dissenting.
I agree with today's ruling that section .320 of the Official English Initiative is unconstitutional because it violates protected freedoms of speech. But I do not agree that this infirmity can be cured by severing the section's first sentence and giving the remaining language of the section-and the entire initiative as well-a new meaning that conflicts with its plain meaning as originally enacted. Nor do I agree that the court's decision to rewrite section .320 can justify its refusal to recognize and address the obvious overbreadth problems left unresolved in the initiative's remaining provisions. Because section .820 is the initiative's keystone provision, because its original purpose and meaning are plainly unconstitutional, and because no court has the power to redraft an invalid statute, I believe that the constitution requires us to strike the initiative in its entirety.
I. SALVAGING THE INITIATIVE
A. Narrowing Construction
As the court recognizes in Part IV.A.2.b of its opinion, the Official English Initiative (OET) was presented to the voters as an English-only law. The core provisions of the initiative are sections .820, "Scope," which describes when English must be used, and 340, "Exceptions," which specifies when languages other than English are allowed.1 Section .820 sweepingly extends the English-only requirement to all functions and actions performed by government officials and all written materials they prepare:
See,. 44.12.320. Seope. The English language is the language to be used by all public agencies in all government functions and actions. The English language shall be used in the preparation of all official public documents and records, including all documents officially compiled, published or recorded by the government.
Section .340 then describes eleven limited purposes for which government officials may use other languages "when necessary"; it also specifies that private citizens who address government officials may communicate in a language other than English, but only if their statements are translated into English:
See. 44.12.340. Exeeptions. (a) The government, as defined in AS 44.12.3830, may use a language other than English when necessary for the following purposes:
(1) to communicate health and safety information or when an emergency requires the use of a language other than English;
(2) to teach another language to students proficient in English;
(8) to teach English to students of limited English proficiency;
(4) to promote. international relations, trade, commerce, tourism or sporting events;
(5) to protect the constitutional and legal rights of criminal defendants;
(6) to serve the needs of the judicial system in civil and eriminal cases in compliance with court rules and orders;
(7) to investigate eriminal activity and protect the rights of crime victims;
*216(8) to the extent necessary to comply with federal law, including the Native American Languages Act;
(9) to attend or observe religious ceremonies;
(10) to use non-English terms of art, names, phrases, or expressions included as part of communications otherwise in English; and
(11) to communicate orally with constituents by elected public officials and their staffs, if the public official or staff member is already proficient in a language other than English.
(b) An individual may provide testimony or make a statement to the government in a language other than English, if the individual is not an officer or employee of the government, and if the testimony or statement is translated into English and included in the records of the government.
All members of the court, including myself, agree on the intended meaning and dominant purpose of these provisions. Today's opinion correctly rejects the argument of Alaskans for a Common Language, Inc. (ACL) that we should read the initiative leniently, so that it applies only to "formal" and "official" documents and records. As the opinion states in concluding that section .320 as a whole cannot be read leniently:
Because the meaning of the first sentence of AS 44.12.3820 appears plain and unambiguous, and because ACL has not offered sufficient evidence of contrary voter intent, we have no basis to find that the voters shared what ACL calls its "common sense" reading of the initiative. The first sentence of section .820 plainly mandates the use of English by government officers and employees in the performance of their jobs, whether communicating with English or non-Emglish speakers, exeept in specific cireumstances [enumerated in AS 44.12.340(2) ]. Accordingly, we reject ACL's contention that the plain language of the first sentence of AS 44.12.8320 permits the "unofficial" or "informal" use of languages other than English by state officials or employees in the performance of their duties.
All members of the court further agree that, so construed, the initiative's requirements impinge on constitutionally protected rights of free speech and are therefore invalid. Yet despite this understanding of section .320's "plain and unambiguous" meaning, the court proceeds to give the second sentence of section .320 precisely the meaning that it just rejected for the section as a whole.
Because the second sentence of section 320 refers to "official" documents and records, the court reasons, it is "capable of a narrow reading," which, in the court's view, plainly contemplates a permissible category of "informal, unofficial written documents" outside the reach of the OEI. The court also points out that the initiative's neutral ballot summary stated that "[sltate records must be in English"; in the court's view, because this wording fails to specify that "all records must be in English," it "at least suggest[(s] that those state records that are not official are not within the reach of the OET." Finally, the court finds nuanced meaning in subtle phrasing differences between section .820's two directives to use English: the first sentence directs that "[the English language is the language to be used," while the second directs that "[the English language shall be used." The court takes this difference in the two otherwise clear directives as showing "a permissive aspect" in the second sentence-an aspect "allowing the use of non-English languages in documents so long as English is also used."
