Nonce So Fast: Software “Structure” and Algorithmic Specificity in Computer-Implemented Means-plus-Function Patent Claims

by Zachary Hadd1

The patent sys­tem is based on a bar­gain between the paten­tee and the pub­lic.2 What the paten­tee receives in this bar­gain is clear: the right to exclude oth­ers from mak­ing, using, or sell­ing the inven­tion cov­ered by the patent.3 In exchange for this exclu­sive right, the paten­tee must ful­fill two duties. The first duty is to pro­vide the pub­lic with a descrip­tion of the inven­tion that enables one skilled in the art to make and use it.4 The sec­ond duty is to clear­ly define what the inven­tion is so that the pub­lic under­stands what tech­nol­o­gy they may prac­tice with­out infring­ing on the patentee’s rights.5 The paten­tee ful­fills both of these duties through the dis­clo­sures that they make in the patent document.

The patent doc­u­ment con­sists of two parts. The first part is the spec­i­fi­ca­tion, which includes a writ­ten descrip­tion and draw­ings of the inven­tion.6 The spec­i­fi­ca­tion pro­vides the enabling dis­clo­sure of the inven­tion and ful­fills the patentee’s first duty under the patent bar­gain.7 The sec­ond part of the patent doc­u­ment is the claims sec­tion.8 The claims define the inven­tion that the paten­tee has the exclu­sive right to prac­tice and ful­fills the patentee’s sec­ond duty under the patent bar­gain.9 If the paten­tee drafts their claim such that the ordi­nar­i­ly skilled per­son can­not deter­mine the patent’s scope of pro­tec­tion with rea­son­able cer­tain­ty, the claim is invalid for indef­i­nite­ness under § 112(b).10

Ordi­nar­i­ly, a paten­tee claims their inven­tion using struc­tur­al lan­guage that describes what the inven­tion is. The paten­tee is also free, how­ev­er, to claim their inven­tion by the func­tion that it per­forms, rather than by the struc­ture that per­forms that func­tion.11 His­tor­i­cal­ly, patent drafters have claimed func­tion­al fea­tures by claim­ing a “means” for per­form­ing the desired func­tion.12

The paten­tee who choos­es this route of func­tion­al claim­ing, how­ev­er, is not absolved of their duty to clear­ly com­mu­ni­cate to the pub­lic what the claim does and does not cov­er.13 Indeed, when Con­gress gave statu­to­ry approval for this practice—known as means-plus-func­tion claim­ing under 35 U.S.C. § 112(f)—it restrict­ed the scope of cov­er­age pro­vid­ed by the claimed “means” to only those struc­tures that the paten­tee dis­closed in the spec­i­fi­ca­tion as cor­re­spond­ing to the claimed func­tion.14 The struc­ture must be clear­ly linked or asso­ci­at­ed with the claimed func­tion.15 If a per­son of ordi­nary skill in the art would be unable to rec­og­nize the struc­ture dis­closed in the spec­i­fi­ca­tion and asso­ciate it with the appro­pri­ate func­tion, the means-plus-func­tion lim­i­ta­tion is indef­i­nite and the claim is invalid.16

Con­gress pro­vid­ed the statu­to­ry basis for means-plus-func­tion claim­ing in the 1952 Patent Act.17 By 1960, near­ly fifty per­cent of issued patents includ­ed at least one claim with a “means for” lim­i­ta­tion.18 This prac­tice con­tin­ued for the next forty years. Through the 1980s, the per­cent­age of issued inde­pen­dent patent claims uti­liz­ing means-plus-func­tion lan­guage hov­ered around—and some­times eclipsed—sixty per­cent.19 But over time, paten­tees rec­og­nized that courts inter­pret­ing these claims hon­ored the statu­to­ry stric­tures of § 112(f) and inter­pret­ed means-plus-func­tion claims as cov­er­ing only those “means” for per­form­ing the func­tion that were described in the patent doc­u­ment. The result was that means-plus-func­tion claims were con­strued nar­row­ly. Poten­tial infringers could avoid lia­bil­i­ty by find­ing a “means” for per­form­ing the claimed func­tion that was not dis­closed in the spec­i­fi­ca­tion. Worse yet, fail­ure to describe in the patent doc­u­ment any struc­ture for per­form­ing the claimed func­tion ren­dered the claim invalid as indef­i­nite. Rec­og­niz­ing that means-plus-func­tion claim­ing was a trap for the unwary drafter, the prac­tice fell out of favor as patent appli­cants sought to avoid the risk of a nar­row claim con­struc­tion or inva­lid­i­ty.20 By 2014, few­er than ten per­cent of issued patent claims were writ­ten in means-plus-func­tion form.21

But even as the prac­tice of means-plus-func­tion claim­ing declined, paten­tees con­tin­ued to claim their inven­tions by the func­tions that those inven­tions per­formed. As patent drafters attempt­ed to pre­serve their abil­i­ty to claim an inven­tion by what it did with­out falling into the “trap” of § 112(f) means-plus-func­tion claim­ing, oth­er forms of “func­tion­al claim­ing” pro­lif­er­at­ed.22 By 2014, approx­i­mate­ly half of the inde­pen­dent claims in issued patents uti­lized func­tion­al lan­guage by claim­ing some form of “struc­ture” for “per­form­ing a giv­en func­tion.”23 More recent­ly, patent drafters have uti­lized “con­fig­ured to” lan­guage in an attempt to cap­ture func­tion­al fea­tures of an inven­tion.24 Indeed, approx­i­mate­ly thir­ty per­cent of issued inde­pen­dent claims recit­ed struc­ture “con­fig­ured to” per­form a giv­en func­tion in 2014.25

Soft­ware patents are one of the bas­tions of func­tion­al claim­ing.26 Fre­quent­ly, soft­ware engi­neers can describe what they want soft­ware to do before they can explain pre­cise­ly how the soft­ware will do it.27 Accord­ing­ly, “[s]oftware patents typ­i­cal­ly … describe, in inten­tion­al­ly vague and broad lan­guage, a par­tic­u­lar goal or objec­tive [of the soft­ware].”28 Unsur­pris­ing­ly, func­tion­al claim­ing is par­tic­u­lar­ly well-suit­ed to the task of pro­tect­ing soft­ware-based inven­tions by describ­ing what the soft­ware does.

