John Bilyew’s petition for habeas corpus alleges that he was denied due process because the Illinois trial court placed on him the burden of proving that he was unfit to stand trial. At the time of Bilyew’s conviction Illinois law provided that “[t]he burden of proving the defendant is not fit is on the defendant if he raises the question * * *.” Ill.Rev.Stat. ch. 38, § 1005-2-l(i) (1973) (repealed by P.A. 81-1217, § 3, effective Dec. 28, 1979). Bilyew’s lawyer raised the question of fitness and thereby assumed the burden of showing that Bilyew was unfit to be tried. The Illinois Supreme Court has since held that due process requires the State to bear the burden of proving fitness once the defendant has raised a bona fide doubt, People v. McCullum, 66 Ill.2d 306, 5 Ill.Dec. 836, 362 N.E.2d 307 (1977), and the Third Circuit also has held that “[e]vidence showing competency must be more persuasive than that showing incompetency.” *1239United States v. DiGilio, 538 F.2d 972, 988 (3d Cir. 1976), certiorari denied, 429 U.S. 1038, 97 S.Ct. 733, 50 L.Ed.2d 749; see also United States v. Makris, 535 F.2d 899, 906 (5th Cir. 1976), certiorari denied, 430 U.S. 954, 97 S.Ct. 1598, 51 L.Ed.2d 803. Nevertheless, the district court denied Bilyew’s petition, holding “that even if petitioner was denied due process for having been convicted pursuant to an unconstitutional statute such error must be deemed harmless beyond a reasonable doubt.” Order at 2. We reverse and remand to the district court for further proceedings.
I
On June 7, 1974, John William Bilyew was indicted for murder, rape, and concealing a homicidal death. Bilyew allegedly was sitting on the front steps of his apartment building in Marion, Illinois, when he saw Frances Buckner, a 13-year-old girl with whom he had spoken once before, walking down the street. Bilyew approached Frances and asked her to come over and sit on his porch. She began to comply, but as they reached his yard he asked her for “some sex.” According to the statement Bilyew later gave to the police, “She goes no, I have to get to the store. I [Bilyew] said why not. It won’t take very long. She goes no, I have to get home. I [Bilyew] put my hands around her neck and choked her. I thought she was dead.” After undressing the body, Bilyew wrapped it in a blanket that he got from his apartment and hid the body beneath a nearby house. Three days later, upon hearing a “voice” commanding him to turn himself in, Bilyew dragged the body into the street, removed the blanket, and threw a rock through a neighbor’s window to attract attention. According to his statement, “I [Bilyew] then went to bed but could not sleep.”
On the other hand, according to the experts who testified at his pretrial fitness hearing, Bilyew has never been fully “awake.” Following a request by Bilyew’s appointed counsel, the trial court appointed two psychiatrists, Drs. Norris and Horecker, and two psychologists, Drs. Rubin and Rad-er, to examine Bilyew, and then held a two-day hearing for the purpose of determining Bilyew’s fitness to stand trial. The experts agreed essentially that Bilyew was at least somewhat retarded and socially maladjusted, but split in their ultimate opinions of whether he was fit to be tried. Drs. Norris and Rubin concluded that Bilyew was not fit, while Drs. Rader and Horecker found that Bilyew was fit. We now review the evidence presented at the fitness hearing in detail to determine whether there is a reasonable possibility Bilyew would have been found unfit to stand trial had the trial judge placed the burden of proof on the State instead of Bilyew.
A
Bilyew’s first witness, Dr. Norris, prepared a written report for the trial judge but was unable to testify at the hearing. Pursuant to a stipulation by the State and Bilyew, the trial judge admitted Dr. Norris’ deposition as evidence at the hearing. In his report and deposition, Dr. Norris stated that he is Chairman of the Department of Psychiatry of the Southern Illinois University School of Medicine and that he examined Bilyew on June 11, 1974 for approximately one and one-half hours. On the •basis of this examination and test scores recorded on other occasions, Dr. Norris concluded that Bilyew was not fit to stand trial.
Specifically, Dr. Norris noted that Bilyew had attended special education classes throughout his school years. While being treated in 1968 at the Bowen Children’s Center in Harrisburg, Illinois, Bilyew was classified as mildly mentally retarded with an IQ of 63. Dr. Norris stated that such a score corresponded to a mental age of roughly nine and one-half to ten years. At this level of acumen, Dr. Norris stated that the typical subject probably would be able to read, would be able to recall past events, and should be geographically oriented. Dep. at 17-18. In 1973, Bilyew was tested at Anna [Illinois] State Hospital, and scored an IQ of 41, equivalent to a mental age of six years. With this lower level of intelligence, the subject would “have limited ca*1240pacity to benefit from education” and would “surely need supervision for the rest of his life unless there was some dramatic change, which is very rare.” Dep. at 17. Dr. Norris concluded that the second, more recent test more closely reflected Bilyew’s 1974 capacity; “at best his intelligence quotient is that of someone who has a mental age of 6 years.” Report at C-105.
