Optionality as Demarking in an L2 Advanced State

Optionality as ‘demarking’ in an L2 advanced state 1. Introduction IL optionality is a pervasive phenomenon from an L2 initial to steady state. Recently, optional suppliance of morphological markings in spontaneous speech by L2 learners of English has been under vigorous investigation in an attempt to account for an apparent discrepancy between highly variable suppliance of affixes which may suggest the absence of relevant functional categories and/or features in IL grammar, on the one hand, and TL-like use of syntactic correlates, implicating the presence of such functional categories and features, on the other (Goad and White 2003; Goad, White, and Steele 2003; Lardière 2000, inter alia). IL optionality, however, is not restricted to spontaneous oral performance, but is also evident in other modes of IL performance. One such case involves persistent and variable placement of an adverb between the verb and the logical object as observed in writing and grammaticality judgment performances by francophone speakers of English (i.e., *SVAO) (White 1990/1991, inter alia). More recently, Robertson and Sorace (1999) investigated German-speaking learners’ V2-constraints phenomena in L2 English writing composition and metalinguistic tasks. An investigation of IL optionality in this domain enables us to focus on the factors involved in the lexicon and syntax components of IL grammar and examine how lexical selections in generating sentences effect IL optionality. This study reports a new IL optionality phenomenon in the domain of metalinguistic performance and considers how, as well as why, it occurs, drawing on grammaticality judgment data on Japanese passives by English- and Chinese-speaking learners. Despite overall acquisition patterns for Japanese passives consistent with the typological similarities and differences in the passive types between Japanese, English and Chinese, both English- and Chinese-speaking highly advanced learners converged in optionally accepting one type of ungrammatical ni indirect passive sentence as grammatical while consistently rating its grammatical counterpart as grammatical. It is argued that this optionality behavior results from learners’ attempt to assign a simpler structure to this type of ni indirect passive sentence when a related but distinct structure for it is available. This attempt, it is argued, is driven by Economy of Representations. 2. Passives in Japanese, English, and Chinese Following Hoshi (1994) and Ting (1995), Japanese, English, and Chinese passives are classified syntactically and semantically as shown in Table 1. Passivization in the table refers to the syntactic passivization operations commonly termed accusative Case absorption and external theta-role suppression. Affectivity signifies that a top-most NP in the passive construction carries a semantic role of experiencer (or affectee) who has been affected by an action or event described by the passive sentence. Since this paper is concerned only with L2 learners’ syntactic knowledge of passives, we focus on passivization in the table. Japanese has three types of passives: the ni direct and the ni yotte passives involve syntactic passivization, whereas the ni indirect does not. English has two types of passives: both the be and the get passives trigger syntactic passivization, but there is no English counterpart for the Japanese ni indirect passive. Chinese also has two types of passives: the bei short passive corresponds to the ni direct and the ni yotte passives in syntactic terms, while the bei long passive resembles the ni indirect passive. Table 1. Classifications of Japanese, English, and Chinese passives Passive types Passivization Affectivity Japanese Ni direct + + Ni indirect – + Ni yotte + – English Be + – Get + + + + Chinese Bei short Bei long – + Example sentences of these passives are given below: (1) Japanese passives: a. Ni direct passive: John-ga sensee-ni sikar-are-ta. John-NOM teacher-by scold-PASS-PAST ‘John was affected by being scolded by the teacher.’ b. Ni indirect passive: Ryoosin-ga sensee-ni John-o sikar-are-ta. (John’s) parents-NOM teacher-OBL John-ACC scold-PASS-PAST ‘John’s parents were affected by the teacher scolding him.’ c. Ni yotte passive: John-ga sensee-ni yotte John-NOM teacher-by sikar-are-ta. scold-PASS-PAST ‘John was scolded by the teacher.’ (2) English be/get passives: John was/got scolded by the teacher. (3) Chinese passives: a. Bei short passive: Zhangsan bei Zhangsan PASS ‘Zhangsan was hit.’ da-le. hit-ASP b. Bei long passive: Zhangsan bei Lisi da-le pigu. Zhangsan PASS Lisi hit-ASP (Zhangsan’s) buttocks ‘Zhangsan was affected by Lisi hitting his buttocks.’ As exemplified in (1a) and (1c), in the ni direct and the ni yotte passives which trigger syntactic passivization, the logical subject of the passivized verb, sensee ‘teacher’, is demoted to an adverbial phrase marked by ni and ni yotte, respectively. The logical object of the passivized verb John, on the other hand, has been moved to a syntactic subject position.1 Notice that the syntactic passivization just outlined fully corresponds to that of the English be and get passives (see (2)). The Chinese bei short passive (3a) also involves the same operations: the logical object of the passivized verb, Zhangsan, has been moved to a pre-verbal syntactic subject position, while the agent of the passivized verb is completely suppressed. Turning to the ni indirect passive (1b) which does not involve syntactic passivization, the logical object of the lower verb, John, is marked by the accusative Case marker -o, indicative of the absence of accusative Case absorption. Although the agent of the lower verb, sensee ‘teacher’, is marked by the oblique Case marker -ni, it retains a syntactic subject status in the lower clause, suggesting the lack of external theta-role suppression. Similarly, in the Chinese bei long passive (3b), the internal argument of the lower verb, pigu ‘buttocks’, appears post-verbally in a syntactic object position, while the external argument of the lower verb, Lisi, is placed pre-verbally in a syntactic subject position. In short, the English be and get passives correspond to the Japanese ni direct and ni yotte passives in syntactic terms, whereas there is no English counterpart for the Japanese ni indirect passive. The Chinese bei short passive corresponds to the Japanese ni direct and ni yotte passives, and the bei long passive is similar to the Japanese ni indirect passive2 in terms of the presence or absence of syntactic passivization. 3. The study 3.1. Hypotheses The Full Transfer Full Access Hypothesis (Schwartz and Sprouse 1996, inter alia) predicts that English-speaking learners will have considerable difficulty acquiring the ni indirect passive in syntactic terms due to the absence of this passive type in English, while they will not have difficulty with the ni direct and the ni yotte passives. Chinesespeaking learners, on the other hand, will acquire the three types of Japanese passives with equal difficulty (or ease) since Chinese passives typologically match those of Japanese. (Hawkins's (2003) Representational Deficit Hypothesis appears to make the same predictions for the general acquisition patterns for Japanese passives by these L1 learners.) The Full Access Hypothesis (Epstein, Flynn, and Martohardjono 1996, inter alia), by contrast, predicts that there will be no difference in the ways in which Japanese passives are acquired regardless of learners’ L1s. 3.2. Subjects The subjects of the study included 81 English-speaking and 85 Chinese-speaking learners of Japanese as well as 31 NS controls. The NNS subjects were assigned to the three proficiency levels of highly advanced, advanced, and intermediate on the basis of 1 The derivation for the ni direct passive is simplified for expository reasons (see Hoshi 1994 for a fuller account). This simplification does not affect the argument in this paper. 2 The range of nouns that appear in the internal argument position of the lower verb in the bei long passive seems more restricted than that in the ni indirect passive (see Shi 1997 and references cited therein). their scores on part of the Japanese Language Proficiency Test (JLPT) they took during the data collection procedure. There was a highly significant difference in the group mean JLPT scores (F(5, 160) = 219.954, p < .0005): the two (English and Chinese) groups within each proficiency level were not different from each other, but differed from the other groups of different proficiency levels. Table 2 summarizes NNS subjects’ background information on Japanese-learning experiences. Notice that their JLPT scores generally correlate with their length of study and visiting experience. Table 2. NNS subjects’ background information on learning Japanese Proficiency No. JLPT scores Length of Visiting experience levels % study (length) Highly advanced 25 100-88 9.74 yrs 92% (4.2 yrs) Advanced 36 84-64 4.16 yrs 75% (20 mons) Intermediate 20 60-32 4.32 yrs 65% (9.9 wks) L1 languages English † Highly advanced 36 100-87.5 5.47 yrs † Advanced 43 83.3-62.5 2.66 yrs † Intermediate 6 58.3-45.8 1.75 yrs † The Chinese subjects had resided in Japan for minimally three to four months at the time of data collection. Due to misunderstanding of a relevant questionnaire item by respondents, exact information was unavailable. Chinese 3.2 Materials The subjects were asked to assess the grammaticality of 84 test sentences (including 22 distractors) using a five-point scale: acceptable, somewhat acceptable, unsure, somewhat unacceptable, and unacceptable. There were 16 (grammatical and ungrammatical) test items in the grammaticality judgment questionnaire that were specifically devised to probe NNS subjects’ syntactic knowledge of Japanese passives by manipulating Case-marking assignments resulting from the presence or absence of syntactic passivization. For each sentence type, two tokens were contrived. A critical triplet of grammatical and ungrammatical ni indirect passive sentences is given below: (4) a. Bill-wa Jane-ni sono himitu-o Bill-TOP Jane-OBL the secret-ACC ‘Bill was affected by Jane knowing the secret.’ sir-are-ta. know-PASS-PAST b. * Bill-wa Bill-TOP sono himitu-ga Jane-ni sir-are-ta. the secret-NOM Jane-by know-PASS-PAST * Bill-wa Bill-TOP Jane-ga sono himitu-o Jane-NOM the secret-ACC c. sir-are-ta. know-PASS-PAST The sentence (4a) represents a grammatical ni indirect passive sentence, while the sentences (4b) and (4c) are ungrammatical. In (4b), syntactic passivization has been applied incorrectly. Thus, the internal argument of the lower verb, sono himitu ‘the secret’, is nominative ga-marked due to NP movement, while the experiencer of the lower verb, Jane, is demoted to the adverbial ni phrase. In (4c), by contrast, syntactic passivization has not been applied correctly as indicated by the theme of the lower verb, sono himitu ‘the secret’, being accusative Case (-o)-marked as well as by the external argument of the lower verb, Jane, being nominative Case (-ga)-marked. This nominative Case -ga marking in the lower clause, however, is in direct conflict with that of the topmost NP John (which is further topicalized in the example), rendering it as ungrammatical. 4. Results General acquisition patterns for the three types of Japanese passives are first reported followed by a closer examination of grammaticality judgments of the ni indirect passive. Figure 1 displays the grammaticality judgments of the three types of grammatical Japanese passive sentences (the ni direct, the ni indirect, and the ni yotte passive) by the English-speaking groups (highly advanced, advanced, and intermediate) and the NS con2.0 1.5 1.0 .5 0.0 -.5 NS control -1.0 High adv. -1.5 Adv. -2.0 Interm. Ni direct Ni indirect Ni yotte Figure 1. Grammaticality judgments of Japanese passives by English-speaking learners trol group. Both highly advanced and advanced groups rated the ni direct passive significantly better than the other two types of passives (F(2, 156) = 15.568, p < .0005; Within-Subjects Contrast, p < .0005). There was, however, no significant difference in grammaticality assessment between the ni indirect and the ni yotte passives for these two groups (Within-Subjects Contrast, p = .399).3 The intermediate group did not rate any one type of passive significantly higher than the other two (F(2, 38) = 2.023, p = .146). 3 A closer inspection of the grammaticality judgments of both grammatical and ungrammatical sentence types of the ni indirect and the ni yotte passives by the English-speaking highly advanced and advanced groups revealed that the ni yotte passive was rated more accurately than the ni indirect passive by the highly advanced group (see Hara 2002 for these results as well as discussion of why the ni yotte passive was rated worse than the ni direct passive despite its exact match with the English be passive). Figure 2 presents the grammaticality judgments of the three types of grammatical passive sentences by the Chinese-speaking groups and the NS control group. Unlike the 2.0 1.5 1.0 .5 0.0 -.5 NS control -1.0 High adv. -1.5 Adv. -2.0 Interm. Ni direct Ni indirect Ni yotte Figure 2. Grammaticality judgments of Japanese passives by Chinese-speaking learners English-speaking groups, neither the highly advanced nor the advanced Chinese-speaking group rated any single type of passive significantly higher than the other two passives (F(2, 164) = .836, p = .435). The intermediate Chinese-speaking group did not make any differential judgments between the three passives, nor did its English-speaking counterpart. Next, since the ni indirect passive appears to have posed the greatest learning difficulty, at least to the English-speaking groups, the grammaticality judgments of both grammatical and ungrammatical ni indirect passive sentences are examined more closely. Figure 3 reports all of the subject groups’ grammaticality assessments of the triplet of ni indirect passive sentences as illustrated in (4): Grammatical (a) and Ungrammatical (c)/(b) in the figure correspond to the sentence types of (4a) and (4c)/(4b), respectively. Since the advanced groups of both L1s did not rate the grammatical sentences very high (not to mention the performance of intermediate groups), we focus only on the highly advanced groups. On the one hand, both highly advanced groups made a reliable distinction in grammaticality assessment between the grammatical (a)-type sentences and the ungrammatical (c)-type sentences (F(2, 48) = 30.983, p < .0005 for the English group and F(2, 70) = 42.476, p < .0005 for