HTLAW Law, made legible.
HTLAW / ArticlesCriminal LawJuly 22, 2026 · Hyder Talpur
Identification evidence / A study in reliability

A face remembered.
A person accused.

Between those two events, the law must test both the procedure and the memory.

A
B
C
D
E
F
Illustration only · No photograph is being identified
In this article 6 chapters + sources

A shopkeeper sees a stranger during a robbery. Weeks later, the shopkeeper picks someone from a lineup.

Does that mean the right person has been identified?

Recognition has a history
  1. SeeOpportunity at the event
  2. DescribeFirst uncontaminated account
  3. SelectA neutral identification test

Already knew the accused? Prove that acquaintance. A stranger’s identification and recognition of a known person are different questions.

Article 22 makes identity relevant. A parade supports the witness’s evidence; it is not a separate proof of guilt.
123

Not necessarily. The court needs two answers: was the lineup fair, and did the shopkeeper get a reliable view of the robber?

An identification parade is a supervised test of recognition. It is meant to prevent police suggestion, accidental selection and familiarity acquired after arrest. It does not prove guilt by itself.

Kanwar Anwaar Ali brings together the safeguards for the test. Mian Sohail Ahmed explains why the witness’s original view needs a separate examination.1

Test the procedure. Then test the memory. Explain how any defect affects reliability; do not simply count objections. The ultimate question remains proof beyond reasonable doubt.

The statutory foundation

Article 22 of the Qanun-e-Shahadat Order 1984 makes facts establishing a disputed identity relevant. It does not specify the lineup’s size, location or arrangement.2

Those details come from judgments, High Court rules and executive instructions. They help ensure that the identification was obtained fairly before the court decides how much to rely on it.

Recognition is not the same as new familiarity

A parade supports and tests the witness’s later evidence. It is not a separate eyewitness account.

If the shopkeeper already knew the accused, a parade ordinarily adds little. The prosecution must instead prove that prior acquaintance.

If the accused was a stranger, a first identification in court is generally unsafe. By then, remand appearances, statements, charge proceedings and trial may have made the face familiar.3

“I recognise that face now” does not necessarily mean “I saw that person commit the robbery.”

02 / Protect the selection

A fair parade should leave the answer to the witness.

Before the parade

Keep prior exposure visible in the record.

Record whether prior exposure occurred and how the suspect’s unseen status was protected.

Event→Later exposure→Selection
The issue is possible suggestion. This illustration does not assign a probability of error.

Safeguards that protect the parade

The underlying appeal in Kanwar Anwaar Ali involved three suspects placed in a joint identification exercise. The accused were acquitted on the doubtful evidence as a whole.

The later proceeding concerned the Magistrate. The Court accepted his explanation, discharged the notice and collected the safeguards. The Magistrate was not the criminal accused.4

Protect the first uncontaminated selection
  1. PromptExplain delay.
  2. UnseenPrevent prior police exposure.
  3. Fair lineupOrdinarily one suspect with nine or ten suitable dummies; separate suspects.
  4. Separate witnessesNo sharing answers before or after turns.
  5. Full recordObjections, mistakes, roles and precautions.
These safeguards reduce identifiable risks. Their importance is assessed against the facts, not by counting boxes.
456

The Court described identification as the “testimony of the senses of the witness”.5 That matters because an honest witness can still make an honest mistake.

Five jobs the procedure must do

  1. Hold the test promptly. Unexplained or unreasonable delay calls for caution. Memory can fade, and later information can affect it.

  2. Keep the suspect unseen. Do not show the suspect to witnesses beforehand. The safeguards reject police-station parades. Record the period of police custody; allow suspects going to remand to cover their faces if they choose.

  3. Avoid making the suspect stand out. Ordinarily, place one suspect among nine or ten suitable non-suspects, often called dummies. Hold separate parades for separate suspects.

  4. Keep witnesses apart. Someone who has finished must not tell the next witness whom they chose or where that person stood.

  5. Record the whole exercise. Identify the dummies, record objections and every right or wrong choice, ask how the selected person is known, obtain the attributed role where roles differ, describe precautions and append the prescribed certificate.6

These are safeguards against identifiable mistakes, not paperwork for its own sake. They protect a reliable prosecution identification as well as an accused person wrongly selected.

03 / Perception has conditions

Confidence at trial cannot supply light at the scene.

Distance

How far away was the face?

Light

What illumination was proved?

Duration

How long could the witness observe?

Attention

What competed for attention?

How memory affects reliability

In Mian Sohail Ahmed, the Court rejected the idea that memory works like a stored video.7

It separated two kinds of conditions:

What was available to memory?
The original view
Light & distance
Duration & delay
Stress & weapon focus
Condition & disguise
A well-run parade cannot repair a poor original view. Later confidence cannot supply missing light, time or attention.
7
  • System variables: things investigators and the Magistrate can control, such as lineup arrangement and prior exposure.
  • Estimator variables: conditions at the event, such as light, distance and the witness’s attention.

