*374*3Defendant Yazan Aledamat was charged with assault with a deadly weapon, specifically a box cutter. A few objects are inherently deadly weapons. Others, including a box cutter, are deadly weapons only if used in a way that makes them deadly weapons. Here, the trial court erroneously permitted the jury to consider the box cutter an inherently deadly weapon. It presented the jury with two possible theories of guilt: (1) that the box cutter was inherently deadly, and (2) that defendant used the box cutter in a deadly way. The first of these theories was erroneous under the facts. A box cutter is, as a matter of law, not inherently deadly. The second theory was correct. We must decide what standard of review applies to this error.
We conclude the usual "beyond a reasonable doubt" standard of review established in Chapman v. California (1967) 386 U.S. 18, 24, 87 S.Ct. 824, 17 L.Ed.2d 705 ( Chapman ) for federal constitutional error applies. The reviewing court must reverse the conviction unless, after examining the entire cause, including the evidence, and considering all relevant circumstances, it determines the error was harmless beyond a reasonable doubt. On this record, applying this standard, we conclude beyond a reasonable doubt that the error *4was harmless. Accordingly, we reverse the judgment of the Court of Appeal, which found the error prejudicial.
I. FACTUAL AND PROCEDURAL HISTORY
The Court of Appeal opinion summarized the facts. "In October 2016, defendant approached a woman working at a lunch truck parked in downtown Los Angeles. He told her that he found her attractive and asked her for her phone number; she declined, explaining that she was married with children. On October 22, 2016, defendant approached the woman's husband, who owned the food truck. Defendant asked, 'Where's your wife?' Defendant then told the man that he wanted to 'fuck' his wife because she was 'very hot' and 'had a big ass and all of that.' When the man turned away to remove his apron, defendant pulled a box cutter out of his pocket and extended the blade; from three or four feet away, defendant thrust the blade at the man at waist level, saying, 'I'll kill you.' Two nearby police officers on horses intervened and arrested defendant." ( People v. Aledamat (2018) 20 Cal.App.5th 1149, 1151-1152, 229 Cal.Rptr.3d 771 ( Aledamat ).)
As relevant to the issue on review, the People charged defendant with assault with a deadly weapon under Penal Code section 245, subdivision (a)(1), and making a criminal threat under Penal Code section 422.1 As to the threat charge, the People *375also alleged that defendant personally used a deadly and dangerous weapon. (§ 12022, subd. (b)(1).) The case went to a jury trial.
The court instructed the jury that, for the assault charge, the People had to prove the following: "The defendant did an act with a deadly weapon other than a firearm that by **281its nature would directly and probably result in the application of force to say [sic ] person; the defendant did that act willfully; when the defendant acted, he was aware of facts that would lead a reasonable person to realize that his act by its nature would directly and probably result in the application of force to someone; and when the defendant acted, he had the present ability to apply force with a deadly weapon other than a firearm to a person." (See CALCRIM No. 875.)
The court defined "a deadly weapon" as "any object, instrument, or weapon that is inherently deadly or one that is used in such a way that it is capable of causing and likely to cause death or ... great bodily injury." (See CALCRIM No. 875.) Regarding the weapon enhancement, the court instructed that "a deadly or dangerous weapon is any object, instrument, or weapon that is inherently dangerous, ... or one that is used in such a way that it is capable of causing or likely to cause death or great bodily injury. In *5deciding whether an object is a deadly weapon, consider all of the surrounding circumstances including when and where the object was possessed and any other evidence that indicates whether the object would be used for a dangerous rather than a harmless purpose." (See CALCRIM No. 3145.) The court did not define what "inherently" deadly or dangerous meant.
In his opening argument to the jury, the prosecutor argued that the box cutter was a "deadly weapon" because "[i]f used in a way to cause harm, it would cause harm." Emphasizing the word "probably," defense counsel argued that defendant's act would not probably result in the application of force to the victim. Defense counsel did not specifically discuss whether the box cutter was a deadly weapon. In his closing argument, the prosecutor argued that the box cutter was an "inherently deadly weapon," noting that "you wouldn't want your children playing with" it.
The jury convicted defendant of both counts and found the weapon allegation true. The court sentenced defendant to prison, and he appealed.
The Court of Appeal affirmed the conviction for making a criminal threat. But it reversed the conviction of assault with a deadly weapon and the true finding on the weapon allegation. It found that the trial court erroneously permitted the jury to find the box cutter to be an inherently deadly weapon. It believed the error required it to reverse the conviction " 'absent a basis in the record to find that the verdict was actually based on a valid ground,' " which "exists only when the jury has 'actually ' relied upon the valid theory." ( Aledamat , supra , 20 Cal.App.5th at p. 1153, 229 Cal.Rptr.3d 771.) It found "no basis in the record for concluding that the jury relied on the alternative definition of 'deadly weapon' (that is, the definition looking to how a noninherently dangerous weapon was actually used)." ( Id . at p. 1154, 229 Cal.Rptr.3d 771.)
The Court of Appeal added that "the rules regarding prejudice that we apply in this case are arguably in tension with more recent cases, such as People v. Merritt (2017) 2 Cal.5th 819 [216 Cal.Rptr.3d 265, 392 P.3d 421], providing that the failure to instruct on the elements of a crime does not require reversal if those omitted elements are 'uncontested' and supported by ' "overwhelming evidence." ' ( Id . at pp. 821-822, 830-832, 216 Cal.Rptr.3d 265, 392 P.3d 421 ; see *376Neder v. United States (1999) 527 U.S. 1, 17-18, 119 S.Ct. 1827, 144 L.Ed.2d 35.) That test would certainly be satisfied here, where defendant never disputed that the box cutter was being used as a deadly weapon and where the evidence of such use is overwhelming." ( Aledamat , supra , 20 Cal.App.5th at p. 1154, 229 Cal.Rptr.3d 771.)
We granted the Attorney General's petition for review to determine the standard of review of error of this kind, and to determine whether the error was prejudicial under this standard.
*6II. DISCUSSION
A. The Error
The jury found defendant guilty of assault with a deadly weapon under section 245, subdivision (a)(1), and, as to the criminal threat charge, it found true that defendant personally used a deadly or dangerous weapon under section 12022, subdivision (b)(1). The court instructed the jury that a weapon **282could be either inherently deadly or deadly in the way defendant used it. The instruction accurately stated the law. However, as the parties agree, the evidence did not support the instruction.
"As used in section 245, subdivision (a)(1), a 'deadly weapon' is 'any object, instrument, or weapon which is used in such a manner as to be capable of producing and likely to produce, death or great bodily injury.' [Citation.] Some few objects, such as dirks and blackjacks, have been held to be deadly weapons as a matter of law; the ordinary use for which they are designed establishes their character as such. ( People v. Graham (1969) 71 Cal.2d 303, 327 [78 Cal.Rptr. 217, 455 P.2d 153] ....) Other objects, while not deadly per se, may be used, under certain circumstances, in a manner likely to produce death or great bodily injury. In determining whether an object not inherently deadly or dangerous is used as such, the trier of fact may consider the nature of the object, the manner in which it is used, and all other facts relevant to the issue." ( People v. Aguilar (1997) 16 Cal.4th 1023, 1028-1029, 68 Cal.Rptr.2d 655, 945 P.2d 1204 ; accord, People v. Perez (2018) 4 Cal.5th 1055, 1065, 232 Cal.Rptr.3d 51, 416 P.3d 42.)
Because a knife can be, and usually is, used for innocent purposes, it is not among the few objects that are inherently deadly weapons. "While a knife is not an inherently dangerous or deadly instrument as a matter of law, it may assume such characteristics, depending upon the manner in which it was used ...." ( People v. McCoy (1944) 25 Cal.2d 177, 188, 153 P.2d 315.) "A box cutter is a type of knife" that, "because it is designed to cut things and not people," is not an inherently deadly weapon as a matter of law under McCoy . ( Aledamat , supra , 20 Cal.App.5th at p. 1153, 229 Cal.Rptr.3d 771 ; see People v. Stutelberg (2018) 29 Cal.App.5th 314, 317, 240 Cal.Rptr.3d 156 ( Stutelberg ) [also involving assault with a box cutter].)2
*377*7Accordingly, as the Court of Appeal held in this case, and as the court in Stutelberg , supra , 29 Cal.App.5th at page 317, 240 Cal.Rptr.3d 156, held, the trial court erred in presenting the jury with two theories by which it could find the box cutter a deadly weapon: (1) inherently or (2) as used. The first theory (inherently) is incorrect, but the second theory (as used) is correct.
