Competing past tense forms in English attrition

Abstract

Aims and objectives:

This paper examines longitudinal speech data from two Korean sisters, focusing on English irregular past tense forms to probe the course of attrition within the framework of the regression hypothesis, which suggests that language is lost in the reverse order of acquisition. During the course of attrition, evidence supporting one of the two irregular past tense acquisition theories (blocking vs competition hypothesis) is manifested.

Methodology:

The loss of English past tense forms of two girls who had lived in Anglophone Hawaii for two years before returning to Korea is tracked using free speech samples.

Data and analysis:

The children’s naturalistic speech data collected over a three-year period after their return is analyzed in terms of accuracy and error types.

Findings/conclusions:

Although the older sister did not exhibit clear signs of attrition, the constitution of the younger sibling’s licit and illicit past tense usage varied every year, reflecting her declining proficiency. The results also show that the path of attrition follows the prediction of the regression hypothesis.

Originality:

The prolonged attrition process of the sisters’ language use, which is unlike acquisition that can happen at a quick rate, demonstrated a relatively large window to witness their reshaping grammar at different interlanguage stages.

Implications:

The sisters’ irregular past forms retreating to a more rudimentary form provided an opportunity to support the competition model of irregular past tense acquisition. Their past tense accuracy and error analysis demonstrated various past tense forms in competition that could produce a different winner over different periods of time.

Keywords

Attrition English irregular past tense regression hypothesis competition model blocking model

Introduction

The phenomenon of individual non-pathological language loss is primarily the result of contact with another language in a new environment. Issues pertaining to language attrition have been consistently gaining scholarly interest in recent years as they contribute valuable insights into human cognition, bilingualism research, and linguistic theory. There are some important generalizations derived from attrition research. First, some aspects of the attriting grammar are more vulnerable than others. For example, Turkish L1 attriters, under the influence of L2 English, showed decline only in the binding properties of the overt pronoun o while null pronoun and reflexive kendisi were maintained (Gürel, 2004). Another finding is that attrition involves the initially more complex and more narrowly distributed rules being replaced by less complex rules with wider distribution. Seliger (1991) demonstrates that a young female’s L1 English dative alternation rule, which is lexically governed, deteriorated under the influence of a more general and syntactic Hebrew rule after the child moved to Israel.

Another important issue related to attrition is whether an order or sequence of language loss follows a predictable path analogous to the stages of acquisition (Brown, 1973). The regression hypothesis, also referred to as first in, last out, suggests that language is lost in the reverse order of the acquisition of the same language (Bardovi-Harlig & Stringer, 2010; Ecke, 2004; Jakobson, 1941) as the most conceptually and/or formally complex structures that are acquired late diminish early, and easy-to-process structures that are learned first tend to remain resistant to loss. While there is no consensus agreement on whether the course of attrition is actually a mirror image of acquisition, several studies have successfully tested this hypothesis.

Cohen (1975) presented both supporting data for the hypothesis and some counter-evidence against it. Data collected from three first grade L1-English children who were L2 learners of Spanish provided a number of examples to support the regression hypothesis in several aspects of their post-summer break Spanish use. One child, for instance, substituted the third person present tense inflection for the first person (e.g. *yo puede instead of yo puedo). The child appeared to have acquired this distinction late, which suggests that this was a case of last learned – first forgotten. However, while two of the three children provided a few cases of unlearning in the reverse order from the original acquisition process, one child did not go through such a process in her attrition. Then, Berman and Olshtain (1983) and Olshtain (1989) reported signs of regression in children who overregularized English past tense morphemes after their return to Hebrew-speaking countries. Kuhberg (1992) also argued that attrition in Turkish children’s deteriorating German was largely a mirror image of acquisition based on his observations on lexical loss and a few morpho-syntactic systems. More recently, Anderson (2001) stated based on the findings of Spanish–English bilingual children’s third person singular form usage that these children had regressed to an earlier Spanish developmental stage. In addition, Keijzer (2010) tested loss of morphology and syntax in Dutch immigrants in Anglophone Canada. Evidence in favor of the regression hypothesis was found, mainly in the morphological domain, such as plural and diminutive morphology on noun phrases.

