1,218 Retrial Applications Concluded in Five Years and 7 Orders to Reopen — 115 of the 121 Dismissed in 2023 Had No Examination of Facts at All

On 10 August 2026, NHK reported on the state compensation case arising from the Matsubase incident. In a killing in Kumamoto 41 years ago, a man acquitted at retrial after serving his sentence had his family sue the state and the prefecture; the appellate court found the prosecution's conduct and the police investigation unlawful, and the prefecture has decided not to appeal.

The system itself is also moving. The Legislative Council's subcommittee on criminal law (retrial) has met from its first session on 21 April 2025 through its eighteenth on 2 February 2026[1].

The subject is mandate 129: “In light of recent circumstances surrounding criminal retrial procedure, and from the standpoint of ensuring that the procedure functions properly as an extraordinary remedy, the Council is asked for its views on the rules governing inspection and copying by defence counsel of records held by prosecutors and not submitted to the court in retrial applications, the rules on appeals against orders to reopen a case, the rules on the exclusion and recusal of judges in retrial applications, and other rules of criminal retrial procedure.”[2]

Seven orders to reopen in five years

Statistics prepared by the Supreme Court General Secretariat were circulated at the subcommittee's first meeting. They show how the system runs.

Retrial applications concluded came to 230 people in 2019, 213 in 2020, 254 in 2021, 252 in 2022 and 269 in 2023[3]. That is 1,218 over five years.

The same table gives the number that produced an order to reopen. Four in 2019, none in 2020, one in 2021, none in 2022 and two in 2023[3]. Seven in five years, 0.6% of those concluded.

Six of the seven came from summary courts

Where they arose is given too. Six in summary courts, in 2019 (four), 2021 (one) and 2023 (one), and one in a district court, in 2023[3]. A note adds that no order to reopen was made in any high court or in the Supreme Court[4].

In five years, no high court and no Supreme Court bench ordered a case reopened. District courts managed one.

The weight of the cases differs. Summary courts have first-instance jurisdiction over offences punishable by a fine or less, offences for which a fine is an alternative penalty and certain other offences, and as a rule cannot impose imprisonment; for a limited set of offences including theft and embezzlement they may impose up to three years[5].

The heavier the sentence that became final, the more closed the door to a retrial. One of the seven came from a district court; the other six from summary courts.

The scale of the convictions being reviewed

The same table gives a comparison. First-instance convictions in ordinary proceedings ran from 51,674 people in 2019 to 44,309 in 2023, and convictions in summary proceedings from 199,764 to 160,617[6].

Against more than 200,000 convictions a year, around 250 retrial applications are concluded and between zero and four orders to reopen are made. Four orders of magnitude apart. A system that weights the finality of judgments will produce a ratio of this kind; the problem lies inside it.

The average time to disposal is under a year

Duration is published as well. In 2023, average time to disposal was 8.8 months in district courts, 8.0 months in high courts and 6.4 months in summary courts[7].

Within the district court figures, of 170 concluded, one took more than three years, 16 took up to three years and 16 up to two years, leaving 137 concluded within a year[7].

Statistically, retrial applications do not take long. The "prolonged proceedings" named in the explanation of the mandate refers to a small number of cases outside this average.

It is short because nothing is examined

The reason the average is short appears in another table.

Of 121 people whose applications were dismissed solely on the ground that they fell outside Article 435 of the Code of Criminal Procedure, six had an examination of facts and 115 did not[8].

95.0% ended without their facts being examined.

What happened in the six is limited as well. By method, no witness was heard and no applicant was heard; all six were classified as other, meaning expert opinion or inspection[8].

Among 121 people, not one case called a witness, and not one heard from the person who filed. The average of 8.8 months sits on top of that.

What the dismissals were for

The grounds are listed. No final judgment: 50. Neither obviousness nor novelty: 42. No obviousness: 28. No ground for retrial asserted: 16. No novelty: 9. Claim fails on its own terms: 6. Where several grounds applied, the case is counted in each[8].