But not all func­tion­al soft­ware claims are cre­at­ed equal. The old­est com­put­er-imple­ment­ed patent claims being lit­i­gat­ed today would have been draft­ed when explic­it “means for” lan­guage was still used in near­ly thir­ty per­cent of issued patent claims.29 These claims pre­sump­tive­ly fall with­in the purview § 112(f), leav­ing it up to the patentee’s lit­i­ga­tors to extract the claims from the teeth of the means-plus-func­tion trap. Even claims draft­ed with an eye toward avoid­ing the nar­row­ing effects of means-plus-func­tion claim­ing are not free and clear of § 112(f), as dis­trict courts and the Fed­er­al Cir­cuit have wres­tled with how to han­dle the clever drafter that, instead of using the word “means,” instead uses terms like “mech­a­nism,” “mod­ule,” and oth­er nonce words. When claims can­not escape the purview of § 112(f), lit­i­gants argue about the pre­cise lev­el of struc­ture that must be dis­closed in the spec­i­fi­ca­tion to pre­serve the claim’s def­i­nite­ness and validity.

This Con­tri­bu­tion will argue that claim­ing soft­ware-imple­ment­ed inven­tions is not only facil­i­tat­ed by—but in fact requires—the use of func­tion­al claim­ing. While Fed­er­al Cir­cuit guid­ance on the issue sug­gests that paten­tees may avoid the ambit of § 112(f) by replac­ing the word “means” with a soft­ware-spe­cif­ic place­hold­er, the pre­emp­tion risk asso­ci­at­ed with soft­ware patents requires a high­er stan­dard. Soft­ware inven­tions claimed using func­tion­al lan­guage should be inter­pret­ed under § 112(f) even if soft­ware-spe­cif­ic “means” are recit­ed in the claim, and paten­tees’ pro­tec­tion should be lim­it­ed to the spe­cif­ic, step-by-step algo­rithm dis­closed in the specification.


The Fed­er­al Cir­cuit has laid out a two-step process for deter­min­ing how to ulti­mate­ly treat a claim sus­pect­ed of falling into means-plus-func­tion ter­ri­to­ry under § 112(f).30 At the first step, a court must deter­mine whether a claim is draft­ed in means-plus-func­tion for­mat. Only if the claim is found to be a means-plus-func­tion claim does the inquiry pro­ceed to the sec­ond step.

Deter­min­ing whether a claim is writ­ten in means-plus-func­tion for­mat in the first part of the inquiry goes beyond sim­ply deter­min­ing if the claim recites the phrase “means for.” A claim that includes the word “means” is pre­sumed to be a means-plus-func­tion claim and sub­ject to inter­pre­ta­tion under § 112(f). A claim that does not include the word “means” is pre­sumed to avoid inter­pre­ta­tion as a means-plus-func­tion claim and to not invoke § 112(f). But both of these pre­sump­tions are rebut­table. The pre­sump­tion that a claim falls under § 112(f) can be rebutted by show­ing that, despite using the word “means,” the claim recites suf­fi­cient struc­ture to per­form the entire­ty of the claimed func­tion.31 The pre­sump­tion that a claim does not fall under § 112(f) can be rebutted by show­ing that the claim term used—while not the word “means”—either fails to recite struc­ture that is suf­fi­cient­ly def­i­nite, or fails to recite struc­ture that can per­form the claimed func­tion.32

When the word “means” is not used, the dis­cus­sion quick­ly turns to whether what­ev­er word is used is a “nonce word” that sim­ply takes the place of the word “means”—perhaps in a delib­er­ate attempt to avoid to the purview of § 112(f). The Fed­er­al Cir­cuit has held that gener­ic terms such as “mech­a­nism,” “ele­ment,” “mod­ule,” and “device” are mere­ly ver­bal con­structs whose use is tan­ta­mount to use of the word “means” because they do not con­note suf­fi­cient­ly def­i­nite struc­ture and there­fore invoke § 112(f).33

Patent appli­cants seek­ing to claim soft­ware inven­tions face par­tic­u­lar dif­fi­cul­ty when try­ing to draft claims using func­tion­al lan­guage with­out invok­ing § 112(f). Refer­ring to the soft­ware as a “mod­ule” or “ele­ment” is almost cer­tain to sub­ject the claim to § 112(f) analy­sis. But soft­ware itself lacks a phys­i­cal form; attempt­ing to describe it in a man­ner that imparts struc­ture can be dif­fi­cult. What is the soft­ware patent appli­cant to do?

The Fed­er­al Cir­cuit recent­ly offered some guid­ance in Zeroclick, LLC v. Apple Inc.34 The Zeroclick court was con­front­ed with claims direct­ed to a “pro­gram” and to a “user inter­face code,” each for per­form­ing var­i­ous func­tions. The dis­trict court had treat­ed the claimed “pro­gram” and “user inter­face code” as nonce words, and accord­ing­ly inter­pret­ed the claims as means-plus-func­tion claims. On appeal, the Fed­er­al Cir­cuit vacat­ed the dis­trict court’s deci­sion on sev­er­al grounds. Most notable was the Fed­er­al Circuit’s expla­na­tion that the claimed “pro­gram” and “user inter­face code” were not used as “gener­ic terms or black box recita­tions of struc­ture or abstrac­tions, but rather as spe­cif­ic ref­er­ences to con­ven­tion­al graph­i­cal user inter­face pro­grams or code, exist­ing in pri­or art at the time of the inven­tions.”35 Impor­tant­ly, the Fed­er­al Cir­cuit did not explic­it­ly say that the terms “pro­gram” and “code” were not nonce terms. The court mere­ly held that the dis­trict court’s dis­missal of them as nonce terms was improper.