In this paper, the regression hypothesis is used as an explanatory framework to account for the course of attrition in two Korean children losing English irregular past tense forms. Then, within the time course itself, two competing models of irregular past tense acquisition presented in the next section, blocking versus competition hypothesis, are tested for their validity and applicability to the data.

Acquiring the English irregular past tense

Regular English verbs are inflected into the past tense with the –ed suffix. Some irregular verbs do not conform to this rule, however, and this leads children to make two types of errors: overregularization and mis-irregularization. Overregularization involves applying the regular past tense suffix to the bare form of irregular verbs, as in breaked and comed instead of broke and came. Marcus et al. (1992) report that overregularization errors occur in about 4.2% of all instances of irregular verb use. In contrast, mis-irregularization is the selection of the wrong irregular past tense form, resulting in deviant forms such as brang instead of brought for the past tense of bring (probably due to analogy of ring–rang), or the deviant wope as the past tense of wipe. However, these types of errors are exceedingly rare. In fact, they represent only 0.2% of all instances of irregular use (Xu & Pinker, 1995).

In the acquisition process, children are known to go through several steps before they can correctly use the irregular past tense forms. First, they start out using the bare form of the verb without any tense marker (1a). Then, there is a period when they accurately use the correct irregular form (1b) before they acquire the –ed morpheme for regular verbs. Once the regular past morpheme is learned, there seems to be a period of mixing the correct irregular and incorrect overregularized form (1c). Finally, with more input and experience, they figure out how to avoid the overregularization and use the irregular form (1d).

(1) a. No inflection: hold

b. Irregular form: held

c. Mixing irregular and overregularized forms: held and holded

d. Irregular form: held

Disagreement on the intermediate stage (1c), where the irregular and overregularized forms coexist, has been an interesting topic in English irregular past tense acquisition (e.g. McClelland & Patterson, 2002a, 2002b; Pinker & Ullman, 2002; Ullman, 1999). When a child acquires English, she presumably hears only the correct form of irregular past tense in the input. The blocking hypothesis (Marcus et al., 1992) claims that the knowledge of the irregular past tense form will block the application of the regular past tense form. However, the child may occasionally fail to retrieve the correct irregular form. Such a failure would provide an opportunity for the regular rule to apply as a default, causing overregularization errors such as comed or holded. Thus, Marcus et al. consider overregularizations as speech errors rather than as grammatical alternatives to the irregular forms. The blocking hypothesis predicts that unless the regular rule had been applied for a prolonged period preceding the acquisition of the correct irregular form, overregularization would be rare because such a retrieval failure, which gives way to the incorrect regular –ed rule, would not frequently occur. Based on the low overregularization rate in their data, Marcus et al. favor the blocking hypothesis.

However, Maratsos (2000) refutes Marcus et al.’s (1992) claim and strengthens the case for competition hypothesis. The competition model maintains that instead of the irregular past form blocking the regular past morpheme, the two forms compete until the correct form wins out. Although Marcus et al. (1992) argue that the competition model would predict a relatively high rate of overregularization, Maratsos accounts for the low overregularization rate by focusing on the period during which the competition is carried out. He demonstrates that although the overall overregularization rate may seem low in the long term, the rate is relatively high during the competition period, which can be very brief in the case of high frequency irregular forms.

English irregular past tense in attrition research

Several longitudinal studies document morpho-syntactic changes in young children’s language when they move to an environment where that language is no longer dominant. Among others, Reetz-Kurashige (1999), Tomiyama (1999), and Yoshitomi (1999) collected English data for at least a year from several young Japanese returnees from the U.S. and reported their use of irregular past tense verbs.

