Error Analysis of Passivized Sentences Made by Japanese High School
Students:
A Voice Production Task
OKADA, Mihoko Graduate Student, Nagoya University
Abstract
Overpassivization is the mistake of overusing passive sentences with intransitive verbs and has been the subject of considerable research as a common problem among English as a Foreign Language (EFL) learners of various native tongues. The “Unaccusative Trap Hypothesis” (UTH) postulates that early-stage language learners do not distinguish unaccusatives from unergatives. Instead, after they correctly differentiate between these two types of intransitives, learners’ overpassivization errors appear to be linked to unaccusatives instead of unergatives. Many researchers have demonstrated that subject animacy influences learners’ choice of voice (active or passive). Research has focused on three factors and their interactions: verb categories, learners’ proficiency levels, and subject animacy. This study used statistical analyses of a Voice Production Task to investigate significant differences between unergative verbs and unaccusative verbs (both alternating and non-alternating). Furthermore, the interactions between these verb types and learners’ proficiency in high school students were considered. The findings demonstrate a statistically significant relationship between overpassivization errors and inanimate subjects by high school students. This indicates that learners overpassivize verbs with inanimate subjects once they acquire passive forms with transitive verbs. Thus, passivized errors of unaccusatives are caused more by subject animacy than the UTH.
1. Background – Overpassivization Errors 1.1 Unaccusatives and Unergatives
Intransitive verbs are split into two types: unaccusative verbs and unergative verbs. These two verb types have subjects with different thematic roles. Unaccusatives have a target subject, while unergatives have an agent subject. This structural difference is widely known as the Unaccusative Hypothesis (UH; Burzio, 1986; Perlmutter, 1978). The UH proposes a noun phrase and verb (NP-V) structure characteristic of unaccusatives by the NP-movement of the subject. The subjects of unaccusatives are moved from the object
position. Moreover, as unaccusative verbs are further divided into two types, depending on the existence of transitive counterparts, there are three kinds of intransitives: (a) alternating unaccusatives with transitive counterparts, (b) non-alternating unaccusatives without transitive counterparts, and (c) unergatives:
a. Alternating unaccusative: The snow melted.
(Park and Lakshmanan, 2007, p. 510) b. Non-alternating unaccusative: The guest arrived.
c. Unergative: The boy jumped. (Oshita, 2001, p. 280) Unaccusative verbs cannot be passivized, despite having a patient or target subject. As Park and Lakshmanan (2007) suggested, overpassivized sentences are common with unaccusative verbs among English as a Foreign Language (EFL) learners, regardless of their native language. Due to the different thematic roles of the subject, unaccusative verbs are reportedly overpassivized more often than unergative verbs. EFL learners tend to use the active voice with unergatives and the passive voice with unaccusatives. In this regard, learners seem to misunderstand the target subject as being the object of the sentence. However, the different thematic roles of the subject are not explicitly taught in Japanese high schools. Thus, EFL learners utilize the same nouns to create passive sentences with these subjects rather than active sentences. Previous research (e.g., Deguchi & Oshita, 2004; Kondo, 2005; No & Chung, 2006) has demonstrated that many EFL learners make more overpassivization errors with alternating unaccusatives, with the existence of their transitive counterparts than non-alternating unaccusatives.
1.2 Unaccusative Trap Hypothesis
Oshita’s (2001) Unaccusative Trap Hypothesis (UTH) argues that low- and intermediate-stage learners use one-argument verbs. Furthermore, intermediate-stage learners begin to overpassivize unaccusatives after their proficiency reaches the level necessary to distinguish between the unaccusative and unergative. That is, overpassivization errors with unaccusative verbs indicate that learners realize the underlying deep structure of unaccusatives. If learners begin to overpassivize only unaccusatives, it indicates that they have begun to implicitly distinguish the lexical meanings of unaccusative and unergative verbs. However, if they passivize both unaccusatives and unergatives, or neither of them, they have not yet begun to distinguish between them.
unergatives causes overpassivization errors. As language proficiency grows, overpassivization errors occur. These errors are accompanied by the realization of this distinction and, ultimately, disappear with the learners’ complete proficiency in distinguishing between the two. Some studies examining the UTH, however, have produced results that do not support the UTH. These contrary findings might be related to the various components of research procedures, among others (Oshita, 2014).