In People v. Guiton (1993) 4 Cal.4th 1116, 17 Cal.Rptr.2d 365, 847 P.2d 45 ( Guiton ), we considered the consequences when a court instructs on two theories of guilt, one correct and the other incorrect.3 We distinguished between two categories of incorrect theories. Under what we called a " 'factually inadequate theory,' " the theory is incorrect only because the evidence does not support it. ( Id . at p. 1128, 17 Cal.Rptr.2d 365, 847 P.2d 45.) We said that "[i]f the inadequacy of proof is purely factual, of a kind the jury is fully equipped to detect, reversal is not required whenever a valid ground for the verdict remains, absent an affirmative indication in the record that the verdict actually did rest on the inadequate ground." ( Id . at p. 1129, 17 Cal.Rptr.2d 365, 847 P.2d 45 ; see People v. Rivera (2019) 7 Cal.5th 306, 329, 247 Cal.Rptr.3d 363, 441 P.3d 359.)
Under what we called a " 'legally inadequate theory,' " the theory is incorrect because it is contrary to law. ( Guiton , supra , 4 Cal.4th at p. 1128, 17 Cal.Rptr.2d 365, 847 P.2d 45.) An example of this second category "is a case where the inadequate theory 'fails **283to come within the statutory definition of the crime.' " ( Ibid . ) As an example, we cited People v. Green (1980) 27 Cal.3d 1, 164 Cal.Rptr. 1, 609 P.2d 468, where the jury was erroneously permitted to consider 90 feet to be sufficient to satisfy the asportation element for kidnapping. "At issue [in Green ] was whether 90 feet was sufficient asportation to satisfy the elements, or the 'statutory definition,' of kidnapping. There was no insufficiency of proof in the sense that there clearly was evidence from which a jury could find that the victim had been asported the 90 feet. Instead, we held that the distance was 'legally insufficient.' ( Green , supra , 27 Cal.3d at p. 67 [164 Cal.Rptr. 1, 609 P.2d 468], italics added.)" ( Guiton , at p. 1128, 17 Cal.Rptr.2d 365, 847 P.2d 45.)
When the theory is legally erroneous-i.e., of a kind the jury is not equipped to detect-a higher standard must be met for the error to be found harmless. "These different tests reflect the view that jurors are 'well equipped' to sort factually valid from invalid theories, but ill equipped to sort legally valid from invalid theories." ( Aledamat , supra , 20 Cal.App.5th at pp. 1153-1154, 229 Cal.Rptr.3d 771, quoting Guiton , supra , 4 Cal.4th at p. 1126, 17 Cal.Rptr.2d 365, 847 P.2d 45.) Or, as the *8Stutelberg court summarized, "A legal error is an incorrect statement of law, whereas a factual error is an otherwise valid legal theory that is not supported by the facts or evidence in a case. [Citation.] Between the two, legal error requires a more stringent standard for prejudice, for jurors are presumed to be less able to identify and ignore an incorrect statement of law due to their lack of formal legal training. [Citation.] Factual errors, on the other hand, are less likely to be prejudicial because jurors are generally able to evaluate the facts of a case and ignore factually inapplicable theories." *378( Stutelberg , supra , 29 Cal.App.5th at p. 318, 240 Cal.Rptr.3d 156.)
As both the Court of Appeal and Stutelberg held, the error here is of the second category: legal error. Courts have held that a knife is not inherently deadly as a matter of law . Only a few items that are designed to be used as deadly weapons are inherently deadly. ( People v. Perez , supra , 4 Cal.5th at p. 1065, 232 Cal.Rptr.3d 51, 416 P.3d 42 ; People v. Aguilar , supra , 16 Cal.4th at p. 1029, 68 Cal.Rptr.2d 655, 945 P.2d 1204.) If the court had instructed the jury on this point, the error would have been purely factual. "But the jurors were never provided with this definition, and they could reasonably classify a box cutter, which is sharp and used for cutting, as inherently dangerous based on the common understanding of the term. This amounts to legal, rather than factual, error." ( Stutelberg , supra , 29 Cal.App.5th at p. 319, 240 Cal.Rptr.3d 156.) "There was no failure of proof-that is, a failure to show through evidence that the box cutter is an 'inherently dangerous' weapon. Instead, a box cutter cannot be an inherently deadly weapon 'as a matter of law.' " ( Aledamat , supra , 20 Cal.App.5th at p. 1154, 229 Cal.Rptr.3d 771.) Because the trial court here did not define what "inherently deadly" meant, the jury would not be equipped to know that, contrary to what the instruction suggested, a box cutter is not an inherently deadly weapon.
Based on the state of the law at the time, in Guiton , we said that legal error is "subject to the rule generally requiring reversal." ( Guiton , supra , 4 Cal.4th at p. 1128, 17 Cal.Rptr.2d 365, 847 P.2d 45.) But we also said that this does not mean that reversal is always required when the error is legal. ( Id . at p. 1129, 17 Cal.Rptr.2d 365, 847 P.2d 45.) Because the error in Guiton was purely factual, and thus subject to the lenient standard of review applicable to factually inadequate theories, we did not need to decide the exact standard of review of cases involving legal error. ( Id . at p. 1130, 17 Cal.Rptr.2d 365, 847 P.2d 45.) We said that "[o]ne way of finding this kind of error harmless has long been recognized. Sometimes it is possible to determine from other portions of the verdict that the jury necessarily found the defendant guilty on a proper theory. [Citations.] [¶] There may be additional ways by which a court can determine that [legal] error ... is harmless. We leave the question to future cases." ( Id . at pp. 1130-1131, 17 Cal.Rptr.2d 365, 847 P.2d 45.)
We now consider this question.
**284*9B. Standard of Review of the Error
"[B]efore a federal constitutional error can be held harmless, the court must be able to declare a belief that it was harmless beyond a reasonable doubt." ( Chapman , supra , 386 U.S. at p. 24, 87 S.Ct. 824.) This harmless error rule applies in a variety of contexts, such as to error in omitting entirely one or more elements of a charged offense. ( People v. Merritt , supra , 2 Cal.5th at p. 822, 216 Cal.Rptr.3d 265, 392 P.3d 421 ( Merritt ).)
In Merritt , the trial court neglected to instruct the jury on most of the elements of the charged offenses of robbery. We found the error, "serious though it was," subject to harmless error review. ( Merritt , supra , 2 Cal.5th at p. 822, 216 Cal.Rptr.3d 265, 392 P.3d 421.) Relying heavily on United States Supreme Court decisions such as Neder v. United States , supra , 527 U.S. 1, 119 S.Ct. 1827 ( Neder ) (omission of an element of the offense) and Hedgpeth , supra , 555 U.S. 57, 129 S.Ct. 530 (alternative-theory error like that of this case), we held the error "is reversible unless harmless beyond a reasonable doubt." ( Merritt , at p. 822, 216 Cal.Rptr.3d 265, 392 P.3d 421.)
