Exposure and connectedness to natural environments: An examination of the measurement invariance of the Nature Exposure Scale (NES) and Connectedness to Nature Scale (CNS) across 65 nations, 40 languages, gender identities, and age groups Viren Swami a,b,*, Mathew P. White c, Martin Voracek c, Ulrich S. Tran c, Toivo Aavik d, Hamed Abdollahpour Ranjbar e, Sulaiman Olanrewaju Adebayo f, Reza Afhami g, Oli Ahmed h, Annie Aimé i, Marwan Akel j,k, Hussam Al Halbusi l, George Alexias m, Khawla F. Ali n, Nursel Alp-Dal o, Anas B. Alsalhani p, Sara Álvarez-Solas q, Ana Carolina Soares Amaral r, Sonny Andrianto s, Trefor Aspden t, Marios Argyrides u, John Jamir Benzon R. Aruta v, Stephen Atkin n, Olusola Ayandele w,x, Migle Baceviciene y, Radvan Bahbouh z, Andrea Ballesio aa, David Barron ab, Ashleigh Bellard ac, SóleySesselja Bender ad, Kerime Derya Beydaǧ ae, Gorana Birovljević af, Marie-Ève Blackburn ag, Teresita Borja-Alvarez ah, Joanna Borowiec ai, Miroslava Bozogáňová aj,ak, Solfrid Bratland-Sanda al, MatthewH.E.M. Browning am, Anna Brytek-Matera an, Marina Burakova ao, Yeliz Çakır-Koçak ap, Pablo Camacho aq, Vittorio Emanuele Camilleri ar, Valentina Cazzato ac,as, Silvia Cerea at,au, Apitchaya Chaiwutikornwanich av, Trawin Chaleeraktrakoon aw, Tim Chambers ax, Qing-Wei Chen ay,az, Xin Chen ba, Chin-Lung Chien bb, Phatthanakit Chobthamkit aw, Bovornpot Choompunuch bc, Emilio J. Compte bd,be, Jennifer Corrigan bf, Getrude Cosmas bg, Richard G. Cowden bh, Kamila Czepczor-Bernat bi, Marcin Czub an, Wanderson Roberto da Silva bj, Mahboubeh Dadfar bk, Simon E. Dalley bl, Lionel Dany ao, Jesus Alfonso D. Datu bm, Pedro Henrique Berbert de Carvalho bn,bo, Gabriel Lins de Holanda Coelho bf, Avila Odia S. De Jesus v, Sonia Harzallah Debbabi bp, Sandesh Dhakal bq, Francesca Di Bernardo br, Donka D. Dimitrova bs,bt, Jacinthe Dion bu, Barnaby Dixson bv, Stacey M. Donofrio bl, Marius Drysch bw, Hongfei Du bx, Angel M. Dzhambov bt,by, Claire El-Jor bz, Violeta Enea ca, Mehmet Eskin cb, Farinaz Farbod cc, Lorleen Farrugia ar, Leonie Fian c, Maryanne L. Fisher cd, Michał Folwarczny ce, David A. Frederick cf, Matthew Fuller-Tyszkiewicz ax, Adrian Furnham cg, Antonio Alías García ch, Shulamit Geller ci, Marta Ghisi at,cj, Alireza Ghorbani ck, Maria Angeles Gomez Martinez cl, Sarah Gradidge a, Sylvie Graf cm, Caterina Grano aa, Gyöngyvér Gyene cn, Souheil Hallit co,cp, Motasem Hamdan cq, Jonathan E. Handelzalts ci,cr, PaulH.P. Hanel cs, Steven R. Hawks ct, Issa Hekmati cu, Mai Helmy cv,cw, Tetiana Hill cx, Farah Hina cy, Geraldine Holenweger cz, Martina Hřebíčková cm, Olasupo Augustine Ijabadeniyi da, Asma Imam cq, Başak İnce db, Natalia Irrazabal dc, * Corresponding author. School of Psychology and Sport Science, Anglia Ruskin University, Cambridge, Cambridgeshire, CB1 1PT, United Kingdom E-mail address: viren.swami@aru.ac.uk (V. Swami). Contents lists available at ScienceDirect Journal of Environmental Psychology journal homepage: www.elsevier.com/locate/jep https://doi.org/10.1016/j.jenvp.2024.102432 Received 20 May 2024; Received in revised form 9 August 2024; Accepted 13 September 2024 Journal of Environmental Psychology 99 (2024) 102432 Available online 16 September 2024 0272-4944/© 2024 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license ( http://creativecommons.org/licenses/by/4.0/ ). mailto:viren.swami@aru.ac.uk www.sciencedirect.com/science/journal/02724944 https://www.elsevier.com/locate/jep https://doi.org/10.1016/j.jenvp.2024.102432 https://doi.org/10.1016/j.jenvp.2024.102432 http://crossmark.crossref.org/dialog/?doi=10.1016/j.jenvp.2024.102432&domain=pdf http://creativecommons.org/licenses/by/4.0/ Rasa Jankauskiene y, Ding-Yu Jiang dd, Micaela Jiménez-Borja ah, Verónica Jiménez-Borja de, Evan M. Johnson cf, Veljko Jovanović df, Marija Jović dg, Marko Jović dh, Alessandra Costa Pereira Junqueira di,dj, Lisa-Marie Kahle dk, Adam Kantanista ai, Ahmet Karakiraz dl, Ayşe Nur Karkin cb, Erich Kasten dm, Salam Khatib dn, Nuannut Khieowan do, Patricia Joseph Kimong bg, Litza Kiropoulos dp, Joshua Knittel cz, Neena Kohli dq, Mirjam Koprivnik dr, Aituar Kospakov ds,dt, Magdalena Król-Zielińska ai, Isabel Krug dp, Garry Kuan du, Yee Cheng Kueh dv, Omar Kujan dw, Miljana Kukić af, Sanjay Kumar dx, Vipul Kumar dy, Nishtha Lamba dz, Mary Anne Lauri ar, Maria Fernanda Laus di,dj, Liza April LeBlanc cd, Hyejoo J. Lee ea, Małgorzata Lipowska eb, Mariusz Lipowski ec, Caterina Lombardo aa, Andrea Lukács ed, Christophe Maïano i,ee, Sadia Malik ef, Mandar Manjary dx,eg, Lidia Márquez Baldó eh, Martha Martinez-Banfi ei,ej, Karlijn Massar ek, Camilla Matera el, Olivia McAnirlin am, Moisés Roberto Mebarak em, Anwar Mechri en, Juliana Fernandes Filgueiras Meireles eo, Norbert Mesko ep, Jacqueline Mills ax, Maya Miyairi eq, Ritu Modi dq, Adriana Modrzejewska er, Justyna Modrzejewska es, Kate E. Mulgrew bv, Taryn A. Myers et, Hikari Namatame eu, Mohammad Zakaria Nassani ev, Amanda Nerini el, Félix Neto ew, Joana Neto ex, Angela Nogueira Neves ey, Siu-Kuen Ng ez, Devi Nithiya fa, Jiaqing O t, Sahar Obeid fb, Camila Oda-Montecinos fc, Peter Olamakinde Olapegba w, Tosin Tunrayo Olonisakin f, Salma Samir Omar fd, Brynja Örlygsdóttir ad, Emrah Özsoy dl, Tobias Otterbring fe, Sabine Pahl c, Maria Serena Panasiti aa,ff, Yonguk Park fg, Muhammad Mainuddin Patwary fh,fi, Tatiana Pethö aj, Nadezhda Petrova fj, Jakob Pietschnig fk, Sadaf Pourmahmoud g, Vishnunarayan Girishan Prabhu fl, Vita Poštuvan fm,fn, Pavol Prokop fo,fp, Virginia L. Ramseyer Winter fq, Magdalena Razmus fr, Taotao Ru ay,az, Mirjana Rupar cm, Reza N. Sahlan fs, Mohammad Salah Hassan ft, Anđela Šalov fu, Saphal Sapkota fv, Jacob Owusu Sarfo fw, Yoko Sawamiya et, Katrin Schaefer fx,fy, Michael Schulte-Mecklenbeck cz,fz, Veya Seekis ga, Kerim Selvi gb, Mehdi Sharifi gc, Anita Shrivastava dz, Rumana Ferdousi Siddique gd, Valdimar Sigurdsson ge, Vineta Silkane gf, Ana Šimunić fu, Govind Singh dq, Alena Slezáčková gg, Christine Sundgot-Borgen gh, Gill Ten Hoor ek, Passagorn Tevichapong gi, Arun Tipandjan gj, Jennifer Todd a,b, Constantinos Togas m, Fernando Tonini dc, Juan Camilo Tovar-Castro gk, Lise Katrine Jepsen Trangsrud al, Pankaj Tripathi dq, Otilia Tudorel gl, Tracy L. Tylka gm, Anar Uyzbayeva dt, Zahir Vally gn, Edmunds Vanags go, Luis Diego Vega gp, Aitor Vicente-Arruebarrena cl, Jose Vidal-Mollón eh, Roosevelt Vilar gq, Hyxia Villegas gp, Mona Vintilă gl, Christoph Wallner bw, Simon Whitebridge n, Sonja Windhager fx,fy, Kah Yan Wong gr, Eric Kenson Yau gs, Yuko Yamamiya gt, Victoria Wai Lan Yeung gs,gu, Marcelo Callegari Zanetti gv, Magdalena Zawisza a, Nadine Zeeni bz, Martina Zvaríková fo, Stefan Stieger gw a School of Psychology, Sport, and Sensory Sciences, Anglia Ruskin University, Cambridge, United Kingdom b Centre for Psychological Medicine, Perdana University, Kuala Lumpur, Malaysia c Department of Cognition, Emotion, and Methods in Psychology, Faculty of Psychology, University of Vienna, Vienna, Austria d Institute of Psychology, University of Tartu, Tartu, Estonia e Department of Psychology, Koç University, Istanbul, Turkiye f Department of Psychology and Behavioural Studies, Ekiti State University, Ado-Ekiti, Nigeria g Department of Art Studies, Tarbiat Modares University, Tehran, Iran h Department of Psychology, University of Chittagong, Chattogram, Bangladesh i Department of Psychoeducation and Psychology, Université Du Québec en Outaouais, Saint-Jérôme, Canada j INSPECT-LB: National Institute of Public Health, Clinical Epidemiology, and Toxicology, Beirut, Lebanon k School of Pharmacy, Lebanese International University, Beirut, Lebanon l Department of Management, Ahmed Bin Mohammad Military College, Doha, Qatar m Faculty of Psychology, Panteion University of Social and Political Sciences, Athens, Greece n Royal College of Surgeons Ireland-Bahrain, Adliya, Bahrain o Department of Midwifery, Faculty of Health Sciences, Munzur University, Tunceli, Turkiye p Department of Oral Medicine and Diagnostic Sciences, Vision College of Dentistry and Nursing, Vision Colleges, Riyadh, Saudi Arabia q Facultad de Ciencias de La Vida, Universidad Regional Amazónica Ikiam, Muyuna, Ecuador r Federal Institute of Education, Science, and Technology of Southeast of Minas Gerais, Barbacena, Brazil s Department of Psychology, Universitas Islam Indonesia, Yogyakarta, Indonesia t Department of Psychology, Aberystwyth University, Ceredigion, United Kingdom V. Swami et al. Journal of Environmental Psychology 99 (2024) 102432 2 u Department of Psychology, Neapolis University Pafos, Paphos, Cyprus v Department of Psychology, De La Salle University, Philippines w Department of General Studies, The Polytechnic, Ibadan, Nigeria x Department of Psychology, University of Ibadan, Ibadan, Nigeria y Health Research and Innovation Science Centre, Faculty of Health Sciences, Klaipeda University, Klaipeda, Lithuania z Department of Psychology, Faculty of Arts, Charles University, Prague, Czech Republic aa Department of Psychology, Sapienza University of Rome, Rome, Italy ab School of Social Sciences, Heriot-Watt University Malaysia, Putrajaya, Malaysia ac School of Psychology, Faculty of Health, Liverpool John Moores University, Liverpool, United Kingdom ad Faculty of Nursing and Midwifery, University of Iceland, Reykjavik, Iceland ae Department of Nursing, Faculty of Health Sciences, Yalova, University, Yalova, Turkiye af Department of Psychology, Faculty of Humanities and Social Sciences, University of Rijeka, Rijeka, Croatia ag ÉCOBES-Research and Transfer, Cégep de Jonquière, Québec, Canada ah Colegio de Ciencias Sociales y Humanidades, Universidad San Francisco de Quito USFQ, Quito, Ecuador ai Department of Physical Education and Lifelong Sports, Poznań University of Physical Education, Poznań, Poland aj Institute of Social Sciences of the Centre of Social and Psychological Sciences, Košice, Slovakia ak Faculty of Humanities and Natural Sciences, Institute of Pedagogy, Andragogy, and Psychology, University of Prěsov, Prěsov, Slovakia al Department of Sports, Physical Education and Outdoor Studies, University of South-Eastern Norway, Bø in Telemark, Norway am Department of Parks, Recreation and Tourism Management, Clemson University, Clemson, SC, USA an Institute of Psychology, University of Wrocław, Wrocław, Poland ao Laboratory of Social Psychology, Aix-Marseille University, Aix-en-Provence, France ap Department of Midwifery, Faculty of Health Sciences, Bartın University, Bartın, Turkiye aq CentroSan Isidoro University Center, Seville, Spain ar Department of Psychology, University of Malta, Msida, Malta as Department of Cognitive Sciences, Psychology, Education, and Cultural Studies, University of Messina, Messina, Italy at Department of General Psychology, University of Padova, Padova, Italy au Department of Biomedical Sciences, University of Padova, Padova, Italy av Faculty of Psychology, Chulalongkorn University, Bangkok, Thailand aw Department of Psychology, Faculty of Liberal Arts, Thammasat University, Pathumthani, Thailand ax School of Psychology, Deakin University, Geelong, Australia ay Lab of Light and Physio-Psychological Health, National Center for International Research on Green Optoelectrics, South China Normal University, Guangzhou, China az Guangdong Provincial Key Laboratory of Optical Information Materials and Technology & Institute of Electronic Paper Displays, South China Academy of Advanced Optoelectronics, South China Normal University, Guangzhou, China ba Department of Psychology, Graduate School of Arts and Science, New York University, New York, USA bb Department of Psychology, Soochow University, Taipei, Taiwan bc Department of Educational Psychology and Guidance, Faculty of Education, Mahasarakham University, Maha Sarakham, Thailand bd School of Psychology, Universidad Adolfo Ibáñez, Penalolen, Chile be Comenzar de