The term “engine,” like the terms “pro­gram” and “code,” is fre­quent­ly used in the soft­ware con­text to describe com­put­er pro­grams. In the com­put­ing con­text, Mer­ri­am-Web­ster defines an “engine” as “com­put­er soft­ware that per­forms a fun­da­men­tal func­tion espe­cial­ly of a larg­er pro­gram.”36 The Fed­er­al Cir­cuit has not yet tak­en up a case in which the pro­pri­ety of treat­ing the term “engine” as a nonce word was at issue. But when the East­ern Dis­trict of Texas con­front­ed that lan­guage in a claim to a net­work-based data­base sys­tem, it found that “the term ‘engine’ con­veys struc­ture to one of ordi­nary skill in the art” in the soft­ware con­text.37 Not­ing that the ordi­nar­i­ly skilled per­son would rec­og­nize an “engine” as a soft­ware appli­ca­tion or soft­ware mod­ule, the dis­trict court passed on the oppor­tu­ni­ty to con­strue the term as a nonce word and found that it did not invoke § 112(f).38

The risk of pre­emp­tion asso­ci­at­ed with soft­ware patents imparts impor­tance to these dis­tinc­tions. If a claim to a soft­ware-imple­ment­ed inven­tion uses func­tion­al lan­guage but is not inter­pret­ed in means-plus-func­tion form, the claim pre­sum­ably cov­ers all means for per­form­ing the claimed func­tion that fall with­in the claim as inter­pret­ed under Phillips v. AWH Corp.39 By not tying the claimed func­tion to any structure—algorithmic, or otherwise—broad soft­ware claims of this sort amount to pure­ly func­tion­al claim­ing and exclude oth­ers based on a desired effect, rather than the means for achiev­ing that effect.

The claim terms at issue in Zeroclick illus­trate this point well. In Zeroclick, the claimed “pro­gram” per­formed the func­tion of oper­at­ing the move­ment of a point­er over a screen in the con­text of a graph­i­cal user inter­face for a com­put­er or mobile phone.40 The Fed­er­al Cir­cuit found error in how the dis­trict court dis­re­gard­ed “pro­gram” as a nonce term.41 But how else could one oper­ate the move­ment of a point­er in a GUI if not by use of a pro­gram of some sort? Sim­i­lar­ly, the claimed “user inter­face code” per­formed two func­tions: detect­ing loca­tions touched by the user’s fin­ger on a screen with­out requir­ing the exer­tion of pres­sure, and deter­min­ing a select­ed oper­a­tion based on the loca­tions touched.42 If not through user inter­face code, how else would one detect a user’s inter­ac­tions with a user inter­face? Is code direct­ed to the user inter­face (i.e., user inter­face code) not the only means?

In both of these exam­ples, the “struc­ture” for per­form­ing the claimed func­tion is nec­es­sary for any imple­men­ta­tion of the claimed func­tion. It is too gen­er­al to have real mean­ing. As Pro­fes­sor Mark Lem­ley argues, the pur­pose of § 112(f) can only be real­ized if its struc­tur­al require­ment demands more than the recita­tion of inher­ent­ly nec­es­sary tech­nol­o­gy.43 While the words “pro­gram” and “user inter­face code” are dif­fer­ent from the word “com­put­er,” the prac­ti­cal effect is the same: all are nec­es­sary for per­form­ing the claimed func­tion. Allow­ing terms such as “pro­gram” and “code” to fall out­side of § 112(f) is tan­ta­mount to pure­ly func­tion­al claim­ing and effec­tive­ly pre­cludes all pos­si­ble man­ner of per­form­ing the claimed function.

This is not a crit­i­cism of func­tion­al claim­ing. In fact, it demon­strates why func­tion­al claim­ing is nec­es­sary to claim soft­ware-imple­ment­ed inven­tions: giv­en the dif­fi­cul­ty of claim­ing these inven­tions in phys­i­cal terms—or com­ing up with any sort of term for the soft­ware that does not envel­op every means for per­form­ing the claimed function—the inven­tion must be described by what it does. Scope lim­i­ta­tion is then achieved through the sec­ond step in the analy­sis of com­put­er-imple­ment­ed means-plus-func­tion claims, which lim­its func­tion­al claims accord­ing to the algo­rithms for per­form­ing the func­tions dis­closed in the patent doc­u­ment. This pre­vents pre­emp­tion and encour­ages robust dis­clo­sure on behalf of the patent applicant.


Dur­ing the sec­ond step of the means-plus-func­tion claim analy­sis, courts exam­ine the writ­ten descrip­tion of the patent for the struc­ture that cor­re­sponds to the claimed “means” and is capa­ble of per­form­ing the claimed func­tion. If the struc­ture is dis­closed such that the ordi­nar­i­ly skilled per­son would know and under­stand the struc­ture, the require­ments of § 112(f) are met. If struc­ture for per­form­ing the claimed func­tion is absent from the spec­i­fi­ca­tion, the claim is indef­i­nite under § 112(b).44 It is well-set­tled that in the con­text of com­put­er-imple­ment­ed inven­tions, the struc­ture cor­re­spond­ing to a means-plus-func­tion claim is an algo­rithm dis­closed in the patent doc­u­ment for per­form­ing the claimed func­tion.45 A com­put­er-imple­ment­ed means-plus-func­tion claim that does not cor­re­spond to an algo­rithm in the writ­ten descrip­tion fails to recite the cor­re­spond­ing struc­ture required by § 112(f) and is there­fore indef­i­nite under § 112(b).46

This require­ment ensures com­pli­ance with § 112(b) and pre­vents the pre­emp­tion of all poten­tial meth­ods of per­form­ing a giv­en func­tion. But because a com­put­er-imple­ment­ed means-plus-func­tion claim with­out an algo­rithm in the spec­i­fi­ca­tion cor­re­spond­ing to the claimed means is indef­i­nite, patent appli­cants and lit­i­gants alike often find them­selves argu­ing about whether the spec­i­fi­ca­tion includes the req­ui­site algo­rith­mic dis­clo­sure.47 So what is suf­fi­cient “algo­rith­mic struc­ture” in the com­put­er-imple­ment­ed means-plus-func­tion con­text? The Fed­er­al Cir­cuit has addressed this ques­tion on numer­ous occa­sions, but the answer remains unclear.

