Justice BENDER
delivered the Opinion of the Court.
In this case, the People appeal a decision by the Mesa County District Court dismissing charges against Glen Hickman. The People charged Hickman with one count of Retaliation Against a Witness or Victim, in violation of section 18-8-706, 6 C.R.S. (1998). Ruling that the statute was unconstitutionally overbroad and vague, the trial court dismissed the charge. We have jurisdiction to consider this appeal pursuant to section 13-4 — 102(l)(b), 5 C.R.S. (1998), and section 16-12-102(1), 6 C.R.S. (1998). We conclude that the term “act of harassment” in the statute is overbroad, but hold that section 18-8-706 is otherwise constitutional. Accordingly, we affirm in part and reverse in part the judgment of the trial court.
I. FACTS AND PROCEEDINGS BELOW
The defendant, Glen Hickman, was married to Keri Johnson for three years, from 1994 through 1997. Johnson had a daughter from a previous relationship, and Hickman and Johnson had a son together. Johnson filed for divorce in January 1997, after Hickman allegedly sexually assaulted Johnson’s daughter. After this time, Hickman saw his son several times a week in supervised visitations. During the divorce proceedings, Hickman and Johnson argued over the custody of and visitation arrangements for their son.
Hickman was initially charged with sexually assaulting a child, his step-daughter. Johnson was a prosecution witness in a hearing concerning this sexual assault charge. Approximately one month after she testified at the hearing and five days before Johnson was scheduled to testify at the jury trial concerning the sexual assault charge, Hickman allegedly set off a firecracker near Johnson’s home. The People contend that he then called her and said, “The next one’s gonna blow your head off,” and, “Hope you sleep well after that.” There was no evidence that the firecracker caused any physical harm or property dam[633]*633age. The defendant was charged with one count of Retaliation Against a Witness or Victim, a class 3 felony, in violation of Colorado Revised Statutes section 18-8-706. This statute provides:
An individual commits retaliation against a witness or victim if such person uses a threat, act of harassment, or act of harm or injury upon any person or property, which action is directed to or committed upon a witness or a victim to any crime, an individual whom the person believes has been or would have been called to testify as a witness or victim, a member of the witness’ family, a member of the victim’s family, an individual in close relationship to the witness or victim, an individual residing in the same household with the witness or victim, as retaliation or retribution against such witness or victim.
§ 18-8-706(1), 6 C.R.S. (1998).
Hickman denies the allegations, but argues that even if they were true, the statute is unconstitutional. Hickman does not argue that the statute is unconstitutional as applied to him, but only that the statute is unconstitutionally overbroad and vague on its face.
The trial court agreed with Hickman’s position and dismissed the retaliation count, reasoning that section 18-8-706 was unconstitutionally overbroad and void for vagueness as a result of amendments made to the statute in 1992. The trial court found that the amendments to the statute created an overbreadth problem because certain terms that were added to and deleted from the statute impermissibly extended the scope of the statute to constitutionally protected communication. The trial court also found that the 1992 amendments to the statute rendered the statute unconstitutionally vague because the amended statute failed to provide adequate notice of what conduct was prohibited by the statute.
In the 1992 amendments, the General Assembly added the terms “threat” and “act of harassment,” and deleted the word “intentionally” along with the phrase “for giving testimony in any official proceeding.” See ch. 73, sec. 20, § 18-8-706, 1992 Colo. Sess. Laws 396, 405.1 Because the added terms were not defined in the statute or clearly limited, the trial court reasoned that the statute could prohibit communications that were perceived as a “threat” or as “harassment,” but were nonetheless constitutionally protected by the First Amendment.2 See U.S. Const, amend. I.
The trial court also found that the deletion of the terms from the statute increased the statute’s sweep, because without an express provision establishing the culpable mental state required for this crime it was unclear exactly what conduct the legislature sought to prohibit. The trial court interpreted these amendments to mean that the prohibited retaliation or retribution was not required to be in response to the person’s “giving testimony in any official proceeding.” Under this statutory interpretation, the accused’s actions might concern a matter unrelated to the person’s testimony in a criminal proceeding.
According to the trial court’s analysis, the amended statute is unconstitutionally over-broad because the statute criminalizes activities that are beyond the scope of governmental regulation. The statute would prohibit, for example: making a threat to report or prosecute perjury by a witness; making a threat to initiate a boycott or other legal collective action process because of false testimony during a trial concerning illegal action by union membership during a strike; or picketing the home or workplace of a witness who lied during a trial with signs calling the [634]*634witness a liar. The trial court additionally-held that the statute was void for vagueness, reasoning that the statute failed to give fair warning to citizens as to what conduct was prohibited and thereby gave law enforcement unbridled discretion to decide whether a crime had been committed. Thus, the trial court held the statute unconstitutional and granted the defendant’s motion to dismiss.
The People argue that the trial court erred in concluding that section 18-8-706 is unconstitutionally overbroad and vague. We agree in part with the People’s contention, and we affirm in part and reverse in part the trial court’s ruling.3
II. ANALYSIS
A. OVERVIEW
Although the trial court examined the statute’s constitutionality only as applied to cases involving retaliation against witnesses who testified in earlier criminal proceedings, the statute has a wider scope. Section 18-8-706 protects persons who fall in any of several classifications, including: (1) a witness to a crime; (2) a victim of a crime; (3) an individual who the defendant believes has been or would have been called to testify as a witness or a victim; and (4) a person in a close personal relationship to the witness or victim of a crime. See § 18-8-706. This statute applies to any actions by an accused that are in retaliation against a person for that person’s membership or perceived membership in one of these classes. See id. We examine section 18-8-706 as applied to all of the classes of persons protected by the statute, not simply witnesses who have already testified in earlier proceedings. For the sake of convenience, we refer to these classes of people as the “persons protected by the statute” or “protected classes.”
