I respectfully dissent. In my view, defendant’s conviction for the attempted murder of Renell T., Jr., is unsupported by substantial evidence.
Defendant fired a single bullet into a moving car, narrowly missing the driver and her infant son, after quarreling with the driver and the driver’s boyfriend. There was ample evidence to support the jury’s finding defendant was trying to kill the driver. The evidence was ample also that he acted recklessly, or even with conscious disregard for life, as to the baby. The evidence was insufficient, however, to permit the jury to infer beyond a reasonable doubt that defendant intended to kill the baby, with whom, as far as the evidence showed, defendant had no quarrel at all. The majority struggles to articulate grounds for upholding the second attempted murder conviction. In the course of that struggle, the majority loses sight of the crucial difference between implied malice, or conscious disregard for life, and express malice, which is the specific intent to kill a person.
The majority reasons that because both Karen and her son were in the line of defendant’s fire, and therefore defendant’s single shot could have killed either, the jury could infer he intended to kill both. That defendant had displayed overt hostility toward Karen and none at all toward the baby is, in the majority’s view, immaterial because conviction of attempted murder does not require proof of a motive for killing. (Maj. opn., ante, at pp. 743-744.) The majority is of course correct that intent may ordinarily be inferred from action. But to support the inference that defendant intended to kill the infant, the majority points to no aspect of defendant’s action other than that he placed the infant’s life in danger by shooting in his direction. The majority thus permits knowing endangerment, which establishes at most implied malice, to serve, by itself, as proof beyond a reasonable doubt of intent to kill. This result is contrary to fundamental concepts of California homicide law recognized in the majority opinion (id., at pp. 739-740) and discussed *750further below (post, at pp. 750-752), in particular the distinction between implied and express malice and the requirement that the latter be proven as an element of attempted murder.
Facts
Karen A. drove her boyfriend, Renell T., Sr. (Renell), to a friend’s house. Their three-month-old baby, Renell T, Jr., sat in a rear-facing infant car seat in the backseat directly behind Karen, and Renell sat in the front passenger seat. Karen parked along the curb on the street in front of the house, and Renell got out of the car. As Karen waited in the car to make sure Renell’s friend was home, she saw defendant, a former friend, approaching from behind. The last time Karen had spoken to defendant was in a telephone conversation approximately eight to nine months before, during which defendant had told her the next time he saw her he would “slap the shit out of [her].”
Defendant walked to the front passenger window of Karen’s car, looked inside and said, “Don’t I know you, bitch?” Turning around from the walkway leading to the house, Renell said, “Well, you don’t know me.” Renell walked back to the car, and as he and defendant confronted each other, defendant lifted his shirt and displayed a handgun in his waistband. Renell said, “It is cool,” and backed away from defendant. A group of men from the street comer began approaching the car, and as Renell entered the vehicle through the front passenger door, defendant and the other men began striking him.
As soon as Renell was securely inside the car, Karen pulled away from the curb. After driving about one car length, Karen looked into her rearview mirror and saw defendant standing directly behind the car holding a gun. She heard a single gunshot; although she did not see defendant pull the trigger, he was the only person she had seen with a gun. As soon as she reached a place of safety, she stopped to check the baby for injuries. The rear window had shattered; the baby was screaming, and his face was “full of glass pieces.” Later, the police determined the bullet had entered through the rear window, passed through the driver’s headrest and lodged in the driver’s door.
Discussion
The only question before the court is the sufficiency of evidence to prove defendant attempted to kill Renell T., Jr., Karen and Renell’s three-month-old infant. (Defendant does not challenge his convictions for attempted murder of *751Karen (Pen. Code, §§ 187, 664),1 for child endangerment (§ 273a, subd. (a)) and assault with a firearm (§ 245, subd. (a)(2)) on the baby, or for shooting at an occupied vehicle (§ 246).) The Court of Appeal held the evidence sufficient on a “kill zone,” or “concurrent intent,” theory. (See People v. Bland (2002) 28 Cal.4th 313, 331 [121 Cal.Rptr.2d 546, 48 P.3d 1107] {Bland).) We granted defendant’s petition for review to consider that question.
