In November 2005, Michael Lejeune pleaded guilty to murder, was convicted upon his plea, and was sentenced to imprisonment for life without the possibility of parole. Years later, Lejeune filed a petition for a writ of habeas corpus, alleging that his plea was invalid because, he said, he never was advised that, if he instead had insisted upon a trial, he could not have been compelled at that trial to testify against himself. Following an evidentiary hearing, the habeas court denied his petition. Lejeune appeals,1 and we vacate the decision of the habeas court and remand for further proceedings consistent with this opinion.
1. To properly form the basis for a judgment of conviction, a guilty plea must be voluntary, knowing, and intelligent. Brady v. United States, 397 U. S. 742, 748 (I) (90 SCt 1463, 25 LE2d 747) (1970). See also Hicks v. State, 281 Ga. 836, 837 (642 SE2d 31) (2007). For a plea to be knowing and intelligent, the accused must have “sufficient awareness of the relevant circumstances and likely consequences” of *292his plea. Brady, 397 U. S. at 748 (I). The circumstances and consequences of which the accused must be aware include the essential constitutional protections that the accused would enjoy if he instead insisted upon a trial, protections that he waives by pleading guilty and consenting to judgment without a trial. See Schneckloth v. Bustamonte, 412 U. S. 218, 238 (II) (C) (93 SCt 2041, 36 LE2d 854) (1973) (“Guilty pleas have been carefully scrutinized to determine whether the accused knew and understood all the rights to which he would be entitled at trial, and that he had intentionally chosen to forgo them.” (Footnote omitted.)). See also Loyd v. State, 288 Ga. 481, 485 (2) (b) (705 SE2d 616) (2011). As the United States Supreme Court explained in Boykin v. Alabama, 395 U. S. 238 (89 SCt 1709, 23 LE2d 274) (1969), among these essential protections is the constitutional privilege against compulsory self-incrimination. See id. at 243. In this case, Lejeune alleged that his plea was invalid because no one advised him of his privilege against self-incrimination.2
In its order denying the petition for a writ of habeas corpus, the habeas court proceeded from the premise that the Warden had the burden of proving that Lejeune entered his guilty plea voluntarily, knowingly, and intelligently, and to carry that burden over the allegations of the petition in this case, the habeas court reasoned, the Warden had to show that Lejeune understood at the time of his plea that, if he had insisted upon a trial, he could not have been compelled at trial to testify for the prosecution. The habeas court concluded in the end that the Warden carried that burden. As a basis for its conclusion, the habeas court appears to have relied in significant part on the fact that the prosecution of Lejeune spanned several years, and it involved numerous pretrial hearings, a trial by jury in March 2005 that ended in a mistrial, and a second trial by jury in November 2005 that was underway when Lejeune pleaded guilty. Whether or not Lejeune was advised of his privilege against self-incrimination at or in connection with the proceeding in which he entered his plea, the habeas court found that he already had an adequate understanding of the constitutional privilege by virtue of earlier events in the course of his prosecution. On appeal, Lejeune contends that a number of these earlier events on which the habeas court relied do not actually support its finding, and we agree.
*293For instance, the habeas court pointed to three pretrial hearings in which Lejeune (through counsel) asserted his privilege against self-incrimination. But as Lejeune notes, a pretrial hearing is not a trial, and without more, the assertion of the privilege in a pretrial hearing would not necessarily put an accused on notice that he would enjoy the same privilege in other sorts of proceedings, such as a trial. See Wilson v. Kemp, 288 Ga. 779, 780 (727 SE2d 90) (2011) (it is the privilege at trial against self-incrimination that is significant for purposes of a guilty plea). See also Campos v. State, 292 Ga. 83, 85 (734 SE2d 359) (2012); Adams v. State, 285 Ga. 744, 746 (1), n. 3 (683 SE2d 586) (2009). But see Hawes v. State, 281 Ga. 822, 825 (642 SE2d 92) (2007). Moreover, the first of these hearings was in December 2001, and the others were in June 2003. Lejeune did not enter his guilty plea until November 2005, nearly two-and-a-half years later. Cf. Bazemore v. State, 273 Ga. 160, 162 (1) (535 SE2d 760) (2000) (advice given to petitioner in connection with 1988 pleas did not show that petitioner was aware of his constitutional rights at the time of his 1990 plea). For these reasons, the pretrial hearings on which the habeas court relied do not support its finding that Lejeune understood at the time of his plea that, if he instead insisted upon a trial, he could not be compelled to incriminate himself.
