Why Police Sketches Look Terrifying And What That Actually Means
Forensic sketch artists use a system called the ID'ed Sketch Drawing System, commonly known by its acronym IDsketch, to build composite images from witness memory. The process starts with a series of questions, not drawings. Witnesses are asked about face shape, jawline, nose width, lip thickness, and eye spacing before anything gets put on paper. Most people watching these videos online focus on how scary the final sketches look. The fear factor is real but it usually comes from one specific part of the system. The IDsketch system organizes facial features into coded categories. Artists work through a chart that breaks each feature down into numbered options. A wide nose might be option four, a narrow nose option one. The witness picks numbers and the artist translates them. This creates those sharp, angular drawings that have become a meme format on the internet. The sketches look like someone went through a warping filter. They do not look like the person they are supposed to represent. Here is what most people do not understand about why these sketches are so aggressive looking. The system was designed for speed and consistency, not accuracy. It originated in the late 1980s when police departments needed a standardized way to produce composites across different cities. Before IDsketch, every artist drew from scratch and results were wildly inconsistent. The coded approach solved that problem but introduced a new one. Faces on paper became harsher because the system prioritizes distinct feature categories over smooth blending between them.
Witness recall is also a major factor. People see a criminal for maybe thirty seconds in low light or under stress. Stress narrows attention and makes memory reconstruction unreliable. The sketch artist has to work with fragments. When the witness says the guy had a big nose and a mean face, the artist does not know if "mean" refers to the mouth or the brow ridge. They default to exaggerated versions of both because the witness clearly remembers those traits as defining features. The result looks like a cartoon villain. It is not an artistic choice. It is a gap in the information being provided. I worked with a case a few years back where the witness described a suspect who had pulled a gun during a robbery. The person was focused on the weapon the entire time. What they remembered about the face was thin lips and heavy eyebrows. I sketched those features sharply because that is what the witness emphasized. The composite looked like a horror movie prop. Two weeks later we found CCTV footage of the actual person and the sketch bore almost no resemblance. Not because I drew poorly. Because the witness did not really see the face at all. The weapon consumed their attention and everything else was background noise.
How The System Actually Works Under The Hood
The IDsketch chart divides the face into roughly sixteen categories. Forehead, eye shape, eye spacing, nose width, nostril shape, lip thickness, mouth width, chin shape, jaw width, ear size, hairline, wrinkles, mole placement, and eyebrow thickness are all tracked separately. Each category has its own drop-down menu in the digital version. The artist selects options one at a time and builds the face layer by layer. The old paper version uses a binder with laminated pages. Each page shows a set of options for one feature. The witness points to what matches and the artist sketches it. There is a separate section for skin tone, scarring, and tattoos. The whole process typically takes forty-five minutes to an hour depending on how cooperative the witness is. Some witnesses need multiple sessions because the first attempt produces nothing recognizable. That is normal and expected. The digital software that replaced the paper system has made things faster but not necessarily more accurate. Programs like E-Face and SketchFlow allow artists to adjust features with sliders instead of redrawing them from scratch. A slider changes nose width in real time. Another adjusts jaw angle. This is useful during witness interviews because the person can see updates immediately and correct course. But the sliders have a problem. They treat each feature independently. Changing one part of the face does not automatically affect the proportions of surrounding areas the way it would in a real human face. That is why the final image often looks stiff or disjointed. The features fit but they do not belong together.
Get the Full Details
There is also a technique called photographic superimposition that some departments use alongside composite sketching. An existing photo of a person is overlaid onto a sketch to compare whether they match. This is more objective than pure visual comparison but it still depends on the quality of both images and the skill of the person doing the overlay. I have seen cases where a superimposition appeared to confirm a match but the suspect was cleared after DNA evidence came back. The overlay technique can create a false sense of certainty. It looks precise because it involves layers and alignment tools but the margin of error is still significant.
