Defendant-petitioner William Duane El-ledge appeals a district court order denying him federal habeas corpus relief; he raises six bases for relief. Because we find that Elledge has shown that his constitutional rights were violated in one respect, we vacate the district court’s judgment and remand with instructions.
Elledge was involved in three killings that occurred in a 36 hour period in August, 1974,1 although only the first murder — of Margaret Anne Strack — is at issue on appeal. Strack was raped and killed in Hollywood, Florida, on August 24, 1974; the two subsequent murders occurred in Jacksonville, Florida. Early in the morning of August 26, 1974, Jacksonville police arrested Elledge for the third homicide. He then was interrogated four times between 4:30 a.m. and 10:30 a.m. that morning; the police properly informed Elledge of his rights- on each occasion. See generally Miranda v. Arizona, 384 U.S. 436, 86 S.Ct. 1602, 16 L.Ed.2d 694 (1966). During the course of their questioning and investigation, the Jacksonville police determined that Elledge was a suspect in the Strack homicide and questioned him about it as well. At the last interrogation, Elledge orally confessed to all three murders. The next day he again confessed to the Strack murder after follow-up interrogation by Hollywood police investigators; this confession was tape recorded.
Elledge’s public defender adopted the following strategy: (1) plead insanity if possible; (2) if that defense was not available, suppress the confession and plead not guilty; and (3) if the confession could not be suppressed, plead guilty and seek mercy. The insanity defense proved fruitless when psychiatrists found Elledge to be sane; efforts to suppress the confessions on the grounds that Elledge was physically coerced because he confessed without sleep or food and was still under the influence of drugs and alcohol also failed. Under advice of counsel, Elledge then entered a plea of guilty, leaving sentencing as the sole issue to be determined.
Although a sentence of death was entered after the first sentencing hearing (in 1975), the Florida Supreme Court overturned that sentence and remanded the case for a new sentencing hearing. Elledge v. State, 346 So.2d 998 (Fla.1977). The same state trial judge presided over the second sentencing hearing; he appointed the same attorney, who by this time was in private practice, to represent Elledge. This second sentencing hearing, held in *14421977, is the sentencing hearing referred to in the balance of this opinion.
Elledge’s taped confession was played at the sentencing hearing; additionally, a variety of witnesses testified. Defendant took the stand in his own behalf and detailed his harsh childhood and early addiction to and abuse of alcohol and drugs. A jury considered the evidence and recommended death; the judge agreed and, thus, entered a sentence of death. A series of unsuccessful state appeals followed.2 After exhausting all state remedies, Elledge filed a petition for a writ of habeas corpus with the United States District Court for the Southern District of Florida. That court held an evidentiary hearing and concluded that, while counsel’s performance at the sentencing hearing was inadequate, no prejudice resulted from counsel’s unreasonable performance. The district court then denied the petition, and this appeal followed.
Elledge makes these contentions on appeal: (a) counsel’s performance in challenging the confession was inadequate and prejudiced Elledge; (b) the district court’s conclusion that no prejudice arose from counsel’s inadequate representation at the sentencing phase was erroneous because the court failed to consider that the evidence not adduced would have altered the entire evidentiary picture; (c) Elledge did not receive an individualized capital sentencing determination because the trial judge refused to consider any nonstatutory factors that mitigated against imposing the death penalty; (d) the death penalty was applied mechanistically under Florida’s felony murder rule without regard for whether El-ledge intended that a life would be taken; (e) the death penalty is applied in an arbitrary and discriminatory manner in Florida as evidenced by empirical studies that indicate a disproportionality in death sentences based on the race and/or sex of the victim; and (f) the trial court’s decision to shackle Elledge at the sentencing hearing was inherently prejudicial.
I. Effectiveness of Representation in Seeking to Suppress the Confessions.
As a threshold matter, effective assistance of counsel is a two-prong issue. The petitioner must establish both that counsel’s performance was not reasonably adequate and that petitioner was prejudiced by that unreasonable performance to the point that he did not receive a fair trial. Strickland v. Washington, 466 U.S. 668, 687, 104 S.Ct. 2062, 2064, 80 L.Ed.2d 674, 693 (1984). Elledge’s basic claim is that his counsel was ineffective because he failed to use the proper theory in challenging the confessions. The two confessions require different analytic frameworks; therefore, we will discuss each separately.
A. The first, untaped confession
Elledge’s counsel sought to suppress the first confession, arguing that it was involuntary because it was physically coerced. Counsel maintained that Elledge had no sleep the night of his arrest and interrogation, had no food and drink during his interrogation, confessed while hung over and under the residual impact of drugs and alcohol, and was in a general “daze” at the time of his confession.3 Counsel did not argue, however, that Elledge’s fifth amendment rights were violated when the police repeatedly reinterrogated and rewarned him of his Miranda rights despite El-ledge’s alleged invocation of his right to silence. This omission was unreasonable representation according to Elledge. We disagree.
The test for the performance prong of Strickland is objective “reasonableness under prevailing professional norms.” Id. at 688, 104 S.Ct. at 2065, 80 L.Ed.2d at 694. A reviewing court conducting such an ex*1443amination must view the performance at the time it occurred, avoid the “distorting effects of hindsight,” id., and “indulge a strong presumption that counsel’s conduct falls within the wide range of reasonable professional assistance....” Id. at 689, 104 S.Ct. at 2066, 80 L.Ed.2d at 694.