Another contention point related to the UTH is its primary premise that passivized errors with unaccusatives become apparent only after this verb class is distinguished from the unergative class in the learners’ interlanguage. That is, it is difficult to tell whether learners are aware of the distinction between the two types of intransitives when they make these errors.
1.3 Subject Animacy
Aissen (1999) argued that the association of the agentive role with a person/animate subject is the most robust generalization in syntactic markedness. Ferreira (1994) further contended that an “agent” is the voluntary cause of an action and, therefore, tends to be animate; thus, the agent typically takes an earlier position in an active sentence.
The different lexical meanings of verbs are accompanied differently in terms of their relation to nouns (located in the positions of subjects and objects of the sentences) and per the different thematic roles of nouns (such as agents–patients, targets–experiencers, and experiencers–targets; Shin, 2011). As the two types of intransitive verbs— unaccusatives and unergatives—are differentiated by the functions and animacy of the subjects, it can be difficult for Japanese EFL learners to capture the nuances of these verb structures. These nuances are not explicitly taught in Japanese high schools, and, therefore, students do not generally encounter them. Students must implicitly learn the thematic relations of nouns to individual verbs as well as the correct choice of voice, judged from the existence of external agents when using unaccusative verbs.
Despite the difficulties inherent in investigating the interaction effects between verb types and subject type, many researchers have indicated that EFL learners who are native speakers of Asian languages tend to overpassivize sentences with inanimate subjects. Some researchers emphasize the influence of learners’ first language (L1), especially Asian languages, in the preference for animate subjects over inanimate ones. Additionally, the passive voice in English differs significantly from that in Japanese, especially regarding inanimate subjects. The passive voice is considered difficult for Japanese EFL learners to master. Hinkel (2002) suggested that Japanese speakers of English prefer to
use active verbs with an animate subject because sentence construction in Japanese favors the animate noun and active verb pair. Okada (2018) also found that in Japanese novels and their English translations, English passive sentences with inanimate subjects corresponded mostly to stative verb sentences with the same subjects in the Japanese originals. Thus, overall, there is a possibility that an EFL learner’s syntactic choice, including overpassivization errors, will be influenced by the animacy of the subject of the sentence. However, as Choi (2019) mentioned, although several researchers have expressed interest in the effect of subject animacy on overpassivization errors, their results on whether animate or inanimate subjects cause more errors are contradictory.
2. Research Questions
In the existing literature reviewed, most Japanese participants were undergraduate or graduate students learning English abroad as a second language or majoring in English in Japan. Thus, they were considered to be at the intermediate or upper stage of English language proficiency. The U-shaped developmental pattern in the UTH indicates that lower-stage proficiency learners make fewer passivization errors, while intermediate learners produce many more errors when they become aware of the verb distinction. Finally, learners use these verbs correctly at the upper or final stage of complete proficiency, and it is necessary to examine how high school students, who are primarily at the low or intermediate stages of proficiency, deal with these intransitive verbs. Furthermore, scores on English proficiency tests should be used to increase the research’s credibility regarding the identified stage of development.
Due to semantic and cultural background differences, little previous research has dealt with the relationship between overpassivization and factors such as subject animacy. Japanese speakers of the English language prefer to use active verbs with an animate subject because Japanese sentence construction favors the animate noun and active verb pair (Hinkel, 2002). On the other hand, Aissen (1999) illustrated that the association of the agentive role with a person/animate subject is universally syntactically unmarked in L1 and second language (L2) speakers. Subsequently, it is necessary to closely examine how subject animacy affects Japanese EFL learners’ overpassivization errors with intransitive verbs.
This study contributes to language teachers and second language acquisition (SLA) researchers by demonstrating what affects overpassivization errors with intransitive verbs made by Japanese high school EFL learners. Subsequently, the following three research questions were developed:
RQ 1: Does the verb’s lexical meaning affect errors involving passivized sentences made by Japanese high school students?
RQ 2: Does the level of English proficiency affect errors involving passivized sentences made by Japanese high school students?
RQ 3: Does the subject’s animacy affect errors involving passivized sentences made by Japanese high school students?
