A man is arrested for a robbery he did not commit. He is told that if he goes to trial and loses he faces ten years, but if he admits the lesser charge today he can go home in two. He thinks there is a one in three chance a jury convicts him. What should he do? Nearly everyone who hears the numbers says the same thing, and it is not "fight for the truth".
That small, ugly calculation is the hidden centre of criminal justice in many countries. A first-year criminal law course teaches the two halves behind it. Substantive criminal law asks what counts as a crime and what mind a person must have had. Procedure asks how the state may investigate, charge, try and convict, and what happens when it gets it wrong. A graduate keeps seeing both at once: a rule about guilt is only as good as the way guilt is proved.
This piece describes how legal systems work. It isn't legal advice, and laws differ by country and change. If you are facing a charge, speak to a qualified lawyer where you live.
Every society has the same problem. Some harms are too serious to leave to the victim and the offender to sort out, so the community, through the state, takes over: it names the conduct, finds the person, and decides on a penalty. That brings two dangers of its own. Power to punish can be turned on enemies, and any process can convict the wrong person. So every modern system answers the same five questions. What conduct is forbidden? What must be proved about the person's mind? What excuses or justifications are allowed? Who investigates and judges? And how sure must the judge be?
The first answer is nearly universal: no crime without a prior law, in Latin nullum crimen sine lege. You cannot be punished for something that was lawful when you did it. The US Constitution bans ex post facto criminal laws (Article I, sections 9 and 10). Germany's Basic Law says an act is punishable only if the law defined it as an offence before it was committed (Article 103(2)). Argentina's constitution (Article 18) says no inhabitant may be punished without a trial based on a law in force before the offence, and Mexico's (Article 14) forbids imposing a penalty by simple analogy. The European Convention on Human Rights (Article 7) and the International Covenant on Civil and Political Rights (Article 15) say the same for dozens of countries. The rule is the reason criminal statutes must be written down and precise.
| Family | How crime is written down | How the trial is run |
|---|---|---|
| United States (common law; England and Wales are the source) | Mostly statutes today, but built on common-law ideas such as mens rea; 50 state codes plus federal law. The Model Penal Code (1962) is an influential, non-binding template. | Adversarial: prosecution and defence present their cases to a neutral judge and, for serious crimes, a jury. About nine in ten convictions come from guilty pleas. |
| Latin America (civil-law codes: Mexico, Argentina, Brazil, Chile, Colombia) | A penal code per country, with constitutional rights written into the constitution. | Inquisitorial, file-based procedure for most of the 20th century, replaced since the 1990s by oral, accusatory trials. Brazil keeps a jury for intentional killings. Mexico, Chile and Colombia have all rewritten procedure. |
| Continental Europe (France, Germany, Spain; the Council of Europe) | Codes. France's current penal code dates from 1994; Germany's from 1871, much amended. The European Convention applies across the continent. | Judges play an active role in finding the facts, and ordinary people sit as lay judges or jurors beside professionals in serious cases. Limited forms of negotiated justice have been added since the 1980s. |
| Mixed and other systems (Scotland, South Africa, Louisiana, Quebec; Islamic and customary law) | Blends of the above. Some states apply criminal rules drawn from religious law; customary systems handle many local disputes. | Varies. Treat each on its own terms rather than as a variation on one of the others. |
Real systems drift towards each other. The "adversarial versus inquisitorial" split is a useful map, not a wall. England, Germany and Mexico all now claim to guarantee a fair, public, reasoned trial; they simply give the judge, prosecutor and defence different jobs inside it.
Most crimes need two things at once. Actus reus is the guilty act (or sometimes a failure to act where there is a duty): the conduct, the circumstances, often a result such as a death. Mens rea is the guilty mind: what the person knew, meant or risked. The old slogan, actus non facit reum nisi mens sit rea, says an act does not make a person guilty unless the mind is guilty too. Shooting someone who stepped in front of a hunter's rifle at the last second is not a murder.
Law does not treat "a guilty mind" as one thing. The American Model Penal Code ladders it into four levels: acting purposely, knowingly, recklessly (consciously running an unjustified risk) and negligently (failing to see a risk a reasonable person would). Continental codes draw a similar line between Vorsatz in German and dolo in Spanish (intent, including accepting a likely result, called dolus eventualis or dolo eventual) and Fahrlässigkeit or imprudencia (negligence). Higher states of mind are punished more; some low-level offences ask for none at all and are called strict liability.