In addition to reviewing Bilyew’s IQ scores, Dr. Norris performed a “Mental Status Examination” which consisted of asking Bilyew a sequence of questions and noting his responses, and having Bilyew perform various simple tasks. Dr. Norris first found that Bilyew was unable to read “in any practical sense.” Report at C-103. He showed Bilyew the four words “John,” “dog,” “man,” and “run,” but Bilyew recognized only “John,” his name. Asked to copy the words onto another sheet of paper, Bilyew copied only “John” and “man,” although Dr. Norris believed he might have gotten Bilyew to copy the other words also had he been persistent. Dep. at 29. “In this area, he [was] even more deficient than [Dr. Norris] would expect from a mental age of 6.” Report at C-103. Similarly, given the choices of summer, winter, or spring, Bilyew was nearly correct to say that the season was summer (it was still spring), but answered incorrectly when asked what year it was. Id. Bilyew was not aware of geography. He recognized the name “California” (where he was born and lived for many years) but did not know where it was; indeed, “he had no idea where he was in the United States.” Dep. at 31. Bilyew also had very little mathematical aptitude. He could subtract three from ten by counting backward on his fingers, but was unable to do so if Dr. Norris held his hands. Bilyew could repeat only four digits forward and two digits backward. Dr. Norris doubted that Bilyew could tell time. Dep. at 66.
More importantly, Dr. Norris found that Bilyew generally “cannot pick up stimuli from the environment and use them.” Dep. at 30. For example, Bilyew could not rer member his lawyer’s name, and although told Dr. Norris’ name several times during the examination was unable to recall it either. Bilyew was capable of remembering only “certain gross kinds of experiences, for example, big things that had happened to him in his lifetime [such as] very emotion[-]laden experiences.” Dep. at 31. But Bilyew did not remember killing Frances Buckner. Dep. at 36, 37.
On the other hand, Bilyew had acquired the ability to speak in a manner that “suggested] someone of higher intelligence.” Report at C-103. Dr. Norris surmised that Bilyew’s intensive special education had allowed him to acquire a greater vocabulary and fluency. Dep. at 24-27. Nevertheless, Dr. Norris concluded that Bilyew’s vocal ability would not facilitate any meaningful communication with defense counsel. “[Bilyew] can assist in this defense in only the most primitive manner. His intellectual capacity is so limited that he would respond only to the simplest of questions. His emotional ability and his history of instability would suggest his inability to respond appropriately to the situation. He has a long history of repetitive social behavior which consistently fails, consistently gets him into difficulty, and yet which he consistently repeats.” Report at C-104.
Dr. Norris found that Bilyew knew what murder and rape are, and noted that he was able to describe that rape involves force, although the word “force” probably was not in his vocabulary. Dep. at 62. Bilyew also knew that murder and rape are wrong, but according to Dr. Norris he was incapable of understanding the seriousness of such crimes. Report at C-104. Although Dr. Norris did not believe Bilyew to be “coldhearted,” Dep. at 37, Bilyew displayed “an extremely shallow feeling and at no time [did Dr. Norris] have any sense that he is displaying a reaction appropriate to the death of the girl.” Report at C-104.
Finally, Dr. Norris found that Bilyew did not understand the consequences of his being found guilty of the alleged crimes. Dr. Norris stated that Bilyew would answer affirmatively to a leading question such as “Do you know you may go to the chair or prison for life or twenty years?” and then afterward ask the doctor, “Yeah, but am I *1241getting out of jail today?” Dep. at 64. Bilyew “would have no idea what twenty years [in prison] meant.” Dep. at 63. “[H]e has no true sense of [imprisonment] as being a possible future event. While the patient is able to recall fairly accurately events in the recent and remote past, he has an extremely limited capacity to make a judgment regarding the significance of these events and certainly he has a most limited capacity to reason and discuss with his counsel the consequences of his behavior.” Report at C-105.
Dr. Norris also noted the Bilyew was not being deceptive in the interview and that he would have detected any attempt by Bilyew to deceive him. Dr. Norris believed Bilyew’s problem to be primarily “organic” rather than “cultural,” Dep. at 44-47, and recommended that Bilyew undergo an electroencephalogram in order to detect a possible temporal lobe epilepsy that may have caused Bilyew to hear the “voice.” Dep. at 38-39; Report at C-105. Although Bilyew had heard “voices” on at least two other occasions, Dr. Norris believed Bilyew’s disability was “not like schizophrenia where * * * there is a major psychosis where it’s present all the time.” Dep. at 39.
