Investigating Tagalog Relative Clause Acquisition, Lecture notes of English

Download Investigating Tagalog Relative Clause Acquisition and more Lecture notes English in PDF only on Docsity! W O R K I N G P A P E R S I N L I N G U I S T I C S The notes and articles in this series are progress reports on work being carried on by students and faculty in the Department. Because these papers are not finished products, readers are asked not to cite from them without noting their preliminary nature. The authors welcome any comments and suggestions that readers might offer. Volume 46(1) February 2015 DEPARTMENT OF LINGUISTICS UNIVERSITY OF HAWAI‘I AT MĀNOA HONOLULU 96822 An Equal Opportunity/Affirmative Action Institution University of Hawai‘i at Mānoa Working Papers in Linguistics 46(1) ii DEPARTMENT OF LINGUISTICS FACULTY 2015 Victoria B. Anderson Andrea Berez Derek Bickerton (Emeritus) Robert A. Blust Lyle Campbell Kenneth W. Cook (Adjunct) Kamil Deen (Graduate Chair) Patricia J. Donegan (Chair) Katie K. Drager Emanuel J. Drechsel (Adjunct) Michael L. Forman (Emeritus) Roderick A. Jacobs (Emeritus) William O’Grady Yuko Otsuka Ann Marie Peters (Emeritus) Kenneth L. Rehg (Adjunct) Lawrence A. Reid (Emeritus) Amy J. Schafer Albert J. Schütz, (Emeritus, Editor) Jacob Terrell James Woodward Jr. (Adjunct) Nozomi Tanaka: Tagalog Relative Clause Production: Data from Adults and Children 3 (7) yabu ŋuma-ŋgu buṛa-n mother.ABS father-ERG see-NFUT ‘Father saw mother.’ (Dixon 1979:61) The studies that regard Tagalog as a nominative-accusative language see aŋ as a nominative marker that marks the syntactically prominent argument (Kroeger 1993), and naŋ as either an oblique or a genitive marker.3 In this analysis, the agent-focus pattern, as shown in (8), is considered to be an active transitive, in which the agent subject bears the nominative case, while the theme-focus pattern, as shown in (9), is considered non-active, with the theme object bearing the nominative case. (8) B<um>ili aŋ babae naŋ isda. <TR.PFV>buy NOM woman GEN fish ‘The woman bought a fish.’ (Aldridge 2004a:59) (9) B<in>ili naŋ babae aŋ isda. <PASS.PFV>buy GEN woman NOM fish ‘The woman bought the fish.’ (Aldridge 2004a:59) The opposing view is the ergative-absolutive approach to Tagalog (Payne 1982; Cooreman et al. 1984; De Guzman 1988; Aldridge 2004a). In this analysis, the theme-focus sentences are considered canonical transitive patterns, where aŋ marks the absolutive object and naŋ marks the ergative subject, as shown in (10). (10) B<in>ili naŋ babae aŋ isda. <TR.PFV>buy ERG woman ABS fish ‘The woman bought the fish.’ (Aldridge 2004a:101) The agent-focus sentences are considered intransitives/antipassives,4 where aŋ marks the absolutive and naŋ marks the oblique, as demonstrated in (11). (11) B<um>ili aŋ babae naŋ isda. <INTR.PFV>buy ABS woman OBL fish ‘The woman bought a fish.’ (Aldridge 2004a:100) Two arguments have been presented for the ergative-absolutive analysis of Tagalog: text frequency and the order of acquisition. Cooreman and colleagues (1984) reported that, in written Tagalog, 24 percent of sentences occurred in the AF pattern, while 76 percent occurred in the TF pattern. McFarland (1984) reported a similar distribution, where 38 percent of the total verbal uses are in the AF form. These studies suggest that the basic pattern is the TF pattern, with the AF being a derived, antipassive form. However, Shibatani (1988) and Maclachlan (1996) both claimed that the disparity between AF and TF is not as great as between active and passive forms in nominative-accusative languages (e.g., English), and that sheer frequency cannot be used to decide whether AF or TF is the basic pattern. As for acquisition, a few studies have suggested that TF is acquired earlier than AF, supporting the ergative analysis. Tucker (1971) tested children’s mastery of agent focus and theme focus among other focus patterns, using two tasks: an imitation task and a sentence-completion task. In the imitation task, children were asked to repeat the sentences read by the researchers. In the sentence-completion task, they were given a sentence in a certain focus pattern and asked to turn the sentence into another focus pattern 3 This is because naŋ is also used for genitive possessors, e.g., libro naŋ babae ‘the girlʼs book.’ 4 Antipassives are seen among ergative languages. The agent is realized as the absolutive—the syntactically prominent argument—and the theme as the oblique (Dixon 1979). Because antipassive clauses are morphologically intransitive, it is possible to think of the absolutive agent as the intransitive subject (Schachter 