KOGAN, Justice.

We have for review a direct appeal by Jerry Stokes from his conviction for first-degree murder and sentence of death, as well as his conviction and sentence for armed robbery. We have jurisdiction. Art. V, § 3(b)(1), Fla. Const. We reverse both convictions and sentences and remand this case to the trial court for proceedings consistent with this opinion.

At 9:02 a.m. on September 22, 1986, the body of Cilla Taylor was found behind the counter of her small Madison County country store. She had been shot once in the face. There were no witnesses to the murder, and no physical evidence was recovered, except tire tracks on the dirt driveway of the store.

Investigators interviewed William Brown, who had driven by Taylor's store just after 8:00 a.m. that morning. He saw a blue Pontiac Grand Prix automobile with a white roof parked outside the store. Brown, who had been late for work and was travelling at least sixty miles per hour at the time he passed the store, viewed the car for approximately two seconds. As an automobile enthusiast, he was able to describe the make, model, color, and approximate age of the car, as well as some distinguishing features such as a C.B. antenna and unusual wire wheel covers missing the center emblems.

Brown was taken to identify fifteen to twenty automobiles by investigators but was unable to positively identify any of those cars as the one he saw at Taylor's store. On October 17, 1986, Brown was hypnotized by Florida Department of Law Enforcement Special Agent J.O. Jackson in an attempt to obtain further details about the car. During the session, which was videotaped, Brown described the car as having, in addition to the details he had previously given, vertical lines on the tail-lights, whitewall tires, and a Florida license plate with the number 286 followed by the letter Z followed by two indecipherable letters.

On October 24, Brown was taken to Valdosta, Georgia to identify a car that had been found by Georgia authorities using Brown's original description. This was the first car Brown was shown following the hypnosis session. He identified the car as the one he had seen outside Cilla Taylor's store on the morning of September 22, 1986. That car, which belonged to the appellant's brother, Garfield Stokes, was a faded blue Pontiac Grand Prix with a white, half-vinyl roof. The wheels had wire wheel covers with the center emblem painted over, and tires that were nearly bald. The tread marks left by the tires were similar to the tread patterns left at Taylor's store. The car had horizontal, rather than vertical lines on the taillights, no whitewall tires, and the license plate was a Georgia tag with the number HZY 821.

Georgia authorities had been questioning appellant in connection with matters unrelated to the Taylor homicide. He admitted that he had a pistol and a shotgun, and police found a pistol in his Toyota automobile. The pistol was examined by a firearms expert who determined that it was the same caliber as the one used to kill Cilla Taylor. The pistol had a similar barrel configuration, although all .22 caliber pistols manufactured by the same company carried the same barrel configuration. The firearms expert also determined the gun was subject to accidental firing once the hammer was pulled back because the weapon had a dangerously light trigger pull.

Appellant was arrested and questioned about the Taylor murder. During questioning, appellant claimed that another man, Willie Thomas, had borrowed the gun and his brother's car. When questioned about sneaker tracks that had been found at the store, appellant responded, "I don't know why you want those shoes. I wasn't wearing them when. . . ." He did not complete the statement, and the sneaker tracks were those of a police officer investigating the crime. The remaining evidence against appellant involved testimony showing that while incarcerated in Valdosta, he confided in a fellow inmate, Lowell Woodson, that there had been a robbery and a lady was shot. He asked Woodson to call his brother, Garfield Stokes, to tell him to get rid of the car or tires.

Appellant was convicted of armed robbery and first-degree murder and sentenced to death, following a jury recommendation. He raises several arguments in connection with both the conviction and sentence. The most significant issue deals with the admission of William Brown's hypnotically refreshed testimony.