B
Dr. Rubin, another defense witness, was the first of the experts to testify at the fitness hearing. Dr. Rubin is a psychologist specializing in sexual maladjustment, an associate professor at the Southern Illinois University medical school, and a former consultant to the Illinois Department of Corrections at Menard and Vienna for the purpose of establishing treatment programs for sexual offenders. Dr. Rubin interviewed Bilyew for some ninety minutes and, like Dr. Norris, concluded that Bilyew was unfit to stand trial.
In his report to the trial judge and testimony, Dr. Rubin found Bilyew to be “functionally illiterate,” “exceedingly socially immature,” and especially frustrated and angry over being unable appropriately to find a sexual partner. Report at C-99, C-102; Record at 106. Since Dr. Rubin’s expertise primarily concerns sexual deviancy, his findings concentrated on this aspect of Bilyew’s character. He found that Bilyew had a history of sexual maladjustment. “Mr. Bilyew readily admits approaching girls in California and asking them to have sex with him. He apparently did so in a very inappropriate manner in that each incident resulted in a complaint being lodged against him.” Report at C-100.
Dr. Rubin’s most relevant finding was that Bilyew would be unable to assist counsel in preparing a defense because “in some areas, especially sociosexual, * * * Mr. Bilyew [does not] honestly know[ ] when he is telling the truth.” Report at C-102. Dr. Rubin found that Bilyew had little understanding of truth by leading him with questions into making obviously inconsistent statements. “[H]e personally felt he was telling the truth on both occasions and both statements even though they followed each other very closely in time and [even though] they were conflicting.” Record at 112. Dr. Rubin found that “Mr. Bilyew will respond in any way that he thinks his questioner would consider appropriate * * *. In certain areas of his behavior (especially social interactions) [Dr. Rubin] believe[d] that Mr. Bilyew has difficulty in distinguishing fact from fantasy.” Id. “[F]antasy and actual facts become so confused that he does not know precisely what occurred and does not know when he’s telling the truth and when he’s not telling the truth.” Record at 112-113. Therefore Dr. Rubin concluded that Bilyew would be unable truthfully to tell his lawyer what happened the day he allegedly killed Frances Buckner. “[H]e will be unable to give accurate information to his attorney or, at the very least, his attorney will not be able to determine what information is accurate.” Id.; see also Record at 109,159. These findings may help to explain an important inconsistency in the record of the case. The statement Bilyew gave to the police, in which Bilyew allegedly said, “I put my hands around her neck and choked her,” contradicts Bilyew’s later statements that he had no memory of the event. If Dr. Rubin’s diagnosis is correct, however, Bilyew’s susceptibility to being led could account for the statement.
*1242Dr. Rubin also found that Bilyew “is naive or unrealistic about a great many aspects of his condition. He does not want to go to prison, but if he does he expects a sentence of 1-14 years. If he is sent to a hospital, he thinks he will be returned to Anna [a minimum security state hospital]. He does not want to have a jury trial because he is afraid to talk about ‘those terrible things’ in front of a group of people. He knows he has been charged with a serious offense, but he wants to be out of jail for his birthday some ten days hence.” Report at C-102.
C
Dr. Rader, the final defense witness, testified second at the hearing. Dr. Rader is a clinical psychologist and professor at Southern Illinois University. During two meetings with Bilyew that together lasted some four and one-half hours, Dr. Rader administered a series of standardized intelligence and personality tests to Bilyew. Dr. Rader found that Bilyew’s overall IQ score was 66, that Bilyew suffered from a memory impairment but was not psychotic, and that Bilyew was competent to be tried notwithstanding “difficulties for a defense lawyer” due to Bilyew’s impairment. Report at C-107 to C — 108; Record at 205.
Dr. Rader’s report to the trial judge explains his competency finding more clearly:
Concerning his competency to stand trial, Mr. Bilyew will obviously be at some disadvantage. He is capable of understanding the basic purposes of the legal proceedings and the potential consequences of a conviction. Whether or not he is truthful in disclaiming any memory for the actual event, his ability to perceive events in general and to recall details of what happened is limited by his intellectual deficit. Events — any events — are likely to be perceived and remembered by him in a rather gross, impressionistic manner. He lacks the introspective capacity and sophistication to be able to describe his own inner states— his feelings, his motives, his train of thought — beyond the most superficial and global characterizations. His understanding is limited by both limited reception of the data of experience and by poor memory for what was originally within the boundaries of his awareness. This is not to say that he is incompetent to assist in his own defense. Rather, I wish to point out that relative to a person of normal intelligence he will be able to offer his lawyers less information upon which to base their case.
Report at C-108.
