By Janet Raloff

“Torture numbers and they will confess to anything,” author Gregg Easterbrook once wrote in a magazine piece on climate change. But his quip could have been thesubtitle for this new book on the abuse of numbers in the courts.

Its authors, mother-daughter mathematicians, belong to a research group devoted to improving the use of statistics in criminal trials. Each chapter focuses on cases exemplifying a particular class of statistical error. Poignant tales detail exonerations resulting from faulty math used in the original trials (errors usually corrected only after intervention by statisticians, some of whom stepped in independently).

In one example, the authors tell the stories of grieving mothers charged with child abuse after each lost several children to crib death. Gross miscalculations of the probability that this might occur repeatedly within a family resulted in murder convictions for several women.

Then there was the 1894 conviction of French army captain Alfred Dreyfus for high treason. The officer had been charged with sending secrets to the enemy in a coded message. The real traitor eventually emerged, but only after a faulty statistical analysis of handwriting and codes had sent Dreyfus to a long imprisonment on an isolated island off French Guiana.

Not all those exonerated by faulty stats are sympathetic characters; in one of the book’s examples a person convicted by miscalculated numbers may have been guilty anyway. Regardless, the authors argue, “the misuse of mathematics can be deadly.” Most important, problems tend to arise from inadvertent errors, they find, “not from any inherent inapplicability of mathematics to justice.”

*Basic Books, 2013, 272 p., $26.99*

*Buy this book from Amazon.com.* Sales generated through the links to Amazon.com contribute to Society for Science & the Public’s programs.