OPINION ON APPELLANT’S PETITION FOR DISCRETIONARY REVIEW
KELLER, Judge.
Appellant was convicted under the “stalking” provision of the 1993 harassment statute.
See
Texas Penal Code, § 42.07(a)(7) (1993).
Relying primarily upon
Kramer v. Price,
712 F.2d 174 (5th Cir.1988),
rehearing en banc granted,
716 F.2d 284 (5th Cir.1983),
grant of relief affirmed,
723 F.2d 1164 (5th Cir.1984), he argued to the Court of Appeals that the stalking provision is unconstitutionally vague on its face and as applied to his conduct.
Addressing the merits of appellant’s facial challenge, the Court of Appeals held that the statute was not vague and affirmed the conviction.
Long v. State,
903 S.W.2d 52 (Tex.App.—Austin 1995).
Because we find the statute to be unconstitutionally vague on its face, we will reverse.
A. Prior Vagueness Precedent
It is well-established that criminal laws must be sufficiently clear ⅜ at least three respects. First, a person of ordinary intelligence must be given a reasonable opportunity to know what is prohibited.
Grayned v. Rockford,
408 U.S. 104, 108, 92 S.Ct. 2294, 2298-2299, 33 L.Ed.2d 222 (1972).
Kramer,
712 F.2d at 176. Second, the law must establish determinate guidelines for law enforcement.
Groyned,
408 U.S. at 108-109, 92 S.Ct. at 2298-2299.
Kramer,
712 F.2d. at 176-177. Finally, where First Amendment freedoms are implicated, the law must be sufficiently definite to avoid chilling protected expression.
Groyned,
408 U.S. at 109, 92 S.Ct. at 2299. “When a statute is capable of reaching First Amendment freedoms, the doctrine of vagueness ‘demands a greater
degree of specificity than in other contexts.’ ”
Kramer,
712 F.2d at 177 (citations omitted). Greater specificity is required to preserve adequately the right of free expression because “[ujncertain meanings inevitably lead citizens to steer far wider of the unlawful zone than if the boundaries of the forbidden areas were clearly marked.”
Grayned,
408 U.S. at 109, 92 S.Ct. at 2299 (internal ellipses and quotation marks omitted). Moreover, when a vagueness challenge involves First Amendment considerations, a criminal law may be held facially invalid even though it may not be -unconstitutional as applied to the defendant’s conduct.
Gooding v. Wilson,
405 U.S. 518, 521, 92 S.Ct. 1103, 1105, 31 L.Ed.2d 408 (1972).
Kramer,
712 F.2d at 176 n. 3.
In
Kramer,
the Fifth Circuit addressed the constitutionality of Texas’ pre-1983 harassment statute. The pre-1983 statute provided in part:
(a) A person commits an offense if he intentionally:
(1) communicates by telephone or in writing in vulgar, profane, obscene, or indecent language or in a coarse and offensive manner and by this action intentionally, knowingly, or recklessly annoys or alarms the recipient;
Kramer,
712 F.2d at 176,
citing
Texas Penal Code § 42.07(a)(l)(pre-1983 version). While the court declined to address the question of overbreadth, it nevertheless indicated that First Amendment considerations were intertwined with the vagueness issue.
Kramer,
712 F.2d at 176 n. 3 & 177. Relying upon
Coates v. Cincinnati,
402 U.S. 611, 91 S.Ct. 1686, 29 L.Ed.2d 214 (1971), the Fifth Circuit held that the words “annoy” and “alarm” were inherently vague and that Texas courts had not construed the terms to lessen their vagueness.
Kramer,
712 F.2d at 177-178. The Fifth Circuit also criticized the statute for failing to specify whose sensitivities were offended.
Id.
at 178. The court further explained that a statute’s vagueness is not lessened'by making the conduct depend upon each complainant’s sensitivity.
Id.
Finally, the court held that the intent requirement did not save the statute because the underlying conduct was still vague.
Id.
Consequently, the Fifth Circuit held the statute to be
facially unconstitutional due to vagueness.
Id.