Dr. Rader believed that Bilyew’s diminished abilities were caused by a brain dysfunction and that Bilyew was already performing at the best of his abilities. Record at 198, 199-200. With an IQ score of 66, Dr. Rader stated that Bilyew’s optimal mental functioning would be equivalent to that of a nine- or ten-year-old. Record at 194.
D
The State then introduced the testimony of one expert and three lay witnesses. The State’s expert was Dr. Horecker, a psychiatrist whose expertise comes from having performed between two and three thousand fitness examinations. His examination of Bilyew lasted some two and one-half hours, after which Dr. Horecker found Bilyew fit to stand trial.
In particular, Dr. Horecker found that Bilyew was “oriented as to time and place and person,” and generally able to recollect and convey his memories to others. Record at 216, 219. Any defect of memory, according to Dr. Horecker, was due to Bilyew’s conscious or unconscious. “denial” of the memory rather than any mental disease or defect. Record at 221, 238. Dr. Horecker found no evidence of brain damage and nothing wrong with Bilyew other than perhaps mild retardation. Record at 242, 248. Although Dr. Horecker performed no tests to measure Bilyew’s IQ, he estimated it to be “at least in the 70’s or perhaps even in the low 80’s.” Record at 239. Dr. Horecker believed that Bilyew might be able to read and found that he was not easily led by questioning. Record at 235, 245. Therefore Dr. Horecker concluded:
*1243CONCLUSSIONS [sic] AND RECOMMENDATIONS: A longitudinal study of the material obtained from Mr. Bilyew indicates that within his personality there are moderately strong passive-aggressive factors. These factors on the one hand cause him to try to get people generally and women particularly to accept him and show him that they approve of him; on the other hand cause him to feel angry and belligerent because unconsiously [sic] the feeling keeps telling him over and over again that nobody cares about him or wants him or approves of him. There is no evidence of a clinical neurosis nor of a psychosis. Mr. Bilyew, in spite of his limited education, is able to handle the english language fairly well, although there are certain words that he does not comprehend. When he is instructed to say that he doesn’t know the meaning of a word, I have found that substituting other words that mean the same thing makes it possible for him to deal with the concepts intellegently [sic] and correctly. The fact that he has an I.Q. below the so called “normal” range, in no way interferes with his being able to appreciate or deal with the factors pertinent to the current situation in which he is involved.
Mr. Bilyew understands the nature of the charges and considers them to be serious charges. He can cooperate with counsel in his own defense. It is my professional judgment that he is fit to stand trial.
Report at C-112 to C-113.
The State also presented three lay witnesses in order to corroborate its contention that Bilyew was fit to be tried. The first witness was the proprietor of a television repair shop, from which Bilyew purchased a length of wire. This witness noticed nothing unusual about Bilyew’s ability to purchase wire, except that he returned it the next day because he already had all he needed. The other two witnesses were jailers at the Williamson County jail who testified that Bilyew performed his jail duties satisfactorily, that he dressed himself, shaved, kept clean, walked without apparent handicap, conversed, and purchased specific brands of cigarettes at the jail commissary.
The record also contains records maintained by Several institutions Bilyew had attended, including a “psychological report” performed in 1968 to determine Bilyew’s eligibility to enroll in special education classes at Marion Junior High School. That report shows that on December 13, 1967, Bilyew scored a “full scale” IQ of 51 on the Weehsler Intelligence Scale for Children. The report indicates that the same examination had been given to Bilyew on May 17, 1965, and at that time he scored a “full scale” IQ of 54.
II
The trial judge determined that Bilyew was fit to stand trial. Unfortunately, it is impossible to know from the wording of her oral opinion whether she would have found Bilyew unfit had the burden of proof been placed on the State instead of Bilyew. The opinion does not particularly mention the burden of proof and does not state whether this was such a close question that placement of the burden of proof might have made a difference. Nevertheless, we do not doubt that the trial judge applied the Illinois statute then in force, Ill.Rev.Stat. ch. 38, § 1005-2-l(i) (1973), which placed upon Bilyew the burden of proving that he was unfit. “In the absence of any indication to the contrary, we must assume that the trial judge, in allocating the burden of proof on fitness, followed the clear direction of section 5-2-l(i).” People v. Hubert, 51 Ill.App.3d 394, 398, 9 Ill.Dec. 398, 366 N.E.2d 909 (1st Dist. 1977) (appeal to the Illinois Supreme Court denied). That the trial judge indeed meant to put the burden on Bilyew is supported by her having Bilyew put on his evidence prior to hearing the State’s evidence and giving Bilyew the first opportunity to make closing arguments. See People v. Hubert, supra, 51 Ill.App.3d at 398-399, 9 Ill.Dec. 398, 366 N.E.2d 909. Moreover, during the initial hearing on Bilyew’s motion to determine fitness, the State’s lawyer told the trial judge that “Chapter 38, Section 1005-2-1” is the only relevant statute (Record at 77), and in final *1244arguments stated, “the burden of proof in this case is upon the defendant.” Record at 293. As already noted the trial judge’s opinion does not reveal whether she would have reached the same or a different result had she put the burden of proof on the State.