The names sound technical. The distinction is simple: a fair test cannot improve the view the witness originally had.

Reconstruct the shopkeeper’s view

Ask aboutWhat the answer helps explain
Stress and immediate dangerWhether fear narrowed attention or affected recall.
A visible weaponWhether the witness watched the weapon rather than the face.
DurationHow much time the witness actually had to observe.
Distance and lightWhether facial features could be seen under the proved conditions.
Age, injury or intoxicationWhether the witness’s condition affected perception.
Disguise or changed appearanceWhether the face later selected was fully visible at the event.
DelayHow long passed before an identification unaffected by later exposure.

A good original view does not excuse a suggestive parade either. Both parts need support.

I

Was the procedure fair?

Neutral administration, separation, suitable dummies and a complete record.

II

Was the memory reliable?

Opportunity to observe, first description, delay and intervening exposure.

The two-stage judicial test

Stage one: examine the test. Compare the first description with the person selected. Then check delay, prior exposure, suitable dummies, separate placement, witness isolation, attributed roles, objections, wrong choices and the certificate.

A vague or missing first description is not automatically fatal. But it raises a practical question: how was this particular suspect chosen for the parade?

Reliability requires both inquiries
Viewing conditions supportedViewing conditions doubtfulFair procedure
Assess weight with the whole record.
Regularity alone cannot cure a weak view.
Suggestive procedure
A good view does not erase later contamination.
Both sources of uncertainty need answers.
Conceptual map, not an outcome predictor. Muhammad Hayat illustrates reliable identification supported by close viewing, proved light and forensic evidence. An irregularity is not automatically fatal.
89

Stage two: examine the witness’s opportunity. Reconstruct distance, light, angle, duration, movement, obstruction, stress, weapons and attention from proved facts.

Compare that account with the first description and the testimony at trial. Later confidence cannot fill an earlier gap.

What failed in Mian Sohail Ahmed?

The procedure had a joint lineup, missing role attribution and police exposure of the suspects.

The original view was also doubtful: night-time conditions, an unproved light source, a distance of fifteen to thirty feet, a fired weapon and uncertain duration.

Outcome: acquittal. The Supreme Court could not safely exclude mistaken identification on that record.8

Safeguards are not a ritual

Now consider a stronger record. In Muhammad Hayat, the report was prompt, there was no prior enmity, and the assailants were seen at close range in motorcycle headlights.

Forensic evidence also linked recovered weapons to the crime. The Supreme Court accepted the identification. The appeal was dismissed; the convictions and sentences stood.9

The location issue needs care

Hayat said Article 22 does not confine a parade to jail premises or demand proof beyond realistic human capacity.

That creates a genuine tension with Kanwar Anwaar Ali’s instruction against police-station parades. It is not general permission to ignore that safeguard.

A court must examine the applicable rules and explain why this procedure avoided suggestion. In Hayat, prompt identification, close viewing, proved light and separate forensic links mattered.

Who did what?

If one stranger fired and another drove the getaway motorcycle, the witness should identify each person’s role.

A group acting together may raise a different question. Either way, the judgment must explain reliability rather than assume it.

And if the prosecution says the shopkeeper already knew the accused, prove the acquaintance. Saying so for the first time at trial does not turn a stranger into a known person.

05 / A different source

Repeating an identification does not make it independent.

Ask where each piece of supporting evidence comes from.

One account, repeated three times

One identification↓Retold to another witness↓Repeated in a later account

These accounts may share the same origin. Their number alone does not establish independent corroboration.

When corroboration really helps

Corroboration is other evidence supporting a claim. Useful support must reduce the particular risk of mistake.

A distinctive early description, reliably traced property, a forensic link, digital location evidence or an independently proved motive may help. Each item needs its own reliable foundation.

More accounts, or another source?
One exposed witness↓
Later retelling → Court identification
Shared origin: not independent confirmation.
Forensic link Reliable location Traced property↓Test the particular risk of mistake
Supporting material must itself be reliable and independent. Several doubtful items do not automatically make a strong case.
Subha Sadiq / What could these witnesses have seen?
  1. The shootingThe act that needed proving.
  2. The witnesses arriveTwo police witnesses arrived afterwards.
  3. A later selectionIt cannot supply a view of the earlier shooting.
Timing limits what a witness can establish. The appeal succeeded on the combined evidential problems, not this illustration alone. 10
10

Several doubtful items do not automatically become one strong case. A weak recovery cannot rescue a suggestive parade simply because both point at the same person.

Another account may still come from the same source

Suppose police show the suspect to the shopkeeper before the parade. The shopkeeper later points to that person in court.

That is the same witness after the same exposure, not independent confirmation.

Likewise, two witnesses who discussed their choices may agree because they influenced each other. A separate forensic link can add something that repeated accounts cannot.

A later choice cannot supply an earlier view

In Subha Sadiq, two police witnesses arrived after the shooting. Their later selection could not establish that they saw the selected person fire.