We subsequently endorsed
Kramer
in holding the pre-1983 harassment statute to be unconstitutional.
May v. State,
765 S.W.2d 438 (Tex.Crim.App.1989).
B. The Present Statute
The question we confront today is whether the 1993 stalking provision suffers from the same flaws that invalidated the pre-1983 harassment statute in
Kramer.
The “stalking” portion of the 1993 harassment statute provides:
(a) A person commits an offense if, with intent to harass, annoy, alarm, abuse, torment, or embarrass another, he: ...
(7)(A) on more than one occasion engages in conduct directed specifically toward the other person, including following that person, that is reasonably likely to harass, annoy, alarm, abuse, torment, or embarrass that person;
(B) on at least one of those occasions by acts or words threatens to inflict bodily injury on that person or to commit an offense against that person, a member of that person’s family, or that person’s property; and
(C) on at least one of those occasions engages in the conduct after the person toward whom the conduct is specifically directed has reported to a law enforcement agency the conduct described by this subdivision.
[[Image here]]
(e) It is an affirmative defense to prosecution under Subsection (a)(7) of this section that the actor was engaged in conduct that consisted of activity in support of constitutionally or statutorily protected rights.
§ 42.07(a)(7) & (e). Appellant concedes that subsection (a)(7)(B) is reasonably specific but complains that (a)(7)(A) is vague and renders the entire statute unconstitutional. The Court of Appeals opinion tends to focus upon the stalking provision as a whole, finding that the combined requirements sufficiently define the offense to provide notice and guidelines for law enforcement.
1.
The “Conduct” Requirement
If (a)(7)(A) is viewed in isolation, it appears to suffer the same flaws denounced by
Kramer.
The words “annoy” and “alarm” remain in the statute although they are now joined by the words “harass,” “abuse,” “torment,” and “embarrass.” But, all these terms are joined with a disjunctive “or,” and thus do nothing to limit the vagueness originally generated by “annoy” and “alarm.” Moreover, the additional terms are themselves susceptible to uncertainties of meaning. The stalking subsection specifies a mental state of “intent” as opposed to mere knowledge or recklessness, but
Kramer
did not turn upon the degree of mental culpability; the mental state did not solve the vagueness of the underlying conduct. The stalking subsection adds a requirement that the conduct be “specifically directed” toward a particular person. But at the same time, (a)(7)(A) covers
any
conduct in which a person could possibly engage. This broad conduct provision is even more expansive than that found in the earlier statute, which merely proscribed “communications by telephone or in writing.” Any limiting function served by the words “specifically directed” is offset by the expansive coverage of all conduct by the stalking provision.
Appellant concedes that (a)(7)(A) contains a “reasonable person” standard, which was absent from the earlier statute. But we are not bound by such a concession. Whether the stalking provision contains a “reasonable person” standard is a question of law, and we are under no obligation to accept a concession on an issue of law even if all parties agree.
See Colorado Republican Federal Campaign Committee v. Federal Election Commission,
— U.S. -, -, 116 S.Ct. 2309, 2319, 135 L.Ed.2d 795 (1996).
Orloff v. Willoughby,
345 U.S. 83, 87, 73 S.Ct. 534, 537, 97 L.Ed. 842 (1953).
To determine whether the offense contains a reasonable person standard, we must first examine the plain meaning of the statute’s words.
Boykin v. State,
818 S.W.2d 782, 785 (Tex.Crim.App.1991). If the words are ambiguous or the plain meaning leads to absurd results, we may then consider extra-textual sources to aid our interpretation.
Id.
The only possible locus for a reasonable person standard would be the words “reasonably likely,” contained in (a)(7)(A). While those words suggest a degree of probability, standing alone they do not explain from whose perspective the probability is determined. Moreover, the subsection also contains the words “that person,” which appear to specify the perspective as being that of the complainant rather than that of a reasonable person. We do not believe that ordinary persons, viewing the language of (a)(7)(A), would find an implied “reasonable person” perspective. At most, the words “reasonably likely” plainly import a minimal causality requirement. For example, a person who delusionally believes that he can annoy another through the use of “telepathic powers,” and actually intends to do so, would not be covered by the statute because such conduct would not be “reasonably likely” to annoy the recipient. Such a minimal causality requirement, however, is a far cry from a reasonable person standard.
