An e-fit is a facial composite built from a witness's memory, not a photograph or a biometric identification. In the "baby April" case, the released image is an investigative lead meant to prompt recognition in the woman who left an infant at a cemetery roughly three decades ago, not evidence that identifies her. Its value depends on the witness's encoding and recall, the operator's interviewing skill, and how the image is distributed and tested.
The phrase in this keyword describes a cold case that has circulated in British news reporting: an infant's remains found at a cemetery, an unknown woman sought decades later, and a computer-generated face issued to the public. The underlying science is stable and well documented, even where the specific case details are not. This explainer covers how composites are made, what accuracy research actually shows, where composites fail, and the specific, practical steps that turn a picture into a usable lead.
How a facial composite is built
A composite starts with an interview, not a sketch. A trained operator uses a structured cognitive interview to help a witness reconstruct the face in memory before any image is drawn or assembled. The operator asks the witness to describe the face feature by feature, in a fixed order, and to rate each feature's confidence.
There are two broad construction methods. In feature-based systems, the witness selects individual features, such as eyes, nose, and jawline, from libraries and assembles them. In evolutionary or "holistic" systems, the witness repeatedly views arrays of whole faces and selects the ones that look most like the target, and the system breeds new arrays from those choices. Holistic methods generally produce more recognizable composites because faces are processed as wholes, not as a bag of parts.
The operator then refines the image, adjusts shading and proportions, and produces a final still. Modern systems can also produce short video sequences that show the face turning. The output is a 2D rendering of a 3D object seen under unknown lighting, so it is inherently an approximation.
A key mechanism explains much of the error. Memory for a briefly seen stranger's face is weak and reconstructive. The witness does not retrieve a stored image. They rebuild the face from fragments, and each rebuild is influenced by expectations, other faces seen since, and the composite process itself. The composite therefore reflects the witness's memory as much as the suspect's actual appearance.
Quantified context: what accuracy research shows
Composite accuracy is modest at best, and the research is consistent on this point. Laboratory studies typically find that a composite is correctly named by someone who knows the target at rates well below the rates for a photograph, and often in the range of a few percent to roughly a quarter of attempts, depending on the method and the delay. Field studies of real cases show similar or worse performance.
Recognition rates improve when the composite is shown to people who know the suspect, and improve further when the image is shown alongside other materials, such as a description or context. They fall sharply when the witness saw the face briefly, under poor lighting, or at an angle. They also fall when the witness is asked to construct the composite months or years after the event, which is common in cold cases.
The number that matters in a cold case is not the laboratory hit rate. It is the probability that the right person sees the image, recognizes it, and contacts investigators with actionable information. That is a different quantity, and it can be improved by distribution even when the image itself is imperfect.
A counterintuitive detail: a composite that looks "too good" can be a liability. Highly polished, photograph-like composites invite overconfidence and can produce false identifications, because the witness and the public treat the image as a portrait rather than a reconstruction. Investigators often prefer a slightly generic image that cues recognition without implying certainty.
Where composites fail: edge cases
Composites fail predictably in identifiable conditions. The witness saw the face for only seconds, or saw it in motion and cannot freeze a still. The face was partly obscured by a hood, mask, or hair. The lighting was dim or strongly directional, which flattens or distorts features. The witness was stressed, injured, or intoxicated at the time of encoding.
Age progression is a separate and weaker problem. When a suspect has aged decades, a composite based on an old memory is not an age-progressed image. Age progression is a forensic art technique that estimates how a face changes with time, and it is inherently speculative. Combining an old memory with decades of aging compounds two sources of error.
Cross-race recognition is another documented weakness. People are generally better at recognizing faces of their own racial or ethnic group, and composites built across racial lines show larger error. This is a group-level statistical effect, not a statement about any individual witness.
The "baby April" framing adds a further limit. The person sought is described as having been at the grave, so the relevant memory may be of a person seen briefly, at a distance, in a cemetery. Those conditions are close to the worst case for composite accuracy. The image is therefore best treated as a low-specificity prompt, not a portrait.
Red flags to watch for in any composite release:
A composite presented as a positive identification rather than a lead. An image released without the context of when and how the witness saw the person. A composite that has been retouched or "enhanced" beyond what the witness actually described. Public pressure to name someone based on resemblance alone. Age-progressed images presented as current likenesses. Any claim that a composite "matches" a suspect without independent evidence.
Misconceptions about e-fits
The most common misconception is that an e-fit is a likeness of a real person. It is a rendering of a witness's memory, filtered through software and an operator. It can resemble the suspect and still be wrong in the details that matter for identification.