*379Defendant argues that the application of Chapman is different for alternative-theory error than for other misdescriptions of the elements of the charged offense. The Court of Appeal agreed. Citing Guiton , supra , 4 Cal.4th 1116, 17 Cal.Rptr.2d 365, 847 P.2d 45, it believed the error requires reversal unless there is a basis in the record to find that "the jury has 'actually ' relied upon the valid theory ...." ( Aledamat , supra , 20 Cal.App.5th at p. 1153, 229 Cal.Rptr.3d 771.) The court further believed that, on this record, it could not find that the jury actually relied on the valid theory. Guiton did not resolve the question; it reserved it for a future case. (See People v. Chun (2009) 45 Cal.4th 1172, 1203, 91 Cal.Rptr.3d 106, 203 P.3d 425 ( Chun ); People v. Cross (2008) 45 Cal.4th 58, 70, 82 Cal.Rptr.3d 373, 190 P.3d 706 (conc. opn. of Baxter, J.).) We conclude that no higher standard of review applies to alternative-theory error than applies to other misdescriptions of the elements. The same beyond a reasonable doubt standard applies to all such misdescriptions, including alternative-theory error. We agree with the recent Court of Appeal decisions of Stutelberg , supra , 29 Cal.App.5th at pages 319-321, 240 Cal.Rptr.3d 156, and People v. Brown , supra , 210 Cal.App.4th at pages 12-13, 147 Cal.Rptr.3d 848, which reached similar conclusions regarding similar error.4
*10Chun , supra , 45 Cal.4th 1172, 91 Cal.Rptr.3d 106, 203 P.3d 425, involved alternative-theory error concerning the element of implied malice in a murder case. Citing authority that included Neder , supra , 527 U.S. 1, 119 S.Ct. 1827, and Hedgpeth , supra , 555 U.S. 57, 129 S.Ct. 530, we reiterated that "[i]nstructional error regarding the elements of the offense requires reversal of the judgment unless the reviewing court concludes beyond a reasonable doubt that the error did not contribute to the verdict." ( Chun , at p. 1201, 91 Cal.Rptr.3d 106, 203 P.3d 425.) In seeking a more precise test, we quoted Justice Scalia's concurring opinion in another case involving misdescription of an element: " 'The error in the present case can be harmless only if the jury verdict on other points effectively embraces this one or if it is impossible, upon the evidence, to have found what the verdict did find without finding this point as well.' " ( Id. at p. 1204, 91 Cal.Rptr.3d 106, 203 P.3d 425, quoting California v. Roy (1996) 519 U.S. 2, 7, 117 S.Ct. 337, 136 L.Ed.2d 266 (conc. opn. of Scalia, J.).) We then said, "Without holding that this is the only way to find error harmless, we think this test works well here, and we will use it.
**285If other aspects of the verdict or the evidence leave no reasonable doubt that the jury made the findings necessary for conscious-disregard-for-life malice, the erroneous felony-murder instruction was harmless." ( Chun , at pp. 1204-1205, 91 Cal.Rptr.3d 106, 203 P.3d 425.)
In People v. Cross , supra , 45 Cal.4th 58, 82 Cal.Rptr.3d 373, 190 P.3d 706, the defendant argued that alternative-theory error can be found harmless only if the *380verdict itself shows that the jury relied on a valid theory. Justice Baxter authored a concurring opinion arguing against this position, which we find persuasive. He "agree[d] with the weight of existing authority, which applies the Chapman harmless-error standard in determining whether the submission to the jury of two legal theories, one valid and one invalid, requires reversal." ( Cross , at p. 70, 82 Cal.Rptr.3d 373, 190 P.3d 706 (conc. opn. of Baxter, J.).) Noting that a related question was then before the high court ( ibid ., citing the case that later became Hedgpeth , supra , 555 U.S. 57, 129 S.Ct. 530 ), he argued that "an instructional error with respect to an element does not become more problematic simply because the jury may potentially have relied on an alternative theory that was entirely error free. Defendant's argument [that a more stringent test applies to alternative-theory error] 'reduces to the strange claim that, because the jury here received both a "good" charge and a "bad" charge on the issue, the error was somehow more pernicious than in [a high court decision]-where the only charge on the critical issue was a mistaken one. That assertion cannot possibly be right, so it is plainly wrong.' ( Quigley v. Vose (1st Cir. 1987) 834 F.2d 14, 16 ; accord, Becht v. U.S. (8th Cir. 2005) 403 F.3d 541, 548 ["it would be 'anomalous' to preclude harmless-error review under Chapman 'because the jury also was given the option to convict based on a constitutionally valid theory ...'] ....)" ( Cross , at p. 71, 82 Cal.Rptr.3d 373, 190 P.3d 706 (conc. opn. of Baxter, J.).)
A few months after Justice Baxter wrote this in Cross , the high court expressed similar views in the case that he mentioned. ( Hedgpeth , supra , 555 U.S. 57, 129 S.Ct. 530.) After citing decisions such as Neder , supra , 527 U.S. 1, 119 S.Ct. 1827, that *11applied the reasonable doubt standard to review of instructional error regarding the elements of the charged offense, the high court said this: "Although these cases did not arise in the context of a jury instructed on multiple theories of guilt, one of which is improper, nothing in them suggests that a different harmless-error analysis should govern in that particular context. ... [¶] In fact, drawing a distinction between alternative-theory error and the instructional errors in [several cases including Neder ] would be 'patently illogical,' given that such a distinction ' "reduces to the strange claim that, because the jury ... received both a 'good' charge and a 'bad' charge on the issue, the error was somehow more pernicious than ... where the only charge on the critical issue was a mistaken one." ' " ( Hedgpeth , at p. 61, 129 S.Ct. 530, quoting Pulido v. Chrones (9th Cir. 2007) 487 F.3d 669, 677-678 (conc. opn. of O'Scannlain, J.), which quoted Quigley v. Vose , supra , 834 F.2d at p. 16, and citing Becht v. U.S. , supra , 403 F.3d 541, 548.)
Hedgpeth , supra , 555 U.S. 57, 129 S.Ct. 530, involved collateral review on federal habeas corpus. But, in another case of alternative-theory error, the high court "clarif[ied]" that harmless-error analysis "applies equally to cases on direct appeal." ( Skilling v. United States (2010) 561 U.S. 358, 414, fn. 46, 130 S.Ct. 2896, 177 L.Ed.2d 619 [remanding the case to the circuit court to determine whether the error was prejudicial].) As the court that reviewed the Skilling case on remand recognized, in Hedgpeth and Skilling , the court "did not specifically identify the harmless-error standard that is applicable to alternative-theory errors, but it cited to a string of cases that apply a common harmless-error standard to other types of instructional errors." ( U.S. v. Skilling (5th Cir. 2011) 638 F.3d 480, 481 ; see id . at p. 482 [applying the beyond a reasonable doubt standard].) Additionally, Hedgpeth 's statement that nothing "suggests that a *381different harmless-error analysis should govern" alternative-theory error ( Hedgpeth , at p. 61, 129 S.Ct. 530 ), leaves no doubt that the same Chapman analysis of harmless error applies to alternative-theory error as applies to other kinds of **286misdescription of the elements. Federal circuit decisions have consistently applied the Chapman test to alternative-theory error. ( U.S. v. Garrido (9th Cir. 2013) 713 F.3d 985, 994 ; Bereano v. U.S. (4th Cir. 2013) 706 F.3d 568, 577-578 ; U.S. v. Jefferson (4th Cir. 2012) 674 F.3d 332, 360-361 ; U.S. v. Ferguson (2d Cir. 2011) 676 F.3d 260, 276-277 ; U.S. v. Black (7th Cir. 2010) 625 F.3d 386, 388.)
Applying a different standard of error in this case would be particularly anomalous. If the trial court had simply instructed the jury that a box cutter was a deadly weapon as a matter of law, and given no correct instruction whatsoever, the error would clearly be subject to Chapman harmless error review. ( People v. Brooks (2017) 3 Cal.5th 1, 69, 219 Cal.Rptr.3d 331, 396 P.3d 480 [misdescription of an element of a charged offense].) But here, the court also provided the jury with a valid theory. Providing the jury with *12both a valid and an invalid theory should not be subject to a higher standard of review than applies when the court provides the jury only with an invalid theory.
Our decisions in In re Martinez (2017) 3 Cal.5th 1216, 226 Cal.Rptr.3d 315, 407 P.3d 1 ( Martinez ) and People v. Chiu (2014) 59 Cal.4th 155, 172 Cal.Rptr.3d 438, 325 P.3d 972 ( Chiu ) are not to the contrary. In both cases, we reviewed alternative-theory error regarding the elements of first degree murder. In Chiu , we said that "[w]hen a trial court instructs a jury on two theories of guilt, one of which was legally correct and one legally incorrect, reversal is required unless there is a basis in the record to find that the verdict was based on a valid ground. [Citations.] Defendant's first degree murder conviction must be reversed unless we conclude beyond a reasonable doubt that the jury based its verdict on the legally valid theory that defendant directly aided and abetted the premeditated murder." ( Chiu , at p. 167, 172 Cal.Rptr.3d 438, 325 P.3d 972.) But we also noted that questions from the jury during deliberations "shows that the jury may have based its verdict of first degree premeditated murder on the [erroneous] theory." ( Ibid . ) Accordingly, we could not "conclude beyond a reasonable doubt that the jury ultimately based its first degree murder verdict on a different theory, i.e., the legally valid theory that defendant directly aided and abetted the murder." ( Id . at p. 168, 172 Cal.Rptr.3d 438, 325 P.3d 972.)