The habeas court also relied on the testimony of attorney Brian Steel — who represented Lejeune in connection with his second trial, but withdrew prior to Lejeune entering his guilty plea — which, according to the habeas court, showed that Steel advised Lejeune of his “Boykin rights.” But as Lejeune argues on appeal, the habeas court appears to have taken a portion of that testimony out of context and misconstrued it. When Steel was deposed in connection with the habeas proceeding, he was asked on direct examination about the consultations that he had with Lejeune in connection with the second trial, especially whether he and Lejeune discussed that Lejeune could choose whether to testify at the trial. In response to these questions, Steel said:
It wasn’t ripe yet because [Lejeune] didn’t actually — it wasn’t the defense case in either trial. I don’t know. I mean, I’m very thorough with my clients. So I don’t know. I know I did discuss what we would call Boykin . . . rights with him because he never accepted a guilty plea when I was defending him.
This is the testimony on which the habeas court relied. But Steel subsequently testified that he never advised Lejeune about the constitutional rights that he would waive by virtue of a guilty plea *294because, during the time Steel represented Lejeune, “it never got that far.” In addition, Steel later testified unequivocally that he had “no memory” of advising Lejeune about his privilege against self-incrimination. And most important, Steel was asked on cross-examination: “And I believe you said on [djirect that you know that you did not discuss the Boykin rights with [Lejeune]; is that also correct?” (Emphasis supplied.) To this question, Steel replied: “I feel it’s very comfortable, yes.” From a consideration of the entirety of his testimony, it seems clear to us that Steel misspoke when he testified on direct examination that he discussed the “Boykin rights” with Lejeune in connection with the second trial. The reliance of the habeas court upon that misstatement was clearly erroneous.
2. Because these findings of the habeas court find no support in the record, Lejeune argues that the Warden failed to prove that his plea was voluntary, knowing, and intelligent. Maybe that is so, but it matters only to the extent that the Warden must bear the burden of proving these things. Beginning with Purvis v. Connell, 227 Ga. 764 (182 SE2d 892) (1971), we have held in a number of cases that, whenever a habeas petitioner alleges that the plea on which his conviction rests was not voluntary, knowing, or intelligent, the respondent bears the burden of disproving those allegations. That was the premise from which the habeas court proceeded in its consideration of the proof in this case, and it is the premise as well of the argument by Lejeune on appeal that he is entitled to habeas relief. But for the reasons that follow, we now conclude that Purvis and its progeny are based on a misunderstanding of Boykin, and they are inconsistent with the historical understanding in Georgia of the writ of habeas corpus. Accordingly, we overrule Purvis and its progeny, and we hold that Lejeune bears the burden as the petitioner of proving that his plea was not voluntary, knowing, or intelligent.