3. Method 3.1 Participants
The research involved 220 second-year students from a Japanese public high school who completed the Voice Production Task (VPT). Ten students were excluded from this study, bringing the final total of the participants to 210. The following were reasons for exclusion: one student used to live abroad, three students who were of foreign nationality spoke languages in addition to Japanese at home, two students were involved in creating the research tasks, and four students were unwilling to participate properly by completing all the answers without changing the original forms. Consent forms were signed by both the students and their parents, and the forms included information on ethical considerations regarding anonymity, confidentiality, and appropriate handling of personal data by the researcher.
3.2 Materials
Table 1 presents the three types of verbs and transitive verbs used in the VPT. All intransitive verbs were chosen from target verbs used in previous studies (Chung, 2014; Deguchi & Oshita, 2004; Granger, 2013; Kondo, 2005; No & Chung, 2006; Owada, 2013; Shin, 2011). The transitive verbs were derived from Hirakawa (1995, 2013), the only studies known by the author to use transitive verbs as the research targets.
The classification criteria for subject animacy were set as follows:
(1) Animate subjects were primarily human beings and animals, including characters in novels and comic books.
(2) Inanimate subjects were mainly objects, natural phenomena, and plants, including the titles of novels and comic books.
All the task sentences were given to the participants, with each sentence consisting of a main clause and an adverbial phrase or clause to clarify the context. Task sentences were extracted from high school textbooks and a dictionary. All 20 verbs were in the 3,000-word list for Japanese EFL learners (Yokokawa, 2006), which is ranked by
familiarity. Taking the average of the rankings of the 20 verbs resulted in a mean of 675.85. Table 2 presents the number of sentences in the task classified by verb types and subject animacy, according to the categorization criteria from the previous section. Table 1
Classifications of Verbs Verb
type
Unaccusative /
unergative Target verbs
A Unaccusative Alternating burn change drop freeze grow B Non-alternating become die disappear exist fall
C Unergative run smile swim walk work
T Transitive build hit love need read
Note. A = alternating unaccusative, B = non-alternating unaccusative, C = unergative, T = transitive.
Table 2
Numbers of Sentences: Verb Type, Voice, and Subject Animacy
Voice Active Passive
Verb type A B C Subtotal T Total
Animate subject 5 5 10 20 4 24
Inanimate subject 5 5 10 6 16
Total 10 10 10 30 10 40
Note. A = alternating unaccusative, B = non-alternating unaccusative, C = unergative, T = transitive.
The direction for the task and an example of the task sentences are as follows: “Write the adequate form of the verb in the parentheses on the blank in the sentence. Be sure not to change the part of speech and use the word in parentheses as a verb.
Example: If one kind of life ____, the food chain breaks. (disappear) [food chain=食物連鎖]
3.3 Procedure
Each half of the VPT used the same verb with either an animate subject or an inanimate subject (the same two unergatives both have animate subjects, as illustrated in Table 2).
Each of the classes was 50-minutes in length, and each half of the task itself lasted approximately 15 minutes. Due to the differences in the amount of time it took for participants to complete the task in each class, the time allotted for the task varied from 15 to 20 minutes. The task administrator verbally explained the task’s aim and procedures. The participants were informed that this task was aimed at increasing their grammatical knowledge and that the task results might affect their class grades. No dictionaries or grammar books were permitted.
The voice forms of the target verbs made by the participants were classified into four categories: (1) grammatical, (2) ungrammatical, (3) unclear, and (4) unanswered. When the target verbs were intransitives, active forms were grammatical and passive ones were ungrammatical, while the target verbs were transitives, the voice forms were judged by the opposite criteria. The judgment criteria for the four categories and the voice forms in this study are as follows:
(1) Active voice.
basic, perfect, progressive (basic and perfect) active voice, -ed form of irregular verbs or -ing form without be, have/has/had + base or -ing form
(2) Passive voice.
be + base form, past participle (including -ed form of irregular verbs), third-person present forms, and past-tense forms
(3) Unclear
past participle of irregular verbs without be, (be+) to + base, simple tense, perfect forms and participles, other parts of speech, make + O + C sentence structure (4) Unanswered
no answer, answers with the words in the [ ] brackets
The results of the VPT were analyzed with the following procedures: 1) the hand-written answers were digitalized into text files, including participant ID numbers and their English proficiency data; 2) the verb forms were classified into one of four categories according to the criteria mentioned above (subject animacy was added to the data set as animate subjects and inanimate subjects).