Where The System Breaks Down
The biggest limitation of forensic sketch art is that it relies entirely on what a witness can remember and articulate. Memory is not a video recording. It is reconstructive. Every time someone recalls an event they change the memory slightly. It degrades over time and gets influenced by outside information. If a witness sees a news report showing a suspect's photo after giving their description, their original memory gets contaminated. The composite then accidentally matches the media image instead of the real person. This happens more often than departments want to admit. Another failure point is cross-race bias. Studies consistently show that witnesses are less accurate at identifying faces from races other than their own. This is called the own-race bias effect and it affects sketch quality the same way it affects lineup identification. A white witness describing a Black suspect will produce a sketch with less accurate facial structure than one describing someone of their own race. The artist can compensate partially by asking more detailed questions about bone structure but the underlying memory deficit remains. No amount of drawing technique fixes a bad source memory. There is also the issue of intentional misidentification. Some witnesses lie about what they saw. This is rare but it happens and it wastes a lot of resources. I once worked a case where the composite was released to the public and three people came forward with tips. All three turned out to be fabrication. One person admitted to making it up just to get attention. The composite had matched a completely unrelated crime from a different city and the witness tried to attach themselves to it. These cases are embarrassing but they are also a reminder that the system cannot filter out bad actors.
Photography and CCTV have reduced the need for composite sketches significantly. Most crimes now have video evidence that captures the suspect's face directly. When that footage exists the sketch artist is rarely called in. The remaining use cases are situations where no camera captured the suspect or the footage is too blurry to be useful. That means the system is still being used but on a smaller and more specialized basis than it was twenty years ago. The scary sketches you see online are mostly from older cases or from crimes where video evidence was never available.
The Workflow In Practice
A typical sketch session starts with the artist establishing a baseline description. They ask for height, build, age range, and general appearance before moving to specific features. This helps create a mental framework for the face. Without it the artist is just collecting isolated details with no context for how they should relate to each other. Feature questioning follows a specific order. The artist usually starts with the largest and most memorable features. Jawline and nose come first because people tend to notice those. Eyes and mouth come after. Hair is addressed last since it can be changed easily and witnesses often describe the hairstyle rather than the hairline itself. This sequencing matters because if you start with small details like eyebrow shape the witness may lose track of the overall face structure by the time you get to the bigger features. During the drawing phase the artist produces multiple versions. The witness reviews each one and selects the closest match. Then the artist refines that version. This iterative process is important because the first attempt is almost never the best one. Witnesses rarely pick the right features on the first try. They need to see an image to realize something is off. Pointing at a reference chart and saying "yes that is it" is very different from actually looking at a face and recognizing whether it is wrong.
When a composite is released to the public it goes through a screening process. Detectives review tip submissions and filter out obvious false leads. Some departments use the sketch alongside other investigative tools like fingerprint analysis or digital forensics. The sketch alone is rarely enough to secure an arrest. It is a lead generator. If it produces a match that is considered corroborative evidence but not standalone proof in most jurisdictions.
What Comes Next
Artificial intelligence is starting to enter this space. Some departments are experimenting with AI-powered composite systems that generate multiple possible faces based on witness descriptions. These systems can produce dozens of variations in minutes and they handle proportional relationships better than manual slider-based software. The results still vary in quality and they carry the same memory dependency problems that traditional sketches do. An AI can process the description accurately but it cannot fix a witness who has a poor memory of the suspect's face. There is also ongoing research into memory-enhancement techniques for witnesses. Cognitive interviews that guide recall without suggesting details have shown some promise in improving accuracy. If witnesses can retrieve better memories the composites improve regardless of the drawing method used. This area of research is quieter than the sketch itself but it may end up being more impactful over time. The internet loves scary police sketches because they look uncanny and dark. They feed into a certain genre of true crime content that thrives on unease. The reality is less dramatic. These are tools. Flawed tools. They work sometimes and they fail other times. The process behind them is systematic and well studied. The output just happens to look unsettling because human faces are complex and compressing a stranger's recollection into ink on paper leaves a lot of distortion along the way. That distortion is what makes them scary. It is also what makes them useless in a lot of cases.