Through the narrow opening created by these infinitesimal textual gaps, the court leaps immediately to the conclusion that, because doubtful meaning should be resolved in favor of constitutionality, it has a duty to adopt this artificially narrow reading of the second sentence. The court makes no effort to first determine whether this meaning is textually or contextually plausible Given the second sentence's newly declared meaning, the court leaps onee more to conclude that the entire initiative can be rescued from unconstitutionality by severing the first sentence of section .320 and allowing the second to stand-again making no attempt to ask first whether the severed statute it adopts makes any sense in light of the primary *217purpose and dominant intent underlying the initiative's original version.
This interpretive process results in a radically rewritten law that bears no realistic semblance to the version originally presented to and enacted by Alaska voters. By the court's own account, the initiative started out, and was sent to the voters, as a comprehensive and inflexible English-only law that covered the entire universe of government communications-spoken and written-and ranked among the nation's "most restrictively worded official-English law[s] to date." 2 Yet as revised by the court, this law has now morphed into a modest and permissive measure that welcomes the use of all languages in all government functions and actions, spoken and written, as long as the government makes sure to keep an English version of "official documents and records" (whatever the court might later define that phrase to mean). Put simply, a law originally meant to say "English only and always (except as nee-essary in a few specified situations)" now says "English sometimes but not always or only-and we can't tell yet exactly when."
In my view, this interpretation makes no sense, and its adoption violates settled principles governing statutory construction and severance of unconstitutional language. Our duty to construe statutes in a way that avoids a finding of unconstitutionality is firmly constrained, as the court notes, "by the constitutionally decreed separation of powers which prohibits this court from enacting legislation or redrafting defective statutes."3 Accordingly, we are authorized to use narrowing constructions as a way of avoiding unconstitutional results only where it is reasonable to do so.4 And we have recognized that a narrowing interpretation will be reasonable only if it can be adopted without doing violence to the manifest legislative intent of the statute at issue.5 To this end, in determining the reasonable meaning of a law, courts regularly look for guidance to the "fundamental canon of statutory interpretation that the words of a statute must be read in their context and with a view to their place in the overall statutory scheme."6 In other words, "we must not be guided by a single sentence or member of a sentence, but look to the provisions of the whole law, and to its object and policy."7
- In reflexively giving the second sentence of section .320 the narrowest meaning it could possibly have in the first sentence's absence, today's opinion completely forgets to apply these prudential rules by ignoring the second sentence's meaning in its original context-the meaning that sentence had as an integral part of section .320 as a whole as that provision appeared 'in its original form. ~The court's interpretation of the second sentence in isolation from the first is consequently unsound-both textually and contextually.
As a textual matter, there is simply no basis for the court's assumption that the second sentence's reference to "official" documents implies a permissible category of "informal, unofficial written documents" outside the reach of the OEI. The court's assumption ignores a simpler and more likely meaning of "official documents and records"-that is, all documents and records prepared or retained by government employees in performing their official functions and actions. Indeed, the restricted meaning proposed by the court quickly becomes implausible when carried over to other parts of the initiative. If we accept the court's proposed interpretation of the word "official," parallel logic would advise us to read section .310's broad declaration *218that "[the English language is the official language of the State of Alaska" 8 as implying the existence of an informal, unofficial language outside the reach of the OEI-a meaning that nobody has even thought to propose.
Moreover, section .820's second sentence incorporates other wording that rules out the court's proposed interpretation of "official" documents and records. Specifically, the second sentence requires English to be used not just in "all official public documents and records," as today's opinion suggests, but rather "in the preparation of all official public documents and records."9 Accordingly, the text of the second sentence evinces an unambiguous intent to extend its English-only requirement to all informal writings that precede the government's "formal" public documents and records.
The second sentence's surrounding context points to the same conclusion. As already noted, the first sentence of section .320 unequivocally extends the initiative's English-only requirement to all actions and functions performed by government officers and employees. Because writing performed in the course of government work falls within the meaning of government "functions and actions," the first sentence of section .320 would normally extend its English-only requirement to all writings produced in the course of government employment, regardless of the second sentence's presence. Given the first sentence's broad reach, the see-ond sentence can best be understood, not as a sentence that covers the act of writing, but rather as one that covers the government's preparation and retention of writings produced by or submitted to the government for public use.
While the two sentences may well overlap to a considerable extent when read in this way, the second sentence nonetheless served a valuable purpose in its original context-that is, as part of an initiative designed to apply a strict English-only requirement. By cementing the point that all writings produced by government functions and actions or submitted from other sources must always stay in English if they are to be kept as public records and documents, this sentence closes any potential gaps left open by the first. When construed in context with the original initiative as a whole, then, including its sibling first sentence, the second sentence originally meant to strengthen, not to dilute, the force of the first.