Reetz-Kurashige (1999) tested 18 Japanese returnee children¹ on two types of storytelling tasks. The study assessed the children individually at two or three different times over 12 to 19 months. Reetz-Kurashige analyzed the returnees’ usage of verb tense and aspect in terms of target-like usage (TLU)². The observed changes in TLU from one test session to the next indicated that seven participants from each storytelling task retained or performed better on English irregular past tense forms on the final test than on the first test. Six children from the first task and five children from the second task showed less than a 20% decline in TLU after 12 to 19 months. The other four and five children from each task either showed more than a 20% decline in TLU or began with less than 60% TLU in the first test.

Yoshitomi (1999) also used TLU to measure the performance of four girls who had lived in the U.S. for 3;3 to 5;5 years. The study was conducted for a year, and the children were tested four times using “story description” and “free interaction” tasks. All participants maintained close to 100% TLU in every “story description” task. Two of the girls also maintained 100% TLU across all the “free interaction” tasks. Of the remaining two, one girl showed improvement over the course of the study, ending up at 100% in the last session, after starting at 59% TLU. The other showed a slight decline from 100% in sessions one and two to 97% in session three and 86% in session four. Thus, none of the participants showed a significant decline in irregular past tense TLU.

Tomiyama (1999) observed an eight-year-old boy who had lived in California for nearly seven years. Data collection began two months after he had returned to Japan and continued to 19 months. His irregular past tense verb data are based on the bilingual syntax measure (BSM), spontaneous data, and storytelling. His accuracy for the BSM was perfect at the two-, nine-, and 11-month points, although he produced only three or four tokens in each session. However, his spontaneous and storytelling data showed signs of decline in correct usage. Spontaneous data show that he used the irregular past tense perfectly at the second month, but his accuracy fell to 89% at the 13th month and to 86% by the 17th month. Moreover, his storytelling data indicated a sharp decline to 69% at the 19th month after perfect accuracy at both the nine- and 13-month points.

The three studies mentioned above all followed returnee children for a minimum of one year and tested the children two or three times. Although Tomiyama (1999) tested her participant 13 times, she reported results from only three sessions. Data from these three studies show that many children retain their English irregular past tense forms even in their second or third year after returning to Japan from the U.S. It must be noted that although English input shrank significantly, efforts to maintain their English were conducted, such as watching videos and enrolling in group English lessons (Tomiyama, 1999), or participating in English-maintenance classes on weekends (Yoshitomi, 1999).

The current study

Some aspects of L1 acquisition are accomplished at a quicker rate than acquisition studies involving limited data collection can accurately observe. This means that we may not have a prolonged time frame for a clear picture of the irregular past tense acquisition process, which is why conflicting views like the blocking and competition hypotheses have arisen to account for the mechanism of language through irregular past tense errors. On the other hand, the language attrition process, which may be halted or at least may take longer than the acquisition process due to various paralinguistic factors such as the high value of the attriting language in the new environment or parents’ investment in maintaining their children’s English, may provide more robust insights into the processing mechanism. According to the regression hypothesis, the expected order of losing the irregular past tense is depicted in (2), which is the reversal of process (1) above. However, stage (2c) will be eliminated from further consideration as it seems impossible to revert to this stage during the attrition process. The reason this stage appears prior to mixed usage of the correct and overregularized forms in the acquisition process is due to the child’s rote repetition of the correct form in the input before they fully understand the regular rule. Once the regular –ed rule is acquired, it is intact as long as the child understands past tense. Thus, the stage in which only the correct irregular form appears (2c) after a stage of mixed usage (2b) is unlikely to occur unless the child loses metalinguistic awareness about tense.