I HYPNOTICALLY REFRESHED TESTIMONY

Citing our decision in Bundy v. State, 471 So.2d 9 (Fla. 1985), cert. denied, 479 U.S. 894, 107 S.Ct. 295, 93 L.Ed.2d 269 (1986) ( Bundy II,) Stokes moved in limine to exclude Brown's posthypnotic descriptions, the identification of Garfield Stokes' car, and the hypnotic session in its entirety. The trial court excluded the session but ruled that because the posthypnotic statements were substantially similar to the prehypnotic statements, the descriptions and the identification were admissible. In Bundy II, we held that hypnotically refreshed testimony is per se inadmissible in criminal trials, but a witness who has been hypnotized is still competent to testify to those facts demonstrably recalled prior to hypnosis. Id. at 18. In that case we were persuaded by opinions from other jurisdictions as well as by the voluminous expert research and writing on the subject. These sources found that hypnotically refreshed testimony was simply too unreliable for use in criminal trials. This case compels us to once more examine the reliability and practical application of posthypnotic testimony.

While there is no consistently agreed upon definition of hypnosis, for our purposes we define hypnosis as "an altered state of awareness or perception." Sies Wester, Judicial Approaches to the Question of Hypnotically Refreshed Testimony: A History and Analysis, 35 De Paul L. Rev. 77, 79 (1985). Compare Black's Law Dictionary 668 (5th ed. 1979); Counsel on Mental Health, Medical Use of Hypnosis, 168 J.A.M.A. 186, 187 (1958). During hypnosis, the subject is placed in an artificially induced state of sleep or trance through a series of relaxation and concentration techniques employed by the hypnotist. Hypnosis has a wide variety of forensic applications and benefits and, under clinical circumstances, can be very worthwhile. Here, we are concerned only with the use of hypnosis to refresh the recollection of a witness to an event or a crime for the purpose of testifying to his or her recollection in court.

In the first American case to address the issue of hypnotism in the courts, the California Supreme Court declared that "the law of the United States does not recognize hypnotism." People v. Ebanks, 117 Cal. 652, 655, 49 P. 1049, 1053 (1897) (summary approval of the trial court's decision to exclude expert testimony concerning hypnotically induced statements). Although this skeptical view was predominant well into the twentieth century, in 1968, the Maryland Court of Special Appeals allowed the use of hypnosis for evidentiary purposes. Harding v. State, 5 Md. App. 230, 246 A.2d 302 (Ct. Spec. App. 1968), cert. denied, 395 U.S. 949, 89 S.Ct. 2030, 23 L.Ed.2d 468 (1969), overruled, Collins v. State, 52 Md. App. 186, 447 A.2d 1272 (Ct. Spec. App. 1982). Since then, hypnosis has taken a roller coaster ride through the courts, finding favor in some states, uncertainty in others, and complete disfavor in still others. This has been due, for the most part, to the numerous evidentiary problems associated with the use of hypnotically refreshed testimony.

In this case, we are concerned with the use of hypnotically refreshed testimony as evidence in a criminal trial. Accordingly, we must focus our attention on the reliability of this evidence rather than the many clinical and forensic benefits associated with hypnosis. The nature of hypnosis and memory reconstruction is such that several problems are raised by its use in court. These problems have been identified and summarized by Professor Bernard L. Diamond, M.D. in his article Inherent Problems in the use of Pretrial Hypnosis on a Prospective Witness, 68 Calif.L.Rev. 313 (1980) [hereinafter Diamond, Inherent Problems]. Dr. Diamond is a professor of law, a clinical professor of psychiatry, and a noted expert in the field of hypnosis. His article delineates several evidentiary problems associated with hypnotically manipulated recall. These concerns have been unveiled through extensive, dilligent research conducted by respected members of the scientific community. First, a hypnotized person is subject to a heightened degree of suggestibility. E.g., Council on Scientific Affairs, Scientific Status of Refreshing Recollection by the Use of Hypnosis, 253 J.A.M.A. 1918, 1922 (1895) [hereinafter A.M.A. Council Report]; Diamond, Inherent Problems, supra, at 333. ("Hypnosis is, almost by definition, a state of increased suggestibility.").