One Court of Appeals, in the context of the harassment statute, has held that the “reasonably likely” language “specifies whose sensitivities are relevant.”
Bader v. State,
773 S.W.2d 769, 770 (Tex.App.—Corpus Christi 1989, pet. for disc. rev. refd). According to that court, it is the recipient of the communication whose sensitivities must be offended.
Id.
Another legal context in which “reasonably likely” commonly appears is in connection with the right against self-incrimination. “Interrogation” has been defined as any words or actions by the police that “they
should have known
are reasonably likely to elicit an incriminating response.”
Rhode Island v. Innis,
446 U.S. 291, 302, 100 S.Ct. 1682, 1690, 64 L.Ed.2d 297 (1980)(emphasis in original).
See also Jones v. State,
742 S.W.2d 398, 407 (Tex.Crim.App.1987). The self-incrimination context appears to undercut application of an objective standard: that “reasonably likely” must be paired with the words “should have known” indicates that the former phrase does not, by itself, incorporate a reasonable person standard.
Even assuming ambiguity in the words “reasonably likely,” the legislative history is inconclusive on the matter. Representative Combs suggested that the statute contained an “objective standard” during questioning on the House floor, but representative Goodman later argued against a proposed amendment in part because it incorporated a “reasonable person” standard.
See
House Floor Debates, March 8, 1993, S.B. 25, Tape 24, Side B and Tape 25, Side A. In committee, Senator Turner discussed the possibility of incorporating a reasonable person standard as
an alternative to
a threat requirement. The particular reasonable person standard proposed was rejected.
See
testimony of John Boston, Texas Criminal Defense Lawyers Assn., and Senator Turner, Senate Criminal Justice Committee, February 9, 1993, S.B. 25, written transcript, pp. 18-20.
The legislative history of the 1983 version of the harassment statute, which first placed the words “reasonably likely” in some of the offense provisions, is silent concerning the subject. A review of Representative Patricia Hill’s testimony indicates that the legislative changes were meant to correct constitutional defects unrelated to the issue in the present case. House Criminal Jurisprudence Committee, H.B. 838, April 11,1983.
Moreover, had the legislature intended to apply a reasonable person standard, they easily could have specified one, or a clear synonym. Examples of clear “reasonable person” language can be found in statutes relating to murder, injury to a child, child abandonment, and tampering with identification numbers.
See
§§ 19.02(a)(1), 22.04(d), 22.041(a), 22.041(e), 31.11(a)(2)(B). We conclude that (a)(7)(A) does not incorporate a reasonable person standard.
Nevertheless, the absence of a reasonable person requirement is not necessarily fatal to the statute’s constitutionality. Other provisions, combined with (a)(7)(A), might sufficiently define the offense to avoid a vagueness problem. In
Colten v. Kentucky,
407 U.S. 104, 92 S.Ct. 1953, 32 L.Ed.2d 584 (1972), the Supreme Court upheld a Kentucky disorderly conduct statute. The statute provided:
(1) A person is guilty of disorderly conduct if, with
intent to
cause public inconvenience, annoyance, or alarm, or recklessly creating a risk thereof, he:
[[Image here]]
(f) Congregates with other persons in a public place and refuses to comply with a lawful order of the police to disperse....
Id.
at 108, 92 S.Ct. at 1956 (ellipses in original). The Court held the statute to be sufficiently clear because a person who stands in a group with others in a public place should understand that he could be convicted under the Kentucky statute “if he fails to obey an order to move on.”
Id.
at 110, 92 S.Ct. at 1957. The refusal to comply with a lawful order to disperse, combined with an intent to annoy, rendered the statute sufficiently clear to defeat a vagueness challenge.
2.