A simplified ladder in the style of the Model Penal Code. Real homicide offences differ by country (many systems treat extreme recklessness as murder) and each also needs an act that causes a death. Illustrative, not a statement of any one law.
Look at what the sliders show. The same death is "no crime", a negligence offence, manslaughter or murder depending only on what was in the head of the person involved, which is why so much of a criminal trial is the proof of a state of mind from outward facts.
Defences come in two kinds, and lawyers keep them apart. A justification says the act was right, or at least permitted: self-defence, defence of others, necessity. An excuse says the act was wrong but the person is not blameworthy: insanity (the English M'Naghten rules of 1843 are the classic test; countries now use a range of formulas), duress, and in many systems youth. German doctrine makes this structure explicit in three steps: a defined offence (Tatbestand), then no justification (Rechtswidrigkeit), then personal guilt (Schuld). Common-law courses reach the same distinctions with less tidiness.
Picture a dispute about what happened. One way to find out is to let the two sides fight it out in front of a referee who does not take sides: this is the adversarial idea, from England and now the United States. The other is to appoint an official to find out the truth and report to the court: this is the inquisitorial idea, developed on the European continent from the medieval church courts and the state courts that followed. In the adversarial picture the parties own the case. In the inquisitorial picture the court owns it.
No country is pure. The comparative-law scholar Mirjan Damaška, in The Faces of Justice and State Authority (1986), argued that what differs is less the label than how authority is organised: whether power is concentrated in a hierarchy of officials or spread among competing parties. Try the roles in three common settings.
Simplified outlines of the usual pattern in serious cases. Details differ by country, court and offence. Latin American systems differ among themselves; the reformed pattern shown is the one adopted by Chile, Colombia and Mexico.
Notice the trade. In a contest, a well-funded defendant can probe the state's case hard, and an under-funded one can be steamrolled. In an inquiry, the court is meant to look for exculpatory evidence as well as incriminating evidence, but a judge who has read the entire file may have formed a view before the trial begins. Many Latin American reformers took exactly that second worry to heart, as the next sections show.
The United States Constitution promises a jury trial. In practice, the Sentencing Commission's figures for the federal courts show that in fiscal year 2023 about 97% of the 64,124 offenders sentenced had pleaded guilty, and only about 3% went to trial. State courts are less well counted, but the figure is commonly put near 90%. The Supreme Court has held that prosecutors may offer lower charges or sentences for a plea, and may threaten harsher charges if a defendant refuses (Bordenkircher v. Hayes, 1978). Critics call this a "trial penalty": defendants who exercise their right to trial risk much heavier sentences. Defenders say plea deals save scarce court time and give both sides certainty.
Other systems face the same pressure and respond differently. Italy introduced patteggiamento (negotiated sentence, with a reduction of up to a third) in its 1988 code. France created the comparution sur reconnaissance préalable de culpabilité ("plaider-coupable") in 2004, and by 2009 about 11% of its correctional-court decisions went that way. Germany, which long insisted that a court must find the truth itself, regularised agreements in 2009 (section 257c of the Code of Criminal Procedure), and in 2013 its Federal Constitutional Court criticised informal "deals" that skipped the safeguards. Latin American reforms generally added abbreviated procedures too; Argentina's juicio abreviado is a well-known example.
Expected trial sentence = chance of conviction × sentence; an acquittal counts as zero. Illustrative numbers, and it ignores risk-aversion, record and other costs. If the offer is larger than the trial sentence the tool caps it.
With the default numbers (a 30% chance of conviction, ten years, a two-year offer) the expected cost of trial is 3.0 years, so pleading looks better unless the chance of conviction is under 20% (2 ÷ 10). A defendant who is truly innocent but cannot be confident of a jury faces the same arithmetic. That is the problem, and it is why criminal-procedure scholars worry as much about the incentive as the outcome.
For most of the 20th century the typical Latin American criminal case was a written file passed between an investigating judge, a prosecutor and a clerk. The judge investigated, then often decided on the paper record; a defendant might never see the judge in a hearing. A wave of reforms, beginning in Central America in the 1980s and 1990s and reaching the largest countries last, replaced this with accusatory systems: the prosecutor investigates, a separate judge guards rights during the investigation, and a different judge or panel decides at an oral, public, adversarial trial.