“Attempted murder requires the specific intent to kill and the commission of a direct but ineffectual act toward accomplishing the intended killing.” (People v. Lee (2003) 31 Cal.4th 613, 623 [3 Cal.Rptr.3d 402, 74 P.3d 176].) “The mental state required for attempted murder has long differed from that required for murder itself.” (Bland, supra, 28 Cal.4th at p. 327.) For murder, malice may be express or implied. “Malice is express when the killer harbors a deliberate intent to unlawfully take away a human life. Malice is implied when the killer lacks an intent to kill but acts with conscious disregard for life, knowing such conduct endangers the life of another.” (People v. Lasko (2000) 23 Cal.4th 101, 104 [96 Cal.Rptr.2d 441, 999 P.2d 666]; see § 188.) To be guilty of attempted murder, the defendant must harbor express malice; implied malice will not suffice. (People v. Swain (1996) 12 Cal.4th 593, 604-605 [49 Cal.Rptr.2d 390, 909 P.2d 994].) Express malice, or intent to kill, requires more than knowingly placing the victim’s life in danger: it requires at the least that the assailant either “ 1 “desire the result,” ’ ” i.e., death, or “ ‘ “know, to a substantial certainty, that the result will occur.” ’ ” (People v. Davenport (1985) 41 Cal.3d 247, 262 [221 Cal.Rptr. 794, 710 P.2d 861].)
We also “distinguish between a completed murder and attempted murder regarding transferred intent.” {Bland, supra, 28 Cal.4th at p. 328.) “In its classic form, the doctrine of transferred intent applies when the defendant intends to kill one person but mistakenly kills another. The intent to kill the intended target is deemed to transfer to the unintended victim so that the defendant is guilty of murder.” {Id. at p. 317.) Transferred intent, however, does not apply to attempted murder: “To be guilty of attempted murder, the defendant must intend to kill the alleged victim, not someone else.” {Id. at p. 328.) Thus the defendant’s mental state “must be judged separately as to each alleged victim.” {Id. at p. 331.) Thus defendant’s specific intent to kill Karen, which was adequately proven by the evidence, does not “transfer” to provide the specific intent to kill her baby; the prosecution was required to prove defendant’s mental state as to the baby individually.
In determining whether the prosecution met its burden, the test is whether a rational jury could have found beyond a reasonable doubt that defendant harbored the requisite specific intent to kill the baby. (People v. Ochoa (1993) *7526 Cal.4th 1199, 1206 [26 Cal.Rptr.2d 23, 864 P.2d 103].) In his briefing, the Attorney General posited two factual theories to support the conviction: (1) that defendant actually targeted and thus intended to kill the baby, and (2) that defendant had the concurrent intent to kill the baby’s mother, his primary target, and the baby, a nontargeted person.
Actual Targeting
The Attorney General contends the evidence is sufficient to support the conviction for the attempted murder of the baby on a theory of actual targeting because defendant knew the baby was seated directly behind Karen when he fired a bullet into the vehicle. The majority agrees.
The specific intent to kill needed for attempted murder may, in many cases, be inferred from the defendant’s acts and the circumstances surrounding the alleged attack. (See People v. Lee (1987) 43 Cal.3d 666, 679 [238 Cal.Rptr. 406, 738 P.2d 752].) Here, the evidence shows only that defendant knew the baby was seated directly behind Karen when he fired one bullet into the car; neither circumstantial nor direct evidence shows that defendant—in contrast to his animosity toward Karen—bore any desire to harm the baby. Certainly in shooting a single bullet in the direction of two people without justification or provocation, defendant acted with, at least, implied malice toward both of them; a jury could also reasonably conclude that he acted with intent to kill one of the two. But this evidence is insufficient to permit a reasonable jury to conclude beyond a reasonable doubt that defendant acted with intent to kill the baby as well as Karen because, under the circumstances of this case, defendant’s single shot provides an inadequate basis for an inference that he actually targeted the baby.
To be sure, motive is not a legal element of the crime of attempted murder; in many cases, the evidence of a defendant’s actions will be sufficient to show express malice without proof the defendant had a motive for killing the victim. In many circumstances, perhaps most, the act of shooting a loaded firearm in a person’s direction from short range would, in itself, support an inference of intent to kill, even were the assailant’s reasons for shooting unknown. In this, I agree with the majority. (Maj. opn., ante, at p. 741.) But the majority draws too broad an inference from defendant’s act. When, as here, the defendant has shot only once in the direction of two people, only one of whom he had any reason to attack, the act of shooting gives rise, at most, to an inference the defendant intended to kill the person he had a motive for killing.