Hickman challenges section 18-8-706 as unconstitutionally overbroad and vague.4 We note as an initial matter that statutes are presumed to be constitutional, and “the party challenging [a statute’s] validity has the burden of proving unconstitutionality beyond a reasonable doubt.” People v. Janousek, 871 P.2d 1189, 1195 (Colo.1994). For us to invalidate a statute, it must be sufficiently infirm so that no limiting construction consistent with the legislature’s intent will preserve its constitutionality. See Whimbush v. People, 869 P.2d 1245, 1248 (Colo.1994).
When construing a statute, we are guided by several basic principles of statutory interpretation. Our goal is to ascertain and carry out the intent of the legislature. See Baer, 973 P.2d at 1228. To this end, “[w]ords and phrases used in statutes are to be interpreted according to their generally accepted meaning.” Janousek, 871 P.2d at 1196. “Where the statutory language leaves doubt as to its meaning, we may examine other sources of legislative intent, including the legislature’s objective, the circumstances surrounding the enactment, and the consequences of a particular construction.” Baer, 973 P.2d at 1228.
Thus, we test section 18-8-706 for unconstitutional overbreadth and vagueness by examining the plain meaning of the words “threat” and “act of harassment” as used in the statute, as well as the effect of the deletion of certain terms from the statute. The [635]*635legislative history of the 1992 amendments to section 18-8-706, in which the General Assembly states its desire to protect potential witnesses and victims from coercive conduct, instructs our analysis. The General Assembly’s intent is indicated by the placement of section 18-8-706 within a piece of broader legislation: (a) the purpose of which is to provide protection for victims and witnesses to a crime; and (b) which is to be cited as the “Colorado Victim and Witness Protection Act of 1984.” See § 18-8-701, 6 C.R.S. (1998).
B. OVERBREADTH
1. Overbreadth Standard
We now turn to the People’s argument that the trial court erred in holding that section 18-8-706 was unconstitutionally over-broad. We determine that, given a sufficiently limiting definition of the term “threat,” the statute is not substantially over-broad to the extent that it prohibits certain threats against persons protected by the statute. We also hold that the phrase “act of harassment” is overbroad and strike it from the statute.
We begin our analysis by reviewing the constitutional principles of over-breadth. A statute is unconstitutionally overbroad if it includes within its proscriptions a substantial amount of constitutionally protected speech. See New York v. Ferber, 458 U.S. 747, 769, 102 S.Ct. 3348, 73 L.Ed.2d 1113 (1982); Baer, 973 P.2d at 1231. If a statute meets this description, the statute must be invalidated unless the court can supply a limiting construction or partial invalidation that narrows the scope of the statute to constitutionally acceptable applications. See Broadrick v. Oklahoma, 413 U.S. 601, 613, 93 S.Ct. 2908, 37 L.Ed.2d 830 (1973); People v. Ryan, 806 P.2d 935, 940 (Colo.1991). “The criterion of ‘substantial overbreadth’ precludes a court from invalidating a statute on its face simply because of the possibility, however slight, that it might be applied in some unconstitutional manner.” Baer, 973 P.2d at 1231 (citing Secretary of State of Maryland v. Munson Co., 467 U.S. 947, 964-65, 104 S.Ct. 2839, 81 L.Ed.2d 786 (1984)). The substantial overbreadth doctrine applies to constitutional challenges of statutes that prohibit “pure speech” as well as “conduct plus speech.” See Ferber, 458 U.S. at 771, 102 S.Ct. 3348 (applying substantial overbreadth doctrine to pornography statute); see also City of Houston v. Hill, 482 U.S. 451, 466-67, 107 S.Ct. 2502, 96 L.Ed.2d 398 (1987) (using substantial over-breadth doctrine in case involving verbal criticism of police officer).
If a statute encompasses protected speech but is not substantially overbroad, then whatever overbreadth may exist should be resolved on a case-by-case basis. See Broadrick, 413 U.S. at 615-16, 93 S.Ct. 2908; Ryan, 806 P.2d at 939-40. The court must weigh the degree to which protected speech may be deterred under the statute against the scope of the unprotected speech being regulated. See Ferber, 458 U.S. at 773, 102 S.Ct. 3348 (finding that legitimate reach of child pornography statute “dwarfs its arguably impermissible applications,” and that such applications were a “tiny fraction” of the materials within the statute’s reach); Watts v. United States, 394 U.S. 705, 707-08, 89 S.Ct. 1399, 22 L.Ed.2d 664 (1969) (upholding statute that reaches protected speech because of “a valid, even an overwhelming” interest in protecting the President from threats). If a person engaged in protected speech is prosecuted under the statute, the court should deem the particular prosgeution invalid, not invalidate the entire statute. See Ferber, 458 U.S. at 773-74, 102 S.Ct. 3348 (upholding statute even though statute might conceivably be applied to protected communications); Watts, 394 U.S. at 707-08, 89 S.Ct. 1399 (upholding statute but overturning conviction under the statute because the particular speech was protected).