The largest group, 50 people, had no final judgment at all. They sought a retrial where no conviction had become final, which is to say they were at the wrong door.

The ministry's explanation put the same point first. “As for retrial applications, a considerable number are inadmissible or fail on their own terms, while only a very small proportion require substantive examination; in considering the shape of the retrial system, it has been pointed out that this reality must be given due weight.”[9]

That explanation matches the statistics. Adding the applications that need no substantive consideration (50 with no final judgment, 6 failing on their own terms and 16 asserting no ground) gives 72. Because a case with several grounds is counted more than once the true number is lower, but a substantial share of the 121 sits here.

Argue from the average and both things disappear

Here is the problem with the numbers.

Look only at "8.8 months" and the system runs fast. Look only at "seven in five years" and the door is shut. Both come out of the same tables, and they refer to different cases.

The denominator contains a substantial number of applications filed at the wrong door. An average that includes them says nothing about cases needing substantive examination. Equally, seven orders is not by itself evidence of a narrow door if most of the denominator never warranted consideration.

What is needed is a figure covering only the cases that require substantive examination. No such figure is published.

Settling in advance what will count as an effect is the method covered in designing outcome indicators（このサイトの記事）.

3.456 million consultations over ten years of support for the needy（このサイトの記事） showed the same shape: entries are counted, and what happened to those who branched off at the entrance exists only as an estimate. A system that does not record its entrance decisions cannot measure its own effect.

Where the three issues in the mandate bite

All three examples in mandate 129 bear on the side that requires substantive examination.

Evidence disclosure, the first, concerns whether defence counsel may see records held by prosecutors and not submitted to the court. Without new evidence there is no way to argue either obviousness or novelty under Article 435(6). In 2023, 79 people were found to lack one or both (42 plus 28 plus 9).

Appeals against orders to reopen, the second, arise only after such an order is made. With seven in five years the population is minute, though the Matsubase case shows that years pass on that side.

Exclusion and recusal of judges, the third, governs who hears the application.

All three concern the 5%, not the 95%. None of them reaches the 115 people whose applications end without any examination of facts.

What 41 years buys

Matsubase sits at the long end. Forty-one years from the incident, a prison term served, an acquittal at retrial, a state compensation suit finding the investigation and the prosecution's conduct unlawful, and finally a decision not to appeal.

The structure of hostage justice（このサイトの記事） dealt with the stage before conviction. A retrial is what follows once conviction is final. Correcting at the exit what was not prevented at the entrance takes forty-one years.

What has to be counted before the next decision

The published statistics show the number of applications, the grounds for dismissal and the time taken. Two things are missing.

The first is how many cases the courts judged to require substantive examination. The ministry calls it a very small proportion without saying how many. Without that denominator, seven is neither narrow nor reasonable; it is simply seven.

The second is how often disclosure was sought and how often it was granted. The first issue in the mandate turns on this, and because no rule currently governs it, no statistic exists.

Eighteen sessions have gone into whether to change the system, and the figures needed to measure the effect afterwards have not been prepared. If an amendment comes, what to count should be settled with it.

Further Reading

• Enzai to Saiban (Wrongful Conviction and the Courts)（外部サイト、新しいタブで開きます） by Kaku Imamura (Kodansha). A defence lawyer who has secured a large number of acquittals sets out, case by case, how wrongful convictions are produced and why they are not corrected. Useful for seeing what happens outside the statistics.

References

Statistical Materials (handout 2 to the first meeting of the Legislative Council subcommittee on criminal law (retrial)) — Supreme Court General Secretariat (2025). Ministry of Justice

Minutes of the First Meeting of the Legislative Council Subcommittee on Criminal Law (Retrial) — Ministry of Justice (2025). Ministry of Justice

Legislative Council: Subcommittee on Criminal Law (Retrial) — Ministry of Justice (2026). Ministry of Justice

Matsubase State Compensation Case: Kumamoto Prefecture Will Not Appeal, Judgment Against the Prefecture to Become Final — NHK (2026). NHK NEWS WEB