Scientific American The system could someday provide fast results at a crime scene Advertisement A sleepy driver on a dark road hears a loud “thump!” and later finds a speck of blood on the front bumper. Or police spot a tiny but suspicious bloodstain at a crime scene. Quickly determining whether such traces come from humans or animals is crucial, but the necessary tests can be time-consuming—and may destroy valuable evidence. Researchers say a new technique could help. To develop a rapid, nondestructive way to identify human blood, State University of New York at Albany forensic chemists Igor Lednev and Ewelina Mistek-Morabito combined spectroscopy and statistics. They shined infrared light on dried samples of human blood and on those of 10 animal species, including common pets such as dogs, cats and ferrets, as well as deer and elk (both of which are often struck by vehicles). The researchers recorded the reflected light, the spectrum of which changes based on blood's composition. They then used these data to train a machine-learning algorithm to quickly and accurately distinguish human from animal sources.