Pinson, Justice. Appellant Sidrick Raymone Melancon, Sr., was convicted of second-degree murder after his ex-girlfriend Sadai Higgenbotham inflicted fatal head trauma on their nine-month-old daughter, Laura Higgenbotham. Although Melancon was not present when Higgenbotham killed Laura, he had earlier told his girlfriend, Gerallyn Long, not to cooperate with a Division of Family and Children Services (DFCS) investigation of Higgenbotham that Long had initiated after seeing a bruise and fingernail mark on Laura’s cheek. One of the State’s theories at trial was that this instruction to Long “caused” Laura’s death. The Court of Appeals approved that theory on appeal, holding that the evidence authorized a jury to find that merely by giving that instruction — almost two months before Laura’s death — Melancon “caused” Laura’s death at Hig- genbotham’s hands, because he “knew” about earlier incidents of harm to Laura and “effectively ended” DFCS’s investigation, thus preventing DFCS from preventing Higgenbotham from killing Laura. For each kind of murder offense set out in our murder statute, the State must prove that the accused “cause[d]” the death of another human being. OCGA § 16-5-1. Although we have explained before that the word “ cause” in that statute means “proximate cause,” our decisions to this point have described this causation in a number of different ways. Those descriptions can be distilled into two components, each of which must be established to prove causation in a murder case: cause in fact, and legal (or “proximate”) cause. Cause in fact refers to the basic requirement that the conduct must have an actual causal relationship to the forbidden result — a relationship that is most often shown through evidence that the result would not have happened “but for” the defendant’s conduct. In other words, cause in fact is the “cause” in the phrase “proximate cause.” Legal cause, on the other hand, puts the “proximate” in “proximate cause.” This inquiry takes as a given that the defendant’s conduct was causally connected to the forbidden result and asks whether the death resulted from the defendant’s conduct in such an unforeseen or attenuated way that the defendant cannot be held accountable for the death. As explained more below, our decisions have described these distinct components of causation as follows: A defendant’s conduct is a cause in fact of a death if his conduct “played a substantial part in bringing about or actually causing” the death (again, typically shown through evidence that the death would not have happened “but for” the defendant’s conduct), or if the defendant’s conduct “materially accelerated the death.” And a defendant’s conduct is a legal cause of a death if the death was “reasonably foreseeable” — that is, a “probable or natural consequence” of the criminal conduct “according to ordinary and usual experience.” Measured against these standards, the evidence in this case was not sufficient to support the particular theory of causation the Court of Appeals addressed. In light of this holding, which we explain in detail below, the judgment of the Court of Appeals is vacated. Because this decision reviews only the theory of causation the Court of Appeals addressed, the case is remanded for the Court of Appeals to apply the framework we set forth below and consider in the first instance whether the evidence of causation in this case was sufficient under a theory other than the one we have rejected here. 1. Background (a) Viewed in the light most favorable to the verdict, the evidence at trial showed the following. At the time of her death, nine-month-old Laura was living in an apartment with Melancon (her father), Higgenbotham (her mother), Long (Melancon’s girlfriend), and Melancon and Long’s three children. When Higgenbotham and Laura first moved in with Melancon and Long, Higgenbotham’s mother (Laura’s grandmother) warned Melancon not to leave Laura alone with Higgenbotham. Melancon exerted significant control over Long and Higgenbotham and physically abused both women. Long often took care of Laura when Higgenbotham was at work. Long believed Higgenbotham was unfit to be a mother and testified at trial about Higgenbotham’s abuse of Laura. She said that Higgenbotham would force feed Laura until she gagged and would muffle Laura’s mouth to silence her crying. On one occasion, Hig- genbotham left for work and left Laura alone in a hot room for thirty minutes. Long once heard Higgenbotham tell Laura to “stop” followed by a slapping sound and Laura crying and, on several occasions, she heard “a loud thump . . . immediately followed by [Laura's] crying.” According to Long, Melancon knew about most of these incidents and “wasn’t surprised” by Higgenbotham’s behavior because “[h]e knew she was capable of it.” In particular, Long testified that Melancon knew about the time Higgenbotham left Laura in a hot room and had seen bruises and scratches on Laura’s body. On June 13, 2017, a little under two months before Laura’s death, Long babysat Laura. During that time, she saw a bruise and a fingernail mark on Laura’s face, and she texted a photo of these injuries to Melancon. Long then asked Melancon for permission to call DFCS about these injuries. He agreed, and Long contacted DFCS and reported the