TY - JOUR
AU - Slocum, Brian G.
AB - Abstract Undoubtedly, the most fundamental axiom of legal interpretation is that words in texts should receive their ‘ordinary meaning’. Notwithstanding its talismanic invocation by the judiciary, the ordinary meaning concept has been poorly developed by courts. In particular, courts rely on a meaning postulate approach where they define words without regard to critical sentential (i.e. sentence level) context. The reality, though, is that meaning is distributed in nature. By ignoring semantic regularities, courts risk determining the ‘possible’ meaning of text rather than its ordinary meaning. Particularly in the area of statutory interpretation, such an approach contravenes many of the judiciary’s other interpretive tools, which consistently favor stability, though narrow interpretations, over change. This article examines a paradigmatic set of Supreme Court cases that illustrate the judiciary’s unsystematic approach to ordinary meaning determinations and advocates an interdisciplinary approach to ordinary meaning that incorporates developing cognitive science and linguistic insights. Specifically, this article argues that courts should consider the generative lexicon (GL) theory of lexical semantics when attempting to define the ordinary meaning of text. GL focuses on capturing valid generalizations about words, and its generative mechanisms seek to produce interpretations that reflect words’ combinatorial properties. Instead of the judiciary’s current ad hoc approach, GL should be used to generate default ordinary meaning interpretations that would be subject to shaping by the judiciary’s other interpretive tools. By adopting GL, and thereby taking seriously cognitive science and linguistic advances, courts would be embracing a true ‘linguistic turn’ in legal interpretation. 1. Introduction The concept that words and phrases are to be given their ‘ordinary meaning’ is considered to be so foundational by courts that it is applied when interpreting virtually every type of legal text.1 With respect to the interpretation of statutes, the focus of this article, the determination of ordinary meaning, as prototypically the first step in the interpretive process, is crucial to the ultimate interpretation chosen by a court.2 The ordinary meaning of the language typically supplies the foundation for the application of the judiciary’s other interpretive tools, as well as the boundaries of interpretive acceptability. Notwithstanding its centrality to legal interpretation, and the purely linguistic nature of it, the ordinary meaning concept has not been developed and systematized by the judiciary. The concept is based on the axiom that legislative expressions should be interpreted according to accepted standards of communication.3 Courts generally do not, however, have good answers to the question of what makes a given meaning the ordinary one, as opposed to an inappropriately broad or narrow or technical meaning.4 In fact, the sources courts consider in determining ordinary meaning often favor inappropriately broad meanings because they capture ‘possible’ rather than ‘ordinary’ meanings, which, due to the creativity of language, may result in unlimited potential meanings. The flaw in the judiciary’s methodology is that by endemically ignoring the semantic contribution of all the lexical items in a sentence, insufficient weight is given to crucial sentential context in determining the semantics of individual words. The judicial determination of ordinary meaning follows a predictable, and problematic, process. Courts typically determine the ordinary meaning of a word by consulting a dictionary definition5 or often by just relying on ‘common sense’ or personal judgments about usage.6 Little needs to be said about the subjective nature of commonsense-based judicial assertions of meaning, and the flawed nature of dictionary usage by the judiciary is well known. Even under the best of circumstances, the use of dictionaries warrants caution.7 A dictionary definition is, after all, just a meaning postulate.8 It offers a way of showing the meaning of a word or expression by means of some other words which ‘say the same thing’, raising the problem of circularity.9 Furthermore, most words are polysemous, having multiple distinct but related senses.10 Due to the polysemous nature of words, the numerosity of available definitions allows judges to engage in ‘dictionary shopping’ in order to find a desired definition.11 Typically, these definitions are chosen without systematic regard to context.12 Although semanticists disagree about many fundamental issues, there is widespread agreement that lexical items carry many dimensions of meaning and that a particular interpretation is brought about in a given context.13 To be sure, courts purport to follow a contextual approach when defining words. The context considered by courts, though, is typically broader than, and does not include, sentential context. Usually, as reflected by the in pari materia canon, courts consider the relevant context to be how the provision fits into the statutory scheme as a whole.14 Courts may also choose meanings based on the ‘purpose or object to accomplish’ of the relevant statute or information gleaned from the statute’s legislative history.15 One might, correctly, characterize these contextual considerations as underscoring courts’ ad hoc textual interpretations with debatable, and often fictional, assertions about statutory purpose, congressional intent, and uniformity with other statutes.16 However courts’ contextual considerations are framed, the judiciary frequently fails to realize that words are polysemous and that the sentential context of a word form may influence its semantic contribution to the sentence.17 Thus, any systematicity in the judiciary’s methodological approach is based on a failure to understand the compositional nature of language (i.e. the way words combine to form meanings).18 Making matters worse, the judiciary’s ad hoc approach to ordinary meaning determinations fails to take advantage of developing cognitive science and linguistic knowledge, even though texts, in comparison to oral communication, bring a greater capacity for precision and scientific objectivity.19 For example, cognitive scientists have shown that the ellipsis (i.e. omission) of words is common, and cognitively unsurprising, in certain situations, such as when the ellipsed information is commonly associated with the related noun.20 These cognitive science findings should have implications for ordinary meaning interpretations because they illustrate the error inherent in adopting a broad dictionary definition without regard to sentential context such as ellipsis, which may indicate that a narrower interpretation should be viewed as the ordinary one. Because of the deficiencies in its current approach to ordinary meaning, the judiciary is plainly in need of a theoretical framework that can account for the polysemous nature of words and the distributed nature of meaning and that will encode selectional knowledge (i.e. which co-occurrences of words or meanings of words are semantically well formed). As the most advanced approach among the formal componential theories of semantics, the generative lexicon (GL) theory of lexical semantics developed by James Pustejovsky can provide such a framework.21 The stated function of GL is to establish a formal account of logical polysemy, which occurs when a word carries different meanings in different contexts.22 GL aims to capture generalizations about words and focuses on the combinatorial properties of words rather than merely developing accurate definitions.23 GL thus offers a generative account of word meaning and interpretation that focuses on the distributed nature of meaning in natural language.24 In producing meanings, GL first employs feature structures that represent the nature of a lexical item’s semantic representation and then applies a number of generative devices that are used to contextually construct semantic expressions.25 Thus, unlike the simple enumeration of meanings in the lexicon, which characterizes the judiciary’s current approach, GL attempts to spread the semantic load across all constituents of the utterance. The goal is to elevate the status of the various sentential components from being passive arguments to active functions in the production of meaning.26 As this article explains, GL offers a descriptive, rather than a normative, theory of language.27 Further, although it focuses on text, GL should not be seen as mandating a textualist interpretive methodology.28 Rather, as conceptualized in this article, GL’s generative mechanisms should be used to produce prototypical readings that would serve as default ordinary meaning interpretations. These default meanings can be shaped by whatever interpretive considerations the court determines are relevant, including legislative history, statutory purpose, and other interpretive tools.29 GL thus addresses only the ordinary meaning concept, which is uncontroversially a key component of how all courts interpret statutes, and does not implicate either textualism or the ‘plain meaning doctrine’, which concern the nonconsideration of contextual evidence.30 Despite this article disclaiming any effort to create a normative theory of interpretation, it is hoped that legal interpretation, like the “linguistic turn” that marked the development of philosophy in the last century, will eventually be viewed as an intrinsically linguistic phenomenon subject to linguistic insights, operations, and advances.31 While statutory interpretation theory and methodology is a growing area of scholarly inquiry, the concept of ordinary meaning (like other linguistic phenomena) receives little attention from legal scholars, who prefer to focus on doctrines and issues that are ideological in nature.32 It is, of course, true that interpretation is inherently discretionary and hence subject to the policy- and ideologically based predilections of judges. This article does not dispute the oft-mentioned criticism that courts manipulate even ideologically neutral interpretive tools, including the ordinary meaning concept, in order to achieve favored results.33 Notwithstanding this legal realist insight, courts purport to take textual language seriously, and this article provides a theory of how they might better do so. Section 2 of this article provides a paradigmatic example of the Supreme Court’s current approach to ordinary meaning interpretation. This section describes the Court’s remarkable and repeated misinterpretation of a deceptively simple statute, 18 USC §924(c)(1)(A), that is underspecified due to ellipsis. It also describes important cognitive science findings that could have assisted the Court in its interpretations. Section 3 explains the components of GL and describes how the theory is based on the application of a set of generative operations over complex lexical entries. Section 4 applies GL to §924(c)(1)(A) and outlines a methodologically proper interpretation. Section 5 proves a typology of the various criticisms of GL made by scholars and argues that these objections do not warrant rejection of GL. This section argues that GL can help distinguish common meanings from situations of lexical creativity and can produce default ordinary meaning interpretations that can be shaped, when appropriate, by pragmatic considerations. Finally, section 6 argues that despite its complexity, GL is superior to any alternatives, including other formal theories of meaning. The superiority of GL is based, in part, on its unique congruence with the current interpretive tools used by courts. 2. The Supreme Court and Ordinary Meaning A Case Study: The Supreme Court’s Misinterpretation of 18 USC §924(c)(1)(A) The best sense of the possible role for GL in ordinary meaning determinations can be had through its application to a paradigmatic set of cases in which the Court failed to consider cognitive science insights and the distributed nature of compositionality.34 The focus will thus be on the Court’s interpretations of a federal statute, 18 USC §924(c)(1)(A), that provides for enhanced punishment if the defendant ‘uses’ a firearm ‘during and in relation to … [a] drug trafficking crime’.35 Although the statute specifies the subject or agent (the defendant), the direct object (the firearm), and requires a connection to a drug trafficking crime, it is underspecified regarding the event; how the defendant must ‘use[]’ the firearm within the meaning of the provision.36 This underspecification, a paradigmatic example of underspecificity due to ellipsis, has challenged courts in their efforts to give the ordinary meaning to the statutory language.37 Remarkably for a routine criminal statute providing a penalty enhancement and presenting no constitutional issues, §924(c)(1)(A) has been interpreted by the Supreme Court on multiple occasions. In Smith v. United States,38 the Court held that the statute does not require that the firearm has been used as a weapon. The Court explained that when a word is not defined by statute, as most are not, courts normally construe it in accordance with its ‘ordinary or natural meaning’.39 The Court stated that exchanging a firearm for drugs ‘can be described as “use” within the everyday meaning of that term’.40 The Court consulted two dictionaries regarding the word use and concluded that it means ‘to employ’ or ‘to derive service from’.41 The Court rejected the argument that uses has a reduced scope in §924(c)(1)(A) because it appears alongside the word firearm.42 The Court reasoned that ‘it is one thing to say that the ordinary meaning of “uses a firearm” includes using a firearm as a weapon’, ‘[b]ut it is quite another to conclude that, as a result, the phrase also excludes any other use’.43 Thus, because ‘one can use a firearm in a number of ways’, ‘[t]hat one example of ‘use’ is the first to come to mind … does not preclude us from recognizing that there are other “uses” that qualify as well’.44 Due to the broad meaning of use, the Court concluded that the statute’s language ‘sweeps broadly, punishing any “us[e]” of a firearm, so long as the use is “during and in relation to” a “drug trafficking offense”’.45 Therefore, the Court reasoned, ‘it is both reasonable and normal to say that [the defendant] “used” his MAC-10 in his drug trafficking offense by trading it for cocaine’.46 In the Court’s view, if Congress had intended that the firearm must have been used as a weapon in order for the enhanced punishment to apply, it could have included the words ‘as a weapon’ in the statute.47 In dissent, Justice Scalia criticized the Court’s failure to properly consider context in determining the ordinary meaning of ‘use’. First, Justice Scalia pointed out the ‘elastic’ nature of the word ‘use’.48 Secondly, Justice Scalia pointed out, correctly, that ‘[t]o use an instrumentality ordinarily means to use it for its intended purpose’.49 Thus, ‘to speak of “using a firearm” is to speak of using it for its distinctive purpose, i.e., as a weapon’.50 In Justice Scalia’s view, the words ‘as a weapon’ were ‘reasonably implicit’ from the context of the statute.51 Two years after Smith, the Supreme Court held in Bailey v. United States52 that §924(c) requires active employment of the firearm by the defendant. In other words, the firearm must be an ‘operative factor in relation to the predicate offense’.53 This is because the dictionary definitions of use (‘[t]o convert to one’s service’, ‘to employ’, ‘to avail oneself of’, and ‘to carry out a purpose or action by means of’) ‘imply action and implementation’.54 Thus, use does not include the ‘action of a defendant who puts a gun into place to protect drugs or to embolden himself’.55 It would, however, include situations where the defendant has a gun on display during a transaction or barters with a firearm without handling it.56 Finally, in 2007 the Court held in Watson v. United States57 that a person who trades drugs for a gun does not ‘use’ a firearm ‘during and in relation to … [a] drug trafficking crime’.58 According to the Court, ‘the meaning of the verb “uses” has to turn on the language as we normally speak it’.59 In the Court’s view, the proper interpretation must ‘appeal to the ordinary’ because ‘there is no other source of a reasonable inference about what Congress understood when writing or what its words will bring to the mind of a careful reader’.60 Based on its own understanding of common usage, the Court reasoned as follows: [t]he Government may say that a person ‘uses’ a firearm simply by receiving it in a barter transaction, but no one else would. A boy who trades an apple to get a granola bar is sensibly said to use the apple, but one would never guess which way this commerce actually flowed from hearing that the boy used the granola.61Interestingly, while the Court in Smith indicated that a broad meaning should be given the provision considering