(2) a. Irregular form: held

b. Mixing irregular and overregularized forms: held and holded

c. Irregular form: held

d. No inflection: hold

The regression hypothesis predicts that a child undergoing attrition will ultimately return to the initial stage of acquisition lacking inflection on the verb stem (2d) even if the child was able to produce the irregular past tense before the onset of attrition (2a). It is therefore important to look at what specific predictions the blocking and competition hypotheses will make regarding the intermediate attrition process (2b). The blocking hypothesis will expect little overregularization during the unlearning process as long as the child is using the irregular form, because the correct form is likely to block overregularization. Only after complete loss of the irregular form, overregularization may appear if the regular –ed form is still present. If overregularized forms co-occur with irregular forms, they are speech errors and the occurrence should be rare. On the contrary, the competition hypothesis can predict a higher rate of overregularization than the blocking hypothesis as the lack of input weakens the irregular form, causing competition between the irregular and overregularized forms which can continue for some time.

This paper aims to probe two general areas of inquiry. First, the course of English irregular past tense attrition is expected to follow the path in (2), as predicted by the regression hypothesis. Second, during the course of attrition, participants’ data is expected to provide evidence supporting one of the irregular past tense acquisition theories: the blocking hypothesis versus the competition hypothesis.

Methods

Participants

Two girls, referred to with pseudonyms, began their participation in a longitudinal project when they were aged 6;9 and 4;11, respectively. Hera and Rita are sisters, and had been living in Honolulu, Hawaii, for two years with their mother before returning to Korea at the beginning of their participation. Both spoke only Korean when they first arrived in Hawaii, but their linguistic dominance switched drastically during their stay; by the time they returned to Korea, they were fluent English speakers with limited proficiency in Korean. They seemed to understand Korean to some degree, but their Korean production was limited to greetings and a few vocabulary items. During the few days before their departure from Hawaii when the first round of the experiment was conducted, they responded only in English to both English and Korean spoken by their mother and the experimenter. Hera and Rita were attending public school and kindergarten, respectively, when they left Hawaii.

Their speech data is analyzed to determine whether their course of attrition provides evidence for the regression hypothesis. It was tacitly assumed that the girls exactly mirrored L1 English children in their acquisition of past tense formation, but the possibility of the children’s course of acquisition diverging from L1 acquisition cannot be ruled out. However, the boundaries of child L2 acquisition are somewhat arbitrary as it can refer to “acquisition by individuals young enough to be within the critical period, but yet with a first language already learned” (Foster-Cohen, 1999) or “successive acquisition of two languages in childhood” (McLaughlin, 1978). What is not included in these definitions is simultaneous acquisition of two languages in childhood, which generally falls under the cover term of bilingualism. The question of what constitutes simultaneous acquisition versus sequential acquisition is quite difficult to answer (Gass, Behney, & Plonsky, 2013). When the two girls in this study arrived in Hawaii, they were age 4;9 and 2;11, respectively, which implies that the sisters, especially the younger one, had not reached native proficiency in their L1 Korean. In addition, they had lost their Korean before returning to Korea and spoke only English. Thus, it becomes difficult to categorize English as their L2. Based on the idea developed by Dulay and Burt (1974a, 1974b, 1975), child L2 acquisition will be treated as similar to child L1 acquisition (L1=L2 hypothesis) in this study.

After their return to Korea, the two girls continued to use English at home, where, after a few hours at school and kindergarten every day, they spent most of their time under the supervision of a monolingual Korean babysitter until their parents came home from work. They had English classes at their schools, but both reported that the classes were too easy. For this reason, two or three times a week after school, the girls were sent to a private institution where they were enrolled in a class with other returnee children from English-dominant countries. When they stopped going to the private institution after a few months, their mother had a native English-speaking tutor come to their home approximately twice a week for two-hour visits to talk and play with them, which lasted up to the second year after their return to Korea. After about two years in Korea, both girls were speaking Korean to their parents, but Hera reported that her interaction with her little sister was still more English-dominant than Korean because she thought Rita was not fully proficient in Korean. However, after another year, they had transformed into fluent Korean speakers. Their mother reported that they preferred using Korean in their daily activities and she was worried about her daughters’ deteriorating English.