Indeed, the court's own expressed understanding of "the plain language of the first sentence of AS 44.12.320 conflicts with its proposed narrow reading of the section's see-ond sentence." The court properly concludes that section .320 is incapable of being read in its entirety to "permit{ ] the 'unofficial' or "informal use of languages other than English by state officials or employees in the performance of their duties." Given this conclusion, the court's proposal to read the see-ond sentence of the section as having precisely that meaning would make no sense in the sentence's original context. Thus, as now adopted, this narrow meaning does violence to the manifest original purpose of section 320.
A different contextual anomaly arises between the court's proposed narrow meaning of section .820 and the plain meaning of section .840. As we have seen, section .320 sets out the "seope" of the initiative's English-only requirement by defining the circumstances in which English must be used, while section .340 defines the permissible uses of other languages by adopting an exclusive list of specific "exceptions" that can be invoked only "when necessary." 10 Because section .340's exceptions encompass situations involving both written and spoken language, the court's proposed reading of seetion .820 would have a paradoxical effect on section .840: as a "permissive" provision "allowing the use of non-English languages in documents so long as English is also used," section .820's provisions governing writings would become broader than section .840's exceptions, thus turning the exceptions into restrictions.
*219In short, because the court's proposed reading of section .320's second sentence is textually implausible, contextually unreasonable, and does violence to the manifest legislative intent of section .820 and the initiative as a whole, that narrow meaning fails to offer a viable path for avoiding the provision's unconstitutionality. Other courts considering nearly identical English-only provisions in the only two other states where they were adopted or proposed have not hesitated to strike them as facially unconstitutional.11 Neither should we. As we have previously ruled on similar occasions, "at some point, it must be assumed that the legislature means what it says." 12
B. Severance
Separate problems arise from the court's attempt to invoke severance as a means to enact its recrafted version of section .320's second sentence. This court's authority to sever unconstitutional provisions from an act derives from the same source as its power to adopt narrowing constructions: the court's duty to uphold a statute as constitutional whenever the result is reasonably possible. As Sutherland explains, "[the courts recognize a duty to sustain an act whenever this may be done by proper construction, and extend the duty to include the obligation to uphold part of an act which is separable from other and repugnant provisions." 13
By the same token, the court's severance powers are restrained by the same constraints that fence its powers to adopt limiting constructions: "If a court finds a statute or portions of it unconstitutional, it has the power to strike it down or sever the invalid portion. It does not have the power to redraft the statute as that is the province of the legislature." 14 Thus, not only must the remaining portion of a severed statute be "valid as a law by itself," 15 but it must also "give effect to the apparent intention: of the legislature" that enacted the original provision.16 Just as a court must do when it considers adopting a narrowing construction, a court contemplating severance must initially determine that "severing the invalid portion will not do violence to the intent of the legislature." 17 If the court determines instead that "by sustaining only a part of a statute, the purpose of the act is changed or altered, the entire act is invalid." 18
In keeping with these principles, our severance cases have often observed that, for purposes of determining whether the legislature would have wanted to enact the remaining portion of a severed statute, the critical inquiry is whether the severed portion remains faithful to the "primary intent," 19 "dominant purpose,"20 "spirit," 21 or "primary goal" 22 of the entire act as originally enacted.23 We have emphasized that "[in *220the final analysis, a court must endeavor to fathom the legislative intent from all sources available to it." 24 This focus makes eminent sense because, unless the remaining provisions are faithful to the intended meaning of the original measure, they will amount to an impermissible judicial revisionrof the original law.25
Yet here, in deciding to sever the first sentence of section .320, reconstrue the see-ond, and leave the rest of the initiative intact, the court never once stops to consider the effects of its ruling on the initiative's original intent and dominant purpose, which openly conflict with the intent of the severed initiative's remaining provisions. After all, as the court itself acknowledges early on in its opinion, the original initiative meant to impose a uniquely stringent and all-encompassing English-only requirement on all government communication. In contrast, the revised law as it now stands freely allows government communication in any language for any purpose, as long as English versions of "official" records and documents are kept. Moreover, as the court admits, its interpretation of seetion .320 makes other provisions of the initiative "superfluous or hortatory'"-a classic sign of changed meaning and improper severance.