bruise and fingernail mark and also told DFCS how Higgenbotham would put her hand over Laura’s mouth to muffle her crying. But the DFCS investigator noted that Long was “not willing to provide a current address,” did not give DFCS her own name, and she gave an outdated (and wrong) address for Hig- genbotham. The next morning, Melancon changed his mind and told Long not to “file a report” with DFCS. But he agreed with Long that Hig- genbotham’s explanation for the injuries — that Laura fell off the bed and hit her face on a dresser — was “bulls**t.” So Melancon had Long look after Laura for five days after the incident, and told Hig- genbotham by text message that he “took [Laura from her] because [Higgenbotham was] abusing her, not taking care of her,” and that he knew Laura “didn’t fall out of no f**king bed.” For Long’s part, she stopped answering DFCS’s calls. She testified that if Melancon hadn’t told her not to “file[] the report,” she would have done so. The DFCS investigator testified to his unsuccessful investigation after Long’s initial report. On June 14, the day after Long’s call, the investigator tried to contact Higgenbotham in person, by phone, and via social media based on the information Long had provided, but to no avail. The investigator also tried to call and text Long that day. Long answered two of these calls, but she did not provide additional information to help locate Higgenbotham and Laura and told the investigator DFCS could “cancel the report.” DFCS did not “cancel the report” as Long had asked, but it ultimately closed the case because the investigator could not find Higgenbotham or Laura. On July 22, about two weeks before Laura’s death, Hig- genbotham took Laura to a family nurse practitioner for a nine- month wellness check. The nurse practitioner did not note any concerns of abuse or neglect. On July 31, about a week before Laura’s death, Long saw new evidence of injuries to Laura, including “many scratches” that Long recorded in a video, which was entered into evidence at trial. Long did not tell Melancon about these scratches or show him the video. On August 3, four days before Laura’s death, Long noticed that Laura could not stand up in her playpen and that her legs were shaking. Long did not know that, at the time, Laura had a leg fracture. A little after midnight on August 4, Laura screamed in her sleep and Melancon took Laura, who had been sleeping in his and Long’s room, to Higgenbotham’s room. Around noon on August 5, Higgenbotham texted Melancon that Laura had been sleeping since he left for work, would not wake up, and “felt hot.” Between 3:30 and 4:00 p.m., with Melancon’s permission, Higgenbotham left for work, leaving Laura alone in her room (Long was home but in her own bedroom). Around 4:00 p.m., Melancon came home and brought Laura, who was strapped tightly into her car seat, to Long in their bedroom. Melancon then left the apartment. Long noticed that Laura was not moving and had one eye open and one eye shut. Long removed Laura from the car seat and realized she was strapped into the car seat because she was unconscious — when Long lifted Laura, her limbs and head dropped immediately. Long saw that one of Laura’s pupils was dilated and the other was small, she was drooling on one side of her mouth, her breathing was labored, her heartbeat was fast, and she had thick blood in her mouth. Long called Melancon and tried CPR on Laura. Melancon told Long not to call 911 because he was nearby and would drive her to urgent care, which he did. When Laura arrived at the urgent care, she appeared gray, was not breathing, and her head was slumped over in the car seat. Medical personnel tried to resuscitate Laura and called 911. Laura was eventually taken by helicopter to the pediatric intensive care unit of a children’s hospital, where she was declared brain dead on August 7. She died later that day. An autopsy showed that Laura died as a result of injuries caused by intentionally inflicted head trauma and blunt force injuries to her head, neck, abdomen, back, arms, and legs, which were all in a state of healing. Laura had recent brain injuries that reflected massive trauma, including brain bleed and retinal hemorrhage, and she had spinal injuries and leg fractures, including some older fractures that were partially healed. She also had injuries consistent with undergoing violent shaking back and forth. A medical expert testified that the nature and location of Laura’s injuries indicated that they were likely inflicted intentionally and that she had been in distress for several hours before being taken to urgent care. (b) After Melancon’s trial,[1] the jury returned verdicts of guilty for second-degree murder, second-degree child cruelty, and two counts of influencing a witness. Melancon was sentenced to 30 years in prison for the second-degree murder conviction and two consecutive sentences of 10 years’ probation for each count of influencing a witness. The second-degree child cruelty count merged with the second-degree murder count. Melancon appealed, arguing in relevant part that the evidence was not sufficient to establish that he caused Laura’s death. The Court of Appeals rejected that argument. See Melancon v. State, 368 Ga. App. 340, 344 (1) (890 SE2d 113) (2023). The court