Procedure

Hera and Rita completed 30 rounds (each round usually consisted of three sessions) of experiments over a 40-month period, all of which were audio recorded. Originally, this was part of a dissertation project (Kang, 2011) on English attrition in Korean–English bilingual children. Most of the data collection was concentrated in the first year of their return (rounds 1–22) when the experimenter was staying in Korea. However, data in the second (rounds 23–27 in bold in Table 1) and third year (rounds 28–30) were collected less frequently, only during the breaks between academic semesters when the experimenter was able to visit Korea from Hawaii. The girls were tested individually and each session, on average, took approximately 30 minutes. Table 1 shows the timeline of the experiments.

Table 1.

Timeline of experiments (weeks after return).

Round	1	2	3	4	5	6	7	8	9	10	11	12	13	14	15
WaR	-1	4	6	9	13	15	17	19	22	27	29	32	35	37	39
Round	16	17	18	19	20	21	22	23	24	25	26	27	28	29	30
WaR	43	45	47	50	52	55	58	76	78	98	105	109	161	167	174

WaR: weeks after return

Material

Two types of experiments regarding their irregular past tense were conducted and reported in Kang (2011): elicitation tasks and judgment tasks. Here, however, the experiment results are not reported as they were not able to effectively reflect the children’s irregular past tense usage. Xu and Pinker (1995, p. 544) also acknowledged that naturalistic data tend to be a better measure of children’s actual competence in irregular past tense verb usage than experimental data. Therefore, in the current paper, only naturalistic speech data is investigated, which consisted of spontaneous speech data collected from the participants while they were talking casually or taking part in other experiments not related to the irregular past tense verbs.

Data analysis

The recorded data was transcribed and analyzed with an exclusive focus on situations that clearly required the past tense. Following Maratsos (2000), the following verbs were excluded from analysis for the reasons below:

No-stem-change irregular verbs. Irregular verbs like cut or put have the same past and present stems, which makes it difficult to tell whether the child is using the past tense or the uninflected form. This is critical, given that children have initial stages of using uninflected verbs.

Forms of the copula be. The be verb is usually classified either as its own special verb category or an auxiliary verb due to its different behavior from main verbs; it can take negation like auxiliary verbs, such as wasn’t, and moves to the front in questions, which is impossible with main verbs. Maratsos (2000) states that “children may well make similar internal decisions, in which case be-forms would not usually comprise candidates for regular past tensing” (2000, pp. 190–191). This process seems to be true in the data here since the children never make overregularization errors with past tense forms of be, although they may occasionally produce errors with plural agreement such as using was with a plural subject.

Do and have. These verbs sometimes function as main verbs, but can also serve as auxiliaries. Although Maratsos (2000) states that the decision to exclude these is more problematic compared to the exclusion of the above two categories, his more conservative scoring matrix will be followed.

Get. Maratsos (2000) points out that numerous uses of got are actually present tense in everyday colloquial speech. “I got it” often means “I have it,” which makes it difficult to tell whether got is intended as past or present in the context. Moreover, got in the sisters’ speech was largely used as the auxiliary in passive voice as in “He got fed by Donald.” (Hera, round 19).

Results

For ease of presentation in demonstrating the timeline, rounds 1 to 22, 23 to 27, and 28 to 30 are referred to as year 1, year 2, and year 3. Errors were classified into four major categories: (1) overregularization; (2) infinitive without inflection; (3) irregular form + regular –ed (e.g. broked from Rita’s round 10); and (4) non-standard past tense or mis-irregularization forms (see “weird” in Xu & Pinker, 1995) that do not include error type 3 (e.g. brang instead of the target brought frequently used by Hera).

Table 2 summarizes the sisters’ past tense use in both irregular and regular verbs. Their use of regular verbs was also probed in order to observe their proficiency in applying the regular past tense rule. It is notable that the sisters’ total token of past tense use significantly decreased in year 3 compared to the first two years. The main reason for such a decrease (besides the relatively small sample size) is because the sisters remained primarily focused on the tasks in year 3, while during the first two years, their focus deviated greatly in the middle of the tasks and they talked a great deal about their day or other topics stemming from the experimental prompts. Some might argue that the decrease in small talk could reflect their loss of interest or growing boredom toward the repeated experiments, but the fact that their small talk was mostly done in Korean in year 3 suggests that avoiding English use may have been a sign of attrition.