Until Michigan v. Mosley, 423 U.S. 96, 102-04, 96 S.Ct. 321, 326, 46 L.Ed.2d 313, 320-22 (Dec. 9, 1975), repeated reinterrogation in conjunction with repeated Miranda warnings was not recognized as a potentially coercive technique. Elledge’s counsel sought to suppress the first confession in March, 1975; obviously, he did not have the benefit of Mosley at that time. Furthermore, as of March, 1975, no Florida courts had held that such procedures were coercive. Reasonably effective representation cannot and does not include a requirement to make arguments based on predictions of how the law may develop. See Sullivan v. Wainwright, 695 F.2d 1306, 1309 (11th Cir.), aff'd, 464 U.S. 109, 104 S.Ct. 450, 78 L.Ed.2d 266 (1983) (per curiam). Thus, Elledge’s claim fails on the first, performance prong of Strickland.
B. The second, taped confession
Even if we assume, arguendo, that counsel’s performance was unreasonable because he did not attack the police’s alleged failure to honor Elledge’s invocation of his right to remain silent,4 we find that no prejudice attached as a result, because' the second confession was admissible — under either of two alternative grounds — even if the first confession was not.
A confession that follows on the heels of an involuntary confession may be tainted thereby and thus inadmissible. See Martin v. Wainwright, 770 F.2d 918, 928 (11th Cir.1985), modified on other grounds, 781 F.2d 185 (11th Cir.1986). The taint transfers, however, only if the first confession was inadmissible as involuntary, i.e., coerced and not the product of a free will. See Oregon v. Elstad, 470 U.S. 298, 105 S.Ct. 1285, 1294, 84 L.Ed.2d 222, 231 (1985); Martin, 770 F.2d at 928. This circuit has held that not honoring a request to stop questioning is no different from failing to give the Miranda warning in the first place; while both are “technical” violations of Miranda, neither violates the fifth amendment. Martin, 770 F.2d at 928-29 (relying on Elstad, 470 U.S. at 308, 105 S.Ct. at 1293, 84 L.Ed.2d at 231). Thus, confessions obtained by such violations, while inadmissible because they run afoul of Miranda’s per se bar, are not “involuntary”5 and do not taint any subsequent confessions. Id. Therefore, the second, taped confession was not made inadmissible even if the first confession resulted from technical Miranda violations.
Furthermore, even if the first confession was both violative of Miranda and involuntary, Elledge cannot prove he was prejudiced by admission of the second confession because it was sufficiently distinct from the first confession and, therefore, admissible. See Elstad, 470 U.S. at 309, 105 S.Ct. at 1294, 84 L.Ed.2d at 232-33 (dicta). When an earlier confession has been coerced and, thus, was involuntary, a court seeking to determine whether a subsequent confession is tainted thereby must look to “the time that passes between confessions, the change in place of interrogations, and the change in identity of the interrogators.” Id.; see Holleman v. Duckworth, 700 F.2d 391, 396 (7th Cir.1983); United States v. Gresham, 585 F.2d 103, 108 (5th. Cir.1978).6
In Elledge’s case, a full day had passed between the confessions; he had *1444slept and eaten; new interrogators were employed (although his original interrogator was present for part of the new questioning); and the interrogation occurred in entirely different and comfortable surroundings. Additionally, his inquisitors did not use the first confession as leverage to coerce the second. See Gresham, 585 F.2d at 108.
Consequently, if counsel had succeeded in suppressing the first confession by means of a Mosley -type attack, the second confession nevertheless would have been admissible. Therefore, even were we to assume that counsel’s failure to adopt such a strategy rendered his performance unreasonable, Elledge cannot demonstrate that any prejudice inhered as a result. Counsel’s performance thus was not ineffective under Strickland.
II. Effectiveness of Representation at the Sentencing Hearing7
The district court held an evidentiary hearing and concluded that Elledge’s counsel rendered unreasonable assistance at the sentencing phase because he failed to investigate and to present a psychiatric mitigating defense based on his client’s mental deficiencies, abject childhood and history of drug and alcohol abuse. That court nevertheless determined that counsel was not ineffective because the failure did not prejudice Elledge: he would have received the same sentence even had such evidence been adduced. See Strickland, 466 U.S. at 695, 104 S.Ct. at 2069, 80 L.Ed.2d at 698. Because these conclusions are subject to plenary review by this court,8 id. at 698, 104 S.Ct. at 2070, 80 L.Ed.2d at 700, we shall discuss each separately.
A. Performance
As discussed earlier, the first prong in an ineffective assistance claim is whether counsel’s performance was professionally unreasonable. A reviewing court, however, must be highly deferential in scrutinizing counsel’s performance; the tendency and temptation to second-guess is high and must be avoided. Strickland, 466 U.S. at 689, 104 S.Ct. at 2065, 80 L.Ed.2d at 694. Thus, we look to the particular facts of the case and determine whether the acts or omissions were “outside the wide range of professionally competent assistance” to the extent that the errors caused the “adversarial testing process” not to work. Id. at 690, 104 S.Ct. at 2066, 80 L.Ed.2d at 695.
In this case, the claim is that counsel failed to investigate Elledge’s past.9 Specifically, the district court found that counsel made no effort either to locate an expert psychiatric witness or to put on background character testimony from family members in mitigation.10 This factual *1445determination is not clearly erroneous.11 See id. at 698, 104 S.Ct. at 2070, 80 L.Ed.2d at 700. As the court noted, “counsel presented a psychiatric defense without a psychiatrist,” relying instead solely on El-ledge’s “free-form” testimony about his background and mental impairments.