Nuevo Treatment Center, Monterrey, Mexico bf School of Applied Psychology, University College Cork, Cork, Ireland bg Faculty of Psychology and Education, Universiti Malaysia Sabah, Kota Kinabalu, Malaysia bh Human Flourishing Program, Harvard University, Cambridge, USA bi Department of Pediatrics, Pediatric Obesity and Metabolic Bone Diseases, Faculty of Medical Sciences in Katowice, Medical University of Silesia, Katowice, Poland bj Graduate Program in Food, Nutrition, and Food Engineering, São Paulo State University, São Paulo, Brazil bk Department of Addiction, School of Behavioral Sciences and Mental Health (Tehran Institute of Psychiatry), Iran University of Medical Sciences, Tehran, Iran bl Department of Psychology, University of Groningen, Groningen, the Netherlands bm Academic Unit of Human Communication, Learning, and Development, Faculty of Education. The University of Hong Kong, China bn Body Image and Eating Disorders Research Group, Federal University of Juiz de Fora, Juiz de Fora, Brazil bo Institute of Psychiatry, University of São Paulo, São Paulo, Brazil bp Faculty of Medicine, University of Sousse, Sousse, Tunisia bq Central Department of Psychology, Tribhuvan University, Kathmandu, Nepal br Department of Psychology, University of Campania Luigi Vanvitelli, Caserta, Italy bs Department of Health Management and Healthcare Economics, Faculty of Public Health, Medical University of Plovdiv, Plovdiv, Bulgaria bt Environmental Health Division, Research Institute at Medical University of Plovdiv, Medical University of Plovdiv, Plovdiv, Bulgaria bu Department of Psychology, Université Du Québec à Trois-Rivières, Trois-Rivières, Canada bv School of Health, University of the Sunshine Coast, Moreton Bay, Australia bw Department of Plastic and Hand Surgery, BG University Hospital Bergmannsheil Bochum, Ruhr University Bochum, Bochum, Germany bx Department of Psychology, Beijing Normal University at Zhuhai, Zhuhai, China by Health and Quality of Life in a Green and Sustainable Environment Research Group, Strategic Research and Innovation Program, Medical University of Plovdiv, Plovdiv, Bulgaria bz Department of Natural Sciences, School of Arts and Sciences, Lebanese American University, Beirut, Lebanon ca Department of Psychology, Alexandru Ioan Cuza University, Iași, Romania cb Department of Psychology, Koc University, Istanbul, Turkiye cc Department of Textile and Fashion Design, Alzahra University, Tehran, Iran cd Department of Psychology, Saint Mary’s University, Halifax, Canada ce Discipline of Marketing, J.E. Cairnes School of Business & Economics, University of Galway, Galway, Ireland cf Crean College of Health and Behavioral Sciences, Chapman University, Orange, USA cg Department of Leadership and Organizational Behaviour, Norwegian Business School, Oslo, Norway ch Department of Education, University of Almería, Almería, Spain ci School of Behavioral Sciences, The Academic College of Tel Aviv-Yaffo, Tel Aviv-Yafo, Israel cj Unità Operativa Complessa (UOC) Hospital Psychology, Padova University Hospital, Padova, Italy ck Department of Social Sciences, Payam Noor University, Tehran, Iran cl Faculty of Psychology, Pontifical University of Salamanca, Salamanca, Spain cm Institute of Psychology, Czech Academy of Sciences, Brno, Czech Republic cn Doctoral School of Psychology, Eötvös Loránd University, Budapest, Hungary co School of Medicine and Medical Sciences, Holy Spirit University of Kaslik, Jounieh, Lebanon cp Applied Science Research Center, Applied Science Private University, Amman, Jordan cq Faculty of Public Health, Al-Quds University, East Jerusalem, Palestine cr Department of Psychiatry, University of Michigan, Ann Arbor, USA cs Department of Psychology, University of Essex, Colchester, United Kingdom ct Department of Kinesiology and Health Science, Utah State University, Logan, USA V. Swami et al. Journal of Environmental Psychology 99 (2024) 102432 3 cu Department of Psychology, Faculty of Human Science, University of Maragheh, Maragheh, Iran cv Department of Psychology, College of Education, Sultan Qaboos University, Muscat, Oman cw Department of Psychology, Faculty of Arts, Menoufia University, Shebin El Kom, Egypt cx Hertfordshire Business School, University of Hertfordshire, Hatfield, United Kingdom cy Department of Psychiatry, University of Cambridge, Cambridge, United Kingdom cz Department of Consumer Behavior, University of Bern, Bern, Switzerland da Department of Sociology and Social Justice, Afe Babalola University, Ado-Ekiti, Nigeria db Centre for Research in Eating and Weight Disorders, Department of Psychological Medicine, Institute of Psychiatry, Psychology, and Neuroscience, King’s College London, London, United Kingdom dc Faculty of Social Sciences, University of Palermo, Buenos Aires, Argentina dd Department of Psychology, National Chung Cheng University, Chia Yi, Taiwan de Colegio de Comunicación y Artes Contemporáneas, Universidad San Francisco de Quito USFQ, Quito, Ecuador df Department of Psychology, Faculty of Philosophy, University of Novi Sad, Novi Sad, Serbia dg Faculty of Organizational Sciences, University of Belgrade, Belgrade, Serbia dh Faculty of Medicine, University of Belgrade, Belgrade, Serbia di Department of Psychology, University of São Paulo, Ribeirão Preto, Brazil dj Department of Nutrition, University of Ribeirão Preto, Ribeirão Preto, Brazil dk Faculty of Life Sciences, Medical School Hamburg, Hamburg, Germany dl Sakarya Business School, Sakarya University, Sakarya, Turkiye dm Practice for Psychotherapy, Am Krautacker 25, Travemünde, Germany dn Faculty of Health Professions, Al-Quds University, East Jerusalem, Palestine do Asian Studies Department, Faculty of International Studies, Prince of Songkla University, Phuket Campus, Phuket, Thailand dp School of Psychological Sciences, University of Melbourne, Melbourne, Australia dq Department of Psychology, University of Allahabad, Prayagraj, India dr Institute of Anton Martin Slomsek, Primary School Montessori, Maribor, Slovenia ds Department of Sociology and Social Work, Al-Farabi Kazakh National University, Almaty, Kazakhstan dt Department of General Education Disciplines, Astana IT University, Astana, Kazakhstan du Exercise and Sport Sciences Programme, School of Health Sciences, Universiti Sains Malaysia, Kubang Kerian, Malaysia dv Biostatistics and Research Methodology Unit, School of Medical Sciences, Universiti Sains Malaysia, Kubang Kerian, Malaysia dw Oral Diagnostics and Surgical Sciences, UWA Dental School, The University of Western Australia, Nedlands, Australia dx Department of Psychology, D.A.V. College, Muzaffarnagar, India dy Department of Psychology, Kashi Naresh Government Post-Graduate College, Gyanpur, India dz Department of Psychology, Middlesex University Dubai, Dubai, United Arab Emirates ea Department of Counselling Psychology and Social Welfare, Handong Global University, Pohang, South Korea eb Institute of Psychology, University of Gdańsk, Gdańsk, Poland ec Faculty of Social and Humanities, University WSB Merito, Gdansk, Poland ed Faculty of Health Sciences, University of Miskolc, Miskolc, Hungary ee Cyberpsychology Laboratory, Department of Psychoeducation and Psychology, Université Du Québec en Outaouais (UQO), Saint-Jérôme, Canada ef Department of Psychology, University of Sargodha, Sargodha, Pakistan eg M.M.D. Public School, Brahmpuri, Muzaffarnagar, India eh Department of Research Methods and Diagnosis in Education, University of València, València, Spain ei Faculty of Legal and Social Sciences, Simón Bolívar University, Barranquilla, Colombia ej Life Science Research Centre, Simón Bolívar University, Barranquilla, Colombia ek Department of Work and Social Psychology, Maastricht University, Maastricht, the Netherlands el Department of Education, Languages, Intercultures, Literatures, and Psychology, University of Florence, Florence, Italy em Department of Psychology, Universidad Del Norte, Barranquilla, Colombia en Faculty of Medicine of Monastir, Eya Medical Centre, Monastir, Tunisia eo Department of Family and Community Medicine, School of Community Medicine, University of Oklahoma, Tulsa, USA ep Institute of Psychology, University of Pécs, Pécs, Hungary eq Department of Health Sciences, DePaul University, Chicago, USA er Department of Medical Anthropology, Faculty of Medical Sciences in Katowice, Medical University of Silesia, Katowice, Poland es Institute of Pedagogy, University of Bielsko-Biala, Bielsko-Biala, Poland et Department of Psychology, Virginia Wesleyan University, Virginia Beach, USA eu Faculty of Human Sciences, University of Tsukuba, Tsukuba, Japan ev Department of Restorative and Prosthetic Dental Sciences, College of Dentistry, Dar Al Uloom University, Riyadh, Saudi Arabia ew Faculty of Psychology and Educational Sciences, University of Porto, Porto, Portugal ex REMIT – Research on Economics, Management and Information Technologies, Universidade Portucalense, Porto, Portugal ey Division of Research, Physical Education College of the Brazilian Army, Rio de Janeiro, Brazil ez Physical Education Unit, Chinese University of Hong Kong, China fa Department of Physiology, Mahatma Gandhi Medical College and Research Institute, Puducherry, India fb Social and Education Sciences Department, School of Arts and Sciences, Lebanese American University, Jbeil, Lebanon fc Institute of Social Sciences, Universidad de O’Higgins, Rancagua, Chile fd Department of Dermatology, Venereology, and Andrology, Alexandria University, Alexandria, Egypt fe Department of Management, University of Agder, Kristiansand, Norway ff Santa Lucia Foundation, Scientific Institute for Research and Healthcare, Rome, Italy fg Department of Psychology, Kyungnam University, Changwon, South Korea fh Environment and Sustainability Research Initiative, Khulna, Bangladesh fi Environmental Science Discipline, Life Science School, Khulna University, Khulna, Bangladesh fj Department of Human Anatomy, Histology and Embryology, Faculty of Medicine, Medical University of Plovdiv, Plovdiv, Bulgaria fk Department of Developmental and Educational Psychology, Faculty of Psychology, University of Vienna, Vienna, Austria fl Industrial and Systems Engineering, University of North Carolina, Charlotte, USA fm Slovene Centre for Suicide Research, Andrej Marusic Institute, University of Primorska, Koper, Slovenia fn Department of Psychology FAMNIT, University of Primorska, Koper, Slovenia fo Department of Environmental Ecology and Landscape Management, Faculty of Natural Sciences, Comenius University in Bratislava, Bratislava, Slovakia fp Institute of Zoology, Slovak Academy of Sciences, Bratislava, Slovakia fq School of Social Work, University of Missouri, Columbia, USA fr Institute of Psychology, Marie Curie-Skłodowska University, Lublin, Poland fs Department of Counseling, School, and Educational Psychology, Graduate School of Education, University at Buffalo-SUNY, Buffalo, USA ft Arabic Program Department, Modern College of Business and Science, Muscat, Oman fu Department of Psychology, University of Zadar, Zadar, Croatia V. Swami et al. Journal of Environmental Psychology 99 (2024) 102432 4 fv KOSHISH-National Mental Health Self-Help Organization, Kusunti, Lalitpur, Nepal fw Department of Health, Physical Education and Recreation, University of Cape Coast, Cape Coast, Ghana fx Department of Evolutionary Anthropology, University of Vienna, Vienna, Austria fy Human Evolution and Archaeological Sciences, University of Vienna, Vienna, Austria fz Max Planck Institute for Human Development, Berlin, Germany ga School of Applied Psychology, Griffith University, Gold Coast, Australia gb Department of Psychology, Eskişehir Osmangazi University, Eskişehir, Turkiye gc Department of Psychology, Islamic Azad University, Bandar Gaz, Iran gd Department of Psychology, University of