Instead of focusing on the initiative's original intent and purpose, the court's opinion seems to suggest that severance is justified because the newly interpreted second sentence of section .320 now has a constitutionally permissible meaning and still serves a useful purpose. But as I have already explained, the court minted its new interpretation of the second sentence after striking the first sentence, and without initially asking whether the interpretation did violence to the initiative's original intent and purpose; had the court stopped to ask, it would have found that its permissive reading of the second sentence conflicts with the basic purpose that the initiative was intended to serve-to impose a strict English-only requirement on all government speech. Because of this conflict, it is "bootstrapping" 26 for the court to use severance as a means to enable it to give the second sentence of section .820-and the redacted section as a whole-a new meaning that the original initiative never meant to enact.
Gottschalk v. State27 illustrates the proposition. There, the legislature had defined a criminal defamation law to include an over-broad culpable mental state requirement. On appeal, the state urged this court to sever the overbroad provision and construe a related provision of the statute as incorporating a constitutionally permissible mens rea requirement.28 Characterizing the state's argument as a request for "radical reconstruction" that asked us "to undertake a wholly inappropriate judicial activity amounting to judicial legislation," 29 we declined the request:
We recognize the rule of construction that where it is reasonably possible to do so, statutes should be construed in a manner consistent with constitutional requirements. Here, however, as in Campbell ..., we are not able to save the statute in question because in doing so we would be stepping over the line of interpretation and *221engaging in legislation."[30]
Today's opinion leans heavily on the initiative's express severance clause and the general severance preference appearing in the Alaska Statutes. The opinion seems to suggest that the electorate's expressed preference for severance somehow confers special powers on the court to rely on the severance doctrine when its use might otherwise be barred. But the court misunderstands the limited role of a severance clause. A general severance law like AS 01.10.0830 31 or a specific severability provision like AS 44.12.390 32 simply works to override the traditional presumption against severance by establishing a presumption in favor of severance.33 Because legislative powers are not delegable and belong to the legislature or the voters, severance clauses do not bestow courts with any substantive authority to sever. As Sutherland explains, "it should be kept in mind that the authority of a court to eliminate invalid elements of an act and yet sustain the valid elements is not derived from the legislature, but rather flows from powers inherent in the judiciary." 34 Nor do severance clauses establish a specific legislative intent as to particular statutory provisions; at most, they merely create a "slight" generalized preference that helps guide the court when more specific evidence concerning the legislature's actual intent is close:
It would seem that the soundest interpretation of this language [discussing preference clauses] is that, whereas a specific severability clause creates a slight presumption in favor of severability, a general clause creates an even weaker presumption. For all practical purposes, the difference between the two is negligible.[35]
In each case, then, courts must look to the totality of the evidence bearing on "the content of the measure and the cireumstances surrounding its proposal" 36 in order to decide the legislature's likely intent on the particular severance question at issue: the final analysis, a court must endeavor to fathom the legislative intent from all sources available to it." 37 And as previously mentioned, this endeavor typically begins by centering on the "primary intent" and "dominant purpose" of the original enactment.38
Today's opinion completely fails to undertake this endeavor. The closest it comes are its brief discussion of the "overarching" symbolic importance that an .official-English measure can have; its description of the fervent plea for severance advanced in ACT's supplemental brief; and its eryptic reference to the initiative's opening provision, AS 44.12.9300, which broadly finds that "English is the common unifying language of the State of Alaska and the United States of America," and then "declare[s] a compelling interest in *222promoting, preserving and strengthening its use." But the court's approach mistakes the initiative's abstract statement of hopes and aspirations for the concrete "content of the measure and the cireumstances surrounding its proposal." 39 The voters' likely intent in enacting the initiative must be realistically gauged by what the initiative's substantive provisions actually do, not by the aspirational goals the voters eventually hope to attain.40
The danger posed by the court's approach les in the inevitable temptation it creates to overreach the limits of judicial power by trying to redraft an initiative to mean something that the court believes the voters would have wanted to enact had they proposed a constitutionally valid measure. The court in effect converts ACL's sincere desire to do something constitutional along the initiative's general lines-its fervent wish to adopt some valid form of "official-English" measure if its "English-only" initiative would not pass muster-into a mandate to engage in judicial legislation. The institutional harm of sue-cumbing to this temptation is that it leads to public laws drafted and enacted by judges-a power that the Constitution does not confer on the judicial branch, but always allows the legislature and the voters to pursue for themselves.