Having found Bilyew fit to stand trial, the trial judge held a hearing on defendant’s motion to suppress the statement given to police. The trial judge rejected Bilyew’s contention that he was incapable of understanding the Miranda warnings and thus unable to waive any of his rights. Bilyew then pled guilty to murder and concealing a homicidal death. He was sentenced to consecutive terms of fifty to one hundred years for murder and one to three years for concealing a homicidal death.
Bilyew appealed, and the three-judge state appellate panel reversed his convictions based on the unconstitutionality of the burden of proof statute. People v. Bilyew, 55 Ill.App.3d 69, 12 Ill.Dec. 781, 370 N.E.2d 585 (5th Dist. 1977). The appeals court relied on People v. McCullum, 66 Ill.2d 306, 5 Ill.Dec. 836, 362 N.E.2d 307 (1977), which had held that the burden of proof statute violated due process. The appeals court noted, “this case is an ideal example of a situation where the allocation of the burden of proof would be most likely to influence the court’s decision for the experts were equally divided on the ultimate issue.” 55 Ill.App.3d at 73, 12 Ill.Dec. 781, 370 N.E.2d 585.
The Illinois Supreme- Court reversed the appeals court and reinstated Bilyew’s convictions. The Supreme Court held that although the burden of proof statute is unconstitutional under People v. McCullum, it is unnecessary to reverse a conviction unless the record contains a “specific indication” “that the unconstitutional burden of proof provision entered into the trial court’s determination.” People v. Bilyew, 73 Ill.2d 294, 300, 304, 22 Ill.Dec. 736, 383 N.E.2d 212 (1978). The Court reasoned that “[t]he statutory allocation of the burden of proof becomes significant in a given case only when the evidence is so nicely balanced that neither fitness nor unfitness is established by a preponderance of the evidence. In such a case, the determination must be adverse to the party bearing the burden. * * In [Bilyew’s] case the trial judge’s lengthy and careful statement of her reasons for finding defendant fit, fairly read, gives no indication that she was concerned about the burden of proof or believed that defendant had not met anything required of him.” 73 Ill.2d at 302, 22 Ill.Dec. 736, 383 N.E.2d 212. Since the trial judge had given credence to some testimony and discredited other testimony, the equipoise of the experts’ ultimate conclusions was deemed superficial. The Supreme Court distinguished McCullum because “[t]here a jury was the fact finder at the fitness hearing * * * [and] a court of review cannot know what entered into the jury’s determination.” 73 Ill.2d at 303, 22 Ill.Dec. 736, 383 N.E.2d 212. Presumably, unlike the trial court here, a jury is incapable of giving “specific indication that the burden of proof was a factor in [its] conclusion.” 73 Ill.2d at 304, 22 Ill.Dec. 736, 383 N.E.2d 212.
Bilyew then filed the underlying petition for a writ of habeas corpus in the district court. The district court did not decide whether the burden of proof statute was unconstitutional, but granted summary judgment in favor of the State because “even if the petitioner was denied due process for having been convicted pursuant to an unconstitutional statute such error must be deemed harmless beyond a reasonable doubt. * * * [A] review of [the] evidence causes this Court to conclude that the circuit judge’s finding of fitness is supported by the record beyond a reasonable doubt.” Order at 2. Bilyew now appeals from that order.
Ill
There is little question that the Fourteenth Amendment requires the State or federal prosecution to shoulder the burden of proving that the defendant is fit to stand trial once the issue of unfitness has been properly raised.1 Every court that has *1245considered the issue has so held. See, e.g., United States v. DiGilio, 538 F.2d 972, 988 (3d Cir. 1976), certiorari denied, 429 U.S. 1038, 97 S.Ct. 733, 50 L.Ed.2d 749; United States v. Makris, 535 F.2d 899, 906 (5th Cir. 1976), certiorari denied, 430 U.S. 954, 97 S.Ct. 1598, 51 L.Ed.2d 803; People v. McCullum, supra; Dixon v. Scott, No. 80 C 2962 (N.D.Ill.1980). The Illinois Supreme Court has put the reason for giving the State the burden succinctly: “Let us assume that defendant is in fact unable to co-operate with counsel and present his case in a rational manner. It would be a strange rule, indeed, to impose upon him the burden of proving his own incompetence, for the very disability which he would be seeking to prove renders him incapable, either logically or legally, of sustaining the burden of proof.” People v. Bender, 20 Ill.2d 45, 53-54, 169 N.E.2d 328 (1960).