the Chinese group; Within-Subjects Contrast, p < .0005 for both groups), suggesting that they accepted the former ((a)-type sentences) as grammatical and rejected the latter ((c)-type) as ungrammatical. The ungrammatical (b)type sentences, on the other hand, were rated significantly differently from both the grammatical (a)-type and the ungrammatical (c)-type sentences by both groups (WithinSubjects Contrasts, p ≤ .011 for the English group and p ≤ .029 for the Chinese group). Their mean judgment scores were, indeed, close to zero. Since, in syntactic terms, this judgment pattern is the only one in which the highly advanced groups clearly diverged from the NS control group (see Hara 2002), it requires 2 1 NS control 0 Eng high adv. Eng adv. Eng interm. -1 Chi high adv. Chi adv. -2 Chi interm. Grammatical (a) Ungrammatical (c) Ungrammatical (b) Figure 3. Grammaticality judgments of the ni indirect passive by all groups closer examination. Table 3 presents the percentage of subjects per group who rated the two ungrammatical (b)-type sentences as either both grammatical, or as one grammatical and the other ungrammatical, or both ungrammatical. In the table, a predicted response (i.e., ungrammatical) is marked by an X, while a response inconsistent with the prediction is signified by a hyphen (–). In order to exclude from this analysis NNS subjects whose syntactic knowledge of the ni indirect passive was not firm, this tabulation included only those who consistently judged the grammatical ni indirect passive sentences as grammatical. Table 3. Percentage of subjects per group in judgment of the ungrammatical (b)-type sentences † ── XX Groups X─ English 33.3% 33.3% 33.3% † (5) (5) (5) highly adv. (15) Chinese 26.3% 47.4% 26.3% highly adv. (19) (5) (9) (5) NS control (30) 3.3% 3.3% 93.3% (1) (1) (28) † Figures in the parentheses represent the numbers of subjects in each group. Out of those highly advanced subjects who rated the grammatical (a)-type sentences consistently as grammatical, two thirds or more of them (66.6% and 73.7% of the English and the Chinese groups respectively) judged the ungrammatical (b)-type sentences as grammatical either on both occasions or once. Hence, these highly advanced learners, while consistently rating the grammatical ni indirect passive as grammatical, at least optionally accepted the ungrammatical (b)-type sentences as grammatical. A third or less of them (33.3% and 26.3% of the English- and the Chinese-speaking groups respectively), on the other hand, consistently judged the ungrammatical (b)-type sentences as ungrammatical.4 This judgment variability is in sharp contrast with that of the NS control group, 93.3% of which rejected the ungrammatical (b)-type sentences consistently as ungrammatical. In sum, the analysis of individual grammaticality judgments of the ni indirect passive sentences reveals that highly advanced English- and Chinese-speaking learners at least optionally accepted the ungrammatical (b)-type sentences as grammatical, while judging the grammatical sentences consistently as grammatical. 5. Discussion The findings for the general acquisition patterns of Japanese passives by English- and Chinese-speaking learners suggest clear L1 effects. English-speaking learners acquired the ni direct passive, which matches in syntactic terms the English get passive, best of all, while they had considerable difficulty with the ni indirect passive for which an English counterpart does not exist. Chinese-speaking learners, on the other hand, learned the three types of passives similarly well (or poorly), presumably because of the close correspondence between Chinese and Japanese passives. These findings are consistent with The Full Transfer Full Access Hypothesis (Schwartz and Sprouse 1996). Contrary to L1 effects as reflected in these general acquisition patterns, the convergence in optional grammaticality judgments of the ni indirect passive sentences by both English- and Chinese-speaking highly advanced learners points to a property of SLA that is universal to both L1 groups. Elaborating on Sorace's (2000) notion of ‘demarking,’ this paper proposes that demarking as an instantiation of Economy of Representations (Cardinaletti and Starke 1994; Chomsky 1995) underlies the observed IL optionality behavior. 