Table 2.

Hera and Rita’s past tense usage data: frequency (%).

		Irregular					Regular
		Correct	Overregularization	Infinitive	Irreg –ed	Non-standard	Correct	Infinitive	Non-standard
Hera	Year 1	710 (96.2)	5 (0.7)	6 (0.8)	2 (0.3)	15 (2)	387 (96.8)	12 (3.1)	1 (0.003)
	Year 2	71 (94.7)	0	0	0	4 (5.3)	53 (100)	0	0
	Year 3	22 (50)	0	0	0	22 (50)	20 (95.2)	1 (4.8)	0
Rita	Year 1	355 (64.2)	161 (29.1)	29 (5.2)	6 (1.1)	2 (0.4)	296 (87.6)	42 (12.4)	0
	Year 2	59 (53.6)	38 (34.5)	11 (10)	1 (0.9)	1 (0.9)	29 (74.4)	10 (25.6)	0
	Year 3	6 (27.3)	7 (31.8)	8 (36.4)	1 (4.5)	0	15 (62.5)	9 (37.5)	0

Hera’s regular verb usage demonstrates that her –ed rule remained intact (above 95%) for all three years. The lone non-standard error in year 1 was sput for the past tense of sprout. Her use of irregular past tense verbs is much more interesting. For the first two years, her accuracy was above 94% with no particular error pattern. However, in year 3, she used 22 tokens of licit irregular forms and 22 tokens of non-standard past tense forms. Surprisingly, all 22 tokens of non-standard forms were brang as the past tense of bring. Hera used brang in the first two years as well; in fact, brang accounted for 17 of the 19 non-standard forms produced in years 1 and 2 (the others were thank for thought and clang for clung), but in year 3 the frequency suddenly skyrocketed considering the significantly smaller sample size. It is clear that although Hera’s irregular past tense accuracy declined significantly during year 3, this result does not definitely indicate overall attrition since all the errors involved one errant form.

Meanwhile, Hera’s little sister, Rita, demonstrated a steady decline in regular past tense accuracy. She started with approximately 88% accuracy during the first year before dropping to around 74% in year 2 and then 63% in year 3. Accordingly, her rate of infinitive errors rose from around 12% in year 1 to nearly 26% in the second year and 38% in the third year. With irregular verbs, her accuracy was approximately 64% and 54% during the first two years before plunging to 27% in year 3. During the first two years, the biggest source of errors was overregularization, but it switched to infinitive errors in the final year.³

As for the course of attrition, Rita’s data demonstrates what is predicted by the regression hypothesis. The order of attrition presented itself as the reverse of acquisition: the infinitive stage (year 3) arrived after the stage where the irregular and overregularized forms had generally coexisted (years 1 and 2).

Discussion

The language use of the older sister, Hera, did not show definite signs of attrition during the three years after her return to Korea. She was able to apply the –ed past tense rule to regular verbs and produce irregular past tense forms correctly, with the exception of the one problematic irregular verb bring. A more detailed investigation of Hera’s use of the past tense of bring provides an interesting aspect of her irregular past tense use. She used two forms of the past tense for bring at least once during each year: the correct form brought and the non-standard form brang⁴. As Table 3 demonstrates, her correct use fluctuated throughout the three years. In year 1, 59% of Hera’s use (19 tokens) were correct and she used brang 13 times (41%). Then, in the second year, her accuracy improved to 85%. Considering that her younger sister Rita, whom she spent most of her English-using time with, did not provide her with the appropriate input she needed (Table 3), it can be speculated that her times at school and with the native speaker tutor positively affected her. However, her drastic drop of accuracy in year 3 (4%), when she no longer had the native speaker tutor but still attended school, suggests that the tutor might have been the main source of correct input responsible for her temporary improvement in year 2. Although Hera was predominantly using the non-standard form in year 3, she used brought once and brang five times within a span of approximately five minutes in round 28 (year 3). All of this evidence suggests that the two forms were in constant competition during the years after her return to Korea; the licit irregular form dominated the competition during the first two years but eventually the illicit non-standard form brang won in the final year (although they were still competing), probably due to a lack of correct input in the attrition process⁵. If this is the case, it follows that year 3 could actually have been the onset of attrition in Hera’s use of the irregular past tense, starting with the victory of brang over brought.