The purpose of the overbreadth doctrine is to protect persons whose speech or conduct is constitutionally protected but who “may well refrain from exercising their rights for fear of criminal sanctions by a statute susceptible of application to protected expression.” Ferber, 458 U.S. at 768, 102 S.Ct. 3348 (quoting Village of Schaumburg v. Citizens for a Better Env’t, 444 U.S. 620, 634, 100 S.Ct. 826, 63 L.Ed.2d 73 (1980)). As we discussed in note 4, supra, we allow persons [636]*636to bring overbreadth attacks even when the person’s conduct is clearly unprotected and could be proscribed by a statute drawn with the requisite specificity. See also Ferber, 458 U.S. at 769, 102 S.Ct. 3348. Because of the effects of striking down a statute on its face as a result of a challenge by a person whose own conduct may not be protected by the First Amendment, courts recognize the overbreadth doctrine as “strong medicine,” and the doctrine is applied only as a “last resort.” Id. at 769, 102 S.Ct. 3348; Ryan, 806 P.2d at 939. If a limiting construction or partial invalidation that confines the statute to sufficiently narrow applications can be applied, the court should construe the statute in light of that limiting construction. See Whimbush, 869 P.2d at 1248.
Summarizing the relevant caselaw, we conclude that overbreadth analysis involves two questions. First, the court must determine if the statute at issue encompasses constitutionally protected communications. Second, if the statute extends to protected communications, the court must determine whether the statute extends to a “substantial” amount of protected communication such that the statute is unconstitutional, or whether unconstitutional applications of the statute should be cured on a case-by-case basis. A court has the responsibility to apply a limiting construction or partial invalidation if doing so will preserve the statute’s constitutionality.
2. The Term “Threat” Does Not Render the Statute Unconstitutionally Overbroad
The trial court agreed with Hickman’s contention that the statute’s prohibition of “threats” against a person protected by the statute is overbroad because the statute sweeps within its scope a substantial amount of protected communications. The trial court found that the statute covered “threats” that are protected speech, including as examples: threats to report or prosecute perjury by a witness; threats to initiate boycotts or other legal collective action in response to false testimony given during a trial concerning illegal action by union membership during a strike; and picketing the home or workplace of a witness with signs calling the witness a liar. Under the trial court’s analysis, because the statute potentially could be applied to a broad range of protected speech, the statute was unconstitutionally overbroad.
We are not persuaded by this reasoning. Instead, we supply a limiting construction for the term threat under section 18-8-706 that narrows the scope of the statute to include only expressions of intent to commit harm or injury to another’s person, property, or rights through the commission of an unlawful act. The statute itself further limits the scope of prohibited speech because the threat must be made against a person protected by the statute for retributive or retaliatory purposes. Based on the limited scope of the statute’s prohibition of threats, we hold that the proscription of threats in section 18-8-706 does not infringe on a substantial amount of protected speech and thus we conclude that the statute is not overbroad.
a. Definition of “Threat” Under Section 18-8-706
Earlier Colorado cases address the meaning of “threat.” In People v. Hines, we held, in the context of our felony menacing statute, that the meaning of threat is “a statement of purpose or intent to cause injury or harm to the person, property, or rights of another, by the commission of an unlawful act.” 780 P.2d 556, 559 (Colo.1989). In Schott v. People, we defined “threat” in the context of a theft by threat statute as a “declaration of purpose or intention to work injury to the person, property, or rights of another by the commission of an unlawful act.” 174 Colo. 15, 18, 482 P.2d 101, 102 (1971) (citing Black’s Law Dictionary 1651 (4th ed.1968)).
“It is to be presumed that a legislature is cognizant of and adopts the construction which prior judicial decisions have placed on particular language when such language is employed in subsequent legislation.” Binkley v. People, 716 P.2d 1111, 1114 (Colo.1986). Because the legislature provided no definition of “threat” under section 18-8-706, we presume that the legislature intended that the term have the meaning adopted by this court in earlier decisions such as Hines and Schott.
[637]*637In addition to our earlier cases and the presumption we make about the legislature’s intent based on those cases, other sources define threat as a communication of an intention to commit harm through the commission of unlawful acts. Webster’s Third Neiv International Dictionary 2382 (1986), defines “threat” as:
[An] expression of an intention to inflict loss or harm on another by illegal means and especially] by means involving coercion or duress of the person threatened.
Black’s Law Dictionary 1480 (6th ed.1990), defines threat as a “declaration of an intention to injure another or his property by some unlawful act.”5
Although our caselaw, the presumption we make about the General Assembly’s intent based on that caselaw, and the dictionary definition of threat all indicate that “threat” refers to the commission of illegal acts, we note that “threat” may be conceived of in a broader fashion, including expressions of intent to commit harm or injury through the commission of an act that is not necessarily illegal. For example, the extortion statute invalidated in Whimbush defined in a broad manner the threats proscribed. See 869 P.2d at 1247. The extortion statute defined the prohibited threats as including threats , to “confine, restrain, or cause economic or bodily injury to the threatened person,” as well as threats to “damage the property, economic well-being, or reputation of the threatened person,” and this definition of threat encompassed a greater spectrum of speech than does the definition we supplied in earlier cases. See id.; see also, e.g., Black’s Law Dictionary 1489 (7th ed.1999) (defining threat as a communicated intent to inflict harm or loss on another, without reference to the commission of unlawful acts); Webster’s, supra, at 2382 (defining threat as “an indication of something impending and [usually] undesirable or unpleasant” and as “an expression of an intention to inflict evil, injury, or damage on another”); Planned Parenthood League of Massachusetts, Inc. v. Blake, 417 Mass. 467, 631 N.E.2d 985, 990 (1994) (accepting definition of threat' as “the intentional exertion of pressure to make another fearful or apprehensive of injury or harm”). Thus, we recognize that threat has several meanings, including some definitions that are broader than the one we have applied in our earlier cases.