Here is what we do know: if the claimed func­tion is basic and con­sists of no more than the sim­ple func­tions that a gen­er­al pur­pose com­put­er can perform—receiving, stor­ing, and pro­cess­ing data—then dis­clos­ing a gen­er­al pur­pose com­put­er in the spec­i­fi­ca­tion is suf­fi­cient struc­ture for the claimed means and sat­is­fies § 112(b) and (f)—even if the spec­i­fi­ca­tion dis­clos­es no algo­rithm.48 But this is the exception—not the rule.49 If the claimed func­tion is not coex­ten­sive with the basic func­tions of a micro­proces­sor and requires spe­cial pro­gram­ming, dis­clo­sure of a mere gen­er­al pur­pose com­put­er is insuf­fi­cient.50 The algo­rithm for per­form­ing the func­tion must be dis­closed.51 But at what lev­el of abstraction?

A paten­tee may dis­close the required algo­rithm in “any under­stand­able terms includ­ing as a math­e­mat­i­cal for­mu­la, in prose, or as a flow chart, or in any oth­er man­ner that pro­vides suf­fi­cient struc­ture.”52 But stat­ing that a gen­er­al pur­pose com­put­er or stan­dard micro­proces­sor with “appro­pri­ate pro­gram­ming” can per­form the claimed func­tion is not suf­fi­cient.53 Like­wise, a gener­ic dis­clo­sure that “soft­ware” can be used to pro­gram a gener­ic com­put­er to per­form the claimed function—without pro­vid­ing some detail about the means to accom­plish the function—is not enough to meet the algo­rithm dis­clo­sure require­ment.54

Aside from these base­line qual­i­fi­ca­tions, one line of Fed­er­al Cir­cuit cas­es indi­cates that, ulti­mate­ly, the stan­dard for def­i­nite­ness in com­put­er-imple­ment­ed means-plus-func­tion claims is not a lofty one.55 In Med­ical Instru­men­ta­tion Diag­nos­tics Corp. v. Elek­ta AB, the Fed­er­al Cir­cuit con­firmed that soft­ware could be the cor­re­spond­ing struc­ture in a com­put­er-imple­ment­ed means-plus-func­tion claim.56 The court went on to sug­gest that as long as one skilled in the art would know the kind of soft­ware pro­gram to use to per­form the claimed func­tion, the spe­cif­ic pro­gram code of the soft­ware need not be dis­closed.57 The Med­ical Instru­men­ta­tion court went so far as to state that even sug­gest­ing that the claimed func­tion “can be per­formed by soft­ware pro­grams known to those of skill in the art” may be suf­fi­cient struc­tur­al dis­clo­sure to sat­is­fy the require­ments of § 112(f).58 And while the Med­ical Instru­men­ta­tion court repeat­ed­ly not­ed the specification’s dis­clo­sure of com­mer­cial­ly avail­able soft­ware, the court lament­ed that there was noth­ing in the spec­i­fi­ca­tion to tie these com­mer­cial­ly avail­able soft­ware pro­grams to the claimed func­tion.59 If the patent-in-suit in Med­ical Instru­men­ta­tion had sim­ply dis­closed that one of these com­mer­cial­ly avail­able soft­ware pro­grams could be used to per­form the claimed func­tion, would the Fed­er­al Cir­cuit have found the require­ments of § 112(f) met? The court seems to contemplate—and even suggest—that the answer is “yes.”60

Despite decades of caselaw on the books, Fed­er­al Cir­cuit prece­dent leaves the own­ers of soft­ware-based patents with uncer­tain­ty regard­ing how their claims will be inter­pret­ed. Still, paten­tees debate the lev­el of speci­fici­ty at which they must describe the soft­ware-imple­ment­ed “struc­ture” of the inven­tion to ensure valid­i­ty. If a paten­tee does find them­selves bound by the stric­tures of § 112(f)—be it inten­tion­al­ly, or by accident—does dis­clos­ing soft­ware that the ordi­nar­i­ly skilled per­son could use to per­form the claimed func­tion guar­an­tee the claim’s def­i­nite­ness? Despite the open-end­ed nature of Fed­er­al Cir­cuit caselaw on these issues, the broad­er con­text of the court’s deci­sions and the under­ly­ing pub­lic pol­i­cy ratio­nales demon­strate that the actu­al bur­dens on soft­ware paten­tees are greater than any­thing the Fed­er­al Cir­cuit has put in writing.

While acknowl­edg­ing that “algo­rithm” has broad mean­ing in the com­put­er arts, the Fed­er­al Cir­cuit defines the term today much the same as it did forty years ago: “a step-by-step pro­ce­dure for accom­plish­ing a giv­en result.”61 With this def­i­n­i­tion in mind, the Fed­er­al Circuit’s prece­dent on com­put­er-imple­ment­ed means-plus-func­tion claim­ing is right­ly read as requir­ing a step-by-step pro­ce­dure in the spec­i­fi­ca­tion for per­form­ing the claimed func­tion with a com­put­er. The scope of those claims is lim­it­ed to that step-by-step procedure.