its purpose of combating the dangerous combination of drugs and guns,62 the Court in Watson declined to give such considerations significance.63 The Supreme Court’s Failure to Recognize Linguistic and Cognitive Science Insights As its interpretations of §924(c)(1)(A) indicate, the Court purports to give textual language its ordinary meaning. Evidence from cognitive science, however, indicates that its approach is flawed. One stark example of the Court’s deficient methodology is its undifferentiated approach to expressions that feature ellipsis and thus require enriched semantic composition. A linguistic ellipsis is a ‘truncated or partial linguistic form’ ‘in which constituents normally occurring in a sentence are superficially absent, licensed by structurally present prior antecedents’.64 An important, and empirically based, principle of grammars is that speakers produce the minimum linguistic information sufficient to achieve the speaker’s communicational needs.65 In its interpretations of §924(c)(1)(A), though, the Court fails to appreciate this insight and thus fails to recognize the increased difficulty of interpretation when faced with legislative ellipsis of important sentential content. An ellipsis is a ubiquitous and natural way to avoid periphrastic constructions. For example, a case of verb phrase ellipsis occurs in the following sentence: Max went to the store, and Oscar did, too.66 Psychologist Matthew Traxler indicates, via corpus analysis, that expressions with fully specified event structures are rare (i.e. are ellipsed) when the event is commonly associated with the noun.67 According to Traxler, in normal usage, a fully specified event structure is used less than five per cent of the time.68 Full event structures tend to occur only with less predictable activities.69 As a result of studies he conducted that examined the processing of expressions argued to require enriched semantic composition, Traxler explains that a challenge for interpreters is that noun phrases denoting entities are difficult to process when they follow verbs that require event complements.70 The verbs begin, finish, and enjoy, for example, require postverbal arguments that semantically represent events.71 Consider the sentence,The sentence is difficult to process because it contains a noun phrase denoting an entity (namely, book) that follows a verb requiring an event complement (namely, began).73 (1) The boy began the book.72 The increased processing difficulty with respect to (1) is due to related issues. First, the ellipsis of a fully specified event structure requires complex mental operations to construct a suitable event sense.74 Comprehenders construct an event sense of the complement by implicitly generating an activity that is commonly associated with the complement noun and compatible with the agent of the clause.75 For example, the above sentence could be interpreted as The boy began reading the book.76 Secondly, the ellipsis of a fully specified event structure may engender competition between alternative interpretations.77 The sentence could either be interpreted as The boy began writing the book or The boy began reading the book.78 As Traxler indicates, though, if the sentence were intended to be interpreted as The boy began to translate the book, the full event structure would most likely have been specified.79 According to Traxler, his findings ‘provide behavioral evidence for the psychological reality of an enriched form of composition’.80 These kinds of cognitive science findings should have implications for ordinary meaning interpretations because they illustrate the intrinsic complexity of the compositional operation described above and indicate that there is systematicity to the data that are relied on by GL.81 Traxler’s findings are also consistent with those of other linguists such as Jack Hawkins, who has shown that Modern English has significantly more ambiguity and vagueness in its syntactic, lexical, and morphological categories than it once did.82 Consistent with GL’s approach to meaning, Hawkins indicates that Modern English relies to a much greater extent on co-occurring categories for the assignment of relevant disambiguating properties.83 Furthermore, ‘dominant meanings are understood more readily than are subordinate meanings in a wide range of tasks’, which should influence which meaning a court designates as the ‘ordinary’ one.84 3. Overview of GL Theory As the above discussion illustrates, cognitive science and linguistic research support a revised approach to ordinary meaning determinations. The Court’s current approach to interpretation simply does not systematically account for the polysemy of words such as use, which carries different meanings in different contexts.85 GL can, however, serve as a method for incorporating Traxler’s insights, as well as those of linguists and others with findings about the systematicity of language, and can offer a formal methodology that will avoid overly broad, or overly narrow, interpretations.86 Before explaining in detail how 18 USC §924(c)(1)(A) would have been interpreted if GL had been applied, it is necessary to give a general overview of how GL assembles meanings into compositional structures. As Jackendoff explains, ‘the learnability of an unlimited variety of word meanings argues that word meanings are composite, built up in terms of a generative system from a finite stock of primitives and principles of combination’.87 Consistent with this composite view of word meanings, GL views the lexicon as the key source for explaining the creative use of language.88 GL presents an analytic framework that accounts for the disambiguation process involved in the determination of meaning.89 The function of GL is the production of a lexicon consisting of complex lexical entries over which a set of generative operations may apply in order to yield interpretations.90 The GL system is based on feature structures, which have a sequence of nested feature structures within them.91 The resulting lexical entries contain the prototypical predicates and relations associated with the given word.92 The word thus carries a set of information (as fields in a template), and operations define how to combine this information. The purpose of these feature structures, along with the generative mechanisms, is to make linguistic theory more explicit and testable.93 The Four Levels of GL Theory The resources available to a word within GL theory consist of the following four levels: LEXICAL TYPING STRUCTURE,ARGUMENT STRUCTURE, EVENT STRUCTURE, and QUALIA STRUCTURE.94 1. Lexical Typing Structure The LEXICAL TYPING STRUCTURE ‘gives an explicit type for a word positioned within a type system for the language’.95 Pustejovsky has created an ontology that divides the domain of individuals into three levels of type structures: (i) natural types, which are natural kind concepts (i.e. ‘things that occur naturally in the world without need for human intervention’);96 (ii) artifactual types, which are not natural types but, rather, concepts making reference to the purpose, function, or origin of the concept; and (iii) complex types, which are concepts integrating reference to the relation between types from the other levels.97 2. Argument Structure The ARGUMENT STRUCTURE specifies the number and type of logical arguments for a predicate.98 It provides a list structure where the type of argument is directly encoded in the ARGUMENT STRUCTURE, as indicated below:  For example, the verb kiss requires two arguments, a kisser and a kissee.99TRUE ARGUMENTS (ARG1 and ARG2) are those that must be syntactically realized.100 They are what we generally think of as the arguments of a lexical item (e.g. to kiss: true arguments = kisser, kissee).101DEFAULT ARGUMENTS (D-ARG1) are not necessarily expressed syntactically but are necessary for the logical well-formedness of the sentence (to paint has the default argument = brush).102SHADOW ARGUMENTS (S-ARG1) also refer to semantic content that is not necessarily expressed but which is expressible under certain circumstances.103 These are arguments in the QUALIA (explained below) that the sentence normally does not indicate at all because the verb is related to the SHADOW ARGUMENT (e.g. to fax: shadow argument = fax machine).104Trueadjuncts are elements in the sentence that are not part of the QUALIA STRUCTURE of the other words in the sentence (e.g. to kiss: true adjunct: at noon).105 3. Event Structure The EVENT STRUCTURE (EVENTSTR) of a lexical item defines what type of event is taking place.106 It defines both the event type of the expression and any subeventual structure it may have, which includes information about aspect and the temporal relationship among an event’s subevents.107 The schemata for an EVENT STRUCTURE is illustrated below:  Events are classified as states, processes, or transitions, with the latter sometimes being divided into accomplishments and achievements.108 A state ‘denotes an event such as be sick, know, love, and resemble’.109 A process denotes an event such as walk, run, push, and drag. By itself, a process contains no information about how long the action lasts and denotes an event that is composed of a series of events. If V is a process, then the entailment X is V-ing entails X has V-ed should hold.110For example, replacing X and V with Finn and run, Finn is running entails Finn has run.111 An accomplishment is one type of transition and refers to the movement from one state to another state.112 Examples of accomplishments include build and destroy.113 When a verb V is an accomplishment, X is V-ing does not entail X has V-ed.114 Thus, Finn is building a house does not entail Finn has built a house.115Accomplishments ‘denote events with slow change and can be modified by temporal adverbs such as in an hour’.116 An achievement is another type of transition, ‘denoting events such as die, find, and arrive’.117Achievements can be modified by point adverbials (e.g. Finn arrived at noon).118 Modification by point adverbials indicates that achievements ‘denote events with instantaneous change’.119 Events can be simple and complex independently of their syntactic representation as either a simple or a complex clause. When complex, Pustejovsky deals with them in terms of subevents.120 The subevents in the EVENT STRUCTURE may have one of the following three restrictions on their temporal ordering.121 In a complex event, e, formed by e1 and e2, e1 precedes e2.122 Thus, the subeventual structure involves a development process and a resulting state (e.g. build).123 This structure is referred to as exhaustive ordered part of and is represented by ‘<α’.124 The second type of subevent involves those that can occur simultaneously (e.g. accompany, which requires the person accompanying and the one being accompanied). Pustejovsky refers to this relationship as exhaustive overlap part of, and he represents it as ‘oα’.125 A third situation occurs when two subevents are basically simultaneous but one starts before the other (e.g. run, where the movement of the legs brings about the movement of the body).126 Pustejovsky refers to these as exhaustive ordered overlap, which is annotated as <oα.127 The relative prominence of the subevents is also indicated. The head of an event is defined as ‘the most prominent subevent of the EVENT STRUCTURE of a predicate, which contributes to the “focus” of the interpretation’.128 4. Qualia Structure The QUALIA STRUCTURE is the ‘set of properties or events associated with a lexical item which best explain what that word means’.129 In other words, it contains the essential attributes of an object or event as defined by the lexical item.130QUALIA ‘not only structure our knowledge of words, but also “suggest” interpretations of words in context’.131 It provides a general strategy for creating increasingly specific concepts with conjunctive properties.132 The QUALIA STRUCTURE specifies the following four basic roles: the CONSTITUTIVE, FORMAL, TELIC, and AGENTIVE.133 Each of the roles, which capture how humans understand objects and relations in the world, provides a distinct and essential component that helps to determine the meaning of a word.134 The schemata for a QUALIASTRUCTURE is illustrated below:  The CONSTITUTIVE role of a lexical item captures the ‘relation between an object and its constituents or proper parts’, which may include a specification of material, weight, and parts and component parts (e.g. a hand has fingers).135 The FORMAL role specifies what ‘distinguishes the object within a larger domain’ and may include information about the orientation, magnitude, shape, dimensionality, color, and position of the object (i.e. whether the object is human, a physical object, a proposition, etc.).136 The TELIC role defines the purpose and function of the object (e.g. dishwasher: TELIC = to clean dishes).137 Finally, the AGENTIVE role describes the factors involved in the object’s origins or ‘coming into being’, including distinctions such as natural kind versus artifact (e.g. computer program: AGENTIVE = to code).138 The Generative Mechanisms of GL Theory A set of generative devices connects the four levels described above and provides for the compositional interpretation of words in context.139 The interpretation of a case of logical polysemy is conditioned by one of the types of generative operations, which include type coercion, co-composition, and selective binding.140 Because it is the theory needed to provide a methodologically sound interpretation of 18 USC §924(c)(1)(A), the co-composition mechanism will be the focus of the discussion below. Verbs are necessarily the focus of GL because, typically, linguistically sensitive theories of lexical structure tend to focus on how verb meanings relate to syntactic forms within a sentence.141 More precisely, the selectional behavior of verbal predicates follows from the type associated with the verb’s arguments.142 The GL model uses a type system, which is a tractable syntactic framework for classifying phrases according to the kinds of values they compute.143 Recall that Pustejovsky has created an ontology that divides the domain of individuals into three levels of type structures.144Natural types reference FORMAL and CONSTITUTIVEQUALIA roles, artifactual types reference TELIC (purpose or function) or AGENTIVE (origin) QUALIA roles, and complex types integrate reference to the relation between types from the other levels.145Artifactual types behave differently from naturals as they carry more information regarding their use and purpose.146 A simple natural physical object, though, can be given a function (i.e. a TELIC role) and transformed into an artificial type.147 Co-composition, also termed accommodation by Pustejovsky, is the mechanism that operates over structures that allow for more than one function application.148 For co-composition to occur, both the verb and the complement must share some QUALIA value. In other words, ‘co-composition occurs when the verb specifies some aspect of its direct object’s QUALIA STRUCTURE in its own lexical structure’.149 Consider the example (given by Pustejovsky, 1995) below:As the example illustrates, bake has a different meaning in (5)(a) than it does in (5)(b). In (5)(a), bake has a change of state sense, while in (5)(b) it has a creative sense. The verb bake has thus been argued to be systematically ambiguous between a creative and change of state sense.150 Pustejovsky, however, argues that the two senses should be treated as instances of the single phenomenon of logical polysemy, where a word has two or more meanings but the different senses are determined by the semantics of the complement noun.151 (2) (a) John baked a potato.   (b) John baked a cake. Consider the lexical representation for the verb bake below:152  According to Pustejovsky, the default sense of bake is a change-of-state interpretation.153 Any other interpretations must occur through the generative mechanism of co-composition, which applies at the level of the entire verb phrase.154 The co-composition mechanism allows for the information carried by a complement to act on the governing verb, by taking the verb as argument and shifting its event type.155 Consider the lexical representation for the noun cake below:156  The co-composition theory accounts for the differences in interpretation of bake in bake a cake and bake a potato by distinguishing artifacts from natural kinds.157 As indicated above, bake is indicated as having a change of state denotation. When combined the complement potato, whose AGENTIVE role is specified as natural kind, the change of state interpretation of bake remains unchanged.158 The structure of cake above, however, shifts the meaning of bake from a change-of-state to a creation sense.159 The QUALIA STRUCTURE for cake specifies that there is an event, the process of baking, associated with this object’s coming into being.160 The AGENTIVE role of the complement noun makes reference to the process denoted by the verb phrase, which gives rise to the co-specification title for the process.161 The creation sense of bake thus arises from the semantics of cake by virtue of its being an artifact (i.e. originating out of the act of baking).162 According to Pustejovsky, it is not, therefore, the verb bake that is polysemous, but rather, the difference in meaning between bake a cake and bake a potato is the result of principles of semantic composition.163 The complements add to the basic meaning of bake through co-composition. 