recounted evidence that authorized the jury to find what it saw as “a natural and continuous sequence leading from Melancon’s interference” to Laura’s death, and also that “had Melancon not interfered to end the DFCS investigation, Higgenbotham’s abuse of Laura would not have continued.” Id. at 345 (1). Although “no direct evidence” showed “what DFCS would have done if Long had cooperated,” the court reasoned that a jury could infer that DFCS would have stopped the abuse based on evidence that Higgenbotham was “actively abusing Laura” while DFCS was taking steps to find her after Long’s first report. Id. at 346 (1). Nor did the court consider Higgenbotham’s later abuse of Laura an “intervening act that broke the causal chain between his interference in the DFCS investigation and Laura’s death.” Id. In the court’s view, “[i]t is reasonably foreseeable that if one interferes to stop a DFCS investigation into child abuse that the abuse will continue to occur,” id., and Melancon also could have reasonably foreseen that Higgenbotham’s continued abuse would “re- sult[] in serious injury to or death of the baby.” Id. at 347 (1). The court concluded with a summary: the evidence was sufficient to find that Melancon committed second-degree murder and cruelty to children in the second degree because “the jury was presented with evidence that Melancon, knowing that Higgenbotham was physically abusive to Laura, took steps to stop DFCS from investigating and acting to prevent further abuse,” and “that Higgenbotham’s continued abuse caused the baby to suffer and ultimately caused her death.” Id.[2] We granted review. 2. Causation Principles For each kind of murder defined by our murder statute, the State must prove (among other things) that the accused “cause[d] the death of another human being.” OCGA § 16-5-1.[3] Although we have generally described this requirement as a single element of “causation” or “proximate cause,” our decisions show that the causation required by our murder statute encompasses two distinct concepts: cause-in-fact, and legal cause. Start with State v. Jackson, 287 Ga. 646, 647 (2) (697 SE2d 757) (2010), a seminal decision addressing “what the term ’causes’ means as used in the felony murder statute.” In that decision, we overruled an earlier decision that had interpreted the term “causes” to mean that the defendant must have “directly caused” the death at issue. See id. at 656 (4), 660 (6) (overruling State v. Crane, 247 Ga. 779 (279 SE2d 695) (1981)). In place of that rule, we held that “causes” requires “proximate cause,” so “the felony murder statute requires only that the defendant’s felonious conduct proximately cause the death of another person.” Id. at 660 (6). Along the way to that holding, we offered varied descriptions of the causation standard. We first noted that “the term ’cause’ is customarily interpreted in almost all legal contexts to mean ‘proximate cause,’” and then recited a definition of that term from Black’s Law Dictionary: “that which, in a natural and continuous sequence, unbroken by any efficient intervening cause, produces injury, and without which the result would not have occurred.” Id. at 648 (2) (quoting Black’s Law Dictionary 1103 (5th ed. 1979)). We then turned to our own decisions. From Skaggs v. State, 278 Ga. 19, 1920 (1) (596 SE2d 159) (2004), we quoted a standard for criminal cases in general: “In a criminal case, proximate cause exists when the accused’s ‘act or omission played a substantial part in bringing about or actually causing the victim’s injury or damage and . . . the injury or damage was either a direct result or a reasonably probable consequence of the act or omission. Jackson, 287 Ga. at 648-649 (2). From James v. State, 250 Ga. 655, 655 (300 SE2d 492) (1983), we quoted a standard for murder cases in general: “Where one inflicts an unlawful injury, such injury is to be accounted as the efficient, proximate cause of death, whenever it shall be made to appear, either that (1) the injury itself constituted the sole proximate cause of the death; or that (2) the injury directly and materially contributed to the happening of a subsequent accruing immediate cause of the death; or that (3) the injury materially accelerated the death, although proximately occasioned by a pre-existing cause.” Jackson, 287 Ga. at 649 (2). From Durden v. State, 250 Ga. 325, 329 (5) (297 SE2d 237) (1982), and Jones v. State, 220 Ga. 899, 902 (3) (142 SE2d 801) (1965), we cobbled together a causation standard that we described as applicable “[i]n the context of this case”: In the context of this case, proximate causation would exist if (to use “the rule” for felony murder that the Court stated a year after deciding Crane) the felony the defendants committed “directly and materially contributed to the happening of a subsequent accruing immediate cause of the death,” Durden, 250 Ga. at 329, 297 SE2d 237, or if (to use language from a case decided 16 years before Crane) “‘the homicide [was] committed within the res ges- tae of the felony’ … and is one of the incidental, probable consequences of the execution of the design to commit the robbery.” Jackson, 287 Ga. at 652 (2). And finally, in distinguishing Cranes “often unhelpful[] direct-indirect dichotomy,” we described “[p]roxi- mate causation” in yet another way, explaining that it “imposes liability for the reasonably foreseeable results of criminal (or, in the civil context, tortious) conduct if there is no sufficient, independent, and unforeseen intervening cause.” Id. at 654 (3). On a first read, it is not clear how exactly one should distill from Jackson’s collection of descriptions a definitive standard or test for determining whether the element of causation has been proven in a murder case.