concurring in part and dissenting in part. Part I is joined by Judge WARDLAW and Judge BERZON. Part II is joined by Judge SCHROEDER:
Writing and mailing off to the media a largely incomprehensible diatribe with *1304vague allusions to evil acts that the author intended to have accomplished by the time his Manifesto was received was the unwise act of a troubled mind. It was not, however, criminal behavior that satisfies either of the two elements required for a conviction under 18 U.S.C. § 876(c). I concur with the majority’s reversal of Havelock’s criminal conviction and with its holding that a communication must be addressed to a natural person. Unfortunately, the majority’s subsequent discussion, which concludes that the identity of the addressee can be discovered anywhere within the body of the document, renders irrelevant the holding on which its decision is based, and totally eliminates one of the two essential elements required by the statute.
The statutory text clearly sets forth two elements. Specifically, it requires that 1) the communication be addressed to a person, and 2) the communication contain a threat to a person. By permitting courts to look to the entire document in order to identify an “addressee,” the majority ignores the context of the statute: the term addressee is used in relation to the posting of mail at the post office or in a mail box. The opinion holds, in effect, that whenever a threat against a person is contained in the body of the communication, the subject or object of the threat may also serve as the addressee. This makes little sense with respect to the particular statute before us. I therefore write separately to express my disagreement with Section II.B of the majority’s opinion.
Before I do so, however, I explain why the confusion and inconsistency created by the majority opinion (which I nevertheless concur in, in part) is unnecessary. There is a simpler and clearer basis for reversing Havelock’s conviction. Beside the majority’s inability to locate an “addressee” buried within the content of his Manifesto, there is another more substantial reason why Havelock’s ramblings do not violate 18 U.S.C. § 876(c). That reason is that Havelock’s Manifesto does not contain a threat. Even if the Manifesto could be construed as referring to an identifiable group of persons attending a particular event, which I seriously doubt, it was intended to arrive after the occurrence of that event. A threat can refer only to a future, not a past, act; a threat is an act that is intended to put its subjects or objects in fear of an event to occur in the future. Put simply, a communication sent when and under the circumstances Havelock’s was cannot induce fear of an impending event, and is therefore not a threat.
In sum, while the majority decides the case in a confused and incorrect manner based on the first element, finding that the communication lacked a natural person addressee, I would decide this case on the more basic question presented by the second element, and hold that the Manifesto did not contain a threat.1
I
The assorted writings mailed to the six news and music outlets which arguably refer to an event that Havelock expected to have taken place before the communica*1305tions were received — the Superbowl — did not meet the legal requirements for a threat, and were thus not punishable by § 876(c).
A threat must “communicate a serious expression of an intent to commit an act of unlawful violence to a particular individual or group of individuals.” Virginia v. Black, 538 U.S. 343, 359, 123 S.Ct. 1536, 155 L.Ed.2d 535 (2003); see also United States v. Stewart, 420 F.3d 1007, 1017 (9th Cir.2005). Havelock’s rambling Manifesto included statements ranging from excoriations against the economic and political system to quotations from popular songs and films to critiques of video games. Among his pronouncements was the declaration that “I will sacrifice your children upon the altar of your excess, and know it has been your own greed, your loftiness, your disdain and mistreatment of your fellow man that has done this.” The Manifesto also included the profession that “I will shed the blood of the innocent. For the blood of the guilty will always remain unclean. And other dregs of society will always quickly lap up the blood of the unclean, take their place, and carry on as if nothing happened. Meet the new boss, same as the old boss.” Included within the Manifesto was also the nonsensical explanation, “Oh yeah, Cthulhu made me kill! la! la! Cthulhu Ftaghn!”