The majority’s argument reduces to this claim: from defendant’s knowledge of the baby’s location and his shooting in the direction of both Karen *753and the baby, the jury could reasonably infer defendant intended to kill them both. In so reasoning, the majority stretches the meaning of intent to kill so far as to make it indistinguishable from the conscious disregard for life that constitutes implied malice. If conscious disregard of a risk of death—shown here by defendant’s act of firing in the baby’s direction—suffices to support an inference of intent to kill, no difference is discemable between the two types of malice.
As we held in Bland, supra, 28 Cal.4th at pages 328-329, intent to kill does not “transfer” from victim to victim for purposes of attempted murder; to prove defendant attempted to kill the baby, the prosecution had to prove he intended to kill the baby. Intent to kill Karen is insufficient for this charge. Defendant’s evident lack of motive for killing the baby, in contrast to his marked animosity toward Karen, is critical not because motive is an element of the crime, but because his lack of motive as to the baby points to the only rational answer to the question, whom did defendant in firing his single shot actually target and intend to kill?
The majority reasons that defendant’s act of firing in the child’s direction, thereby placing the child’s life at serious risk, shows defendant harbored animus toward the child, from which the jury could find he desired the child’s death. (Maj. opn., ante, at pp. 743, 748.) But if, as the majority argues, the act of placing a person’s life at risk, in itself, shows the intent to kill that person, nothing differentiates the two types of malice, a conclusion contrary to the fundamental California law of homicide embodied in section 188. The suggestion that endangerment also shows animus adds nothing to the analysis, for the inference of intent to kill is still being drawn from the act of endangerment.
In short, the majority would sanction an inference of intent to kill solely from an act knowingly endangering the victim. But, as noted earlier, intent to kill requires that the assailant either “ ‘ “desire” ’ ” the victim’s death or “ 1 “know, to a substantial certainty, that the result will occur.” ’ ” (People v. Davenport, supra, 41 Cal.3d at p. 262.) Here there was no evidence, other than his act endangering the baby, that defendant desired the baby dead. Nor was defendant’s means of attack so powerful that it made substantially certain the death of both Karen and her infant son. If, as the majority would have it, proof of knowing endangerment suffices to support a finding of intent to kill—without any evidence the assailant either wanted to kill the victim or acted so as to make the victim’s death substantially certain—then the distinction between implied and express malice has been effectively obliterated.
The majority’s reasoning potentially opens the door to an unlimited number of attempted murder convictions based on a single act intended to kill *754a single person—without proof the defendant used means intended to create a kill zone around the target. How, one must ask, is the number of attempted murder convictions arising from a single shot limited under the majority’s reasoning? If assailant D shoots a handgun once at close range in the direction of a targeted victim, VI, who is standing in a close crowd of strangers, V2 through V10, could a jury find D intended to kill all 10 victims, even in the absence of evidence D had any reason to want V2 through V10 dead? To the suggestion D bore no animus against anyone but VI, the majority would presumably respond that “his very act of discharging a firearm into the [crowd] from close range and narrowly missing [V2 through V10] could itself support such an inference.” (Maj. opn., ante, at p. 744.) The majority’s reasoning cannot be correct, for it results in the absurd conclusion that an assailant has tried to murder everyone his act endangers.
People v. Chinchilla (1997) 52 Cal.App.4th 683 [60 Cal.Rptr.2d 761], upon which the majority relies, is distinguishable. The appellate court there affirmed two convictions for attempted murder arising out of the defendant’s having fired one bullet at two pursuing police officers, one of whom was crouched in front of the other. (Id. at pp. 687, 690-691.) Mr. Chinchilla had equal reason to try to kill each of the pursuing officers. In the present case, in contrast, defendant had threatened Karen with violence but, as far as the evidence showed, had no reason to attack her infant child. Even if a jury could reasonably find, from his act of shooting once, that Mr. Chinchilla intended to kill both officers, the evidence of defendant’s single shot in this case did not justify a parallel inference.2
Despite claiming evidence of motive is unnecessary in these circumstances, the majority repeatedly suggests defendant might have wanted to kill Karen and Renell’s baby because, according to him, Karen was his former girlfriend (rather than just a friend as she testified). (Maj. opn., ante, at pp. 742, 744, 747, 748.) The jury, however, obviously did not believe defendant’s version of events; if they had, they would not have convicted him at all. Even defendant’s testimony, moreover, fails to support the inference the majority puts forward; defendant testified his argument with Karen the day before the confrontation with Renell arose because Karen was reluctant to give him a ride in her car, as she sometimes did. His testimony indicates he was angry at Karen for her reluctance and for the language she used toward him in the ensuing argument. Nothing suggests he was angry with or about the baby she had recently had with Renell.
