— I dissent.
I agree that CALJIC No. 2.62 was improperly given for lack of evidentiary support, but I must depart from the holding of the majority that the error was not prejudicial.
I
The reasonable probability test of People v. Watson (1956) 46 Cal.2d 818, 836 [299 P.2d 243], governs, of course, in determining whether there has been a miscarriage of justice requiring reversal of the judgment under article VI, section 13, of the California Constitution. In applying that test, we have often emphasized that when a case is “closely balanced” it is more likely to be “reasonably probable that a verdict more favorable to the [appealing party] might have resulted if the error had not occurred.” (People v. Hannon (1977) 19 Cal.3d 588, 603 [138 Cal.Rptr. 885, 564 P.2d 1203]; accord, People v. Wagner (1975) 13 Cal.3d 612, 621 [119 Cal.Rptr. 457, 532 P.2d 105], and cases there cited.)
In Hannon the trial court, over the defendant’s objection, instructed the jury that it could consider certain evidence as a circumstance tending to show a consciousness of guilt. (19 Cal.3d at pp. 596-597.) On appeal from the conviction, we held there was no evidence in the record to support the disputed instruction and hence the trial court erred in giving it. (Id., at p. 600.) In assessing the effect of that error, we stressed that the People’s case rested in the final analysis on the strength of the identification testimony of one witness, which the defendant had sought to weaken through cross-examination and presentation of an alibi. (Id., at p. 602.) Because “an impermissible impact may have resulted in the minds of the jurors” from the erroneous instruction, we held the closeness of the case to be determinative of prejudice. (Id., at p. 603; see also People v. Anderson (1965) 63 Cal.2d 351, 360 [46 Cal.Rptr. 763, 406 P.2d 43]; People v. Underwood (1964) 61 Cal.2d 113, 125-126 [37 Cal.Rptr. 313, 389 P.2d 937].)
As shown by Hannon, we must be especially vigilant in appraising prejudice in close criminal cases, where an erroneous instruction might easily tip the delicate balance of evidence against the defendant.
*687II
In its discussion of the facts the majority acknowledge the weaknesses of the People’s evidence, yet decline to hold the error prejudicial. (Ante, pp. 683-684.) My examination of the record persuades me otherwise: if ever there were a close case, this is it.
Remarkably absent from the majority’s analysis of the evidence is any consideration of the undeniable fact that the robber, gave the names “Saddler” and “Brown” to Mrs. Overfield. Yet it is highly unlikely, to say the least, that a person contemplating commission of a crime would volunteer his own name to an eyewitness-victim; and it is equally incredible to conceive that Saddler, after some deliberation, would then offer “Brown” as the alleged maiden name of his nonexistent wife when that name was linked to him through a close relative residing under the same roof as he. This conspicuous fact, I submit, goes far to weaken the credibility of Mrs. Overfield’s identification of Saddler as the robber.
As in Hannon, the People’s case ultimately hinged on the strength of the identification testimony of its key witness, Mrs. Overfield. Contrary to the majority’s characterization of her testimony as “positive,” the record is replete with instances of her acknowledging lack of clear memory of the events in question. She testified on numerous occasions that she simply could not remember because “it’s been a long time” — seven and one-half months elapsed between the date of the robbery and her appearance at trial — and she frankly admitted she had “been trying to forget this since December.”
The majority next assert the circumstances of the robbery afforded Mrs. Overfield ample time to observe and study the person who robbed her; yet she testified she did not' know how long she talked to her assailant, she had “no idea” how often she looked directly at him, and, not being in the habit of staring at people, she merely “glanced” at him “occasionally.” Moreover, Mrs. Overfield admitted she noticed “nothing unusual” about the robber’s facial features, and supplied Officer Jones with only a general description of his height, weight, and age. Indeed, when asked the question on cross-examination, she conceded she was not a good judge of age or height. In an in-depth interview conducted by Detective Sessions 13 days after the incident, Mrs. Overfield was asked whether her assailant “had any outstanding physical features which would positively or possibly bring a more closer identification to the individual.” She could recall none and allowed that “she did not particularly pay attention to his features or his clothing.”
*688Neither is the photographic identification, on which the majority rely in part, persuasive. Although it is true Mrs. Overfield viewed a total of sixty-eight photos on three occasions, the second of these arrays consisted of only two photos, one of which was of a person of Mexican descent. More importantly, Mrs. Overfield testified that the accuracy of her eventual identification of Saddler’s photo was confirmed in her mind when she asked Detective Sessions whether the person she identified was named “Saddler.”
Finally, the identification testimony was without any corroboration whatever, for no reliance could reasonably be placed on so ambiguous a circumstance as the fact that Saddler occasionally smoked the popular brand of cigarette found at the scene. On the other hand, Mrs. Overfield’s testimony was directly controverted by Saddler’s alibi defense, presented by several witnesses. In Hannon the People’s case likewise rested solely on the identification testimony of one witness, that testimony was weakened on cross-examination, and the defendant presented an alibi. Hannon is therefore directly in point, and I deem it to be persuasive on the issue of prejudice.