This type of danger, and the need to avoid it, was recognized and aptly described in State v. Zarnke,41 a decision of the Wisconsin Supreme Court in a criminal appeal raising severance issues similar to the ones we considered in Gottschalk v. State.42 In Zarnke, the court addressed a challenge to a recently enacted criminal law that shifted the burden of proof from the state by requiring the defendant to prove the absence of guilty knowledge.43 The trial court had dismissed the state's charge against Zarnke, concluding that the law was unconstitutional under the legal theory charged in his case.44 In the intermediate court of appeals, the state conceded the law's unconstitutionality but convinced the intermediate court to sever the provision of the law imposing the unconstitutional burden and to construe the law's remaining provisions as incorporating the conventional requirement giving the state the burden of proving the element of guilty knowledge beyond a reasonable doubt.45
The Wisconsin Supreme Court granted Zarnke's petition to review the intermediate court's severance ruling.46 The state then renewed the severance argument it had raised below, but it advanced a new theory to support the argument, urging the court to rule that, even though the legislature had clearly intended to adopt the current law's allocation of burdens, it was nevertheless proper to sever the invalid provision and reinterpret the law's remaining provisions because the legislature's underlying intent had been to enact the most rigorous guilty-knowledge requirement that the constitution would permit.47
*223The supreme court reversed the intermediate court's severance ruling.48 Although it accepted the state's description of the legislature's underlying intentions, the court roundly rejected the state's new theory of severance and reinterpretation, explaining its ruling as follows:
At oral argument, the State suggested that the legislature's explicit intent as evinced by legislative history is not what appears to be most clear from a reading of that history. Instead, the State suggests that we should consider the legislature's implicit intent, which it believes was really an intent to enact legislation that would allow it to legislate to the limits of the constitution....
We might agree with the State that the legislature's implicit intent was to draft a statute that went to the limits of the constitution. However, that the legislature intends to pass statutes which are constitutional is always our starting point in such an inquiry as this. But were we to rewrite a statute whenever it failed constitutional muster in order to save it, using any means possible, the legislature would soon realize that it need not be concerned with constitutional limitations: the judiciary could always be relied upon to mend and mold its language to fit within constitutional constraints.[49]
Here, neither the initiative's severability clause nor ACL's commendable aspirations can dispel the conclusion that the weak and largely symbolic official-English law the court now adopts diverges radically from the restrictive and inflexible intentions manifested in the original initiative's deliberately sweeping and restrictive English-only requirements. As today's opinion makes clear, if the initiative's sponsors had wanted to propose a moderate and permissive official-English initiative, they would have had numerous examples to use as models-indeed, they still have those models and are free to propose them. Instead, the initiative's sponsors chose to propose a carefully crafted and elaborately structured measure that, the court concedes, clearly and unambiguously mirrored the nation's "most restrictively worded official-English law{s] to date." 50
Today the court claims to have transformed the original initiative from a divisive, zero-tolerance English-only mandate into a unifying and permissive symbol of our common linguistic bond; yet in the same breath, the court declares that it sees no real change in the original initiative's basic meaning and primary intent. The court's goal may be laudable, but in my view, its vision fails. As I see it, the court's action is judicial legislation, pure and simple. I would hold that the original initiative "means what it says." 51
II. OVERBREADTH
Having reshaped section .820 to fit constitutional limits, the court all but ignores the overbreadth claims that address the initiative's other provisions; the court dismissively finds that these claims pose no realistic danger of chilling free speech and can safely be left for later challenges on an as-applied basis. I think that the court is correct in recognizing that its reinterpretation of seetion .320 will ultimately render many of the initiative's remaining provisions "superfluous or hortatory.". But it seems wrong to assume that these provisions have been sufficiently disarmed to pose no lasting danger of a chilling effect. Because I would hold that the initiative must be struck in its entirety and because the court declines to address the remaining overbreadth claims at all, it would be pointless to discuss the claims in detail. But since I think that the court's resolution of the case should require it to address these claims, it seems appropriate to outline my reasons for believing that these claims warrant the court's immediate attention.
As a preliminary matter, I would note that to the extent that these claims have become hortatory and superfluous, their current status results from the court's decision that the *224first sentence of section .320 can properly be severed. Before severing an invalid provision and declaring that the remainder of the statute can stand, the court has a duty to review the statute's remaining provisions and to determine whether they are actually valid and capable of standing on their own.52 Because the court acknowledges that its ruling on section .820 renders other provisions of the initiative hortatory and superfluous, it should identify and strike those provisions as an integral part of its ruling on severance.
More importantly, in their present nebulous state, at least some of these provisions continue to pose the same obvious danger of chilling free speech that led the court to strike the first sentence of section .820. Like landmines left behind in the aftermath of a war, these provisions will continue to be a clear and present danger until they are identified, examined, and defused.