Although both the narrow and broad definitions of threat are reasonable interpretations of the term, if a “statute is capable of alternative constructions, one of which is constitutional, then the constitutional interpretation must be adopted.” People v. McBurney, 750 P.2d 916, 920 (Colo.1988). As we explain in our further discussions in subsection (c), infra, a definition of threat that limits the term to only those statements of intent to cause harm or injury through the commission of unlawful acts bolsters the statute’s constitutionality. By defining threat in this fashion, we save the statute from overbreadth because the statute does not encompass a substantial amount of protected speech.
In sum, Colorado caselaw defines threat and provides a basis for presuming that the General Assembly intended to use this definition, and we find support for this definition in other sources. Our analysis of the constitutionality of section 18-8-706 also suggests that threat should be interpreted in a narrow fashion. Thus, we construe threat in section 18-8-706 to mean an expression of an intent or statement of purpose to commit harm or injury to another’s person, property, or rights through the commission of unlawful acts.
b. First Amendment Protections for “Threats”
Having construed threat under section 18-8-706, we now turn to the question of whether the statute infringes upon communications that are protected by the First Amendment. After reviewing a substantial [638]*638body of caselaw addressing the types ’ of threats that are and are not protected by the First Amendment, we conclude that although the statute infringes on protected communications, it does not substantially burden protected speech.
The Supreme Court has made clear that some threats are communications without First Amendment protections. In cases involving threats against the President, the Supreme Court has acknowledged that “true threats” are not protected by the First Amendment. See Watts, 394 U.S. at 707-08, 89 S.Ct. 1399 (1969)(overturning conviction under statute prohibiting threats against the President but upholding the statute on its face). The Court has explained that “threats of violence” fall outside First Amendment guarantees in order to protect individuals from the fear of violence, from the disruption that fear engenders, and from the possibility that the threatened violence will occur. See R.A.V. v. City of St. Paul, 505 U.S. 377, 388, 112 S.Ct. 2538, 120 L.Ed.2d 305 (1992). In Madsen v. Women’s Health Center, Inc., 512 U.S. 753, 773, 114 S.Ct. 2516, 129 L.Ed.2d 593 (1994), the Court noted that, “Clearly, threats to patients or their families, however communicated, are proscribable under the First Amendment.”
Similarly, we have recognized that some threats are not protected by the First Amendment. In Baer, for example, we upheld Colorado’s harassment-by-stalking statute against an overbreadth challenge, finding that “credible threats” do not have First Amendment protections.6 973 P.2d at 1231-32. In Janousek, we held that a person “has no constitutionally protected right to make threats of violence to a public servant.” 871 P.2d at 1193.7
The category of unprotected threats, however, cannot conveniently be defined as “threats to commit immediate bodily harm” or “threats of violence,” because some unprotected threats do not meet these definitions. Extortionate threats, for example, often involve threats that are neither threats to commit violence nor threats to commit actions that are per se illegal.8 Even so, extortion statutes that prohibit a wide range of such threats are routinely upheld as constitutional, often without explanation of what characteristics distinguish these threats from protected speech. See, e.g., United States v. Hutson, 843 F.2d 1232, 1235 (9th Cir.1988) (“Because the statute in the present case is limited to extortionate threats, it does not regulate speech relating to social or political conflict, where threats to engage in behavior that may be unlawful may nevertheless be part of the marketplace of ideas.”); United States v. Quinn, 514 F.2d 1250, 1268 (5th Cir.1975) (“It may categorically be stated that extortionate speech has no more constitutional protection than that uttered by a robber while ordering his victim to hand over the money, which is no protection at all.”).
While some threats have no First Amendment protections, other threats are protected speech. In Aguilar v. People, we [639]*639recognized that “constitutionally protected speech may be threatening.” 886 P.2d 725, 728 (Colo.1994). The “mere advocacy” of force or violence does not remove speech from the protections of the First Amendment. Brandenburg v. Ohio, 395 U.S. 444, 447-49, 89 S.Ct. 1827, 23 L.Ed.2d 430 (1969) (finding that advocating violence as moral propriety or moral necessity “ ‘is not the same as preparing a group for violent action and steeling it to such action’ ”) (quoting Noto v. United States, 367 U.S. 290, 297-98, 81 S.Ct. 1517, 6 L.Ed.2d 836 (1961)). Threats of violence that are not true threats, but which are “political hyperbole,” are protected speech. Watts, 394 U.S. at 706-08, 89 S.Ct. 1399 (finding that statement, “If they ever make me carry a rifle the first man I want to get in my sights is L.B.J.,” was protected “political hyperbole”). Thus, even threats of violence or force may be protected speech.
