Build Up Big-Team Science

Are some of science’s biggest questions simply unanswerable without redefining how research is done? This is the question that motivated the researchers who would later establish the ManyBabies Consortium: a grass-roots network of some 450 collaborators from more than 200 institutions who pool resources to complete massive studies on infant development (see, for example, ref. 1). Human infants are perhaps the most powerful learning machines on the planet — and understanding how that learning occurs could inform artificial intelligence, public policy, education and more. Yet a full understanding of infant learning seemed difficult (if not impossible) under the current research model.

Consider the question of what captures infants’ attention. Surely the probability that an infant will pay attention to, say, a rabbit, depends on presentation (for example, by a mother or a stranger), the child’s previous experiences with mammals, what else is present alongside the rabbit, and much more. Unpacking this effectively would require dozens of experimental conditions and hundreds of infant participants. But most research projects are run by individual principal investigators and a shifting population of PhD students, meaning that data-collection efforts typically recruit fewer than 25 infants for each condition being tested².

But what if researchers worked interdependently and distributed work across many laboratories? Such consortia might be able to answer questions that no individual lab could tackle alone. In a proof-of-concept study, the ManyBabies Consortium used word of mouth, social media and e-mail lists to amass a team of 69 labs to test whether infants across several world regions prefer ‘baby talk’: the high-pitched, sing-song speech that adults in many cultures use with babies. Data from 2,329 infants in 16 countries provided a resounding yes, demonstrating that infants even prefer baby talk that is not in their native language³. This study, the largest of its kind, was cited more than 100 times within a year of its publication, according to Google Scholar.

…

Read the whole story (subscription may be required): Nature