The defendant was convicted after a jury trial on an indictment charging armed robbery. He claims error in the judge’s use of two examples in his charge to explain reasonable doubt and in his failure to strike certain testimony of a police officer alluding to the defendant’s arrest “on a warrant for a previous offense.”
There was evidence before the jury that on November 17, 1978, at about 8:00 p.m. a robbery occurred at a fast food restaurant in Lowell. The only persons present in the restaurant at the time of the robbery were four young women employees. The police were called immediately after the robber had left. The women gave descriptions of him which, with minor variations, were the same. Two of the women then went to a police station where, after going through several books of photographs, they selected one of the defendant. On the following day a third witness select*209ed the same photograph. The defendant was arrested on December 9, 1978. A new photograph was taken of him and placed in a book containing other photographs. The three women were given the book to examine, and each selected the new photograph as that of the robber. Each made an in-court identification of the defendant at trial. The fourth witness, while unable to identify the defendant, not having observed his face, testified to the description of the robber. Although the closing arguments of counsel were not recorded, it is clear from defense counsel’s cross-examination of the Commonwealth’s witnesses and from the alibi testimony of the witnesses for the defense, that the issue being seriously contested was that of identification.
1. In his charge the judge clearly stated the burden of the Commonwealth to prove every essential element of the crime charged beyond a reasonable doubt. He defined reasonable doubt in the following words:
“And that really means that you have got to be sure in your own mind that that’s what happened — all right? — to what we call a moral certainty. In good conscience, you can make that determination of guilt. And that has to be beyond a reasonable doubt.”
He then went on as follows:
“Now, the words ‘reasonable doubt’ we’ve got to wrestle with a little bit. There are certain things that are not subject to absolute, mathematical certainty; and there’s certain things that you can’t be absolutely positive of, which is a little bit different from being sure, and without a reasonable kind of a doubt.
“The milkman normally delivers milk, and in this case it’s normally delivered at six o’clock. There’s a rattle of bottles and he’s taken your note, if you have left one, and responded accordingly. If you put a note out last night for two quarts of milk and a jar of cream . . . and *210there’s a noise at six o’clock the following morning and a rattling of bottles, and then the milk is there when you go out to pick it up just as you left the order, you have the right to presume and find that the milkman left the bottles there. You didn’t see it — all right? — you don’t know for sure; something else may have happened. But, if, under the circumstance, without having heard anything else, there is no question to be determined whether or not your regular milkman delivered these bottles, and you heard nothing to the contrary, and there is a decision to be made on that particular subject, and one of you were of the position that: Hey! I don’t know for sure. I didn’t see it. It may have been a different company and they are using my old company’s bottles. That person would be throwing in a doubt that wasn’t reasonable, in my opinion. They don’t find that as a particular fact, and is hanging on it: it wasn’t proved to me, and I can’t be certain. To a mathematical certainty, it hasn’t been proven to me.”
After a few transitional words, the judge related this example:
“I will run through a progression of things in relation to reasonable doubt. Somewhere along the line, in this progression of things, the doubt that a person has, I think, would become unreasonable. Just when that would take place would be a determination that you will have to make. But, let’s say that you have — that a friend of yours or a family member has been found in the Mystic River dead, and you are talking to another mutual friend, and that mutual friend says to you that that was a suicide; and your opinion is: There’s no way. That’s Charlie — or whoever it happened to be — no way would he commit suicide.
*211“Well, it turns out that your friend is able to point out to you that he was drowned. Then it turns out. So you say: Well, if he was drowned, he could have been pushed off or fallen off a boat or whatever. Then your friend and you learn that a body was seen hurtling from the bridge that fits the description of your friend. You still don’t know. He may have been pushed. It may be a robbery-type of thing. You find that he has his wallet and all his money is intact. It turns out that still he could have been pushed, or still it could have been an accident.
“ It turns out later that there are seventeen people who were traveling on the Mystic River Bridge on this particular occasion and saw a person standing there with nobody else being around him. You are still not convinced. He just wouldn’t do that, and I’m going to hire a detective to find out more. Then it turns out there was a note — a typical suicide note: I am doing this. I apologize to everybody; and all this and that, and it’s signed by your friend, and there’s no question in your mind that is your friend’s signature — all right? You still have some doubts whether it was a suicide or not: He had no reason to do that. You find out later that he was found with the evidence, and that he was just caught. He is a bank employee and there has been an investigation which determined that he embezzled some $200,000 — all right? — and that, you find, is a fact, and you still have doubts about whether or not there was a suicide. Somewhere along the line, you are bringing your — if you still have that doubt — still using your heart and your intuition, which is different from making a judgment on the facts that you have heard — all right? — at that stage of these proceedings, that would be an unreasonable doubt.
“You have to make that determination on your own. I just want to make sure that you know there is a difference — that there is a difference between relying upon the hunches and the facts; and there is a difference be*212tween moral certainty — being sure — and being positive to a mathematical certainty.”