The Manifesto clearly contains statements alluding to acts of violence, from running people over with a Hummer to killing the children of his unidentified perceived oppressor. None of this language identifies a person or group of persons who might constitute the objects or subjects of any threat. In the final lines of the Manifesto, preceding quotations from the Old Testament and the punk rock group Bad Religion, is the only oblique reference to the Superbowl and those in attendance. I do not believe that, by its reference to a t-shirt containing the message “I survived the Superbowl,” the Manifesto sufficiently identifies an intent on the part of Havelock to commit a violent act regarding that event. Because some of my colleagues appear to disagree, however, I explain here why even if it did, we could not hold that the Manifesto contains a threat. In any event, it is clear that no other part of the Manifesto identifies any particular individual or group of individuals as the intended objects of a threat or identifies any time, place, or event at which any act of violence will occur.
A threat is “an expression of an intention to inflict evil, injury, or damage on another,” Planned Parenthood of Columbia v. American Coalition of Life Activists, 290 .F.3d 1058 (9th Cir.2002) (en banc), or an “indication of impending danger or harm.” United States v. Davila, 461 F.3d 298, 302 (2nd Cir.2006). Threats are by definition prospective. See Webster’s New World Dictionary 1394 (3rd Coll. ed. 1988) (defining a “threat” as “an expression of intention to hurt”). They may include announcements of future or impending action, but not statements regarding past events or retrospective harm. As the Supreme Court has recognized, the prohibition on “threats protects] individuals from the fear of violence and from the disruption that fear engenders,” Black, 538 U.S. at 360, 123 S.Ct. 1536 (alteration in original) (internal citation omitted). Havelock’s statements were intended to be, and could only have been, understood by the recipients as a post hoc confession of his already completed actions. Having deposited the Manifesto in a United States Postal Service mailbox on the very day of the Superbowl, a mere thirty minutes before his intended rampage at the stadium, Ha*1306velock’s Manifesto could not have reached its intended destinations in time to put the recipients in fear of imminent danger or to constitute a threat of future action. To the contrary, the mailings sent by Havelock containing his Manifesto could only have arrived after the events alluded to within its pages had already taken place or after any possibility of committing the acts no longer existed.
When determining whether a defendant’s speech represents a threat our analysis is “not confine[d] ... to the defendant’s statements alone,” United States v. Bagdasarian, 652 F.3d 1113, 1123 (9th Cir.2011), but must consider the context in which the communication was made as well. Considering the context, it would be impossible to conclude that Havelock’s often-rambling statements constituted a threat, no matter how troubling those statements may have been. It is clear from the fact that none of the statements could have been read by anyone until after the only event referred to in that document that could arguably be the subject of a threat had already taken place that the Manifesto lacked any capacity to threaten any injury or express any intent to cause the victims future harm or fear of future harm. In short: under all the circumstances, and in context, the language of the Manifesto could not constitute a threat.
Finally, Havelock’s related writings make it clear that he intended the media packets and letters mailed the day of the Superbowl to be examined and understood in the wake of the already completed actions that he did not expect to survive. It is therefore not possible that he could have intended to put the recipients in fear of any future actions on his part because, as Havelock acknowledged within his letters, he would be dead by the time that the communications reached their various destinations. Section 876(c) is a specific intent crime, and therefore requires a “subjective, specific intent to threaten.” United States v. Twine, 853 F.2d 676 (9th Cir.1988); See also Black, 538 U.S. at 359-60, 123 S.Ct. 1536; Bagdasarian, 652 F.3d at 1118 (9th Cir.2011). Havelock’s statements and the circumstances surrounding the mailing of the copies of his Manifesto clearly demonstrate that he intended them to be read after his imagined rampage had occurred and he was deceased.
Among the evidence that Havelock intended his writings to be received after his demise was a letter mailed to the Office of Personnel Management at the same time as the media packets, in which he stated “[b]y the time this letter reaches you, I will probably be deceased or no longer able to sign any further needed paperwork,” and directed them to transfer his retirement funds to the mother of his children. Consistent with his later statement that he planned to commit “suicide by cop,” Havelock had with him in his car a letter to the police on which he had handwritten the command “DO NOT RESUSCITATE.” In a letter mailed to his parents, he asked that his body be cremated and hoped that his “tragic end” would be eye opening as he planned to “go onto [sic] a better place.” Indeed, the Manifesto itself states near its conclusion: “I’m going to talk to God.” These and other statements clearly demonstrate that Havelock did not intend to survive the events of Superbowl Sunday, and any unlawful acts referred to in the letters sent on that day were not intended to put its recipients in fear of future actions on his part. Nor, in context of when they were received, could they have done so. In addition to the sheer impossibility that a Manifesto received after Superbowl Sunday could instill *1307in its recipients a fear of harm that would result from events that would already have occurred, or could no longer occur, the lack of any intent to cause an apprehension of future harm — shown from both words and context — precludes a finding that the Manifesto was, in fact, a threat.