The real question in this case, then, is whether misallocation of the burden of proof can be overlooked as “harmless error.” In order to determine if placing the burden of proof on Bilyew was harmless, we first must specify exactly the “error” that is to be avoided. Ordinarily, the error question is whether there is a reasonable possibility that the error affected the determination of guilt, i.e., might the defendant have avoided being found guilty absent the constitutional blunder? Chapman v. California, 386 U.S. 18, 87 S.Ct. 824, 17 L.Ed.2d 705; United States ex rel. Gorham v. Franzen, 675 F.2d 932, 938 (7th Cir. 1982). In this ease, however, the error would be the trial of a defendant who is unfit to be tried. The two errors are certainly connected; if the State tries an unfit defendant, we can be much less sure that the defendant in fact committed the crime because by definition the defendant will have been unable to assist counsel in producing a defense. But the trial of an unfit defendant cannot be upheld simply because we feel sure that he in fact is the culprit. As Chief Justice Burger wrote for a unanimous Supreme Court, “the prohibition [against trials of incompetents] is fundamental to an adversarial system of justice.” Drope v. Missouri, 420 U.S. 162, 172, 95 S.Ct. 896, 904, 43 L.Ed.2d 103. The adversarial system of justice cannot be sidetracked in order to convict unfit, albeit probably guilty, defendants.
Thus the harmless error question is whether there is a reasonable possibility that by using a constitutional burden of proof the trial judge would have found Bilyew unfit. The Appellate Court of Illinois so held (55 Ill.App.3d at 73, 12 Ill.Dec. 781, 370 N.E.2d 585) and we agree. The Illinois Supreme Court, to the contrary, found that the burden of proof error was harmless because “the record contains no specific indication that the burden of proof was a factor in the trial judge’s conclusion.” 73 Ill.2d at 304, 22 Ill.Dec. 736, 383 N.E.2d 212.
The Illinois Supreme Court’s presumption that placement of the burden of proof is harmless error stems from its conception of the importance of the burden. In the Court’s view, quoting in part from DiGilio, supra, 538 F.2d at 988, “the burden of proof becomes significant in a given case only when the evidence is so nicely balanced that neither fitness nor unfitness is established by a preponderance of the evidence.” 73 Ill.2d at 300, 22 Ill.Dec. 736, 383 N.E.2d 212. Still quoting from the Third Circuit’s opinion in DiGilio, the Illinois Supreme Court stated such cases would be “rare.” 73 Ill.2d at 301, 22 Ill.Dec. 736, 383 N.E.2d 212 (again quoting 538 F.2d at 988). The Illinois Supreme Court offered no further reasoning or factual basis for its conclusion that close cases where the burden of proof would matter are infrequent. DiGilio qualifies its “rare” conclusion as “in theory” (538 F.2d at 988) and vacated DiGilio’s judgment of sentence for a new competency hearing unless the trial judge could retroactively determine DiGilio’s competency. 538 F.2d at 989.
*1246We doubt the soundness of the “theory” that concludes that an unconstitutional burden of proof can be disregarded unless the trial judge has made a point of stating that he or she has relied upon it.2 Even if cases where the burden of proof makes a difference are in fact rare, there is no need to preclude a defendant from pointing to facts in the hearing record to support his argument that the burden of proof did make a difference. Surely the hearing record, in conjunction with the trial judge’s opinion, is a more reliable indicator of the closeness of the question than merely this trial judge’s listing or failure to list which procedural statutes she relied on in reaching her decision. When the trial judge is informed by the prosecution that the relevant statute places the burden of proof on the defendant and she does not say in her opinion whether she relied on the burden of proof, then we certainly cannot presume, as did the Illinois Supreme Court, that she did not rely on it. The situation is analogous to where a jury has been instructed to use the same statute and decides the issue. In either case, we do not know from the judicial or jury decision alone whether the burden of proof was decisive. Yet as noted above, had the fitness hearing been held before a jury, the Illinois Supreme Court would have reversed Bilyew’s convictions and remanded for a new hearing to determine fitness. 73 Ill.2d at 303, 22 Ill.Dec. 736, 383 N.E.2d 212 (distinguishing McCullum, supra). See also People v. Tamayo, 56 Ill.App.3d 800, 804, 14 Ill.Dee. 423, 372 N.E.2d 434 (2d Dist. 1978) (“this court is compelled to reverse and remand this cause for a new fitness hearing since it is not apparent on this record under which standard of proof the defendant’s fitness to stand trial was determined by the trial court”); People v. Hubert, 51 111. App.3d 394, 399, 9 Ill.Dee. 398, 366 N.E.2d 909 (1st Dist. 1977) (per then Justice Bua) (“A court of review may not properly speculate as to what result the trier of fact, whether a judge or jury, would have reached had the burden of proof been allocated differently.”).