Just as some threats of violence may be protected communications, threats to commit minor crimes or “insubstantial harm” also may be protected speech in some contexts. See Whimbush, 869 P.2d at 1249-50 (citing State v. Robertson, 293 Or. 402, 649 P.2d 569, 580-81 (1982)).9 The concern in these cases is that in some instances where a statute purports to reach only threats to commit illegal activities, the statute still encompasses a substantial amount of protected speech. The Robertson court, for example, reasoned that some threats to commit minor or insubstantial crimes may be legitimate speech, especially in political contexts. See 649 P.2d at 583-84. Thus, as we stated in Whimbush, “a statute does not escape potential overbreadth solely because it is limited to threats involving unlawful conduct,” because such threats may be protected speech. 869 P.2d at 1249-50 (citing Robertson, 649 P.2d at 587).
Along with some threats to commit violence or to commit minor crimes, other types of “threats” may have constitutional protection even if the threat has a coercive effect on other people. Threats of social ostracism in the context of an economic boycott are protected. See NAACP v. Claiborne Hardware Co., 458 U.S. 886, 910, 102 S.Ct. 3409, 73 L.Ed.2d 1215 (1982) (“Speech does not lose its protected character ... simply because it may embarrass others or coerce them into action.”). Expressions made with the intent to exert a coercive influence over others are not necessarily outside the protections of the First Amendment. See Organization for a Better Austin v. Keefe, 402 U.S. 415, 419, 91 S.Ct. 1575, 29 L.Ed.2d 1 (1971) (finding that “offensive” and coercive speech may be protected). In Whimbush, we invalidated an extortion statute that prohibited a wide variety of threats, including some forms of.protected speech.10 See 869 P.2d at 1248. The extortion statute defined “threats” so broadly that it encompassed a substantial amount of protected speech, such as threats to engage in legal collective action in support of group demands. See id.
Based on our review of the general principles of the easelaw addressing First Amendment protections for threats, we conclude that the term threat under section 18-8-706, as we have construed it, extends to protected speech. Threat means an expression of intent to commit harm or injury to another’s person, property, or rights through the commission of an unlawful act. This definition includes various threats not entitled to First Amendment protection, such as threats to kill a person and threats to retaliate forcibly against governmental witnesses. However, [640]*640this definition of threat also encompasses threats that have First Amendment protection, such as threats to block an intersection as part of a political protest or to obstruct the entrance to a storefront as part of a boycott. Thus, standing alone, the term “threat,” as we define it under section 18-8-706, extends to protected speech as well as unprotected communications.
We note that the ambiguous nature of a term like threat, which involves both protected and unprotected communications, presents a special difficulty for a legislature. It is apparent from our review of the law concerning First Amendment protections for threats that some threats are protected and some are not, but it is not evident how to define precisely the differences between the two categories. As a result of this ambiguity, any attempt to compose a statute prohibiting unprotected speech will almost necessarily encompass protected speech. It is difficult, perhaps even impossible, to construct a statute that applies to a maximum amount of the unprotected elements of speech from a category that contains both protected and unprotected communications, without also infringing on protected speech as well. Cf. Whimbush, 869 P.2d at 1250 (“If the General Assembly chooses to reenact the criminal extortion statute, it must make many policy choices to define the scope of the statute and the nature of the conduct to be prohibited.”); see also Kent Greenawalt, Speech, Crime, & the Uses of Language 91 (1989) (“Finding the appropriate language to reach just [proscribable] threats is no easy task.”). This difficulty highlights the need to test a statute’s over-breadth by determining whether a substantial amount of protected speech is burdened by the statute. Requiring a ’legislature to construct a statute so that a significant amount of unprotected speech but no protected speech is covered by the statute places an unreasonable and unnecessary burden on the legislature.
c. Test for Substantial Overbreadth
We now ask whether section 18-8-706 as a whole infringes on a substantial amount of protected communications, and we conclude that it does not.
Viewing the statute in light of our definition of threat, we see that section 18-8-706 has three essential elements. First, there must be a communicated intent to commit harm or injury to another’s person, property, or rights by the commission of unlawful acts. Second, the threat must be directed toward a person protected by the statute. Third, the threat must be made for retributive or retaliatory purposes based on the victim’s membership in one of the protected classes. To retaliate means “to put or inflict in return ... a wrong!;] ... to return like for like: make requital; especially]: to return evil for evil.” Webster’s, swpra, at 1938. Retribution is “the dispensing or receiving of reward or punishment according to the deserts of the individual ... [or] something given or exacted in recompense.” Id. at 1940.11 Each of these terms thus defines retaliation under the statute as occurring for a specific reason: that is, the defendant must specifically intend to inflict punishment or exact revenge on a person as a result of that person’s actual or perceived relationship to criminal proceedings. Thus, the statute requires that the prohibited threat be made for retributive or retaliatory purposes.
While some instances of protected speech may involve threats that meet all of the elements of section 18-8-706, such instances cannot be considered a “substantial” amount of the speech covered by the statute because the bulk of the speech covered by the statute is not protected. For example, threats to kill or injure a witness in retaliation for the [641]*641witness’s testimony are not protected speech. Threats to injure a potential witness’s family, made with the intent of discouraging testimony, are not protected speech. We reason that these and similar forms of unprotected speech comprise the vast majority of proscribed speech, even though we concede that some forms of protected speech conceivably will meet all three elements of the statute.