Dhaka, Dhaka, Bangladesh ge Department of Business Administration, Reykjavik University, Reykjavik, Iceland gf Faculty of Social Sciences, Vidzeme University of Applied Sciences, Valmiera, Latvia gg Department of Medical Psychology and Psychosomatics, Faculty of Medicine, Masaryk University, Brno, Czech Republic gh Regional Department for Eating Disorders, Oslo University Hospital, Oslo, Norway gi Department of Psychology, Faculty of Humanities, Chiang Mai University, Chiang Mai, Thailand gj International Centre for Psychological Counselling and Social Research, Puducherry, India gk Department of Psychology, University of the Andes, Bogotá, Colombia gl Department of Psychology, West University of Timişoara, Timişoara, Romania gm Department of Psychology, The Ohio State University, Columbus, USA gn Department of Clinical Psychology, United Arab Emirates University, Al Ain, United Arab Emirates go Faculty of Education, Psychology, and Art, University of Latvia, R̄ıga, Latvia gp Vice-rectory for Teaching, Research, and Extension, Universida Latina de Costa Rica, San José, Costa Rica gq Department of Psychology, Massey University, Auckland, New Zealand gr School of Psychology, University of Nottingham Malaysia, Semenyih, Malaysia gs Department of Psychology, Lingnan University, China gt Department of Undergraduate Studies, Temple University, Japan Campus, Tokyo, Japan gu Wofoo Joseph Lee Consulting and Counselling Psychology Research Centre, Lingnan University, China gv Department of Physical Education, São Judas Tadeu University, São Paulo, Brazil gw Department of Psychology and Psychodynamics, Karl Landsteiner University of Health Sciences, Krems an der Donau, Austria A R T I C L E I N F O Handling Editor: L. McCunn Keywords: Nature exposure scale Connectedness to nature scale Measurement invariance Cross-cultural Multi-group confirmatory factor analysis (MG- CFA) A B S T R A C T Detachment from nature is contributing to the environmental crisis and reversing this trend requires detailed monitoring and targeted interventions to reconnect people to nature. Most tools measuring nature exposure and attachment were developed in high-income countries and little is known about their robustness across national and linguistic groups. Therefore, we used data from the Body Image in Nature Survey to assess measurement invariance of the Nature Exposure Scale (NES) and the Connectedness to Nature Scale (CNS) across 65 nations, 40 languages, gender identities, and age groups (N = 56,968). While multi-group confirmatory factor analysis (MG-CFA) of the NES supported full scalar invariance across gender identities and age groups, only partial scalar invariance was supported across national and linguistic groups. MG-CFA of the CNS also supported full scalar invariance across gender identities and age groups, but only partial scalar invariance of a 7-item version of the CNS across national and linguistic groups. Nation-level associations between NES and CNS scores were negli- gible, likely reflecting a lack of conceptual clarity over what the NES is measuring. Individual-level associations between both measures and sociodemographic variables were weak. Findings suggest that the CNS-7 may be a useful tool to measure nature connectedness globally, but measures other than the NES may be needed to capture nature exposure cross-culturally. 1. Introduction At the turn of the century, the United Nations Millennium Declara- tion recognised that insufficient respect for nature is a fundamental challenge for international relations and global sustainable development (United Nations General Assembly, 2000). Despite ongoing efforts, anthropogenic-related climate change, biodiversity loss, and land, water, and air pollution are accelerating (e.g., Goudie, 2019). Some observers have suggested that this is, at least in part, a consequence of a growing detachment from the natural world, especially among increasingly urbanised populations (Beery et al., 2023; Soga & Gaston, 2016). Reversing these trends requires an understanding of the drivers and barriers of ecological and pro-environmental behaviours, and using this knowledge to promote widespread behaviour change (Grilli & Curtis, 2021; Schultz & Kaiser, 2012). One important avenue of research concerns people’s physical contact with, and psychological connectedness to, the natural world. In terms of the former, the relationship between recreational nature exposure (e.g., leisure visits to parks, woodlands, and beaches) and a range of public and private pro-environmental behaviours has been shown to be consistently positive (De Ville et al., 2021; Martin et al., 2020). Regarding psychological connectedness, robust evidence shows that people who feel more connected to the natural world – independent of their actual exposure – have more positive attitudes towards the natural environment, ecological behaviours, and nature protection (for meta-analyses, see Barragan-Jason et al., 2022; Whitburn et al., 2020). They also exhibit better well-being and mental health (Capaldi et al., 2014; White et al., 2021). The mechanisms underpinning these relationships are thought to reflect a combination of genetic inheritance (Kellert & Wilson, 1993), personal experience and associative learning (e.g., Yannick & de Block, 2011), and salient sociocultural norms (Bourassa, 1990). In addition, nature exposure and nature connectedness are likely to be mutually reinforcing. Positive contact with the natural world can increase feelings of connectedness to nature (e.g., Fränkel et al., 2019; Lengieza & Swim, 2021; Martin et al., 2020; Swami, Barron, et al., 2020), with a meta-analysis of experimental manipulations and field interventions reporting a moderate positive mean effect of nature contact on nature connectedness (g = .44, 95% CI = .31, .58; Sheffield et al., 2022). Conversely, greater nature connectedness can encourage people to seek more nature exposure (Martin et al., 2020; Stehl et al., 2024). Positive experiences may then strengthen nature connectedness over time. Although improving our understanding of these processes remains important, our objective here was to consider how physical contact and psychological connectedness with nature are measured and, in partic- ular, how generalisable existing measures are across linguistic and V. Swami et al. Journal of Environmental Psychology 99 (2024) 102432 5 national contexts. Given that the environmental crisis is a global phe- nomenon (Goudie, 2019), adequate measurement and monitoring across contexts requires instruments that can be deployed reliably in multiple settings to make robust comparisons. Crucial to this issue is the concept of measurement invariance, the notion that a measurement tool should measure the same underlying construct in the same way across different groups (Swami & Barron, 2019; Vandenberg & Lance, 2000), which in turn ensures that measurement biases leading to artefactual, inaccurate, or irreplicable results are avoided (Fischer et al., 2023). Measurement invariance can be determined at different levels, with scalar or partial scalar invariance typically considered a minimum threshold for comparison of latent means (Chen, 2007). To date, however, determining measurement invariance of key na- ture exposure and nature connectedness instruments has been hampered because most research – including the development of instruments to measure these constructs – has been conducted in a small handful of countries in the Global North (Tirri et al., 2021; Zhang et al., 2020). Here, we aimed to address this shortcoming. Specifically, using data from 65 countries, we aimed to explore the measurement invariance of two well-known instruments across multiple linguistic and national contexts. The first was a measure of nature exposure – the Nature Exposure Scale (NES; Kamitsis & Francis, 2013) – while the second was a measure of nature connectedness, the Connectedness to Nature Scale (CNS; Mayer & Frantz, 2004). We briefly review what is currently known about these instruments. 1.1. Nature exposure Many definitions of nature exposure, or nature contact, exist (Holland et al., 2021). Much of the research linking nature exposure to health and well-being uses remote sensing data to estimate the per- centage of vegetation around the home using various radial buffers (e.g., Browning & Lee, 2017) or distance to local green and blue spaces (e.g., Geary et al., 2023). Others consider vegetation around other core lo- cations, such as work/school (e.g., Dadvand et al., 2015), or explore more deliberative, intentional nature contact, such as leisure visits to natural settings (e.g., Garrett et al., 2023). Finally, many of the benefi- cial effects of nature contact on health and pro-environmentalism may depend on psychological awareness of exposure, or a certain degree of mindfulness of this contact with nature (e.g., Macaulay et al., 2022; Richardson, Hamlin, Butler, et al., 2022). One self-report measure that attempts to address all three aspects – that is, everyday nature around the home/work, recreational visits, and nature awareness – is the 4-item Nature Exposure Scale (NES, Kamitsis & Francis, 2013; see Appendix 1 for items). Although the instrument has been utilised in diverse national groups (e.g., Arroz et al., 2022; Bace- viciene et al., 2021; Picanço et al., 2024; Stieger et al., 2022), its factorial validity has been infrequently assessed. Studies with adults from the United States (Swami et al., 2016), Portugal (Arroz et al., 2022; Picanço et al., 2024), and Lithuania (Matukyniene et al., 2021) suggest scores are unidimensional, whereas a study with an online sample (na- tionality unreported) found that it was necessary to drop one item (Item #1) to achieve unidimensionality (Wood et al., 2019). The equivocal nature of findings vis-à-vis the factorial validity of the NES may reflect the fact that a single instrument is trying to measure very different types of exposure using different response outcomes. Thus, while some authors have suggested that overall scores on the 4- item NES demonstrate adequate indices of face validity (e.g., Picanço et al., 2024; Swami et al., 2019), others have implied that it is only the two items that assess direct contact with nature that truly assesses nature exposure (Goh et al., 2023). As scholars increasingly seek brief self-report measures of nature exposure, a fuller understanding of the factorial validity of the NES, including item behaviour, is vital (Swami, 2024). In the same vein, more can be done to understand the psycho- metric properties of this instrument beyond singular national groups, including in terms of gendered identities, age groups, and languages. This is particularly important as some work has suggested that de- mographic factors affect responding on the NES (Picanço et al., 2024). 