A survey by Nature of 1,576 researchers reveals that over 70% have failed to reproduce another scientist's experiments, and over half have failed to reproduce their own. While 52% agree there is a reproducibility crisis, less than 31% believe that failure to reproduce published results means the result is likely wrong. Most researchers still trust the published literature. The survey found that 73% believe at least half of the papers in their field can be trusted, with physicists and chemists showing the most confidence. The survey highlights conflicting attitudes towards reproducibility. Researchers often assume that failed replications are due to valid, but boring, reasons. Incentives to publish positive replications are low, and journals are reluctant to publish negative findings. However, 24% of respondents were able to publish successful replications, and 13% published failed ones. Many researchers have taken steps to improve reproducibility, with 33% having done so in the past five years. Common strategies include having lab members redo experiments, improving documentation, and standardizing experimental methods. However, efforts to reproduce work can be time-consuming and costly, leading some researchers to perform checks only for innovative projects or unexpected results. Factors contributing to irreproducible research include pressure to publish, selective reporting, insufficient replication in the lab, and poor oversight. Pre-registration is one approach to improving reproducibility, but it is not widely used. Other strategies include better experimental design, improved statistics, and better mentorship. The survey suggests that journals, funders, and research institutions should take steps to improve reproducibility. Over 80% of respondents believe that funders and publishers should do more to improve reproducibility. The results indicate that awareness of the issue is growing, and that now is the time to take action.A survey by Nature of 1,576 researchers reveals that over 70% have failed to reproduce another scientist's experiments, and over half have failed to reproduce their own. While 52% agree there is a reproducibility crisis, less than 31% believe that failure to reproduce published results means the result is likely wrong. Most researchers still trust the published literature. The survey found that 73% believe at least half of the papers in their field can be trusted, with physicists and chemists showing the most confidence. The survey highlights conflicting attitudes towards reproducibility. Researchers often assume that failed replications are due to valid, but boring, reasons. Incentives to publish positive replications are low, and journals are reluctant to publish negative findings. However, 24% of respondents were able to publish successful replications, and 13% published failed ones. Many researchers have taken steps to improve reproducibility, with 33% having done so in the past five years. Common strategies include having lab members redo experiments, improving documentation, and standardizing experimental methods. However, efforts to reproduce work can be time-consuming and costly, leading some researchers to perform checks only for innovative projects or unexpected results. Factors contributing to irreproducible research include pressure to publish, selective reporting, insufficient replication in the lab, and poor oversight. Pre-registration is one approach to improving reproducibility, but it is not widely used. Other strategies include better experimental design, improved statistics, and better mentorship. The survey suggests that journals, funders, and research institutions should take steps to improve reproducibility. Over 80% of respondents believe that funders and publishers should do more to improve reproducibility. The results indicate that awareness of the issue is growing, and that now is the time to take action.