For the reasons explained above, I would hold that the Manifesto did not contain a threat and would on that ground reverse Havelock’s conviction for violating 18 U.S.C. § 876(c).
II
Returning to the ground on which the majority decides to reverse Havelock’s conviction, I start with the language of the statute.
Whoever knowingly so deposits or causes to be delivered as aforesaid, any communication ... addressed to any other person and containing any threat to kidnap any person or any threat to injure the person of the addressee or of another, shall be fined under this title or imprisoned not more than five years, or both.
18 U.S.C. § 876(c) (emphasis added).
The “as aforesaid” refers to subsection (a), which states “[wjhoever knowingly deposits in any post office or authorized depository for mail matter, to be sent or delivered by the Postal Service or knowingly causes to be delivered by the Postal Service according to the direction thereon.” § 876(a). The subsection at issue here explicitly punishes the mailing of any threatening communication in the prescribed manner if it is “addressed to any other person,” and “contain[sj any threat” to a person. § 876(c).
The language of the statute unmistakably distinguishes between the addressee portion of the communication and the communication’s content. The majority initially recognizes appropriately that “person,” in both sections of the statute, must be a natural person. However, by subsequently concluding that a court may look to the content of the communication to determine whom it is “addressed to” the majority renders its holding without substance and effectively excises the “addressed to” requirement from the statute. Under the majority’s interpretation, the “addressed to” element becomes superfluous because once a threat to a person is identified within the content of the letter the communication will necessarily contain a subject or object of the threat, and that subject or object will necessarily become the addressee of the communication. If there is no such subject or object of the threat, then whether or not there is an addressee of the communication who is a natural person is irrelevant because the communication fails to violate the statute due to the absence of a threat. Indeed, in the instant case the majority holds that there is no addressee because it fails to identify a target of any threat within Havelock’s rambling Manifesto, and without such a target there can be no threat. Thus, the significant number of pages that the majority devotes to explaining why § 876(c) is applicable only when the addressee is a natural person is of no significance under its rule that the identity of that natural person may be located anywhere within the body of the communication. Either there is such a person identified in the body of the communication and there is both an addressee and a threat, or there is no person so identified and thus no addressee and no threat. By failing to preserve the statutory distinction between the two elements, the addressee and the contents of the communication, the majority has re-written the *1308statute, eliminating one of the two necessary elements.
The majority begins its analysis by recognizing two potential definitions of the term “address.” Neither definition supports the majority’s ultimate reading of the statute, but even the broader definition of “to address,” meaning “to speak, write, or otherwise communicate directly to,” Maj. Op. at 1293, does not control the question of who is the addressee of a letter deposited in the mail. Section 876, adopted in 1932, shortly after the kidnapping of the Lindbergh baby, is a statute that punishes individuals who mail various forms of threatening communications addressed to identifiable persons by means of the postal system. In subsection (a), the statute clearly details the mailing requirements that are necessary to bring a communication under the province of the statute. These requirements are incorporated into all of the subsequent subsections. The issue here involves the meaning of the statutory use of the word “addressee,” which is a term used to designate the person to whom the Post Office is to deliver an envelope mailed through its system. Keeping in mind that this is a statute intended to punish the sending of certain communications through the postal system, the only reasonable interpretation of the term “addressee” is the party named on the outer envelope to whom the Post Office would deliver a communication that has been mailed pursuant to the terms described within the statute. Even if the court were permitted to look beyond the outer envelope to the salutation line of a communication to determine the addressee, that would clearly be the outer limit under any definition of the words of the statute. To go further, as the majority does here, and to examine the entire contents of the communication, in order to seek the identities of those who may be the potential targets of the threatening communication and to label them the addressees, distorts the language of the statute beyond its plain meaning and intention as well as eliminates the first of its two requirements, that there be an addressee as well as a threat to injure any person.