The limited scope of the statute excludes many forms of protected speech that the trial court assumed would be covered. Threats to expose a witness’s perjury, for example, would not be encompassed under section 18-8-706 because the “threat” in that case does not rely on the commission of an illegal act. Threats to initiate a boycott or other legal collective action in response to untruthful testimony about union members also would not be covered by the statute because the threatened actions are not illegal.
Our construction of section 18-8-706 also excludes a category of protected “threats” that might otherwise contribute to the statute’s overbreadth. For example, because the legislature applied a broader definition of threat than that applied by this court in earlier cases like Hines and Schott, the extortion statute invalidated in Whimbush encompassed protected threats to engage in legal collective action activities. See 869 P.2d at 1248. Section 18-8-706 does not encompass these or other protected threats because our construction renders the statute considerably narrower than statutes like the extortion statute addressed in Whimbush, and threats to engage in otherwise legal activities are not covered by this statute. Thus, we are not persuaded that a substantial amount of the speech prohibited by the statute is protected speech.
If any prosecution involves an instance of speech that meets all the elements of this statute but the threat is actually protected under the First Amendment, courts may cure the overbreadth on a case-by-ease basis. In Ferber, for example, the Court acknowledged that the child pornography statute at issue might conceivably be applied to protected materials such as medical texts or National Geographic magazines. See 468 U.S. at 773-74, 102 S.Ct. 3348. However, the Court doubted that these applications could amount to more than a “tiny fraction” of the materials within the statute’s reach, so the Court held that any unconstitutional applications of the statute should be cured on a case-by-case basis. See id. at 773, 102 S.Ct. 3348.
Similarly, because we determine that only a small fraction of the speech encompassed by the statute is protected, a court may cure possible applications to protected speech on a case-by-case basis. Hence, in the few instances when protected speech might be prosecuted under the statute, a court may invalidate the particular prosecution.
Although we commented in Whimbush that “a statute does not escape potential overbreadth solely because it is limited to threats involving unlawful conduct,” 869 P.2d at 1249-50, we are not required to find section 18-8-706 overbroad on this basis. Our holding that section 18-8-706 is not over-broad does not rest solely on the fact that we have narrowly construed the term threat. In addition to being limited only to threats to commit unlawful acts, section 18-8-706 further limits the scope of prohibited threats by describing a specific class of persons to whom the threat must be directed and by requiring that the threat be made for retaliatory or retributive purposes. Thus, our holding has not saved section 18-8-706 from overbreadth solely because it is limited to threats to commit illegal conduct.
In sum, we construe the term threat under section 18-8-706 to identify only a limited category of expressions, and the statute further narrows the category of prohibited communications to only those threats made against a person protected by the statute for retributive or retaliatory purposes. Because only a small fraction of the speech prohibited by this section of the statute is protected speech, we hold that section 18-8-706 is not overbroad insofar as the statute prohibits a person from making threats against a person protected by the statute for retributive or retaliatory purposes.
S. “Act of Harassment” Is Unconstitutionally Overbroad
We agree with the trial court that “act of harassment” as used in section 18-8-706 en[642]*642compasses a substantial amount of constitutionally protected communications. Following the standard for overbreadth we outline in section 11(B)(1), supra, we hold that section 18-8-706 substantially burdens protected communications and that the phrase cannot be sufficiently limited. Therefore, we strike this prohibition from the statute.
a. “Act of Harassment” Encompasses Protected Communications
We turn to the definition of “act of harassment” under the statute. The trial court found that “act of harassment” was undefined by the statute, and that the phrase encompasses speech and conduct that are constitutionally protected. We agree.
We look to the plain meaning of the terms in order to determine whether they encompass protected communications. The term “harassment” is synonymous with “vex,” “trouble,” or “annoy.” See Webster’s, supra, at 1031; see also Black’s Law Dictionary 721 (7th ed.1999) (defining harassment as conduct that is directed at a specific person that “annoys, alarms, or causes substantial emotional distress and serves no legitimate purpose”).
This broad meaning of the term applies to a wide range of communications and conduct, many of which are protected by the First Amendment. For example, forecasting a change in weather, engaging in a political discussion, or discouraging a witness from lying on the stand might “vex,” “trouble,” or “annoy” a person protected by section 18-8-706, but such communications are protected by the First Amendment. See, e.g., People v. Smith, 862 P.2d 939, 942 (Colo.1993) (finding that subsection of harassment statute was “anything but narrowly drawn” and concluding that statute covered protected speech); Bolles v. People, 189 Colo. 394, 398, 541 P.2d 80, 83 (1975) (finding that forecasting the weather or predicting political trends, for instance, could “alarm” a person, but are still protected speech). Thus, we conclude that “act of harassment” encompasses protected communications.
b. Test for Substantial Overbreadth
Having determined that “act of harassment” encompasses constitutionally protected communications, we turn to the question of whether this overbreadth is substantial. Because no limiting construction can narrow the statute to permissible applications, we conclude that the phrase “act of harassment” in section 18-8-706 is substantially overbroad, and we strike it from the statute.