It is true that a paten­tee may “express that algo­rithm in any under­stand­able terms.”62 The algo­rithm need not be dis­closed in the form of com­put­er code and need not even be “high­ly detailed.”63 In Typhoon Touch Techs., Inc. v. Dell, Inc., the court char­ac­ter­ized the state of Fed­er­al Cir­cuit prece­dent as requir­ing that “the patent need only dis­close suf­fi­cient struc­ture for a per­son of skill in the field to pro­vide an oper­a­tive soft­ware pro­gram for the spec­i­fied func­tion.”64 In light of the court’s seem­ing­ly for­giv­ing approach in Med­ical Instru­men­ta­tion, it is easy to read this pas­sage as bless­ing any com­put­er-imple­ment­ed means-plus-func­tion claim as long as an ordi­nar­i­ly skilled arti­san could pro­vide some soft­ware pro­gram for per­form­ing the func­tion. This would seem to cov­er com­mer­cial­ly avail­able soft­ware avail­able to the skilled arti­san, at least as long as that soft­ware is dis­closed in the specification.

How­ev­er, a more care­ful read­ing of Typhoon Touch illus­trates why this is not the case. The patent must dis­close suf­fi­cient struc­ture for the skilled arti­san to pro­vide an oper­a­tive soft­ware pro­gram for the claimed func­tion.65 The struc­ture is the algo­rithm.66 The algo­rithm is a step-by-step pro­ce­dure for per­form­ing the func­tion.67 The instruc­tion of Typhoon Touch, then, is that the step-by-step pro­ce­dure must be dis­closed with such speci­fici­ty that the skilled arti­san could pro­vide soft­ware to per­form the function.

This inter­pre­ta­tion makes sense in light of both the Fed­er­al Circuit’s deci­sion in Typhoon Touch and the court’s prece­dent lead­ing up to that case. The Typhoon Touch court ulti­mate­ly found that the claims in the patent-in-suit were def­i­nite and valid after return­ing to the spec­i­fi­ca­tion, iden­ti­fy­ing a four-step algo­rithm for per­form­ing the claimed func­tion, and find­ing that those steps could be read­i­ly imple­ment­ed by those skilled in the art using known com­put­er-imple­ment­ed operations—even after the patentee’s state­ment that “the spe­cif­ic algo­rithm con­not­ing the struc­ture of the means for [the func­tion] is not explic­it­ly dis­closed in the spec­i­fi­ca­tion.”68

This line of rea­son­ing is actu­al­ly a com­mon thread through the Fed­er­al Circuit’s prece­dent on com­put­er-imple­ment­ed means-plus-func­tion claim­ing. Before the Fed­er­al Circuit’s com­put­er-imple­ment­ed means-plus-func­tion doc­trine was dis­tilled down to the “struc­ture is the algo­rithm,”69 the court’s under­stand­ing of what § 112(f) required in the con­text of com­put­er-imple­ment­ed inven­tions was root­ed in an under­stand­ing of the tech­nol­o­gy itself. The court acknowl­edged that a micro­proces­sor pro­grammed to car­ry out an algo­rithm cre­ates a “new machine”: “[t]he instruc­tions of the soft­ware pro­gram in effect ‘cre­ate a spe­cial pur­pose machine for car­ry­ing out the par­tic­u­lar algo­rithm.’”70 In oth­er words, the algo­rithm and the soft­ware that instructs the com­put­er to per­form the algo­rithm are two dif­fer­ent things. And the notion that the algo­rithm is struc­tur­al is more than mere metaphor: the Fed­er­al Cir­cuit ground­ed this prece­dent on the man­ner in which “[t]he instruc­tions of the soft­ware pro­gram that car­ry out the algo­rithm elec­tri­cal­ly change the gen­er­al pur­pose com­put­er by cre­at­ing elec­tri­cal paths with­in the device.”71 The court envi­sioned the microprocessor’s indi­vid­ual tran­sis­tors open­ing and clos­ing as elec­tron­ic switch­es in accor­dance with the instruc­tions of the soft­ware pro­gram, there­by car­ry­ing out the desired func­tion accord­ing to the algo­rithm.72 The algo­rithm dic­tates the actu­al, phys­i­cal struc­ture of the microprocessor.

The pur­pose of the algo­rith­mic dis­clo­sure require­ment is to ful­fill the statu­to­ry bur­den of § 112(b) and “par­tic­u­lar­ly point[] out and dis­tinct­ly claim[] … the inven­tion.”73 Does dis­clos­ing an algo­rithm real­ly ful­fill that end any bet­ter than dis­clos­ing a spe­cif­ic soft­ware pro­gram? After all, the algo­rithm is sup­posed to pro­vide the cor­re­spond­ing struc­ture to the claim. But it is “[t]he instruc­tions of the soft­ware pro­gram [that] cause the switch­es [of the tran­sis­tor] to either open or close.”74 The soft­ware pro­gram cre­ates the elec­tri­cal paths with­in the com­put­er.75 The soft­ware pro­gram thus actu­al­ly “create[s] a spe­cial pur­pose machine for car­ry­ing out the par­tic­u­lar algo­rithm.”76 Since “gen­er­al pur­pose com­put­ers can be pro­grammed to per­form very dif­fer­ent tasks in very dif­fer­ent ways,”77 wouldn’t dis­clo­sure of the software—the spe­cif­ic pro­gram­ming used to per­form the claimed task—actually point out and claim the inven­tion as par­tic­u­lar­ly and dis­tinct­ly as possible?

Pos­si­bly. But the require­ment that the algo­rithm itself be disclosed—and that the require­ment can­not be met by only dis­clos­ing the soft­ware that per­forms the algorithm—is still the bet­ter approach for two reasons.