By the majority’s own definition, to “address” a letter to someone means to “communicate directly to” that individual. One may “communicate directly to” another through the mail by placing the intended recipient’s name and address on the outer envelope and causing the post office to deliver that letter to the person so identified. One may arguably in some limited circumstances, even mail a letter to a corporate entity or the office of an official and designate a natural person as the intended recipient in the salutation line, although I need not decide that question here. Either of these methods may conceivably be viewed as a valid attempt to “communicate directly to” another person through the mail. In any ordinary usage of the term, however, one would not be understood as having “communicated directly to” some individual whose name is mentioned or discussed only within the body of a document that is formally addressed to another — or addressed to no one at all. One fails to “communicate directly to” another person when one mails a letter to an artificial entity, fails to address that person in the salutation, and mentions his name only in the body of the letter as an individual whom one intends to harm. This manner of communication is by no means “direct” and any individual identified only by examining the contents of such a communication cannot be referred to as the “addressee.” An individual so identified may be the *1309object of the threat; he may be the subject of the threat; but an individual not intended to be the recipient of the letter cannot, in any way, be termed the addressee of the communication. That person may properly be understood to satisfy the second element of the statute, but he cannot satisfy the first.2
To assist it in arriving at its untenable conclusion that the addressee may be contained in the body of the letter, the majority selectively adopts the Fourth Circuit’s reasoning in United States v. Rendelman, 641 F.3d 36 (4th Cir.2011). In Rendelman, the court considered whether the sending of two letters, whose outer envelopes were each addressed to the U.S. Marshals Service and which contained a threat to harm the President of the United States, was a punishable act under § 876(c) and subject to the ten year enhancement as a “communication addressed to an official covered by 18 U.S.C. § 1114.”3 Id. The Rendelman court recognized that the defendant’s letters were “addressed to” the Marshals Service, because they specified the Marshals Service on the outer envelope. The court concluded that the defendant’s mailings satisfied the “person” requirement of § 876(c) by holding that a communication addressed to the Marshals Service “can reasonably be understood as addressed to the United States Marshal himself — a natural person.” Id. at 46. While we need not adopt the Fourth Circuit’s conclusion as to the natural person status of the Marshals Service, it is plain that the court identified a “person” as the addressee based on the name supplied on the outer envelopes of the charged communications, and that it did not identify the addressee through investigating the contents of the mailing itself. This holding, therefore, does not provide support for the majority’s conclusion that it may scour the contents of a communication to identify the natural person to whom it is addressed.