1994). University of Hawai‘i at Mānoa Working Papers in Linguistics 46(1) 4 by supplying the verb in a correct form, following a noun phrase given by the researchers.5 The results from both tasks showed that theme focus was the easiest of all the focus patterns for the children. Segalowitz and Galang (1978) also conducted a sentence-completion task, but in a slightly different design. In this task, children were asked to describe pictures, instead of to transform a sentence from one structure to another. Their results showed that children did equally well in both AF and TF. They also conducted a comprehension task, in which children were asked to select an appropriate picture based on the sentence they had heard. When the sentence was produced in an NP-initial word order, children did equally well in both AF and TF; however, when the sentence had a canonical, verb-initial word order, they performed better when the sentence was in a TF pattern. Although Segalowitz and Galang (1978) presented a slightly more complicated picture than Tucker (1971) did, both studies seem to show that children know TF better than AF, as their performance was not better with AF than TF in Segalowitz and Galang’s test. Because children typically acquire the basic form (e.g., active sentences in English) earlier than the derived form (e.g., passive sentences in English), these findings suggest that TF is the basic pattern in Tagalog. However, Maclachlan (1996) claimed that both AF and TF are acquired relatively early compared to passives in English and Inuktitut, suggesting that the previous acquisition data do not clearly show which form is more basic. The present study therefore looks at the acquisition of relativization in order to gain empirical insight into whether children treat Tagalog as a nominative-accusative language or an ergative-absolutive language. Section 4 explains relativization in Tagalog in relation to previous studies on acquisition of nominative-accusative languages and ergative-absolutive languages. As discussed in section 4.1, acquisition of relativization has been shown to be asymmetrical between subject-relative clauses and object-relative clauses. The majority of acquisition studies in nominative-accusative languages suggest a subject-relative advantage in acquisition; that is, children acquire subject-relative clauses earlier (e.g., Cho 1999; Hsu et al. 2009). However, studies on ergative-absolutive languages have found that the subject advantage may be weaker (Polinsky et al. 2012; Gagliardi et al. 2013) or even completely reversed (Gutierrez-Mangado 2011), showing a clearly different pattern from nominative-accusative languages. Therefore, it is of interest to inquire whether children acquiring Tagalog show the nominative- accusative pattern or the ergative-absolutive pattern. 4. RELATIVIZATION IN TAGALOG. Relative clauses in Tagalog begin with the “linker” -ŋ, a complementizer-like item, and follow the noun that they modify.6 (12) Relativization of the sole argument of an intransitive verb lalaki=ŋ [RC t<um>akbo] man=LK <AF.PFV>run ‘the man that ran’ An important characteristic of Tagalog relative clauses is that only focused (aŋ-marked) arguments can be relativized. This constraint is irrelevant for intransitive patterns such as (12), where the verb has just one argument. However, the focus requirement has very tangible consequences for verbs that take both an agent argument and a theme argument. (13) Agent-focus pattern a. Relativization of the agent argument (acceptable) lalaki=ŋ [RC k<um>ain naŋ isda] man=LK <AF.PFV>eat CASE fish ‘the man that ate fish’ 5 This requires children to produce a sentence in NP-initial word order, which is possible (as long as the first NP is in focus) but not canonical in Tagalog. 6 These examples show only head-initial relative clauses. Head-final and head-internal relative clauses also exist in Tagalog (Aldridge 2004b). Nozomi Tanaka: Tagalog Relative Clause Production: Data from Adults and Children 5 b. Relativization of the theme argument (unacceptable) *isda=ŋ [RC k<um>ain aŋ lalake] fish=LK <AF.PFV>eat CASE man ‘the fish that the man ate’ (14) Theme-focus pattern a. Relativization of the theme argument (acceptable) isda=ŋ [RC k<in>ain naŋ lalake] fish=LK <TF.PFV>eat CASE man ‘the fish that the man ate’ b. Relativization of the agent argument (unacceptable) *lalaki=ŋ [RC k<in>ain aŋ isda] man= LK <TF.PFV>eat CASE fish ‘the man that the fish ate’ As illustrated in (13–14), only the agent argument can be relativized in the agent-focus pattern and only the theme argument can be relativized in the theme-focus pattern. In this paper, the pattern in (13a) is referred to as an agent-focus relative clause (ARC) and that in (14a) as a theme-focus relative clause (TRC). The existence of such contrasts and constraints raises two questions, both for the study of syntax and for the study of language acquisition. First, do children acquiring Tagalog exhibit a preference for either RC type in their production? Second, if children do show a preference, does it fit the prediction made by one or the other analysis of Tagalog syntax? 