The “Report” Requirement
In addition to (a)(7)(A), the stalking provision before us contains a “threat” requirement, (a)(7)(B), a “report” requirement (a)(7)(C), and an affirmative defense for constitutionally (and statutorily) protected activities, (e). The plain language of the report requirement requires merely that the defendant engage in relevant conduct after the victim has reported previous relevant conduct to a law enforcement agency. The wording of the statute does not require the defendant to
know
that the victim has made such a report. Hence, (a)(7)(C) does little or nothing to inform an ordinary person that his conduct is forbidden because the subsection contains no culpable mental state.
Under some circumstances, a mental state may be required even though not expressly prescribed by a statute’s plain wording. § 6.02(b).
But in the present case a mental state is prescribed in the statute: the intent to harass, annoy, alarm, abuse, torment, or embarrass. Moreover, § 6.02(b) does not apply if the offense “plainly dispenses with any mental element.” Former § 42.07 specifies a mental state in subsection (a) proper. Subsection (a) contains seven numerical divisions, each of which defines a separate offense. Divisions (3), (5), and (6) each contain an
additional
culpable mental state connected with some of the acts outlined in those divisions.
The legislature obviously knew how to prescribe an additional mental state in connection with offense divisions of the harassment statute. By prescribing an additional mental state in connection with other divisions, but omitting a mental state in (7)(C), the legislature plainly dispensed with any additional mental state in (7)(C) that might otherwise be required. The legislative history confirms our reading of (a)(7)(C). The report requirement was originally a stalking bill amendment authored by Representative De La Garza. He explained that the purpose of the report requirement was “to show an official record” of stalking. House Floor Debate, S.B. 25, March 8,1993, tape 25, side A.
The absence of a mental element renders the report requirement useless as a clarifying element in an otherwise vague law. The report requirement involves no conduct of the defendant; the alleged victim is the one who does the reporting. If the defendant is unaware of the report, then it cannot provide the requisite notice that he has violated the law. The outcome of
Colten
would surely have been different if, instead of requiring an order to the defendant to move on, the Kentucky statute had merely required the officer to radio a complaint to the police station and such complaint could be radioed without the defendant’s knowledge.
3.
The “Threat” Requirement
The threat requirement appears to have potential for clarifying the statute and placing it beyond the reach of First Amendment concerns. In the present statute, however, those purposes are fatally undermined by the threat requirement’s relationship to the conduct requirement in (a)(7)(A).
The stalking offense requires at least two instances of conduct, but only one of those instances need be a threat. Hence, the offense of stalking could occur if one of the two predicate acts is described
solely
by (a)(7)(A), which we have decided suffers from the same vagueness problems as the harassment statute addressed in
Kramer
and
May.
Moreover, the stalking statute does not require that the predicate acts be related to each other in any fashion. The absence of any nexus between the threat requirement in (a)(7)(B) and the conduct requirement in (a)(7)(A) strongly supports viewing the two subsections separately and requiring each to pass constitutional muster on its own.
The lack of any connection between the predicate acts distinguishes the present case from
Colten.
The Kentucky statute in that
ease involved two elements of conduct: (1) congregating in a public place, and (2) refusing to comply with a lawful order to disperse. Both of these elements of conduct involve what is in essence a single act: a gathering of people that refuses to disperse.
An examination of decisions from other jurisdictions in which stalking and harassment statutes have been upheld illustrates the deficiencies of the statute before us. Illinois supplies a nexus requirement not contained in the Texas statute: the Illinois statute requires a threat and relevant conduct
in furtherance of the threat. People v. Bailey,
167 Ill.2d 210, 212 Ill.Dec. 608, 614 & 617-618, 657 N.E.2d 953, 959 & 962-963 (1995). Some other states require that the conduct evidence a “continuity of purpose.”
Culbreath v. State,
667 So.2d 156, 158 (Ala.Crim.App.1995).
Bouters v. State,
659 So.2d 235, 236 (Fla.1995),
cert. denied,
— U.S. -, 116 S.Ct. 245, 133 L.Ed.2d 171 (1995).