In Martinez , we applied the same standard to collateral review of cases containing the same error: " Chiu error requires reversal unless the reviewing court concludes beyond a reasonable doubt that the jury actually relied on a legally valid theory in convicting the defendant of first degree murder." ( Martinez , supra , 3 Cal.5th at p. 1218, 226 Cal.Rptr.3d 315, 407 P.3d 1.) But we also noted that the prosecutor had relied heavily on the invalid theory in argument to the jury, and that "an inquiry by the jury during its deliberations suggested that it was considering the" invalid theory. ( Id . at p. 1227, 226 Cal.Rptr.3d 315, 407 P.3d 1.) For these reasons, "we conclude[d] that the Attorney General has not shown beyond a reasonable doubt that the jury relied on a legally valid theory in convicting Martinez of first degree murder." ( Ibid . )
Defendant argues that, by focusing on what the jury actually did, Chiu and Martinez stated a standard different, and higher, than Chapman 's reasonable doubt standard. But Chiu and Martinez were only a *382specific application of the more general reasonable doubt test stated in cases like Neder , supra , 527 U.S. 1, 119 S.Ct. 1827, and Merritt , supra , 2 Cal.5th 819, 216 Cal.Rptr.3d 265, 392 P.3d 421. The test stated in Chiu and Martinez was taken from Chun , supra , 45 Cal.4th at pages 1203 to 1205, 91 Cal.Rptr.3d 106, 203 P.3d 425. (See Chiu , supra , 59 Cal.4th at p. 167, 172 Cal.Rptr.3d 438, 325 P.3d 972.) Chun also stated that the error "requires reversal of the judgment unless the reviewing court concludes beyond a reasonable doubt that the error did not contribute to the verdict." ( Chun , at p. 1201, 91 Cal.Rptr.3d 106, 203 P.3d 425.) Finding beyond a reasonable doubt that the error did not contribute to the verdict is *13essentially the same as finding the error harmless beyond a reasonable doubt. In Chun , we said that one way, but not necessarily the only way, the Chapman test could be satisfied was to apply Justice Scalia's test of whether " 'it is impossible, upon the evidence, to have found what **287the verdict did find without finding' " the missing point as well. ( Id . at p. 1204, 91 Cal.Rptr.3d 106, 203 P.3d 425.) We equated this impossibility with a conclusion that the jury actually made the necessary finding. ( Id . at pp. 1204-1205, 91 Cal.Rptr.3d 106, 203 P.3d 425.)
In determining this impossibility or, more generally, whether the error was harmless, the reviewing court is not limited to a review of the verdict itself. An examination of the actual verdict may be sufficient to demonstrate harmlessness, but it is not necessary. In both Chiu and Martinez , we examined the record and found that it affirmatively showed the jury might have based its verdict on the invalid theory. Because no other basis to find the error harmless beyond a reasonable doubt was at issue, we did not explore whether other ways of finding the error harmless existed. Those cases merely provide one way in which a court might evaluate harmlessness. They do not preclude other ways.
For these reasons, we conclude that alternative-theory error is subject to the more general Chapman harmless error test. The reviewing court must reverse the conviction unless, after examining the entire cause, including the evidence, and considering all relevant circumstances, it determines the error was harmless beyond a reasonable doubt. We disapprove of any interpretation of People v. Green , supra , 27 Cal.3d 1, 164 Cal.Rptr. 1, 609 P.2d 468, that limits the reviewing court to an examination of the jury's findings as reflected in the verdict itself or that is otherwise inconsistent with this opinion.
We now apply this standard to this case.
C. Application of the Standard to This Case
A number of circumstances convince us beyond a reasonable doubt that the error was harmless. It is clear the error did not contribute to the verdict.
The argument that the error was prejudicial supposes that, under the instructions, the jury would believe there were two separate ways it could find the box cutter to have been a deadly weapon. The first method would be simply to find it was inherently deadly without considering any of the surrounding circumstances. The second method would be to consider how defendant used it. Technically, this is correct. The court used the disjunctive "or," which, out of context, would seem to permit such separation. In context, however, it is unlikely the jury would so view the instructions.
The instruction referred to an object that is "inherently deadly or one that is used in such a way that it is capable of causing and likely to cause death *14or ... great bodily injury." This juxtaposition at *383least indicates what the "inherently deadly" language was driving at. Additionally, the jury was also instructed: "In deciding whether an object is a deadly weapon, consider all of the surrounding circumstances including when and where the object was possessed and any other evidence that indicates whether the object would be used for a dangerous rather than a harmless purpose." This part of the instruction suggested the question was unitary, that is, that the jury had to consider all of the circumstances in deciding whether the object was a deadly weapon, either inherently or as used. The jury would likely view the "inherently deadly" language in light of this additional instruction that it had to consider all of the circumstances. Given this additional instruction, it seems unlikely the jury would simply view the box cutter as inherently deadly without considering the circumstances, including how defendant used it.
The arguments of counsel support this conclusion. At one point, the prosecutor stated that the box cutter was inherently deadly because "you wouldn't want your children playing with" it, without further explaining the term. But no one ever suggested to the jury that there were two separate ways it could decide whether the box cutter was a deadly weapon. Defense counsel argued that defendant did not use the box cutter in a way that would probably result in the application of force, that is, that defendant did not assault the victim at all-an argument the jury necessarily rejected when it found defendant guilty of that crime. But counsel never argued that, if he did assault the victim with the box cutter, the box cutter was not a **288deadly weapon. Although defense counsel did not expressly concede that the box cutter was a deadly weapon, he did not contest the point.
Contesting the point would have been futile based on the record here. A box cutter is not inherently deadly because it is not designed for that purpose. But if used to assault someone, i.e., used as a weapon , a box cutter is potentially deadly even if not designed for that purpose. (See People v. Graham , supra , 71 Cal.2d at pp. 327-328, 78 Cal.Rptr. 217, 455 P.2d 153 [explaining that when a sharp or heavy object "is capable of being used in a 'dangerous or deadly' manner, and it may be fairly inferred from the evidence that its possessor intended on a particular occasion to use it as a weapon should the circumstances require, we believe that its character as a 'dangerous or deadly weapon' may be thus established"].) Counsel could readily believe it would be pointless for him to argue that even if (contrary to the argument counsel did make) the jury found defendant assaulted the victim with the box cutter, it was not a deadly weapon. This is particularly so in light of defendant's statement, "I'll kill you."
A nonexclusive way the error can be found harmless beyond a reasonable doubt, one that "work[ed] well" in Chun , supra , 45 Cal.4th at page 1205, 91 Cal.Rptr.3d 106, 203 P.3d 425, and *15that also works well here, is the test derived from Justice Scalia's concurring opinion in California v. Roy , supra , 519 U.S. at page 7, 117 S.Ct. 337. The reviewing court examines what the jury necessarily did find and asks whether it would be impossible, on the evidence, for the jury to find that without also finding the missing fact as well. ( Chun , at pp. 1204-1205, 91 Cal.Rptr.3d 106, 203 P.3d 425.) Here, under the instructions, the jury necessarily found the following: (1) defendant did an act with a deadly weapon (either inherently or as used) that by its nature would directly and probably result in the application of force; (2) defendant was aware of facts that would lead a reasonable person to realize *384that his act by its nature would directly and probably result in the application of force to someone; and (3) defendant had the present ability to apply force with a deadly weapon to a person.
Additionally, the jury must have considered the term "inherently deadly" to mean something . As the Stutelberg court explained, the theoretical risk is that, because the court did not define the term, the jury might have applied its common understanding to find the box cutter deadly because it is sharp and used for cutting. ( Stutelberg , supra , 29 Cal.App.5th at p. 319, 240 Cal.Rptr.3d 156 ; cf. People v. Pruett (1997) 57 Cal.App.4th 77, 86, 66 Cal.Rptr.2d 750 [the trial court did not err in failing to define what is a deadly weapon because "[j]urors can certainly employ common sense and experience to determine whether or not such a knife is a 'deadly' instrument"].) But if the jury did so, it would necessarily find the box cutter deadly in the colloquial sense of the term-i.e., readily capable of inflicting deadly harm-and that defendant used it as a weapon.