We accept the district court’s view that counsel’s failure at least to interrogate El-ledge’s relatives and to seek an expert witness was outside the range of competent assistance. See Tyler v. Kemp, 755 F.2d 741, 744-45 (11th Cir.1985), cert. denied sub nom. James v. Tyler, 474 U.S. 1026, 106 S.Ct. 582, 88 L.Ed.2d 564 (1985), overruled in part (on other grounds), Peek v. Kemp, 784 F.2d 1479, 1494 n. 15 (11th Cir.1986); King v. Strickland, 748 F.2d 1462, 1463-64 (11th Cir.1984), cert. denied, 471 U.S. 1016, 105 S.Ct. 2020, 85 L.Ed.2d 301 (1985). An attorney has a duty to undertake reasonable investigation or “to make a reasonable decision that makes particular investigations unnecessary.” Strickland, 466 U.S. at 691, 104 S.Ct. at 2066, 80 L.Ed.2d at 695. Elledge made his counsel aware of his unhappy and abused past; yet counsel did not even interrogate Elledge’s family members to ascertain the veracity of the account or their willingness to testify. He also did not seek out potentially helpful expert witnesses.
Briefly stated, counsel’s total failure to investigate possible witnesses, both expert and lay, when he was aware of Elledge’s past and knew that mitigation was his client’s sole defense, was unprofessional performance.
B. Prejudice
1. Whether a favorable psychiatrist could have been found with reasonable diligence
The district court stated that “counsel for the respondent conceded that a psychiatrist such as Dr. Dorothy Lewis, who testified in Mr. Elledge’s defense during the evidentiary hearing before this Court, could have been located in 1977 to testify during the sentencing proceeding....”12 This led the court to conclude that counsel’s performance fell below the standard set out in Tyler v. Kemp, 755 F.2d 741, 744-45 (11th Cir.1985).13 The court went on to conclude, however, that even had counsel produced such a witness the death sentence nevertheless would have been imposed. Consequently, Elledge was not prejudiced; counsel was not ineffective under Strickland; and the sixth amendment was not violated.
Although the district court’s conclusion that the sixth amendment had not been *1446violated is correct, the analytic framework the court used to determine that a favorable witness could have been located was inaccurate. The record reveals that the State's counsel merely acknowledged that Dr. Lewis was extant in 1977 and had formulated her clinical theories at that time. The State never conceded that a reasonably diligent investigation would have uncovered either Dr. Lewis or a similar expert who would have testified favorably at Elledge's sentencing.
Specifically, the district court concluded that, in 1977, counsel "could have located" Dr. Lewis. The test, however, is not simply whether counsel "could have located" a witness similar to the one eventually produced. Instead, the court must determine whether it is reasonably likely that a reasonable attorney, operating under the circumstances of the case and acting in a reasonably professional manner, would have located such a witness.14
In other words, Strickland requires only that counsel conduct a reasonable investigation. Strickland, 466 U.S. at 691, 104 S.Ct. at 1066. To prove that he was prejudiced by counsel's failure to investigate and to produce a certain type of expert witness, a habeas petitioner must dem-onstrat~ a reasonable likelihood that an ordinarily competent attorney conducting a reasonable investigation would have found an expert similar to the one eventually produced. If such a result was not reasonably probable, the petitioner was not prejudiced by counsel's failure to investigate. Merely proving that someone-years later-located an expert who will testify favorably is irrelevant unless the petitioner, the eventual expert, counsel or some other person can establish a reasonable likelihood that a similar expert could have been found at the pertinent time by an ordinarily competent attorney using reasonably diligent effort.15
In deciding whether a petitioner has met this burden a court must look to all the circumstances of the case and consider all the evidence presented.16 See Strickland, 466 U.S. at 695, 104 S.Ct. at 2069, 80 L.Ed.2d at 698. In Elledge's case, he has made no showing that it was reasonably probable that an ordinary, reasonable lawyer, operating under the time and monetary constraints Elledge's counsel faced and using reasonable diligence, would have discovered a psychiatrist who would have *1447testified as did Dr. Lewis.17 Accordingly, Elledge cannot demonstrate that he was prejudiced by counsel’s failure to investigate his mental condition and produce a favorable expert witness.
2. Effect of the psychiatric and family testimony on the ultimate sentence.
Even if Elledge’s counsel had produced Dr. Lewis and Elledge’s family members at the sentencing phase, we agree with the district court that Elledge was not prejudiced thereby: he nevertheless would have received the death penalty.
The value of Dr. Lewis’s testimony was undercut in part by the revelation that her analysis largely relied on Elledge’s recitations and had not been fully corroborated by independent follow-up investigation. In addition, the two court-appointed psychiatrists who examined Elledge each ' gave damaging evaluations that would have diluted Dr. Lewis’s impact.18 Moreover, much of the testimony elicited from El-ledge’s brother and sister could be used against him; e.g., their descriptions of his early violent temper, his sister’s explanation of his alleged incestuous assault on her, his brother’s description of Elledge as “a mean guy,” and their emergence as normal citizens even though they had been subjected to similar abuse and neglect. The family testimony also was cumulative to a degree since Elledge had testified to many of the particulars in question.
As a final point, the aggravating circumstances of the case were substantial. It cannot be gainsaid that the cruelty of the rape and the murder made it more difficult for Elledge to alter the final sentence by adducing mitigating circumstances. It is proper for a reviewing court, in deciding \yhether the additional evidence would have altered the eventual sentence, to consider the strength of the case presented against the defendant.19 See Strickland, 466 U.S. at 696, 104 S.Ct. at 2069, 80 L.Ed.2d at 699 (“a verdict or conclusion only weakly supported by the record is more likely to have been affected by errors than one with overwhelming record support.”)