Luplow v. State,
897 P.2d 463, 465 (Wyo.1995). Ohio requires that the incidents be closely related in time.
State v. Dario,
106 Ohio App.3d 232, 665 N.E.2d 759, 761-763 (1995). Indiana defines “harassment” as conduct which would cause a reasonable person to suffer emotional distress but defines the offense of “stalking” as harassment that would cause a reasonable person to feel “terrorized, frightened, intimidated, or threatened.”
Johnson v. State,
648 N.E.2d 666 (Ind.App. 1 Dist.1995). This language appears to punish as stalking a series of acts where each of the predicate acts in isolation may cause only minor emotional distress such as annoyance but where the series of acts taken together causes more serious emotional distress, i.e. being terrorized, frightened, intimidated, or threatened.
Some jurisdictions have supplied a general mental state provision and/or a reasonable person provision that specifies a more serious state of emotional distress than mere annoyance. Some require a fear of death or bodily injury.
Culbreath,
667 So.2d at 158-159.
Johnson v. State,
264 Ga. 590, 449 S.E.2d 94, 96 (1994).
Woolfolk v. Commonwealth,
18 Va.App. 840, 447 S.E.2d 530, 533 & 535 (1994).
See also Champagne v. Gintick,
871 F.Supp. 1527, 1528 & 1535 (D.Conn. 1994)(fear for physical safety). Others require “substantial emotional distress.”
Bouters,
659 So.2d at 236 & 238.
State v. Martel,
273 Mont. 143, 902 P.2d 14, 18-19 (1995).
Commonwealth v. Schierscher,
447 Pa.Super. 61, 668 A.2d 164, 172 (1995).
Luplow
897 P.2d at 465. Other jurisdictions have construed their statutes to require fear or serious alarm.
CISPES v. FBI,
770 F.2d 468, 471 & 476 (5th Cir.1985).
State v. Moyle,
299 Or. 691, 705 P.2d 740, 742
&
748 (1985).
State v. Lee,
82 Wash.App. 298, 917 P.2d 159, 162, 168 (1996). Florida has expressly “excised” the words “offend” and “annoy” from its telephone harassment statute.
Gilbreath v. State,
650 So.2d 10, 13 (Fla.1995), cert.
denied,
— U.S. -, 115 S.Ct. 1966, 131 L.Ed.2d 857 (1995).
Other courts have upheld their statutes by construing them to punish conduct only when the actor’s
sole
intent was to annoy or cause other emotional distress to the recipient.
McKillop v. State,
857 P.2d 358, 364 (Alaska App.1993).
State v. Richards,
127 Idaho 31, 896 P.2d 357, 362 (1995).
See also People v. Shack,
86 N.Y.2d 529, 634 N.Y.S.2d 660, 663-664, 658 N.E.2d 706, 709-710 (1995)(“with no purpose of legitimate communication” excludes First Amendment activities from the statute).
But see Langford v. City of Omaha,
755 F.Supp. 1460, 1461 n. 2 & 1464 (D.Neb.1989),
appeal dismissed,
978 F.2d 1263 (8th Cir.l992)(“without the purpose of legitimate communication” is unconstitutionally vague). A few states have upheld telephone harassment statutes by emphasizing the limited nature of the forum.
Yates v. Commonwealth,
753 S.W.2d 874, 875-876 (Ky.App.1988)(private telephone is not a public forum).
State v. Kipf
234 Neb. 227, 450 N.W.2d 397, 409 (1990)(privacy interests).
Shack,
634 N.Y.S.2d at 665-666, 658 N.E.2d 706 (privacy interests).
Also of interest, Texas has a statutory civil cause of action for stalking. The civil cause of action contains several restrictions not contained in the criminal statute, including (1) that the harassing behavior cause the claimant to reasonably fear for his safety or the safety of a family member and (2) that
the claimant convey a clear demand to the defendant to stop his harassing behavior.