3.4 Analyses
3.4.1 Accuracy Rates
Table 3 presents the number of valid answers given by the 210 participants, including participants with no English-language proficiency data. “Unclear” and “unanswered” responses were entered as missing values based on the criteria in section 3.3. The
accuracy rates for the verb types and animacy are also illustrated in Table 3, excluding 567 missing values out of a total of 8,400 responses (5.7%).
Table 3
Accuracy Rates of Verb Type, Voice, and Subject Animacy, Excluding Those Unclear or Unanswered
Voice Active Passive Total
Verb type A B C Subtotal T
Animate subject 91% 94% 98% 95% 10% 81%
Inanimate subject 87% 90% 88% 24% 65%
Total 89% 92% 98% 93% 19% 75%
Note. A = alternating unaccusatives, B = non-alternating unaccusatives, C = unergatives, T = transitives.
Table 4
Descriptive Statistics for the Voice Production Task
Source N = 209 M SD Proficiency 44.78 4.13 Verb A 0.89 0.32 Verb B 0.92 0.27 Verb C 0.98 0.13 Verb T 0.19 0.39 Animate subject 0.81 0.39 Inanimate subject 0.65 0.48
Note. verb A = alternating unaccusatives, verb B = non-alternating unaccusatives, verb C = unergatives, verb T = transitives. ( ) = missing values or unclear/unanswered answers.
Table 4 illustrates the English proficiency levels and descriptive statistics for the VPT completed by the 209 high school students, further excluding one participant who did not have a proficiency score. The participants’ deviation values were scaled to a mean of 50 on a nationwide college preparatory exam and collected to compare the participants’ English proficiency levels because few participants had taken the TOEIC or TOEFL. If the participants had missing values (ratings) for any verb type or subject animacy, those answers were excluded from the descriptive statistics. However, for the generalized linear mixed model (GLMM; discussed in the next section), the ratings were computed using R
software, with missing values treated as missing completely at random using the full information maximum likelihood method.
3.4.2 Generalized Linear Mixed Models
The three research questions were examined using statistical analyses of the results conducted using a GLMM. The lme4 package and lmerTest package of the statistical software R (ver. 3.6.3) were used to calculate a maximum likelihood estimate, standard errors, z values, p values, and correlations were calculated and described with appropriate plots.
The participants’ answers (ratings) were analyzed as dependent variables with verb types, subject animacy, and proficiency for the statistical analyses. The answers determined to be grammatical were denoted with a “1” and the ungrammatical ones with a “0” based on the criteria in section 3.3. The answers determined to be “unclear” or “unanswered” were considered as missing values and were left blank. Participants and sentence items were input as crossed random effects. English-language proficiency data constituted a continuous set and was centralized. When necessary, a participant random slope and an item random slope for proficiency, verb type, and subject animacy were included to identify the best model with the lowest Akaike’s information criterion (AIC). This research adopted the model with interactions between proficiency, verb type, and subject animacy with by-subject random slope for verb type as the most adequate model with the lowest AIC as follows:
Model=glmer(rating~(1+verb|subject)+(1|item)+proficiency*verb*animacy, data=dat, family=binomial, glmerControl(optimizer ="bobyqa"))
Furthermore, to compare the effects of verb type and subject animacy, the same model was repeated with the addition of the command below to change the reference level to that of either the non-alternating unaccusatives (B) or the unergatives (C) and that of inanimate subjects. However, since the results between subject animacies were revealed to be the same, but with opposite (positive or negative) tendencies, Table 5 excluded the results for the inanimate subject that was referenced in the pairwise interactions between verb types and animacy as well as the three-way interactions (proficiency, verb type, and subject animacy).