IV
Therefore, the task before the district court was to examine the hearing record to determine whether there is a reasonable possibility that Bilyew would have been found unfit had the State been given the burden of proving him fit. Because the evidence was very close, there clearly is such a reasonable possibility. Four experts examined Bilyew. Two found that Bilyew was not fit to stand trial, one found that he was fit but would present “difficulties for a defense lawyer,” and one simply found that he was fit. Even under a microscope, the evidence appears to have been closely balanced. For example, Dr. Norris, who found Bilyew to be unfit, had the best qualifications among the four experts since he was Chairman of the Psychiatry Department of the Southern Illinois University medical school. The 1968 and 1973 IQ scores of 63 and 41, respectively, that Dr. Norris relied on in part were hardly less probative than the 1974 IQ score of 66 relied on by Dr. Rader. Similarly, Dr. Horecker’s unqualified testimony that Bilyew was fit to be tried might be given greater or less weight depending on the value assigned to his experience of performing literally thousands of fitness examinations. He did not conduct an IQ test on Bilyew.
The trial judge’s conclusion was not necessarily in error since, as in most close decisions, there was substantial evidence to support either conclusion. But the conclusion might well have been different had the burden of proof been placed as required by due process. Thus the trial 'court’s use of an unconstitutional burden of proof cannot be deemed harmless error in this case. And the district court’s finding that the trial judge’s fitness determination was “supported by the record” is not to the point.
Consistently with the Third Circuit’s action in DiGilio, supra, 538 F.2d at 989, we therefore remand to the district court to determine whether it is still possible to hold a meaningful retrospective hearing to find whether Bilyew was fit to stand trial at the time of the state proceedings in 1974. Al*1247though there are obvious hazards attendant with retrospective competency hearings, see Drope v. Missouri, 420 U.S. 162, 182, 95 S.Ct. 896, 909, 43 L.Ed.2d 103 and Pate v. Robinson, 383 U.S. 375, 387, 86 S.Ct. 836, 843, 15 L.Ed.2d 815, the mere passage of time may not make the effort meaningless. “The passage of even a considerable amount of time may not be an insurmountable obstacle if there is sufficient evidence in the record derived from knowledge contemporaneous to trial.” United States v. Makris, 535 F.2d 899, 904 (5th Cir. 1976), certiorari denied, 430 U.S. 954, 97 S.Ct. 1598, 51 L.Ed.2d 803 (two and one-half year delay from trial to retrospective competency hearing), citing Conner v. Wingo, 429 F.2d 630 (6th Cir. 1970), certiorari denied, 406 U.S. 921, 92 S.Ct. 1779, 32 L.Ed.2d 121 (seven-year delay from trial to competency hearing); see also Bruce v. Estelle, 536 F.2d 1051 (5th Cir. 1976), certiorari denied, 429 U.S. 1053, 97 S.Ct. 767, 50 L.Ed.2d 770 (nine-year delay from trial to competency hearing). “A reliable reconstruction of petitioner’s mental status [as of the time of trial] depends less on time than on the state of the record. Especially where medical information substantially contemporaneous to trial is available, the chances for an accurate assessment increase.” 536 F.2d at 1057, citing Holloway v. United States, 343 F.2d 265 (D.C.Cir.1964). Since Bilyew was examined by four doctors prior to trial and some of Bilyew’s institutional records are still available, the district judge may be able to decide on the present record Bilyew’s fitness at the time of the state proceedings. In that case he should hold a hearing to determine whether Bilyew was fit to be tried as of 1974. Bilyew’s possible lack of fitness appears to be of a permanent rather than transitory nature; therefore further medical or psychiatric testing may be relevant and advisable to ascertain his condition at the time of the original state court proceedings.
If the district court finds that such a hearing would still be meaningful, it should conduct the hearing, placing the burden of proof upon the State to show by a preponderance of the evidence that in 1974 Bilyew was fit to be tried.3 If the district court finds that it is too late to hold a retrospective fitness hearing on this record and possibly additional evidence, or holds the fitness hearing and finds that Bilyew was unfit in 1974 to stand trial, then the district court must grant Bilyew’s request for a writ of habeas corpus. In that event, it should give the State an opportunity to try to prove Bilyew’s present fitness in the Circuit Court of Williamson County, resulting in a new trial if it carries that burden.
We reverse and remand for further proceedings consistent with this opinion.