All of the statutes described above contain additional limiting elements not found in the Texas criminal statute. Many limit the vagueness of the conduct described by linking it to more specific conduct through a continuity of purpose requirement. Others tie the conduct together with a more specific mental state than a mere intent to annoy, such as intent to place in fear of bodily injury, or with a more intense mental state, such as intent to frighten. In addition to creating greater specificity, these kinds of limiting elements help to avoid a vagueness problem by taking the First Amendment out of the picture. Conduct which alone would constitute protected activities may be actionable if it is part of a common plan that includes activity that is clearly unprotected. And, while conduct does not lose First Amendment protection merely because the actor intends to annoy the recipient, such conduct is much less likely to enjoy protection where the actor intends to “frighten” the recipient, and such conduct is unlikely to enjoy any protection where the actor intends to place the recipient in fear of death or bodily injury. If the First Amendment can be removed from the arena, a stalking statute can be evaluated under more deferential due process standards, and is thus more likely to survive scrutiny.
The threat requirement in the present statute, however, does not successfully remove the First Amendment from the purview of the offense. In the absence of any nexus between the threat requirement and the conduct requirement, there is a real likelihood that the statute could chill the exercise of protected First Amendment expression. If the threat is the first predicate act, then any future conduct that is annoying (or harassing, alarming, etc.) could trigger the application of the statute — so long as both predicate acts were directed at the same person — even though the second predicate act consists entirely of protected First Amendment activity. This situation in essence holds protected activity hostage once a person engages in one proscribed act that is not protected. For example, a political protester who crosses the line and makes a threat might find himself forever barred from engaging in peaceful, legitimate expression for fear of subjecting himself to punishment for stalking. The legislature may legitimately punish the threat,
but it should not be permitted to hold someone’s First Amendment rights forever hostage after he makes such a threat.
Moreover, we believe that the First Amendment does not permit a legislature to create a new offense by simply adding unrelated protected activity to an ordinary criminal law. For example, the legislature could not pass a law making it a crime to (1) on one occasion physically assault a government official, and (2) on a separate occasion criticize the official’s policies. While physical assaults can surely be punished, criticism of official policies constitutes protected expression and does not become actionable simply because it is incorporated into an offense definition alongside unprotected activity. The same analysis should apply where the unrelated activity, while not clearly protected, would itself be impermissibly vague under First Amendment standards. In fact, under the present stalking provision, the assault/criticism scenario could satisfy the elements of the statute.
Another problem posed by adding protected activity to an otherwise ordinary criminal law is the potential for creating a greater offense whose only distinction from the lesser offense is the protected activity. Suppose, for example, the legislature decided to punish the hypothetical assauH/criticism crime outlined above — a crime to physically assault a public official
and
criticize the official’s policies — as a felony. The legislature might also decide to punish the mere physical assault of a public official as a misdemeanor. The only difference between the lesser and greater offenses in this example is that the greater offense contains an additional predicate act involving protected activity. Surely, the legislature may not be permitted to create such greater offenses whose only distinguishing characteristic — and hence sole justification for imposing a greater punishment — consists of activity protected by the First Amendment.
The stalking provision poses a similar problem. Stalking is punished as a Class A misdemeanor or as a felony if it is a repeat offense. § 42.07(d). Another provision exists that punishes a threat to inflict bodily injury or to commit a felony against a person. § 42.07(a)(2). An offense under (a)(2) is a Class B misdemeanor. § 42.07(c). Hence, a threat to commit
any
offense, including misdemeanors and felonies, becomes a higher grade offense than a threat to commit a felony merely through the addition of a predicate act meeting the requirements of (a)(7)(A), which we have already decided does not by itself confer adequate notice of the activity purportedly covered or give adequate guidance to law enforcement officials.
Ordinary people and law enforcement officials are left to distinguish between the two offenses based upon a vague provision that is capable of reaching into areas of First Amendment expression.
4.
The Affirmative Defense for Protected Activity
The remaining question is whether subsection (e)’s affirmative defense for constitutionally protected activity can save the statute. While posturing constitutionally protected activity as an affirmative defense appears unusual, most states that have adopted stalking laws appear to have some exemption for constitutionally protected activity.
See
Boychuck, M. Katherine.
Are Stalking Laws Unconstitutionally Vague or Overbroad?
