ECHO Corpus: Cross-cultural database of 15000 nonverbal vocalisations from speakers of 22 languages

Research on human non-verbal vocalisations such as screams, cries and laughter is rapidly expanding, yet most studies remain restricted to single cultures and heavily overrepresent English speakers. Here, we introduce the ECHO corpus – a large-scale cross-cultural database of non-verbal vocalisations produced by nearly 1050 vocalisers aged 18-84 years from 52 countries, including native speakers of 22 different languages spanning 10 language families. The corpus comprises over 15000 volitional vocalizations expressing sixteen affective states, from pain, aggression and alarm to amusement, relief and sensual pleasure, together with baseline speech from the same speakers. We provide extensive metadata on vocalisers (e.g., age, sex, countries of birth and residence, spoken languages, and media usage) and validate the dataset through acoustic quality assessment, analyses of fundamental frequency, and cross-cultural perception experiments demonstrating high recognition of vocalisation contexts. This first-of-its-kind resource provides a representative foundation for investigating how humans vocally encode and decode emotions and motivations across diverse cultures, revealing the mechanisms and functions of human vocal communication. The ECHO corpus is openly available and his content is described in the README file.