The first involves the “equiv­a­lents” pro­vi­sion of § 112(f). Under § 112(f), means-plus-func­tion claims cov­er the cor­re­spond­ing struc­ture in the spec­i­fi­ca­tion and equiv­a­lents there­of.78 If software—even if dis­closed as a spe­cif­ic program—were allowed to define the cor­re­spond­ing struc­ture, courts would be left to deter­mine the equiv­a­lents of the spe­cif­ic pro­gram dis­closed. This would open the door for paten­tees to con­vince courts that a broad swath of soft­ware programs—even those that use unre­lat­ed algo­rith­mic logic—were equiv­a­lent to the soft­ware dis­closed by the paten­tee and there­fore fall with­in the scope of the patent’s pro­tec­tion.79 A paten­tee suc­cess­ful in this strat­e­gy would inch clos­er to—and per­haps break through—the bar­ri­er to pure­ly func­tion­al claim­ing that § 112(f) sought to erect.

The sec­ond rea­son is direct­ed to ful­fill­ing the notice func­tion of the patent sys­tem. Patents are pub­lic doc­u­ments and serve to inform oth­ers in the same tech­no­log­i­cal space as the paten­tee about those tech­nolo­gies the paten­tee has the exclu­sive right to prac­tice. If a paten­tee dis­clos­es a soft­ware pro­gram as the means for per­form­ing the claimed func­tion, then all poten­tial infringers will be on notice that using that pro­gram to per­form the claimed func­tion would fall with­in the patentee’s rights. But what about alter­na­tives that are not dis­closed in the patent doc­u­ment? Are poten­tial infringers expect­ed to dig into the source code of the named soft­ware to deter­mine the inner work­ings of the pro­gram, deter­mine all soft­ware that may be deemed equiv­a­lent, and avoid the entire bunch? This seems like an unre­al­is­tic bur­den to place on the public.

Requir­ing dis­clo­sure of an algorithm—a step-by-step process for per­form­ing the claimed function—resolves both of these issues sim­ply. Because an algo­rithm will gen­er­al­ly be a broad­er con­cept than even the code used to imple­ment the algo­rithm, the scope of the patentee’s inven­tion will be broad­er. But this means that it will also be eas­i­er to iden­ti­fy what falls with­in the doc­trine of equiv­a­lents, as well as algo­rith­mic means that are not equiv­a­lent to what was dis­closed by the paten­tee. The notice require­ment is also eas­i­ly ful­filled: the algo­rithm dis­closed by the paten­tee is pro­tect­ed. Oth­er means for per­form­ing the claimed func­tion, using dif­fer­ent algo­rithms, are not.


In Williamson v. Cit­rix Online, LLC, the Fed­er­al Cir­cuit not­ed the recent “pro­lif­er­a­tion of func­tion­al claim­ing unteth­ered to § 112, para. 6 and free of the stric­tures set forth in the statute.”80 It has long been the Fed­er­al Circuit’s goal to “avoid pure func­tion­al claim­ing.”81 Pure­ly func­tion­al claim­ing pre­empts all pos­si­ble ways of per­form­ing a giv­en func­tion, grants paten­tees pro­tec­tion for accom­plish­ing the func­tion using means that they did not con­ceive of, and ulti­mate­ly hin­ders innovation.

Inven­tors of soft­ware-imple­ment­ed inven­tions do face an uphill bat­tle when seek­ing patent pro­tec­tion. But deci­sions like that of the Fed­er­al Cir­cuit in Zeroclick and the East­ern Dis­trict of Texas in Stra­gent risk allow­ing the pen­du­lum to swing too far in the oppo­site direc­tion. Claim terms like “pro­gram,” “code,” and “engine” should be inter­pret­ed as nonce words devoid of any struc­ture and sub­ject claims to means-plus-func­tion inter­pre­ta­tion under 35 U.S.C. § 112(f).

For com­put­er-imple­ment­ed means-plus-func­tion claims, the Fed­er­al Cir­cuit has been clear on one point: the algo­rithm is the struc­ture. But in order to under­stand what the Fed­er­al Cir­cuit bless­es as an algo­rithm, cas­es like Med­ical Instru­men­ta­tion and Typhoon Touch must be read in con­text. Despite pas­sages in these cas­es sug­gest­ing that the dis­clo­sure of soft­ware usable for per­form­ing the claimed func­tion pro­tects the claim’s def­i­nite­ness, the touch­stone of the Fed­er­al Circuit’s inquiry is its under­stand­ing of an algo­rithm from 40 years ago: “a step-by-step pro­ce­dure for accom­plish­ing a giv­en result.”82 In order to clear­ly define the metes and bounds of the patent­ed inven­tion, as well as equiv­a­lent alter­na­tives, the step-by-step pro­ce­dure for achiev­ing the claimed func­tion must be dis­closed with specificity.

1. Zachary Hadd is a J.D. Can­di­date (2021) at New York Uni­ver­si­ty School of Law. This piece is a com­men­tary on the 2020 prob­lem pre­sent­ed at the Giles Suther­land Rich Memo­r­i­al Moot Court Com­pe­ti­tion in Boston, MA. Oral argu­ment was held remote­ly due to ongo­ing efforts to stem the spread of COVID-19. The prob­lem addressed the def­i­nite­ness require­ments for com­put­er-imple­ment­ed means-plus-func­tion claims. The views expressed in this arti­cle do not nec­es­sar­i­ly rep­re­sent the views of the author on this point of law. Rather, this arti­cle is a dis­til­la­tion of the argu­ments made by the team on one side of the issue at the Giles Suther­land Rich Memo­r­i­al Moot Court Competition.

2. See 35 U.S.C. § 101 (“Who­ev­er invents or dis­cov­ers any new and use­ful process, machine, man­u­fac­ture, or com­po­si­tion of mat­ter, or any new and use­ful improve­ment there­of, may obtain a patent there­for, sub­ject to the con­di­tions and require­ments of this title.”); 35 U.S.C. § 112 (defin­ing dis­clo­sure requirements).

3. See 35 U.S.C. § 271(a) (defin­ing the statu­to­ry con­di­tions of patent infringement).

4. 35 U.S.C. § 112(a).