Our law appears always to have recognized a presumption of regularity with respect to the final judgments of courts of general jurisdiction, see LeMaster v. Orr, 101 Ga. 762, 764 (1) (29 SE 32) (1897), and criminal judgments of conviction were no different. See Wells v. Pridgen, 154 Ga. 397, 399 (114 SE 355) (1922). When a judgment of conviction was assailed by way of a petition for a writ of habeas corpus, our law presumed the regularity of the judgment, and it was understood that the petitioner bore the burden of overcoming the presumption. See, e.g., Gay v. Balkcom, 219 Ga. 554, 601 (134 SE2d 600) (1964); Stanforth v. Balkcom, 217 Ga. 816, 816 (125 SE2d 505) (1962); Solesbee v. Balkcom, 207 Ga. 352, 353 (1) (61 SE2d 471) (1950); Wilcoxon v. Aldredge, 193 Ga. 661, 668 (19 SE2d 499) (1942). And this Court historically treated the presumption as especially warranted when the judgment of conviction was based upon a plea of *295guilty. See, e.g., Sharpe v. Smith, 225 Ga. 52, 54 (6) (165 SE2d 656) (1969) (“Since there is a presumption in favor of the validity of a sentence, especially where based upon a plea of guilty, the burden of overcoming this presumption is upon the prisoner.” (Citation omitted.)); Dutton v. Parker, 222 Ga. 532, 533 (150 SE2d 833) (1966) (“There is a presumption in favor of the validity of sentences and this is especially true where, as here, they are based on pleas of guilty.” (Citations omitted.)). Generally speaking, this Court has held to the presumption of regularity, and we have continued in most every context to put the burden upon the petitioner in habeas to show the irregularity of his conviction. See, e.g., Humphrey v. Walker, 294 Ga. 855, 859-860 (II) (A) (757 SE2d 68) (2014); St. Lawrence v. Bartley, 269 Ga. 94, 97 (1) (495 SE2d 18) (1998); Turpin v. Todd, 268 Ga. 820, 828-830 (2) (b) (493 SE2d 900) (1997); Gaither v. Gibby, 267 Ga. 96, 97 (1) (475 SE2d 603) (1996).
But beginning with Purvis, we departed from the usual and settled rule in habeas cases in which the petitioner claims that his plea was not voluntary, knowing, and intelligent because he entered it without an adequate understanding of an essential constitutional protection, such as the privilege against self-incrimination. Just a few years before Purvis, the United States Supreme Court held in Boykin that the United States Constitution requires the State to bear the burden of showing on direct review that a plea was voluntary, knowing, and intelligent. See 395 U. S. at 242-244. In Purvis, we extended this allocation of the burden to habeas cases. Relying exclusively on Boykin, we held in Purvis that the United States Constitution forbids Georgia courts to indulge the usual presumption of regularity in a habeas case in which the petitioner contends that his plea was not voluntary, knowing, and intelligent:
We are aware of cases in this state holding that since there is a presumption in favor of the validity of a sentence . . . especially where based upon a plea of guilty, the burden of overcoming this is upon the prisoner. However, this presumption can no longer be indulged with the advent of the Boykin case ....
227 Ga. at 767 (citations omitted). We failed in Purvis, however, to acknowledge that Boykin was no habeas case — it was a direct appeal from a judgment of conviction, see 395 U. S. at 240-241 —■ and the United States Supreme Court said nothing in Boykin about the burden in habeas proceedings or the presumption of regularity with respect to final judgments of conviction. Considering that failure, the soundness of our reasoning in Purvis always was debatable. In the *296years after Purvis, although we adhered to its holding in a handful of cases, we never elaborated on its reasoning, nor did we articulate any alternative basis for its holding. Instead, in those few cases in which we mentioned Purvis, we simply restated its holding without additional comment. See, e.g., Knight v. Sikes, 269 Ga. 814, 816 (2) (504 SE2d 686) (1998); Bowers v. Moore, 266 Ga. 893, 895 (1) (471 SE2d 869) (1996); Glover v. Jones, 245 Ga. 848, 849 (268 SE2d 156) (1980); Roberts v. Greenway, 233 Ga. 473, 475 (2) (211 SE2d 764) (1975).
In the meantime, the United States Supreme Court decided Parke v. Raley, 506 U. S. 20 (113 SCt 517, 121 LE2d 391) (1992), making clear in its decision that Boykin did not, in fact, abrogate the presumption of regularity that attaches to final judgments, and nothing about Boykin requires that the State bear the burden of proving the voluntariness of a plea in the context of a collateral attack upon a final judgment:
To import Boykin’s presumption of invalidity into this very different context [of a collateral attack on a conviction] would, in our view, improperly ignore another presumption deeply rooted in our jurisprudence: the “presumption of regularity” that attaches to final judgments, even when the question is waiver of constitutional rights.