Ill
The prejudicial effect of giving CALJIC No. 2.62 here is dramatically emphasized by the procedural history of the case. Saddler was previously tried on the same charge; after deliberating all of one day and part of the next the jury reported itself hopelessly deadlocked at seven votes to five, and a mistrial was declared. The event is significant because CALJIC No. 2.62 was not given at the first trial, while the remainder of the instructions were essentially identical. The majority acknowledge this difference, but fail to justify their cavalier dismissal of its relevance, {ante, p. 684.) As will appear, the evidence at the second trial was substantially the same as — if not less than — the evidence at the first. Accordingly, the most likely explanation for the verdict of guilt was the giving of CALJIC No. 2.62.
My examination of the record reveals that the People and Saddler presented essentially the same witnesses at both trials. There were, moreover, two witnesses for the People at the first trial who did not testify at the second. Another distinction between the trials was that 22 exhibits for the People were admitted into evidence in the first trial, but only 9 in the second. Notably, a pocket knife belonging to Saddler was admitted in the first trial but excluded in the second: in the absence of a showing that the knife offered closely resembled the knife used in the robbeiy, the *689court ruled that its potential for prejudicial effect on the jury outweighted any probative value. (Evid. Code, § 352.)
In view of these differences in witnesses and exhibits between the two trials, a reasonable inference may be drawn that the quantum of evidence at the second trial was no greater, and apparently less, than that at the first trial. Moreover, the three and one-half months intervening between the trials render it probable that the witnesses’ recollection would have been less reliable at the second trial. For this reason their testimony would tend to have been less credible, as evidenced by Mrs. Overfield’s repeated admissions of a failing memory by the time of her second appearance.
The foregoing unusual procedural history therefore dictates a finding of prejudicial error for reasons wholly apart from the fact this was a close case: it is reasonably probable from the record that the erroneous instruction was the decisive factor in the second verdict. Thus there is present not one, but two, indicia of the “reasonable probability” which requires reversal under the Watson test.
IV
Finally, an error in the majority opinion is its assertion that CALJIC No. 17.31 is relevant in determining whether prejudice resulted from giving an insruction unsupported by any evidence. {Ante, pp. 683-684.) The majority correctly states that CALJIC No. 17.31 would not itself save an otherwise improper instruction; but I disagree that it may be given any weight whatever in this determination.
The majority quote the words of CALJIC No. 17.31 but fail to grasp their true significance. In almost every trial the contending parties introduce different versions of the facts; indeed, a single party may also do so in support of alternative theories of the complaint or the defense. The jury will ultimately believe one of those versions and reject the others; yet because the court cannot ascertain what the jury may find, it must instruct on the legal consequences of every factual determination the record would support — i.e., on “the law applicable to any state of facts that may be logically deducible from any of the evidence, regardless of the weight or credibility of that evidence, which are within the sole province of the jury. . . .” (People v. Spraic (1927) 87 Cal.App. 724, 732-733 [262 P. 795]; accord, People v. Palmer (1946) 76 Cal.App.2d 679, 686 [173 P.2d 680].) The sole and simple purpose of CALJIC No. 17.31 is *690to caution the jury that it should disregard any instruction relating to a state of facts that it does not find to be true.1
This purpose is plainly written, moreover, on the face of the instruction itself. It provides in relevant part that “You have been instructed as to all the rules of law that may be necessary for you to reach a verdict. Whether some of the instructions will apply will depend upon your determination of the facts. You will disregard any instruction which applies to a state of facts which you determine does not exist.” (Italics added.) The emphasized language clearly focuses the jury’s attention on purported facts that it determines from the evidence are not true.
None of this, however, is pertinent to the issue before us. As the majority acknowledge, it is for the court and not the jury to determine whether there is any evidence to support a requested instruction. Here the error occurred when the court gave an instruction devoid of such support — i.e., an instruction on a state of facts that the jury could not have found to be true. In assessing the probability of prejudice from that error, it is obviously irrelevant that the jury was also told to disregard any instruction on a state of facts it could have found true but did not. CALJIC No. 2.62 is simply not such an instruction.
Not only does the majority’s reliance on CALJIC No. 17.31 contribute to their erroneous affirmance of this tainted conviction, it will return to haunt us in many cases yet to come. The instruction, as noted, is given as a matter of routine in virtually every jury trial. Under the majority’s view, however, it will inevitably be invoked by the People to minimize the prejudicial effect of any instruction — not simply No. 2.62 — given without factual support. In such a context the Watson test is in danger of becoming a casualty.
I would reverse the judgment.
The modest intent of this instruction is confirmed by its placement in the CALJIC collection, i.e., among the general “cautionary” instructions given in virtually every case. (See, e.g., Nos. 17.30 [jury not to take cue from the judge], 17.40 [jury’s duty to deliberate], 17.41 [how jurors should approach their task].)