5. See Warn­er-Jenk­in­son Co. v. Hilton Davis Chem­i­cal Co., 520 U.S. 17, 33 (1997) (dis­cussing “the role of claims in defin­ing an inven­tion and pro­vid­ing pub­lic notice”); 35 U.S.C. § 112(b) (“The spec­i­fi­ca­tion shall con­clude with one or more claims par­tic­u­lar­ly point­ing out and dis­tinct­ly claim­ing the sub­ject mat­ter which the inven­tor or a joint inven­tor regards as the invention.”).

6. See, e.g., Gen­er­al Foods Corp. v. Stu­di­enge­sellschaft Köh­le mbH, 972 F.2d 1272, 1274 (Fed. Cir. 1992) (describ­ing the spec­i­fi­ca­tion and the claims as the “two pri­ma­ry parts” of the patent document).

7. See id.; 35 U.S.C. § 112(a).

8. Gen­er­al Foods, 972 F.2d at 1274.

9. See id.; 35 U.S.C. § 112(b).

10. Nau­tilus, Inc. v. Biosig Instru­ments, Inc., 572 U.S. 898, 901 (2014).

11. See 35 U.S.C. § 112(f) (“An ele­ment in a claim for a com­bi­na­tion may be expressed as a means or step for per­form­ing a spec­i­fied func­tion with­out the recital of struc­ture, mate­r­i­al, or acts in sup­port there­of, and such claim shall be con­strued to cov­er the cor­re­spond­ing struc­ture, mate­r­i­al, or acts described in the spec­i­fi­ca­tion and equiv­a­lents thereof.”).

12. See Warn­er-Jenk­in­son Co. v. Hilton Davis Chem­i­cal Co., 520 U.S. 17, 18 (1997) (explain­ing the his­to­ry of means-plus-func­tion claim­ing rel­a­tive to the 1952 Patent Act); Con­ti­nen­tal Paper Bag Co. v. East­ern Paper Bag Co., 210 U.S. 405, 417 (1908) (eval­u­at­ing func­tion­al “means for” lan­guage in a claim to a paper bag machine).

13. See Northrop Grum­man Corp. v. Intel Corp., 325 F.3d 1346, 1350–51 (Fed. Cir. 2003) (explain­ing claim con­struc­tion of means-plus-func­tion claims).

14. See id.

15. See Med. Instru­men­ta­tion & Diag­nos­tics Corp. v. Elek­ta AB, 344 F.3d 1205, 1219 (Fed. Cir. 2003) (“The require­ment that struc­ture must be clear­ly linked or asso­ci­at­ed with the claimed func­tion is the quid pro quo for the con­ve­nience of claim­ing in func­tion­al terms.”).

16. See Noah Sys., Inc. v. Intu­it Inc., 675 F.3d 1302, 1311-12 (Fed. Cir. 2012) (cita­tions omitted).

17. 35 U.S.C. § 112(f).

18. See Den­nis Crouch, The Fre­quen­cy of Means-Plus-Func­tion Claims, Patent­lyO (Jul. 25, 2011),

19. See Den­nis Crouch, Func­tion­al Claim Lan­guage in Issued Patents, Patent­lyO (Jan. 23, 2014),

20. See id.

21. See id.

22. See id.

23. For exam­ple, “a fas­ten­er for cou­pling a first part to a sec­ond part.”

24. For exam­ple, “a fas­ten­er con­fig­ured to cou­ple a first part to a sec­ond part.”

25See Crouch, supra note 19.

26. See Intel­lec­tu­al Ven­tures I LLC v. Syman­tec Corp., 838 F.3d 1307, 1327–28 (Fed. Cir. 2016) (May­er, J., con­cur­ring) (not­ing “the vast num­ber of soft­ware patents—most of which are replete with broad, func­tion­al claims”).

27. See id. at 1327 (“Engi­neers can describe what they want soft­ware to do—in terms that have been suf­fi­cient for the PTO—well before they have made it work.” (quot­ing Wendy Seltzer, Soft­ware Patents and/or Soft­ware Devel­op­ment, 78 Brook. L. Rev. 929, 972 (2013))).

28. Id. at 1327 (cit­ing Dan L. Burk & Mark A. Lem­ley, Is Patent Law Tech­nol­o­gy-Spe­cif­ic?, 17 Berke­ley Tech. L. J. 1155, 1164–65 (2002)).

29. See Crouch, supra note 19.

30. See Apple Inc. v. Motoro­la, Inc., 757 F.3d 1286, 1296 (Fed. Cir. 2014), over­ruled on oth­er grounds by Williamson v. Cit­rix Online, LLC, 792 F.3d 1339 (Fed. Cir. 2015).

31. Per­son­al­ized Media Commc’ns, LLC v. Int’l Trade Com­m’n, 161 F.3d 696, 703–04 (Fed. Cir. 1998) (cit­ing Sage Prods. v. Devon Indus., Inc., 126 F.3d 1420, 1427–28 (Fed. Cir. 1997)).

32. Williamson, 792 F.3d at 1349 (quot­ing Watts v. XL Sys., Inc., 232 F.3d 877, 880 (Fed. Cir. 2000)).

33. Id. at 1350 (cit­ing Mass. Inst. of Tech. v. Aba­cus Soft­ware, 462 F.3d 1344, 1354 (Fed. Cir. 2006)).

34. Zeroclick, LLC v. Apple Inc., 891 F.3d 1003, 1008 (Fed. Cir. 2018).

35. Id.

36. Engine,,

37. Stra­gent, LLC v., Inc., No. 6:10CV225 LED-JDL, 2011 WL 13152568, at *2, *4 (E.D. Tex. June 27, 2011).

38. Id. at *4.

39. Phillips v. AWH Corp., 415 F.3d 1303 (Fed. Cir. 2005).

40. Zeroclick, LLC v. Apple Inc., 891 F.3d 1003, 1005 (Fed. Cir. 2018).

41. Id. at 1008.

42. Id. at 1006.

43. Mark A. Lem­ley, Soft­ware Patents and the Return of Func­tion­al Claim­ing, 2013 Wis. L. Rev. 905, 946 (2012).

44. See Atmel Corp. v. Info. Stor­age Devices, Inc., 198 F.3d 1374, 1381–82 (Fed. Cir. 1999).

45. See Har­ris Corp. v. Eric­s­son Inc., 417 F.3d 1241, 1253 (Fed. Cir. 2005) (“A com­put­er-imple­ment­ed means-plus-func­tion term is lim­it­ed to the cor­re­spond­ing struc­ture dis­closed in the spec­i­fi­ca­tion and equiv­a­lents there­of, and the cor­re­spond­ing struc­ture is the algorithm.”).