4. Application of GL Theory to 18 USC §924(c)(1)(A) Applying GL to 18 USC §924(c)(1)(A) requires the creation of attribute structures for use and firearm, as well as the generative mechanism of co-compositionality. First, though, it is necessary to consider the nature of the verb use. The Supreme Court has conceded that the ‘word “use” poses some interpretational difficulties because of the different meanings attributable to it’.164 These ‘difficulties’ for the Court stem from its fundamental misunderstanding of compositionality and the nature of a verb such as use.165 The verb use is designated by Pustejovsky as a light verb and by Ritter and Rosen as a weak verb.166 The light and weak verb categorizations are meant to indicate the underspecified nature of use.167 These categorizations indicate that verbs such as use have great utility because they can occur with a variety of different objects and permit an economy of expression that allows fundamental information to be ellipsed, such as the nature of the particular activity that is denoted by the verb.168 The Supreme Court’s failure to recognize the underspecified nature of use when interpreting 18 USC §924(c)(1)(A) is tied to its failure to recognize the existence of ellipsis and its nature, specifically, that the ellipsed information is supplied by the complement.169 In combination with the object of the sentence, use forms a meaning that is more specific than either itself or the object.170 Consider the following examples:According to Pustejovksy, in (3)(a) ‘our knowledge of knives as tools which can cut permits an economy of expression, whereby mention of the particular activity of cutting may be illipsed’.172 Similarly, in (3)(b), ‘contact lenses are visual aids, and the use of them refers to the act of wearing them’.173 In the same way, we should interpret (3)(c) as being a ‘near paraphrase’ of ‘“My wife travels on the subway every day”’.174 (3) (a) John used the new knife on the turkey.   (b) Mary has used soft contact lenses since college.   (c) My wife uses the subway every day.171 As the examples above illustrate, because of the underspecified nature of use, co-composition is necessary in order to construct the meaning of a sentence. Consider the structure for use below:  Regarding the EVENT STRUCTURE, as explained in section 3, use should be classified as a process because it does not inherently last for a well-delineated period of time.175 For example,does not specify how long the using action lasts unless we add a phrase such as for a few minutes. Another indication that use is a process is that Finn is using the camera entails Finn has used the camera.176 The uppercase E1 on the left-hand side of the equal sign indicates an event.177 The lowercase e1 is a variable that refers to the event.178 The combination of the lowercase e1, the colon, and the word process shows that the event is denoted by a process.179  (4) Finn uses the camera Regarding the ARGUMENT STRUCTURE, both the arguments denoted by the boxes and are true arguments, meaning that they are ‘syntactically realized parameters of the lexical item’.180Usables include nouns that have a function or purpose. As indicated in (3) above, these functions or purposes are wide and varied, expressing the different ways that usables help accomplish the act expressed in their TELIC quale.181 Notice that, reflecting the unspecified nature of use, the allowable arguments are quite broad. ARG1, representing the subject, only requires that a physical object be the user. Thus, inanimate objects are allowed as subjects, as (5) illustrates:  (5) The washer used electricity to run. ARG2 is even less constrained, allowing animate, inanimate, and incorporeal direct objects.182 In the feature structure, there is no modal modifying the object’s TELIC predicate in the verb’s FORMAL predicate, reflecting that using an entity represents performing the action in the entity’s TELICQUALIA. ARG2 is an entity rather than a physobj because the following type of sentence is possible:  (6) I wish I spoke French well enough to use it.183 As explained above, the AGENTIVE quale of the QUALIA STRUCTURE indicates how the event came into being.184 Pustejovsky typically makes the value of the AGENTIVE quale predicate for verbs the name of the lexical entry following by _act, expressing that an event comes into being through performing the event.185 The e1 found in both the EVENT STRUCTURE and the AGENTIVE quale is a variable that refers to the event of using.186 The boxed numerals and following e1 in the use_act predicate of the AGENTIVE quale co-index the first and second arguments of the ARGUMENT STRUCTURE, respectively (i.e. the physobj subject and the entity object).187 The box around a numeral indicates that any feature structure information following the boxed numeral combines with feature structure information following an identical boxed numeral elsewhere in the feature structure.188 There is no feature structure information following the boxed numeral in the AGENTIVE quale for use, but there is the feature structure information entity following the boxed numeral in the ARGSTR of use.189 In fact, the information after the boxed numerals has combined so that it is entity in both places.190 As shown by the feature structure for use, the following formallows a wide range of different interpretations, depending on the choice of direct object (and, sometimes, subject). Of course, we are most interested in the feature structure for firearm, which is below:191  The EVENT STRUCTURE indicates that the function of a firearm is an accomplishment.192 The first predicate of the QUALIA STRUCTURE (the FORMAL predicate) indicates that firearm is a physical object x.193 An example of one type of TELIC, that of purposeTELIC, is found with objects that are used in the performance of an activity, such as the noun firearm. The TELICSTRUCTURE for firearm indicates that its purpose is to be used as a weapon x directed at a physical object y.194 The third predicate (the AGENTIVE predicate) associated with the QUALIA STRUCTURE for firearm is the weapon predicate in which the human argument a of the ARGUMENT STRUCTURE (ARGSTR) uses the firearm as a weapon.195  (7) use DETERMINER NOUN As stated above, for co-composition to occur, both the verb and the complement must share some QUALIA value.196 As indicated in section 3, the main feature of co-composition is the incorporation of the QUALIA STRUCTURE of the direct object, or subject if appropriate, into the lexical structure of the verb.197 Typically, the TELIC predicate of the direct object is incorporated into the arguments of the FORMAL predicate of the verb.198 As applied to use and firearm, this means thatcannot mean that Finn used the firearm as a monetary device. Thus, contrary to the Supreme Court’s interpretation of 18 USC §924(c)(1)(A), (8) is not synonymous withInstead, the default interpretation of (8) must be that Finn used the firearm as a weapon.200 As the above discussion illustrates, by relying on the attribute structures and the co-composition mechanism, GL is able to give a formal means of understanding how ordinary meaning interpretations of sentences can be generated.  (8) Finn used the firearm  (9) Finn used money.199 One might object that the TELIC and AGENTIVE features of firearm must be shoot and shoot_act, respectively, thus requiring that the interpretation of (8) be that Finn used the firearm to shoot.201 This version of the attribute features of firearm would, arguably, illustrate that the interpretation generated by GL is obviously too narrow considering the context of 18 USC §924(c)(1)(A), which was not intended to require that any actual shooting occur.202 Perhaps, this issue regarding the attribute structure for firearm reveals the discretion inherent in defining the attribute features for words, but it does not seriously undermine the efficacy of GL.203 Even if shoot and shoot_act are more appropriate than weapon and weapon_act, there is no reason that the TELIC and AGENTIVE STRUCTURES require that a shooting actually occur. Rather, they should be seen as being satisfied if the firearm is used agentively in relation to its TELIC function. Thus, if the firearm is used as a weapon and the potentiality exists that it will shoot, its TELIC function is satisfied even if it is denominated as shoot.204 … . The application of GL to 18 USC §924(c)(1)(A) underscores Justice Scalia’s argument in Smith v. United States205 that the Court relied on a dubious presumption when it stated that Congress could have included the words ‘as a weapon’ in the statute if it had intended that the firearm must have been used as a weapon.206 The foundation of the Court’s error was its reliance on the multiple dictionary definitions of use to create an overly broad interpretation of the statute and the corresponding failure to recognize event ellipsis. Instead, the Court should have recognized that use has little semantic content of its own and yields divergent interpretations through its attachment to direct objects.207 Further undermining the Court’s ordinary meaning methodology of primarily relying on a dictionary definition for use was its arbitrary holding in Watson v. United States208 that a person who trades drugs for a gun does not ‘use’ a firearm ‘during and in relation to … [a] drug trafficking crime’.209 A pragmatic-based limitation might have been justifiable, but the Court based its holding on semantics.210 Unfortunately, the Court failed to recognize that unlike the verbs sell or give, use is not unidirectional.211 As the Court noted in its earlier decision in Bailey v. United States,212 one of the dictionary definitions of use is ‘[t]o convert to one’s service’, and another was ‘to avail oneself of’.213 Certainly, the receipt of a firearm as one’s possession means that the item has been ‘convert[ed] to one’s service’.214 The Court’s methodology of choosing broad dictionary interpretations and deciding on an ad hoc, and often linguistically suspect, basis whether the broad interpretation should be narrowed is unpredictable and unprincipled. By applying GL, the Court would have flipped its erroneous presumption that Congress was required to add the phrase as a weapon. The Court would have first generated an ordinary meaning for ‘uses a firearm’ and then analyzed whether there was any clear evidence that Congress intended to adopt a broader, and highly anomalous, interpretation that would cancel the default interpretation and license the interpretation use the firearm as currency.215 Certainly, idiosyncratic or anomalous uses of objects can be imagined (e.g. using a briefcase as a shield), but it is erroneous to assume in the absence of evidence to the contrary, when the ordinary meaning of the statute is being sought, that an unconventional meaning was intended.216 Although Justice Scalia’s dissenting opinion in Smith reached the correct result regarding the interpretation of §924(c) and insightfully identified the crucial flaw in the Court’s reasoning, Justice Scalia’s approach should not be seen as an effective substitute for GL. Importantly, his opinion did not offer a comprehensive systematic approach to ordinary meaning determinations that can be applied in future cases.217 Indeed, Justice Scalia’s approach to ordinary meaning, especially his reliance on dictionaries, has been shown to be inconsistent and invoked in order to produce desired results.218 By applying GL in Smith, the Court would have reached the same conclusion as Justice Scalia but with reasoning based on a consistent, systematic, and replicable theory of language. 5. Challenges to GL Theory As demonstrated above, Pustejovsky’s co-composition mechanism for analyzing polysemy provides a compelling account of §924(c)(1)(A). His generative mechanisms are consistent with his claim that ‘the notion of word sense only emerges at the level of the sentence, namely its “sense in context”’.219 As such, the sense of ‘governing’ expressions (such as the verb use) is inherited partly from their lexical entries but partly also from the semantics of the expressions that they govern in context.220 This account persuasively explains why the common judicial practice of deriving the meaning of a word in isolation through a dictionary definition is problematic.221 Even if a court chooses the dictionary meaning as only an initial step in the interpretive process and allows for the possibility that pragmatic contextual information may narrow the dictionary definition, the selection of the dictionary definition ignores important sentential context and will frequently lead to erroneous interpretations.222 Despite its undeniable perspicacity in illustrating the Supreme Court’s faulty approach to the interpretation of §924(c), one foundational issue about the legitimacy of GL as an account of polysemy is whether it would be fatally undermined if it, at least occasionally, makes clearly erroneous predictions about proper default interpretations. Indeed, a criticism of GL is that it generates inaccurate interpretations because its purported semantic and syntactic regularities are not sufficiently generalizable across the English corpus. Jerry Fodor and Ernest Lepore (FL), for example, criticize lexical theories that isolate the internal constitution of word meaning.223 FL dispute that Pustejovsky has identified any regularities in the language. In fact, they argue that he can ‘offer[] no convincing cases where the meaning of a governing expression is modulated by the lexical content of the expressions that it governs’ because ‘this never happens’.224 Three related claims are derived from these critiques. One is that Pustejovsky’s purported regularities overgenerate meanings because they erroneously predict semanticality with regard to language that is unsemantical. Another criticism, a mirror image of the first, is that Pustejovsky’s purported regularities undergenerate meanings because they erroneously predict unsemanticality with regard to language that is semantical. The third is that GL generates erroneous interpretations of language that is semantical.225 This section addresses these criticisms in turn and also discusses the relationship between underspecificity and vagueness. GL Theory and the Overgeneration of Meanings The claim that GL erroneously predicts semanticality with regard to language that is unsemantical is debatable and, in any case, not relevant to the usefulness of GL as a mechanism for generating ordinary meaning interpretations of normative legal texts. Consider the following sentences:Recall that Pustejovsky rejects the notion that the verb bake is systematically ambiguous between a creative and change of state sense. He argues that the two senses should be treated as instances of the single phenomenon of logical polysemy, where a word has two or more meanings but the different senses are determined by the semantics of the complement noun.227 Pustejovsky claims that due to the co-composition mechanism, the semantics of the direct objects in (10)(a) and (10)(b) result in a noncreative sense of bake for (10)(a) and a creative sense of bake for (10)(b).228 While the default change of state sense of bake is appropriate for potato in (10)(a), the creation sense of bake in (10)(b) arises from the semantics of cake by virtue of its being an artifact (i.e. originating out of the act of baking).229 Thus, according to Pustejovsky, ‘[t]o classify the verb bake as having both senses lexically specified is to miss the semantic generalization …’.230 (10) (a) John baked a potato.   (b) John baked a cake.   (c) John baked a trolley car.226 FL argue that Pustejovsky’s theory overgenerates meanings due to the strength of the proposed generative mechanisms. They argue that it is unclear how ‘a combination of a verb with a[] [noun phrase] argument could ever fail to be interpretable’.231 Pustejovksy’s currently constructed co-composition mechanism cannot account for the data above because if bake a cake is heard as creative because cake is marked as denoting an artifact, then bake a trolley car should be heard as creative too.232 FL argue that other mechanisms are therefore necessary. For example, with regard to a verb such as use which relies on the TELIC role of the governed expression, a device is needed which ‘picks the TELIC role when there is one and that yields unsemanticality when there isn’t one’.233 One problem with FL’s argument is that it is not clear that GL would not yield unsemanticality with regard to bake a trolley car. Notwithstanding that the alleged overinclusiveness of the co-composition mechanism is relevant to semantics in general, this overinclusiveness is not important to the interpretation of normative legal texts. American courts recognize the doctrine of scrivener’s error, which allows courts to correct the obvious clerical or typographical errors made by the legislature when drafting language.234 Courts also recognize the doctrine of absurdity, which allows them to avoid a literal interpretation of text if the court views the interpretation to be absurd.235 Neither of these doctrines seriously undermines, though, the fundamental notion that statutory text is, by definition, semantical.236 As the judicially created rule against surplusage illustrates, if text is placed in a statute, it has a meaning.237 Even if the possibility that Pustejovsky’s generative mechanisms may carry over to unsemantical constructs is relevant, it is not clear that GL really would attribute a creation sense to John baked a trolley car. Wilks argues that FL’s analysis ‘completely ignore[es] the fact that the whole JP analysis is based on the “natural” assumption that potatoes and cakes both share some attribute like FOOD (as trolley cars do not), which is the only way the discussion can get under way: being a FOOD is a necessary condition for this analysis of “bake” to get under way’.238 Wilks’ response, as a defense of GL, is problematic, however, to the extent it relies on world rather than lexical knowledge that is encoded in the QUALIA STRUCTURE. Recall the QUALIA STRUCTURE for cake: It does not indicate the attribute food, making Wilks’ solution one that is pragmatic rather than semantic because it suggests that the relevant information would be based on world knowledge rather than encoded in the lexicon.239 A lexical response to FL’s critique is possible, though, because FL’s analysis ignores the entirety of an object’s QUALIA STRUCTURE. The TELIC role, which defines the purpose and function of the object, of cake is eat, which indicates that it is an edible object.240  Presumably, the TELIC role of trolley car would be something quite different, such as ride. ‘Co-composition occurs when the verb specifies some aspect of its direct object’s QUALIA STRUCTURE in its own lexical structure.’241 As the structure below illustrates, bake specifies a bake_act, as does cake.  Trolley car, however, would not specify a bake_act. A lexical explanation is thus consistent with GL. GL Theory and the Undergeneration of Meanings Interestingly, FL do not consider the possibility that their example John baked a trolley car might be a semantically meaningful sentence.242 Elsewhere in their response, though, FL claim that Pustejovsky’s purported regularities are underinclusive because they erroneously predict unsemanticality with regard to language that is, in fact, semantical. FL claim that ‘use raises a thorny problem for [Pustejovsky’s] account since it’s perfectly possible to use things that don’t have uses: a rock to break a window, snow to make a snowball’.243 Thus, FL wonder ‘what happens if a verb makes a demand on an argument that the lexical entry of the argument doesn’t satisfy’.244 It seems apparent, though, that GL can handle standard word senses in nonstandard contexts. Recall that Introduction is applicable when the selecting type (i.e. the verb) is richer than the type of its argument.245 In response to FL’s criticism about GL’s underinclusiveness, Pustejovsky concedes that while ‘functionality plays a crucial role in the process of individuation of artifacts, but not of natural kinds’, ‘[e]xceptions arise when new concepts are referred to, such as when the object is construed relative to a specific activity’.246 Thus, while the phrase  (11) enjoy the rockmight be considered semantically ill-formed as a general matter, a novel context can give it meaning.247 According to Pustejovsky, in the sentence‘rock takes on a new meaning, by virtue of having telicity associated with it …’.248 Notice too that the meaning of (12) is derived by co-composition with the semantics of the subject NP. Traxler comments in his study that ‘coerced senses are computed from a broader range of properties than the QUALIA STRUCTURE of the complement noun’ because ‘comprehenders appear capable of using properties that do not appear to be exclusively derived from the complement noun but instead appear to be associated with the agent of the sentence or other discourse elements’.249  (12) The climber enjoyed the rock Still, while GL may be able to cope with standard word senses in unusual contexts, the same is not true with regard to nonstandard word senses.250 A ‘nonstandard’ usage is defined by the linguist Kilgarriff as not matching a literal reading of any of the word’s dictionary definitions.251 Kilgarriff conducted a study that examined the ‘nonstandard’ uses of words found in the HECTOR lexical database. As discussed above with example (12), GL can handle combinatorial uniqueness. For example, in the sentencethe meaning of sack is nonstandard because the cease-employing meaning is specified in HECTOR as taking a person as its direct object.252 The GL mechanisms could, however, only account for a small proportion of unique word meanings.253 For example, it was inapplicable in settings where handbag denoted a musical genre.254  (13) Mr. Bruce George has called for the firm to be sacked from duty Kilgarriff’s findings are not damaging to the use of GL in legal interpretation. Kilgarriff views his study as identifying the scope of the usefulness of GL rather than undermining its core, and this conclusion is certainly valid with respect to GL’s application to legal texts.255 Recall that ordinary meaning must mean the common meaning, as opposed to a technical or creative meaning.256 Kilgarriff believes that while linguistic creativity is always possible, words have normal uses, and ‘much of the time, speakers simply proceed according to the norms’.257 Kilgarriff, thus, agrees with GL’s premise that the lexicon is rife with generalizations and that facts about a word that cannot usefully be viewed as instances of a generalization are vastly outnumbered by those that can.258 GL Theory and the Generation of Erroneous Meanings While concerns that GL over- or undergenerates meanings are largely either misplaced or not relevant to the application of GL to legal texts, assertions that GL provides the erroneous sense for some standard senses in typical contexts require a more nuanced response that recognizes the inherent limitations of language. The main objection of critics is to the way in which under GL ‘certain inferences seem to come out as necessary entailments as a result of the application of the generative operations over lexical entries’.259 In other words, some linguists have argued that there are exceptions to all of GL’s generated meanings and that the existence of these exceptions is delegitimizing to the theory.260 1 Errors, Default Rules, and Pragmatics Consider the following sentences:FL argue that contrary to Pustejovsky’s claim that bake in (14)(b) should be given a creative sense, it is actually ambiguous between a creative and a non-creative (i.e. ‘warm up’) sense of bake.261 They claim that one can do to a (preexistent) cake just what one does to a (preexistent) potato: put it in the oven and (noncreatively) bake it. Since bake a cake is ambiguous and cake is univocal, it must be that bake is lexically ambiguous (specifically, polysemous) after all, contrary to [Pustejovsky’s] analysis.262The same argument is made that Pustejovsky’s co-composition mechanism makes an incorrect prediction with regard to (14)(c).263 Pustejovsky’s account would predict a creative interpretation of bake based on pizza being an artifact in the same way as cake.264 But is a creative reading more accurate than a noncreative reading considering the possibility that John may be merely heating a frozen pizza? (14) (a) John baked a potato.   (b) John baked a cake.   (c) John baked a pizza. One linguist, Yorick Wilks, responds to FL’s argument regarding (18)(b) by asserting that ‘warming up a (preexistent) cake is not baking it … That activity would be referred to as warming a cake up, or through, never as baking’.265 Perhaps, Wilks would make the same argument with respect to (14)(c). Pustejovsky, however, disclaims that his generative mechanisms should be viewed as producing necessary entailments. In contrast to Wilks’ reasonable, though perhaps debatable, position, Pustejovsky responds that ‘[c]o-composition is not an obligatory semantic rule, and the expression is formally ambiguous’.266 Some may argue that Pustejovsky’s concession that co-composition is not an obligatory semantic rule greatly undermines his GL project. After all, Pustejovsky’s GL account of word meaning is meant to generate default interpretations, but his mechanisms do not purport to identify interpretations that should be seen as ambiguous after their application. Pustejovsky himself believes that data should not be analyzed in a post hoc fashion, but his theory necessarily must allow for analysis subsequent to the application of his generative mechanisms.267 In considering the accuracy of GL, it is necessary to realize that the theory provides only a default semantic-based meaning and that no semantic theory can disclaim the potential for pragmatic inferences to shape meaning.268 To aver that language cannot be interpreted without reference to context does not entail that words contain no context-independent linguistically encoded meaning.269 Indeed, it has been argued that, due to the autoglottic space, statutory language should normally be regarded as ‘unsponsored’ with a form of sentence meaning rather than utterer’s meaning.270 In any case, word meanings have a ‘prototype structure’, where the most typical and salient uses of a word are based on a clear and shared semantic model, and, of course, GL is a theory that seeks to formalize combinatorial meanings and distinctions that are real, systematic, and part of the language itself.271 Pustejovsky himself indicates that GL is designed to generate default interpretations.272 Obviously, as the examples in (14) illustrate, default meanings, like word meanings in general, are subject to change for various reasons, including advances in food preparation. The general strategy of GL is thus to create default meanings by assuming an underspecified lexical representation (i.e. the attribute structures) which is then contextually interpreted on the basis of the sentential context incorporated into the GL framework (i.e. the generative mechanisms). Of course, the theory assumes that pragmatic inferences may shape the default meaning. As distinct from the semantic issues that concern GL, pragmatics involves information that is ‘generated by, or at least made relevant by, the act of uttering the sentence’.273 Pragmatics is widely recognized to be intrinsic to interpretation because it cannot be maintained that compositionality, when defined narrowly as the principle that the meaning of an expression is determined by the meanings of its constituent expressions, is always sufficient to render the meaning of an expression. Instead, certain aspects of language necessarily require reference to the users of the language, and, in general, semantic meaning is always subject to pragmatic shaping.274 Some utterances or inscriptions, such as indexicals (also known as deixis), have a content that is universally recognized to be relative to the context of utterance. Examples of indexicals include terms such as I, now, here, and today, corresponding to person, place, and time. Consider the sentenceObviously, a factually rich environment is necessary in order to reach an accurate understanding of the sentence. Examples of pragmatic enrichment include presuppositions, implicatures (a means of communicating more than is literally said), as well as discrepancies between speaker meaning and sentence meaning, such as sentences that connote irony or sarcasm.275  (15) I am not here right now. Pustejovsky concedes, as he must, that the results of his generative mechanisms can be shaped by the application of pragmatic evidence.276 Note one of Pustejovksy’s descriptions of word meaning: ‘polysemy is the result of both compositional operations in the semantics, such as coercion and co-composition, and of contextual effects, such as the structure of rhetorical relations in discourse and pragmatic constraints on co-reference’.277 GL does not explicitly specify how pragmatics intersects with its semantics-based default interpretations and arguably should not be obligated to do so since it does not purport to offer any novel pragmatic theories.278 Instead, it is important to recall that in cases of legal interpretation, GL should only be used to generate semantic-based default interpretations. Criticisms of GL must, therefore, be tempered by the realization that GL contemplates the application of traditional pragmatic inferences that may play a role in shaping meaning. 2. The General Problem of Underspecification Discussion of GL’s accuracy must also include consideration of the underspecificity of language, especially legal language. Linguists generally agree that natural languages are underdetermined and ‘all linguistic forms are “incomplete” and selective and must be disambiguated against the context in which they appear’.279 Difficulties in understanding in ordinary language use can frequently be readily solved in a communicative situation, but the interpretation of legal texts does not provide such possibilities.280 Because autoglottic space (a distance between the writing and the author(s)) is an inherent characteristic of legal texts, courts must create interpretive tools, such as the ordinary meaning concept, that are based on generalized, and often fictional, assumptions about authorial intent.281 Beyond the general difficulty of interpreting texts, normative legal texts often present additional challenges. Due to the nature of legislative drafting incentives, including the frequent legislative desire for maximal interpretive flexibility, indeterminacy is particularly prevalent.282 One reason for the indeterminacy is the use of hedging terms, which include such words as reasonable and other modifiers (such as just, due, proper, and fair).283 Apart from hedging terms, indeterminacy is also caused by the ellipsis of important information.284 While a generative theory of semantics such as GL by definition assigns meanings to phrases and sentences, greatly underspecified language presents particular difficulties. One of the central aspects of GL is that despite the potential indeterminacy caused by, for example, the ellipsis of important sentential context, observed syntactic regularities can provide the information necessary to attribute the correct default meaning to the language in question.285 Nevertheless, there are limits to precisification. As a result, no generative theory can be expected to provide accurate meanings to greatly underspecified language.286 The problematic nature of the kind of underspecificity that involves vagueness is well known in the law.287 With vagueness, a ‘lexeme can be used with multiple different interpretations, but those interpretations are instantiations of a single general sense, rather than distinct polysemous senses’.288 Two separate types of vagueness need to be distinguished. One is lexical vagueness, which occurs when a term presents borderline difficulties.289 This type of vagueness is widespread because ‘[m]ost expressions in natural languages are vague—that is, the denotations of most expressions are fuzzy around the edges’.290 When, for example, is something green instead of brown or blue? Consider the following paradigmatic example of statutory indeterminacy:Example (16), an oft-used example of an ambiguous statute, is properly seen as an example of vagueness borne from a term with borderline difficulties. Is a motorized wheelchair a vehicle? What about a motorcycle or a Segway? What about a nonfunctioning statue of a World War II jeep? And so on.  (16) A statute provides: “No vehicles are allowed in Memorial Park on weekends.”291 Another type of vagueness occurs at the phrasal and sentential level as a species of wanted or unwanted generality in the sense that the term or provision is underspecific because it fails to ‘give enough detail for the purpose at hand’.292 For example, the statementis vague in the lack-of-detail sense.293 Consider the following example of statutory indeterminacy based on a lack of detail:If (18) is not to be read literally as granting authority to detain ‘aliens’ indefinitely, it suffers from a lack of detail. The Director of Homeland Security is given authority to detain ‘aliens’ ‘beyond the [90 day] removal period’, but no information is provided regarding the duration of this power. Can ‘aliens’ be detained indefinitely, and, if not, from where would a restriction emanate? Despite the possible vagueness of (18), the Supreme Court held that the provision was ambiguous.295 The Court’s mischaracterization of the indeterminacy as ambiguity rather than vagueness is highlighted by its arbitrary selection of six months as the extent of the government’s detention authority.296  (17) some event will happen at some time  (18) A federal statute provides that certain ‘aliens’ who are eligible to be deported from the United States but whose countries of origin will not accept their return ‘may be detained beyond the [90 day] removal period … .’294 GL is a theory largely designed to formalize consideration of combinatorial meanings and, as such, does not purport to resolve the type of vagueness that occurs when terms are fuzzy around the edges.297 Furthermore, vagueness can present problems for any interpretive theory. Semantic representations are inherently underspecified in the sense that they are ‘more abstract than the specific interpretations obtained in various contexts’.298 In this light, the line between interpretable and underspecified language should be seen to depend on the purpose of the interpretation. If the purpose is to generate the ordinary meaning in a default interpretation context, a broader range of language is interpretable than if the purpose is to generate the only possible meaning.299 For example, consider the sentenceFL argues that (19) is a near paraphrase of ‘My wife travels on the subway every day’, which is ‘an interpretation that is made possible by our knowledge of what the function of a subway is’.300 FL reasons that ‘if lexical entries really do express meanings then what’s in the semantic lexicon is information about words, and not information about the world—that’s why it’s perfectly possible both to understand Sarah likes to use the subway and to wonder what she likes to use it for’.301  (19) My wife uses the subway every day. Pustejovsky is correct that FL ‘confuses the set of possible, identifiable inferences with those that are automatic and unsurprising’.302 Must we assume, if no further contextual information is available, that pen in (20)(a) and gun in (20)(b) might have been used in the same sense?If the objective is to estimate the probable, ordinary meaning interpretation, surely the answer is ‘no’. We can expect that Mary used the pen to write and the gun as a weapon. Thus, while it is possible of course to wonder what Sarah likes to use the subway for, it is not doubtful that the expected meaning is that Sarah rides the subway.303 Such an interpretation is consistent with the telicity attributable to subway.304 If one cannot predict that the ellipsed information explaining Sarah’s use of the subway involves her riding it, and must assume instead that an idiosyncratic use is just as likely, it would seem that the notion of expected meaning does not exist.305 This is not, however, a necessary conclusion. (20) (a) Mary used a pen.   (b) Mary used a gun. In contrast to (20)(a) and (20)(b), considering the lack of telicity normally associated with rock outside of an enriched context, the sentenceis underspecified with regard to how Mary used the rock. As FL argue, it is, of course, possible to use things that do not have uses, but when the object does not have an associated telicity, it is the contextual information that will give use its sense.306 If there is no contextual information accompanying (21), any default interpretation would necessarily be speculative and not an occasion where GL could provide interpretive assistance. (21) Mary used the rock 6. GL Theory and the Judiciary The judiciary’s prototypical ordinary meaning methodology may be fundamentally flawed, and GL may be able to withstand scholarly criticism and offer a persuasive tool for ordinary meaning interpretations, but judicial adoption of the theory must still depend on it being seen as superior to other courses of action. Indeed, judicial adoption of GL may appear to be a long shot. Arguably, its undeniable complexity, especially in comparison with alternative methodologies, lessens its feasibility as a tool to be used by the judiciary when making ordinary meaning determinations. Even proposing judicial adoption of such a complex, formal theory of meaning may strike some as naively ambitious. None the less, this section argues that judicial adoption GL is superior to alternative courts of action, including other formal theories of meaning, and that GL is uniquely compatible with the judiciary’s existing interpretive tools. GL Theory versus the Alternatives One alternative to judicial application of GL is for judges to simply adopt a few, comparatively simple linguistic principles as a way of improving on their current methodology.307 Perhaps, for example, judges could endeavor to more thoughtfully consider their usage of dictionaries by consulting multiple editions and definitions and avoid temporal errors by considering the ways in which the meanings of words change over time.308 Doing so would, of course, be preferable to current practice. Considering the centrality and importance of ordinary meaning determinations, though, such a compromise should be rejected. This article does not argue that courts should create formal linguistic theories but, rather, apply a very sophisticated, but learnable by a layperson, theory that will facilitate an objective and rigorous consideration of sentential context. A few simple linguistic principles and vague exhortations to ‘consider context’ should not be seen as an adequate substitute for a formal theory like GL that can guide judicial discretion. In addition, the application of a formal theory is more verifiable than nonformal principles and can make a judge’s ordinary meaning reasoning much more transparent. Another possible alternative to judicial application of GL is the appointment of linguists as ‘expert advisers’, ‘consultants’, or ‘confidential advisers’ through the ‘inherent powers of a court’.309 If judges are unwilling to independently study the theories and findings of linguists, they can simply hire linguists to do the difficult work of theory identification and application.310 Such an option is undoubtedly less likely than judicial adoption of GL, however. While the linguist as an adviser option is not inconsistent with this article’s argument for application of GL in ordinary meaning determinations, such determinations are so frequently required that the linguist position would be full time. Hence, if GL is to be applied to statutes in some official way, it will almost certainly be by judges. In addition to its superiority over informal theories and options, GL is superior to other formal linguistic theories in serving as a tool for ordinary meaning interpretations. Various such theories have been proposed by linguists in nonlegal publications. Typically, these theories are heavily pragmatic in nature.311 For example, two-level semantics, created by Manfred Bierwisch, ‘provides a model for the interaction of word knowledge and world knowledge in actual contexts of use’.312 Another competing theory is relevance-theoretic (RT) pragmatics, which, in contrast to GL, assumes that communication is essentially an inferential process.313 In comparison to the other competing theories, GL is the most elaborate formalized componential model and focuses more on polysemous patterns in the lexicon.314 The competing theories claim, correctly, that GL, ‘although incorporating a considerable amount of world knowledge into the lexicon, still leaves a lot of work for pragmatics to do in finding the interpretation intended by a speaker’.315 By integrating pragmatics into their models, these theories claim that they provide more unified, often simpler, accounts of meaning that avoid problems of misinterpretation.316 Nevertheless, it is the formal features of GL and the exclusive focus on semantics rather than pragmatics that make it the most useful method of developing ordinary meaning interpretations of legal texts. Most importantly, GL, as a default interpretation concept, does not require an enriched context, whereas the other approaches depend on one.317 In other words, the other theories depend heavily on considerable contextual information. Arguing for application of one of the pragmatic-based theories in lieu of GL misses the key contribution of GL to lexical semantics: a default interpretation based on sentential context alone. If legal texts always involved the kind of enriched contextual environment necessary for pragmatic analysis, the GL structure of a default interpretation that can be shaped by context would, arguably, be inappropriate. In fact, just the opposite is true. A legal text is not a contextualized lengthy conversation. When interpreting legal texts, the specific provisions themselves are of paramount importance and often the only evidence considered by the court.318 Not surprisingly, one of the advocates of RT argues that RT principles can be applied successfully to legal language but notes that ‘RT [as] a neo-Gricean theory focuses on conversation …’.319 This linguist correctly observes that ‘in the legal domain RT should lend itself more readily to theorising about the language of the courtroom, situations of lawyer-client consultations, or minority groups’ linguistic performance than texts’.320 GL Theory and the Judiciary’s Current Interpretive Tools In addition to its superiority in generating interpretations in nonenriched contexts, GL is also a desirable theory because it can be successfully integrated into the judiciary’s current interpretive framework with minimal disruption to interpretive doctrines other than ordinary meaning. Certainly, this article has argued that the judiciary’s meaning postulate approach to ordinary meaning determinations misses generalizations across relational properties of lexical items, and its ad hoc application of its other interpretive tools cannot cure this fundamental flaw.321 The adoption of GL, however, does not entail the abandonment of the judiciary’s other interpretive tools. Rather, in contrast to a pragmatic-based linguistic theory of meaning such as RT which makes no explicit allowance for input from other interpretive tools, GL contemplates the influence of other interpretive principles, particularly those of a pragmatic nature.322 Recall that this article proposes that GL be used to establish default defeasible ordinary meanings that can be shaped by courts’ traditional interpretive tools.323 GL is particularly congruent with the current interpretive regime because, while it provides a lexically based default interpretation, the current interpretive tools are either pragmatic in nature or present no conflict with GL principles.324 Consider a couple of the major categories of interpretive tools. Substantive canons, also known as ‘normative canons’ among other terms, are commonly described as policy-based directives about how statutory uncertainty should be resolved.325 These interpretive tools are applicable in cases involving a wide range of issues, including constitutional concerns, retroactive effects, divestment of judicial review or habeas corpus jurisdiction, extraterritorial application, negative impact on American Indians or immigrants, or various federalism concerns.326 Substantive canons, unlike GL, do not purport to offer theories about the meanings of words but, rather, direct courts to interpret statutes narrowly when Congress has not been sufficiently clear (based on the ordinary meaning of the language) that a broad interpretation was intended.327 Similar to substantive canons, textual canons do not present any conflicts with GL. Textual canons ‘set forth inferences that are usually drawn from the drafter’s choice of words, their grammatical placement in sentences, and their relationship to other parts of the “whole” statute’.328 Courts assume, for example, that Congress does not include superfluous statutory language,329 that when it enumerates things it does not intend to include things not listed,330 and that it intends for similar language within the same statutory section to have a consistent meaning.331 Although courts may apply textual canons when determining the ordinary meaning of language, inferences drawn from grammatical rules and nonsentential context can be applied consistently with GL. Unlike some other proposed interpretive theories, a proposal that judges should apply GL does not raise systemic co-ordination problems because the efficacy of the GL approach is not dependent on co-ordination by courts.332 The failure of some judges to apply GL does not, from a systemic standpoint, undermine the efforts of those courts that do apply the theory. Furthermore, not only is GL compatible with the judiciary’s current interpretive tools, the theory is more compatible than is the judiciary’s current ordinary meaning methodology.333 As argued in this article, the judiciary’s meaning postulate approach to ordinary meaning is, perhaps unwittingly, designed to capture ‘possible’ rather than ‘ordinary’ meaning.334 The current approach is especially troubling considering that the creative nature of natural languages means that new senses and possibilities can be expected to develop, resulting in surprising possibilities that Congress may not have intended.335 Such a methodology is inconsistent with the judiciary’s other interpretive tools which, as David Shapiro has explained, favor continuity, via narrow interpretations, over change.336 GL, by favoring prototypical, rather than unusual meanings, thus better fits the historical desire of courts to adopt broad interpretations cautiously. 7. Conclusion The judiciary’s unsystematic approach to ordinary meaning is unsurprising. Undoubtedly, the current approach is ad hoc because courts view the process of lexical semantics in overly simplistic terms. Typically, courts fail to appreciate that the combinatorial aspects of language create subtle interactions in meaning that make interpretation difficult (sometimes deceptively so). Nevertheless, linguistics and the law are possible because courts purport to take language seriously. The judiciary should, thus, be receptive to evidence that they chronically ignore the polymorphic nature of words and the distributed nature of meaning. The incompatibility of the judiciary’s current ordinary meaning methodology, which often features inappropriately broad interpretations that overextend the domain of statutes, and its natural inclination to favor continuity over change should stimulate the desire of courts to incorporate linguistic and cognitive insights into their interpretive approaches. As this article has shown, the flaw in the judiciary’s interpretive methodology can be addressed through a more sophisticated understanding of the way in which the meaning of any word is in some sense different in every distinct context in which it appears. GL can provide a theoretical framework for encoding selectional knowledge in natural language. Application of GL should cause the Court to change its practice from allowing pragmatic evidence to sometimes cancel a broad definition chosen on (typically faulty) semantic grounds to allowing pragmatic evidence to sometimes cancel a narrow definition chosen on semantic grounds. By adopting GL, the judiciary would be demonstrating a desire to bring important systemization to its ordinary meaning determinations and, at the same time, signaling a commitment to a true ‘linguistic turn’ in legal interpretation. The author would like to thank Steve Calandrillo, Miriam Cherry, Omar Dajani, William N. Eskridge, Brian Foley, Dr. Randal N M Graham, Linda Jellum, Brian Landsberg, Stephen Legomsky, Thom Main, James Pustejovsky, Sanford Shane, David L. Shapiro, John Sprankling, and Ted Sichelman for their views on an earlier draft of this article. The author would especially like to thank Professor Raul Aranovich and Professor John Hawkins and the students in their workshops at University of California, Davis, for their advice and assistance in writing this article. 1 See, e.g. Estelle v. McGuire 502 US 62, 72 (1991) (explaining that jury instructions are to be interpreted according to their ‘common and ordinary meaning’); Asgrow Seed Co v. Winterboer 513 US 179, 187 (1995) (‘When terms used in a statute are undefined, we give them their ordinary meaning’); Sanchez-Llamas v. Oregon 548 US 331, 346 (2006) (indicating that ‘international agreements’ are to be interpreted in accordance with their ordinary meaning); Baze v. Rees 553 US 35, 97 (2008) (Thomas J, concurring) (referring to the ‘ordinary meaning of the word “cruel”’ in connection with the Eighth Amendment); LM Solan ‘Contract as Agreement’ (2007) 83 Notre Dame L Rev 353, 383 (explaining the presumption that the ordinary meaning of a contract’s terms was the meaning intended by the parties). The concept might, in fact, be a constitutionally required mode of interpretation, at least in some circumstances. R Dickerson The Interpretation and Application of Statutes 10-11 (1975) (explaining that the ordinary meaning concept is a ‘constitutional assumption’). 