88 Nw. U.L.Rev. 769, 791 (1994). One state has held that an exemption clause removed First Amendment implications from its stalking statute.
Luplow,
897 P.2d at 467. A couple of other state courts have held that such a clause, in combination with other limiting factors, achieves that purpose.
Bouters,
659 So.2d at 237.
People v. White,
212 Mich.App. 298, 536 N.W.2d 876, 883 (1995).
We do not believe that the affirmative defense'can save the present statute. As the Fifth Circuit has noted, a general savings provision “cannot substantively operate to save an otherwise invalid statute, since it is a mere restatement of well-settled constitutional restrictions on the construction of statutory enactments.”
CISPES,
770 F.2d at 474. While the provision would permit the defendant to introduce evidence before the jury regarding the constitutional nature of his conduct, it would relegate the First Amendment issue to a “case-by-case adjudication,” creating another vagueness problem.
Moyle,
705 P.2d at 748 n. 12. In essence, (a)(7)(A), as modified by the affirmative defense, would read something like “it is a crime to intentionally annoy someone unless by that conduct the actor engages in activity protected by the First Amendment.” Application of the affirmative defense to subsection (a)(7)(A) on a case-by-case basis would require people of ordinary intelligence — and law enforcement officials — to be First Amendment scholars. Arguably, people are always “on notice” that constitutionally protected conduct is exempt from prosecution, and law enforcement officials could always look to the First Amendment to determine when a law should not be enforced because it would interfere with constitutionally protected activity. But, the mere existence of the First Amendment has never been held automatically to cure vagueness problems implicating First Amendment freedoms. Because First Amendment doctrines are often intricate and/or amorphous, people should not be charged with notice of First Amendment jurisprudence, and a First Amendment defense cannot by itself provide adequate guidelines for law enforcement.
Moreover, an attempt to charge people with notice of First Amendment caselaw would undoubtedly serve to chill free expression.
Nevertheless, a savings clause may give courts more latitude to place a narrowing construction upon a statute that would save it from a constitutional challenge.
CISPES,
770 F.2d at 474.
Earley v. State,
789 P.2d 374, 376 n. 2 (Alaska App.1990)(protected speech excluded from definition of “noise”).
State v. Bagley,
164 Wis.2d 255, 474 N.W.2d 761, 764 (App.1991)(affirmative defense for protected speech reinforces conclusion that “interfere” refers only to physical disturbance). And we have a general duty to employ reasonable narrowing constructions to avoid constitutional violations.
Morehead v. State,
807 S.W.2d 577, 581 (Tex.Crim.App.1991).
Olvera v. State,
806 S.W.2d 546, 552 (Tex.Crim.App.1991). But, “a court may not assume the legislative prerogative and rewrite a statute in order to save it if the statute is not readily subject to a narrowing construction.”
Olvera,
806 S.W.2d at 552.
See also Morehead,
807 S.W.2d at 581. Nor may we sever from the statute an element of an offense where such action would broaden the scope of the statute and violate legislative intent.
Howard v. State,
617 S.W.2d 191, 192 n. 1 (Tex.Crim.App.1979).
Although we have never expressly considered whether a narrowing construction could be applied in the vagueness context (as opposed to overbreadth), at least one intermediate appellate court has indicated that a narrowing construction may be used to remove First Amendment considerations from the vagueness issue.
Al-Omari v. State,
673 S.W.2d 892, 896 (Tex.App.—Beaumont 1983, pet. ref d). We believe that view is a logical one. If a narrowing construction can remove the First Amendment from the arena, then the statute would be examined under more deferential vagueness standards and a facial challenge could be mounted successfully only if the statute were vague in all of its applications.
See Village of Hoffman Estates v. Flipside, Hoffman Estates, Inc.,
455 U.S. 489, 494-495, 102 S.Ct. 1186, 1191, 71 L.Ed.2d 362 (1982). Hence, while a narrowing construction would not by itself resolve the vagueness issue, it could change the character of the vagueness question and make obtaining relief more difficult.