For example, in the absence of any further provision requiring the exclusive use of English, section .840(a)'s list of exceptions would at first blush seem meaningless and presumably harmless. Yet as long as they continue to stand as part of the literal law, these "exceptions" can hardly be counted on to be harmless. As already mentioned, even in section .820(a)'s absence, section .840(a) is capable of being read to stand on its own as an exclusive list of circumstances in which languages other than English may be used by the government. While we might be able to count on courts to understand that today's ruling makes these provisions superfluous, many members of the public-both within government service and outside of it-will not share that understanding and may conform their conduct to the literal terms of this statutory language.
Section .840(b) creates a separate problem by allowing members of the public to communicate with government officials in languages other than English, but only if their communications can be translated into English:
An individual may provide testimony or make a statement to the government in a language other than English, if the individual is not an officer or employee of the government, and if the testimony or statement is translated into English and included in the records of the government.
Though originally designed as a narrow exception that would override section .320(b)'s English-only requirement by allowing English to be used if a translator could be found, now this subsection will surely be read and applied as a stand-alone provision that requires non-English communications to be translated even though the rule that justified the exception has been struck. As a logical matter, the exception should disappear with the rule; yet today's opinion seems to recognize section .840 as continuing to exist.
Section .8350 vastly compounds the problem created by the holdover status of subsection .840(b):
See. 44.12.350. Public accountability. All costs related to the preparation, translation, printing, or recording of documents, records, brochures, pamphlets, flyers, or other material in languages other than English shall be defined as a separate line item in the budget of every governmental agency, department, or office.
Though styled as an "accountability" provision, this section actually erects an appropriation requirement. By demanding that the costs of preparing all non-English government writings be accounted for by "separate line item in the budget of every governmental agency," subsection .350(b) ultimately means that no government writings can be prepared in a language other than English unless line-item funds have been appropriat, ed for the preparation. Thus, although the court now recognizes that section .320 creates a permissible "category of unofficial or informal documents" that includes writings like a "note in Spanish from a teacher," "a letter from a city councillor in Yupik," or "a fisheries notice to be posted in English and Yupik," section .350 may well require all government workers, before sending or accepting such notes, letters, and notices, to locate the line-item money. And if the mon*225ey is not there, those writings would not be permitted. Although this provision certainly might have fit well with the initiative's original role as an English-only law, it has no legitimate justification in the reconstructed version and can only invite mischief.
The lingering concerns posed by these provisions are further exacerbated by the overarching language of section .380:
See. 44.12.8380. Private cause of action authorized. Any person may bring suit against any governmental entity to enforce the provisions of AS 44.12.300-44.12.390.
In effect, this provision gives every person in Alaska a wildeard to sue-or to threaten suit-for enforcement of the initiative's requirements.
Although this private-prosecution function might have played a vital role in ensuring that the original initiative's broad requirements were enforced, now that the initiative no longer has any hard and fast requirements, it seems fair to wonder what will be left to enforce. The question is hardly moot: there is no reason to expect that Alaskans who hold strong views favoring the adoption of English as our only official language will hesitate to sue, or to threaten suit, based on their personal impressions of what the initiative means and requires. In many cases, these impressions will now reflect mispercep-tions stemming from the initiative's unsettled meaning in its just-adopted form. Because the initiative's newly proclaimed meaning creates no obvious rights that would be capable of being privately enforced, the only suits likely to be threatened under section .380 are suits triggered by the problematic provisions described above. The only effect section .380 seems capable of producing, then, is a chilling effect. And since the threat of suit can be as effective in chilling free speech as the suit itself, courts are not likely to see any sign of this chilling effect.
We have previously held that "[a] statute regulating speech is overbroad, and thus unconstitutional, 'when constitutionally protected conduct as well as conduct which the state can legitimately regulate are included within the ambit of [a] statute's prohibition.' " 53 We have further held that a court must engage in an overbreadth analysis when an individual whose own speech may be prohibited challenges a statute on its face " 'because it also threatens others not before the court-those who desire to engage in legally protected expression but who may refrain from doing so rather than risk prosecution or undertake to have the law declared partially invalid." " 54 The challenger must show that there is "'a realistic danger that the statute itself will significantly compromise 'recognized First Amendment protections of parties not before the Court for it to be facially challenged on overbreadth grounds. " 55
In considering whether the combined ef-feet of the provisions discussed above cere-ates a realistic danger that free speech will be chilled, we must start by bearing in mind the importance of our right to speak freely and by recognizing that, when First Amendment freedoms are at stake, "[tlhe threat of sanctions may deter their exercise almost as potently as the actual application of sanctions." 56 We must likewise consider the importance of the particular speech rights threatened, the seope of the threat, the class of persons affected, and the likelihood that the presence of a chilling effect might remain undetected.