The details are best learned country by country. Chile's new code was introduced in stages from 2000 and was complete by about 2005 to 2006. Colombia's 2004 law began operating by judicial district in 2005. Mexico's constitutional reform of 18 June 2008 set an eight-year transition, a single national code of criminal procedure was published on 5 March 2014, and the deadline fell in 2016. Argentina's federal accusatory code (Law 27,063, 2014) was held back for years and has been switched on district by district since 2019, beginning with Salta and Jujuy; the rollout has continued since, with some dates moved. Mexico's own constitution now states that criminal proceedings "will be oral and adversarial" (Article 20).
Dates are for the national reforms named in the text; individual states, provinces and districts moved at different times, and the dates for Colombia and Argentina are the start of phasing in, not completion.
Why do it? Reformers argued that the old system was slow, secret and left suspects in jail awaiting trial; open hearings let the public see how decisions were reached. Critics and evaluators have found mixed results. Trials are faster and more transparent in many places, but new systems have struggled with training, funding and with a heavy reliance on the same plea deals the reform was supposed to limit. The honest summary is that a procedural code can change what happens in a courtroom much faster than it can change a country's policing and prisons.
Every system will sometimes convict the wrong person. How often is unknowable, because the cases we learn about are only those that get overturned. In the United States, the National Registry of Exonerations listed 3,698 known exonerations since 1989 as of June 2025. In its first 873 cases, mistaken eyewitness identification played a part in 43%, perjury or false accusation in 51%, and false confessions in 16%. In a famous 2014 study in the Proceedings of the National Academy of Sciences, Samuel Gross and colleagues estimated that at least 4.1% of defendants sentenced to death in the US would be exonerated if they stayed on death row indefinitely, and the authors warn that this cannot be extended to other crimes.
Wrongful convictions are not only an American problem. Statistical error has produced them too. In 1999 an English solicitor, Sally Clark, was convicted of murdering her two baby sons after a paediatrician told the jury that the chance of two cot deaths in a family like hers was "1 in 73 million" (he had squared a single-death figure, wrongly treating the two deaths as independent). The Royal Statistical Society protested, and the convictions were quashed in 2003, after it emerged that evidence of infection in one baby had not been disclosed. The deeper error is the prosecutor's fallacy: confusing "the chance of this evidence if the person is innocent" with "the chance the person is innocent, given this evidence".
Assumes the source is in the pool, everyone in the pool was equally likely beforehand, and the match is the only evidence. Real cases carry other evidence; the point is the shape of the arithmetic. Illustrative.
A one-in-a-million match found by searching a city of a million people leaves roughly even odds that the person matched is the source, not a million to one. Reformers' answers are practical rather than philosophical: record interrogations, run lineups so that the officer does not know which person is the suspect, require corroboration of informants, preserve biological evidence, and give convicted people a real route to review. Many of these ideas come from psychology research on memory and interrogation (see Cognitive Psychology, 35/22).
Criminal law decides when the state may punish; criminal procedure decides how it must prove it. Guilt needs an act and a mind, and the state must have named the crime first. Systems differ most in who runs the trial, a referee between contesting sides or a court that inquires, and most real cases are decided before a trial through a plea. Latin America has spent thirty years moving from file to courtroom, and every system still has to guard against the persuasive, wrong conviction.
Try this: next time you read about a verdict, ask three questions. What did the prosecution have to prove about the person's mind? Was it a trial, or a plea? And how many other people could have fit the evidence?
Sources. US Sentencing Commission, 2023 federal sentencing statistics (97.2% pleas of 64,124 cases); Bordenkircher v. Hayes, 434 U.S. 357 (1978); National Registry of Exonerations (3,698 as of 28 June 2025; percentages from the first 873 cases, per its Wikipedia summary); Gross, O'Brien, Hu & Kennedy, PNAS 111(20), 2014; Sally Clark case (convicted 1999, quashed 2003); Basic Law Art. 103(2); Argentina's constitution Art. 18; Mexico's constitution Arts. 14 and 20; Brazil's constitution Art. 5 XXXVIII; Mexican reform of 18 June 2008 and the CNPP of 5 March 2014; Argentina Law 27,063 (2014) and the federal rollout from 2019. Plea bargaining in France, Germany and Italy: Wikipedia summary of national codes. Model Penal Code (American Law Institute, 1962). Damaška (1986). Dates for Chile and Colombia are approximate and come from a World Bank-style summary of the regional reforms.