1.2. Connectedness to nature Although many measures of nature connectedness exist (Martin & Czellar, 2016; Richardson et al., 2019), one of the most widely used is the Connectedness to Nature Scale (CNS; Mayer & Frantz, 2004; cited more than 3500 times based on Google Scholar citations up to August 2024). In the original study reporting on the development of the 14-item CNS (see Appendix 1 for items), Mayer and Frantz (2004) reported – based on exploratory factor analyses (EFAs) – that scores were unidi- mensional in two community and three college samples from the United States. The unidimensional model of the 14-item CNS scores has also been supported in other national and linguistic contexts, such as Brazil (Pessoa et al., 2016), China (Li & Wu, 2016), Italy (Lovati et al., 2023), and Spain (Mattas-Terrón & Elósegui-Bandera, 2012). However, not all studies have demonstrated adequate fit of the unidimensional model of CNS scores, and in some national contexts, a unidimensional model was only supported once several items were eliminated: one item in Spain (Olivos et al., 2013), three items in France (Navarro et al., 2017) and Australia (Pearce et al., 2022), four each in Kenya (Marczak & Sorokowski, 2018) and South Korea (Gim et al., 2019), and seven items in Poland (Strzelecka et al., 2023). Likewise, Anđić and Šuperina (2021) reported difficulties translating four CNS items into Croatian, resulting in a 10-item, single factor, instrument. In Turkey, the CNS reduced to two dimensions reflecting integration with nature (two items) and feeling part of nature (six items; Bektaş et al., 2017). These equivocal findings may reflect the fact items on the CNS contain two verbal structures: items that include the word “feel” as an emotional component and other items that more closely reflect a cognitive belief in one’s connection to nature (Lee & Oh, 2021; Perrin & Benassi, 2009). Alternatively, it is possible that some CNS items either function poorly or are redundant in some national contexts. Based on Item Response Theory, Pasca et al. (2017) suggested that seven CNS items (Items #1, 3, 4, 8, 12, 13, and 14) were either redundant or lacked adequate fit in a sample of Spanish adults. In subsequent analyses, the same authors also suggested that a truncated, 7-item version of the CNS (i.e., the CNS-7) had adequate composite reliability, although factorial validity was not assessed. In a more recent study with Brazilian uni- versity students, Rosa et al. (2022) reported that scores on the CNS-7 were unidimensional and that the CNS-7 had slightly improved fit indices compared to the full version. To date, however, assessments of the measurement invariance of the CNS across national groups in the same study, conducted at the same time, remain rare. One study using samples from seven nations (Spain, the Netherlands, Turkey, Portugal, Germany, France, and Hungary) utilising the CNS-7 supported metric, but not scalar invariance, once the loading associated with Item #7 was relaxed (Navarro et al., 2022). In terms of the 14-item CNS, Pasca et al. (2018) examined measurement invariance across samples from Spain and the United States (the latter representing data from Mayer & Frantz, 2004). Their analyses indicated support for configural, but not metric, invariance. Based on Item Response Theory, Pasca et al. (2018) further noted that seven of the CNS items showed differential functioning across groups. These studies suggest that the latent connectedness to nature construct is not equiv- alent across national groups (Navarro et al., 2022). Beyond invariance across national groups, very little work has assessed invariance of the CNS across other sociodemographic charac- teristics. For instance, only two studies have examined the measurement invariance of the CNS across gender identities. In samples of Italian adults, Di Fabio and Rosen (2019) and Lovati et al. (2023), respectively, reported that the CNS achieved full scalar invariance across women and men. However, it remains possible that gendered experiences – partic- ularly across national or cultural groups – shape one’s understanding V. Swami et al. Journal of Environmental Psychology 99 (2024) 102432 6 and manifestations of connectedness to nature. Drawing on gender socialisation theories, for instance, McCright (2010) suggested that women, more so than men, are expected to demonstrate an ethic of care for the natural environment and exhibit both greater environmental concern (e.g., Xiao & McCright, 2015) and dispositional empathy with nature (Tam, 2013). As such, there is a need to more carefully and comprehensively assess the extent to which the CNS is invariant across gender identities in multiple cultural contexts. Likewise, we are unaware of any previous work that has examined invariance of the CNS across age groups. Existing studies have reported equivocal results in terms of the associ- ation between CNS scores and age, with some studies reporting that CNS scores increase with older age (Swami et al., 2016) and other studies reporting no significant association (e.g., Swami et al., 2016). Other work has suggested that connectedness to nature dips in adolescence before returning to pre-adolescent levels in early adulthood (Richardson et al., 2019), where it then remains relatively stable (Anderson & Krettenauer, 2021). More work is needed to explore these patterns across different cultural contexts. 1.3. The present study Large, multinational studies offer the best opportunity to deal with many of the issues noted above, particularly given that research on nature exposure and connectedness to nature often centres the experi- ences of respondents in the Global North (Barragan-Jason et al., 2023; Soga & Gaston, 2023). Thus, in the present study, we utilised data from the Body Image in Nature Survey (BINS; Swami, Tran et al., 2022), a collaborative, 253 researcher-crowdsourced project that gathered CNS and NES data between 2020 and 2022 from participants in 65 nations across 40 language groups with variance across gender identities and age groups. In terms of the NES, we considered whether a unidimen- sional model with all four items, as well as multidimensional models, would offer optimal fit. Given that there are few assessments of the factorial validity of this instrument and the limits of cross-sectional data for establishing the dimensionality of measures, we do not advance any specific hypotheses here. In terms of the CNS, we adopted an exploratory framework, considering the extent to which either the full 14-item CNS, or the truncated CNS-7, would balance item retention and measurement invariance across groups. As a preliminary hypothesis, we expected that the CNS-7 would demonstrate superior fit compared to the full CNS and would also evidence scalar or partial scalar invariance across groups. A second objective was to assess whether, and the extent to which, nature exposure is significantly associated with connectedness to nature across nations. Our expectation, based on previous work, was of a small, positive correlation in the r ~ .30 range (Sheffield et al., 2022; Swami et al., 2016; Swami, Barron, et al., 2020). Finally, we also assessed the extent to which sociodemographic variables included in the BINS (i.e., financial security, urbanicity, educational qualifications, marital status, and racialised status) were associated with both nature exposure and connectedness to nature. Although this aspect of our study was more exploratory, based on the available evidence, we expected that greater nature exposure and connectedness to nature would be significantly associated with greater financial security, rural residence (e.g., Carrus et al., 2020; Richardson et al., 2019), higher educational qualifications (e.g., Nesbitt et al., 2019), being married/in a committed relationship (Pasanen et al., 2023), and racialised majority status (e.g., Murdock, 2019). 2. Materials and methods 2.1. Overview of the Body Image in Nature Survey Full details of the Body Image in Nature Survey (BINS) are published elsewhere (Swami, Tran et al., 2022). Data were collected between November 2020 and February 2022 with community sampling, with the majority of recruitment taking place online. The overall project received ethics approval from the School Research Ethics Panel at the first au- thor’s institution (approval code: PSY-S19-015) and, unless exempt by national laws, all collaborating teams additionally obtained ethics approval from local institutional ethics committees or review boards. A list of nations, associated sample sizes, data collection methods, ethics approvals, and survey languages is presented in Supplementary Table S1. Sample sizes ranged from 204 in the United Arab Emirates (Arabic) to 3275 in Thailand. 2.2. Participants The BINS dataset consists of 56,968 respondents, of whom 58.9% were women, 40.5% were men, and 0.6% reported another gender identity. Ages ranged from 18 to 99 years (M = 33.10, SD = 13.79). In terms of financial security compared to others of their age in their country, 25.5% and 49.6% felt more or equally secure, respectively, with 24.9% feeling less secure. Most (84.5%) lived in an urban rather than a rural (15.5%) area, and the majority reported at least completing secondary education (72.6%). In total, 53.0% were in a committed relationship including marriage. The majority (74.2%) self-identified as being part of a racialised majority in their country, whereas 11.3% identified as part of a racialised minority group (13.5% were uncertain and race data were not collected in France due to prohibition of the collection and storage of race-related data). Table 1 presents detailed sample description data for all individual nations. In six countries, data was collected in either two (Canada, Iceland, India, the Philippines, the United Arab Emirates [UAE]) or three (China) languages. 2.3. Measures 2.3.1. Nature exposure The 4-item self-reported Nature Exposure Scale (NES; Kamitsis & Francis, 2013) covers perceptions about everyday nature exposure, frequency of more distal visits (“nature exposure”; Items #1 and #3), and attention paid to nature in both settings (“noticing nature”; Items #2 and #4; see Appendix 1 for English wording). Response anchors varied depending on the item, but all used 5-point scales. The NES was translated for use in the present project using the back-translation pro- cedure (Brislin, 1986; for further information, see Swami, Tran et al., 2022) unless it was presented in English or a validated, localised version was available for use. A list of the 40 languages in which the BINS survey package was presented is reported in Supplementary Table S1 and all translations are available from the first author. 2.3.2. Connectedness to nature The 14-item Connectedness to Nature Scale (CNS; Mayer & Frantz, 2004; items in English are presented in Appendix 1) uses a 5-point response scale: 1 (strongly disagree) to 5 (strongly agree). Unless pre- sented in English, or where a previously validated translation was not available, the CNS was also translated for use in the BINS using the parallel back-translation procedure (see Supplementary Table S1). 2.3.3. Urbanicity To assess urbanicity, participants were asked about their current place of residence, with response options adapted from Pedersen and Mortensen (2001) as follows: capital city, capital city suburbs, provincial city (more than 100,000 residents), provincial town (more than 10,000 residents), and rural areas. Response options were assigned values 1 to 5 (in the above order) for statistical analysis and collapsed into urban versus rural for descriptive purposes. This measure of urbanicity has been used in previous cross-national work (Swami et al., 2020). 