APPENDIX
The “theoretical” explanation for why who has the burden of proof rarely makes any difference may be explained by examining the simple mathematical model to which the “theoretical” explanation is parallel. See, e.g, Kaye, The Limits of the
*1248Preponderance of the Evidence Standard: Justifiably Naked Statistical Evidence and Multiple Causation, 1982 A.B.F. Res.J. 487; Kornstein, A Bayesian Model of Harmless Error, 5 J.Legal Stud. 121 (1976). The model supposes that the evidence at a fitness hearing can be toted on a scale of probabilities from zero to one, with zero representing certain unfitness and one representing certain fitness. Expressing the Illinois Supreme Court’s view within the terms of the model, the burden of proof makes a difference only when the probabilities of fitness and unfitness are in equipoise, that is, when the evidence suggests a one-half probability that the defendant is fit. Since the number one half is infinitely rare among the numbers between zero and one, the occasions when the burden of proof makes a difference must be rare also.
There are at least three objections to such a theory of the burden of proof. First, factfinders do not discern a continuous spectrum of evidentiary weights. While there is an unaccountably infinite supply of numbers between zero and one, a judge or a jury can experience only a small, finite number of degrees of certainty that a defendant is fit to stand trial. A defendant may be absolutely fit, most likely fit, hard-to-tell-but-probably fit, etc., but before long any additional gradations of fitness are imperceptible. Thus cases when the evidence of fitness and unfitness seem in balance are not unique among some infinite variety of evidentiary balances, but instead are among a much smaller number of possibilities that may be perceived by the factfinder.
Second, the statement that cases where the evidence is equipoised are “rare” assumes without justification that the weights of evidence are distributed in a fashion that makes equipoise rare. The contrary assumption, that there are relatively many close cases, similarly lacks justification. Perhaps the Illinois Supreme Court intended to state an empirical judgment about the relative frequency of hard cases in which the burden of proof is determinative, but instead the Court appears simply to have swallowed the unjustified logic that when probabilities are evenly distributed equipoise occurs infrequently.
Finally, the allocation of the burden of proof probably is important throughout the hearing, not just at the end when all the evidence has been received and is being weighed by the factfinder. The party with the burden of proof must put on its evidence first, and goes first with closing argument. While these procedural differences do not by themselves violate due process, they suggest a weakness in a model of burdens of proof that requires the evidence be evaluated all at once at the close of the proceedings. More likely is that the fact-finder (to the extent it is familiar with and knows who has the burden of proof) makes use of the burden throughout its reception of the evidence, viewing the burdened party’s evidence more skeptically as it comes in. Of course, the burden of proof should be applied only at the close of the hearing, see, e.g., McCormick on Evidence at 784, but in determining generally whether burdens of proof do make a difference we must consider reality.
We conclude that models such as the above are not much help in deciding whether burdens of proof often make a difference. See also L. J. Cohen, The Probable and the Provable 49-120 (1977) (catalog of paradoxes inherent in probabilistic analysis of legal proof); G. Shafer, A Mathematical Theory of Evidence (1976) (criticising Bayesian probability theories of evidence). If the evidence is closely balanced, then common sense indicates there is a reasonable possibility that who bears the burden of proof will determine the outcome.
. The test for determining whether a defendant is fit “must be whether he has sufficient present ability to consult with his lawyer with *1245a reasonable degree of rational understanding — and whether he has a rational as well as factual understanding of the proceedings against him.” Dusky v. United States, 362 U.S. 402, 402, 80 S.Ct. 788, 789, 4 L.Ed.2d 824 (per curiam). See also Ill.Rev.Stat. ch. 38, § 1005-2-l(a) (1973) (repealed by P.A. 81-1217, § 3, effective Dec. 28, 1979).
. See Appendix hereto.
. Ordinarily in a habeas corpus proceeding, the burden of proof is on the petitioner to show that his conviction is illegal. E.g., United States v. Hollis, 569 F.2d 199, 205 (3d Cir. 1977); Spanbauer v. Burke, 374 F.2d 67, 74 (7th Cir. 1966), certiorari denied, 389 U.S. 861, 88 S.Ct. Ill, 19 L.Ed.2d 127; Piner v. United States, 222 F.2d 199, 203 (7th Cir. 1955). Bilyew has borne that burden by proving that the burden of proof statute used to determine his fitness was unconstitutional and that there is a reasonable possibility that placement of the fitness burden made a difference. If we were also to place upon Bilyew the burden of showing his unfitness in a new hearing before the district court, then the error complained of would become practically unreviewable; the district court would only be repeating the mistake of the state court and doubtless would reach the same result. “The rationale for [placing the burden of proof on a habeas petitioner] is that criminal proceedings usually may be presumed to have been proper and legal and, consequently, a habeas petitioner must show that such a presumption is incorrect in his own case.” 569 F.2d at 205-206. Bilyew has demonstrated the irregularity in the state court proceedings, however, and the presumption of correctness consequently is inapplicable. The only purpose for holding another hearing if the district judge cannot decide fitness on the present record is because Bilyew has shown a likelihood that with a constitutional burden of proof he would be found unfit. Any new hearing should give Bilyew the opportunity to test that proposition.