dat$verb=relevel(dat$verb, ref = “B”), or dat$verb=relevel(dat$verb, ref = “C”) dat$animacy=relevel(dat$animacy, ref = “inanimate subject”)
4. Results and Discussion
4.1 Overpassivization Errors Related to the Lexical Meanings of Verbs
The accuracy rate of the alternating unaccusatives (A) was 89%, while that of non-alternating unaccusatives (B) was 92%. Statistical significance was not obtained between the two types of unaccusatives as illustrated in Table 5. The accuracy rates were lower for the two types of unaccusatives than for the unergatives (C), which produced an accuracy rate of 98%. The GLMM analysis (excluding the one participant without proficiency data) demonstrated that the participants made more passivization errors with unaccusatives (A) and (B) than with unergatives (C). This result aligns with previous studies. In this study, the participants were asked to provide appropriate forms of the given verbs so that the researcher could determine if they understand the lexical meanings of the verbs.
The results of the analysis of variance (ANOVA; Type III Wald chi-square test) of the global model containing all three factors and their interaction demonstrated that the simple effect of verbs was significant (χ2(3) = 224.31, p < .001). Of all the pairwise
interactions between the two effects, the interactions among proficiency and verb type (χ2(3) = 45.56, p < .001), proficiency and subject animacy (χ2(1) = 11.01, p < .001), and
verb type and subject animacy (χ2(2) = 7.85, p = .020) were statistically significant. The
interactions of all three effects (proficiency, verb type, and subject animacy) were also statistically significant (χ2(2) = 18.31, p < .001).
The GLMM analysis revealed statistically significant interactions between the transitives and all three intransitive verbs. It appears that Japanese high school students overpassivized more unaccusatives than unergatives when given a task focusing only on the verb forms. This may be partly due to the tenets expressed in the UH and the UTH. As discussed in Section 1, the EFL learners’ language development stage—especially understanding passive voice, subject animacy, and L1—is strongly related to overpassivization errors.
Table 5
GLMM Results of the Interactions Among Verb Type, Proficiency, and Subject Animacy as Fixed Effects and Participants and Items as Random Effects
Parameter Animacy Estimate SE z value Pr(>|z|)
Verb A (verb B referenced) Animate-ref -1.002 0.607 -1.653 .098 . Inanimate-ref -0.147 0.585 -0.251 .802 Verb A (verb C referenced) Animate-ref -1.984 0.564 -3.517 <.001 *** Verb B (verb C referenced) Animate-ref -0.982 0.585 -1.677 .094 . Verb T (verb A referenced) Animate-ref -6.017 0.646 -9.310 <.001 *** Inanimate-ref -4.576 0.583 -7.849 <.001 *** Verb T (verb B referenced) Animate-ref -7.023 0.670 -10.481 <.001 *** Inanimate-ref -4.724 0.586 -8.066 <.001 *** Verb T (verb C referenced) Animate-ref -8.004 0.624 -12.830 <.001 *** Proficiency -0.024 0.016 -1.456 .145 Proficiency : verb A (verb B referenced) Animate-ref -0.047 0.044 -1.067 .286 Inanimate-ref -0.073 0.037 -1.979 .048 * Proficiency : verb A (verb C referenced) Animate-ref 0.001 0.048 0.015 .988 Proficiency : verb B (verb C referenced) Animate-ref 0.048 0.054 0.892 .373 Proficiency : verb T (verb A referenced) Animate-ref 0.240 0.062 3.893 <.001 *** Inanimate-ref 0.447 0.057 7.814 <.001 *** Proficiency : verb T (verb B referenced) Animate-ref 0.193 0.066 2.914 .004 ** Inanimate-ref 0.375 0.058 6.489 <.001 *** Proficiency : verb T (verb C referenced) Animate-ref 0.241 0.064 3.772 <.001 *** Animacy -0.033 0.353 -0.093 .926 Verb A : animacy (verb B referenced) Animate-ref 0.856 0.822 1.041 .298 Verb T : animacy (verb A referenced) Animate-ref 1.441 0.815 1.769 .077 . Verb T : animacy (verb B referenced) Animate-ref 2.298 0.829 2.772 .006 ** Proficiency : animacy -0.077 0.023 -3.318 <.001 ***
Note. verb A = alternating unaccusatives, verb B = non-alternating unaccusatives, verb C = unergatives, verb T = transitives, Animate-ref = animate subjects referenced, Inanimate-ref = inanimate subjects referenced. The results of simple effects and interaction “Proficiency”, “Animacy”, and “Proficiency : animacy” were obtained through an analysis with contrast cording treatment to recenter the categorical
variables: four types of verbs and both animacies. Statistical significance is represented by the codes ‘***’ (p < .001), ‘**’ (p < .01), ‘*’ (p < .05), and ‘.’ (p < .10).