4.1 RELATIVIZATION IN NOMINATIVE-ACCUSATIVE LANGUAGES. It is known that the case system of a language has some effects on relativization, in terms of typology, processing, and language acquisition. On one hand, a preference for subject-relative clauses (i.e., relativization of the subject) over object- relative clauses (i.e., relativization of the direct object argument) is well-attested in nominative-accusative languages, such as English (e.g., Diessel and Tomasello 2005), Dutch (e.g., Frazier 1987), Swedish (e.g., Håkansson and Hansson 2000), Greek (e.g., Stavrakaki 2001), Persian (e.g., Rahmany et al. 2011), Hungarian (e.g., MacWhinney and Pleh 1988), Hebrew (e.g., Friedmann et al. 2009), Mandarin Chinese (e.g., Hsu et al. 2009), Japanese (e.g., Ishizuka 2005), and Korean (e.g., Cho 1999). Consider the English subject-relative clause, exemplified in (15), which is comprehended and produced with higher accuracy, read faster, and acquired earlier than the object-relative clause, of which (16) is an example (Brown 1972; Diessel and Tomasello 2005; Gibson et al. 2005; among others).7 (15) the boy that __ hugged the girl (16) the boy that the girl hugged __ In Tagalog, the direct object of the AF construction cannot be directly extracted. The only way to have the direct object as the head of the relative clause is to use the TF pattern (the passive, in this analysis), in which the direct object in an active sentence is promoted to a grammatical subject—the syntactically prominent argument. Therefore, in a strict sense, Tagalog only has “subject-relative clauses” (Keenan and Comrie 1977). However, languages without such a restriction are also known to utilize the passive construction when the extraction of the direct object is difficult; some production studies show that, in order to avoid direct object-relative clauses, speakers sometimes use passive relative clauses. Nevertheless, they are still more likely to choose active subject-relative clauses over passive relative 7 Mak and colleagues (2002) as well as Kidd and colleagues (2007) argue that object-relative clauses become easier for both adults and children when the head is inanimate, suggesting that the difficulty may be due to an animacy effect. University of Hawai‘i at Mānoa Working Papers in Linguistics 46(1) 8 Each item was accompanied by an auditory prompt in Tagalog describing the event in the picture.10 For example, in the case of the ARC item in figure 1, participants heard, “A boy is hugging a girl. Another boy is hugging a monkey. Who has the arrow mark?” The targeted response was “the boy that is hugging the girl.” Similarly, the presentation of the TRC item in figure 2 was accompanied by the auditory prompt, “A boy is hugging a girl. A monkey is hugging another girl. Who has the arrow mark?” The targeted response in this case was “the girl that the boy is hugging.” Participants’ responses were audio-recorded and transcribed for later analysis. 5.3 RESULTS. The analysis focused on NP [V NP] patterns containing a focus-marked verb and a gap corresponding to the aŋ-marked nominal—the basic relative clause template in Tagalog. For the purposes of analysis, responses that omitted the linker -ŋ, which was frequently absent from children’s responses, were retained in the data. The excluded data (7% of the adults’ and 58% of the children’s total responses) included headless relative clauses,11 relative clauses without an NP inside, relative clauses with a bare verb, relative clauses containing an aŋ-marked nominal, multiple RCs, cleft sentences, and so on.12 These responses were excluded either because they were uninterpretable or were not relative clauses. The responses included in the analysis were coded for membership in three classes: targeted responses, head errors, and reversal errors. Targeted responses consisted of RCs with the appropriate verbal affixation (AF for ARC; TF for TRC) and case marking. I classified a response as a head error if the referent of the head was changed, along with its thematic role. In the following example of a head error, the targeted head is the girl (the theme) in (17a), but the response selects