2.3.4. Financial security Following previous cross-national work (Swami et al., 2012, 2020), participants were asked to self-report how financially secure they felt V. Swami et al. Journal of Environmental Psychology 99 (2024) 102432 7 Table 1 Sample Descriptions of Data from the Body Image in Nature Survey (BINS). Nation Sample size Mean age (SD) % Women Mean financial security (SD) %Urban residence %Secondary/tertiary education %In committed relationship or married %Racialised minority Argentina 670 35.36 (13.6) 57 2.13 (.7) 98 81 50 9 Australia 1038 35.23 (13.1) 71 1.90 (.8) 93 77 55 18 Austria 1279 41.99 (16.5) 54 2.08 (.7) 67 62 63 9 Bahrain 441 30.47 (9.8) 74 1.98 (.6) 98 87 51 8 Bangladesh 460 29.30 (8.6) 42 1.78 (.8) 88 80 51 13 Bosnia & Herzegovina 406 43.93 (10.9) 64 2.15 (.7) 87 90 70 16 Brazil 1462 36.77 (12.0) 58 2.21 (.7) 99 86 66 12 Bulgaria 248 33.52 (14.1) 62 2.16 (.6) 92 54 52 4 Canada (English) 336 24.61 (10.0) 83 2.10 (.7) 82 36 48 14 Canada (French) 806 38.22 (12.8) 88 2.29 (.7) 78 95 72 7 Chile 422 36.14 (13.6) 79 2.28 (.8) 94 73 41 8 China (Cantonese) 409 20.50 (5.9) 58 2.18 (.7) 100 96 2 2 China (English) 349 21.93 (5.3) 65 1.79 (.7) 97 62 26 6 China (Mandarin) 1231 35.00 (7.3) 69 1.82 (.6) 95 92 86 4 Colombia 793 27.15 (11.5) 60 2.01 (.8) 96 57 22 7 Croatia 898 39.10 (12.1) 59 2.08 (.7) 71 91 69 2 Cyprus 363 34.31 (9.6) 65 2.09 (.7) 87 69 64 4 Czechia 700 38.10 (17.0) 66 2.29 (.6) 82 75 62 2 Ecuador 863 30.97 (12.3) 53 1.81 (.8) 86 65 33 11 Egypt 1627 23.62 (8.7) 72 2.06 (.6) 98 86 27 6 Estonia 449 38.93 (14.1) 63 2.10 (.7) 80 64 58 2 France 562 36.01 (14.2) 76 2.08 (.7) 64 67 47 NA Germany 620 31.01 (11.9) 62 2.18 (.8) 83 64 58 12 Ghana 434 21.97 (4.5) 41 2.08 (.8) 84 72 32 26 Greece 556 31.49 (11.8) 65 2.03 (.7) 91 63 55 5 Hungary 654 32.80 (13.4) 69 2.07 (.6) 72 69 63 2 Iceland (English) 1149 38.50 (17.5) 50 2.27 (.7) 92 61 65 11 Iceland (Icelandic) 432 54.91 (15.5) 54 2.05 (.6) 75 81 78 3 India (Hindi) 1664 32.07 (11.8) 45 2.14 (.8) 73 78 45 13 India (Tamil) 376 36.78 (12.1) 52 1.71 (.6) 57 65 70 37 Indonesia 292 19.79 (3.2) 72 1.76 (.5) 87 43 14 3 Iran 1318 33.46 (11.3) 60 1.99 (.6) 95 82 61 29 Iraq 405 34.13 (12.1) 33 1.49 (.5) 100 97 45 53 Ireland 351 33.73 (12.4) 50 2.11 (.8) 76 80 62 5 Israel 493 30.77 (11.6) 62 2.13 (.7) 87 67 32 7 Italy 2307 33.17 (14.0) 62 1.95 (.6) 81 67 61 6 (continued on next page) V. Swami et al. Journal of Environmental Psychology 99 (2024) 102432 8 Table 1 (continued ) Nation Sample size Mean age (SD) % Women Mean financial security (SD) %Urban residence %Secondary/tertiary education %In committed relationship or married %Racialised minority Japan 360 49.44 (16.6) 100 1.79 (.6) 90 81 61 8 Kazakhstan 380 30.07 (11.3) 53 2.04 (.6) 94 76 48 11 Latvia 827 41.04 (12.8) 66 2.02 (.7) 74 82 69 4 Lebanon 1295 25.74 (12.3) 67 1.93 (.7) 70 63 33 16 Lithuania 491 40.34 (12.8) 51 2.05 (.6) 72 84 74 3 Malaysia 1193 27.81 (8.7) 69 1.74 (.6) 76 84 29 30 Malta 347 35.52 (15.4) 72 2.10 (.7) 78 71 60 7 Nepal 353 25.78 (6.0) 50 1.77 (.7) 82 98 28 5 Netherlands 1004 46.81 (16.3) 53 2.05 (.6) 61 98 69 9 Nigeria 1274 31.64 (9.2) 34 1.85 (.8) 93 64 63 14 Norway 360 41.24 (11.6) 77 2.17 (.7) 78 92 77 4 Pakistan 267 20.59 (2.7) 28 2.16 (.9) 100 47 83 49 Palestine 401 27.64 (9.5) 25 2.01 (.6) 81 90 42 7 Philippines (English) 350 24.87 (11.2) 0 2.03 (.7) 97 56 24 13 Philippines (Tagalog) 504 37.43 (11.9) 73 1.83 (.7) 97 89 65 16 Poland 1954 30.51 (11.9) 62 1.99 (.7) 74 63 56 3 Portugal 363 36.53 (17.9) 68 2.05 (.7) 85 81 37 5 Romania 1819 26.94 (10.8) 53 2.05 (.7) 80 49 60 5 Russia 206 39.94 (11.8) 71 1.84 (.5) 97 84 67 8 Saudi Arabia 380 28.02 (9.7) 55 2.03 (.7) 94 83 33 20 Serbia 650 30.72 (11.3) 56 2.20 (.7) 95 65 65 10 Slovakia 814 37.79 (14.7) 54 1.92 (.6) 65 75 67 4 Slovenia 452 36.84 (14.9) 59 2.16 (.7) 49 87 66 2 South Africa 318 35.15 (16.1) 53 1.74 (.8) 78 73 45 31 South Korea 381 27.60 (9.7) 48 1.89 (.6) 98 54 43 52 Spain 1266 34.54 (16.3) 52 2.17 (.8) 88 82 43 5 Switzerland 377 46.48 (15.2) 52 1.98 (.7) 62 51 66 5 Taiwan 529 41.36 (13.6) 60 2.48 (.7) 90 92 67 7 Thailand 3275 25.85 (10.8) 62 1.76 (.6) 87 45 23 6 Tunisia 374 41.62 (15.2) 55 2.10 (.6) 96 90 63 0 Türkiye 2518 31.63 (11.5) 57 1.98 (.8) 97 61 57 14 Ukraine 141 39.00 (11.7) 59 1.74 (.6) 95 87 71 9 United Arab Emirates (Arabic) 204 26.37 (6.7) 73 2.07 (.4) 99 35 39 10 United Arab Emirates (English) 904 27.50 (11.8) 36 2.13 (.8) 98 73 43 31 United Kingdom 1243 37.99 (13.9) 54 2.03 (.7) 84 87 68 23 United States of America 2531 35.35 (12.7) 62 1.93 (.7) 85 82 61 20 Note. SD = standard deviation. V. Swami et al. Journal of Environmental Psychology 99 (2024) 102432 9 relative to others of their own age in their country of residence (1 = less secure, 2 = same, 3 = more secure). 2.3.5. Demographics Highest educational qualification was assessed as follows: 1 = no formal education, 2 = primary education, 3 = secondary education, 4 = still in full-time education, 5 = undergraduate degree, 6 = postgraduate degree, 7 = other; marital status was assessed as: 1 = single, 2 = single but in a committed relationship, 3 = married, 4 = other; and racialised status relative to their country of residence was assessed as: 1 = ethnic/racial majority, 2 = ethnic/racial minority, 3 = not sure. The latter item provides a common metric of categorising ethnicity/race across diverse nations (Swami, Barron, et al., 2020). For descriptive purposes at the national level and for analyses, response options for highest educational quali- fication were collapsed into secondary/tertiary (secondary education, undergraduate degree, postgraduate degree) versus other (all remaining categories) and response options of racialised status were collapsed into racialised minority (racial minority) versus other (all remaining categories). 2.4. Procedures, ethics, and data sharing Full procedural information about the BINS is provided in Swami et al. (2022). The BINS project was conducted in accordance with the principles of the Declaration of Helsinki and following all local institu- tional guidelines. In brief, once local ethics approval had been obtained or collaborators confirmed that approval was not required as per na- tional laws (see Supplementary Table S1), researchers recruited partic- ipants from the community in their respective nations between November 2020 and February 2022. Inclusion criteria were being ≥ 18 years of age, a resident and citizen of the particular nation in which recruitment took place, and being able to complete a survey in the language in which it was presented. In all but nine locales (see Sup- plementary Table S1), data collection was conducted online. All par- ticipants were presented with a standardised information sheet and provided (digital or written) informed consent before completing an anonymous version of the BINS package. Upon completion, participants received debriefing information, which included contact information for the first author as well as a local researcher. The BINS data and our analytic codes are available on the Open Science Framework at htt ps://osf.io/rfhwe/?view_only=dc87d4d3088b4f62922177fbbe06e8b6. 2.5. Analytic strategy The general analytic plan, including structural and measurement invariance analyses of the key variables of the BINS (including the NES and CNS) is described in the BINS study protocol (Swami et al., 2022). Further analyses not covered in the study protocol were not preregis- tered separately. The analysis proceeded in four steps, in a similar fashion for both the NES and CNS: first, CFA models were fitted to the total sample to determine the structure of the NES and test both the full CNS (henceforth “CNS-14”) and the CNS-7 for unidimensionality. For the NES, unidi- mensional and 2-factor models were fitted, testing for the possible scale multidimensionality (nature exposure in everyday life and environ- ments: Items #1 and #2; nature exposure outside everyday environ- ments: Items #3 and #4). As the items further allude to a conditional structure (with Items #1 and #3 assessing levels of “nature exposure” in each domain and Items #2 and #4 assessing “noticing nature” in each domain), a unidimensional model with correlated error variances be- tween Items #1 and #2, and Items #3 and #4, respectively was also fitted on the data (an analogous 2-factor model could not be tested, as it was under-identified with only 4 items). The best-fitting model was then used for further analysis. In the unidimensional model of the CNS, error variances of Items #5 and #7, and Items #5 and #10 were allowed to correlate (following Rosa et al., 2022). We expected a better model fit for the CNS-7 than the CNS-14, but followed this strategy to determine which version of the instrument should be used in further analysis. Second, in the six countries where data was collected in multiple languages, measurement invariance of the NES and CNS was tested with multi-group CFAs (MG-CFAs) for cross-language survey presentation. Data from the same nations were merged in the subsequent measure- ment invariance analysis of nations only if scalar measurement invari- ance held within nations (i.e., all linguistic versions of the scales showed satisfactory invariance). This was only true for China. Where this was not the case, data for the two different language versions of the survey were kept separate in further analysis (i.e., Canada, Iceland, India, the Philippines, and the UAE). We therefore use the term “national groups” rather than “countries” where we are reporting results that include multiple languages within countries. Third, measurement invariance of the CNS and NES was tested with MG-CFAs for: national groups, languages, gender identities (women vs. men vs. other gender identities), and age groups based on Arnett (2000) and Erikson (1968), namely young adulthood (18–24 years), middle adulthood (25–44 years) and older adulthood (≥ 45 years). Note that testing measurement invariance of languages implied mixing different national groups with the same language in analysis (e.g., English was the survey language in the UK, the USA, but, inter alia, also in one of the Chinese samples). Assuming scalar or at least partial scalar measurement invariance, we then examined latent means in these groups and, assuming at least metric invariance, we examined associations between the NES and CNS across national groups, using factor scores. Fourth, assuming scalar or at least partial scalar measurement invariance, sociodemographic correlates of the NES and CNS factor scores were investigated with multilevel models (MLMs) across national groups. Predictors were financial security, urbanicity (urban vs. rural), education (secondary/tertiary vs. other), marital status (committed/ married vs. other), and racialised identity (racialised minority vs. other). Mplus 8.8 (Muthén & Muthén, 2022) was used for the CFAs, MG-CFAs, and MLMs. Significance was set to p < .05. For the structural analyses, the weighted mean- and variance-adjusted weighted least squares estimator (WLSMV) was used to account for the ordered-categorical item response formats of the NES and CNS. WLSMV estimates one loading parameter for each item, but #response options – 1 threshold parameters (one for each transition of one response option to the next) instead of a single intercept parameter per item. To account for missing data (0.4% in the NES, 1.4% in the CNS-14), full information maximum likelihood was used. Measurement invariance analyses tested for configural invariance (i. e., same loading patterns across groups), metric invariance (i.e., same unstandardised loadings across groups), and scalar invariance (i.e., same unstandardised loadings and threshold parameters across groups). Based on the configural invariance models, reliability estimates (ω total) are presented for all groups. If full scalar measurement invariance could not be assumed, we aimed for partial scalar measurement invariance instead (i.e., equal item parameters across some groups and items, but not all). For this, the alignment method (e.g., Asparouhov & Muthén, 2023) was used for guidance. Alignment does not require exact measurement invariance, but instead seeks a solution that minimises the differences in loadings and threshold parameters across groups, while still retaining identical fit to the configural invariance model. The method provides for each item parameter information on the groups for which invariance holds and an R2 measure that indicates the amount of invariance (typically, the more invariant, the higher the R2; however, in some special cases, R2 is a poor measure of invariance; Asparouhov & Muthén, 2023). This information was used to (a) select two items per scale (an- chor items) for which invariance was assumed (two anchor items being sufficient for the comparison of latent means; Pokropek et al., 2019) and to (b) exclude groups that violated invariance the most. Additionally, we