2 For example, Zuni Pub. Sch. Dist. No. 89 v. Dept of Educ. 550 US 81, 90 (2007) (explaining that the Court was ‘depart[ing] from a normal order of discussion, namely an order that first considers Zuni's statutory language argument[,]’ ‘because of the technical nature of the language in question’). 3 B Jackson Making Sense in Law: Linguistic, Psychological and Semiotic Perspectives (Deborah Charles Publications 1995) 44 (stating that ‘ordinary content’ is the ‘content which would be assigned to a sentence by the “sensible speaker/hearer”’—an ‘idealized normative language user’); A Phillips Lawyers’ Language: How and Why Legal LanguageIs Different (Routledge 2003) 99 (stating that ‘what judges seem to have in mind using natural meaning is the meaning which is unforced, where the sense of the words is not distorted to fit the Procrustean bed of an argument’). 4 This article primarily focuses on the problematic nature of overly broad meanings rather than the ordinary meaning/technical meaning distinction, which has been discussed by others. For example, MJ Morrison ‘Excursions into the Nature of Legal Language’ (1989) 37 Cleveland State L Rev 271, 287–90 (discussing the distinction between technical and ordinary meanings). In fact, if a court is convinced that a technical meaning is appropriate, the principles introduced in this article can assist in excluding overly broad or narrow technical interpretations. 5 SA Thumma and JL Kirchmeier ‘The Lexicon Has Become a Fortress: The United States Supreme Court’s Use of Dictionaries’ (1999) 47 Buffalo L Rev 227, 252–60 (documenting the increased use of dictionaries by the Court). 6 For example, Carachuri-Rosendo v. Holder, 130 S Ct 2577, 2585 (2010) (looking at the ‘commonsense conception’ of statutory terms and concluding that the government argued ‘for a result that the English language tells us not to expect’). 7 CD Cunningham, JN Levi, GM Green and JP Kaplan ‘Plain Meaning and Hard Cases’ (1994) 103 Yale L J 1561, 1563 (explaining that ‘compared to analysis of a trained linguist, dictionaries are a crude and frequently unreliable aid to word meaning and usage’); C Goddard ‘Can Linguists Help Judges Know What They Mean? Linguistic Semantics in the Court-Room’ (1996) 3 Forensic Linguistics 250, 251 (explaining that ‘linguists generally take a dim view of the accuracy of dictionaries’). 8 A meaning postulate involves the definition of a word in terms of its relations with other words. ML Murphy and A Koskela Key Terms in Semantics (Continuum International Publishing Group 2010) 100. 9 Cf. Goddard, above n 7 at 252–53 (explaining that ‘any worthwhile explanation of a meaning must be framed in terms which are simpler and more easily understood than the word being defined’). 10 K Willems ‘Logical Polysemy and Variable Verb Valency’ (2006) 28 Language Sciences 580, 583 (explaining that words are inherently polysemous). 11 E Aprill ‘The Law of the Word: Dictionary Shopping in the Supreme Court’ (1998) 30 Arizona State L J 275, 297 (stating that the level of ‘linguistic analysis’ performed by courts rarely rises above ‘dictionary shopping’). 12 Context is crucial because dictionary meanings are fundamentally indeterminate. A Kilgarriff ‘Generative Lexicon Meets Corpus Data: The Case of Non-standard Word Uses’ in P Bouillon and F Busa (eds) The Language of Word Meaning (Cambridge University Press 2001) 314 (explaining that ‘any two dictionaries will differ in the meanings they say words have at innumerable points’). 13 J Pustejovsky The Generative Lexicon (Cambridge: The MIT Press, 1995) 59 [‘If lexical items are to be thought of as carrying several parameters (or dimensions) of interpretation, then the question immediately arises as to how a particular interpretation is arrived at in a given context.’]. 14 United Sav. Ass’n of Texas v. Timbers of Inwood Forest Assocs., Ltd 484 US 365, 371 (1988) (‘Statutory construction … is a holistic endeavor. A provision that may seem ambiguous in isolation is often clarified by the remainder of the statutory scheme … .’). 15 The broad purpose of the statute has long been thought to be relevant to the meaning of the text. KN Llewellyn ‘Remarks on the Theory of Appellate Decision and the Rules or Canons About How Statutes Are To Be Construed’ (1949) 3 Vanderbilt L Rev 395, 400 (‘If a statute is to make sense, it must be read in the light of some assumed purpose.’). As well, courts have long considered the legislative history of a provision when interpreting it. A Vermeule ‘Legislative History and the Limits of Judicial Competence: The Untold Story of Holy Trinity Church’ (1998) 50 Stanford L Rev 1833 [analyzing Church of Holy Trinity v. United States 143 US 457 (1892), generally recognized as the first case to sanction the use of legislative history]. 16 BG Slocum ‘The Importance of Being Ambiguous: Substantive Canons, Stare Decisis and the Central Role of Ambiguity Determinations in the Administrative State’ (2010) 69 Maryland L Rev 791, 815–29 (discussing the fictional nature of judicial claims that interpretive rules are based on congressional intent). 17 Cf. DA Cruse Lexical Semantics (Cambridge University Press 1986) 51 (explaining how ‘the meaning of any word form is in some sense different in every distinct context in which it occurs’). 18 ‘The principle of compositionality states that the meaning of a complex linguistic expression is built up from the meanings of its composite parts in a rule-governed fashion.’ Murphy and Koskela, above n 8 at 36. 19 Jackson, above n 3 at 86 (discussing how written communication allows for greater scientific objectivity). 20 For an explanation of ellipsis and the cognitive science explaining it, see below n 64–84 and accompanying text. 21 D Geeraerts Theories of Lexical Semantics (Oxford University Press 2010) 151 (referring to GL as the ‘most productive and sophisticated type of post-generativist formal decompositional semantics’). GL offers a structuralist conception of meaning because it views language ‘as a system, and not just as a loose bag of words’. Ibid at 48 (comparing language to a chess game). 22 For a more detailed explanation of the logical polysemy concept, see below n 85. 23 Kilgarriff, above n 12 at 326 (observing that GL’s aim to capture generalizations about words will ‘tend to make the lexicon smaller’). 24 ‘A linguistic theory is generative if it consists of a finite set of rules and principles that generate all the possible expressions in a language while ruling out those that are not possible.’ Murphy and Koskela, above n 8 at 74. 25 Kilgarriff, above n 12 at 314 (explaining that a dictionary would specify as separate senses ‘those productive uses that the GL would explain as the outcome of generative processes’). 26 J Pustejovsky Introduction to Generative Lexicon (2006) 1 (unpublished manuscript, on file with author); see also Pustejovsky, above n 13 at 42 (explaining that an assumption of GL is that ‘distinctions in interpretation are real, systematic, and part of the language itself’). 27 Although statutory interpretation is the focus of this article, the methodology introduced can be readily applied in other contexts. 28 JF Manning ‘What Divides Textualists from Purposivists’ (2006) 106 Columbia L Rev 70 (describing the textualist methodology). 29 See below section 6(B) (describing the various interpretive tools that may be applied consistently with GL). 30 See generally AW Murphy ‘Old Maxims Never Die: The “Plain-Meaning Rule” and Statutory Interpretation in the “Modern” Federal Courts’ (1975) 75 Columbia L Rev 1299 (describing the plain meaning doctrine). 31 M Galdia Legal Linguistics (Peter Lang Pub, Inc. 2009) 51–2 (explaining that the linguistic turn ‘did not take place in the legal science’). 32 For one notable exception to the general scholarly avoidance, see Cunningham et al., above n 7 (focusing on whether linguists can determine empirically whether a phrase is ambiguous as a matter of ordinary language). 33 LM Solan The Language of Judges (University Of Chicago Press 1993) 62 (arguing that ‘judges do not make good linguists because they are using linguistic principles to accomplish an agenda distinct from the principles about which they write’). 34 This article does not provide a typology of cases in which the application of GL would have changed the interpretation a court selected. Instead, this article provides an in-depth application of GL to one particular interpretive problem. An adequate understanding of GL and its application should suffice to convince readers that the theory would be important in a nonnegligible number of cases. 35 The statute also applies to anyone who ‘carries’ or, since an amendment to the statute in 1998, ‘possesses’ a firearm ‘during and in relation to … [a] drug trafficking crime’. 18 USC §924(c)(1)(A). In the cases described in this section, the defendants were all charged with ‘use’ of the firearm. 36 An ‘event’ is a type of situation in which something happens. Murphy and Koskela, above n 8 at 65, 156. See below n 106–28 and accompanying text (discussing in more detail events and the role they play in GL). 37 For a description of ellipsis, see below n 64–6 and accompanying text. 38 508 US 223 (1993). 39 Ibid at 228. Because the defendant traded a ‘machinegun’, the sentence was 30 years. Ibid at 226. 40 Ibid. 41 Ibid at 229. 42 Ibid. 43 Ibid at 230. 44 Ibid. 45 Ibid at 229. 46 Ibid at 230. 47 Ibid at 229. 48 Ibid at 241–42 [quoting the Court’s decision in Deal v. United States 508 US 129, 132 (1993), that the ‘meaning of a word cannot be determined in isolation’]. 49 Ibid at 242. Justice Scalia pointed out that ‘[w]hen someone asks, “Do you use a cane?,” he is not inquiring whether you have your grandfather’s silver-handled walking stick on display in the hall; he wants to know whether you walk with a cane’. Ibid. 50 Ibid. 51 Ibid at 244. 52 516 US 137 (1995). 53 Ibid at 143. 54 Ibid at 145. 55 Ibid. 56 Ibid. 57 552 US 74 (2007). 58 Ibid at 79. 59 Ibid. 60 Ibid. 61 Ibid. 62 508 US at 240. 63 552 US at 585. 64 R May Encyclopedia of Cognitive Science (Macmillan Publishers London 2002) 1094. 65 JA Hawkins Efficiency and Complexity in Grammars (Oxford University Press 2004) 38. 66 Ibid. The sentence without ellipsis would read: Max went to the store, and Oscar went to the store, too. 67 MJ Traxler, B McElree, RS Williams and MJ Pickering ‘Context Effects in Coercion: Evidence from Eye Movements’ (2005) 53 J Memory and Language 1, 2. 68 Ibid. 69 Ibid. 70 Ibid. (‘When the complement of these verbs is a noun phrase [NP] with a default interpretation as an entity, it cannot be combined straightforwardly with the verb to form an interpretable verb phrase.’) 71 Ibid at 1–2 (explaining that these ‘common and seemingly simple expressions appear to require a richer form of composition’). 72 This is a familiar example used by Traxler and in the GL literature. For example, Pustejovsky, above n 13 at 116. 73 Traxler, above n 67 at 1–2. 74 Ibid at 2 (explaining that the ‘observed coercion cost is specific to the type of NP complement and not because verbs like begin engender processing difficulty regardless of the semantic type of its complement’). 75 Ibid (explaining that the interpreter must engage in complement coercion and shift the complement from an entity to an event). Traxler describes complement coercion as a ‘process by which a context-sensitive interpretation for a complement expression emerges from the interaction of semantic properties in the sentence and discourse’. Ibid. Coercion is one of the generative mechanisms of GL. See below n 140 and accompanying text. 76 In contrast, The author began the book is typically interpreted as The author began writing the book. Traxler, above n 67 at 2. 77 Ibid. 78 Ibid. 79 Ibid (stating that ‘[f]ull event structures tend to occur only with less predictable activities like translate the book’). 80 Ibid at 5. A second study has confirmed the hypothesis that the increased processing cost reflects the online construction of an event sense for the complement. S Frisson and B McElree ‘Complement Coercion Is Not Modulated by Competition: Evidence from Eye Movements’ (2008) 34 J Experimental Psychology: Learning, Memory and Cognition 1, 1; see also B McElree, L Pylkkanen, M Pickering and M Traxler ‘A Time Course Analysis of Enriched Composition’ (2006) 13 Psychonomic Bulletin and Rev 53 (finding that enriched composition requires the online deployment of complex compositional operations). 81 Traxler, above n 67 at 3. Traxler argues that his studies ‘are at odds with approaches that would solely attribute the cost of processing coercing expressions to inferring or selecting the action inherent in the event sense’. Ibid at 17. 82 Hawkins, above n 65 at 267. 83 Ibid. 84 E Klepousniotou, D Titone and C Romero ‘Making Sense of Word Senses: The Comprehension of Polysemy Depends on Sense Overlap’ (2008) 34 J Experimental Psychology: Learning, Memory, and Cognition 1534, 1534. Of course, a ‘subordinate context may increase activation of subordinate meanings[.]’ Ibid. 85 See below n 164–8 and accompanying text (describing the nature of the verb use). Pustejovsky defines ‘logical polysemy as a complementary ambiguity where there is no change in lexical category, and the multiple senses of the word have overlapping, dependent, or shared meanings’. Pustejovsky, above n 13 at 28; see also Geeraerts, above n 21 at 148 (referring to logical polysemy as regular polysemy and describing it as referring to the ‘existence of polysemous patterns in the lexicon’). By ‘no change in lexical category’, Pustejovsky is referring to a situation where the two senses are not homonyms. HM de Melo Moura ‘Lexicon and Context in the Production of Meaning’ (2002) 12 Revue De Semamtiqur Et Pragmatique 17, 18. 86 In addition to GL, courts can of course apply their traditional interpretive tools to shape the presumptive meanings created by GL. See below section 6 (explaining how GL interacts with the traditional interpretive tools used by courts). 87 R Jackendoff Meaning and the Lexicon: The Parallel Architecture 1975–2010 (Oxford University Press 2010) 2. In contrast to the composite view of word meanings, Jackendoff explains that ‘in classical semantics, word meanings (except for words with logical properties) are typically taken to be atomic’. Ibid. ‘An atomic semantic entity is one that is not analyzed in terms of smaller components that it is made of.’ Murphy and Koskela, above n 8 at 21. 88 Pustejovsky explains the two major lines of inquiry of GL as follows: ‘(1) How is it that we are able to deploy a finite number of words in our language in an unbounded number of contexts? (2) Is lexical information and the representations used in composing meanings separable from our commonsense knowledge?’ Pustejovsky, above n 26 at 1; see also J Pustejovsky ‘Type Theory and Lexical Decomposition’ (2006) 7 J Cognitive Science 39, 59 (‘Overall, GL is concerned with explaining the creative use of language; we consider the lexicon to be the key repository holding much of the information underlying this phenomenon.’). 89 Willems, above n 10 at 589–90 (explaining that GL ‘presents an analytic framework which can account for the disambiguation process involved and which is compatible with the compositional approach’). 90 Pustejovsky, above n 26 at 7. 91 Geeraerts, above n 21 at 149. 92 Ibid at 149–51. 93 Pustejovsky, above n 13 at 61. 94 Ibid at 6–7 (arguing that ‘lexical semantics must study all syntactic categories in order to characterize the semantics of natural language’). 95 Pustejovsky, above n 88 at 60. 96 Murphy and Koskela, above n 8 at 110. 97 Pustejovksy, above n 88 at 61. See also below n 107–29 and accompanying text (discussing GL’s conception of types in more detail). 98 Pustejovsky, above n 13 at 62–7. 99 MM Abdulkhaliq, English Noun Classes and Generative Lexical Mechanisms (2001) 6 (unpublished PhD dissertation, Stanford University, on file with Stanford University Library). 100 Pustejovsky, above n 13 at 63 (describing them as ‘[s]yntactically realized parameters of the lexical item’). 101 Abdulkhaliq, above n 99 at 6. 102 Pustejovsky, above n 13 at 63 (describing them as ‘[p]arameters which participate in the logical expressions of the qualia, but which are not necessarily expressed syntactically’). The following is Pustejovsky’s example: ‘Mary buttered her toast with an expensive butter’. Ibid at 64. 103 Ibid at 63–4 (describing them as parameters which ‘can be expressed only by operations of subtyping or discourse specification’). 104 Abdulkhaliq, above n 99 at 6. 105 Ibid at 7; see also Pustejovksy, above n 88 at 64 (describing them as ‘[p]arameters which modify the logical expression, but are part of the situational interpretation, and are not tied to any particular lexical item’s semantic representation’). 