[4] But on a closer review of Jackson and our other causation decisions, two distinct concepts of causation emerge: (1) cause in fact (sometimes known as “actual cause”), and (2) legal (or “proximate”) cause. (a) Cause in Fact As a general matter, “cause in fact” refers to the basic requirement that the defendant’s conduct must have an actual causal relationship to the forbidden result (here, the death of another human being). See Burns v. State, 240 Ga. 827, 828 (242 SE2d 579) (1978) (explaining that a “causal relationship” between the defendant’s act and the victim’s death is required). The cause-in-fact requirement shows up in our decisions in the form of various statements that define the necessary causal relationship itself. We have said, for example, that the defendant’s conduct “causes” a death if it “played a substantial part in bringing about or actually causing the death.” Jackson, 287 Ga, at 648-649 (2) (quoting Skaggs, 278 Ga. at 19-20 (1)); Chaney v. State, 281 Ga. 481, 482 (1) (640 SE2d 37) (2007) (same).[5] That showing is ordinarily made through evidence from which a jury can infer that the forbidden result would not have happened “but for” the defendant’s conduct. See, e.g., Daddario v. State, 307 Ga. 179, 186-187 (2) (a) (2019) (rejecting argument that evidence of causation was not sufficient because evidence was presented showing that the victim’s severe injuries would not have occurred “but for” the defendant’s act of molestation). See also Bell v. State, 317 Ga. 519, 523 (893 SE2d 918) (2023) (explaining conduct is considered the “cause” where “without [it] the result would not have occurred” (quoting Jackson, 287 Ga. at 648 (2))). And we have made clear that such a showing will establish cause in fact whether or not the defendant’s conduct was the “sole” or “immediate” cause of death. Jackson, 287 Ga. at 649 (2) (explaining that a defendant’s injury-creating conduct causes a death if “the injury itself constituted the sole proximate cause of the death” or “the injury directly and materially contributed to the happening of a subsequent accruing immediate cause of the death”). See also Wayne R. LaFave, 1 Subst. Crim. L. § 6.4 (b) (3d ed.) (“In order that conduct be the actual cause of a particular result, it is almost always sufficient that the result would not have happened in the absence of the conduct; or, putting it another way, that ‘but for’ the antecedent conduct the result would not have occurred.”).[6] Finally, if the evidence shows that the death would have eventually happened even absent the defendant’s conduct, we have said that the defendant’s conduct is still a cause if it “materially accelerated” the death. Jackson, 287 Ga. at 649 (2) (quoting Wilson, 190 Ga. at 829)). See also Wayne R. LaFave, 2 Subst. Crim. L. § 6.4 (3d ed.) (agreeing that “one who hastens the victim’s death is a cause of his death”). (b) Legal Cause In addition to cause in fact, our decisions describe a separate requirement that we refer to here as legal cause.[7] It takes as a given that the defendant’s conduct was causally connected to the forbidden result — in a “but for” sense or otherwise — and asks whether the death resulted from the defendant’s conduct in such an unforeseen or attenuated way that the defendant cannot be held accountable for the death. See Bell, 317 Ga. at 523 (citing Johnson v. Avis Rent A Car Systems, LLC, 311 Ga. 588, 593 (858 SE2d 23) (2021) (“The requirement of proximate cause constitutes a limit on legal liability; it is a policy decision that, for a variety of reasons, e.g., intervening act, the defendant’s conduct and the plaintiff’s injury are too remote for the law to countenance recovery.”)). See also LaFave, 1 Subst. Crim. L. § 6.4 (b) (3d ed.) (explaining that “even when cause in fact is established, it must be determined that any variation between the result intended (with intent crimes) or hazarded and (with reckless or negligent crimes) the result actually achieved is not so extraordinary that it would be unfair to hold the defendant responsible for the actual result”). To that end, the legal-cause requirement shows up in our causation decisions as language specifying the requisite foreseeability or likelihood of the death happening in the way it did, given the defendant’s conduct. As we put it recently, legal cause “requires that the death actually happened in a way that was a reasonably foreseeable result of the criminal conduct—that is, the death must also have been a ‘probable or natural consequence’ of the criminal conduct.” Eubanks v. State, 317 Ga. 563, 569 (2) (a) (ii) (894 SE2d 27) (2023). See also Jackson, 287 Ga. at 654 (3) (“Proximate causation imposes liability for the reasonably foreseeable results of criminal (or, in the civil context, tortious) conduct if there is no sufficient, independent, and unforeseen