"No reasonable jury that made all of these findings could have failed to find" that defendant used the box cutter in a way that is capable of causing or likely to cause death or great bodily injury. ( Merritt , supra , 2 Cal.5th at p. 832, 216 Cal.Rptr.3d 265, 392 P.3d 421.) For all of these reasons, the error was harmless beyond a reasonable doubt.
D. The Instructions
As this case demonstrates, the standard instructions on assault with a deadly weapon and use of a deadly and dangerous weapon are problematic. (See CALCRIM Nos. 875, 3145.) They do not define what is an inherently deadly weapon. Worse, without modification, they provide the jury with the "inherently deadly" theory even in those cases (i.e., most of them) in which the weapon is not inherently deadly as a matter of law. We suggest the instructions be modified to avoid these problems in the future.
In most cases, the inherently deadly language is inapplicable, for most objects are not inherently deadly even if they may be used in a way that *16makes them deadly. The inherently deadly language is also generally unnecessary. For the most part, those objects that are designed for use as a deadly **289weapon will be also used in a way that makes them deadly weapons.5 Accordingly, the standard instruction might be improved by simply deleting any reference in the usual case to inherently deadly weapons.
But because, under current law, some objects, such as dirks and blackjacks, are inherently deadly, instructing on that theory might be appropriate in some cases. (But see fn. 5.) If the prosecution believes the weapon used in a given case is inherently deadly, and it believes modifying the instruction would be useful, it may request the court to add that theory of the case to the instructions. On such a request, the court should consider whether the evidence would support a finding that the weapon is inherently deadly. If so, the court would have discretion to instruct on that theory. If it does so, however, it should also define what is meant by inherently deadly, i.e., an object that is designed for use as a deadly weapon. (See *385People v. Perez , supra , 4 Cal.5th at p. 1065, 232 Cal.Rptr.3d 51, 416 P.3d 42.)
III. CONCLUSION
We reverse the judgment of the Court of Appeal and remand the matter to that court for further proceedings consistent with this opinion.
We Concur:
CANTIL-SAKAUYE, C. J.
CORRIGAN, J.
KRUGER, J.
Concurring and Dissenting Opinion by Justice Liu
I agree with today's opinion that alternative-theory error is subject to the Chapman beyond-a-reasonable-doubt harmless error standard ( Chapman v. California (1967) 386 U.S. 18, 24, 87 S.Ct. 824, 17 L.Ed.2d 705 ) and that our decisions in People v. Chun (2009) 45 Cal.4th 1172, 91 Cal.Rptr.3d 106, 203 P.3d 425, People v. Chiu (2014) 59 Cal.4th 155, 172 Cal.Rptr.3d 438, 325 P.3d 972, and In re Martinez (2017) 3 Cal.5th 1216, 226 Cal.Rptr.3d 315, 407 P.3d 1 are "not to the contrary." (Maj. opn., ante , 251 Cal.Rptr.3d at p. 381, 447 P.3d at p. 286; see id. at pp. 380-382, 447 P.3d at pp. 285-287.)
I part ways with today's opinion, however, with respect to its conclusion that in light of what "the jury necessarily did find ... it would be impossible, on the evidence, for the jury to find that without also finding the missing fact as well." (Maj. opn., ante , 251 Cal.Rptr.3d at p. 383, 447 P.3d at p. 288.) Justice Cuéllar's concurring and dissenting opinion carefully explains why we cannot be confident beyond a reasonable doubt that the jury in this case found the box cutter to be a deadly *17weapon on a valid legal theory. In particular, the trial court's own equivocation and the prosecutor's repeated conflation of the deadly-weapon and force requirements in closing argument could well have misled the jury. (Conc. & dis. opn. of Cuéllar, J., post , 251 Cal.Rptr.3d at pp. 388-389, 447 P.3d at pp. 292-293.) Because of these confusing statements, and because the trial court erred in providing the "inherently deadly" instruction to the jury, the jury may have convicted defendant for conduct that does not constitute the crime of assault with a deadly weapon. Although the jury "must have considered the term 'inherently deadly' to mean something " (maj. opn., ante , 251 Cal.Rptr.3d at p. 384, 447 P.3d at p. 288), it is quite possible that the jury understood "inherently deadly" to mean that the box cutter itself was readily capable of causing deadly harm, without finding that defendant in fact used the box cutter in a manner likely to cause death or great bodily injury. Accordingly, I would affirm the judgment of the Court of Appeal.
Concurring and Dissenting Opinion by Justice Cuéllar
In our constitutional system the right to trial by jury means "the jury, rather than the judge, reach[es] the requisite finding of 'guilty.' " ( **290Sullivan v. Louisiana (1993) 508 U.S. 275, 277, 113 S.Ct. 2078, 124 L.Ed.2d 182 ; see Cal. Const., art. I, § 16.) That's a principle that can be simple to state but difficult to honor, especially when harmless error review is at stake. Because virtually all forms of harmless error review risk infringing on "the jury's factfinding role and affect[ing] the jury's deliberative process in ways that are, strictly speaking, not readily calculable," courts performing harmless error review are walking a tightrope - where they must weigh how an error affected the proceedings without displacing the jury as finder of fact. ( *386Neder v. United States (1999) 527 U.S. 1, 18, 119 S.Ct. 1827, 144 L.Ed.2d 35 ( Neder ).)
That's why caution's been the watchword when we've stepped onto that tightrope. Like the United States Supreme Court, to date we've found instructional error harmless only when we can conclude "beyond a reasonable doubt" either that the jury necessarily relied on a valid legal theory (see People v. Chun (2009) 45 Cal.4th 1172, 1205, 91 Cal.Rptr.3d 106, 203 P.3d 425 ( Chun ); see maj. opn., ante , 251 Cal.Rptr.3d at p. 384, 447 P.3d at p. 288) or that the element omitted or misdescribed "was uncontested and supported by overwhelming evidence , such that the jury verdict would have been the same absent the error" ( Neder , supra , 527 U.S. at p. 17, 119 S.Ct. 1827, italics added; see, e.g., maj. opn., ante , 251 Cal.Rptr.3d at pp. 383-384, 447 P.3d at p. 288; People v. Merritt (2017) 2 Cal.5th 819, 824, 216 Cal.Rptr.3d 265, 392 P.3d 421 ( Merritt ) [finding error harmless where the defense expressly conceded a robbery occurred and there was overwhelming video evidence of the only contested issue]; cf. People v. Canizales (2019) 7 Cal.5th 591, 616, 248 Cal.Rptr.3d 370, 442 P.3d 686 ( Canizales ) [examining both the "strong" and the "conflicting evidence" on a contested issue and noting "both the prosecutor's closing argument and the attempted murder instruction" had "the potential to cause confusion"]; People v. Mil (2012) 53 Cal.4th 400, 417, 135 Cal.Rptr.3d 339, 266 P.3d 1030 ( Mil ) [reversing burglary and robbery special *18circumstances because the defendant "contested whether he acted with reckless indifference to human life" and "the record support[ed] a reasonable doubt as to that element"] ). These tests - permutations of harmless-error review tailored for instructional error - demand searching inquiry, and rightly so: They help us maintain the critical equilibrium between constitutional guarantees and " 'society's interest in punishing the guilty.' " (Id. at p. 383, 447 P.3d at p. 287-288.)
The majority loses that balance today. Tumbling headlong into the jury's factfinding role, the majority fails to live up to the "more general Chapman [v. California ] harmless error test" it purports to apply. (Maj. opn., ante , 251 Cal.Rptr.3d at p. 382, 447 P.3d at p. 287; see Chapman v. California (1967) 386 U.S. 18, 87 S.Ct. 824, 17 L.Ed.2d 705 ( Chapman ).) The result, unfortunately, is an error of our own -- one hardly harmless to the defendant in this case. With respect, I dissent.
I.