Elledge nevertheless maintains that if the psychiatric mitigating circumstances had been presented and if those circumstances had been credited by the judge and jury, the “entire evidentiary picture” would have been altered. See id. at 696, 104 S.Ct. at 2069, 80 L.Ed.2d at 699. Psychiatric mitigating evidence has this potential because it may impact the causal relationship that can exist between mental illness and homicidal behavior. Thus, psychiatric mitigating evidence not only can act in mitigation, it also could significantly weaken the aggravating factors. See Huckaby v. State, 343 So.2d 29, 33-34 (Fla.1977), cert. *1448denied, 434 U.s. 920, 98 S.Ct. 393, 54 L.Ed.2d 276 (1977).
This argument fails, however, because a careful reading of the district court's order shows that the court-acting within its discretion as factfinder-gave little weight to the testimony of Dr. Lewis as well as that of Elledge's family members. The district court simply found that no significant mitigating evidence was adduced. Further, when the court weighed the value of Dr. Lewis's testimony, it found that the aggravating factors outweighed those presented in mitigation. See, e.g., Mann v. State, 453 So.2d 784, 785 (Fla.1984), cert. denied, 469 U.S. 1181, 105 S.Ct. 940, 83 L.Ed.2d 953 (1985); Adams v. State, 412 So.2d 850, 854, 857 (Fla.1982) (three mitigating factors, including that the capital offense was committed while defendant was under the influence of extreme mental or emotional disturbance, outweighed by three aggravating circumstances), cert. denied, - U.S. , 106 S.Ct. 1506, 89 L.Ed.2d 906 (1986). There is no indication that the district court applied anything other than the "reasonable probability" standard or failed to weigh the overall impact of the evidence on the total evidentiary picture.
The foregoing discussion illustrates that Elledge has not demonstrated a reasonable probability that, if adduced at trial, the psychiatric and background evidence presented in his habeas proceeding would have caused the sentencer to conclude "that the balance of aggravating and mitigating circumstances did not warrant death." Strickland, 466 U.S. at 695, 104 S.Ct. at 2069, 80 L.Ed.2d at 698.
III. Consideration of Nonstatutory Mitigating Circumstances
Elledge asserts that he was denied the individualized sentencing determination demanded by the eighth amendment because the sentencing judge restricted his consideration solely to statutory mitigating factors. See generally Lockett v. Ohio, 438 U.S. 586, 98 S.Ct. 2954, 57 L.Ed.2d 973 (1978).
The confusion in Florida law concerning the consideration of nonstatutory mitigating circumstances-confusion that lasted from 1972 to 1978-has been discussed at length by this court. Hitchcock v. Wainwright, 770 F.2d 1514, 1516 (11th Cir.1985) (en banc), rev'd sub nom. Hitchcock v. Dugger, - U.S. -, 107 S.Ct. 1821, 95 L.Ed.2d 347 (1987). The period of confusion, which was exacerbated by the decision in Cooper v. State, 336 So.2d 1133 (Fla.1976), cert. denied, 431 U.S. 925, 97 S.Ct. 2200, 53 L.Ed.2d 239 (1977), finally was put to rest in Songer v. State, 365 So.2d 696 (Fla.1978), cert. denied, 441 U.S. 956, 99 S.Ct. 2185, 60 L.Ed.2d 1060 (1979), after the United States Supreme Court decided in Lockett that the eighth and fourteenth amendments require the sentencer to consider all mitigating evidence. Hitchcock, 770 F.2d at 1516.
Elledge, who was sentenced after Cooper and before Songer, concedes that the trial court allowed him to introduce non-statutory mitigating evidence; he also acknowledges that his attorney did not feel precluded from presenting such evidence. Elledge maintains, however, that the trial judge did not consider the nonstatutory evidence that was adduced and that the jury was improperly instructed to consider only the statutorily enumerated factors.
Recently, in Hitchcock v. Dugger, - U.S. -, 107 S.Ct. 1821, 95 L.Ed.2d 347 (1987), the Supreme Court overturned a Florida-imposed death penalty, basing its decision on a somewhat similar Lockett claim. The sentencing proceedings in Hitchcock -including the judge's instruction to the jury, the arguments to the jury, and the trial court's admission of evidence of nonstatutory mitigation-resemble El-ledge's proceedings.
We nevertheless find Hitchcock inapplicable to these facts. Hitchcock did not create a per se rule of reversal when the tria1 judge gives a particular jury instruction.20 Instead, the Court focused on the *1449specific facts of the sentencing proceeding and emphasized that both the judge and the jury believed themselves to be limited to statutory mitigating factors. See id,., 107 S.Ct. at 1823. Although parts of the jury instructions in Elledge’s case were virtually identical to those in Hitchcock, the trial judge in Hitchcock, when sentencing the defendant, indicated he considered solely the statutory factors: “there were insufficient mitigating circumstances as enumerated in Fla. Statute 921.141(b) to outweigh the aggravating circumstances.” Id., 107 S.Ct. at 1824 (emphasis in the original). In the case before us, there is no such statement indicating an improperly narrow focus by Elledge’s trial judge.