After holding that the Marshals Service satisfied the statutory requirement that the addressee be a “person,” the Fourth Circuit went on to consider an entirely different subject: the question of enhancement. It held that the defendant was subject to enhanced penalties because “the threat contained in the ... [l]etter was sufficiently alleged as being ‘addressed to’, i.e., ‘directed to,’ the President ... even though the letter was not mailed to [him].” Id. at 47.4 The majority relies on this *1310statement to support its conclusion that a court may look to the contents of a communication to identify an addressee. What the Rendelman court identified within the text of the defendant’s letters, however, was the object of the threat, not the addressee of the communication. It is of course necessary to look to the content of the communication to find if the communication contains a threat, and if so, to whom or to what that threat is directed.5 We are not limited to the outside of the envelope or the salutation line of the letter to determine if there is a threat, and a threat to a person “contained” in a communication can certainly be found only in the content of the communication. This truism is of no relevance, however, to the majority’s conclusion that it may explore the contents of the communication in search of the addressee of the communication in which the threat is contained. In contrast to the majority’s decision, the Rendelman court looked to the content of the communication to identify the object of the sender’s threat, not the identity of the addressee of the communication. The majority selectively invokes the language of the Rendelman decision to assist it in reaching its desired conclusion, but Rendelman, to the extent that the court interpreted the portion of § 876(c) that is at issue in this case, simply does not support the majority’s holding.6
*1311The majority opinion also cites Rendelman as support for its conclusion that the contents of a letter are fair game in determining who the writing is “addressed to” because “[a]t its essence, § 876(c) criminalizes the use of the postal system to deliver a threatening communication.” Maj. Op. at 1302 (quoting Rendelman, 641 F.3d at 48). Unfortunately we are not asked to divine a statute’s essence but to interpret its text. The text of § 876(c) does not merely criminalize the use of the United States Postal Service to deliver threatening communications, but requires that such communications be deposited in the mail “addressed to any other person” (other than the sender) before such acts are subject to criminal penalties. § 876(c). In Section II.A of its opinion the majority appears to recognize this, but in Section II.B, perhaps for practical reasons at which we are left to guess, it goes off on an entirely different tack. The erroneous reading of the statute offered by the majority in Section II.B allows for the criminalization of a wider array of bad acts consistent with their view of what the statute should have achieved, but their reading is well beyond the scope of conduct that is actually circumscribed by the statutory text.
In its awkward attempt to explain the “true meaning” of the statute, the majority violates the “cardinal principle of statutory construction ... to give effect, if possible, to every clause and word of a statute,” United States v. Menasche, 348 U.S. 528, 75 S.Ct. 513, 520, 99 L.Ed. 615 (1955) (internal quotations and citations omitted). See also Duncan v. Walker, 533 U.S. 167, 121 S.Ct. 2120, 2125, 150 L.Ed.2d 251 (2001). The Rendelman court and today’s majority choose to ignore this axiom because it does not square with their own view as to how the statute should have been written. Once they have deemed its true purpose to be to punish the use of the postal system for the delivery of all threatening communications, any reading of the statute that falls short of this purpose is unacceptable. In subverting the language of the statute to meet the ends it desires, the majority ignores not only the command that we give all words meaning, but also the central principle in the interpretation of penal statutes that requires us to interpret criminal laws narrowly. See Regents of Univ. of California v. Public Emp’t Relations Bd., 485 U.S. 589, 604, 108 S.Ct. 1404, 99 L.Ed.2d 664 (1988) (Stevens, J., dissenting); United States v. Bass, 404 U.S. 336, 347-48, 92 S.Ct. 515, 30 L.Ed.2d 488 (1971). Today’s opinion violates both commands by ignoring the statutory language and re-writing the statute to encompass a broader range of conduct than is criminalized under the clear language of the statute. If the majority had grounded *1312its analysis on the recognition that “[w]e are not at liberty to rewrite the statute to reflect a meaning we deem more desirable .... [but rather] must give effect to the text [C]ongress enacted,” Ali v. Fed. Bureau of Prisons, 552 U.S. 214, 128 S.Ct. 831, 841, 169 L.Ed.2d 680 (2008), it would have given the “addressed to” element the meaning it deserves.
For the above reasons I concur in part and dissent in part with the majority opinion; but concur only because a majority of this en banc court fails to adopt what I believe to be the preferable ground for reversing Havelock’s conviction.
. Havelock also argued that his Manifesto was political speech and as such it deserves First Amendment protection. As the Supreme Court has recognized, constitutional protections afforded to speech do not extend to "true threats,” which have been defined as threats communicated by the speaker with the specific intent to threaten. Virginia v. Black, 538 U.S. 343, 123 S.Ct. 1536, 155 L.Ed.2d 535 (2003). Because Havelock’s writings were not a threat in any sense of the word, I need not reach the constitutional question he raises.
