 4 3 6 the impact of the invasive/problem taxon on wild taxa or habitats? If yes, the study will be 4 3 7 included. If no, the study will be excluded. trait/functional diversity), community composition, community structure (e.g., 4 7 5 trophic structure), area covered (e.g., by different habitat types), and physical habitat structure (e.g. rugosity, height, basal area).

7 7
Eligible types of study design 4 7 8 Included: Studies with After, Before-After, Control-Impact, Before-After-Control-Impact, where there is no test of a specific intervention by humans, or pure ecology (e.g.,  1. Does this study measure the effect of an intervention that is or was under human control on 4 9 0 human behaviour (actual or intentional) which is likely to protect, manage, restore or 4 9 1 reduce threats to wild taxa or habitats (including mitigating the impact of 4 9 2 invasive/problem taxon on wild taxa or habitats)? If yes, go to 2. If no, the study will be 2. Could the intervention be put in place by a conservationist, manager or decision maker to 4 9 5 change human behaviour? If yes, the study will be included. If no, the study will be 4 9 6 excluded.

9 7
Eligible populations or subjects 4 9 8 Included: Actual or intentional human behaviour including self-reported behaviours.

9 9
Change in human behaviour must be linked to outcomes for wild taxa or habitats. occurred under a conservation program (e.g., we would exclude a study demonstrating  resulting in the conservation, management, and restoration of wild taxa or habitats.

0 7
Interventions which are particularly likely to have a behaviour change outcome include, • New policies for conservation/protection.

2 7
Excluded: Impacts from climatic or other natural events. Studies with no intervention, e.g.

2 8
correlating human personality traits with likelihood of conservation-related behaviours. From each of the studies that were identified by searchers as potentially eligible, the following 5 4 8 metadata were extracted and recorded using a template file: -Link to the article (URL, if available) of providing nest boxes on the breeding success of blue tits") The metadata were extracted largely by the searchers, but, for some languages where the 5 7 2 searchers were not available, by other collaborators who are native speakers of the language 5 7 3 and are at least undertaking or have a bachelor's, but often higher research, degree in a relevant discipline (see Table S2 for more detail). They were all requested to first read and fully before starting data coding.

7 7
For all studies that were validated as eligible (see Study validity assessment), the recorded The eligibility of each study that was identified as being potentially eligible was validated by at 5 8 9 least one experienced literature searcher (assessors) at the Conservation Evidence project (see  native speakers of each non-English language. This process was conducted by assessing the 5 9 3 English-language title, abstract and one-sentence summary of each study identified by the involved direct discussions between the relevant searchers and assessors to obtain clarification 5 9 6 on the details of each study. Those studies that were deemed ineligible by the assessors were 5 9 7 excluded from the final list of eligible studies in each language. 5 9 8 5 9 9 Limitations 6 0 0 Although, as described above, we adopted a search strategy that allowed us to identify eligible Although we identified 465 journals in 16 languages, we were only able to screen 326 of 6 1 2 them, as we prioritised journals ranked as "very relevant" and "relevant" for some to the search process. Therefore, we assessed whether our choice of journals screened in 6 1 5 each language was appropriate for identifying the most eligible studies in the language, by 6 1 6 examining the "rank-abundance" curve for each language, where the x axis of the curve highest % of eligible studies was given rank 1), and the y axis was the % of eligible studies. If a curve reached zero (i.e., there were almost no eligible studies) in lower-ranked journals, that language (see Fig. S1 for the result). Searches for some long-running journals only went back ten years from the latest volume, We tried to identify as many eligible studies as possible in each language, by making sure 6 3 0 that (i) every searcher was well qualified and trained before starting the searches (see Although we did our best to train searchers to fully understand the eligibility criteria (see increases in the number of eligible non-English-language studies in each journal (Fig. 1), as 6 4 9 the same journal was searched by a single searcher, nor (ii) the spatial and taxonomic We focused only on studies published in peer-reviewed academic journals and thus did not 6 5 6 consider the effects of publication bias, caused by ignoring grey literature, within each 6 5 7 language, while recognizing that important scientific knowledge may also be published in this study are limited to peer-reviewed studies published in academic journals. To compare study characteristics (i.e., study design, study location and study species) between 6 6 2 eligible English-and non-English-language studies, we used English-language studies stored coordinates-mean coordinates where a study had multiple sites, study design, scientific and Data S5) from the database on 11/12/2020. Again, here we defined a paper published in a 6 7 3 peer-reviewed journal as a study. The Conservation Evidence database also stores some 6 7 4 non-English-language peer-reviewed studies, most of which were identified incidentally by the 6 7 5 project. Those non-English-language studies were also incorporated into our dataset of 6 7 6 non-English-language studies, if they were in any of the 16 languages covered in this study (a 6 7 7 total of 74 non-English-language studies, see records with "Source" being "Ad hoc" in Data We first calculated the proportion of eligible studies for each non-English-language journal, by To test whether the number of eligible non-English-language studies had changed over time, Building a tool to overcome barriers in research-implementation spaces: The Conservation and addressing the prevalence and bias of study designs in the environmental and social  London, UK: The Royal Society; 2020. future generations to deliver evidence-based conservation and ecosystem management. Ecological Solutions and Evidence. 2021;2(1):e12032. doi: Co., Ltd.; 1922. interventions on bats: a protocol for subject-wide evidence synthesis. OSF Registries.  2018;558(7710):364-6. doi: 10.1038/d41586-018-05472-8.  Taxonomic information from around the web. R package version 0.9.98: https://github.com/ropensci/taxize; 2020.  brief introduction to mixed effects modelling and multi-model inference in ecology. PeerJ. Intelligence Agency: https://www.cia.gov/the-world-factbook/; 2021.      https://cran.r-project.org/web/packages/mcmcplots/index.html2018. This study, while providing important insights into the role of non-English-language sources in 1 0 5 1 developing a specific database, lacks assessments of (i) temporal changes in 1 0 5 2 non-English-language sources, and (ii) differences in study quality between languages. These  In contrast, our study revealed that the number of relevant non-English-language studies is 1 0 5 9 increasing in many languages, showing that non-English-language science will continue 1 0 6 0 playing a crucial role, while there seems to be a trade-off between evidence quality and  Before-After-Control-Impact, Control-Impact, Before-After, and After) in each study as the 1 0 6 8 response variable, and languages (16 non-English languages and "English -official" 1 0 6 9 (English-language studies conducted in countries where English is an official language), 1 0 7 0 compared to "English -others" (English-language studies conducted in the other countries) as   any English language studies (grid cells in black). The inset is a hexbin chart showing a