Moreover, we have in this record an affirmative statement by the state trial judge that indicates he did not feel he was limited to the statutory mitigating factors. Specifically, in discussing whether a particular piece of evidence, which was of a nonstat-utory mitigating nature, should be allowed into evidence, the judge said, “I don’t think I am limited on mitigation to a specific statute.” Tr. of Sentencing Proceedings, Vol. 8 at 246.
The differing factual setting presented here persuades us that Hitchcock is not dispositive of this case. Even assuming that the instruction to the jury was erroneous under Hitchcock, the sentencing jury in Florida’s trifurcated capital scheme is merely advisory. The trial judge, alone, makes the ultimate decision as to sentencing in capital cases. See Proffitt v. Florida, 428 U.S. 242, 252, 96 S.Ct. 2960, 2966, 45 L.Ed.2d 913 (1976). When the trial judge has the proper view of the law — as is evident from the record here — and imposes sentence based not only on statutory, but also on nonstatutory, factors, the resulting sentence meets the constitutional parameters outlined in Lockett.
Elledge argues, however, that his trial judge’s improper view of the law is demonstrated by an earlier case in which the judge who sentenced Elledge was adjudicated by the Florida Supreme Court to have held “the mistaken belief that he could not consider nonstatutory mitigating circumstances.” Jacobs v. State, 396 So.2d 713, 718 (Fla.1981). Elledge also cites the record of a second earlier case, State v. Rose, which allegedly indicates that the same trial judge held an improper view. The death penalty in Rose eventually was overturned on other grounds. 425 So.2d 521 (Fla.1983), cert. denied, 471 U.S. 1143, 105 S.Ct. 2689, 86 L.Ed.2d 706 (1985).
While this raises a concern about the trial judge’s perceptions, we are unpersuaded. The two trials Elledge points to preceded Elledge’s case; had they bracketed his trial they would be more conclusive. Additionally, the Florida Supreme Court, which reversed the trial judge for his mistaken views in Jacobs, was presented on appeal with the argument that the judge held that same, improper view in Elledge’s trial. Although obviously aware of the possible problem — to the point of having overturned a death sentence on those grounds in Jacobs —the state supreme court found that the judge held no such erroneous perception in Elledge’s case. We agree.
In sum, the record reflects no manifestations that the trial judge held an improper view toward nonstatutory mitigating evidence.
IV. The Enmund v. Florida claim
Elledge maintains that his death sentence violated his eighth amendment rights as articulated in Enmund v. Florida, 458 U.S. 782, 102 S.Ct. 3368, 73 L.Ed.2d 1140 (1982). The key to his argument is an attempt to distinguish “causation” from “culpability.”
Active participation in the murder vitiates the Enmund concerns. Hitchcock v. Wainwright, 745 F.2d 1332, 1340 (11th Cir. 1984), aff'd, 770 F.2d 1514 (11th Cir.1985) (en banc), rev’d on other grounds sub. nom., — U.S. -, 107 S.Ct. 1821, 95 L.Ed.2d 347 (1987); Drake v. Francis, 727 F.2d 990, 997 (11th Cir.1984), aff'd in part, rev’d in part, remanded in part (all on other grounds) sub nom. Drake v. Kemp, *1450762 F.2d 1449 (11th Cir.1985) (en banc). There was no Enmund-type violation in this case.21
V. Disparate Application of the Death Penalty
Elledge argues that the death pen- . alty is applied in an unconstitutionally discriminatory manner in Florida. He points to empirical studies which indicate that, when the victim is white, the death sentence is imposed in Florida at an aberra-tionally higher rate.
The Supreme Court recently rejected this same claim as based on a study of Georgia’s application of the death penalty. See McCleskey v. Kemp, — U.S. -, 107 S.Ct. 1756, 95 L.Ed.2d 262 (1987). The Court concluded that a petitioner who raises such an allegation “must prove that the decisionmakers in his case acted with discriminatory purpose”. Id., 107 S.Ct. at 1766 (emphasis in original). Elledge’s claim cannot pass this scrutiny.
VI. Shackling
Just prior to selecting the jury that would sentence Elledge, the state trial judge announced that a law enforcement official had informed the judge of two matters relating to Elledge. First, while incarcerated, Elledge had stated that, “because he had nothing to lose,” he intended to assault the courtroom bailiff. Second, the law enforcement official also had informed the judge that, while in jail, Elledge had become proficient at karate. Accordingly, the trial judge — acting on his own accord— ordered Elledge placed in leg irons for the duration of the sentencing phase.
Defense counsel objected to the shackling order; the trial judge overruled the objection. Elledge then asked if he could “say something” and the trial judge refused. Elledge was shackled and remained so for the duration of his sentencing trial.
There is a troublesome question as to whether the appearance in shackles of a defendant whom the jury has just convicted of a gruesome crime is so inherently prejudicial that he is thereby denied his constitutional right to a fair capital sentencing proceeding.
Initially, the prejudice perceived when a defendant is seen in shackles by the jury involves the presumption of innocence. The issue has usually arisen in the context of a determination of guilt or innocence. Courts focus upon the prejudicial impact restraints have on the defendant's presumption of innocence. See, e.g., Allen v. Montgomery, 728 F.2d 1409, 1413 (11th Cir.1984); Collins v. State, 164 Ga.App. 482, 297 S.E.2d 503, 505 (1982); State v. Tolley, 290 N.C. 349, 226 S.E.2d 353, 367 (1976). Were this the sole rationale which prohibits courtroom shackling, the defendant here would have no case. The jury knows he is not innocent. Having just convicted him of a crime that makes him a candidate for capital punishment, he is no longer entitled to a presumption of innocence.
