Knowledge is being lost at an "astonishing rate" mainly because scientists are so disorganised, a study has found.
Evaluation of more than 500 randomly selected studies showed that 20 years after publication, 80 per cent of their original data had vanished.
Mundane issues of organisation, such as out-of-date email addresses and obsolete storage devices were chiefly to blame, researchers said.
The key problem was said to be that authors are expected to store and look after their own data.
"I think nobody expects that you'd be able to get data from a 50-year-old paper, but to find that almost all the data sets are gone at 20 years was a bit of a surprise," said Dr Timothy Vines of the University of British Columbia, Canada, who led the study.
"Publicly funded science generates an extraordinary amount of data each year.
"Much of these data are unique to a time and place, and are thus irreplaceable, and many other data sets are expensive to regenerate.
"The current system of leaving data with authors means that almost all of it is lost over time.
"The data are thus unavailable for future researchers to check old results or use for entirely new purposes.
"Losing data is a waste of research funds, and it limits how we can do science.
"Scientific data are being lost at an astonishing rate, and concerted action - particularly by journals - is needed to make sure it is saved for future researchers."
The new assessment, published in the journal Current Biology, focused on papers reporting the length measurements of plants and animals.
Those papers were selected because length measurements have been collected in exactly the same way for decades, making comparisons over time easier.
The analysis found that the odds of obtaining an original data set for any one paper fell by 17 per cent every year.
Dr Vines' team urged scientific journals to insist that study authors share their data on a public archive before a paper can be published.