Our cases consistently invalidate as substantially overbroad statutes that prohibit speech made with the intent of “harassing,” “annoying,” or “alarming” others. In Smith, for example, we invalidated a subsection of the harassment statute that prohibited “all repeated communications containing ‘offensively coarse language’ if made with the intent to annoy, harass, or alarm” because of the amount of protected speech covered by the statute. See 862 P.2d at 942. In Bolles, we invalidated a statute that prohibited certain communications made with the intent to “harass, annoy, or alarm another person,” because a significant amount of the speech covered by the statute was protected speech. See 189 Colo. at 399, 541 P.2d at 84.
Similarly, the phrase “act of harassment” in section 18-8-706 encompasses a substantial amount of protected communications. For example, if a defendant calls a witness a liar or proposes to turn the witness in for perjury as retribution for the witness’s testimony, these communications may “vex,” “trouble,” or “annoy” the witness, and would thus be pro.secutable under the statute. Alternatively, if a union leader tells a manager that workers will strike if the manager gives false testimony concerning the union’s members in a labor dispute, this could certainly “trouble” or “annoy” the manager and be prosecuted under the statute. However, these communications are protected speech, even if they were intended to harass a person protected by the statute.12 These and other forms of protected speech that could be prosecuted under the statute as acts of harassment amount to more than a “tiny fraction” [643]*643of the communications prohibited under section 18-8-706.
Even though the statute requires that an act of harassment be directed at a person protected by the statute and be made for retributive or retaliatory purposes, the statute’s broad scope is not adequately limited by these requirements. A considerable amount of protected speech is encompassed by prohibiting acts of harassment directed towards persons protected by the statute for retributive or retaliatory purposes. As our examples above demonstrate, retaliatory communications designed to “harass” a person protected by the statute may nonetheless be protected speech. Thus, the requirements that the act of harassment be directed at a person protected by the statute for retributive or retaliatory purposes do not limit the statute to permissible applications.
We conclude that the proscription of an “act of harassment” under section 18-8-706 includes a substantial amount of protected communications and that we cannot supply a limiting construction for the statute. We also conclude that the statute does not sufficiently limit the scope of protected communications to which it applies. Thus, we hold that the phrase “act of harassment” in the statute is unconstitutionally overbroad and strike it from the statute.
Although we find that the term “act of harassment” in section 18-8-706 is unconstitutionally overbroad, we need not invalidate the entire statute if only one part of a statute is objectionable. A court may sever one section of a statute from the whole if “partial invalidation will rid the statute of the constitutional infirmity of overbreadth.” Ryan, 806 P.2d at 940. The General Assembly has provided by statute that we engage in such partial invalidation when necessary to save a statute from complete invalidation. See § 2-4-204, 1 C.R.S. (1999).13 Therefore, we hold that section 18-8-706 is invalid only insofar as it prohibits an act of harassment” against persons protected by the statute, and we strike this phrase from the statute. This partial invalidation does not alter the basic prohibition against making a threat or committing an “act of harm or injury” against a person protected by the statute.
C. VAGUENESS
The People also argue that the trial court erred when it determined that section 18-8-706 was void for vagueness. We agree. We hold that this statute, as construed, provides adequate notice of the prohibited behavior and is not void for vagueness.
Vague laws are unconstitutional because they offend due process. See Baer, 973 P.2d at 1233. A vague law offends due process because it fails to give fair notice of the conduct prohibited and does not supply adequate standards to prevent arbitrary and discriminatory enforcement. See id. A law is void for vagueness if its prohibitions are not clearly defined and it may be reasonably susceptible to more than one interpretation by a person of common intelligence. See id. “A penal statute must be sufficiently definite to give fair warning of proscribed conduct so that persons may guide their actions accordingly[,] ... [and] must define an offense with sufficient clarity to prevent arbitrary and discriminatory enforcement of the statute.” Janousek, 871 P.2d at 1195. However, “[a] law is unconstitutional only if it ‘is vague, not in the sense that it requires a person to conform his conduct to an imprecise but comprehensible normative standard, but rather in the sense that no standard of conduct is specified at all.’” Board of Educ. v. Wilder, 960 P.2d 695, 703 (Colo.1998) (quoting Coates v. City of Cincinnati, 402 U.S. 611, 614, 91 S.Ct. 1686, 29 L.Ed.2d 214 (1971)). We may sustain a facial vagueness challenge only “where the enactment is ‘impermissibly vague in all of its applications.’ ” Baer, 973 P.2d at 1233 (quoting Village of Hoffman [644]*644Estates v. Flipside, Hoffman Estates, Inc., 455 U.S. 489, 497, 102 S.Ct. 1186, 71 L.Ed.2d 362 (1982)).
The trial court determined that section 18-8-706 was vague because the terms “threat” and “act of harassment” were not adequately defined and were susceptible to multiple common sense interpretations. In the trial court’s view, the statute failed to provide fair warning concerning what conduct was proscribed and therefore did not comport with due process. We disagree.
The statute is sufficiently specific to provide the constitutionally required guidance to individuals seeking to comply with the law and to law enforcement officers applying the statute.14 We believe that our construction of the statute sufficiently delineates the type of behavior — threats of illegal action or harmful or injurious conduct in retaliation or retribution for an individual’s status as one of the persons protected by the statute — -for a person of normal intelligence to have fair notice of what acts are proscribed. The prosecution must prove: (1) the defendant threatened or committed harmful or injurious conduct towards a person protected by the statute; (2) the defendant made a threat or engaged in the conduct because the defendant knew or believed that the victim of the threat or conduct was a member of the class of protected persons; and (3) the defendant intended the threat or conduct as retaliation or retribution for the defendant’s perception of the victim’s relationship to criminal proceedings. These requirements sufficiently define the conduct prohibited under section 18-8-706 so that persons may conform their actions to the statute, and they present a standard for determining whether enforcement of the statute has been arbitrary and discriminatory. Hence, we conclude that section 18-8-706 is not unconstitutionally vague.