intervening cause.”); Skaggs, 278 Ga. at 20 (1) (“In cases of felony murder, ‘for example, legal cause will not be present where there intervenes (1) a coincidence that is not reasonably foreseeable . . . or (2) an abnormal response.’” (quoting 1 LaFave, Substantive Criminal Law, § 6.4 (h), p. 495 (2d ed. 2003)). And “probable” means just that: “a person is not responsible for a consequence which is merely possible, according to occasional experience, but only for a consequence which is probable, according to ordinary and usual experience.” Bell, 317 Ga. at 523 (cleaned up). See also id. (explaining that “probable” means “not unlikely” or “such a chance of harmful result that a prudent man would foresee the risk at issue here, i.e. the risk of serious injury or death”). Our causation decisions also describe legal cause when they address “intervening acts” — that is, links in the causal chain, like the conduct of another person or other causal forces that happen between the defendant’s conduct and the death. See Eubanks, 317 Ga. at 570-571 (2) (a) (ii); Wilson v. State, 315 Ga. 728, 809 (4) (883 SE2d 802) (2023) (“[T]he defendant is liable for the reasonably foreseeable results of [his] criminal conduct if there is no sufficient, independent, and unforeseen intervening cause.” (cleaned up)); Jackson, 287 Ga. at 654 (3). As with legal cause generally, the inquiry about intervening acts asks whether an intervening act was itself a natural or probable consequence of the defendant’s conduct. See Eubanks, 317 Ga. at 570 (2) (a) (ii) & n.3. That would be true when, for example, the intervening act “ensue[s] in the ordinary course of events,” or was “set in motion by the original wrong-doer,” as when the intervening act was undertaken “by someone other than the defendant who could reasonably be expected to take that action in response to the criminal conduct.” Id. at 570 (2) (a) (ii) (cleaned up). See Bell, 317 Ga. at 522-523; Skaggs, 278 Ga. at 20 (1). When an intervening act was a natural or probable consequence of the defendant’s conduct, a finding of legal cause is not precluded. And this is so even when the intervening act was “not intended by the defendant,” like when the victim is “especially vulnerable” or suffers a “medical complication as a result of the defendant’s conduct.” Eubanks, 317 Ga. at 571 (2) (a) (ii). * So the standard for proving the element of causation for a murder charge can be distilled from our decisions as follows. As we held in Jackson, proving that a defendant “caused” the death of another human being requires proof of “proximate cause.” Jackson, 287 Ga. at 647; id. at 654 (3). This showing has two components: cause in fact and legal cause. See e.g., Jackson, 287 Ga. at 648-649 (2) (“In a criminal case, proximate cause exists when the accused’s act or omission played a substantial part in bringing about or actually causing the victim’s injury or damage and the injury or damage was either a direct result or a reasonably probable consequence of the act or omission.” (quoting Skaggs, 278 Ga. at 19-20 (1)) (cleaned up and emphasis added)). See also LaFave, 1 Subst. Crim. L. § 6.4(a) (3d ed.) (describing cause in fact as “the word ’cause’ in the phrase . . . ‘proximate cause,’” and legal cause as “the word . . . ‘proximate’ in the phrase . . . “proximate cause”). Cf. McGrath v. State, 277 Ga. App. 825, 828 (627 SE2d 866) (2006) (proving causation in a vehicular homicide case requires showing that “the defendant’s conduct was the ‘legal’ or ‘proximate’ cause, as well as the cause in fact, of the death” (cleaned up)); Miller v. State, 236 Ga. App. 825, 828 (513 SE2d 27) (1999) (same). A defendant’s conduct is a cause in fact of a death if the defendant’s conduct “played a substantial part in bringing about or actually causing” the death — typically shown through evidence that the death would not have happened “but for” the defendant’s conduct — or if the defendant’s conduct “materially accelerated the death.” Jackson, 287 Ga. at 648-649 (2); Daddario, 307 Ga. at 186-187 (2) (a); James, 250 Ga. at 655. And a defendant’s conduct is a legal cause of a death if the death was “reasonably foreseeable” — that is, a “probable or natural consequence” of the criminal conduct “according to ordinary and usual experience,” not a “merely possible” result. Eubanks, 317 Ga. at 569 (2) (a) (ii); Bell, 317 Ga. at 523. Determining whether the State has proved these requirements in any given case is “fact-intensive” and demands “mixed considerations of logic, common sense, justice, policy, and precedent,” so questions of causation are “generally left to the jury at trial.” Jackson, 287 Ga. at 652 (2). 3. This Case Having clarified the standard for proving causation under our murder statute, we can turn to this case. On appeal from his conviction for second-degree murder, see OCGA § 16-5-1 (d), Melancon contends that the State failed to present sufficient evidence that he caused the victim’s death.