506 U. S. at 29 (II) (B) (citation omitted).3 ****8 After Parke, it was apparent that the rule of Purvis and its progeny was based on a misunderstanding of federal constitutional law. But when Parke was brought to our attention in Byrd v. Shaffer, 271 Ga. 691 (523 SE2d 875) (1999), a majority of the Court decided nevertheless to adhere to Purvis and its progeny. The majority did so, however, without any meaningful *297explanation of its decision. The majority instead just stated summarily that the respondent bears the burden in habeas of proving that a plea was voluntary, knowing, and intelligent, citing authority derived from Purvis,4 and without any citation to Parke 5 See Byrd, 271 Ga. at 692-693 (2). Justices Carley and Hines dissented, relying on Parke, and arguing that “the petitioner in a habeas corpus proceeding bears the burden of overcoming the presumption of the validity of the conviction and sentence entered on his guilty plea.” Id. at 696 (Carley, J., dissenting) (citations omitted). Since Byrd, we have reaffirmed in a number of cases that the respondent has the burden in habeas to prove that a plea was voluntary, knowing, and intelligent, but we never have articulated any basis for that rule other than our original reasoning in Purvis. See, e.g., Tyner v. State, 289 Ga. 592, 593 (2) (714 SE2d 577) (2011); Wilson v. Kemp, 288 Ga. 779, 779 (727 SE2d 90) (2011); Sentinel Offender Svcs. v. Harrelson, 286 Ga. 665, 666 (1) (690 SE2d 831) (2010); Sanders v. Holder, 285 Ga. 760, 761 (684 SE2d 239) (2009); Bullard v. Thomas, 285 Ga. 545, 545 (1) (678 SE2d 897) (2009); State v. Hemdani, 282 Ga. 511, 511 & n. 1 (651 SE2d 734) (2007); Arnold v. Howerton, 282 Ga. 66, 66 (646 SE2d 75) (2007); Hawes v. State, 281 Ga. 822, 822-823 (642 SE2d 92) (2007); Beckworth v. State, 281 Ga. 41, 42 (635 SE2d 769) (2006); Green v. State, 279 Ga. 687, 688 (620 SE2d 788) (2005); Baisden v. State, 279 Ga. 702, 702 (620 SE2d 369) (2005); State v. Futch, 279 Ga. 300, 300 (1) (612 SE2d 796) (2005); Foskey v. Battle, 277 Ga. 480, 482 (1) (591 SE2d 802) (2004); Britt v. Smith, 274 Ga. 611, 616-617 (1) (556 SE2d 435) (2001); Wetherington v. Carlisle, 273 Ga. 854, 855 (547 SE2d 559) (2001).
As we noted earlier, that reasoning is quite clearly erroneous. Throughout our history, Georgia law has presumed the regularity of final judgments of conviction, even when those judgments were challenged by way of a petition for a writ of habeas corpus. In Purvis, we departed from the presumption of regularity, but only because we concluded from Boykin that the United States Constitution required such a departure.6 56 Parke made clear that we had misread Boykin, and the United States Constitution does not, in fact, abrogate the presumption of regularity. As a rule of federal constitutional law, the rule *298of Purvis and its progeny is simply wrong, and the dissent does not even attempt to argue otherwise.