*755The inference the majority would draw as to why defendant might wish to harm the infant is thus entirely speculative. It could be true (if one disbelieves Karen and believes defendant as to their prior relationship), but no evidence to that effect appears in the record. Speculation does not constitute substantial evidence. (People v. Lewis (2001) 26 Cal.4th 334, 369 [110 Cal.Rptr.2d 272, 28 P.3d 34].)
Finally, even indulging this speculation and assuming defendant wanted to kill Renell T., Jr., because he was another man’s son, in order to find defendant intended to kill both Karen and the baby one would also have to infer he intended somehow to hit and kill both with his single shot. Though the majority does not fully articulate its factual theory, its repeated invocation of the “large-caliber” bullet used here (maj. opn., ante, at pp. 743, 746, 747) is presumably intended to suggest defendant intended to shoot Karen through her baby. Of course, a single bullet can hit and even kill two people, but here (where the baby presented a notably small target and the bullet would, in addition, have had to traverse the infant car seat and the driver’s seat without deflection) there was no evidence defendant was capable, or believed himself capable, of such a feat of marksmanship. Again, an appellate court’s speculation cannot substitute for evidence at trial.
Concurrent Intent to Kill
Although the majority purports not to rely on this point (maj. opn., ante, at p. 746), the Attorney General alternatively contends the evidence is sufficient to support defendant’s conviction for the attempted murder of the baby on a concurrent intent theory because defendant intentionally created a “kill zone,” from which the jury could reasonably infer he concurrently intended to kill both Karen, his intended target, and her baby.
We have explained that multiple attempted murder convictions may be sustained on a “kill zone,” or “concurrent intent,” theory when the evidence shows the defendant used lethal force of a type and extent calculated to kill everyone in an area, including but not limited to the victim shown to be the defendant’s primary target, as a means of accomplishing the killing of the primary target. Under these circumstances, the fact finder could rationally infer the defendant intended to kill not only his or her primary target, but also concurrently intended to kill all those in the zone of fatal harm. (Bland, supra, 28 Cal.4th at pp. 329-330.) A kill zone, or concurrent intent, analysis, therefore, focuses on (1) whether the fact finder can rationally infer from the type and extent of force employed in the defendant’s attack on the primary *756target that the defendant intentionally created a zone of fatal harm, and (2) whether the nontargeted alleged attempted murder victim inhabited that zone of harm. (Harrison v. State (2004) 382 Md. 477, 495 [855 A.2d 1220, 1231].)
In Bland, we illustrated the operation of the kill zone, or concurrent intent, theory of attempted murder with several examples: “ ‘[A]n assailant who places a bomb on a commercial airplane intending to harm a primary target on board ensures by this method of attack that all passengers will be killed. Similarly, consider a defendant who intends to kill A and, in order to ensure A’s death, drives by a group consisting of A, B, and C, and attacks the group with automatic weapon fire or an explosive device devastating enough to kill everyone in the group. The defendant has intentionally created a “kill zone” to ensure the death of his primary victim, and the trier of fact may reasonably infer from the method employed an intent to kill others concurrent with the intent to kill the primary victim.’ ” (Bland, supra, 28 Cal.4th at pp. 329-330, quoting Ford v. State (1992) 330 Md. 682 [625 A.2d 984].) In Bland itself, we explained, the evidence strongly supported an inference of concurrent intent to kill: the defendant and a fellow gang member intentionally created a zone of fatal harm when they fired a “flurry of bullets” into a fleeing car, justifying convictions for attempted murder of the passengers. {Bland, supra, at p. 331.)