At the conclusion of the charge defense counsel objected to the two portions of the charge set out above, stating that the “milkman” portion of the charge was “confusing” when given in a charge on reasonable doubt. He indicated to the judge that such an analogy is usually given in the context of a charge on drawing inferences. His objection to the portion of the charge concerning “a fellow being thrown off a bridge” was general. The position taken by defense counsel was that the charge was so far off the mark that it could not be corrected and for that reason a mistrial should have been declared.
The judge stated he would give further instructions and denied the defendant’s motion for a mistrial. The further charge as it related to the parts of the main charge set out above were:
“Ladies and gentlemen, any analogies that I had made to you for the purpose of having you understand the law as it exists; any illustrations I could give you, I would have been attempting to keep them as far removed from the facts that you have to wrestle with as possible.
“So, when I talk about a milkman or if I talk about what may have been a suicide, I am attempting to use things that are as far removed from the facts in this case as possible. Now, it turns out, in the course of talking about the possibility of suicide, in making that talk, I did mention the word ‘robbery’; and there’s no way it was intended to be about this case. It just happened to do with the various aspects of the possibility of a suicide or not.
*213“This illustration that I was telling you about the milk bottles, that illustration is really not applicable to this particular case. That is an illustration that can normally be used — that you can infer and find something as a fact, even though you didn’t see it happen. If, in the milkman situation, there is a question of whether or not the milk was delivered and you heard something else — you heard somebody say: ‘I didn’t deliver the milk that morning. It was done by another company who recently purchased our company.’ And, without the customer knowing it, the ownership has changed; if there is a question now of whether or not the milk had been delivered, there can be some doubt or no doubt. It’s up to you to make the determination as to how it was delivered.”
At the outset we note that the objections made to the questioned portions of the charge were sketchy but in our view sufficient to focus the judge’s attention on the charge’s defects. The objections prompted the judge to instruct the jury further despite the prosecutor’s plea: “I hope you’re not going on reasonable doubt again.” The judge apparently was reminded by counsel’s objections that the “milkman” charge is ordinarily given to illustrate reasonable inferences rather than reasonable doubt as he told the jury that “that illustration is really not applicable to this particular case . . . .” Had the judge stopped there, or, better yet, instructed the jury to disregard the “milkman” illustration, it would probably have been removed from the jury’s and our consideration. He did not so instruct the jury but appears to have attempted to convert the example to one illustrating how doubt might be formed in the minds of the jurors. At most, the further charge left both the “milkman” and “bridge” examples intact.
This charge was given nearly two years after the decision in Commonwealth v. Ferreira, 373 Mass. 116 (1977). Jury instructions given after the date of Ferreira will be *214scrutinized even more carefully than those given before. Commonwealth v. Garcia, 379 Mass. 422, 441 (1980). The example of Charlie on the bridge is particularly flawed. It not only “trivialize[d] the awesome duty of the jury to determine whether the defendant’s guilt was proved beyond a reasonable doubt” (Commonwealth v. Ferreira, 373 Mass. at 129); it also related the jury’s decision making function to the highly personalized approach to decision making (or lack thereof) of Charlie’s reluctant friend. See Commonwealth v. Garcia, supra at 440, and cases cited.
At the very least the examples would have confused or mystified the jury as to the meaning of proof beyond a reasonable doubt. They could even have gone so far as to relate the progression of the story of Charlie on the bridge to the build up of the evidence in the case presented against the defendant. At each stage of that story Charlie’s reluctant friend is given an additional fact, and the judge twice informed the jury that at some point in the story the friend’s doubt would become unreasonable. We note that the evidence in the case built progressively from the testimony of one eye witness to that of three, with each identifying the defendant as the robber, and finally to that of a fourth witness who corroborated the description of the first three. The jury could have believed they were being told that any doubt they had of the defendant’s guilt would be unreasonable by the time the third or fourth witness had testified.
We are of the opinion that the charge on reasonable doubt suffered from defects more serious than those found to call for reversal in Ferreira and Garcia, and therefore conclude that the judgment must be reversed. See United States v. Pinkney, 551 F.2d 1241, 1243-1246 (D.C. Cir. 1976); compare Commonwealth v. Hughes, 380 Mass. 596, 598-602 (1980). We do not agree with the Commonwealth’s view that the case against the defendant was so overwhelming that the defects complained of could not have contributed to the guilty verdict. In a case such as this, which is almost wholly dependent on eyeball identification, it cannot be said that the evidence was so overwhelming that a *215serious defect in the very heart of the charge could be harmless. See Commonwealth v. Garcia, supra at 441.
2. We need not address the question whether it was error for the judge not to attempt to cure any prejudicial effect of a police officer’s testimony that “I arrested Roger Pomerleau on a warrant for a previous offense,” as the Commonwealth concedes it was erroneous not to strike this statement, and therefore the matter should not arise at any retrial.
Judgment reversed.
Verdict set aside.