D. REQUIREMENT THAT PROHIBITED THREATS OR CONDUCT BE MADE FOR RETRIBUTIVE OR RETALIATORY PURPOSES
The trial court concluded that the General Assembly contributed to the statute’s overbreadth and vagueness by deleting the word “intentionally” and the phrase “for giving testimony in any official proceeding” from the statute. The trial court reasoned that the 1992 amendments removed the specific intent requirement from the statute. The trial court explained that this lack of a culpability requirement “cast doubt on exactly what the legislature wanted the statute to prohibit,” and that the deletions also “did not help narrow” the sweep of the statute. We disagree with the trial court and hold that the statute requires, as a culpable mental state, that the defendant act intentionally and that this statute is a specific intent offense.
To determine whether this offense is a specific intent crime, we refer to the Criminal Code for guidance. When a criminal statute includes the word “intentionally,” that particular offense is “declared to be a specific intent crime.” § 18-1-501(5), 6 C.R.S. (1999). “A person acts ‘intentionally’ or ‘with intent’ when his conscious objective is to cause the result proscribed by the defining offense.” Id. The Criminal Code also states that when a criminal statute does not expressly designate a culpable mental state, but the proscribed conduct “necessarily involves” a culpable mental state, a culpable mental state may be required for the commission of the offense or some elements of the offense. § 18-1-503(2); see also People v. Gross, 830 P.2d 933, 940 (Colo.1992) (finding that because a crime ordinarily requires an act and simultaneously an accompanying culpable mental state, a court should not necessarily construe legislative silence on the element of intent in a criminal statute as an indication that no culpable mental state is required).
Even though the legislature deleted the word “intentionally” from the statute, the statute nonetheless requires intentional conduct. In the amended statute, the prohibited threat or conduct must still be made “as retaliation or retribution.” As we reasoned in section 11(B)(1)(c) above, both the terms [645]*645“retribution” and “retaliation” by definition require intentional conduct. A person can act in a retributive or retaliatory fashion only if she has a conscious objective to do so. Thus, the statute by the plain meaning of its terms requires the defendant to have as her conscious objective the intent to inflict harm on a specific person for a specific reason— that is, the specific intent to retaliate or to seek retribution against a person protected by the statute because of that person’s relationship to a criminal proceeding. Because the statute requires that the defendant act with a specifically defined conscious objective, we infer the culpable mental state of “intentional” even though the legislature deleted the term from the statute. Thus, we hold that section 18-8-706 is a specific intent offense.
As additional support for our holding that the statute’s terms contain a specific intent requirement, we turn to the legislative history of the amendments. According to the testimony in the committee meetings concerning the 1992 amendments, the proponents of the amendment noted that the statute as originally written required the witness already to have testified before the prohibited conduct occurred and, therefore, this statute did not include as criminal acts threats against a person who was a victim or witness who had not yet testified.15 See Hearing on H.B. 1078 Before the House Judiciary Committee, 58th General Assembly, 2d Sess. (Jan. 23,1992) (statement of Bonnie Benedet-ti, Colo. Dist. Attorneys Council). By amending the statute, the legislature sought to include within the statute’s sweep threats made against persons who had not yet testified and thereby expanded the class of persons protected by the statute.16 Nothing in the legislative history supports the interpretation that by deleting the word “intentionally” the legislature eliminated the requirement that the threat or conduct be retaliation or retribution for a person’s relationship to criminal proceedings.
Further bolstering our holding is the statute’s placement in the Revised Statutes. Section 18-8-706 is part of the “Colorado Victim and Witness Protection Act of 1984.” § 18-8-701, 6 C.R.S. (1998). This Act is designed to protect a person involved in a trial from bribery, intimidation, and tampering.17 Section 18-8-706 itself is entitled “Retaliation against a witness or victim.” Given the location of section 18-8-706 in this title, we infer that the legislature’s intended purpose is to protect people who are or who are thought to be actual or potential witnesses to criminal proceedings.
[646]*646Based on our review of the statutory language, the legislative history of the 1992 amendments, and the placement of this statute within the Colorado Revised Statutes, the statute still contains a specific intent requirement, which both narrows the statute’s sweep and identifies the specific types of behavior prohibited by the statute. Thus, the 1992 changes to section 18-8-706 do not cause any impermissible overbreadth or vagueness.
III. CONCLUSION
We conclude that section 18-8-706, after our partial invalidation, is not unconstitutionally vague or overbroad. We construe the word “threat” to refer to an expressed intent to commit harm through the commission of illegal acts. We conclude that the term “act of harassment” is overbroad and strike it from the statute, but hold that the statute is otherwise valid insofar as it penalizes retributive or retaliatory “threats” against persons protected by the statute. Accordingly, we reverse in part and affirm in part the judgment of the trial court, and we remand the case to the trial court for further proceedings consistent with this opinion.
Chief Justice MULLARKEY concurs in part and dissents in part, and Justice HOBBS and Justice RICE join in this concurrence and dissent.