To find Yazan Aledamat guilty of assault with a deadly weapon, the jury had to decide not only that he had the box cutter in his hand, but that he used it as a deadly weapon. The trial judge issued form instructions from CALCRIM No. 875, which provides that the prosecution must prove:
(1) that the defendant did an act with a deadly weapon other than a firearm that by its nature would directly and probably result in the application of force to a person;
(2) that the defendant did that act willfully;
(3) that when the defendant acted, he was aware of facts that would lead a reasonable person to realize that his act by its nature would directly and probably result in the application of force to someone; and
(4) that when the defendant acted, he had the present ability to apply force with a deadly weapon other than a firearm. ( CALCRIM No. 875.)
Here's what's also covered by those jury instructions: to "apply force" means "to touch in a harmful or offensive manner," which can include "the slightest touching" if done in a rude or angry way. As long as *387the prosecution can prove the defendant's act would probably result in the application of force, it is not required to prove "that the defendant actually touched someone." ( **291CALCRIM No. 875. ) A deadly weapon other than a firearm, moreover, "is any object, instrument, or weapon that is inherently deadly or one that is used in such a way that it is capable of causing and likely to cause death or great bodily injury." (Ibid. ) Great bodily injury, in turn, "means *19significant or substantial physical injury. It is injury that is greater than minor or moderate harm." (Ibid .)
Presented to the jury during trial, these instructions permit the jury to conclude a box cutter can be a deadly weapon. But to reach that conclusion, a jury must find the box cutter is either "inherently deadly" or was "used in such a way that it is capable of causing and likely to cause death or great bodily injury." ( CALCRIM No. 875.) Yet as the majority explains, this instruction is incorrect, at least in this case: as a matter of law, a box cutter is not inherently deadly. (Maj. opn., ante , 251 Cal.Rptr.3d at p. 376, 447 P.3d at pp. 281-282.) It is, however, capable of causing death or great bodily injury and, depending on how it is used, may even be likely to cause death or great bodily injury. (Ibid. ) So for a proper conviction of assault with a deadly weapon on these facts, the jury would have needed to rely on the second prong of the deadly weapon definition - the legally valid theory in this case.
The majority plays up that there's more than "one way" (maj. opn., ante , 251 Cal.Rptr.3d at pp. 382, 383-384, 447 P.3d at pp. 286-287, 288) a reviewing court can conclude " 'beyond a reasonable doubt that the error did not contribute to the verdict' " (id . at p. 382, 447 P.3d at p. 286, quoting Chun , supra , 45 Cal.4th at p. 1201, 91 Cal.Rptr.3d 106, 203 P.3d 425 ). I agree. But when, we cannot "conclude, beyond a reasonable doubt, that the jury based its verdict on a legally valid theory" ( Chun , supra , 45 Cal.4th at p. 1203, 91 Cal.Rptr.3d 106, 203 P.3d 425 ), we may find the instructional error harmless only if we can determine beyond a reasonable doubt that "the jury verdict would have been the same absent the error." ( Neder , supra , 527 U.S. at p. 17, 119 S.Ct. 1827.) That's the situation here.
In other words, we ask whether it is clear beyond a reasonable doubt the jury would have found Aledamat guilty of assault with a deadly weapon had it been given only the correct instruction, which required it to find that he used a box cutter in a manner "capable of causing and likely to cause death or great bodily injury." ( CALCRIM No. 875.) Or, " 'in typical appellate-court fashion,' " we ask the inverse: " 'whether the record contains evidence that could rationally lead to a contrary finding with respect to the [deadly weapon] element.' " ( Merritt , supra , 2 Cal.5th at p. 832, 216 Cal.Rptr.3d 265, 392 P.3d 421, quoting Neder , supra , 527 U.S. at p. 19, 119 S.Ct. 1827.) And Neder itself provided a useful rubric to evaluate instructional error, explaining that Chapman 's mandate can be satisfied where the element omitted or misdescribed "was uncontested and supported by overwhelming evidence." ( Neder , at p. 17, 119 S.Ct. 1827.)
The majority seems not to have made this inquiry. Indeed, its primary quarrel is with the very notion that the instruction was error, let alone prejudicial error. (Maj. opn., ante , 251 Cal.Rptr.3d at pp. 382-383, 447 P.3d at p. 287.) It argues, in essence, that the jury "likely" understood the term "inherently deadly" to mean something approximating the correct instruction - "capable of and likely to *20cause great bodily injury." (Id. at pp. 382, 383, 447 P.3d at pp. 287, 287.) That is, a jury is "unlikely" to find the disjunctive "or" - "inherently *388deadly or ... capable of causing and likely to cause" great bodily injury ( CALCRIM No. 875, italics added) - to present two alternatives. (Maj. opn., ante , 251 Cal.Rptr.3d at pp. 382-383, 447 P.3d at pp. 287-288.) Instead, the majority says, the jury will understand "inherently deadly" to be defined by the words surrounding it, notwithstanding the disjunctive. (Id. at pp. 382-383, 447 P.3d at p. 287 [explaining the "juxtaposition" of the words "at least indicates what the 'inherently deadly' language was driving at"]; id. at p. 384, 447 P.3d at p. 288 ["[T]he jury must have considered the term 'inherently deadly' to mean something "].) In other words, if the jury doesn't understand that "inherently deadly" is a shortcut, it will hold the prosecution to roughly the same standard that the correct instruction does.
Three problems mar this argument. First, it is pure conjecture. Nothing in the record **292suggests, let alone compels us to conclude, that the jury read the instructions in the way the majority speculates. It is just as likely - and more consistent with principles of English usage - that the "juxtaposition" of two disjunctive clauses suggests just that (maj. opn., ante , 251 Cal.Rptr.3d at pp. 382-383, 447 P.3d at p. 287): They are juxtaposed because they are distinct. One does not define the other, in whole or in part.
Second, the argument is at odds with the majority's characterization of the instructions as "problematic" and its suggestion that they be "modified to avoid these problems in the future." (Maj. opn., ante , 251 Cal.Rptr.3d at p. 384, 447 P.3d at p. 288.) Why alter anything if, as the majority implies, jurors can more or less figure it out on their own? (Id. at p. 384, 447 P.3d at p. 288 [citing People v. Pruett (1997) 57 Cal.App.4th 77, 86, 66 Cal.Rptr.2d 750 for the proposition that jurors can determine whether a knife is a deadly instrument based on "common sense"].) (Indeed, jurors can guess approximate meanings for most legal principles and elements, from robbery to "deadly weapon"; I imagine the majority would not suggest we dispense with those instructions out of an abiding belief that common sense will suffice.)
Third - and perhaps most importantly - in this case there is evidence not only that the jury may have misunderstood the task before it, but that it was affirmatively (though inadvertently) misled.
The confusion began when the judge instructed the jury on the elements. On beginning to read the definition of great bodily injury - and before reaching the definition of a deadly weapon - the judge stopped, said the jury "[didn't] need that definition" and asked the jury to cross it out. The jurors affirmed that they had crossed out the definition. Moving on to the deadly weapon instruction, the judge realized the definition of "great bodily injury" was relevant. He stopped, had a short exchange on the record with counsel, *21told the jury to ignore their earlier strike-through of the definition - "And despite the fact I told you to cross it out, I [now] want you to consider it. Okay? If you want to write it in, you can." - and allowed the jury a brief break. He then recited the criminal threat instructions without revisiting the remainder of the deadly weapon instruction.
The majority highlights a feature of the jury instructions. The jury was instructed to consider "all of the surrounding circumstances" in deciding whether an object is a deadly weapon - " 'including when and where the object was possessed and any other evidence that indicates whether the object would be used for a dangerous rather than a harmless purpose.' " (Maj. opn., ante , 251 Cal.Rptr.3d at p. 383, 447 P.3d at p. 287.) "Given this additional instruction," the majority writes, "it seems unlikely the jury would simply view the box cutter as *389inherently deadly without considering the circumstances, including how the defendant used it." (Id. at p. 383, 447 P.3d at p. 287.)
But the additional instruction didn't apply to the assault charge. The judge read it in instructing the jury on the deadly weapon enhancement, and expressly said it became relevant only if the jury "[found] the defendant guilty of the crimes charged in counts 1 and 2." What's more, during deliberations the jury asked the court "how to deal with the issue of the allegation, the deadly or dangerous weapon allegation." It is far from obvious that the jury understood its obligation to make a finding on the deadliness of the box cutter. (See Canizales , supra , 7 Cal.5th at p. 617, 248 Cal.Rptr.3d 370, 442 P.3d 686 [noting that the jury requested a readback of testimony that "suggests the jurors at one point were focused on testimony that would have supported the [defense] theory"].)