[8] The Court of Appeals disagreed, concluding that the evidence authorized the jury to find that Melancon caused Laura’s death at Higgenbotham’s hands by telling Long to stop cooperating with DFCS. To review that conclusion, we must consider whether a rational trier of fact could have found, beyond a reasonable doubt, that Melancon’s conduct was the proximate cause — both cause in fact and legal cause — of Laura’s death. See Jackson v. Virginia, 443 U.S. 307, 318-19 (99 SCt 2781, 61 LEd2d 560) (1979) (explaining that the “critical inquiry on review of the sufficiency of the evidence to support a criminal conviction” is “whether, after viewing the evidence in the light most favorable to the prosecution, any rational trier of fact could have found the essential elements of the crime beyond a reasonable doubt”). In doing so, as with any due- process sufficiency review, we view the evidence presented at trial in the light most favorable to the verdict, id., “leav[ing] to the jury the resolution of conflicts or inconsistencies in the evidence, credibility of witnesses, and reasonable inferences to be derived from the facts.” Perkins v. State, 313 Ga. 885, 891 (2) (a) (873 SE2d 185) (2022) (citation and punctuation omitted). (a) Cause in Fact Although the Court of Appeals did not expressly address cause in fact, its reasoning indicates that it determined that a rational juror could have found that Melancon’s conduct was a but-for cause of Laura’s death. The Court of Appeals accepted the State’s theory, presented in its indictment of Melancon, that Melancon’s instructing Long to stop cooperating with DFCS (after she initially reported Higgenbotham’s abuse of Laura) caused Laura’s death. Melancon, 368 Ga. App. at 347 (1). In support of that theory, the Court reasoned that a jury could find “that, had Melancon not interfered to end the DFCS investigation, Higgenbotham’s abuse of Laura would not have continued.” Id. at 345 (1). In other words, the Court thought the evidence supported a finding that but for Melancon’s instruction to Long, Higgenbotham’s abuse would have stopped, and she would not have killed Laura. At the outset, this particular kind of cause-in-fact theory faces an uphill battle. To establish that Melancon’s instruction was a cause in fact of Laura’s death, the State had to prove that it “played a substantial part in bringing about or actually causing” that death. Jackson, 287 Ga. at 653 (2) (quoting Skaggs, 278 Ga. at 19-20 (1)). In a case where the defendant’s conduct was not the most immediate cause of death, this showing is typically made through evidence that he nonetheless “directly and materially contributed” in some affirmative way to that immediate cause. For instance, in Skaggs, we held that kicking the victim in the face with his steel-toed boots, resulting in the victim falling headfirst onto the concrete, caused the victim’s death some days later from trauma to skull and brain because the fall “was the direct and immediate result” of the defendant’s actions and “[t]he only intervening force was gravity.” Skaggs, 278 Ga. at 20 (1). See also Larkin v. State, 247 Ga. 586, 587 (278 SE2d 365) (1981) (upholding felony murder conviction because defendant “caused his mother-in-law’s death” when he stabbed her after she tried to stop him from assaulting his wife, and his mother-in-law later died from a complication of surgery to re-stitch the knife wound (citation omitted)). And in Calhoun v. State, 38 Ga. 146, 150-151 (2) (a) (839 SE2d 612) (2020), we held that a defendant caused the death by fleeing, where a police officer’s use of a “PIT” maneuver to end a high-speed car chase resulted in the defendant’s car crashing and killing his passenger. By contrast, the theory here — that Melancon caused Laura’s death by stopping Long from cooperating with DFCS — is that the defendant’s conduct merely foreclosed one conceivable way that a third party might have intervened to stop another third party from killing the victim. For such a preventing-someone-from- preventing-murder theory to work (if it can work), the State would have to establish not only that the defendant’s conduct in fact prevented the hypothetical intervening third-party conduct, but also that the intervention would have in fact succeeded in stopping the murder. Without such evidence, it cannot be said that the defendant’s conduct — preventing the putative intervention — played a substantial part in bringing about the death, much less that the death would not have happened but for the defendant’s conduct. And that’s precisely the kind of evidence that is missing here. The Court of Appeals was right that some evidence showed that but for Melancon’s instruction to Long, DFCS could have found Higgenbotham and Laura. Long testified that she would have “filed the report” about Laura’s fingernail mark and bruise if Melancon had not stopped her, and the jury could have inferred from her testimony that she would have told DFCS where to find Higgenbotham and Laura (even though she did not share their whereabouts when she first called DFCS with Melancon’s approval). But as the Court of Appeals also acknowledged, the jury heard no evidence at all about what DFCS would have done next. There was no evidence in the record, for instance, that based on allegations of a bruise and fingernail mark, DFCS would have removed Laura from Higgenbotham’s custody or taken some other action that would eliminate the possibility of further abuse. The Court of Appeals suggested that the jury could nonetheless “infer” that DFCS “would have intervened in a way to stop the ongoing abuse,” Melancon, 368 Ga. App. at 346 (1), but inferences require some basis in evidence. Without any evidence even suggesting how DFCS would have responded to Long’s report and cooperation, or the likelihood that it would have done so based on the evidence of which it was aware, any conclusion about further actions DFCS might have taken — and whether any such actions would have prevented Higgenbotham from inflicting fatal head trauma to Laura months later — would be speculation, not infer- ence.