Instead, our dissenting colleagues would adhere to Purvis and its progeny as a matter of stare decisis. But even the venerable doctrine of stare decisis does not permit us to persist in an error of federal constitutional law. See Oregon v. Hass, 420 U. S. 714, 719 (II) (95 SCt 1215, 43 LE2d 570) (1975) (on questions of federal constitutional law, state courts must adhere to the decisions of the United States Supreme Court). And in any event, “stare decisis is not an inexorable command.” State v. Jackson, 287 Ga. 646, 658 (5) (697 SE2d 757) (2010) (citation and punctuation omitted). “When we consider whether an earlier decision ought to be reexamined, we consider a number of factors, including the age of the precedent, the reliance interests involved, the workability of the prior decision, and most importantly, the soundness of its reasoning.” Smith v. State, 295 Ga. 120, 122 (757 SE2d 865) (2014) (citation and punctuation omitted). We also consider the ease with which the People and their elected representatives might overrule our precedents, if they think them incorrect. See Georgia Dept. of Natural Resources v. Center for a Sustainable Coast, 294 Ga. 593, 601 (2) (755 SE2d 184) (2014) (“The doctrine of stare decisis is always important, but it is less compelling when, as in this case, the issue is the meaning of a constitutional provision. That is because it is much harder for the democratic process to correct or alter our interpretation of the Constitution than our interpretation of a statute or regulation.” (Citation omitted.)).
A consideration of the factors that inform the application of stare decisis leads to the conclusion that Purvis and its progeny ought to be overruled. First, because those precedents are based on a misunderstanding of federal constitutional law, their error is not capable of correction by the People of this State. Second, the reasoning of those precedents is quite obviously unsound, as shown by Parke. Third, when the courts speak of reliance interests in the context of stare decisis, they refer to contract interests, property rights, and other substantive rights. Jackson, 287 Ga. at 658 (5). See also State v. Hudson, 293 Ga. 656, 661 (748 SE2d 910) (2013). The rule of Purvis and its progeny implicates no such substantive rights. Fourth, the usual rule in habeas cases — that the petitioner bears the burden of proof — is more workable than the rule of Purvis and its progeny, inasmuch as, “in establishing a Boykin violation[,] the [petitioner] is the one best situated to know whether his or her rights were infringed in the earlier proceedings.” Nash v. State, 271 Ga. 281, 285 (519 SE2d 893) (1999) (citation and punctuation omitted). The only factor that points toward continuing adherence to Purvis and its progeny is the age of Purvis. But without more, that we have been wrong for many *299years and in many cases is no reason to persist in the error. See, e.g., Center for a Sustainable Coast, 294 Ga. at 601 (2) (overruling precedent that was “less than 20 years old,” and citing cases in which the Court overruled precedents that were 29 and 90 years old); Hudson, 293 Ga. at 656-657 (unanimous decision overruling 38-year-old precedent to account for more recent developments in federal constitutional law). We now overrule Purvis and its progeny.7
3. “We recognize that, given the clear, though incorrect, mandate of our overruled case law, [Lejeune] may be caught somewhat by surprise with this opinion.” Sosniak v. State, 292 Ga. 35, 40 (3) (734 SE2d 362) (2012). In the proceedings below, Lejeune and the habeas court both proceeded from the premise that the Warden had the burden of proof. That premise was wrong, and it is Lejeune who has the burden of proving that his plea was not voluntary, knowing, and intelligent. To afford Lejeune a fair opportunity to carry that burden, and to permit the habeas court in the first instance to consider the evidence with a proper understanding of the burden (and without the findings we have held to be clearly erroneous in Division 1), we vacate the decision of the habeas court, and we remand for a new evidentiary hearing consistent with this opinion.8
*300 Judgment vacated and case remanded with direction.
All the Justices concur, except Hines, P. J., Benham and Hunstein, JJ., who dissent.Lejeune timely filed an application for a certificate of probable cause to appeal from the decision of the habeas court, see OCGA § 9-14-52, and we granted that application.
Lejeune never alleged that his plea was not a voluntary expression of his own choice, that he did not understand the nature of the case against him, nor that he entered his plea without understanding the other constitutional rights that he would have been afforded at trial. His habeas petition was based exclusively on the allegation that he never was adequately advised of his privilege against self-incrimination.