the boy (the agent) as head in (17b). (17) a. Targeted (TRC): “the girl [that the boy is hugging]” b. Response (ARC): “the boy [that is hugging the girl]” A response was treated as a reversal error if the correct head was selected, but its thematic role was changed from agent to theme or vice versa. In the following example, the relative clause is restructured in a way that changes the thematic role of the girl from theme to agent, as shown in 18b. (18) a. Targeted (TRC): “the girl [that the boy is hugging]” b. Response (ARC): “the girl [that is hugging the boy]” The 65 adult responses included in the analysis comprised 32 from the ARC condition and 33 from the TRC condition. As can be seen in table 2, adults did equally well in both conditions, with no significant difference between the two relative clause types.13 10 The auditory prompt had an argument focus that matched that of the targeted RC: agent focus when an ARC was targeted and theme focus when a TRC was targeted. 11 A headless relative clause refers to a grammatical relative clause that does not have an overt NP as a head. This corresponds to “the one that hugged the girl” (ARC) or “the one that the girl hugged” (TRC) in English. The adult data included one headless ARC and the child data, seven headless relative clauses (five ARC, two TRC). The inclusion of these relative clauses increases the adults’ rate of targeted responses for ARC from 93.75% (as seen in table 2) to 93.94% (31 out of 33) and the children’s rate for ARC from 80.00% (as shown in table 3) to 83.33% (25 out of 30). However, this does not influence the interpretation of the results. 12 The rate of exclusion and the kinds of sentences that are excluded are in line with what are reported as non- target responses in previous studies (e.g., Hsu et al. 2009; Kim 2013). Most of the excluded data used the focus marker aŋ inside of the relative clauses. Only children made such errors, and they did so in both ARC and TRC conditions. Despite the pervasiveness of this type of response, it is unclear why children make such errors; therefore, I chose to exclude such responses from the analysis. 13 Because of the exclusion of some data, the number of responses from each participant was different. Therefore, the mean of the means of each adult participant gives slightly different rates of targeted responses: 91.67% for the ARC condition (SD = 0.14) and 92.38% for the TRC condition (SD = 0.14). Nozomi Tanaka: Tagalog Relative Clause Production: Data from Adults and Children 9 TABLE 2. Results from adults (n = 7). Response Types Frequency Rate ARC Targeted 30/32 93.75% Head errors (TRC) 0/32 0.00% Reversal errors (TRC) 2/32 6.25% TRC Targeted 31/33 93.94% Head errors (ARC) 0/33 0% Reversal errors (ARC) 2/33 6.06% The children provided 25 responses from the ARC condition and 23 responses from the TRC condition for the analysis. As summarized in table 3, the rate of targeted responses14 was significantly higher in the ARC condition (β = 1.54 ± 0.62, p < .05).15 The most common mistake on TRC items involved head errors, which made up 47.83 percent of all responses in this condition. The second most common mistake involved reversals, which constituted 13.04 percent of all responses. Both error types produce ARC in place of TRC. This means that 58.33 percent of the responses in the TRC condition were realized as ARC. In contrast, the only common mistake on ARC items involved reversal errors (i.e., producing TRC in place of ARC), which made up 20.00 percent of all responses in this condition. This means that more TRCs were turned into ARCs than vice versa. TABLE 3. Results from children (n = 10). Response Types Frequency Rate ARC Targeted 20/25 80.00% Head errors (TRC) 0/25 0.00% Reversal errors (TRC) 5/25 20.00% TRC Targeted 9/23 39.13% Head errors (ARC) 11/23 47.83% Reversal errors (ARC) 3/23 13.04% The two participant groups did not show a significant difference in their responses to the agent relative clause items; however, they showed a significant difference in their responses to the theme relative clause items (β = 4.12 ± 1.41, p <. 01). This indicates that the children’s performance significantly differs from the adults’, but only in the TRC condition. These findings appear to point toward an advantage for ARCs. However, if we look at the results from individual participants, an interesting picture emerges. Based on the results shown in table 4, the individual children fell into two distinct groups. Group 1 failed to produce a single correct TRC, but nonetheless showed largely successful performance on ARC patterns. In contrast, group 2 produced far more targeted TRC responses than ARC responses. This suggests that not all the children show the ARC advantage. Overall, then, we see indications of a bimodal distribution, with one group of children doing better on ARCs and the other manifesting an advantage for TRCs, although a study with a larger sample size is needed to confirm this distribution and test its statistical significance. 