A second misconception is that a good composite is a strong evidence. It is an investigative tool that generates leads. It does not establish identity, and it cannot substitute for forensic DNA, fingerprints, or corroborated witness accounts. In a murder case, the composite's role is to find people who can provide those things.
A third misconception is that composites are obsolete because of DNA and cameras. They remain useful precisely when there is no camera footage and no database match, which is common in decades-old cases. The composite is a way to reach the one person who recognizes the face, often a neighbor, coworker, or family member.
A fourth misconception is that the public should treat a composite like a wanted poster. Resemblance is not evidence. Acting on resemblance alone can harm innocent people and waste investigative resources. The correct response to recognizing a composite is to contact the investigating authority with specific, checkable information.
Comparison: composite methods and alternatives
| Method | How it works | Typical strength | Typical weakness | Best use |
|---|---|---|---|---|
| Feature-based composite | Witness picks individual features from libraries | Fast, operator-friendly | Poor holistic likeness | Quick leads, minor cases |
| Evolutionary/holistic composite | Witness selects whole faces; system iterates | Better overall resemblance | Requires trained operator and time | Serious cases with cooperative witness |
| Forensic sketch artist | Artist draws from description and feedback | Can capture expression and character | Depends heavily on artist skill | Sensitive cases, media releases |
| Age progression | Artist estimates aging from a known photo | Useful for long-term missing persons | Speculative over decades | Missing persons, long-term fugitives |
| DNA phenotyping | Predicts traits from genetic markers | Independent of witness memory | Cannot produce a usable face; broad traits only | Investigative exclusion, not identification |
The comparison makes one point clear. No method produces a photograph. Each produces a probabilistic cue, and the cue's value depends on how it is used.
Timeline: from memory to public release
| Stage | What happens | Typical timing | Key risk |
|---|---|---|---|
| Encoding | Witness sees the person | Seconds to minutes | Poor conditions, stress |
| Interview | Structured cognitive interview | Days to weeks after | Leading questions, contamination |
| Construction | Composite assembled and refined | Hours to days | Operator skill, system limits |
| Verification | Image tested with people who know the suspect | Days | Overfitting to one description |
| Release | Image distributed to public and media | Case-dependent, often months to years | Overconfidence, false leads |
| Follow-up | Tips screened and investigated | Weeks to months | Tip volume, resource limits |
In cold cases, the gap between encoding and release can span decades. That gap is the single largest source of error, because memory decays and is overwritten by later experience. The timeline also shows why early, well-conducted interviews matter more than later image polish.
Practical actions: turning an e-fit into a lead
For investigators, the practical steps are procedural. Conduct the witness interview before showing any other faces, to avoid contamination. Use a holistic composite method where resources allow. Record the witness's confidence for each feature, and preserve the original description. Test the composite with people who know the suspect before public release, and treat a failure to recognize as informative rather than as a reason to retouch the image.
For the public, the steps are simpler and more cautious. Look at the image as a prompt, not a portrait. Ask whether you know someone whose overall appearance and context match, not whether a single feature matches. If you have specific, checkable information, contact the investigating authority directly rather than posting names on social media. Do not confront anyone based on resemblance.
Specific thresholds help. Treat a composite as a lead only when it is paired with at least one independent fact, such as a location, a vehicle, a time, or a relationship. Treat a tip as actionable only when it can be checked against records, not just against the image. And treat any public identification as premature until investigators confirm it through independent evidence.
For journalists and community members sharing the image, context is the safeguard. State when the witness saw the person, under what conditions, and that the image is a reconstruction. That framing reduces the risk of false identification while preserving the image's value as a recognition cue.
Limitations: what the image cannot tell you
The composite cannot tell you the person's name, age, or identity. It cannot establish that the person committed any act. It cannot be compared to a suspect by a member of the public with any reliability, because resemblance is not identity. It cannot be age-progressed reliably across three decades, and it cannot correct for poor encoding conditions.
The evidence base is also limited in a specific way. Laboratory studies of composite accuracy use controlled conditions and small samples, and field studies are few and often retrospective. The research supports general conclusions, such as the superiority of holistic methods and the harm of delayed interviews, but it does not support precise per-case probabilities. Any claim that a given composite has a specific percent chance of being correct is not supported by the literature.
What the evidence does support is a disciplined process. Early structured interviews, holistic construction, cautious release with context, and independent verification are the practices most likely to produce a useful lead. The image itself is the beginning of an investigation, not the end of one.
This article is general educational information about facial composite science and investigative practice, not a diagnosis, legal opinion, or identification of any person. It does not promise that any composite will lead to an arrest or conviction. For questions about a specific case, contact the investigating authority or a qualified local professional; in an emergency, contact local emergency services.