[9] In short, this record did not authorize a jury to conclude beyond a reasonable doubt that DFCS would have prevented Laura’s death at Higgenbotham’s hands but for Melancon telling Long not to “file a report” about Laura’s fingernail mark and bruise two months earlier. Because the evidence did not authorize the jury to find that Laura’s death would not have happened but for Melancon telling Long not to cooperate, and the State has not advanced any theory other than but-for causation (for example, that Melancon’s conduct somehow accelerated Laura’s death), we conclude that the evidence was not sufficient to establish that instructing Long not to file a report was a cause in fact of Laura’s death. (b) Legal Cause As explained above, the analysis of legal cause takes as a given that the defendant’s conduct was causally connected to the forbidden result and asks whether the death that resulted from the defendant’s conduct was reasonably foreseeable. So here, we ask whether, assuming Melancon’s instruction to Long was an actual cause of Laura’s death, her death by head trauma at Higgenbotham’s hands two months later was a reasonably foreseeable result of that instruction.[10] As we discussed above, that standard is met only if the death is a “probable or natural consequence” of the conduct in question. Eubanks, 317 Ga. at 569 (2) (a) (ii). A death can be a probable or natural consequence even if an intervening act happened between the defendant’s conduct and the death, as long as the intervening act, too, was a probable or natural consequence of that conduct. See Id. at 570 (2) (a) (ii). In all events, however, a defendant “is not responsible for a consequence which is merely possible, according to occasional experience, but only for a consequence which is probable, according to ordinary and usual experience.” Bell, 317 Ga. at 523. Measured against this standard, the evidence here does not support a finding that Melancon’s conduct was a legal cause of Laura’s death. When Melancon told Long not to “file the report” about the fingernail mark and bruise on Laura, certain consequences may have been reasonably foreseeable: Given the evidence of his controlling relationship with Long, one could conclude that she was likely going to heed his instruction, and indeed, she asked DFCS to “cancel the report.” It could also be fair to conclude that ending Long’s cooperation with DFCS might well prevent DFCS from finding Higgenbotham and Laura. And although Melancon addressed this particular incident by putting Laura in Long’s care for several days, given evidence that Melancon may have known about not only the bruise and fingernail mark but also earlier incidents of concern, the evidence arguably supported a conclusion that more incidents of harm to Laura going forward would be a reasonably foreseeable (i.e., natural or probable) consequence of his conduct aimed at stopping possible action by DFCS. But the legal-cause inquiry here is not merely whether it was reasonably foreseeable that Higgenbotham’s abuse of Laura would continue if Melancon stopped Long from filing the report about the bruise and fingernail mark. The question is whether the intervening act of Higgenbotham inflicting fatal injuries, including massive head trauma on Laura almost two months later was a natural or probable consequence of that instruction. And we see no basis in the record for that conclusion. We do not deny that it could be “possible, according to occasional experience,” that declining to help authorities investigate allegations of child abuse of this nature could result in the authorities not taking some action that conceivably would have prevented some future act of abuse serious enough to cause the child’s death. See Bell, 317 Ga. at 523. But nothing in this record indicates that such a consequence is “probable, according to ordinary and usual experience” as a general matter, or was so in the particular circumstances of this case. Id. As the State concedes, there is no argument that Melancon’s instruction “set in motion” Higgenbotham’s later, fatal abuse of Laura, or that he reasonably could have expected it as some kind of “response” to his conduct. See Eubanks, 317 Ga. at 570 (2) (a) (ii) (citing Davis v. State, 290 Ga. 757, 760 (4) (725 SE2d 280) (2012) (defendant proximately caused death where defendant’s brother fatally shot drug dealer); Robinson v. State, 298 Ga. 455, 456, 458-459 (1) (782 SE2d 657) (2016) (defendant proxi- mately caused death when store owner they were robbing shot and killed his accomplice)). Nor can we see how Laura’s death would be expected to “ensue in the ordinary course of events” from stopping Long’s cooperation with the authorities given Melancon’s knowledge of, at most, a small number of milder (relative to the massive and fatal trauma that caused Laura’s death) incidents of harm in the past. The record shows that Long and Melancon were concerned about Higgenbotham’s ability to parent and her treatment of Laura, but there was no evidence that Higgenbotham had inflicted any potentially fatal injuries before Laura’s death, much less that Melancon knew about any such incidents.