This heightened suggestibility leads to other problems which tend to render hypnotically refreshed testimony less reliable than testimony of a witness whose memory has not been refreshed through the use of hypnosis. For example, many researchers have concluded that a hypnotist, no matter how skilled, cannot avoid implanting intentional or inadvertant suggestions in the mind of the hypnotized subject. This occurs as much through nonverbal body language as through verbal cues. E. Hilgard, The Experience of Hypnosis 9 (1968); Diamond, Inherent Problems, supra, at 333. Furthermore, a hypnotic subject cannot, upon awakening, distinguish between his own thoughts and feelings and those which were implanted during the hypnosis session.

Another serious problem associated with the use of hypnotically refreshed testimony involves the tendency of the hypnotic subject to "confabulate," or invent details that he or she does not actually recall. Much research into the effects of hypnosis on the human memory has revealed that a hypnosis subject will invent or fabricate facts that he or she does not actually remember. Worse still, the subject is unable to distinguish between these confabulations and the true facts. In other words, hypnosis tends to force the subject to invent memories and to believe that they are true. Thus, neither the hypnotist nor the subject is able to separate fact from fantasy when the hypnosis session is completed. Diamond, Inherent Problems, supra, at 335-37.

Perhaps the most serious evidentiary problem associated with hypnosis involves the phenomenon known as "memory hardening." Memory hardening affects ones ability to resolve doubts and uncertainties resulting in the subject becoming certain of his or her memories regardless of the accuracy of those memories. A subject becomes certain of his or her recall of the events without foundation for that confidence. This memory hardening "creates special barriers to the court's truth-seeking ability." Note, The Admissibility of Posthypnotic Testimony: Constitutional Considerations and the Defendant's Right to Testify, 16 Fla. St. U.L. Rev. 185, 189 (1988). Hypnosis tends to "bolster a witness whose credibility would easily have been destroyed by cross-examination but who now becomes quite impervious to such efforts, repeating one particular version of his story with great conviction." Orne, The Use and Misuse of Hypnosis in Court, 27 Int'l J. Clinical Experimental Hypnosis 311, 332, (1979). Thus, a witness who has been hypnotized prior to testifying becomes very difficult to cross-examine on any subject discussed in the hypnosis session, raising questions which involve a defendant's sixth amendment right to confront witnesses against him. The defendant must face a "witness whose natural recollection may have been altered by suggestion or confabulation, but who nevertheless has a firm conviction as to its truth." Falk, Posthypnotic Testimony — Witness Competency and the Fulcrum of Procedural Safeguards, 57 St. John's L. Rev. 30, 54 (1979) (footnote omitted). The task of cross-examining such a witness therefore becomes an exercise in futility.

Recognizing these evidentiary concerns, courts have developed four different approaches to the admissibility of hypnotically refreshed testimony. These are per se admissibility of hypnotically refreshed testimony; conditional admissibility, providing that several procedural safeguards have been fulfilled; per se inadmissibility; and a balancing approach in accordance with rule 403 of the Federal Rules of Evidence. Taking these approaches chronologically, we begin with the per se admissible approach adopted by the Maryland Court of Special Appeals in Harding v. State.

In Harding, the Maryland court held that all hypnotically refreshed testimony is admissible in the same manner as other testimony. The court stated the general proposition that the fact that a witness has been hypnotized goes to the weight, not admissibility, of the evidence. This case sparked a number of decisions allowing the admission of hypnotically refreshed testimony with only the standard procedural safeguards used to ensure the reliability of normal testimony and evidence. See, e.g., Kline v. Ford Motor Co., 523 F.2d 1067 (9th Cir. 1975); People v. Smrekar, 68 Ill. App.3d 379, 385 N.E.2d 848 (1979); State v. McQueen, 295 N.C. 96, 244 S.E.2d 414 (1978), overruled, State v. Peoples, 311 N.C. 515, 319 S.E.2d 177 (1984); State v. Brom, 8 Or. App. 598, 494 P.2d 434 (Or. 1972). It is important to note that the psychiatrist who hypnotized the witness in the Harding trial testified that hypnosis does not dispose the subject to suggestion. Diamond, Inherent Problems, supra, at 322. Though contrary to the great weight of scientific evidence, see A.M.A. Council Report, supra, at 1922, this misconception was highly influential in the Harding decision and those decisions which followed.