14 As with the adults, the mean of the means for each child participant is slightly different: 70.00% for ARC (SD = 0.42), and 40.00% for TRC (SD = 0.52). 15 I report fixed effect coefficients, β, from mixed effects logistic regression, which includes participants and items as random effects. The estimates of the two random effects were close to zero, indicating that the probability of targeted responses does not depend on individual participants or items. University of Hawai‘i at Mānoa Working Papers in Linguistics 46(1) 10 TABLE 4. Results from individual children who produced targeted responses. ARC TRC Participants Targeted Errors Targeted Errors G ro up 1 1 5 0 0 5 2 3 0 0 2 3 3 0 0 2 4 3 0 0 1 5 2 0 0 4 6 1 1 0 0 G ro up 2 7 2 2 2 0 8 0 2 3 0 9 0 0 2 0 10 1 1 2 0 20 5 9 14 6. GENERAL DISCUSSION AND CONCLUSION. The aggregated results from the elicited production task show an overall higher rate of success for ARCs. However, some children performed well in the ARC condition while others performed better in the TRC condition, which indicates that at least some children treat Tagalog as an ergative-absolutive language. Such a bimodal distribution is not reported in any of the previous studies, and may be due to the morphosyntactic properties of Tagalog. As noted in sections 2 and 3, Tagalog syntax is unusual in its treatment of syntactic prominence. Regardless of which argument is selected for focus, including use as head of a relative clause, special verbal affixation is required. There is no morphologically unmarked form and no apparent default option. It is possible that this property of Tagalog opens the door to the variation in relative clause preferences observed in this study. As noted in sections 5.1 and 5.3, some of the participants did not provide useful data, and some collected data were excluded from the analysis. Some of the children did not complete the task because they were uncomfortable around strangers or did not engage in the task, despite the researchers’ attempts to build rapport. These issues might be avoided by testing older children. I would also like to point out some of the problems specific to this type of research, mainly in regard to (1) the experimental setting and (2) the participants’ language experience. First, although the ideal environment for experimental work is a carefully controlled environment—a quiet room with necessary equipment where participants can work alone without distraction—this creates a challenge for field psycholinguistics. This study’s participants were all familiar with the formal education setting and technology, and they were all tested at their daycare center or university, in a location where we could use a laptop. This is a far easier research environment than that of some other experimental work in field research settings (see Clemens et al. 2014). Even in this case, however, it was impossible to avoid all distractions, unlike in a laboratory designed for linguistic experiments. Second, because English is one of the official languages of the Philippines and is used commonly in the society, all the participants had some exposure to English at home, school, and elsewhere, and some of them were quite proficient bilinguals (much like the participants in Clemens et al. 2014). Although a language history survey confirmed the participants’ dominance in Tagalog, the possibility of influence from English remains strong. It would be possible to recruit participants outside of Manila and of lower socioeconomic status, whose exposure to English would likely be less, but this would probably entail a trade-off in the desirable qualities for an experimental setting. These problems, however, should not discourage researchers from doing experimental work in field research settings. Rather, such research should be encouraged so that we can develop new methodologies, as Clemens and colleagues (2014) suggested. Because research in psycholinguistics and acquisition generally lacks variety in languages and populations (Anand et al. 2011), an increase in field-based experimental work would also contribute to the cross-linguistic diversity of psycholinguistic and acquisition studies. However, as Clemens and her co-authors noted, we should always be aware of the