4.2 Overpassivization Errors and Learners’ Proficiency Levels
The average correct answer rate for every intransitive was high, approximately 90%. As illustrated in Figure 1, the higher the student’s proficiency, the lower the accuracy rate for all the intransitives, especially the alternating unaccusatives (A). The GLMM analysis did not reveal statistically significant interactions between language proficiency and the three verbs. These insignificant interactions between learners’ proficiency levels and overpassivization errors do not support the UTH postulation of a relationship between these syntactic errors and the developmental learning stages in distinguishing the unaccusatives from the unergatives (Oshita, 2001).
While the effects of proficiency on altering unaccusatives (A) and non-altering unaccusatives (B) were not different from those on unergatives (C) for animate subjects, the effects of proficiency on altering unaccusatives (A) were marginally significantly different from those on non-altering unaccusatives (B) in the case of inanimate subjects. The effects of proficiency on transitives (T) were significantly different from those of all types of intransitives (A), (B), and (C). This suggests that students gradually learn to correctly use active forms for unaccusative verbs with animate subjects. On the other hand, they do not learn to correctly choose voice forms with inanimate subjects.
Proficiency : verb A : animacy (verb B referenced) Animate-ref -0.026 0.057 -0.452 .651 Proficiency : verb T : animacy (verb A referenced) Animate-ref 0.208 0.052 4.004 <.001 *** Proficiency : verb T : animacy (verb B referenced) Animate-ref 0.182 0.057 3.170 .002 **
Unrelated to proficiency, EFL learners tend to correctly use active forms for unergatives, which have only animate subjects in the task sentences. The current study demonstrated that, as the acquisition of the passives proceeds, the altering (A) and non-altering (B) unaccusatives tend to be overpassivized more than unergatives (C).
Figure 1
Interactions of the Four Verb Types, Proficiency, and Subject Animacy
Note. Although the task sentences in this case study included no instances of unergatives with inanimate subjects, the plot in Figure 1 depicts a line for the verb in the inanimate subject section. This is because the plot command was used to draw a plot to clearly illustrate the results of the three effects—proficiency, verb type, and subject animacy— as dependent variables. This regression model object with the type “pred” is a plot to illustrate the predicted values (marginal effects).
Significant interactions between proficiency and verb types were found among the transitives (T) and the three intransitive verbs, indicating that learners develop their skills to use passive forms with transitives while their skills regarding the use of active forms of intransitive verbs do not develop in the same way. The correlation analysis results in
Figure 2 also support the tendency for proficiency and transitives (T) to correlate (r = .53). On the other hand, proficiency does not correlate with the intransitives. The higher the scores of transitive verbs, the more overpassivization errors occur with the unaccusatives. This phenomenon is evident from the correlations between the accuracy rate of transitives to alternating unaccusatives (A) (r = -.46) and to non-alternating unaccusatives (B) (r = -.28). The data in this study demonstrate that, contrary to the UTH predictions, overpassivization errors by high school students may be caused by the level of acquisition of the passive form.
Figure 2
Correlations Among the Four Verb Types and Proficiency
Note. VerbA = alternating unaccusatives, verbB = non-alternating unaccusatives, verbC = unergatives, verbT = transitive verbs
4.3 Animacy Effect on Voice Choice
When focusing on the pairwise interaction of verb types and subject animacy, a significant interaction was found for transitives (T) and non-alternating unaccusatives (B). The alternating unaccusatives (A) did not demonstrate a statistical difference with transitives (T) in how animate and inanimate subject sentences are passivized. While inanimate subjects are more prone to overpassivization in both verb types, only the alternating unaccusatives (A) demonstrated a gradual shift toward overpassivization for
the animate subjects. This is both an interesting similarity and a significant difference between the two unaccusatives. Despite this observation, the effect of proficiency on altering unaccusatives (A) was marginally significantly different from that on non-alternating unaccusatives (B) in the case of inanimate subjects. Meanwhile, there was no statistically significant difference between the two unaccusatives for animate subjects. The interactions among proficiency, verb types, and subject animacy with the transitives (T) and the unaccusatives (A) and (B) were statically significant. This suggests that learners overpassivize intransitives with inanimate subjects compared to animate subjects. However, they correctly passivize with inanimate subjects of transitives more frequently as their proficiency increases. Both findings indicate that inanimate subjects are more prone to passivization, whether they are correct with transitives or incorrect with intransitives.