A narrowing construction is most viable when a court can reasonably read the statute as excluding all expressive aspects from a particular kind of activity. Thus, noise can be limited to volume in a statute
prohibiting unreasonably loud noise, and interference can be limited to physical disturbance in a statute prohibiting interference with hunting and fishing.
See Earley
and
Bagley,
cited above. But, the stalking provision clearly contemplates the use of words as included within conduct. Subsection (a)(7)(B) states that “on at least one of those occasions [from (a)(7)(A) ] by acts or
words
threatens....” While that wording appears in (a)(7)(B), it seems clear from this phrase that “conduct” in (a)(7)(A) includes both acts and words. Moreover, words that might, alone, be protected may be rendered unprotected when connected with other words or activity. Senator Moncrief, author of the stalking bill, anticipated that the addition of a threat to otherwise protected activity might render the activity subject to prosecution. Hearings, House Criminal Jurisprudence Committee, February 15, 1993, tape 2, side B. To exclude entirely speech and expressive conduct from (a)(7)(A) would limit the statute in a way that the legislature did not appear to contemplate. Moreover, given the wide variety of intentionally annoying conduct that would contain expressive elements, and the difficulty of sorting expressive from nonex-pressive conduct in the stalking/harassment context, the vagueness problem would remain.
Nor can we reasonably construe the statute to contain a nexus requirement. There are no words present in the statute that would remotely suggest one. Further, a very minimal nexus requirement would probably not, by itself, result in a statute that would pass constitutional muster. For example requiring a threat and another instance of annoying conduct to relate to the same overall purpose would implicate the scenario in which a political protester crosses the line but then decides to confine future protests to peaceful demonstrations. A more ambitious nexus requirement would add significant content not now present in the statute and could be fashioned in a number of different ways. We will not rewrite the statutory language.
In the past, we have approved a narrowing construction that increased the intensity of the conduct under scrutiny. In
Morehead,
this Court construed the words “obstructs” and “interferes” as applying only to conduct which
substantially
impaired the ordinary conduct of lawful meetings. 807 S.W.2d at 577 (emphasis in original). We found that narrowing construction to be consonant with the “meeting disruption” statute’s apparent legislative purpose.
Id.
However, we did not attempt to save the harassment statute in
May
by intensifying the emotional terms present, and we do not believe it appropriate to do so here. While “annoy” and “embarrass” probably describe the least intense emotional states available under the statute, we think that “harass,” “alarm,” “abuse,” and “torment” are of low enough emotional intensity that they too implicate First Amendment freedoms. The legislature obviously intended low intensity emotions to trigger the statute. And, we think a constitutional statute can be drafted in which these low intensity emotional terms play a role. In the absence of statutory definitions, therefore, we will not interpret any of these terms as carrying an emotional intensity not suggested by the words themselves. To do otherwise would significantly alter the meaning of the statute the legislature drafted.
Finally, we are not in a position to excise a vague term from the statute. To do so, we would have to strike provision. (a)(7)(A). That action would leave (a)(7)(B) as the operative provision of the statute. Such an alteration would be significant and would broaden the reach of the statute by requiring only one instance of conduct. Further, because a felony threat is already subject to punishment as a class B misdemeanor, striking (a)(7)(A) from the stalking statute would result in an offense that imposes greater punishment (class A misdemeanor) for a misdemeanor threat than the legislature has expressly authorized for felony threats. Such a result was clearly not intended by the legislature.
C. Conclusion
In assessing the constitutionality of the statute before us, we have discussed several potential ways to clarify the statutory language. We express no opinion concerning whether any single limiting factor discussed would be sufficient to render the statute
constitutional. That may depend upon the nature of the limiting factor itself. We do note that many of the other jurisdictions utilize more than one additional limiting factor not contained in the Texas criminal stalking provision.
See
above authorities. Any such changes are the province of the legislature. We hold the 1993 stalking provision to be unconstitutionally vague on its face.
The judgment of the Court of Appeals is reversed, and the case is remanded to the trial court to enter an order dismissing the prosecution.
WHITE and MANSFIELD, JJ., not participating.