Arguments could be made that the jury’s view of such a convicted murderer would have no effect on the sentencing or, indeed, may benefit the defendant. A jury may be more inclined to give a life sentence if it feels a defendant can be securely confined, as' evidenced by the shackles. Seeing that the defendant can be properly restrained, it is not necessary to give the death sentence in order to protect against future harm. On the other hand, a jury might view the shackles as first hand evidence of future dangerousness and uncontrollable behavior which if unmanageable in the courtroom may also be unmanageable in prison, leaving death as a proper decision.
With no definitive answer as to how the shackling at sentencing would affect the jury, it might be appropriate to consider the issue on a case-by-case basis and make a judicial evaluation as to the effect on the jury in each particular case. Much as we might be inclined to follow this path, it *1451would not seem to be available if controlling precedents are faithfully followed. The problem is that the Supreme Court has not bottomed the prohibition against shackling on presumption of innocence alone. When a broader concern is brought into play, there seems to be no reason to restrict the principles to the guilt-innocence phase of the trial. The language of the Supreme Court and this court in opinions regarding shackling, gives full consideration to that broader concern.
The Supreme Court has characterized shackling as an “inherently prejudicial practice.” Holbrook v. Flynn, 475 U.S. 560,-, 106 S.Ct. 1340, 1845, 89 L.Ed.2d 525, 534 (1986).22 “Not only is it possible that the sight of shackles and gags might have a significant effect on the jury’s feelings about the defendant, but the use of this technique is itself something of an affront to the very dignity and decorum of judicial proceedings that the judge is seeking to uphold.” Illinois v. Allen, 397 U.S. 337, 344, 90 S.Ct. 1057, 1061, 25 L.Ed.2d 353 (1970). When shackling occurs, it must be subjected to “close judicial scrutiny,” Estelle v. Williams, 425 U.S. 501, 503-04, 96 S.Ct. 1691, 1692-93, 48 L.Ed.2d 126 (1976), to determine if there was an “essential state interest” furthered, Holbrook, 475 U.S. at —, 106 S.Ct. at 1345, 89 L.Ed.2d at 534, by compelling a defendant to wear shackles and whether less restrictive, less prejudicial methods of restraint were considered or could have been employed. Holbrook, 475 U.S. 560, 106 S.Ct. 1340, 89 L.Ed.2d 525; Woodard v. Perrin, 692 F.2d 220, 221 (1st Cir.1982); Hardin v. Estelle, 365 F.Supp. 39, 47 (U.D.Tex.), aff'd on other grounds, 484 F.2d 944 (5th Cir.1973).
This Court has fully adopted the broad concerns reflected in the Supreme Court opinions. Allen, 728 F.2d at 1413-14; Zygadlo v. Wainwright, 720 F.2d 1221, 1223-24 (11th Cir.1983), cert. denied, 466 U.S. 941, 104 S.Ct. 1921, 80 L.Ed.2d 468 (1984).
Putting the case in the same posture as it would be had the shackling occurred at the guilt-innocence stage of the trial, it is apparent that resentencing is required. Before selecting the jury to sentence El-ledge, the trial court suddenly announced that it had decided to shackle Elledge:
I received information from Chief Miro yesterday, the Chief of Detention for the Broward Sheriff’s Office.
He told me he received information from two of the Lieutenants — one in the jail, and one in the Detective Division — that Mr. Elledge had become a karate expert, while in prison — apparently; and he was going to attack my Bailiff, either on the way to the courtroom or in the courtroom, today, because he had nothing to lose — he is now serving two life sentences, plus fifty years for rape; and he is facing a death penalty, or another life sentence, arising out of this situation; that he has been adjudged guilty of these particular crimes.
In an abundance of precaution, I entered an Order to leave Mr. Elledge in leg irons here in these proceedings.
MR. McCAIN: For the record, the defense would object to it, in the presence of the jury, having the belief it would unduly influence and prejudice the jury against the defendant.
THE COURT: Fine. I think we ought to put it on the record.
THE DEFENDANT: May I say something, your Honor.
THE COURT: No sir. We will be in recess.
First, Elledge was denied the required procedure when the court refused him an adequate opportunity to challenge the untested information that served as the basis for the shackling. Under the decided case law, the court should have given the defense a reasonable opportunity to meet the surprise information or at the very least should have allowed Elledge the opportunity to speak with his attorney, who *1452then could have possibly made a more specific argument. Zygadlo v. Wainwright, 720 F.2d at 1223-24 (noting that due process may require an evidentiary hearing if the factual basis for security procedures was in dispute); State v. Moen, 94 Idaho 477, 491 P.2d 858, 860-61 (1971) (defendant should be afforded a reasonable opportunity to meet the information); Tolley, 226 S.E.2d at 368 (evidentiary hearing is required). Moreover, in Bowers v. Maryland, 306 Md. 120, 507 A.2d 1072, 1078-81, cert. denied, - U.S.-, 107 S.Ct. 292, 93 L.Ed.2d 265 (1986), the only case that squarely addresses shackling at the capital sentencing stage, the State’s highest court found no error with the shackling after the defendant was given an opportunity to contest the necessity of shackling.