Hinkel (2002) studied overpassivization in relation to subject animacy with native Chinese, Korean, Japanese, and Spanish-speaking university students using the grammaticality judgment task. Hinkel’s (2002) findings argued that these learners tend to passivize sentences with inanimate subjects. The reason for this phenomenon appeared to be that, in Japanese and some other Asian languages, the subject should have higher animacy than the object. Similarly, Aissen (1999) asserted that the sentence structure “subject agent (animate) + active form” is universal across languages with an unmarked argument structure, including English and the learners’ target language.
5. Conclusion and Future Implications
This study—a VPT with the allotted verbs and clear animacy of subjects—was conducted with sentences extracted from high school textbooks and a dictionary. The results revealed the main effects of verbs correlated among the verb types. In terms of the effect of the participants’ proficiency on passivized errors, this study demonstrated that, as the acquisition of the passives proceeds, the alternating unaccusatives (A) and non-alternating unaccusatives (B) tend to be overpassivized more commonly than the unergatives (C).
The GLMM analysis revealed statistically significant interactions between the choice of voice (active intransitives or passive transitives) and subject animacy. Researchers’ opinions differ on whether subject animacy induces more overpassivization in EFL learners. Although subject animacy has been claimed to affect overpassivization errors because of differences in language characteristics between the learners’ native languages, some researchers (e.g., Aissen, 1999) have asserted that the sentence structure “animate
subject + active voice” is universal and an unmarked argument structure, including in English.
A new hypothesis can be considered based on this research’s findings regarding Japanese high school students’ overpassivization errors: as the learners’ acquisition of passive forms proceeds alongside their language proficiency, the more they passivize active forms with inanimate subjects. This may be partly due to their L1 features and partly due to universal language features. This finding demonstrates that passivized errors of unaccusatives are caused more by subject animacy than the UTH. To study this hypothesis further, four implications for future research are proposed. First, other types of tasks, such as grammaticality judgment tasks, can be used. Second, longitudinal studies should be conducted to reduce the obstacles in acquiring adequate research results, such as the difference in participants’ learning environments. Third, participants with broader levels of language proficiency and types of educational environment could be utilized. Finally, in future research, participants with various first languages should be investigated to further solidify the hypothesis. Using these guidelines, further contributions of future research in overpassivization errors can be expected to benefit the SLA field.
Acknowledgments
I would like to thank Associate Professor Remi Murao at Nagoya University Graduate School for her constructive comments on an earlier draft of the paper and the two anonymous reviewers for their valuable feedback.
Note
Parts of this article were presented at the LET Chubu 95th (Autumn) Conference on-line on December 13, 2020.
References
Aissen, J. (1999). Markedness and subject choice in Optimality Theory. Natural Language & Linguistic Theory, 17(4), 673–711.
https://doi.org/10.1023/A:1006335629372
Burzio, L. (1986). Italian syntax: A government-binding approach. Kluwer Academic Publishers. http://doi.org/10.1007/978-94-009-4522-7
Choi, J. (2019). Overpassivization of English unaccusatives revisited: A look behind the scenes. 영어학 [English Language Studies], 19(3), 407–431.