looked at the reliability of the scales in each group and the contributions of each group to the overall χ2 values of the MG-CFAs. If composite re- liabilities were low and/or the contributions to the χ2 value were high, V. Swami et al. Journal of Environmental Psychology 99 (2024) 102432 10 https://osf.io/rfhwe/?view_only=dc87d4d3088b4f62922177fbbe06e8b6 https://osf.io/rfhwe/?view_only=dc87d4d3088b4f62922177fbbe06e8b6 compared to sample size, groups were excluded from the final partial scalar measurement models. For the assessment of model fit, commonly used fit indices were consulted: the comparative fit index (CFI; good/acceptable fit: ≳ .95/ .90), the Tucker-Lewis index (TLI; ≳ .95/.90), the root-mean square error of approximation (RMSEA; good fit: ≲ .06) and its 90% confidence interval, and the standardised root mean square residual (SRMR; good fit ≲ .08; Hu & Bentler, 1999). The cut-off for the RMSEA was set to .15 for MG-CFAs with more than 10 groups (Rutkowski & Svetina, 2014). For the measurement invariance analyses, we present ΔCFI and ΔRMSEA values, and Δχ2 tests, but primarily interpreted the former two as they were not affected by sample size. For the comparison of metric to configural, and scalar to metric invariance models, the cut-offs ΔCFI ≲ .020/.010 and ΔRMSEA ≲ .030/.015 were used to indicate good fit of the respective stricter model (Rutkowski & Svetina, 2014). For the MLMs, Bayesian estimation (using diffuse priors as specified in Mplus default settings) was used. This allowed obtaining correctly standardised (cf. van Assen et al., 2022) parameter estimates that were interpreted as measures of effect size (comparable to Pearson’s r). For the dichotomous predictors, these estimates were further transformed into the metric of Cohen’s d as well. The predictors of financial security, urbanicity, education, marital status, and racialised identity were groupmean-centred on Level 1, and their cluster-level means were further used as Level-2 predictors. Thereby, associations on the indi- vidual level (Level 1) and on the cluster-level (national groups; Level 2) could be optimally distinguished. However, to avoid overfitting on Level 2, we only kept significant Level-2 predictors in the final models. 3. Results 3.1. Nature exposure scale 3.1.1. Structural analysis in the total sample The unidimensional model had poor to borderline fit to the data, judging by its CFI value (and disregarding TLI and RMSEA values, because of the small degrees of freedom of this model; Table 2). Items #1 and #3 had lower standardised loadings (.54 and .67) than Items #2 and #4 (.84 and .72; all ps < .001). Including correlated errors between Items #1 and #2, and #3 and #4, respectively, to accommodate the model for their conditional structure improved the model fit consider- ably (to keep this model identified, the strength of the residual associ- ation between the two items needed to be constrained to equality across pairs). The standardised items loadings in this model were .44, .86, .60, and .67, with residual correlations between Items #1 and #2, and #3 and #4, of .35 and .27 (all ps < .001). In contrast, a correlated 2-factor model (setting equality constraints for the similar loadings in each pair of Items #1 and #3, and #2 and #4 to obtain admissible parameter estimates of the residual variances) resulted in a poorer model fit (Table 2). In this model, Items #1 and #2 loaded on one factor (nature exposure in everyday life and environ- ments; standardised loadings = .62 and .87) and Items #3 and #4 on another (nature exposure outside everyday environments; .63 and .89). The latent factor intercorrelation was .76 (all ps < .001). Thus, even though there was some indication of potential multidi- mensionality in the NES, a unidimensional model that accommodated the conditional structure of the items fitted the data best and was, therefore, used in all subsequent analyses. However, according to their threshold parameters, some response options were relatively uninfor- mative, because item categories were so close to one another (which was also reflected in the relative sparseness of response option endorse- ment). Thus, response options 1 and 2, and 4 and 5 were each combined for Items #1 and #3, and response options 1 and 2 for Item #4. Even though this measure only slightly increased some model fit indices and slightly decreased others (Table 2), it ensured that sparseness of data did not further complicate the subsequent multigroup analyses. Thus, response options were also combined in all multigroup analyses. 3.1.2. Invariance of the cross-language results in the six countries with multiple survey languages The MG-CFA invariance test results are presented in Supplementary Table S2. Configural and metric invariance was found for all six coun- tries, but scalar invariance only for China. Thus, only the data from Table 2 Analyses of the NES in the total sample and invariance of the NES concerning national groups, language, gender identity, and age. Grouping variable and type of model χ2(df) CFI TLI RMSEA 90% CI SRMR Model comparisons ΔCFI ΔRMSEA Configural Metric Total sample 1F 6369.45(2) .944 .831 .236 [.232, .241] .043 1FCE 1754.73(1) .984 .907 .175 [.169, .182] .021 1FCE + CRO 1493.96(1) .984 .902 .162 [.155, .169] .027 2FConL 4434.23(2) .961 .882 .197 [.192, .202] .042 2FConL + CRO 3367.73(2) .963 .889 .172 [.167, .177] .048 National groups Configural invariance 1423.31(139) .987 .961 .107 [.102, .112] .034 Metric invariance 6951.05(346) .934 .920 .153 [.150, .156] .066 .053 .046 5300.73(207) Scalar invariance 19559.39(691) .812 .886 .183 [.181, .185] .084 .124 .030 17980.50(552) 13946.75(345) Language Configural invariance 1579.37(79) .985 .954 .115 [.111, .120] .032 Metric invariance 5970.23(196) .942 .929 .144 [.141, .147] .055 .043 .029 4266.41(117) Scalar invariance 17242.02(430) .831 .906 .166 [.164, .168] .072 .023 .022 15544.82(351) 12088.40(234) Gender identity Configural invariance 1052.28(5) .989 .959 .105 [.100, .110] .027 Metric invariance 863.35(11) .991 .985 .064 [.060, .068] .027 − .002 − .041 23.19(6) Scalar invariance 989.69(23) .989 .992 .047 [.045, .050] .028 .002 − .017 134.03(18) 120.26(12) Age Configural invariance 1206.13(5) .987 .952 .112 [.107, .118] .027 Metric invariance 1048.65(11) .989 .981 .070 [.067, .074] .028 − .002 − .042 150.04(6) Scalar invariance 1473.44(23) .984 .987 .058 [.055, .060] .031 .005 − .012 536.33(18) 414.80(12) Note. 1F = 1-factor model; 1FCE = 1-factor model with correlated errors between Items #1 and #2, and #3 and #4; 1FCE + CRO = 1FCE model with combined response options in Items #1, #3, and #4 (see main text for details); 2FConL = correlated 2-factors model with equality constraints on the loadings of Items #1 and #3, and #2 and #4; 2FConL + CRO = 2FConL model with combined response options in Items #1, #3, and #4 (see main text for details). All ps of χ2 and Δχ2 tests (comparisons of the multigroup models) were <.001. Gender identity compared groups of women, men, and other gender identity, age compared groups of partic- ipants with 18–24 years, 25–44 years, ≥45 years of age. V. Swami et al. Journal of Environmental Psychology 99 (2024) 102432 11 China (i.e., Mandarin, Cantonese, and English versions) were pooled for the analysis of national groups. This analysis therefore included 70 “national groups” from 65 countries. 3.1.3. Invariance across national groups, languages, gender identities, and age groups 3.1.3.1. Overall findings. The results of the MG-CFA analyses are pre- sented in Table 2. The NES showed configural, but neither metric nor scalar invariance, across the 70 national groups and 40 languages. Fig. 1. Ordering and Magnitude of Standardised Latent Mean Differences (Cohen’s d) in the NES Between National Groups (as Compared to the UK; Left) and Languages (as Compared to English; Right). Note. Data from UAE (Arabic version), Ecuador, Iran, Iraq, and Spain were excluded either due to poor scale reliability or poor fit in the partial measurement model (see main text). V. Swami et al. Journal of Environmental Psychology 99 (2024) 102432 12 However, it showed configural, metric, and scalar invariance across gender identities and age groups. Scale reliability was not strong. Me- dian scale reliability (ω total) across all national groups was .65, ranging from just .16 (UAE [Arabic version] though only reliability below .44) to .87 (Saudi Arabia; P25 = .60, P75 = .74). Nevertheless, with only 4 items and two correlated errors, scale reliability was judged to be sufficient in all national groups (except the UAE Arabic version) for further analyses for current purposes. Data from the UAE (Arabic version) were excluded in all further analyses of national groups and in the partial measurement model for languages. 3.1.3.2. National groups. The alignment method (Table S3) suggested the following descending ordering of the four NES items concerning their invariance: #2, #4, #3, #1. Item #3 had a higher summed R2 value Fig. 2. Ordering and Magnitude of Standardised Latent Mean Differences (Cohen’s d) in the CNS-7 Between National Groups (as Compared to the UK; Left) and Languages (as Compared to English; Right). Note. Data from Iraq and Bosnia and Herzegovina were excluded due to the poor scale reliabilities in these countries. V. Swami et al. Journal of Environmental Psychology 99 (2024) 102432 13 than Item #4; however, Item #4 had a higher loading and exhibited invariance in its parameters in a larger number of national groups than Item #3. Items #2 and #4 were, therefore, used as item anchors in a partial scalar measurement model. Note that this indicated that the two “nature noticing” items (#2 and #4) were more invariant across na- tional groups than the two “nature exposure” items (#1 and #3). In addition to the UAE (Arabic version), data from Ecuador, Iran, Iraq, and Spain were also subsequently excluded, as they disproportionately decreased the fit of the partial scalar measurement model, relative to their sample sizes. For the remaining 65 national groups (including 61 single language countries and four multi-language countries), the partial scalar mea- surement model (using items #2 and #4 as anchors) had acceptable fit, χ2 = 7722.24, df = 449, CFI = .913, TLI = .924, RMSEA = .141 (.138, .144), SRMR = .051. Compared to the UK as a reference category (for purely nominal reasons [the first author is based there]), the largest positive differences in self-reported nature exposure, as reflected by the latent means on the NES, were for (in descending order) Bosnia and Herzegovina, Croatia, and Lithuania, and were in the range of d = .98 to 1.18 (see Fig. 2, left panel; for individual Cohen’s d values, see Table S4). The largest negative differences compared to the UK were observed for Lebanon, South Korea, and, Brazil and were in the range of d = − .97 to − 2.21. 3.1.3.3. Languages. Using Items #2 and #4 again as anchor items, the partial scalar measurement model had an acceptable fit to the data of the 40 language groups, χ2 = 6959.85, df = 274, CFI = .926, TLI = .935, RMSEA = .131 (.128, .134), SRMR = .048. Here, the largest positive differences in country-level latent means on the NES, compared to En- glish (the original scale language), were observed for (in descending order) Bosnian, Lithuanian, and Croatian and were in the range of d = 1.04 to 1.18 (see Fig. 2, right panel; for individual Cohen’s d values, see Table S5). The largest negative differences were observed for Cantonese, Korean, and Portuguese and were in the range of d = − .87 to − 1.68. These rankings mostly matched the rankings of the analysis of national groups (see above; but note that languages could contain a mix of different national groups, if the language in which the scale was pre- sented there was the same; this specifically applied to English). 