The record does not show whether El-ledge was going to admit these allegations, contest them as false, or provide rebuttal evidence to indicate that he was behaving well as a prisoner, the last option recently recognized as mitigating evidence which must be considered by the sentencer. Skipper v. South Carolina, 476 U.S. 1, 106 S.Ct. 1669, 90 L.Ed.2d 1 (1986). Only with an adversary process can the reliable evidence be sorted out from the unreliable. Future dangerousness can be determined only “when the convicted felon has the opportunity to present his own side of the case.” Barefoot v. Estelle, 463 U.S. 880, 901, 103 S.Ct. 3383, 3398, 77 L.Ed.2d 1090 (1983).
The second problem with the shackling decision is that the State at no time made any showing that the shackling was necessary to further an essential state interest. Needless to say, the security and safety of a state’s courtrooms is an essential state interest. Holbrook, 475 U.S. at -, 106 S.Ct. at 1345, 89 L.Ed.2d at 534. There is no indication that the trial court considered alternative restraints, as the court is required to do. Id. at-, 106 S.Ct. at 1342-44, at 530-32; Woodard, 692 F.2d at 221; Hardin, 365 F.Supp. at 47. Safeguards deemed necessary by other courts dealing with the issue are absent in Elledge’s case. The trial court never “polled the jurors to determine whether any of them would be prejudiced by the fact that the defendant was under restraints.” Woodard, 692 F.2d at 222. See also Bowers, 507 A.2d at 1081 (voir dire adequate to screen out one juror who indicated shackling would influence him). The trial court also gave no specific cautionary instruction. See, e.g., Billups v. Garrison, 718 F.2d 665, 668 (4th Cir.1983); Commonwealth v. Brown, 364 Mass. 471, 305 N.E.2d 830, 834 (Mass.1973).
Since these matters could not now be satisfactorily determined by an evidentiary hearing in federal district court after the passage of ten years since sentencing, it is quite appropriate that a writ of habeas corpus issue as to the capital sentence, with the State to have the right to conduct a constitutional sentencing procedure.
Conclusion
The district court’s judgment is VACATED and the case is REMANDED to the district court with directions to issue the writ of habeas corpus setting aside the sentence of death in the absence of a new sentencing trial.
. The facts and details of this crime spree are set forth in Elledge v. State, 346 So.2d 998 (Fla. 1977).
. The Florida Supreme Court affirmed Elledge’s death sentence. Elledge v. State, 408 So.2d 1021 (Fla. 1981), cert. denied, 459 U.S. 981, 103 S.Ct. 316, 74 L.Ed.2d 293 (1982). Elledge then sought and was denied collateral relief pursuant to Fla.R.Crim.P. Rule 3.850. The Florida Supreme Court affirmed. Elledge v. Graham, 432 So.2d 35 (Fla.), cert. denied, 464 U.S. 986, 104 S.Ct. 436, 78 L.Ed.2d 368 (1983).
. Elledge maintains that this daze was the result of three people dying in such a short period of time.
. Elledge now contends that the police repeatedly reinterrogated him — albeit after warning him of his Miranda rights on each occasion — after only short periods of time had passed.
. Whether a statement is "voluntary" requires application of pre-Miranda due process volun-tariness analysis. Martin, 770 F.2d at 928. That is, the court must look at all the circumstances to determine whether the confession was made of the accused’s free will. Id. at 925.
. In Bonner v. City of Prichard, 661 F.2d 1206 (11th Cir.1981) (en banc), this court adopted as precedent all decisions of the former Fifth Circuit Court of Appeals decided prior to October 1, 1981.
. The district court considered this to be El-ledge’s sole claim with "arguable” merit.
. The underlying factual findings of the district court are presumptively correct unless clearly erroneous, and the factual findings of the state court are subject to the 28 U.S.C. Sec. 2254(d) presumption of correctness.
. Strickland requires only that counsel conduct a reasonable investigation or make a reasonable decision not to investigate. Strickland, 466 U.S. at 691, 104 S.Ct. at 1066. Establishing what a reasonable investigation would have entailed often will be pivotal. Testimony from lawyers normally will be necessary to establish what the reasonable level and quality of investigation would be. In some extreme cases, such as where no investigation occurred, a judge can conclude, even in the absence of expert testimony, that the investigation — or lack thereof — was deficient. We devote little attention to this first step, in this case, because the court correctly concluded that, under these circumstances, some investigation was required.
.Elledge's counsel’s performance was not inadequate, according to the state, since counsel was aware that an unbroken trail of psychiatric evaluation unanimously had not found Elledge to suffer from an organic brain dysfunction. Given this diagnostic history, the State claims it was a reasonable decision to conduct no investigation. Were these the sole facts presented we might agree; but other relevant facts require a different conclusion.
Elledge’s counsel was convinced in 1975 that Elledge was "crazy”; he continued to hold that conviction in 1977 (i.e., prior to the second sentencing hearing) despite two court-ordered psychiatric evaluations to the contrary. Counsel then discovered, in talking with Elledge, that prison authorities had been treating Elledge with Mellaril and Dilantin while he was incarcerated pending the 1977 sentencing hearing. Mellaril is an antipsychotic medication and Di*1445lantin is an antiepileptic medication. Despite this new information, which supported his already held conviction that his client was "crazy,” counsel did not seek new expert advice. Counsel’s failure to pursue the issue, once its viability was renewed, was inadequate professional performance.