106 J Pustejovsky ‘The Syntax of Event Structure’ in B Levin and S Pinker (eds) Lexical and Conceptual Semantics (Oxford: Blackwell, 1992) 47. 107 Ibid. 108 Abdulkhaliq, above n 99 at 11. 109 Ibid. Pustejovsky explains that a state denotes a ‘single event evaluated relative to no other event’. Pustejovsky, above n 106 at 50. 110 Abdulkhaliq, above n 99 at 11. An entailment ‘is the propositional relation in which if one proposition is true, then it is always the case that the related proposition is true’. Murphy and Koskela, above n 8 at 63. 111 Abdulkhaliq, above n 99 at 11. 112 Ibid. 113 Ibid. 114 Ibid at 12. 115 Ibid. 116 Ibid. 117 Ibid. 118 Ibid. 119 Ibid. 120 Pustejovsky, above n 13 at 68. See also IL Falkum ‘Generativity, Relevance and the Problem of Polysemy’ (2007) 19 UCL Working Papers in Linguistics 205, 209. 121 Ibid. 122 Ibid at 69. 123 Falkum, above n 120 at 209. 124 Pustejovsky, above n 13 at 69. 125 Ibid at 69. 126 Falkum, above n 120 at 209. 127 Pustejovsky, above n 13 at 70. 128 Ibid at 72. 129 Ibid at 77. 130 Falkum, above n 120 at 208. 131 Pustejovsky, above n 13 at 87. 132 J Pustejovksy ‘Generativity and Explanation in Semantics: A Reply to Fodor and Lepore’ (1998) 29 Linguistic Inquiry, 289, 294. 133 Pustejovksy, above n 13 at 76 (stating that the qualia structure ‘emphasizes aspects of a word’s meaning that have been largely ignored by formal treatments of lexical semantics’). 134 Ibid (arguing that ‘[b]y looking at modes of explanation for a word, we permit a much richer description of meaning than either a simple decompositional view or a purely relational approach to word meaning would allow’). 135 Ibid at 85. 136 Ibid. See also Falkum, above n 120 at 210. 137 Pustejovsky, above n 13 at 86. See also Abdulkhaliq, above n 99 at 9. 138 Ibid (explaining that ‘Pustejovsky typically makes the value of the AGENTIVE quale predicate for verbs the name of the lexical entry followed by _act, expressing that an event comes into being through performing the event’). 139 Pustejovsky, above n 13 at 105. 140 Ibid at 105–31; Pustejovsky, above n 88 at 23. 141 Pustejovsky, above n 88 at 40. 142 Ibid at 41. 143 See generally B Pierce Types and Programming Languages (The MIT Press 2002) (describing type systems). 144 See above n 95–7 and accompanying text (describing the three levels). 145 Pustejovsky, above n 88 at 61. 146 Ibid at 65 (explaining that artifacts refer to ‘artifactually constructed objects, or natural objects that have been assigned or assume some function or use’). See also J Pustejovsky and A Rumshisky ‘Between Chaos and Structure: Interpreting Lexical Data Through a Theoretical Lens’ (2008) 21 Intl J Lexicography 337, 340–1. 147 J Pustejovsky and E Jezek ‘Semantic Coercion in Language: Beyond Distributional Analysis’ (2008) 20 Italian J Linguistics/Rivista di Linguistica 175 [stating that a ‘simple natural physical object … can be given a function (i.e., a TELIC role), and transformed into an artifactual type’]. For an example, see below n 247–and accompanying text (analyzing enjoy the rock). 148 Pustejovsky, above n 13 at 122. 149 Abdulkhaliq, above n 99 at 19–20 (indicating that ‘[p]rime cases of Co-composition involve weak verbs such as open, break, and use’). In the same way, in Accommodation, ‘[t]he type a function requires is inherited through the type of the argument’. Pustejovsky, above n 88 at 68. 150 BT Atkins, J Kegl and B Levin ‘Anatomy of a Verb Entry: From Linguistic Theory to Lexicographic Practice’ (1988) 1 Intl J Lexicography 84. 151 Pustejovsky, above n 13 at 123 (explaining that ‘there is only one sense for bake, and … any other readings are derived through generative mechanisms in composition with its arguments’). 152 Ibid (providing a structural representation for bake). 153 Pustejovksy, above n 13 at 124. 154 Ibid. 155 Ibid (explaining that ‘the creation sense of bake is contributed in part by the meaning of a cake, by virtue of it being an artifact’). 156 Ibid at 123 (giving a structural representation for cake). 157 Ibid at 124. 158 Ibid at 123 [explaining ‘why nouns such as cake, bread and cookie “shift” the meaning of the verb bake, while other nouns (such as potato and garlic) do not’]. 159 Ibid. 160 Ibid at 124. 161 Ibid. 162 Ibid. 163 Ibid. 164 Bailey v. United States 516 US 137 (1995) (indicating that ‘use’ must be given its ‘ordinary or natural’ meaning). 165 See above n 26 (explaining that compositionality stands for the principle that the meaning of a complex linguistic expression is built up from the meanings of its composite parts in a rule-governed fashion). Plainly, the Court never recognized that use is the type of verb that depends heavily on its arguments for semantic content. Abdulkhaliq, above n 99 at 20. 166 Pustejovsky, above n 13 at 87 (‘The utility of a verb such as use is that it is semantically light …, what I will refer to as simply underspecified, with respect to the particular activity being performed.’); E. Ritter and S. Rosen ‘Strong and Weak Predicates: Reducing the Lexical Burden’ (1996) 26 Linguistic Analysis 29; see also Pustejovsky, above n 132 at 305 [explaining that ‘verbs that allow a multiplicity of expressions … have a complex relational type (i.e., polymorphic type)’]. 167 In his dissent in Smith v. United States 508 US 223 (1993), Justice Scalia also recognized the unspecified nature of use. Ibid at 242. 168 Pustejovsky, above n 132 at 303. 169 See above n 70–84 (discussing cognitive studies on the impact of ellipsis on interpretation). 170 Abdulkhaliq, above n 99 at 68 (stating that weak verbs ‘combine with their objects to form a complex semantic structure’, but by themselves the verbs ‘would not have sufficient semantic resources to create’ a complex semantic structure). 171 The examples are taken from Pustejovksy, above n 13 at 87. 172 Ibid. 173 Ibid. 174 Ibid. Pustejovsky explains that for the examples in (7), ‘[t]he factors allowing us to determine which sense is appropriate for any of these cases are twofold: (1) the qualia structure for each phrase in the construction; and (2) a richer mode of composition, which is able to take advantage of this qualia information’. Ibid. 175 See above n 110 and accompanying text. 176 Ibid. Note that, according to B Comrie Tense (Cambridge University Press 1985) 130, the definition of English past tense does not exclude the possibility that the event in question continues in the present and will continue into the future. 177 Abdulkhaliq, above n 99 at 12. 178 Ibid. 179 Ibid. 180 Pustejovsky, above n 13 at 63. 181 See above n 171 and accompanying text. 182 The D-ARG, which explains the purpose of the use, can be omitted. 183 This example is taken from Abdulkhaliq, above n 99 at 84. 184 See above n 138 and accompanying text. 185 Pustejovsky often omits the other qualia from the qualia structure of verbs when they are unnecessary or inappropriate. 186 Abdulkhaliq, above n 99 at 9. 187 Ibid. 188 Ibid at 9–10. 189 Ibid at 10. 190 Ibid. 191 Not including an event structure for nouns, but including them for verbs, adjectives, and nominalizations, is an abbreviatory convention common in GL theory. Ibid at 13 (‘Not indicating an event structure is another abbreviatory convention frequent in Generative Lexicon theory.’). 192 See above n 116 and accompanying text (explaining accomplishments). 193 Abdulkhaliq, above n 99 at 9. 194 See above n 137 and accompanying text (explaining that the TELIC role indicates the purpose and function of an object). 195 See above n 138 and accompanying text (explaining the AGENTIVE predicate). 196 See above n 148–63 and accompanying text (explaining the co-composition generative mechanism). 197 Abdulkhaliq, above n 99 at 68. 198 Ibid. 199 See above section 2(A) (explaining the Supreme Court’s flawed interpretation of use). 200 See below n 268–72 and accompanying text (explaining that GL creates default interpretations, not necessarily final interpretations). 201 This objection, if true, might undermine GL to some degree but would not validate the Supreme Court’s multiple misinterpretations of use. 202 One possibility is that considering that the statute also penalizes a defendant who possesses or carries a firearm, see above n 35, the ordinary meaning of uses is that the defendant must shoot the gun. This argument carries some weight, but it is important that the ordinary meaning generated by GL be plausible regardless of the specific statutory context. 203 Due to the underspecified nature of natural languages, discretion is necessarily endemic to any interpretive endeavor. See below section 5(C)(ii). (discussing the underspecified nature of language). 204 The TELIC structure indicates the purpose or use of an object, and as long as the object is being used consistently with the indicated purpose or use, the GL framework is being used properly. 205 508 US 223 (1993). 206 Ibid at 229. 207 The Supreme Court thus had the wrong focus. The question was not whether use should have a technical or ordinary meaning, or the ordinary meaning of ‘use’ in isolation, but rather how use and a direct object combine to form a compositional meaning. 208 552 US 74 (2007). 209 Ibid at 79. The Court’s interpretation only makes sense if, in the language of §924(c)(1)(A), any ‘relation to … [a] drug trafficking crime’ is deemed to have been terminated immediately after the sale occurs, which would be an unjustifiably narrow interpretation. 210 See below section 6(B) (describing pragmatic considerations). 211 Cf. Pustejovsky, above n 132 at 304 [explaining that ‘certain verbs of transfer such as get or rent can refer to both directions of transfer, thus allowing multiple expressions such as John got a book for Mary (give) and John got a book from Mary (take). In contrast, verbs such as give and sell, which are fully specified for the directionality of the transfer, do not allow the same range of prepositional phrases …’]. 212 516 US 137 (1995). 213 Ibid at 145. 214 The application of GL to §924(c)(1)(A) thus demonstrates that Watson was correctly decided but for the wrong reasons. If Smith had been correctly decided, and the firearm had to be used as a weapon, there would have been no need to decide the interpretive issue in Watson. Arguably, the Court’s broad holding in Smith also influenced it to make a questionable interpretation in Bailey when it held that use does not include the ‘action of a defendant who puts a gun into place to protect drugs or to embolden himself’. 516 US 137, 145 (1995). See above n 204 and accompanying text (arguing that the statute is satisfied when the firearm is used agentively in relation to its TELIC function). 215 See below n 273–81 and accompanying text (discussing the role of pragmatics in forming interpretations). It is not the case that GL is the only theory capable of illustrating the errors inherent in the Supreme Court’s interpretations of 18 USC §924(c)(1)(A). For example, a simple way to see the case would be in terms of Inferential Role Semantics, where the following rule would apply: X uses Y → X uses Y to do whatever is normally done with Y to do whatever is normally done with Y. Y Wilks ‘The “Fodor”-FODOR Fallacy Bites Back’ in P Bouillon and F Busa (eds) The Language of Word Meaning (Cambridge University Press 2001) 80. Other theories would likewise undermine the Court’s approach to ordinary meaning. GL is adopted in the article because it is the most advanced approach among the formal componential theories of semantics and thus the most appropriate for providing default ordinary meaning interpretations. See above n 21 and accompanying text. 216 See below n 268–72 and accompanying text (explaining that the function of GL is to generate default ordinary meaning interpretations, which precludes adoption of idiosyncratic meanings). 217 See above n 48–51 and accompanying text (describing Justice Scalia’s dissenting opinion in Smith). 218 Aprill, above n 11, at 315–30; see also M McGowan ‘Do as I Do, Not as I Say: An Empirical Investigation of Justice Scalia’s Ordinary Meaning Method of Statutory Interpretation’ (2008) 78 Mississippi L J 129 (criticizing Justice Scalia’s ordinary meaning approach as inconsistent). 219 Pustejovsky, above n 13 at 293. 220 See above section 3(B) (describing GL’s generative mechanisms that operate in connection with the features of the relevant lexical items). 221 See above n 5–13 and accompanying text (describing the judiciary’s common practice of relying on dictionary definitions to define lexical items in isolation). 222 See above section 2(B) (describing how the Supreme Court’s approach to ordinary meaning interpretation ignores cognitive science insights). 223 JA Fodor and E Lepore ‘The Emptiness of the Lexicon: Reflections on James Pustejovsky’s The Generative Lexicon’ (1998) 29 Linguistic Inquiry, 269–88. FL believe in holism, which is the proposition that not all the properties of a given system can be determined or explained by its component parts alone. M Silcox ‘Semantic Holism vs. Semantic Atomism’ (2005) 25 Language and Communication 335, 335 (‘Semantic holism is the thesis that the smallest independently significant bearer of linguistic meaning is an entire human language.’). 224 Fodor and Lepore, above n 223 at 281. 225 Thus, the claim is that the phrase or sentence at issue is semantical and GL would recognize it as such but would give the phrase or sentence an erroneous meaning. 226 These examples are taken from Fodor and Lepore, above n 223 at 280. 227 See above n 156–69 and accompanying text (discussing the examples in (10)(a) and (10)(b)). 228 Ibid. 229 Ibid. 230 Pustejovsky, above n 13 at 98 (explaining that the semantic generalization is ‘that certain objects come into being by virtue of an activity which can otherwise simply change the internal state of an object. The sense is the same, but manifests itself with overall effects differently, resulting in the “sense in context” phenomenon’). 231 Fodor and Lepore, above n 223 at 282. 232 FL make the same argument with respect to bake a knife. Ibid at 280. 233 Ibid at 282. 234 U.S. Nat’l Bank of Or. v. Indep. Ins. Agents of Am., Inc 508 US 439, 462 (1993) (correcting misplaced punctuation). 235 Green v. Bock Laundry Mach. Co 490 US 504 (1989) (recognizing the existence of the absurdity doctrine). 236 Besides the absurdity and scriverner’s error doctrines, another possibility is the void-for-vagueness doctrine. A judge applying void-for-vagueness doctrine does not, however, find the statutory language to be unsemantical but, rather, that it ‘fails to provide a person of ordinary intelligence fair notice of what is prohibited, or is so standardless that it authorizes or encourages seriously discriminatory enforcement’. U.S. v. Williams 553 US 285, 304 (2008). 237 The rule against surplusage stands for the principle that courts aim to give an independent meaning to every term in a statute. Duncan v. Walker 533 US 167, 174 (2001) (stating that it is ‘a cardinal principle of statutory construction’ that ‘a statute ought, upon the whole, to be so construed that, if it can be prevented, no clause sentence or world shall be superfluous, void, or insignificant’). 238 Wilks, above n 215 at 79. 239 See below n 273–78 and accompanying text (explaining the distinction between pragmatics, which often relies on extra-linguistic world knowledge, and semantics, which involves lexically encoded information). 240 See above n 137 and accompanying text (explaining the TELIC role). 241 Abdulkhaliq, above n 99 at 19–20. 242 For example, the trolley car could be the shape of a gingerbread cookie, although that is not likely the interpretation intended by FL. 243 Fodor and Lepore, above n 223 at 281. 244 Ibid. 245 This generative mechanism was refined, if not introduced, by Pustejovsky subsequent to FL’s criticism. 246 Pustejovsky, above n 132 at 294. 247 Presumably, this new information would be added to the word’s QUALIA structure, which raises a new objection. FL state, correctly, that ‘the more context sensitive the meanings of governing expressions are supposed to be, the richer must be the lexical entries for the expressions that they govern’. Fodor and Lepore, above n 223 at 279. There is no evidence, however, that the GL system would become unmanageable. In fact, GL may make the lexicon smaller. Kilgarriff, above n 12 at 325 (explaining that ‘a strategy for capturing generalizations … will tend to make the lexicon smaller: it will take less bytes to express the same set of facts. GL is associated with a compact lexicon, in this sense’). 