Once the gate was opened to all hypnotically refreshed testimony, some abuse necessarily followed. Several cases arose where detectives and hypnotists exerted pressure on hypnotic subjects or introduced suggestions to the subject designed to compel the identification of a specified suspect. These cases led some courts to modify the rule of per se admissibility into one of conditioned admissibility. In one such case, following the suggestion of Dr. Martin Orne, a noted expert in the field of hypnosis research and editor of the International Journal of Clinical and Experimental Hypnosis, the Supreme Court of New Jersey allowed the introduction of hypnotically refreshed testimony only upon the satisfaction of several procedural safeguards. State v. Hurd, 86 N.J. 525, 432 A.2d 86 (1981).

In Hurd, the victim of a violent sexual assault was hypnotized by a psychiatrist in hopes that the victim could identify the perpetrator. During the hypnosis session and during the posthypnotic interview that followed, the hypnotist and a police detective pressured the victim into identifying her ex-husband. The court found that the hypnotically refreshed testimony could not be admitted unless the hypnosis session had complied with certain procedures. These procedures were based on several premises: 1) that courts have an aversion to the exclusion of relevant evidence; 2) that courts recognize the inherent risks of hypnotically refreshed testimony; and 3) that procedural safeguards could be fashioned to minimize or negate those risks. The court adopted safeguards in Hurd which required the following:

"(1) The hypnotic session should be conducted by a licensed psychiatrist or psychologist trained in the use of hypnosis.

(2) The qualified professional conducting the hypnotic session should be independent of and not responsible to the prosecutor, investigator or the defense.

(3) Any information given to the hypnotist by law enforcement personnel prior to the hypnotic session must be in written form so that subsequently the extent of the information the subject may have received from the hypnotist may be determined.

(4) Before induction of hypnosis, the hypnotist should obtain from the subject a detailed description of the facts as the subject remembers them, carefully avoiding adding any new elements to the witness' description of the events.

(5) All contacts between the hypnotist and the subject should be recorded so that a permanent record is available for comparison and study to establish that the witness has not received information or suggestion which might later be reported as having been first described by the subject during hypnosis. Videotape should be employed if possible, but should not be mandatory.

(6) Only the hypnotist and the subject should be present during any phase of the hypnotic session, including the prehypnotic testing and post-hypnotic interview."

These safeguards were not intended to ensure the reliability of the hypnotically refreshed testimony, but merely to curb the potential for abuse that had arisen under the per se admissibility approach. Some courts remained concerned that the reliability of this evidence was at least questionable, and at worst doubtful. This doubt was instilled because many of the evidentiary problems discussed above are not addressed by these procedural safeguards. Moreover, the process of determining compliance with these safeguards would prove difficult. Practically speaking, application of the Hurd rules troubled several states' courts. Accordingly, the courts in these states decided that hypnotically refreshed testimony was, by its nature, too unreliable for use in the courtroom.

By the time the California Supreme Court decided People v. Shirley, 31 Cal.3d 18, 181 Cal.Rptr. 243, 723 P.2d 1354, cert. denied, 459 U.S. 860, 103 S.Ct. 133, 74 L.Ed.2d 114 (1982), scientific opinion on the use of hypnotically refreshed memory in court had changed dramatically. Although the courts in some states had already reached the conclusion that hypnotically refreshed testimony was per se inadmissible, State v. Mack, 292 N.W.2d 764 (Minn. 1980); State v. Mena, 128 Ariz. 226, 624 P.2d 1274 (1981); Commonwealth v. Nazarovitch, 496 Pa. 97, 436 A.2d 170 (1981); State v. Palmer, 210 Neb. 206, 313 N.W.2d 648 (1981), cert. denied, 484 U.S. 872, 108 S.Ct. 206, 98 L.Ed.2d 157 (1987), the California court's decision in Shirley became the premier opinion discussing the unreliability of posthypnotic testimony.

