In July 1994, a woman was kidnapped from her apartment building in Waukegan, Illinois, driven roughly a mile away, and raped by two men. Hours later, police stopped a car near the scene. Its driver, 20-year-old Angel Gonzalez, had no criminal record. He had arrived in the United States from Mexico less than two years earlier and was not conversant in English.
The victim identified him at the scene, a identification made under conditions criminologists would today flag as inherently unreliable: high stress, poor lighting, a brief window of observation, and a suggestive one-on-one “show up ” rather than a properly administered lineup. After an overnight interrogation conducted in a mix of English and Spanish, detectives reported that Gonzalez eventually offered a confession, emerging “piecemeal” after roughly two hours of questioning.
That confession became the evidentiary core of the prosecution’s case. Yet on close examination, it collapsed under its own inconsistencies. Gonzalez identified the wrong location for the assault. He described a single crime scene where the victim recalled two. He supplied details, such as both assailants grabbing and covering the victim’s mouth — that the victim herself never corroborated. Meanwhile, four independent witnesses, including his girlfriend and her sister, placed him elsewhere for the relevant hours.
None of this prevented conviction. In June 1995, a Lake County jury found Gonzalez guilty of rape and kidnapping. He was sentenced to 55 years.
The slow correction of a fast judgment
It took the better part of two decades, and the sustained intervention of the Innocence Project beginning in 2012, before DNA testing was authorized on the case evidence. When results returned in January 2015, they identified the genetic profiles of two men and excluded Gonzalez entirely. His convictions were vacated within weeks. He had served 20 years for a crime he did not commit, and was later awarded $220,000 in state compensation, followed by a $9.5 million settlement in a federal civil rights suit.
Reading the case against the literature
Gonzalez’s conviction is not an aberration; it is a near-textbook illustration of the two contributing factors the Innocence Project’s own case data identify as the most recurrent drivers of wrongful conviction nationally: eyewitness misidentification and false or coerced confessions. What makes this case analytically instructive is not merely that these two factors were present, but that they were mutually reinforcing. The initial identification, however fragile, appears to have anchored investigators’ certainty going into the interrogation, a phenomenon the literature on confirmation bias in criminal investigations describes as tunnel vision. Once that certainty was established, an internally inconsistent confession from a young, non-native English speaker with no prior contact with the justice system was treated not as a signal to reexamine the theory of the case, but as corroboration of it.
This is precisely the mechanism by which procedurally “valid” convictions can be substantively wrong: each piece of evidence, examined in isolation, met a low threshold of admissibility, while the conjunction of unreliable identification and an unreliable confession was allowed to substitute for the kind of physical, corroborated evidence the case ultimately turned on when DNA testing was finally performed.
For anyone working on criminal justice reform, evidentiary reliability, or comparative wrongful conviction scholarship, cases like Gonzalez’s underscore a recurring argument in the field: safeguards such as recorded interrogations, blind or double-blind lineup administration, and statutory rights to post-conviction forensic testing are not procedural formalities. They are the structural conditions under which the difference between a conviction and the truth can still be recovered sometimes, as here, only after two decades.