3.1.3.4. Gender identities and age groups. Men reported lower nature exposure (i.e., had lower NES latent means) than women (Cohen’s d = − .14, p < .001), as did those with other gender identities (d = − .25, p < .001). Reported nature exposure also increased with age: age groups 25–44 years vs. 18–24 years differed by d = .14 (p < .001), whereas age groups ≥ 45 years vs. 18–24 years by d = .48 (p < .001). 3.2. Connectedness to nature scale 3.2.1. Structural analysis of the CNS-14 and CNS-7 in the total sample A unidimensional model had a poor fit on the CNS-14, χ2 = 56589.17, df = 75, CFI = .921, TLI = .904, RMSEA = .115 (.114, .116), SRMR = .048, compared to the CNS-7, χ2 = 5583.20, df = 12, CFI = .987, TLI = .978, RMSEA = .090 (.088, .092), SRMR = .017. Addi- tionally, the standardised loadings of Items #12 and #13, and of the negatively worded Items #4 and #14, were low (≤ .36). Hence, the CNS- 7 was used in all subsequent analyses. 3.2.2. Invariance of the cross-language results in the six countries with multiple survey languages The MG-CFA invariance test results of the cross-language survey presentation of the CNS-7 in Canada, China, Iceland, India, the Philippines, and the UAE are presented in Supplementary Table S6. Full scalar invariance could be assumed for all countries, except Canada and Iceland, for which metric invariance was upheld. In all countries, except Canada and Iceland, the available data were pooled for the analysis of national groups. Thus, this analysis included 67 national groups from 65 nations. 3.2.3. Invariance across national groups, languages, gender identities, and age 3.2.3.1. Overall findings. The results of the MG-CFA analyses are pre- sented in Table 3. The CNS-7 showed configural, metric, and scalar invariance across gender identities and age groups, but only configural and metric invariance for the 67 national groups and 40 languages. Median scale reliability (ω total) across all national groups was consid- erably higher than the NES (.92), ranging from .64 (Iraq) to .97 (Spain; P25 = .89, P75 = .93). Bosnia and Herzegovina (.73) was the only other country below .80. 3.2.3.2. National groups. Using the alignment method, we obtained information on items and item parameters that were most invariant among the 67 national groups (Table S7). According to the summed R2 values across the loading and all threshold parameters per item (to get an indication of overall item invariance), Items #7, #5, and #2 (in this order) appeared to be the most invariant. However, it was Items #2 and #7 whose loadings were also invariant amongst the largest number of national groups (Item #5 had the lowest number here). Hence, we opted for using Items #2 and #7 as anchor items in a partial scalar measure- ment model. Considering the low scale reliabilities in the Iraq and Bosnia and Herzegovina data, we excluded these national groups from this analysis. Data of these two countries were also excluded in all further analyses of national groups and in the partial measurement model of languages. The partial scalar measurement model had a good fit to the data, χ2 = 10264.83, df = 1164, CFI = .982, TLI = .979, RMSEA = .095 (.094, .097), SRMR = .026. Comparing all other national groups to the UK, the ordering and magnitude of standardised latent mean differences (Cohen’s d) are provided in Fig. 2 (individual Cohen’s d values are provided in Table S8). The largest positive differences were observed for (in descending order) Nepal, Iran, and South Africa and were in the range of d = 1.20 to 1.39 (see Fig. 1, left panel), suggesting that par- ticipants from these nations reported higher connectedness to nature compared to the United Kingdom. The largest negative differences were observed for Israel, Japan, and Spain and were in the range of d = − .30 to − .61. 3.2.3.3. Languages. Using Items #2 and #7 as anchor items again, the partial scalar measurement model had a good fit on the data of the 40 language groups, χ2 = 8666.51, df = 716, CFI = .985, TLI = .983, RMSEA = .088 (.086, .090), SRMR = .023. Comparing all other lan- guages to English (again the original scale language), the ordering and magnitude of standardised latent mean differences (Cohen’s d) are provided in Fig. 2 (individual Cohen’s d values are provided in Ta- ble S9). The largest positive differences were observed for (in descend- ing order) Nepali, Bangla, and Farsi and were in the range of d = .63 to .88 (see Fig. 2, right panel). The largest negative differences were observed for Dutch, Hebrew, and Japanese and were in the range of d = − .40 to − .69. These rankings mostly matched the rankings of the analysis of national groups (see above). 3.2.3.4. Gender identities and age groups. Men reported slightly lower connectedness to nature (CNS-7 latent means) than women (Cohen’s d = − .08, p < .001). Other gender identities reported similar connected- ness to women (d = .02, p = .79). Connectedness to nature also increased with age: age groups 25–44 years vs. 18–24 years differed by d = .17 (p < .001), whereas age groups ≥ 45 years vs. 18–24 years by d = .36 (p < .001). V. Swami et al. Journal of Environmental Psychology 99 (2024) 102432 14 3.3. Associations between NES and CNS-7 factor scores within national groups In the 65 national groups for which both CNS-7 and NES factor scores could be computed, the association between nature exposure and connectedness to nature ranged from r = − .13 (Cyprus) to .22 (Malta), with r = .03 in median. That is, the median correlation between these two constructs across the 65 national groups was practically nil. 3.4. Sociodemographic correlates of NES and CNS-7 factor scores At the individual (Level-1) level, higher financial security, living in rural (vs. urban) settings, secondary/tertiary (vs. other) educational qualification, and being in a committed relationship or married (vs. other) were all associated with both greater nature exposure and connectedness to nature (see Table 4). Being a member of a racialised minority (vs. other) was associated with lower nature exposure. All as- sociations at this individual level were, however, small and statistically explained only small amounts of the Level-1 variance. Expressed as Cohen’s ds, the effect sizes for the dichotomous predictors of nature exposure and connectedness to nature, respectively, were as follows: urban vs. rural living setting, − .25/-.10; secondary/tertiary vs. other educational qualification, .02/.05; and committed/married vs. other, .07/.11. For racial minority (vs. other), Cohen’s d was − .04 for nature exposure and non-significant for nature connectedness. At the national group cluster-level (Level 2), connectedness to nature was not associated with any of the sociodemographic variables. However, higher cluster-level means of nature exposure were associated with lower cluster-level means (i.e., lower prevalence rates) of living in urban (vs. rural) settings and of being a member of a racial minority (vs. other), and thus included in the final model. In other words, countries (including those with multiple languages) with more rural respondents had higher nature exposure and those with more racial minority re- spondents had lower overall nature exposure, echoing the Level-1 re- sults. These cluster-level associations were sizeable and, combined, explained 26% of the Level-2 variance. 4. Discussion Here, we used the BINS dataset – with data from 56,968 respondents across 65 nations and 40 languages – to conduct the most comprehen- sive assessment of the factorial validity and measurement invariance of the NES and CNS. In terms of the NES, reliability across countries was highly variable with many countries falling below the usual thresholds of acceptability. A unidimensional model of the NES did, however, show full scalar invariance across gender identities and age groups; partial scalar invariance was also found for all languages and all but five na- tional groups. In terms of the CNS, our results are consistent with pre- vious work suggesting that the 14-item, unidimensional model of this instrument has poor factorial validity (Pasca et al., 2017; Rosa et al., 2022). Conversely, the CNS-7 showed full scalar invariance across gender identities and age groups, and partial scalar invariance across all languages and all but two national groups. Associations between nature exposure and connectedness to nature across nations were negligible, Table 3 Invariance of the CNS-7 concerning national groups, language, gender identity, and age. Grouping variable χ2(df) CFI TLI RMSEA 90% CI SRMR Model comparisons ΔCFI ΔRMSEA Configural Metric National groups Configural invariance 8001.09(804) .986 .976 .103 [.101, .105] .024 Metric invariance 9076.08(1200) .985 .982 .088 [.086, .090] .028 .001 − .015 2679.88(396) Scalar invariance 33207.52(2520) .941 .967 .120 [.119, .121] .044 .044 .032 25335.58(1716) 25786.21(1320) Language Configural invariance 6625.49(480) .988 .980 .095 [.093, .097] .020 Metric invariance 7376.45(714) .987 .985 .081 [.079, .083] .024 − .005 − .014 2167.62(234) Scalar invariance 27691.05(1494) .951 .972 .111 [.110, .112] .038 .036 .020 21099.45(1014) 21599.26(780) Gender identity Configural invariance 5361.43(36) .988 .979 .088 [.086, .090] .017 Metric invariance 3851.56(48) .992 .989 .065 [.063, .066] .017 − .004 − .023 120.35(12) Scalar invariance 4026.67(88) .991 .994 .049 [.047, .050] .018 .001 − .016 766.62(52) 710.21(40) Age Configural invariance 5784.07(36) .987 .977 .092 [.090, .094] .017 Metric invariance 3981.35(48) .991 .988 .066 [.064, .067] .017 − .004 − .026 150.24(12) Scalar invariance 4236.02(88) .991 .993 .050 [.049, .051] .018 .000 − .016 837.15(52) 757.37(40) Note. All ps of χ2 and Δχ2 tests (model comparisons) were <.001. Gender identity compared groups of women, men, and other gender identity, age compared groups of participants with 18–24 years, 25–44 years, ≥45 years of age. Table 4 Sociodemographic correlates of nature exposure and connectedness to nature. Predictor Nature exposure Connectedness to nature Estimate (posterior SD) 95% credibility interval p (one-tailed) Estimate (posterior SD) 95% credibility interval p (one-tailed) Level 1: Individual level Financial security 0.07 (0.004) [0.06, 0.08] < 0.001 0.02 (0.005) [0.01, 0.03] < 0.001 Urbanicity ¡0.12 (0.004) [-0.13, -0.12] < 0.001 ¡0.05 (0.004) [-0.06, -0.04] < 0.001 Education 0.01 (0.004) [0.005, 0.02] < 0.001 0.03 (0.004) [0.02, 0.04] < 0.001 Marital status 0.04 (0.005) [0.03, 0.05] < 0.001 0.05 (0.004) [0.05, 0.06] < 0.001 Racialised status ¡0.02 (0.005) [-0.03, -0.01] < 0.001 .002 (.004) [-.005, .01] .29 Level 2: Cluster-level means Urbanicity ¡0.35 (0.098) [-0.51, -0.14] < 0.001 Racialised status ¡0.30 (0.103) [-0.46, -0.05] < 0.001 Random Effects Intercept residual variance .74 (.09) [.56, .92] <.001 1.00 (.00) NA <.001 Explained variance (Level 1/Level 2) 2%/26% .7%/0% Note. Estimates are on a standardised scale. SD = standard deviation. Significant (one-sided p < .025) estimates are highlighted in boldface. V. Swami et al. Journal of Environmental Psychology 99 (2024) 102432 15 although we did find significant (albeit weak) individual level associa- tions between both constructs and several sociodemographic factors. 