To be sure, Parke involved a recidivist sentencing proceeding in which an earlier judgment of conviction fell under collateral attack, and Parke does not directly address the burden of proof with respect to the voluntariness of a plea in habeas. But like the collateral attack in Parke, a proceeding in habeas corpus “is a collateral attack that is separate and distinct from direct review, and occurs only after a prisoner has failed to obtain relief by direct appeal. It is not an extension of direct appeal: Habeas corpus always has been a collateral remedy, providing an avenue for upsetting judgments that have otherwise become final. . . .” Gibson v. Turpin, 270 Ga. 855, 857 (1) (513 SE2d 186) (1999) (citation omitted; emphasis in original). More important, in speaking of the presumption of regularity in Parke, the United States Supreme Court made clear that it was speaking of the same presumption long recognized in habeas proceedings: “Although we are perhaps most familiar with this principle in habeas corpus actions, it has long been applied equally to other forms of collateral attack.” Parke, 506 U. S. at 29-30 (II) (B) (citations omitted; emphasis supplied). A number of other courts have recognized that the holding of Parke extends to habeas proceedings. See, e.g., Skaftouros v. United States, 667 F3d 144, 158 (B) (1) (2d Cir. 2011); Little v. Crawford, 449 F3d 1075, 1080 (III) (A) (9th Cir. 2006); United States v. Custis, 988 F2d 1355, 1363 (III) (4th Cir. 1993).
The Byrd majority cited Bowers, 266 Ga. at 895 (1). Like Byrd, Bowers summarily stated that the burden is on the State to show that a plea is voluntary, and Bowers cited only Roberts, 233 Ga. at 475 (2), for that proposition. Roberts relied on Purvis.
Justice Hunstein concurred, acknowledging that no constitutional law required the State to bear the burden in habeas, but accepting that allocation of the burden as reflecting “the better policy position.” Byrd, 271 Ga. at 693.
To the extent that the dissent points to the Georgia Constitution, neither Purvis nor any of its progeny suggests that the Purvis rule is one of state constitutional law.
The dissent suggests that our overruling of Purvis and its progeny is some sort of novel departure from the usual rule of stare decisis. But this Court - unanimously, in mosteases -has not hesitated to overrule longstanding precedents when it has become apparent that they are out of step with recent developments in federal constitutional law, even when our precedents were not themselves, strictly speaking, precedents of federal constitutional law, and instead were only influenced by our understanding of federal constitutional law. See, e.g., Hudson, 293 Ga. at 656-657; Sosniak v. State, 292 Ga. 35, 37-40 (2) (734 SE2d 362) (2012). The rule oí Purvis and its progeny was not just influenced by our (mis)understanding of federal constitutional law; that rule is a rule of federal constitutional law, having been adopted only because we thought that Boykin required it.
We express no opinion about the nature or quality of the evidence required to carry the burden, and in particular, the extent to which a plea transcript alone, extrinsic evidence, or some combination of the two may be sufficient. See Parke, 506 U. S. at 30 (suggesting that extant transcript that is “suspiciously silent on the question [of] whether the defendant waived constitutional rights” might be enough to overcome the presumption of regularity and make out a case that the plea was not voluntary, knowing, and intelligent). We also need not determine at this point whether we should reconsider our recent precedents holding that the failure to advise a pleading defendant of one of the “three Boykin rights” can never be deemed harmless error, precedents which appear inconsistent with an earlier decision of this Court, see Goodman v. Davis, 249 Ga. 11, 13-14 (1) (287 SE2d 26) (1982), and which also have been subject to some criticism more recently. See Tyner, 289 Ga. at 595-596 (4) (opinion of Nahmias, J.); Wilson, 288 Ga. at 781-782 (Carley, P. J., dissenting). To reach and decide the merits in this case, the dissent would sweep away Goodman as an “anomaly,” without any discussion of the soundness of its reasoning (including its conformity to the decisions of the United States Supreme Court) or the other factors that inform the application of stare decisis as they relate to Goodman, a precedent that has been on the books for nearly as long as Purvis. The inconsistency of our more recent cases with Goodman will have to be resolved some day, and perhaps in this case, but that is better done on a record following a hearing at which everyone understood the proper allocation *300of the burden. We remand for development of such a record, and we decline at this point to further address the apparent inconsistency between Goodman and the other cases on which the dissent relies.