Forensic science plays an important role in the investigation of serious crimes. One of the first significant achievements in the field was the development of techniques for identifying individuals by their fingerprints. In the 19th century, it was discovered that almost any contact between a finger and a fixed surface left a latent mark that could be made visible by a variety of procedures (e.g., the use of a fine powder). In 1894 in England the Troup Committee, a group established by the Home Secretary to determine the best means of personal identification, accepted that no two individuals had the same fingerprints—a proposition that has never been seriously refuted. In 1900 another committee recommended the use of fingerprints for criminal identification. Fingerprint evidence was first accepted in an Argentine court in the 1890s and in an English court in 1902. Many other countries soon adopted systems of fingerprint identification as well.

Fingerprinting was originally used to establish and to make readily available the criminal records of individual offenders, but it quickly came to be widely used as a means of identifying the perpetrators of particular criminal acts. Most major police forces maintain collections of fingerprints that are taken from known criminals in order to identify them later should they commit other crimes. The FBI, for example, reportedly held millions of prints in its electronic database at the beginning of the 21st century. Fingerprints found at crime scenes thus can be matched with fingerprints in such collections.
Historically, searching fingerprint collections was a time-consuming manual task, relying on various systems of classification. The development in the 1980s of computerized databases for the electronic storage and rapid searching of fingerprint collections has enabled researchers to match prints much more quickly.

Although the science of fingerprinting is popularly perceived as error-free, some critics have charged that it is not an exact science—in part because prints are rarely pristine when gathered at a crime scene—and that some defendants have been convicted on the basis of mistaken fingerprint identification. For example, in 2004 the FBI used a fingerprint to link Brandon Mayfield, an American attorney, to a train bombing in Madrid; however, he was vindicated after a review revealed that the fingerprint, used to obtain a warrant for his arrest, did not belong to him. According to the British standard, if a set of fingerprints found at a crime scene is incomplete, it may be said to match another set (e.g., a set stored in a fingerprint database) if the two sets share at least 16 characteristics. However, no particular number of characteristics is accepted everywhere, and some jurisdictions require as few as 12 characteristics to reach a conclusive identification.
A broad range of other scientific techniques are available to law enforcement agencies attempting to identify suspects or establish beyond doubt the connection between a suspect and a crime. Since becoming reliably available in the late 1980s, DNA fingerprinting of biological evidence (e.g., hair, sperm, and blood) can exclude a suspect absolutely or establish guilt with a very high degree of probability. Many other substances, such as fibres, paper, glass, and paint, can yield considerable information under microscopic or chemical analysis. Fibres discovered on the victim or at the scene of the crime can be tested to determine whether they are similar to those in the clothing of the suspect. Documents can be revealed as forgeries on the evidence that the paper on which they were written was manufactured by a technique not available at the time to which it allegedly dates. The refractive index of even small particles of glass may be measured to show that a given item or fragment of glass was part of a particular batch manufactured at a particular time and place. Computer networks allow investigators to search increasingly large bodies of data on material samples, though the creation of such databases is time-consuming and costly.