In Shirley, the victim of an alleged sexual battery had little or no memory of the events surrounding the attack, particularly regarding the issues of consent and the use of force. At a preliminary hearing, the victim gave ambiguous testimony which seemed to indicate that she had consented to the defendant's actions. After this hearing, but prior to trial, the victim was hypnotized by a district attorney. Although she gave statements which were helpful to the prosecution, they contradicted her prehypnotic testimony. The defendant moved to suppress the statements, claiming they were "manufactured" by the hypnosis. Shirley, 31 Cal.3d at 29, 181 Cal.Rptr. at 248, 723 P.2d at 1359. The trial court denied the motion and allowed the statements, along with any impeachment testimony and other evidence concerning the inconsistent statements and the unreliability of hypnosis.

The California Supreme Court reversed Shirley's conviction, holding that "the testimony of a witness who has undergone hypnosis for the purpose of restoring his memory of the events in issue is inadmissible as to all matters relating to those events, from the time of the hypnotic session forward." Id. at 66-67, 181 Cal.Rptr. at 273, 723 P.2d at 1384. The court in Shirley presented an extensive history of hypnosis in the courts and then examined the different judicial approaches to the problems presented by hypnosis. Finally, the court based its holding on the principles set forth in Frye v. United States, 293 F. 1013 (D.C. Cir. 1923). In that case, the federal circuit court held that scientific evidence is not admissible unless "the thing from which the deduction is made [is] sufficiently established to have gained general acceptance in the particular field in which it belongs." Id. at 1014. The court held that truth serum tests should be excluded because they "have not attained sufficient scientific and psychological accuracy nor general recognition as being capable of definite and certain interpretation." Id. at 1026. The underlying theory for this rule is that a courtroom is not a laboratory, and as such it is not the place to conduct scientific experiments. If the scientific community considers a procedure or process unreliable for its own purposes, then the procedure must be considered less reliable for courtroom use.

Acknowledging this, the court in Shirley examined the extensive scientific literature, research, and testimony on the reliability of hypnotically refreshed memory, and determined that the scientific community was divided on the subject. Indeed, although some experts profess the belief that hypnotically refreshed testimony is reliable, many more experts have arrived at the opposite conclusion, successfully rebutting the arguments proffered by the proponents of admitting hypnotically refreshed testimony. Upon completion of its review of the available research from the scientific community, the court in Shirley determined that hypnotically refreshed memory had not gained sufficient acceptance to hurdle the Frye test. Shirley, 31 Cal.3d at 66, 181 Cal.Rptr. at 273, 723 P.2d at 1384.

The Shirley rule of per se inadmissibility has come under significant criticism from commentators for being "a draconian device." Nonetheless, courts have followed that rule more than any other judicial approach put forth. Most of the criticism stems from the belief that the per se rule of inadmissibility is overinclusive, rendering evidence inadmissible that would otherwise be reliable. While those who have been critical of Shirley acknowledge the extensive evidentiary problems raised by hypnosis, they believe that the problems are sufficiently overcome by the procedural safeguards set forth in Hurd or by the use of a balancing test which would weigh the probative value of the testimony against the danger of unfair prejudice.

This balancing test is drawn from Federal Rule of Evidence 403, which excludes relevant evidence if its probative value is "substantially outweighed by the danger of unfair prejudice, confusion of the issues, or misleading the jury, or by considerations of undue delay, waste of time, or needless presentation of cumulative evidence." Fed.R.Evid. 403. This approach is a relatively recent development which has yet to find much favor in the courts.