There's more. In closing arguments the prosecutor exacerbated this confusion, suggesting that the mere existence of the box cutter was sufficient to satisfy the deadly weapon allegation: "You don't have to actually inflict injury on the person. What [Aledamat] did was sufficient; he committed a crime, a crime of assault with a deadly weapon. And the added allegation is that he used a box cutter ." (Italics added.) She then conflated the deadly weapon and force requirements, saying, "Ladies and gentlemen, you wouldn't want your children using a box cutter, would you? This is a deadly weapon. If **293used in a way to cause harm, it would cause harm. It's not whether he did cause harm; it's could he; could he have caused harm with that box cutter? The answer: Absolutely."
That statement is wrong. Mangling the recitation of the applicable deadly weapon instruction, the prosecutor's statement confuses the minimal force requirement with the requirement that Aledamat have used the box cutter in a way "likely to cause death or great bodily injury." ( CALCRIM No. 875.) Nevertheless, the majority charges that defense counsel failed to "contest the *22point" (maj. opn., ante , 251 Cal.Rptr.3d at p. 383, 447 P.3d at pp. 287-288) - seemingly recognizing that Neder compels an inquiry of whether the mistaken instruction was contested.
True: We and the United States Supreme Court have said that "removing an element of the crime from the jury's consideration" can be harmless "where the defendant concedes or admits that element." ( People v. Flood (1998) 18 Cal.4th 470, 504, 76 Cal.Rptr.2d 180, 957 P.2d 869 ; see Hurst v. Florida (2016) --- U.S. ----, 136 S. Ct. 616, 623, 193 L.Ed.2d 504 [describing Neder as "holding that the failure to submit an uncontested element of an offense to a jury may be harmless," italics added]; Connecticut v. Johnson (1983) 460 U.S. 73, 87, 103 S.Ct. 969, 74 L.Ed.2d 823 (plur. opn. of Blackmun, J.) [stating that an instructional error "may be harmless" if "defendant himself has taken the issue ... away from the jury"].) But the burden of proof in a criminal trial lies solely with the People. (See Flood , at p. 481, 76 Cal.Rptr.2d 180, 957 P.2d 869.) That burden "is not relieved by a defendant's tactical decision not to contest an essential element of the offense." ( Estelle v. McGuire (1991) 502 U.S. 62, 69, 112 S.Ct. 475, 116 L.Ed.2d 385.) Moreover still, where the defense contested Aledamat's slightest application of force, it defies logic that the defense would not also contest an element predicated - at least in this case - on an even greater application of force. It is at the very least inconsistent with what we have previously required. (See, e.g., Merritt , supra , 2 Cal.5th at p. 824, 216 Cal.Rptr.3d 265, 392 P.3d 421 [finding the defense conceded that a robbery occurred when counsel *390said there was "no question these people were robbed, okay. Our only contention is with element number one that it was not the defendant."].)
And counsel did contest the point - explicitly, at the outset of his closing argument, when he told the jury that one of the "two main questions" that they "need[ed] to answer" was "with respect to the great bodily injury."
On rebuttal, the prosecutor continued to conflate the minimal touching requirement with the dangerousness of the box cutter itself. "As I said before," she said, "you wouldn't want your children playing with this (indicating). It's inherently a deadly weapon. It's by definition the reason this law was created ." (Italics added.) By that time, there was no opportunity for defense counsel to "contest" the point.
It's quite plausible that the jury took the prosecutor at her parting words: the box cutter is "inherently a deadly weapon" and "by definition the reason this law" - assault with a deadly weapon - "was created." The majority all but acknowledges the only possible understanding of these words, noting that "no one ever suggested to the jury that there were two separate ways it could decide whether the box cutter was a deadly weapon." (Maj. opn., ante , 251 Cal.Rptr.3d at p. 383, 447 P.3d at p. 287.) That's true. The prosecutor suggested only one method of finding the *23box cutter was a deadly weapon, and that method was incorrect as a matter of law. That the defense spent so little time discussing the element - which, again, was not its burden to disprove - further risked letting the case pivot on the prosecutor's easy reassurance that "inherently deadly" was, in essence, just a matter of common sense - what objects you wouldn't let your children play with. (See In re Martinez (2017) 3 Cal.5th 1216, 1226-1227, 226 Cal.Rptr.3d 315, 407 P.3d 1 [concluding the jury could have convicted on the invalid theory where the prosecutor argued that theory to the jury "at length during closing argument and rebuttal"].) This is not the kind of record to give one confidence in the majority's argument that the jury understood "inherently deadly" to mean something approximating deadly as-used. Far more likely on this record **294is that the jury would "quickly and easily have found the element satisfied" by "relying on the instructional misdefinition" available to it. ( People v. Harris (1994) 9 Cal.4th 407, 445, 37 Cal.Rptr.2d 200, 886 P.2d 1193 (conc. opn. of Mosk, J.); see Canizales , supra , 7 Cal.5th at p. 614, 248 Cal.Rptr.3d 370, 442 P.3d 686 [concluding there was "a reasonable likelihood that the jury understood the kill zone instruction in a legally impermissible manner" because the prosecution's definition "was significantly broader than a proper understanding of the theory permits" and thus "had the potential to mislead the jury"].) So based on the jury instructions and counsel's arguments we can't conclude "that the jury verdict would have been the same absent the error." ( Neder , supra , 527 U.S. at p. 17, 119 S.Ct. 1827.)
II.
The majority's harmless-error analysis makes scant reference to the evidence in the record. Relying on that record, however, is how we evaluate whether the evidence to support the correct theory - that Aledamat used the box cutter in a way "capable of causing and likely to cause death or great bodily injury" ( CALCRIM No. 875 ) - was so strong that we can safely conclude the instructional error did not contribute to the verdict. (See Neder , supra , 527 U.S. at p. 17, 119 S.Ct. 1827.)
Consider, for instance, what we decided in the recent case of *391In re B.M. (2018) 6 Cal.5th 528, 241 Cal.Rptr.3d 543, 431 P.3d 1180 ( B.M . ). We analyzed the sufficiency of the evidence supporting a conviction for assault with a deadly weapon - in that case, a butter knife. We explained that "for an object to qualify as a deadly weapon based on how it was used, the defendant must have used the object in a manner not only capable of producing but also likely to produce death or great bodily injury." ( Id. at p. 530, 241 Cal.Rptr.3d 543, 431 P.3d 1180, italics omitted.) Though we avoided defining "likely," we explained that we have previously treated the term to mean at least "probable" ( id. at p. 533, 241 Cal.Rptr.3d 543, 431 P.3d 1180 ) - based on "how the defendant actually 'used' the object," rather than on conjecture as to how the object could have been used or what injury might have been inflicted had the object been used differently ( id. at p. 534, 241 Cal.Rptr.3d 543, 431 P.3d 1180 ). *24As an example, we discussed People v. Duke (1985) 174 Cal.App.3d 296, 302, 219 Cal.Rptr. 873, in which the defendant used a headlock to hold his victim. The victim said the headlock made her feel choked but did not cut off her breathing; the defendant's grip was " 'firm,' " but the victim did not testify that he tightened his grip. ( B.M. , supra , 6 Cal.5th at p. 534, 241 Cal.Rptr.3d 543, 431 P.3d 1180.) We favorably cited the Duke court's explanation for reversing the conviction: "[T]he fact that appellant could have easily broken [the victim's] neck or could have choked her to the point of cutting off her breathing by exerting greater pressure on her neck or windpipe will not support the conviction of felony assault." ( B.M. , at p. 534, 241 Cal.Rptr.3d 543, 431 P.3d 1180, quoting Duke , at p. 303, 219 Cal.Rptr. 873, internal quotation marks omitted.) It would involve "gross speculation on the part of the jury as to what the appellant would have done if he had not stopped." ( B.M. , at p. 534, 241 Cal.Rptr.3d 543, 431 P.3d 1180, quoting Duke , at p. 303, 219 Cal.Rptr. 873, internal quotation marks omitted.) We emphasized in B.M. , however, that it is appropriate to consider "what harm could have resulted from the way the object was actually used." ( B.M. , at p. 535, 241 Cal.Rptr.3d 543, 431 P.3d 1180.) We further held that "the extent of actual injury or lack of injury" is relevant, in that it may "suggest that the nature of the object or the way it was used was not capable of producing or likely to produce death or serious harm." ( Ibid. )
Against that backdrop, here is what the jury heard in Aledamat's trial: Aledamat approached a food truck in downtown Los Angeles and made crude remarks about the truck owner's wife. The owner, standing on the sidewalk, reacted in shock, and removed his apron. From approximately three or four feet away, Aledamat took a step back, pulled from his right pocket a box cutter, blade extended - how far it extended, how large it was, or whether it was locked in its casing, no one explained - and "thrust" or "pointed"
**295it from his waist towards the owner, saying, "I'll kill you." It was clear that Aledamat had moved his arm toward the truck owner. But there was no indication that Aledamat jabbed the box cutter at the owner, or that he flailed his arms around or advanced as though to cut him. Naturally, the owner jerked back. The jury heard nothing about what the owner was wearing or how close the box cutter actually got to his clothing or body. The box cutter did not touch him.
