(dissenting, with whom Liacos, C.J., joins). To obtain a conviction of murder, the Commonwealth was required to prove beyond a reasonable doubt that the defendant’s killing of his wife was not voluntary manslaughter. This is so because, as the court recognizes, the evidence warranted a finding of voluntary manslaughter; that is, there was “evidence of provocation deemed adequate in law to cause the accused to lose his self-control in the heat of passion, and . . . the killing followed the provocation before sufficient time had elapsed for the accused’s temper to cool.” Ante at 237, quoting Commonwealth v. Schnopps, 383 Mass. 178, 180 (1981), S.C., 390 Mass. 722 (1984).
The jury heard the testimony of several witnesses. The testimony of five of those witnesses was erroneously admitted over the defendant’s objection. Those witnesses testified that in the weeks before her death the victim had told them that the defendant had threatened and beaten her. Following the judge’s final charge to the jury, in which for the first time he *247instructed the jurors to disregard that testimony, the jury returned a verdict of guilty of murder. Presumably, the jury were satisfied beyond a reasonable doubt that the killing was not the product of a sudden loss of self-control in the heat of passion provoked by the defendant’s confirmation of his earlier suspicion of his wife’s (the victim’s) infidelity. The critical question is whether the erroneous admission of that testimony might have influenced the jury’s decision. Unless the court can rightly say that no such possibility exists, we cannot be confident that the defendant had a fair trial, and the conviction of murder must be reversed.
In order to reach the right answer to that critical question, it is necessary to focus on the standard, which this court has previously established, by which the effect of the erroneous admission of the five witnesses’ hearsay testimony must be measured. “Because the error here was not a constitutional one, our inquiry is whether the improper admission of the evidence constituted prejudicial error. See Commonwealth v. Schulze, 389 Mass. 735, 741 (1983). Under this standard we consider whether ‘the error possibly weakened [the defendant’s] case in some significant way so as to require a new trial.’ Id” Commonwealth v. Daggett, 416 Mass. 347, 352 n.5 (1993). The ultimate question is whether the court can fairly say “that the jury could not have been influenced by [the erroneously admitted evidence].” Commonwealth v. Stone, 321 Mass. 471, 474 (1947). See Commonwealth v. Welcome, 348 Mass. 68, 70 (1964) (“It is very possible that the finding made by the judge was not at all affected by what he [improperly] heard of the defendant’s previous difficulties. We can only speculate upon the effect of that evidence. We are not in a position to say that it had none and such doubts as we entertain can only be resolved in favor of the defendant. Commonwealth v. Stone, 321 Mass. 471”). It is inappropriate, therefore, for the court to consider “the probable effect on the jury of the improper evidence,” ante at 241, or to consider whether “[i]t seems likely . . . [that the jury] would have relied primarily on the abundant admissible evidence of threats and marital discord,” ante at 241. Instead, the court’s concentration should be on the question whether the jury could have been — might have been — influenced by the inadmissible hearsay evidence of the defendant’s assaultive behavior toward the victim over a several week period before *248the killing. Did the error possibly weaken the defendant’s case in some significant way? I think it did. Is the court in a position to say that the inadmissible hearsay evidence had no effect on the verdict? I think not.
As the court states, ante at 239, “the Commonwealth’s case contained compelling admissible evidence of considerable marital discord” evidenced, in part at least, by loud arguments between the victim and the defendant. Such discord says little, if anything, about whether the killing was the result of the defendant’s sudden loss of control provoked by his confirmation of the victim’s unfaithfulness. Furthermore, did the jury believe the testimony that the defendant had expressed an intent to kill his wife and himself? Did they believe the testimony that (1) the defendant had driven his automobile into the victim’s automobile with enough force to break a taillight, and (2) had “attempted” to hit the victim? If they did believe it, did they attach importance to it, or did they simply focus on the inadmissible hearsay? Of course, it is possible that the jury believed that testimony and without reference to the inadmissible hearsay concluded that the killing was not manslaughter but was murder. It is also possible, however, that, because of disbelief of at least some of that testimony or because of doubt about what it proved relative to whether the killing was manslaughter or murder, the jury concluded beyond a reasonable doubt that it was murder only by taking into account the testimony of the five witnesses relative to threats and beatings over an extended period of time. That testimony, inadmissible as it was, strongly tended to show that the killing was other than a sudden impulsive reaction to a recent provocation. Despite the judge’s instructions, the jury may have relied on it to reach their verdict. Not only the court, but I, too, “harbor some doubt of the complete efficacy of a curative instruction striking portions of testimony heard by the jury three or four days earlier.” Ante at 241. Whether the verdict in this case was influenced by inadmissible hearsay evidence is a matter of doubt. That doubt should be resolved in the defendant’s favor. “Whether guilty or innocent, he was entitled to be tried in accordance with law.” Commonwealth v. Stone, supra at 474.