4.1. Nature exposure We tested both a unidimensional model of the NES with all four items and a correlated 2-factor model, with equality constraints set for the two items that most closely assess “nature exposure” (#1 and #3) and “noticing nature” (#2 and #4), respectively. In the total sample, the 2- factor model had poor fit to the data, with item loadings reflecting “nature exposure” in everyday environments (#1 and #2) and outside everyday environments (#3 and #4), rather than an “exposure-noticing” split. In contrast, a unidimensional model had adequate fit to the data, although we did find that the “nature exposure” items had lower standardised loadings than the “noticing nature” items. Including correlated errors between two item pairs (#1 and #2, and #3 and #4, respectively) substantially improved fit of the unidimensional model. In short, although the NES does present some indication of possible sta- tistical multidimensionality, this does not obviously align with face validity of the items, and a unidimensional model presented the best fit to the data anyway. Although this unidimensional model of the NES showed configural invariance, it did not show metric or scalar invariance across national groups and languages. This suggests that, while there may be a near- universally plausible basic organisation of the nature exposure construct – as measured using the NES – each item does not contribute to the latent construct in the same way across nations or languages, making it a problematic measure to use in cross-cultural research. It was none- theless possible to achieve partial scalar invariance across all but five of the national groups (i.e., UAE [in Arabic], Ecuador, Iran, Iraq, and Spain) represented in the analyses and all 40 languages, suggesting that it may tap a common latent construct across national groups and lan- guages, albeit with some variation in meaning. There were also large inter-nation and inter-language differences in NES scores. These results may reflect actual cross-national differences in exposure to natural environments, as evidenced by the large variations seen in Fig. 1. Content-wise, these differences appeared to be particu- larly driven by differences in “noticing nature” (Items #2 and #4 were the primary contributors to latent NES scores) but may have also been affected by differences in the meaning of “exposure to nature” across national groups (McPhie & Clarke, 2020). Here, we highlight Brazilian participants as an outlier in terms of their low latent NES scores (see Fig. 1). Reasons for this are unclear but may reflect difficulties that Brazilian participants experience in noticing natural environments, despite their proximity to such environments (see Profice et al., 2023). Alternatively, it may reflect low levels of nature exposure that many urban dwellers in low- and lower-middle income nations have due to a lack of time, money, and nearby natural environments (Awoyemi et al., 2024) or a perceived lack of safety in natural environments (e.g., due to social unrest). Conversely, there was evidence that the unidimensional model of the NES achieved full scalar invariance across gender identities and age groups. In terms of gender identities, women reported significantly greater nature exposure than men and individuals who identified their gender in another way, although effect sizes were small. This is a curious finding, particularly as studies in Western nations have generally iden- tified a gender gap in nature contact, with women visiting natural en- vironments less frequently than men (e.g., Boyd et al., 2018), possibly because of experiences of fear and vulnerability in natural environments or due to societal and gendered norms that mean women often feel a lack of entitlement to leisure time in general (Day, 2000). It is possible that this discrepancy across studies is due to the inclusion of “noticing na- ture” in the NES and possibly women having a greater tendency to “notice nature” because they may be more sensitive to particular settings that may evoke fear, anxiety, and negative emotions (van den Berg & ter Heijne, 2005). In terms of age, self-reported nature exposure using the NES generally increased from emerging to older adulthood. It is likely that the latter group has greater time and opportunities to engage with nature (Freeman et al., 2019) and may also have a fuller understanding of the natural world, which in turn may enhance their abilities to “notice” the natural environment (Ojala, 2009). 4.2. Connectedness to nature The CNS results indicated that a unidimensional model including all 14 items had poor fit. Although the 14-item model demonstrated adequate fit in some previous research (e.g., Li & Wu, 2016; Mayer & Frantz, 2004; Pessoa et al., 2016), our findings corroborate more recent difficulties replicating this model (e.g., Navarro et al., 2017; Olivos et al., 2013; Pearce et al., 2022). As Pasca et al. (2017) have suggested, it is highly likely that some CNS items either perform poorly – for lin- guistic, conceptual, or semantic reasons – or are redundant in some national contexts. Adopting an Item Response Theory approach also suggested that some items may inadequately discriminate between in- dividuals who vary in their degree of connectedness to nature (Pasca et al., 2017). For these reasons, Pasca et al. (2017) recommended using a truncated version of the CNS that includes only seven of the 14 original items, which has been found to have a unidimensional factor structure in previous work (Rosa et al., 2022). Using this CNS-7, our results suggested that configural and metric invariance was supported across nations, languages, gender identities, and age groups. In terms of national groups and languages, the lack of full scalar invariance is consistent with previous work in seven European nations, where only configural and metric invariance was found (Navarro et al., 2022). One possible reason for the lack of scalar invariance of the CNS-7 across national groups and languages is that this instrument – and indeed the construct of connectedness to nature itself – is steeped in a boundary distinction between humans and the natural environment (i.e., that one can be distinct or disconnected from nature). As Fletcher (2017, p. 228) has suggested, however, this view is grounded in a “culturally specific … conceptual dichotomy between opposing realms of ‘nature’ and ‘culture’” that is characteristic of wealthy, Western nations. In other words, the suggestion that humans can be separate from nature – or, indeed that humans are not nature – may be a cultural worldview with limited global application, which in turn may explain the lack of scalar invariance in relation to the CNS-7 in the present study. Having said this, full scalar invariance is often an unrealistic goal for datasets with a large number of groups (Marsh et al., 2018). Impor- tantly, our results also suggested that it was possible to achieve partial scalar invariance of the CNS-7 across all but two national groups (i.e., Iraq and Bosnia and Herzegovina, respectively) and all 40 languages represented in the BINS. This is important theoretically because it sug- gests that the CNS-7 can be used to measure a latent construct of connectedness to nature that may have near-universal applicability, albeit with some possible loss of meaning. From a practical point-of-view, achieving partial scalar invariance means that we were able to compare latent CNS-7 means across national groups and lan- guages, with our results showing large cross-national and cross-language differences in CNS-7 latent means. Understanding why such differences exist is, however, more difficult and only preliminary explanations can be put forward based on the present data. Richardson, Hamlin, Elliott, and White (2022), for instance, sug- gested that lower country-level nature connectedness may reflect his- torical interactions with, and attitudes towards, the natural world. Observing that all six of the countries with the lowest nature connect- edness scores in an 18-country study by White et al. (2021) were at some time subject to British rule, Richardson, Hamlin, Elliott, and White (2022) wondered whether early industrialisation and urbanisation of the UK, and resource extraction and exploitation of its colonies, encouraged seeing nature as external and separate to human lives. However, although the UK again had very low levels of CNS in the V. Swami et al. Journal of Environmental Psychology 99 (2024) 102432 16 present research, as did some of its former colonies (e.g., Australia, Canada [English], Ireland, and the USA), other former British colonies, notably on the African continent, including South Africa, Nigeria, and Ghana, were among the countries with the highest connectedness to nature. An alternative suggestion, therefore, is that it is the English language version that results in lower CNS scores, an idea supported by the English versus French/Icelandic data from Canada and Iceland data, but undermined by the fact that these three African countries also used the English language version. More broadly, therefore, it appears that the variation in connected- ness to nature across nations and languages may reflect differences in the ways that connectedness to nature is constructed, negotiated, and experienced by different communities (see McPhie & Clarke, 2020). That is, what it means to be “connected to nature” or experience a sense of oneness with nature likely varies across nations and/or cultures, which results in the type of variation that can be seen in Fig. 2 when individuals are asked to self-report their experiences. It is also possible that cultural practices and traditions that foster both connectedness to nature, as well as the ability to articulate that connectedness, varies across national groups (Keaulana et al., 2021). For instance, there is some evidence to suggest that individuals in Nepal and Bangladesh – two nations that had very high latent CNS-7 scores in the present study – live in ways that are intimately connected to the natural world (Widdop Quinton & Khatun, 2020). Further work is clearly needed to explore these issues in greater depth. In terms of gender identities, we found that the CNS-7 showed full scalar invariance across women, men, and individuals who identified their gender in another way. The CNS-7 taps a common underlying construct of connectedness to nature that is not differentially affected by gender identities. Women reported higher nature connectedness than men, consistent with previous work showing that women have greater environmental concern (e.g., Xiao & McCright, 2015) and dispositional empathy with nature (Tam, 2013), though effect sizes were relatively small. Additionally, we found that the CNS-7 was fully invariant across age groups, with connectedness to nature generally increasing from emerging adulthood to older adulthood, though again group differences were relatively small. 4.3. Correlates of nature exposure and connectedness Contrary to our hypothesis, the median correlation between nature exposure and nature connectedness, as measured using the NES and CNS, across nations was practically nil. This finding stands in contrast to previous cross-sectional (e.g., Fränkel et al., 2019; Martin et al., 2020) and experimental studies (Sheffield et al., 2022) that have shown these constructs to be weakly-to-moderately correlated. Moreover, where studies have assessed these constructs using the NES and CNS, respec- tively, reported associations have tended to be moderate-to-large (Baceviciene et al., 2021; Picanço et al., 2024; Swami et al., 2016; Swami, Barron, et al., 2020). In these studies, however, the significant associations may have been inflated through the use of ecological cor- relations in singular nations. Instead, the present results suggest that the true variation in the as- sociation between nature exposure (measured using the NES) and connectedness to nature (measured using the CNS-7) across nations may be relatively wide. This raises questions about what exactly is being measured by the NES, and how. While we acknowledge the rationale for the scale to want to incorporate the three types of exposure explored in the literature (local, active visits, and awareness) into a single measure, the way this is operationalised is logically problematic because there