While the balancing approach solves the primary concern with the per se rule of exclusion, namely that it is too inflexible, the balancing approach may well take the concept of flexibility too far. By definition, any rule 403 analysis must be conducted on a case-by-case basis. A balancing of the probative value of evidence against the danger of unfair prejudice necessarily requires the court to closely examine the facts of each case to determine with precision the probative value of the proffered posthypnotic testimony and to carefully weigh the probative value against the dangers of unfair prejudice.

Although judges must consistently make this determination in trial, the weighing process becomes significantly more complicated when considering the translucent nature of hypnosis and hypnotically refreshed testimony. Doubtless such a determination would require the parties to call numerous expert witnesses to advocate or oppose the use of the testimony. This foreshadows an extremely expensive and time-consuming procedure preceding each trial in which posthypnotic testimony is sought to be introduced. Moreover, the balancing approach provides no guidelines for judges attempting to balance the probative value against the danger of unfair prejudice. Thus, while sufficiently flexible to allow admission of relevant, reliable posthypnotic testimony and to exclude testimony which is not reliable, the balancing approach is impractical and difficult to apply.

Upon consideration of the approaches to this problem, we believe that the test espoused in Frye properly addresses the issue of the admissibility of posthypnotic testimony. We acknowledge that the Frye rule has come under some criticism since its inception in 1923 as too harsh and inflexible, see McCormick on Evidence § 203 (2d ed. 1972); however, we believe that the problems associated with the other recognized judicial approaches foreclose their use. See Bundy II; Bundy v. State, 455 So.2d 330 (Fla. 1984) ( Bundy I), cert. denied, 476 U.S. 1109, 106 S.Ct. 1958, 90 L.Ed.2d 366 (1986).

Because our determination centers on the application of the Frye rule, we must examine the scientific research and literature to see whether opinions within the scientific community on the use of hypnotically refreshed testimony have changed since the Shirley decision was published in 1982. Although research into the question of the reliability of hypnotic memory continues, a review of the available literature shows the views of the scientific community have either remained divided or are leaning towards disapproval of hypnosis as a reliable means of accurately enhancing memory.

Dr. Martin Orne, the psychiatrist and hypnosis expert who proffered the procedural safeguards to the New Jersey court in Hurd, now contends that these safeguards are insufficient to protect against the inherent unreliability of hypnotically refreshed testimony. M. Orne, Hypnotically Induced Testimony, in Eyewitness Testimony: Psychological Perspectives 210 (Wells E. Loftus eds. 1984). Dr. Orne states that "[t]he present state of scientific knowledge is consistent with the rules of a number of state supreme courts that memories retrieved through hypnosis are sufficiently unreliable that their use is precluded as eyewitness testimony in criminal trials. . . ." Id. at 204. Dr. Orne concludes that "hypnotically induced testimony is not reliable and ought not be permitted to form the basis of testimony in court." Id.

In 1985, the American Medical Association Council on Scientific Affairs commissioned a panel report defining the scientific status of hypnotically refreshed memory. A.M.A. Council Report, supra. The council's report reviews the literature and research available at the time and concludes that the vast majority of credible research finds the use of hypnosis to refresh memory to be completely unreliable. The panel makes several recommendations. The first urges that hypnosis be used only in the investigative process, with no thought of preserving the hypnotic interview or using the subject's testimony at trial, Id. at 1922. The panel further recommended that the guidelines established in Hurd should be followed completely, even for investigative hypnosis. Id. at 1922-23.

Since 1985, when the A.M.A. Council Report was issued, research into the accuracy of hypnotically enhanced memory has slowed. Although work continues in the area, most credible sources have rejected hypnosis as an accurate and reliable means of refreshing recollection. Thus, it is clear that most members of the scientific community are still of the opinion that hypnotically refreshed memory is not reliable for use as testimony in court. Accordingly, we hold that the testimony of a witness who has undergone hypnosis for the purpose of refreshing his or her memory of the events at issue is inadmissible as to all additional facts relating to those events from the time of the hypnotic session forward. A witness who has been hypnotized may testify to statements made before the hypnotic session, if they are properly recorded. Any hypnosis session shall act as a time barrier, after which no identifications or statements may be admitted. See State v. Coe, 109 Wn.2d 832, 838, 750 P.2d 208, 211 (1988). As this rule is applied to this case, William Brown may testify to the description of the car given to police before the hypnotic session, if that description was properly recorded. However, Brown's hypnotic and posthypnotic statements are inadmissible, as is his posthypnotic identification of Garfield Stokes' car as the one he saw in front of Cilla Taylor's store.

