dissenting:
In this felony murder ease, everyone agrees with the proposition, drawn from United States v. Heinlein, 160 U.S.App.D.C. 157, 168, 490 F.2d 725, 736 (1973), that there must be enough of a causal connection between the felony and the homicide that “the killing can be said to have occurred as a part of the perpetration of the crime. ” (Emphasis added.) In the case of a robbery, “perpetration” includes flight from the scene carrying the stolen goods. See Head v. United States, 451 A.2d 615, 625 (D.C.1982). We therefore confront this question: how much “a part of the perpetration” of the robbery — including flight from the robbery — must the homicide be?
The answer is straightforward: the felony must be a legal cause of the homicide. See People v. Aaron, 409 Mich. 672, 299 N.W.2d 304, 312-16 (1980); W. LaFave & A. Scott, 2 Substantive Criminal Law § 7.5(d), at 213 (1986); II Model Penal Code and Commentaries § 210.2, at 29-43 (1980); see generally Marshall v. United States, 623 A.2d 551, 559-61 (D.C.1992) (Ferren, J., dissenting in part and concurring in the result only). This means the death must have been a “natural and probable consequence” of the felony, i.e., “proximately caused” by acts within the scope of the felony. See id.
Legal causation is not necessarily limited to causation by one actor or event; there can be multiple legal causes. I am therefore persuaded that the felony murder rule will be unfairly harsh and unjust — especially in an attenuated case, such as this, where flight from the scene is deemed part of the crime— unless a jury reasonably can find, and is *441instructed it must find, that the felony at least was a “substantial factor” (among others) in the chain of causation from felony to death; a minor factor cannot be a legal cause. See Baylor v. United States, 407 A.2d 664, 670 (D.C.1979) (medical malpractice not intervening cause of victim’s death in homicide case where defendant’s action — a blow to the side rupturing spleen — “contributed substantially to death”); cf. Dalo v. Kivitz, 596 A.2d 35, 41-42 (D.C.1991) (in legal malpractice action, liability predicated on finding injury proximately caused by tortious act that played “central role,” or was “substantial factor,” in injury) (quoting District of Columbia v. Freeman, 477 A.2d 713, 716 & n. 9 (D.C.1984)); Lacy v. District of Columbia, 424 A.2d 317, 319 (D.C.1980) (adopting “substantial factor” test for proximate cause in civil action for assault and battery). Although legal causation analysis is not necessarily the same under criminal and tort law, the foregoing case law recognizes that, under either regime, every cause resulting in liability, whether criminal or civil, must contribute substantially to the result.
The trial court’s instructions initially may have satisfied the “substantial factor” causation test by telling the jury that, for conviction, the robbery (including transportation of the stolen goods) must have been “ongoing” at the time of the murder and that the speeding (i.e., the “specific actions leading to the fatal injuries”) must have comprised “part of the defendant’s efforts to successfully complete the robbery.” Ante at 432. This first part of the instructions in effect told the jury that, to find guilt, it must find that the speeding was part of a conscious “effort[ ]” to complete the robbery; this language arguably expressed a “substantial factor” requirement.
But then the court pulled back. It added that the speeding (“specific actions leading to the fatal injuries”) must have been “motivated at least in -part by the defendant’s desire to avoid apprehension with the stolen property.” (Emphasis added.) Appellants contend, and I agree, that this follow-up language diluted whatever “conscious effort” implication there was in the preceding part of the instruction so that the jury, in effect, was told it could find causation even if the robbery comprised only a minor — not substantial — “part” of the reasons for the speeding that directly led to the death.
In its brief, and under direct questioning at oral argument on appeal, the government did not shy away from appellants’ reading of the instruction. First, the government acknowledged it did not believe there is a “substantial factor” component to the legal cause of felony murder. Second, the government candidly conceded that the trial judge himself interpreted the instructions as appellant does, and that for this reason the prosecutor was permitted to make a “minor factor” argument to the jury, as follows: “it doesn’t matter, as long as you find that in any part, in any way, they were motivated to avoid police apprehension because sitting right there in that vehicle were the proceeds of a robbery that they committed six minutes before.” (Emphasis added.)
If the substantial factor test applies — as I believe it does (and as my colleagues intimate they would approve if not prefer, see ante at 434)1 — then the trial court erred in failing to make that test clear to the jury. I cannot agree with the majority that the short, “five to ten minute interval between the robbery assault and appellants’ sighting by the police” necessarily made the other two identified causal factors — appellants’ “possession of a stolen vehicle and their having just ignored two stop signs,” ante at 432 — so insignificant in the decision to flee from the police that the jury, even as instructed, inevitably had to find that the robbery was a substantial factor in the child’s death, not a lesser one. Certainly the jury on this record could have found the required substantial factor, but the factual record and the instruction — as the government concedes — allowed less.
*442In applying the trial court’s dilution language (“at least in part”) to the facts here, see ante at 434-435, the majority implicitly accepts a “substantial factor” requirement and finds “no error” based on “harmless error” reasoning. Especially, however, in a multiple legal cause situation such as this case appears to be, it is not for this court to cure the instructional lapse — called to the judge’s attention at trial — by ruling on the basis of how we, as appellate judges, think the jury must have performed its task. Because the jury, if properly instructed, could have ruled either way on this record, not necessarily as my colleagues believe it did, the instructional error was not harmless. See White v. United States, 613 A.2d 869, 877 (D.C.1992) (en banc) (instructional error subject to “harmless error” analysis).
I would reverse and remand for a new trial with proper felony murder instructions expressly incorporating the “substantial factor” test. Respectfully, therefore, I must dissent.
. The “substantial factor” test — regularly employed by juries in this jurisdiction, especially in tort cases — does not require the factfinder to “quantify the respective significance — to 'evaluate the strength’ — of different causal factors" or to "assign a percentage causal weight approaching if not exceeding fifty percent.” See ante at 434. The majority’s worries in this regard are unfounded.