5.1. Demarking Sorace (2000) makes reference to demarking in accounting for Italian and English bilingual speakers’ optional use of an overt subject pronoun instead of a null subject pronoun where the latter is appropriate in a monolingual grammar of Italian (but not vice versa). For instance, English-speaking near-native speakers of Italian (Italian as L2) and Italian-speaking near-native speakers of English (Italian as L1) variably utter: Lei ha deciso di fare una passeggiata (She decided to go for a walk) where pro ha deciso di fare una passeggiata (Ø decided to go for a walk) is appropriate (Sorace 2000: 719, (1b) and (1c)). In contrast to this optional overuse of overt subject pronouns, these same speakers do not overgeneralize null subject pronouns where overt subject pronouns are appropriate. Thus, in response to the question Perchè Maria è uscita? (Why did Maria 4 Highly similar individual results obtain even if we include those subjects who accepted the grammatical sentences of the ni indirect passive correctly only once as well as those who did so both times (see Hara and Ma under review). leave?), they do not utter: * Perchè pro è venuto a prenderla (Because Ø came to pick her up); but they utter: Perchè Gianni è venuto a prenderla (Because Gianni came to pick her up) (Sorace 2000: 720, (3a) and (3b)). Following Cardinaletti and Starke (1994), Sorace takes an overt pronoun to be an unmarked option and a null pronoun a marked one. Since demarking is defined as the phenomenon in which a marked linguistic option is destabilized, and an unmarked option is used where the former is appropriate (Sorace 2000: 724), the unidirectionality of overgeneralization of overt subject pronouns (but not that of null subject pronouns) is captured. The marked option of null subject pronouns is destabilized, and the unmarked option of overt subject pronouns is overgeneralized where the former should be used.5 This paper proposes that demarking is an instantiation of a general linguistic principle, that of Economy of Representations (Cardinaletti and Starke 1994) which states that “a smaller structure is obligatorily chosen, if possible” (p. 89). That is, to the extent that unmarked linguistic options are characterized as linguistic forms with a smaller structure in comparison to their marked counterparts, demarking can be conceived of as a specific instance of Economy of Representations. Thus, informally stated, a linguistic form that is unmarked and smaller in structure may be chosen, if possible, over the one that is marked and more complex in structure. 5.2. A Minimalist analysis of Japanese passives Hoshi's (1994) analysis of Japanese passives adopted in Table 1 is basically in the GB framework. More recently, Watanabe (1996) recasts essential syntactic properties of passives in the Minimalist framework, and his analysis enables us to determine their markedness statuses in structural terms. Watanabe proposes the following categorial structures of the ni direct and the ni indirect passives (slightly modified for expository purposes)6; rare in (5a) and (5b) are passive verbs of the ni direct and the ni indirect passives: (5) a. Ni direct passive: [VP [AgrP VP Agr] rare] b. Ni indirect passive: [VP Experiencer [AgrP VP Agr] rare] [ACC] In Watanabe’s theory, the ni direct and the ni indirect passives are derived exactly in the same way by means of the general Case checking mechanism that he proposes, and the syntactic differences between the two types of passives result merely from which categorial structure their respective passive verb rare takes. 5 Indeed, there is a problem with Sorace's (2000) application of demarking in this case. It appears that the property destabilized is that of the unmarked option, i.e., the [+Topic Shift] feature of overt pronouns (see Hara and Ma under review, for discussion). 6 Watanabe (1996) discusses the ni yotte passive, not the ni direct passive: I assume, as he implies, that the analysis of the former carries over to the latter. Specifically, in Watanabe's (1996) Agr-based Case checking theory7, the Agr head serves as a garbage can to clean up a Case feature from a Case feature-bearing element like DP. Importantly, to complete the Case checking process, there needs to be an additional follow-up Case checker, i.e., another functional head immediately above the AgrP, for the Case feature of a Case-bearing head such as V0 to be cleaned up. Without such a functional head right above the AgrP, Case absorption ensues. Notice that in both (5a) and (5b), there is no functional head that serves as a follow-up Case checker right above the AgrP. Thus, this structural configuration results in accusative Case absorption on the lower verb in both the ni direct and the ni indirect passives. Subsequently, in (5a) of the ni direct passive, the [NOM] Case feature that the internal argument of the lower passivized verb bears is checked as it moves to the spec of the Agr-sP by the standard nominative Case checking process. In (5b) of the ni indirect passive, in contrast, the [ACC] Case feature that the internal argument of the lower verb carries is checked under the spec-head agreement when this [ACC]-bearing DP moves to the spec of the Agr-oP, and as the passive verb rare that also bears the [ACC] Case feature adjoins to the Agr-o head. This (ordinary) accusative Case-checking is made possible by the property of the ni indirect passive verb rare bearing the accusative Case feature