The burden and standard of proof of the affirmative defence of limitation, and some aspects of the affirmative defence of withdrawal, were considered in Smith v United States, USSC No 11-8976, 9 January 2013.
Subject to any statutory definition to the contrary, an affirmative defence is one that affirms that even though the actus reus and the mens rea may have been proved, nevertheless the defendant is not guilty.
Examples of affirmative defences are infancy, insanity, self-defence, coercion, and limitation.
This last one, limitation, amounts to the defendant saying, "Yes, I committed the offence, but that was so long ago that the law prevents me being charged." Not all offences are limited in this way, and limitation periods are established by legislation. Terminology can be a bit confusing, so when I say the offence was inside the limitation period that means the defendant can be prosecuted, and when I say it was outside the limitation period that means the defendant cannot be prosecuted.
In Smith the relevant offences were conspiracies to commit various serious offences. A limitation period applied. The defendant relied on the limitation period, saying his participation in the conspiracies had ceased outside that period because he had withdrawn his participation. At trial the issue of withdrawal only arose as a result of a question from the jury, and the judge gave a direction that the defendant had to prove withdrawal to the standard of the balance of probabilities. The Supreme Court held that this did not violate the Due Process clause.
Placing the burden of proof of withdrawal on the defendant did not violate the right to be presumed innocent because, as the Court had said in Patterson v New York, 432 U.S. 197 (1977),
Of course legislatures could change that rule, as indeed most have, so that it is now usual for defences, whether they be affirmative or not, to be for the defendant to raise but for the prosecutor to disprove. The defendant under this usual approach has only the "evidential burden" of pointing to some evidence sufficient to put the existence of the defence into issue.
But in the eighteenth century, as Blackstone described in his Commentaries, the defendant had to prove the defence. In Smith the Court, in its unanimous opinion delivered by Scalia J (who did not on this occasion have the opportunity to indulge his appetite for the purple prose of vigorous dissent), pointed out that the burden on the defence to establish withdrawal on the balance of probabilities was justified because it was the defendant who would have the relevant evidence on the issue. The prosecutor could not be expected to prove a negative. The defence here did not seek to negative any element of the offence that the prosecutor had to prove.
To negative his participation in the conspiracies the defendant had to prove that he took steps "to dissociate from his confederates". The conspiracy was a continuing offence, and the prosecutor had proved that it continued inside the limitation period, but the prosecutor could not be expected to prove that the defendant did not withdraw before then. Participation in the conspiracy continues even if the defendant is inactive after joining it, until he withdraws or until the conspiracy terminates.
I should emphasise that the law is not necessarily the same outside the United States. The common law has developed since the eighteenth century and legislation usually reflects those developments: a defence carries an evidential burden for the defendant and then a legal burden of disproof on the prosecutor. The usual exception is insanity, where the legal burden is on the defendant to the standard of the balance of probabilities. But where the defence of withdrawal is still a common law defence the policy of placing the legal burden on the defendant, on the issue of the taking of sufficient steps to withdraw, may need some consideration.
Withdrawal cannot be effective if the offending cannot be undone. A conspiracy is complete upon agreement, and it continues while that agreement exists even if the parties to it change. Once the defendant has entered the conspiracy it is too late for him to withdraw. And in relation to attempts, it is too late to withdraw once a sufficiently proximate act has been committed. And incitement is committed when the incitement occurs – it is too late to withdraw after then. But liability can arise from assistance or encouragement, and it may be possible for the defendant to withdraw his assistance and encouragement. Liability can also arise from participation in a common purpose, and a defendant may be able to withdraw by negativing his participation. More than verbal withdrawal may be required, for example where the defendant has provided others with equipment needed to commit an offence. Even verbal withdrawal may require more than mere dissociation, and the defendant may have to try to persuade the principal offender not to continue.
In Ngawaka v R  NZCA 249 (6 October 2004) the Court of Appeal approved a statement of the law on withdrawal by Hammond J in R v Pink  2 NZLR 860:
Hammond J continued:
Update: the New Zealand Supreme Court has held that withdrawal is a common law defence and that Hammond J's points in Pink need modification: Ahsin v R  NZSC 153, especially at  per McGrath, Glazebrook and Tipping JJ.