Here, the right threatened-the right to communicate with government-is among the most vital of potential free-speech rights; the threatened deprivation could result in a lack of access to government services; and the two classes at risk are non-English speaking citizens and government workers-both of which have a membership that is extremely vulnerable to deterrence by the *226threat and would be particularly unlikely to seek redress if their rights were improperly chilled.
Considering these factors, as well as the systemic nature of the problem, I think that there are compelling reasons to believe that a realistic danger of a chilling effect does in fact exist. I would therefore conclude that, unless the court strikes the initiative in its entirety, it must review and decide the remaining overbreadth claims now.
III. CONCLUSION
Because the initiative was carefully crafted in its entirety to prohibit or impermissibly chill the right to free speech, I would hold that it must be declared invalid in its entirety. Although I agree with the court's ruling that section .820 is unconstitutional, I dissent from its decision that the rest of the initiative can be saved.
. AS 44.12.320, .340.
. Ruiz v. Hull, 191 Ariz. 441, 957 P.2d 984, 994 (1998) (quoting Michele Arington, Note, EncusH-Laws anp Direct Leerstation: Taz Barzte In THE States Over Lancuace Minority Ricets, 7 J.L. & Por. 325, 337 (1991)).
. State v. Campbell, 536 P.2d 105, 111 (Alaska 1975), overruled on other grounds by Kimoktoak v. State, 584 P.2d 25, 31 (Alaska 1978).
. See Bonjour v. Bonjour, 592 P.2d 1233, 1237 (Alaska 1979) (citing Larson v. State, 564 P.2d 365, 372 (Alaska 1977)).
. State v. Blank, 90 P.3d 156, 162 (Alaska 2004); see also Bonjour, 592 P.2d at 1237.
. Davis v. Michigan Dep't of Treasury, 489 U.S. 803, 809, 109 S.Ct. 1500, 103 L.Ed.2d 891 (1989) (citing United States v. Morton, 467 U.S. 822, 828, 104 S.Ct. 2769, 81 L.Ed.2d 680 (1984)).
. Allied Chem. Workers v. Pittsburgh Plate Glass Co., 404 U.S. 157, 185, 92 S.Ct. 383, 30 L.Ed.2d 341 (1971) (citations omitted).
. AS 44.12.310 (emphasis added).
. AS 44.12.320 (emphasis added).
. AS 44.12.340(a).
. See Yniguez v. Arizonans for Official English, 69 F.3d 920, 936 (9th Cir.1995) (en banc), vacated as moot sub nom. Arizonans for Official English v. Arizona, 520 U.S. 43, 48-49, 117 S.Ct. 1055, 137 L.Ed.2d 170 (1997); In re Initiative Petition No. 366, 46 P.3d 123, 129 (Okla.2002).
. Gottschalk v. State, 575 P.2d 289, 296 (Alaska 1978) (quoting Campbell, 536 P.2d at 111).
. Norman J. Statutes amp Statutory Construction § 44:1, at 549 (6th ed., rev.2001).
. Id. § 44:1, at 549-50 (footnote omitted); see also Gottschalk, 575 P.2d at 296.
. Singer, supra note 13, § 44:4, at 561-62.
. Duryee v. United States Dep't of the Treasury, 6 F.Supp.2d 700, 706 (S.D.Ohio 1995); see also Alaska Airlines, Inc. v. Brock, 480 U.S. 678, 685, 107 S.Ct. 1476, 94 L.Ed.2d 661 (1987) (court has duty to ensure the severed statute will "function in a manner consistent with the intent" of the legislature) (emphasis in original).
. Singer, supra note 13, § 44:3, at 554.
. Id. § 44:7, at 583.
. Lynden Transp., Inc. v. State, 532 P.2d 700, 715 (Alaska 1975).
. Id. at 719 (Rabinowitz, C.J., concurring in part, dissenting in part).
. McAlpine v. Univ. of Alaska, 762 P.2d 81, 94 (Alaska 1988); see also Singer, supra note 13, § 44:3, at 558-59 (Likewise the so-called 'spirit' of an enactment is a term meaning essentially the legislative intent, although it may be somewhat broader, involving also the 'purpose' of the enactment." (footnotes omitted)).
. McAlpine, 762 P.2d at 95.
. Conversely, when approving severance we have sometimes emphasized that the "deleted subsection is a minor part of the overall act." *220Sonneman v. Hickel, 836 P.2d 936, 941 (Alaska 1992).
. Lynden Transp., 532 P.2d at 715.
. See, e.g., Women's Emergency Network v. Bush, 323 F.3d 937, 948-49 (11th Cir.2003) (holding that the unconstitutional portion of a Florida law authorizing the state's "Choose Life" special license plate program-which provided that collected fees would be distributed to organizations that promoted adoption over abortion-was not severable because the legislative purpose of the program was to promote adoption instead of abortion}.