Chung, T. (2014). Multiple factors in the L2 acquisition of English unaccusative verbs. International Review of Applied Linguistics in Language Teaching, 52(1), 59–87. https://doi.org/10.1515/iral-2014-0003
Deguchi, A., & Oshita, H. (2004). Meaning, proficiency, and error types: Variations in nonnative acquisition of unaccusative verbs. EUROSLA Yearbook, 4, 41–65. https://doi.org/10.1075/eurosla.4.04deg
Ferreira, F. (1994). Choice of passive voice is affected by verb type and animacy. Journal of Memory and Language, 33, 715–736.
https://doi.org/10.1006/jmla.1994.1034
Granger, S. (2013). The passive in learner English: Corpus insights and implications for pedagogical grammar. Learner Corpus Studies in Asia and the World, 1, 5–15. http://www.lib.kobe-u.ac.jp/repository/81006671.pdf
Hinkel, E. (2002). Why English passive is difficult to teach (and learn). In E. Hinkel & S. Fotos (Eds.), New perspectives on grammar teaching (pp. 233–260). Lawrence Erlbaum Associates. https://doi.org/10.4324/9781410605030
Hirakawa, M. (1995). L2 acquisition of English unaccusative constructions. In D. MacLaughlin & S. McEwen (Eds.), Proceedings of the 19th Boston University conference on language development (Vol. 1, pp. 291–303). Cascadilla Press. https://www.jstor.org/stable/44485737?seq=1
Hirakawa, M. (2013). Alternations and argument structure in second language English: Knowledge of two types of intransitive verbs. In M. Whong, K. H. Gil, & H. Marsden (Eds.), Universal grammar and the second language classroom (pp. 117– 137). Springer. https://link.springer.com/chapter/10.1007/978-94-007-6362-3_7 Kondo, T. (2005). Overpassivization in second language acquisition. International
Review of Applied Linguistics in Language Teaching, 43(2), 129–161. https://doi.org/10.1515/iral.2005.43.2.129
No, G., & Chung, T. (2006). Multiple effects and the learnability of English unaccusatives. English Teaching, 61(1), 19–39.
http://www.dbpia.co.kr/pdf/pdfView?nodeId=NODE07096611
Okada, M. (2018). Nihon bungaku sakuhin no eiyakusho ni okeru museibutushugo hyougen to gensho no nihongo kouzou to no hikeku [A comparative study of inanimate subject expressions in Japanese novels’ English translated versions and their original Japanese constructions]. Journal of the Chubu English Language Education Society, 47, 197–204. https://doi.org/10.20713/celes.47.0_197
Second Language Acquisition, 23, 279–304. https://doi.org/10.1017/S0272263101002078
Oshita, H. (2014). U-shaped development: definition, exploration, and falsifiable hypotheses. In J. Connor-Linton & L. Wander Amoroso (Eds.), Measured language—Quantitative approaches to acquisition, assessment, and variation (pp. 155–170). Georgetown University Press.
https://repository.library.georgetown.edu/bitstream/handle/10822/1044614/9781 626160378_GURT_2014.pdf?sequence=1&isAllowed=y
Owada, K. (2013). A corpus study of verb patterns of unaccusative verbs “appear,” “happen,” and “occur” in L2 English written by Japanese learners of English. Pan-Pacific Association of Applied Linguistics, 17(1), 29–38.
https://files.eric.ed.gov/fulltext/EJ1026105.pdf
Park, K. S., & Lakshmanan, U. (2007). The unaccusative-unergative distinction in resultatives: Evidence from Korean L2 learners of English. In A. Belikova, L. Meroni, & M. Umeda (Eds.), Proceedings of the 2nd conference on Generative Approaches to Language Acquisition North America (GALANA) (pp. 328–338). http://www.lingref.com/cpp/galana/2/paper1573.pdf
Perlmutter, D. M. (1978). Impersonal passives and the unaccusative hypothesis. In Proceedings of the 4th Annual Meeting of the Berkeley Linguistics Society (pp.
157–190).
http://journals.linguisticsociety.org/proceedings/index.php/BLS/article/view/219 8
Shin, J. (2011). Overpassivization errors in Korean college students’ English writings. Korean Journal of Applied Linguistics, 27(3), 255–273.
https://www.researchgate.net/publication/235333180_Overpassivization_errors_i n_Korean_college_students%27_English_Writings
Yokokawa, H. (Ed.). (2006). Kyoiku kenkyuu no tame no dainigengo database: Nihonjin eigo gakushusha no eitango shinmitusdo (Moji hen) [Database for teaching and research of second language: English vocabulary familiarity of Japanese EFL learners (Visual version)]. Kuroshio Publishers.