.When a defendant informs counsel that "certain investigations would be fruitless or even harmful, counsel's failure to pursue those investigations may not later be challenged as unreasonable.” Strickland, 466 U.S. v. 691, 104 S.Ct. at 1066. The record is not completely clear whether Elledge’s counsel ever asked his client about possible testimony by family members. All indications, however, are that no such questioning occurred. Elledge did eventually imply that he did not feel his family would be particularly interested. Although the record is somewhat opaque, we feel the district court was not clearly wrong in deciding that counsel was aware that family members existed who could corroborate, and perhaps expand on, Elledge’s testimony. The court also was not clearly erroneous in concluding that Elledge had not dissuaded counsel from undertaking an investigation to determine the willingness of such family members to testify. We note that Elledge’s brother and sister later stated that had they been approached in 1977 they would have testified.
. Dr. Dorothy Lewis testified in Elledge’s defense at the federal habeas hearing. She found that he suffered from organic brain dysfunction, episodic rages and paranoid behavior. She also stated that in her opinion Elledge became especially violent when he felt he was teased, tricked or in danger. Her conclusion was that the combination of organic brain dysfunction, psychotic paranoia and childhood abuse caused a disorder that affected Elledge during the Strack homicide.
. The court also found that the failure to present the witnesses was not a decision of trial strategy. See Thomas v. Wainwright, 767 F.2d 738, 744 (11th Cir.1985), cert. denied, — U.S. -, 106 S.Ct. 1241, 89 L.Ed.2d 349 (1986).
. In making this determination, the court must look to what constitutes a reasonable effort; i.e., what a competent attorney would have done given the constraints of time and money under the facts of the case. Merely producing Dr. Lewis, several years later, is of no moment. We note that Dr. Lewis was practicing psychiatry in New Haven, Connecticut in 1977; it strikes us as extremely doubtful that Elledge's counsel would have discovered Dr. Lewis in 1977 with a reasonably diligent investigation. We also note that Dr. Lewis neither testified that she would have come to Florida to testify in 1977 nor that she knew of anyone in Florida who shared her views at that time. We make no comment on whether a similar expert was reasonably discoverable-we merely hold that the record before us reveals no evidence demonstrating such a likelihood.
. Simply put, the Strickland test requires a habeas petitioner in Elledge's position to show: (a) that it was professionally unreasonable for counsel not to investigate; (b) what kind of, and how much, investigation an ordinary, reasonable lawyer would have undertaken; (c) that it is reasonably probable that a reasonable investigation would have turned up an expert who would have presented testimony similar to that which was eventually adduced; and (d) that it is reasonably probable that this testimony would have affected the sentence eventually imposed. Failure to meet any of these steps defeats the ineffectiveness claim.
. We stress that a reviewing court need not address each aspect of the Strickland analysis before concluding that counsel was not ineffective. Strickland, 466 U.S. at 697, 104 S.Ct. at 2069-70. Once the court determines that a ha-beas petitioner cannot meet any one of the various steps, the court need go no further. Id. For example, a district court may simply elect not to address whether an expert was reasonably available and instead conclude that counsel was not ineffective because no prejudice inhered in that such testimony would not have affected the ultimate sentence imposed.
In this case, however, the district court addressed the "availability" issue. Although the "availability" step was not outcome determinative in this case-because the district court properly concluded that Elledge nevertheless would have received the death penalty-it is important for us to set out the proper framework for future cases where it may be.
.To prove it is reasonably probable that such an expert could have been found by a competent lawyer exercising a reasonable amount of diligence, a petitioner could present testimony from (a) members of the bar relating to the amount of investigation that is reasonable in such a situation and the ease or difficulty in finding such experts at that time, (b) psychiatrists, or other experts in the field, relating to how widely the proposed theory was accepted at the time the investigation occurred and the ease an attorney would have had in getting such experts, and (c) any other relevant testimony that would tend to demonstrate it was reasonably probable that reasonable diligence would uncover an expert similar to the one eventually located.
We emphasize that the duty is only to conduct a reasonable investigation. Counsel is not required to "shop” for a psychiatrist who will testify in a particular way. See Barfield v. Harris, 540 F.Supp. 451, 457-59 (E.D.N.C.1982), aff’d, 719 F.2d 58 (4th Cir.1983), cert. denied, 467 U.S. 1210, 104 S.Ct. 2401, 81 L.Ed.2d 357 (1984).
. Of all the several experts who examined El-ledge over the years, only Dr. Lewis found him to be operating under extreme emotional or emotional disturbance caused by organic brain damage. The great weight of expert testimony clearly cut against Dr. Lewis’s testimony and made it less persuasive.
. We note and reject Elledge’s claim that the district court misapplied Strickland when it stated that in light of the extreme aggravating circumstances of the Strack murder, “only the most convincing evidence in support of mitigating circumstances could create a reasonable probability ... that the balance of mitigating circumstances did not warrant death." This argument places far too much emphasis on one sentence in a long order; the order, taken as a whole, makes clear that the court applied the proper standard.
. The judge's instruction in Hitchcock involved a statement to the jury that "the mitigating circumstances which you may consider shall be the following," followed by a listing of the statutory factors. Hitchcock, 107 S.Ct. at 1824. The *1449state trial judge in Elledge’s case used a virtually identical instruction.
. We also note that the trial judge implicitly may have found Elledge intended to murder Strack when the judge found that the murder was committed to avoid arrest for rape. Such a factual finding, of course, would carry a presumption of correctness. Cabana v. Bullock, 474 U.S. 376, 106 S.Ct. 689, 697-98, 88 L.Ed.2d 704, 717-18 (1986).
. Nothing in Holbrook indicates that the Supreme Court did not intend its ruling to apply to the penalty phase of a capital case; furthermore, it is unreasonable to believe that the court made its rule in Holbrook unaware that capital case trials are bifurcated. We think Hol-brook means what it says.