[11] So although the evidence may have offered a basis for believing that Laura’s later death at Hig- genbotham’s hands was a possible consequence of preventing Long from cooperating with DFCS, the record does not support a finding that Laura’s death by violent head trauma “could reasonably have been anticipated, apprehended, or foreseen” as the consequence of that instruction. See Eubanks, 317 Ga. at 569-570 (2) (a) (ii) (cleaned up). The attenuated nature of the State’s theory of causation also informs this conclusion. The State’s theory was that it was reasonably foreseeable to Melancon that when he told Long not to cooperate with DFCS, (1) Long would comply, (2) DFCS would not be able to find Higgenbotham and Laura, and that (3) Higgenbotham would ultimately inflict injuries that caused Laura’s death. All else equal, legal cause generally will be harder to establish when it turns on whether it was reasonably foreseeable to the defendant what third parties would do in response to his conduct — and even more so when, like here, the chain of causation involves multiple third parties acting in particular ways over an extended period of time. See Eubanks, 317 Ga. at 569 (2) (a) (ii) (explaining that in a hypothetical case in which someone dies because they tried to flee from an ongoing robbery by climbing from a third-floor balcony to the balcony below, slipped, and then fell thirty feet to the ground, “proximate cause would be a closer question” (citing Stafford v. State, 312 Ga. 811 (865 SE2d 116) (2021) (decided on other grounds)). That is not to say that a finding of legal cause is foreclosed merely because a causal chain is made up of a series of events that must happen in sequence, or includes third parties, but as a matter of logic and probability, such attenuated chains of causation typically will be harder to characterize as “natural or probable.” And when the ultimate result becomes “too remote” a possibility in such a case — even though the attenuated chain of causation ultimately happened — legal cause stands as a limit on legal liability. Bell, 317 Ga. at 523 (quoting Johnson, 311 Ga. at 593)); See also LaFave, 1 Subst. Crim. L. § 6.4 (b) (3d ed.). Just so here. Finally, the ramifications of the State’s theory of causation here warrant mention. Reduced to its simplest form, that theory is that the defendant caused a child’s death at the hands of another by stopping the reporting of an earlier allegation of that other person’s act of abuse against that child. If that were a sufficient theory of causation as a general matter, any failure to report an allegation of almost any type of child abuse likely could be prosecuted as second- degree murder if the child later dies as a result of abuse. OCGA § 16-5-1 (d) (“A person commits the offense of murder in the second degree when, in the commission of cruelty to children in the second degree, he or she causes the death of another human being irrespective of malice.”). And there is no particular reason that this theory would not transfer to failures to report conduct outside the context of child abuse. Such an expansive construction of our murder statute’s causation element strikes us as untenable, especially since our legislature has enacted a statute that (1) makes reporting of suspected child abuse mandatory for only a subset of people (like doctors, teachers, and police), and (2) makes the failure to report such abuse only a misdemeanor. See OCGA §§ 19-7-5 (c) (1), (h). Because what “causes” means as it is used in OCGA § 16-5-1 (d) is, at bottom, a question of statutory construction, and statutory context is an important indicator of meaning, the serious tension (if not conflict) with the mandatory reporting statute that this particular theory of causation creates is a final reason to reject it here. For all of these reasons, the evidence here was not sufficient to establish that Melancon’s instruction to Long to not cooperate with DFCS was a legal cause of Laura’s murder two months later at Higgenbotham’s hands. 4. Conclusion In sum, the evidence here did not authorize a jury to conclude that Melancon telling Long not to “file a report” with DFCS was either a cause in fact or legal cause of Laura’s death. Because both cause in fact and legal cause are necessary components of the element of causation under OCGA § 16-5-1 (b), the Court of Appeals erred in concluding that there was sufficient evidence of causation based on the theory of causation it reviewed. That said, our decision today addresses only the theory of causation the Court of Appeals reviewed below, which was one of three theories the State included in Melancon’s indictment.[12] Thus, on remand, the Court of Appeals should apply the framework set forth above and consider in the first instance whether the evidence of causation in this case was sufficient under a theory other than the one we have rejected here. Judgment vacated and case remanded. All the Justices concur.
