Sometime over that winter night in Colorado, the unknown male burst into the Bennett family’s home. He had a hammer, and he had something sharp.

The killer murdered 27-year-old Bruce Bennett, his wife Debra, and their seven-year-old daughter Melissa with the weapons. He raped two of the victims. Only the younger daughter, three-year-old Vanessa, survived the blunt-force trauma he left in his wake.

The Aurora Police Department has investigated the case since the bodies were discovered the morning of Jan. 16, 1984. The last in a series of attacks over a two-week period, the killer has never been found. Some hope in solving the cold case was mustered when a DNA profile from the Bennett crime scene was developed in 2002. But no match was made in any database. The case stayed cold.

Now the police have turned to DNA phenotyping, using some of the clues in the genetic profile, to estimate the look of the killer, they announced last week.

Parabon NanoLabs, a Virginia-based company which specializes in phenotyping, put together a “Snapshot” predictive sketch of what the murderer would look like at age 25 and age 55 – an estimation of what he would look like today, if still alive.

“This is the first time we have had some idea of who we’re looking for. He is no longer invisible,” said Steve Conner, the Aurora detective leading the investigation into the Bennett killings. “With release of these Snapshot composites, we hope people familiar with the case and the area at that time might be reminded of something or someone significant to their investigation.”

Phenotyping has been used in a handful of cases where traditional leads have been exhausted. The first investigative usage was announced in January 2015, with the double-homicide of a mother and daughter in Columbia, South Carolina.

But it’s been used in ways other than reconstructing the faces of killers. The face of a murdered woman found in Galveston, Tx. in 1988 was reconstructed last month in an attempt to identify her, according to a local newspaper. The genetic sketch cast new light on that murder case, since the victim was found without a head, and was too decomposed for fingerprints to be taken. The victim was originally described publicly as being white – and not Asian as she was eventually determined to be.

Aurora and other agencies warn that the phenotyping images are “scientific approximations” – and are not exact replicas of the person’s appearance. Notable disparities can be created by environmental factors outside of the genetics, including drug use and drinking, smoking, weight gain, facial scars, and facial hair.

Bennett Family circa 1984. (Courtesy of Parabon Nanolabs)

Parabon hired an expert forensic artist last year to help produce tailored images based off the genetic clues, they told Forensic Magazine in an interview last summer.

The company said in a statement that the Bennett case is a good example of how the technology can assist detectives who are stuck at a dead end.

“When investigators find themselves ‘chasing a ghost,’ Snapshot can provide a wealth of information to make the search for a suspect or person of interest more efficient,” said Ellen Greytak, Parabon’s director of bioinformatics.


Phenotyping has been greeted with enthusiasm by some forensic experts – but with skepticism by others.

Last week, as Aurora police were revealing what they believe could be the face of the Bennett family’s killer, an analysis of the phenotyping technology was published in the journal Cell.

The piece evaluated the burgeoning phenotyping field – and had criticisms of some of the claims of the technology.

Eye color predictions are not 100 percent, according to the piece. The work of Susan Walsh at Indiana University-Purdue University Indianapolis, along with colleagues in the Netherlands, improved the predictive models to 95 percent for brown and blue eyes – but had not yet developed better predictive models for other eye colors, they said.

Scientists in Poland led by Wojciech Branicki have pushed predictive models for hair color to 75 percent accuracy, the piece added.

But advanced facial-feature predictions produced by a team of scientists from Pennsylvania State University and KU Leuven in Belgium can only model as much as 45 percent of the thousands of facial features determined by the complex genetic palette, the author concludes.

The piece warned against unnamed private companies doing DNA phenotyping, since their proprietary software and methods had not been published in scientific journals for peer review, the author contends.

Another concern is potentially racial profiling. The author consulted Bridget Algee-Hewitt, research fellow at Stanford University and incoming professor of anthropology at Florida State University. Algee-Hewitt, who spoke at length with Forensic Magazine earlier this year, told the journal that the 13 DNA loci currently evaluated for law enforcement purposes contain racial information that could be misleading, considering the complexity of ancestry and popular associations with what comes with those identities.

“Likely of African origin” may mean something – and look something – completely different than what we perceive to be “African-American” or “black,” Algee-Hewitt told Cell.

“If you allow ancestry information to be extracted from the (DNA profile)… then you can start identifying individuals based on features that you might think are associated, or know statistically to be associated, with a particular ancestry group,” the anthropologist told Cell.


Image Caption: Predictive sketches of what the killer might have looked like at age 25 and age 55. (Courtesy of the Aurora Police Dept.)