Here, defendant did not fire multiple shots at Karen; he fired one bullet in Karen’s direction at a distance of about a car’s length from the rear of the moving vehicle. By firing a single shot, defendant did not use a type or degree of force reasonably calculated to kill everyone in the vehicle. Defendant’s method of attack was not comparable to the “kill zone” examples and decisions we cited in Bland: detonating a bomb on a commercial airplane (Bland, supra, 28 Cal.4th at pp. 329-330), using an explosive device or automatic weapon fire against a group of people (id. at p. 330), spraying wall-piercing bullets at occupied houses (ibid., citing People v. Vang (2001) 87 Cal.App.4th 554, 563-565 [104 Cal.Rptr.2d 704]) or mailing poisoned candy to a household (Bland, supra, at p. 331, citing People v. Gaither (1959) 173 Cal.App.2d 662, 666-667 [343 P.2d 799]). Nor was defendant’s method of attack comparable to the firing of multiple gunshots into a fleeing car by the defendant in Bland itself. {Bland, supra, at pp. 330-331.) That the bullet came close to hitting Karen and her baby does not, without more, establish that by firing a single shot in the direction of Karen, his intended target, defendant intentionally created a zone of fatal harm around Karen such that he may be deemed to have intended to ensure her death by killing the baby as well.
*757The Attorney General, nevertheless, insists defendant intentionally created a zone of fatal harm by firing one bullet into the car because “[t]he baby and Karen were positioned in the car in such a way that [defendant], firing from the rear, could not have killed Karen without shooting through the baby first.” This argument finds no support in the record; the evidence shows neither that the bullet necessarily had to pass through the baby to kill Karen,3 nor that the ammunition defendant used was of a kind likely to kill two persons in the manner the Attorney General suggests.
Although the majority disavows any reliance on the kill zone theory, the import of the majority opinion is that an act aimed at killing one person creates a kill zone that includes everyone who could have been killed by the act, regardless of whether the assailant used means actually calculated to kill everyone in the target’s vicinity. If a single shot with a handgun constitutes not only an attempt on the life of a person at whom the jury could find the shot was actually aimed, but also an attempt on the life of anyone else nearby, the careful analysis in Bland was unnecessary: the limited concurrent intent theory of Bland would be obviated, subsumed in a much broader endangerment theory.
The majority’s expansion of attempted murder liability to cover mere endangerment is unnecessary in order to ensure assailants are appropriately punished for acts that place victims’ lives in danger. Unjustified shooting in other people’s direction, even when not intended to kill them, will ordinarily subject the shooter to liability for assault or a related offense. Here, as noted, defendant was convicted of assault with a firearm, child endangerment and shooting at an occupied vehicle in addition to the attempted murder counts. Sentencing on these felony convictions was stayed under section 654, but if defendant were not sentenced for attempted murder of the infant he could be sentenced on one or more of these charges, which carry substantial punishments. (See § 245, subd. (a)(2) [sentence of two, three or four years in prison]; § 246 [three, five or seven years]; § 273a, subd. (a) [two, four or six years]; § 12022.5, subd. (a) [firearm use enhancement of three, four or 10 years].)
*758Conclusion
Because the record contains no substantial evidence from which a jury rationally could infer either that defendant actually targeted Karen’s child, as well as Karen, when he shot once at the car, or that he employed a means of attack calculated to kill everyone surrounding Karen, the evidence is insufficient to show defendant had the specific intent to kill the infant, as required to sustain a conviction for attempted murder. Defendant’s conviction on that count should therefore be reversed.
Moreno, J., concurred.
All further statutory references are to the Penal Code.
For the proposition that the intent to kill both victims could be inferred from defendant’s single shot in their direction, the majority also cites People v. Lee, supra, 43 Cal.3d at page 679, People v. Villegas (2001) 92 Cal.App.4th 1217, 1224-1225 [113 Cal.Rptr.2d 1], and People v. Lashley (1991) 1 Cal.App.4th 938, 946 [2 Cal.Rptr.2d 629], None of those decisions, however, addresses whether or under what circumstances a person may be held liable for two or more counts of attempted murder for the act of firing a single shot.
Indeed, photographic evidence showed that the top of the rear-facing infant car seat in which the baby sat reached just to the bottom of the driver’s headrest, a few inches below the point where the bullet hit that headrest. Unless the head of the three-month-old baby extended some inches above his car seat, therefore, the evidence left it entirely possible for a gunman to shoot Karen in the head without hitting the baby. While the evidence thus tends to show defendant recklessly placed the baby at grave risk by firing in his direction, it does not show he intended to shoot the baby as a means of shooting Karen. Implied malice was amply proven; express malice was not.