unlike the ni direct passive verb (see (5a) and (5b)). Therefore, accusative Case absorption takes place in the lower clause of both the ni direct and the ni indirect passive. This is reflected in the identical categorial layers that both passive verbs select, i.e., the AgrP and the VP. Subsequently, accusative Casechecking takes place in the higher clause of the ni indirect passive due to the presence of the [ACC] Case feature its passive verb rare bears. This results in the disguised property of the ni indirect passive not involving accusative Case absorption. This difference in the passive verb properties between the ni direct and the ni indirect passives is reflected in the different structures that each of the passive verbs requires above the AgrP in (5), i.e., only its canonical functional layers for the ni direct passive, while the experiencer DP as well as the [ACC] Case feature for the ni indirect passive in addition to its canonical functional layers. 5.3 Optional judgments of the ni indirect passive as demarking As sketched in section 5.2, the ni direct passive verb selects a smaller structure than the ni indirect passive verb. This difference in the properties of the passive verbs provides a structural characterization for a commonly held view that in a rather unspecified sense, the ni direct passive is an unmarked type of passive, and the ni indirect passive marked (Watanabe 1996: 140). The ni direct passive is unmarked because its verb rare selects a smaller structure of categorial layers, while the ni indirect passive is marked since its verb rare selects a greater structure of categorial layers, implicating a more complex derivation. This enables us to establish a direct link between demarking and Economy of Representations. Under these conceptualizations, demarking in the case of Japanese passives is stated as follows: The ni direct passive verb, the unmarked option, is smaller 7 Watanabe (1996) maintains that his Agr-based Case theory can carry over to Chomsky's (1995) more recent Agr-less theory (see the Appendix of Chapter one in Watanabe 1996). In this paper Watanabe’s original proposal is followed. in structure than the ni indirect passive verb, the marked one. The former may be chosen, if possible, where the latter is appropriate. We are now in a position to interpret English- and Chinese-speaking highly advanced learners’ optional acceptance of the ungrammatical ni indirect passive sentences like (6) (repeated from (4b)) as grammatical, while they consistently rated the grammatical version as grammatical: (6) * Bill-wa sono himitu-ga Jane-ni Bill-TOP the secret-NOM Jane-by ‘Bill was affected by Jane knowing the secret.’ sir-are-ta. know-PASS-PAST I propose that sentence (6), when accepted as grammatical, was analyzed by these highly advanced learners as follows (irrelevant details of the structure are eliminated for ease of exposition): (7) [TopicP John [IP kuruma no madoi [VP [AgrP [VP kodomo-ni ti tj] Agr] warj-are]-ta]] By treating an initial NP John as a topic (which is base-generated in sentence-initial position; see Miyagawa 1989), these highly advanced learners analyzed the remaining part of (7) as an instance of the ni direct passive. Notice that the structure below the IP of (7) is identical with the structural representation of the ni direct passive as illustrated in (5a). Why did these learners analyze the ni indirect passive sentence (6) as a ni direct passive? I contend that demarking drove this analysis: when possible, they selected the ni direct passive verb, an unmarked, simpler one, where the ni indirect passive verb might have been used. I assert that adoption of this analysis was further encouraged by the fact that sentence (6) is ungrammatical as a ni indirect passive. Thus, instead of marking it as ungrammatical, they attempted to construct a grammatical representation of it. This view is supported by their grammaticality judgment patterns for the other two types of ni indirect passive sentences (grammatical (4a) and ungrammatical (4c)). They clearly accepted the former as grammatical and consistently rejected the latter as ungrammatical, because these ni indirect passive sentences could not be analyzed as a grammatical ni direct passive; hence, an alternative analysis of them as a ni direct passive was not available, or at least unmotivated. 6. Conclusion This paper has reported new evidence of optionality in reflected IL behavior, not in spontaneous production. It has been argued that this optional behavior results from demarking, which, in turn, is taken as an instantiation of Economy of Representations. With related but distinct lexical items available, one of which is smaller in structure than the other, the less complex item is optionally chosen if possible. Thus, the locus of the observed optionality is reduced to selection of lexical items in the IL lexicon.