.The statute contemplates the possibility that a communication may be addressed to one person and contain a threat to another by criminalizing threats to “injure the person of the addressee or of another." § 876(c). It is therefore not necessary that the subject or object of the threat also be the addressee of the communication, although if they were one and the same the conduct would be penalized in the same manner. The threat may be directed to the addressee or to another, but in either case the communication itself must be delivered through the mail to a natural person who is the addressee. It is necessary under the clear language of the statute that the communication contain the name of an addressee to whom it is mailed through the postal system as well as a threat to a person, and we cannot merely satisfy ourselves, as the majority does, with the presence of the latter element and not the former.
. Section 1114 specifically identifies “any officer or employee of the United States or of any agency in any branch of the United States Government (including any member of the uniformed services).” 18 U.S.C. § 1114.
. This statement in Rendelman, as well as its holding on the enhancement element, was achieved through a disingenuous substitution of statutory language. Section 876(c) explicitly requires that the communication not the threat be addressed to a person. By substituting one term — threat—for another — communication — the Rendelman court partially altered *1310the meaning of the statute and eliminated, for purposes of enhancement, the need for an addressee that would satisfy the requirement that the communication be addressed to an official covered by § 1114. The Rendelman court bifurcated the statute as it relates to the enhancement element by contending that the second use of the phrase “addressed to” in § 876(c) means something different than the first. This is plainly inconsistent with the language and structure of the statute, in which the requisite manner of mailing is fully articulated only once, in § 876(a), and then incorporated by reference in the subsequent sections. In practical terms, read properly, the statute provides for an enhancement if the letter containing the threat is addressed to a designated official and contains a threat to any person. As Rendelman reads the enhancement provision, it applies if the letter is addressed to any person but the threat contained in the letter is made to a designated official. All of this is irrelevant, however, as the Fourth Circuit held in the part of its opinion deciding when a statutory offense is committed that an addressee is a person who could be identified by the name on the outside of the envelope. That is directly contrary to the proposition for which the majority cites the case.
. It is important to note that § 876(c) and (d) provide for enhanced punishment based on the addressee of the communication, not the target of the threat. As a result, a perpetrator could have his punishment enhanced for addressing a letter to a federal judge threatening to harm the judge or the judge’s family, but would not receive an enhancement if he addressed a letter to the judge's family and threatened to harm the judge. This result may seem anomalous, and may even tempt a court, such as the Fourth Circuit, to substitute the term “threat” for “addressee” in order to achieve a result it believes better reflects the types of harms it feels merit greater punishment. The punishment structure of the statute makes complete sense, however, if we accept that Congress may have been more concerned with the harm that would result from judges or other federal officials being unduly influenced by threats to their loved ones (or to themselves) than it was with the possibility that those unaffiliated with the operation of the federal government may receive threatening communications that expressed an intent to harm a federal official. Moreover, when Congress was concerned with threats of certain types of harm inflicted against certain categories of people, it explicitly identified such threats. See, e.g., § 876(d). Therefore, when Congress failed to impose punishment based on the nature of the threat we must assume that it meant what it wrote and intended punishment based on the addressee rather than the object of the threat.
. The majority also cites to United States v. Williams, 376 F.3d 1048 (10th Cir.2004), for its conclusion that you may look to the contents of a letter to identify the addressee, but *1311this reliance is also in error. In Williams, the Tenth Circuit held that one may look to the salutation line as well as the envelope to identify the addressee, a question I would leave open here. The Tenth Circuit, as cited by the majority, Maj. Op. 1294, went on to state the uncontested fact that "[t]he word 'communication' includes the contents of a letter.” This statement of fact does not at all speak to the propriety of looking to the contents of a letter to identify the addressee, and cannot reasonably be used to support the majority’s ultimate conclusion that an addressee may be located anywhere within the pages of the communication. The Second Circuit in United States v. Davila, 461 F.3d 298 (2nd Cir.2006), held the defendant's mailing punishable under § 876(c), but in doing so, the court looked only to its outer envelope to identify the addressee. Therefore, despite the majority’s citations to other circuits that have purportedly adopted the same “commonsense” understanding that we may look to the contents of the communication to identify an addressee, the Ninth Circuit now stands alone in so holding.