A police officer approached on horse and noticed Aledamat was holding the box cutter in his right hand "in a forward direction, about waistline." The officer said Aledamat was not lunging; he was "just holding" the box cutter. When Aledamat saw the officer, he retracted the box cutter and placed it back in his pocket. The truck *392owner reported to the police that Aledamat had "pulled a knife" on him.
On these facts, it may well have been reasonable for the jury to convict Aledamat of using the box cutter in a way "capable of causing and likely to cause death or great bodily injury," where "great bodily injury," in turn, is *25"significant or substantial." ( CALCRIM No. 875.) Aledamat did, after all, say, "I'll kill you" while wielding a sharp blade. But "our task in analyzing the prejudice from the instructional error is [to determine] whether any rational fact finder could have come to the opposite conclusion." ( Mil , supra, 53 Cal.4th at p. 418, 135 Cal.Rptr.3d 339, 266 P.3d 1030, first italics added.)
Even a brief survey of the evidence presented at trial reveals that the answer is yes. The People admitted in rebuttal that Aledamat had not "lunged" at the victim; he had "thrust" the box cutter out from his waist from a distance of several feet. (Cf. B.M. , supra , 6 Cal.5th at p. 536, 241 Cal.Rptr.3d 543, 431 P.3d 1180 [defendant used the knife only on the victim's legs, which were covered in a blanket, and did not attempt to use the knife on any exposed or vulnerable part of the victim's body].) The jury received no information about how far the blade was extended, whether the blade was locked - such that it would have stayed protruded had it made contact with the victim - or whether, at the time, it was spring-operated to snap back into its casing. (Cf. People v. Stutelberg (2018) 29 Cal.App.5th 314, 322, 240 Cal.Rptr.3d 156 [finding instructional error prejudicial in part because "[t]he exact manner in which Stutelberg used the box cutter against [a victim]" was "unclear"].) Nor did the jury hear testimony about what the victim was wearing, which has some bearing on whether a single thrust likely would cause serious bodily injury. (Cf. B.M. , at p. 536, 241 Cal.Rptr.3d 543, 431 P.3d 1180 ["[T]he moderate pressure that B.M. applied with the knife was insufficient to pierce the blanket, much less cause serious bodily injury to [the victim]."].) And there was no evidence to suggest Aledamat reared for another thrust, that he advanced on the victim after his initial threat, or that he wielded the box cutter uncontrollably. (Cf. id. at p. 538, 241 Cal.Rptr.3d 543, 431 P.3d 1180 [noting that the butter knife "was not applied to any vulnerable part of [the victim's] body" and there was "no evidence that B.M. wielded the knife wildly or uncontrollably"]; Stutelberg , at p. 322, 240 Cal.Rptr.3d 156 ["The jury could reasonably conclude that his 'flicking' motion was more of a threat, as opposed to an act likely to cause death or great bodily injury."].) Indeed, the truck owner refused to confirm his previous statement to the police, claiming instead not to know what he had said and repeating only that Aledamat "pulled out [the box cutter] when we were close to each other."
Only by casting aside this record evidence and supplanting it with its own reasoning can the majority justify its conclusion. It assumes the jury understood "inherently deadly" - the inapt instruction - to mean "something ," and assumes that something is a "common understanding" that a box cutter is deadly. (Maj. opn., ante , 251 Cal.Rptr.3d at p. 384, 447 P.3d at p. 288.) But it overlooks the reasonable possibility that the "something " the jury understood was not "capable of causing and likely to cause death or great bodily injury" ( CALCRIM No. 875 ), but instead likely to cut your child during ill-advised play. It assumes that finding the box cutter deadly "in the colloquial sense of the term" is sufficient because a box cutter is "readily capable of inflicting deadly harm."
**296(Maj.
*26opn., ante , 251 Cal.Rptr.3d at p. 384, 447 P.3d at p. 288.) But it both misstates the standard *393- which requires the likelihood of deadly harm - and misses the holdings of B.M. , which require a jury to look at how the weapon was actually used in context. ( B.M. , supra , 6 Cal.5th at p. 535, 241 Cal.Rptr.3d 543, 431 P.3d 1180.)
These assumptions aside, the evidence is sufficient for a reasonable jury to have found Aledamat guilty under the correct definition of a deadly weapon. To its credit, the majority concedes that's not the standard here. (Maj. opn., ante , 251 Cal.Rptr.3d at pp. 381-383, 447 P.3d at pp. 286-287; Chun , supra , 45 Cal.4th at p. 1201, 91 Cal.Rptr.3d 106, 203 P.3d 425 [stating that we reverse unless it is clear "beyond a reasonable doubt that the error did not contribute to the verdict"].) Yet, for whatever reason, that's the standard it ends up applying. (See, e.g., maj. opn., ante , 251 Cal.Rptr.3d at pp. 382-383, 447 P.3d at p. 287 [speculating that it is "unlikely" the jury would have viewed the instructions as presenting two alternative methods of finding the box cutter "deadly"]; id. at pp. 383-384, 447 P.3d at pp. 287-288 [arguing it would have been futile for defense counsel to contest deadliness because a box cutter is "potentially" deadly].)
III.
We do not undermine a defendant's Sixth Amendment right to a fair trial by jury if we hold an instructional error harmless where the record demonstrates that the jury actually relied on a different legal theory, untouched by error. ( Martinez , supra , 3 Cal.5th at p. 1226, 226 Cal.Rptr.3d 315, 407 P.3d 1 ; Chun , supra , 45 Cal.4th at p. 1205, 91 Cal.Rptr.3d 106, 203 P.3d 425.) Nor do we undermine that guarantee under the "unusual circumstances" in which "each element was undisputed, the defense was not prevented from contesting any of the omitted elements, and overwhelming evidence supports the omitted element." ( Mil , supra , 53 Cal.4th at p. 414, 135 Cal.Rptr.3d 339, 266 P.3d 1030.) In those limited cases, harmless-error review serves the useful purpose of preventing us from "setting aside convictions for small errors or defects that have little, if any, likelihood of having changed the result of the trial." (Ibid. , quoting Neder, supra , 527 U.S. at p. 19, 119 S.Ct. 1827, internal quotation marks omitted.)
This is not such a case, and today the majority dispenses with the guardrails that help us discern as much. Aledamat contested the element of force in the assault, which necessarily extends to the greater degree of force required to convict him of using a deadly weapon. The People presented little or no evidence that Aledamat used the box cutter in a way likely to cause death or great bodily injury, and further confused the jury by referring to the box cutter as "inherently deadly" and suggesting heuristics for assessing its dangerousness. (Cf. People v. Marsh (2019) 37 Cal.App.5th 474, 490, 249 Cal.Rptr.3d 749 [finding no prejudice where "the prosecutor only presented the [deadly-as-used] theory"].) On this record, I cannot conclude beyond a reasonable doubt that the jury verdict would have been the same absent the error. (See Neder , supra , 527 U.S. at p. 19, 119 S.Ct. 1827 ; Chun , supra , 45 Cal.4th at p. 1215, 91 Cal.Rptr.3d 106, 203 P.3d 425.)
*27No doubt we'll continue doing our utmost to tread carefully when deciding whether an error was harmless under the Chapman standard. But today the majority loses its footing. I dissent with respect.
I Concur:
GROBAN, J.