The Court also found problems with descriptions of the suspect and dummies, role attribution and possible post-arrest exposure.10

Outcome: the petition became an appeal, which was allowed. The conviction and sentence were set aside with the benefit of doubt.

This was an assessment of the combined record, not a rule that every parade error requires acquittal. The witness must first have had the relevant opportunity to observe.

06 / Reasons that can be reviewed

The judgment should reconstruct the identification.

What courts should record

The prosecution should be able to reconstruct the process from records made at the time:

Who protects which part?
  1. 01PoliceKeep the suspect unseen.
  2. 02MagistrateRun a neutral test; record wrong choices too.
  3. 03ProsecutorDisclose the complete record.
  4. 04JudgeGive separate reasons on procedure, memory and support.
Reliability protects the accused and the victim. The final assessment remains subject to proof beyond reasonable doubt.
  • Arrest and custody: when the suspect was arrested, where kept and when each witness first saw the suspect afterwards.
  • Lineup: how dummies were selected and whether separate parades were held.
  • Selection: what each witness said the person did, including objections and mistakes.

Strongly worded testimony does not repair an unexplained gap.

The defence should explain why the defect matters. Delay permits decay or contamination. Dissimilar dummies invite selection. Missing roles matter when different acts are alleged.

The judge should give separate findings on procedure, the witness’s ability to observe and independent support.

Some defects prevent meaningful reliance. Others affect weight in a case with strong independent proof. Explain which kind this is, and why.

Why this discipline matters

Each participant protects a different part of the evidence.

Police keep the suspect unseen. The Magistrate runs a neutral test. The prosecutor discloses the complete record, mistakes included. The judge tests memory in context.

Later care cannot always undo earlier exposure. The first uncontaminated memory may already have been lost.

Reliability also protects victims. Convicting the wrong person leaves the actual offender free and makes correction harder. It is not a concern belonging only to the defence.

What follows

Return to the shopkeeper. A confident selection is a reason to investigate, not a reason to stop asking questions.

Was the test fair? Was the memory reliable? Does independent evidence help?

Promptness, separation, suitable dummies, secrecy, role attribution, a complete record and proved viewing conditions all serve those questions.

Neither automatic trust nor automatic rejection is the answer. The conclusion must follow from the effect of the procedure and the original viewing conditions on reliability, measured against proof beyond reasonable doubt.

The source record

Bibliography

Cases

  • In the matter of Kanwar Anwaar Ali, Special Judicial Magistrate, Criminal Miscellaneous Application No. 183 of 2019 in Criminal Appeal No. 259 of 2018, Supreme Court of Pakistan, order dated 22 February 2019, reported at PLD 2019 SC 488. Official judgment.
  • Mian Sohail Ahmed v The State, 2019 SCMR 956. Official judgment.
  • Muhammad Hayat and another v The State, 2021 SCMR 92.
  • Subha Sadiq v The State, 2025 SCMR 50 / 2024 SCP 369.
  • Muhammad Yaqoob v The State, 1989 PCrLJ 2227.

Legislation

  • Qanun-e-Shahadat Order, 1984, art 22.

Other Materials

Footnotes

  1. In the matter of Kanwar Anwaar Ali, Special Judicial Magistrate, Criminal Miscellaneous Application No. 183 of 2019 in Criminal Appeal No. 259 of 2018, Supreme Court of Pakistan, order dated 22 February 2019, reported at PLD 2019 SC 488; Mian Sohail Ahmed v The State, 2019 SCMR 956. ↩

  2. Qanun-e-Shahadat Order, 1984, art 22. The provision is in force. Its official consolidated text is available through the Pakistan Code. ↩

  3. In the matter of Kanwar Anwaar Ali, 2019 SCP 124, paras 6 and 7 of the consolidated order. ↩

  4. Ibid, paras 3 to 9. The Supreme Court’s official judgment is available here. ↩

  5. Ibid, reproducing and approving the formulation in Muhammad Yaqoob v The State, 1989 PCrLJ 2227. ↩

  6. In the matter of Kanwar Anwaar Ali, 2019 SCP 124, consolidated safeguards at para 3. The Court described the list as illustrative of all necessary precautions and directed its circulation to criminal courts. ↩

  7. Mian Sohail Ahmed v The State, 2019 SCMR 956, paras 12 to 15. ↩

  8. Ibid, paras 5 to 8 and 16 to 17. The official Supreme Court judgment is available here. ↩

  9. Muhammad Hayat and another v The State, 2021 SCMR 92, para 4 (SC). Appeal dismissed; convictions and sentences maintained. The judgment discusses Police Rules 1934, r 26.32, and Qanun-e-Shahadat Order 1984, art 22. ↩

  10. Subha Sadiq v The State, 2025 SCMR 50 / 2024 SCP 369, paras 4 and 6-9 (SC). Petition converted into appeal and allowed; conviction and sentence set aside. Paragraph 9 reproduces the short order dated 22 April 2024. ↩