46. See EON Corp. IP Hold­ings LLC v. AT & T Mobil­i­ty LLC, 785 F.3d 616, 622 (Fed. Cir. 2015) (affirm­ing inva­lid­i­ty of means-plus-func­tion claims with no cor­re­spond­ing algo­rith­mic in the spec­i­fi­ca­tion); Aris­to­crat Techs. Austl. Pty Ltd. v. Int’l Game Tech., 521 F.3d 1328, 1338 (Fed. Cir. 2008) (same).

47. See, e.g., Har­ris, 417 F.3d at 1253–54 (con­sid­er­ing argu­ments regard­ing algo­rith­mic disclosure).

48. See EON, 785 F.3d at 621 (cit­ing In re Katz Inter­ac­tive Call Pro­cess­ing Patent Lit­ig., 639 F.3d 1303 (Fed. Cir. 2011)).

49. See id. (“[T]he ‘nar­row’ Katz excep­tion . . . .”).

50. See id. at 622.

51. See id. at 623.

52. Fin­is­ar Corp. v. DirecTV Grp., Inc., 523 F.3d 1323, 1340 (Fed. Cir. 2008) (inter­nal cita­tions omitted).

53. See Aris­to­crat Techs. Austl. Pty Ltd. v. Int’l Game Tech., 521 F.3d 1328, 1334 (Fed. Cir. 2008) (“The term ‘appro­pri­ate pro­gram­ming’ sim­ply ref­er­ences a com­put­er that is pro­grammed so that it per­forms the func­tion in ques­tion, which is to say that the func­tion is per­formed by a com­put­er that is capa­ble of per­form­ing the function.”)

54. See Fin­is­ar, 523 F.3d at 1340–41 (“Sim­ply recit­ing ‘soft­ware’ with­out pro­vid­ing some detail about the means to accom­plish the func­tion is not enough [to ensure definiteness].”).

55. See id. at 1341 (cit­ing Med. Instru­men­ta­tion & Diag­nos­tics Corp. v. Elek­ta AB, 344 F.3d 1205, 1214 (Fed. Cir. 2003)).

56. Med. Instru­men­ta­tion, 344 F.3d at 1214.

57. Id.

58. Id. at 1217.

59. Id. at 1217–18.

60. See Aris­to­crat Techs. Austl. Pty Ltd. v. Int’l Game Tech., 521 F.3d 1328, 1337 (Fed. Cir. 2008) (“[T]he prop­er inquiry for pur­pos­es of sec­tion 112 para­graph 6 analy­sis is to ‘look at the dis­clo­sure of the patent and deter­mine if one of skill in the art would have under­stood that dis­clo­sure to encom­pass soft­ware [to per­form the func­tion] and been able to imple­ment such a pro­gram, not sim­ply whether one of skill in the art would have been able to write such a soft­ware pro­gram.’” (quot­ing Med. Instru­men­ta­tion, 344 F.3d at 1212)).

61. Typhoon Touch Techs., Inc. v. Dell, Inc., 659 F.3d 1376, 1385 (Fed. Cir. 2011) (quot­ing In re Free­man, 573 F.2d 1237, 1246 (C.C.P.A. 1978)).

62. Fin­is­ar Corp. v. DirecTV Grp., Inc., 523 F.3d 1323, 1340 (Fed. Cir. 2008) (quot­ing In re Free­man, 573 F.2d at 1245–46).

63. Typhoon Touch, 659 F.3d at 1385–86 (quot­ing Aris­to­crat, 521 F.3d at 1338).

64. Id. at 1385 (cit­ing Fin­is­ar, 523 F.3d at 1340).

65. See id.

66. Har­ris Corp. v. Eric­s­son Inc., 417 F.3d 1241, 1253 (Fed. Cir. 2005).

67. Typhoon Touch, 659 F.3d at 1385.

68. Id. at 1386.

69. Har­ris, 417 F.3d at 1253.

70. Aris­to­crat Techs. Austl. Pty Ltd. v. Int’l Game Tech., 521 F.3d 1328, 1333 (Fed. Cir. 2008) (cit­ing WMS Gam­ing, Inc. v. Int’l Game Tech., 184 F.3d 1339, 1348 (Fed. Cir. 1999)).

71. WMS, 184 F.3d at 1348.

72. Id. at 1348 n.3 (cit­ing Neil Ran­dall, Dis­sect­ing the Heart of Your Com­put­er, PC Mag., June 9, 1998, at 254–55).

73. 35 U.S.C. § 112(b)

74. WMS, 184 F.3d at 1348 n.3.

75. Id. at 1348.

76. Id.

77. Aris­to­crat Techs. Austl. Pty Ltd. v. Int’l Game Tech., 521 F.3d 1328, 1333 (Fed. Cir. 2008).

78. 35 U.S.C. § 112(f).

79. See Lem­ley, supra note 43, at 951.

80. Williamson v. Cit­rix Online, LLC, 792 F.3d 1339, 1349 (Fed. Cir. 2015).

81. Aris­to­crat, 521 F.3d at 1333.

82. Typhoon Touch Techs., Inc. v. Dell, Inc., 659 F.3d 1376, 1385 (Fed. Cir. 2011) (quot­ing In re Free­man, 573 F.2d 1237, 1246 (C.C.P.A. 1978)).