II OTHER GUILT-PHASE ISSUES

We turn now to the other issues raised by appellant. First, he alleges violation of State v. Neil, 457 So.2d 481 (Fla. 1984) as refined by State v. Slappy, 522 So.2d 18 (Fla.), cert. denied, ___ U.S. ___, 108 S.Ct. 2873, 101 L.Ed.2d 909 (1988), in the state's use of peremptory strikes. Appellant is a black man accused of murdering a white woman. The record shows that the state used six out of seven peremptory challenges to exclude black prospective jurors. Following objections to several of these challenges on the basis of racial discrimination, the trial court declined to conduct any inquiry into their propriety. The state conceded at oral argument that appellant had met the threshold of establishing the likelihood that the peremptory strikes were used in a racially discriminatory manner. Thus the prosecution was required to carry its burden of demonstrating race-neutral reasons for the peremptory strikes. Slappy, 522 So.2d at 22. Assuming that appellant had met the threshold, we believe the trial court did not conduct an adequate inquiry into the state's reasons for the use of peremptory challenges against black prospective jurors. It is incumbent upon trial judges to conduct such an inquiry when the threshold showing of racial discriminatory use of peremptory challenges has been met. We decline the state's invitation to extract from the record the reasons it now believes justify the use of peremptory strikes. The proper time to do this was during voir dire, not on appeal. The proper tribunal to conduct the inquiry was the trial court, not the appellate court.

Appellant next argues the trial court erred in denying his motions for judgment of acquittal on the murder and the robbery counts. He contends that the purely circumstantial case against him was filled with too many improper inferences resulting in too many incorrect conclusions. We find no merit in this contention. However, for the reasons expressed above concerning the admission of posthypnotic testimony, we must reverse appellant's conviction for armed robbery.

III PENALTY PHASE ISSUES

In sentencing appellant to death, the trial court found two aggravating circumstances. The first aggravating factor, that the murder was committed during the course of a robbery, is supported by the record. The trial court also found that the murder was committed in a cold, calculated, and premeditated manner, without pretense of moral justification. The evidence of this factor is that appellant's gun (assuming that it was the murder weapon) must be cocked before firing. The trial court thus inferred that appellant would have had to go through two distinct motions to shoot Cilla Taylor. However, the record also demonstrates that appellant's pistol (again, assuming it was the murder weapon) was, due to a dangerously light trigger pull, vulnerable to accidental firing. This infers the possibility that the gun could have fired accidentally, without appellant taking the second action of pulling the trigger. Thus, the record is inconclusive as to whether the murder was cold, calculated, and premeditated. Moreover, we have stated that "[t]his aggravating factor is reserved primarily for execution or contract murders or witness-elimination killings." Hansbrough v. State, 509 So.2d 1081, 1086 (Fla. 1987); Bates v. State, 465 So.2d 490 (Fla. 1985). No evidence in the record supports this allegation, and therefore the aggravating circumstance of cold, calculated, and premeditated must be stricken.

Because this case must be remanded for a new trial, we need not determine whether the death penalty is proportionate in this case. Therefore, for the reasons stated herein, we reverse the convictions and sentences for first degree murder and robbery and remand both counts for a new trial.

It is so ordered.

EHRLICH, C.J., and BARKETT and GRIMES, JJ., concur.

SHAW, J., concurs in result only with an opinion.

OVERTON, J., concurs in result only.

McDONALD, J., concurs in guilt and concurs in result only with penalty.