. See Campbell, 536 P.2d at 111 ("We cannot 'bootstrap' the wrongful intent requirement into the statute by the wholesale implication of other necessary elements. At some point, it must be assumed that the legislature means what it says and that, in fact, in this instance, it was trying to delineate a new statutory offense. There is nothing, of course, inherently unconstitutional about purposeful omission of [a particular] requirement.").
. Gottschalk, 575 P.2d at 294-95.
. Id. at 296.
. Id.
30. Id.; cf. State v. Zarnke, 224 Wis.2d 116, 589 N.W.2d 370, 377 (1999) ("While when necessary, we have at times severed portions of a statute's language, and at other times have read into a deficient statute a constitutional requirement, the State's request that we save all of [the challenged law] would require this court to combine two distinct saving doctrines, which we, are not inclined to do under the circumstances of this case.").
. Alaska's general severance 01.10.030, provides as follows: statute, AS
Any law heretofore or hereafter enacted by the Alaska legislature which lacks a severability clause shall be construed as though it contained the clause in the following language: "If any provision of this Act, or the application thereof to any person or circumstance is held invalid, the remainder of this Act and the application to other persons or circumstances shall not be affected thereby."
. The initiative's 44.12.390, states: severability clause, AS
The provisions of AS 44.12.300-44.12.390 are independent and severable, and if any provision of AS 44.12.300-44.12.390, or the applicability of any provision to any person or circumstance, shall be held to be invalid by a court of competent jurisdiction, the remainder of AS 44.12.300-44.12.390 shall not be affected and shall be given effect to the fullest extent practicable.
. Lynden Transp., 532 P.2d at 711-12.
. Singer, supra note 13, § 44:8, at 589.
35. Lynden Transp., 532 P.2d at 712-13.
. McAlpine, 762 P.2d at 94-95.
. Lynden Transp., 532 P.2d at 715.
. Id. at 715, 719 (Rabinowitz, C.J., concurring in part, dissenting in part). '
. McAlpine, 762 P.2d at 94-95.
. The court protests that, "[in truth, the court simply strikes the first sentence of section .320 and gives to the second sentence of that section its plain meaning." This protest rings hollow because it misses the point: as the court tacitly acknowledges in declining to give the entirety of the original version of section .320 the same permissive meaning it now attributes to the section's second sentence, the second sentence's "'plain meaning" in the severed version of the section is a meaning that the sentence could not plausibly have been intended to have in its original context. The court's "sever and reinterpret" approach is precisely the bootstrapping that Gottschalk and similar cases forbid. In effect, the court starts with a law that says "No language but English may ever be used." The court severs some text: "NMe-language English may ever be used." It then gives the remainder-"English may be used"-its plain meaning as severed and declares that, because the remainder is capable of standing on its own and-just like the original version-helps to promote English, the severed version has essentially the same purpose as the first and was therefore intended by the original drafters. This strikes me as a considerable reach.
. Zarnke, 589 N.W.2d at 377.
. Gottschalk, 575 P.2d at 296.
. Zarnke, 589 N.W.2d at 372.
. Id.
. Id. at 372-73.
, Id. at 373.
. Id. at 379.
. Id. at 373.
49. Id. at 379 (emphasis in original) (citation omitted).
. Ruiz, 957 P.2d at 994 (quoting Arington, supra note 2, at 337).
. Campbell, 536 P.2d at 111, overruled on other grounds by Kimoktoak, 584 P.2d at 31.
. See, eg., Singer, supra note 13, § 44:4, at 561-62 (remaining portion of severed statute must be "valid as a law by itself"); McAlpine, 762 P.2d at 94-95 (remaining portion of severed statute must be capable of standing on its own).
. Turney v. State, 936 P.2d 533, 539 (Alaska 1997) (second alteration in original) (quoting Marks v. City of Anchorage, 500 P.2d 644, 646 (Alaska 1972)).
. Bd. of Airport Comm'rs v. Jews for Jesus, Inc., 482 U.S. 569, 574, 107 S.Ct. 2568, 96 L.Ed.2d 500 (1987) (quoting Brockett v. Spokane Arcades, Inc., 472 U.S. 491, 503, 105 S.Ct. 2794, 86 L.Ed.2d 394 (1985)).
. Id. (quoting City Council of Los Angeles v. Taxpayers for Vincent, 466 U.S. 789, 801, 104 S.Ct. 2118, 80 L.Ed.2d 772 (1984)).
. Marks, 500 P.2d at 647 (citation omitted).