248 Pustejovsky, above n 132 at 294. Although the concept of co-composition with the subject NP was generally unexplored in Pustejovsky’s early work, he acknowledged the phenomenon and incorporates it into his later work in GL. Pustejovsky, above n 13 at 69 (using examples of subject control); see also K Bibok ‘Word Meaning and Lexical Pragmatics’ (2004) 51 Acta Linguistica Hungarica 265, 283–84 (discussing why it is necessary in some cases to take into account nouns in the subject position). One example of the subject being crucial to the interpretation would be in the following (familiar) examples: the author began the book or the translator began the book. The interpretations would be the author began writing the book and the translator began translating the book. Ibid at 283, 300 (explaining that ‘if in the position of syntactic subject of an utterance there is a noun which refers to a person’s profession, the hearer may infer another event connected to book’). In contrast, in the example considered earlier, Mary began a book, the interpretation would be Mary began reading a book. 249 Traxler, above n 67 at 3. 250 Contrast the nonstandard uses of a word limitation with GL’s ability to capture combinatorial creativity, as the enjoyed the rock example illustrates. 251 A nonstandard use is defined as not matching a literal reading of any of the word’s definitions from the HECTOR lexical database, which comprises a dictionary of several hundred words and a corpus in which all the occurrences of those words have been manually sense tagged by professional lexicographers. Kilgarriff, above n 12 at 315. 252 Ibid at 320. Here, of course, the object is the company. 253 Of the 2276 corpus instances examined, 41 instances were found to be nonstandard word uses. Ibid at 322. Of the 41, only 2 were ‘plausible candidates for GL treatment’. Ibid. 254 Ibid at 320 (analyzing handbag data from the British National Corpus). 255 According to Kilgarriff with regard to nonstandard uses of words, in order to discern the intended meaning, a prominent role in the analysis should be taken by extensive knowledge of how words have deviated from their norms before. Ibid at 326. This is because nonstandard uses generally build on similar uses. Ibid at 325. Thus, ‘an appropriate model must contain both particular knowledge about some nonstandard interpretation, and reasoning to make the nonstandard interpretation fit the current context’. Ibid at 324. 256 See above n 1–4 and accompanying text (explaining the ordinary meaning concept). 257 Kilgarriff, above n 12 at 326 (explaining that ‘it is always open to language users to exploit the potential, carrying just a strand across to some new meaning’). 258 Ibid at 325. 259 Falkum, above n 120 at 218. R Blutner ‘Lexical Semantics and Pragmatics’ (2002) 10 Linguistische Berichte 10, 43, and Fodor and Lepore, above n 223 at 274, make the same claim. For a definition of entailment, see above n 111. 260 Falkum, above n 120 at 220. 261 Fodor and Lepore, above n 223 at 280. 262 Ibid. 263 Falkum, above n 120 at 221. 264 Ibid (arguing that GL’s lexical entries ‘do not specify sufficient conditions for the application of a concept’). 265 Wilks, above n 215 at 79 (arguing that a ‘“baked potato” is fine but a “baked cake” is not’). 266 Pustejovsky, above n 132 at 301, n. 3. 267 Pustejovsky, above n 13 at 42 (arguing against the notion that GL accounts for data in a post hoc fashion). 268 See below n 273–81 and accompanying text (describing pragmatics). 269 A context-independent linguistically encoded meaning is distinguished from an encyclopedic meaning, which is dependent on world knowledge (i.e. pragmatics). Murphy and Koskela, above n 8 at 62–3. 270 Jackson, above n 3 at 87. 271 Pustejovsky, above n 13 at 42; see also Goddard, above n 7 at 254 (explaining that a ‘word can normally be extended away from its clear core of meaning in directions that not all speakers will agree on and which are very difficult to pin down’); Jackson, above n 3 at 42 (referring to the ‘semic core’ of a word as its literal meaning). 272 Pustejovksy, above n 132 at 304 (explaining that GL’s interpretations are ‘facilitated by default interpretations of properties and activities associated with objects’), 307 (indicating that criticisms of GL confuse ‘the set of possible identifiable inferences with those that are automatic and unsurprising’); Pustejovksy, above n 13 at 87 (indicating that the GL mechanisms ‘“suggest” interpretations of words in context’), 100 (referring to ‘defeasible’ interpretations), 232–39 (noting the role for pragmatics in connection with GL). 273 K Bach ‘Semantic, Pragmatic’, in JK Campbell, M O’Rourke and D Shier (eds) Meaning and Truth (New York: Seven Bridges Press, 2002) 284. The precise point where semantics ends and pragmatics beings is a source of scholarly debate. One useful view is that ‘lexical pragmatics investigates the mechanisms by which linguistically-specified word meanings are modified in use’. R Blutner ‘Lexical Pragmatics’ in L Cummings (ed.) The Pragmatics Encyclopedia (London: Routledge, 2009) 1. Thus, whereas semantic information is encoded in what is uttered, pragmatic information is generated by the act of uttering it. 274 SC Levinson Pragmatics (Cambridge University Press 1983) 3; Bach, above n 273 at 293 (explaining that some sentences are semantically incomplete and require pragmatic supplementation). 275 Levinson, above n 274 at 20. 276 Falkum’s assertion, above n 120 at 221, that ‘Pustejovsky’s lexical entries do not specify sufficient conditions for the application of a concept’ is thus correct but does not undermine GL because GL only purports to provide default interpretations that can be shaped by context. 277 Pustejovsky, above n 13 at 236 (indicating that ‘lexical semantics can be made sensitive to sentence level compositional processes as well as discourse level inference mechanisms’). 278 Ibid (acknowledging the role of pragmatics in interpretation but not offering any pragmatic theories). 279 I Witczak-Plisiecka ‘A Note on the Linguistic (In)Determinacy in the Legal Context’ (2009) 5 Lodz Papers in Pragmatics 201, 205; see also T Wasow, A Perfors and D Beaver ‘The Puzzle of Ambiguity’ in Morphology and The Web of Grammar: Essays in Memory of Steven G. Lapointe (CSLI Publications 2005) 4–7 (pointing to evidence that English is ‘massively ambiguous’). 280 Galdia, above n 31 at 93 (indicating that ‘[l]aw does provide us with such comfort’). 281 Jackson, above n 3 at 86 (explaining that ‘autoglottic space’ is the feature which makes possible deductive logic). 282 V Bhatia, J Engberg, M Gotti and D Heller ‘Introduction’ in Vagueness in Normative Texts (Bern: Peter Lang, 2005). 283 M Paradix ‘Just Reasonable, Can Linguistic Analysis Help Us Know What It Is To Be Reasonable?’ (2007) 47 Jurimetrics J 169, 170. 284 Of course, as this article argues, in many cases ellipsis should not cause interpretive problems. See above n 68–70 and accompanying text (describing studies indicating that full event structures tend to occur only with less predictable activities). 285 Pustejovsky, above n 132 at 300, 308 (indicating that GL seeks to model systematicity in the lexicon). 286 Thus, there is a distinction between the kind of ellipsis which does not prevent accurate interpretations and the kind of underspecified language which does. See above n 66–85 and accompanying text (discussing ellipsis). 287 HLA Hart The Concept of Law (Oxford University Press 1961) 297 (arguing that all descriptions are necessarily incomplete because ‘all concepts are associated with an indefinite number of semantic features, some of which have not yet been noticed, and perhaps never will be noticed by anyone’). 288 Murphy and Koskela, above n 8 at 122–23. For example, cousin can refer to either a male or a female and is thus vague with respect to gender. Ibid at 123. 289 R Sorensen ‘The Ambiguity of Vagueness and Precision’ (1989) 70 Pacific Philosophical Q, 174, 174 (calling this sense of vagueness the ‘borderliner sense’). 290 Wasow et al., above n 279 at 2. 291 This is a well-known example that is hypothetical but is based on an actual case (which had a somewhat different statute). For an interesting discussion of the hypothetical, see F Schauer ‘A Critical Guide to Vehicles in the Park’ (2008) 83 New York U L Rev 1109. 292 Sorensen, above n 289 at 175. 293 Ibid. 294 18 USC §1231(a)(6). 295 Zadvydas v. Davis 533 US 678, 697 (2001); see also BG Slocum ‘Canons, The Plenary Power Doctrine, and Immigration Law’ (2007) 34 Florida State U L Rev 363, 394–97 (discussing the case). The Court focused on the alleged ambiguity of ‘may’, illustrating again the judiciary’s unfortunate habit of defining words without consideration of sentential context. Ibid. The question in the case was, in reality, essentially a judgment call: should the lack of specificity regarding the detainment period be seen as an indication that Congress desired the Attorney General to have authority to indefinitely detain? 296 533 US at 701 (creating the ‘presumptively reasonable period of detention’ in order to give the government and courts guidance). 297 See above n 21–30 and accompanying text (describing GL). 298 J Jayez ‘Underspecification, Context Selection, and Generativity’ in P Bouillon and F Busa (eds) The Language of Word Meaning (Cambridge University Press 1999) 124 (arguing that the ‘way in which underspecified representations give rise to more precise interpretations in particular contexts is not always clear’). 299 Recall that GL is being used to generate default interpretations that can be shaped, if appropriate, by context. See above n 272–77 and accompanying text. 300 Fodor and Lepore, above n 223 at 274. 301 Ibid. 302 Pustejovsky, above n 132 at 307. 303 This is especially true considering that the subway is used every day. A common attack on GL by scholars is to posit unusual interpretations as a way of undermining the interpretations generated by GL. Thus, Jayez argues that I am waiting for the bus could mean I am waiting for the bus to take a picture of it. Jayez, above n 298 at 143 (arguing that the TELIC role notion is ‘loose’). These criticisms miss the point of GL and certainly miss it as this article advocates it should be used in interpreting legal texts. Jayez argues that it ‘seems difficult to predict all and only the correct interpretations by postulating all-purpose types or features …’. Ibid at 141. Of course, GL only provides default interpretations based on the ordinary meaning of the phrase or sentence and allows the interpretations to be subject to pragmatic shaping. The only valid course of attack, then, is to allege that GL fails to generate the ordinary meanings of words, not that it fails to predict idiosyncratic meanings. 304 See above n 137 and accompanying text (describing the concept of telicity). 305 This would not be a valid conclusion. See above n 66–80 and accompanying text (discussing cognitive science studies of ellipsis). 306 Fodor and Lepore, above n 223 at 281. Interestingly, FL concede that ‘if X uses Y, where Y is something that has a conventional/typical use, then the invited inference is that X uses Y for what Y is used for’. Ibid. 307 For example, C Hoffman ‘Parse the Sentence First: Curbing the Urge to Resort to the Dictionary When Interpreting Legal Texts’ (2002–03) 6 New York U J Legislation and Public Policy 401, 402 (proposing that judges ‘rely on their linguistic intuitions’ and ‘parse the statutory sentence as a first step in [their] linguistic analysis’). 308 BG Slocum ‘Overlooked Temporal Issues in Statutory Interpretation’ (2008) 81 Temple L Rev 635 (criticizing courts for not considering temporal issues when interpreting statutes). 309 D Wu ‘Patent Litigation: What About Qualification Standards for Court Appointed Experts’ (2010) Boston College Intellectual Property and Technology Forum 91501, *22. 310 Cf. CD Cunningham and CJ Fillmore ‘Using Common Sense: A Linguistic Perspective on Judicial Interpretations of “Use a Firearm”’ (1995) 73 Washington U L Q 1159 (illustrating how a linguist would approach an interpretive issue). 311 See above n 272–81 and accompanying text (describing pragmatics). 312 Bibok, above n 248 at 291 (claiming that two-level semantics ‘could direct the attention of GL to various ways of being underspecified and the more comprehensive handling of some polysemic words’). 313 Falkum, above n 120 at 225. Blutner, too, favors a theory of radical semantic underspecification in the lexicon with a theory of pragmatic strengthening that is based on conversational implicatures. Blutner, above n 259. 314 Geeraerts, above n 21 at 147. 315 Falkum, above n 120 at 223. 316 Ibid at 224 (arguing that a ‘wholly pragmatic account is to be preferred …because it would give a simpler, more unified account overall (with a much leaner semantics)’). 317 Ibid [explaining that ‘[i]n order to recover the complete (truth-evaluable) proposition the speaker has expressed from the linguistically decoded semantic representation, it must be inferentially enriched (by disambiguation, assignment of referents, free enrichment) in a context of available assumptions’]. 318 Note that even legislative history, which is itself controversial as a tool of statutory interpretation, is merely another text, or texts, to be considered by the court. 319 I Witczak-Plisiecka ‘Relevance-Theoretic Perspective on Legal Language’ in CP-L Auría et al. (eds) Actas del Congreso Internacional AELFE. Proceedings of the 5th International AEL Conference: Academic and Professional Communication in the 21stCentury: Genres,Rhetoric and theConstruction ofDisciplinaryKnowledge (Zaragoza: Prensas Universitarias de Zaragoza 2006) 183; see also Falkum, above n 120 at 227 (conceding that ‘there is not much doubt that Pustejovsky is right in assuming that certain interpretations come more readily to mind than others in a “null” context’). 320 Witczak-Plisiecka, above n 319 at 183. 321 See above n 4–18 and accompanying text. 322 See above n 279–81 and accompanying text. 323 See above n 276–78 and accompanying text. 324 Other scholars have considered the relationship between pragmatics and the judiciary’s interpretive rules. GP Miller ‘Pragmatics and the Maxims of Interpretation’ [1990] Wisconsin L Rev 1179; MBW Sinclair ‘Law and Language: The Role of Pragmatics in Statutory Interpretation’ (1985) 46 U Pittsburgh L Rev 373. 325 Slocum, above n 16 at 811–12. 326 WN Eskridge, Jr, et al.Cases and Materials on Legislation: Statutes and the Creation of Public Policy (3rd edn Thomson West 2001) 818 (listing many of the substantive canons). 327 Slocum, above n 16 at 816–17. 328 WN Eskridge, Jr, and PP Frickey Cases and Materials on Legislation: Statutes and the Creation of Public Policy (2nd edn Thomson West 1995) 634. The key assumption underlying these canons is that when Congress enacts statutory language it does so according to accepted precepts of grammar and logic. Ibid at 329. 329 For example, Lindsey Coal Mining Co v. Chater 90 F 3d 688, 692 (3rd Cir 1996) (indicating that surplusage ‘cannot have been Congress’s intent’). 330 For example, Carlson v. Reed 249 F 3d 876, 882 (9th Cir 2001) (‘Applying the canon expressio unius est exclusio alterius, Congress’s failure to expressly require R nonimmigrants to maintain a foreign residence must have been deliberate.’). 331 For example, Ingersoll v. Magone 53 F 1008, 1010 (2nd Cir 1893) (indicating that the ‘principle of noscitur a sociis, show[s] the intent of congress’). 332 A Vermeule Judging under Uncertainty (Harvard University Press 2006) 118–37 (discussing the limitations of proposed methodologies that require universal adoption by the judiciary for their benefits to be realized). 333 See above n 2–12 and accompanying text (describing the judiciary’s current approach to ordinary meaning determinations). 334 This position does not deny, however, that courts may, for reasons other than an accurate determination of the ordinary meaning of the text, choose narrower interpretations than would be called for by an application of GL. 335 Klepousniotou et al., above n 84 at 1534 (explaining that almost any word can become polysemous and have its core meaning extended); Willems, above n 10 at 583 (explaining that words are inherently polysemous); see also Murphy and Koskela, above n 8 at 122 (explaining that polysemy is distinguished from homonymy, ‘which involves distinct semantically unrelated lexemes that coincidentally share the same form’); Pustejovsky, above n 132 at 291 [noting that ‘language is the instrument it is since the same expression can serve countless purposes because the semantic features (meanings) change in context’]. Courts may claim that they aim to capture the original intent of the enacting Congress, but, in reality, often commit temporal errors by allowing, in various ways, the meaning of statutes to change over time. Slocum, above n 308. 336 DL Shapiro ‘Continuity and Change in Statutory Interpretation’ (1992) 67 New York U L Rev 921 (arguing that courts are naturally conservative and seek legal coherence by minimizing disruption of the status quo). © The Author 2012. Published by Oxford University Press. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com.
TI - Linguistics and ‘Ordinary Meaning’ Determinations
JF - Statute Law Review
DO - 10.1093/slr/hmr025
DA - 2012-02-01
UR - https://www.deepdyve.com/lp/oxford-university-press/linguistics-and-ordinary-meaning-determinations-J3MNTWrH0E
SP - 39
EP - 83
VL - 33
IS - 1
DP - DeepDyve
ER -