Table 3.

Hera and Rita’s use of the past tense of bring: frequency (%).

	Hera		Rita
	brought	brang	brought	brang	branged	bring	bringed
Year 1	19 (59.4)	13 (40.6)	14 (16.9)	0	0	2 (2.4)	67 (80.7)
Year 2	22 (84.6)	4 (15.4)	2 (10)	1 (5)	1 (5)	3 (15)	13 (65)
Year 3	1 (4.3)	22 (95.7)	0	1 (8.3)	0	6 (50)	5 (41.7)

Unlike her older sister, Rita showed numerous signs of attrition. It must be noted that while Hera can be considered to have completely acquired both regular and irregular forms as her accuracy of both forms reached at least 96% in the first year, Rita’s deviant irregular past tense forms in year 1 indicate incomplete English acquisition. If Rita did not reach native-level fluency at the onset of attrition, her errors could reflect either incomplete acquisition or attrition. It would be necessary to demonstrate that the deviant forms found in the data are the results of attrition rather than incomplete acquisition. First, regarding her regular past tense usage, Rita seemed to acquire the regular –ed rule demonstrating 88% accuracy. In fact, the rule was productive enough to be applied to a Korean word in code switching. The excerpt in (3) shows her ability to inflect regular past tense verbs such as kick and fart in the correct form. A few seconds later, she also applied the same strategy to the Korean noun banggu (gas/fart) to use it as a past tense verb. Thus, her productivity with regular inflection points toward some degree of native-like proficiency. Then, her regular past tense accuracy consistently declining over the next two years (from 88% to 74%, and then to 63%) while infinitive errors increase can be viewed as a result of attrition.

(3) the bear is, is looking at something and, but a friend, uh, he kicked his tail and, and the bear farted … (after four turns) … and did you know that my daddy banggu-ed every time?

(Rita, round 16)

Similarly, Rita’s correct use of the irregular past tense sharply dropped in the third year (27%) after a 10% decline in the second year (54%) from the first year (64%). During the first two years, overregularization had been the biggest source of errors (around 30%) followed by infinitive forms (5% in year 1 and 10% in year 2). However, in the third year, the rate of infinitive errors increased a great deal and occurred more frequently (approximately 36%) than overregularizations (approximately 32%). Thus, the biggest source of errors, which had been overregularization during the first two years, switched to infinitive errors (Table 4). Obviously, Rita never accomplished native-like competence with the irregular past tense, but her language usage patterns demonstrated in Table 4 seem to suggest that her deviant forms in year 3 are still due to attrition rather than incomplete acquisition. Overregularization and infinitive errors which respectively constitute approximately 29% and 5% of her irregular past tense forms in year 1 are clearly attributable to incomplete acquisition. However, there must be another factor reshaping Rita’s interlanguage grammar, resulting in her language usage in the following two years: attrition. While overregularization errors constantly make up around 30% of her usage throughout the three years, deteriorating irregular forms (and regular forms discussed in the previous paragraph) retreat to a more rudimentary grammar: the infinitive form (Francis, 2011).

Table 4.

Rita’s competing irregular past tense forms (%).

	Competition: winner in bold
Year 1	irregular form (64.2)	vs overregularization (29.1)	vs infinitive (5.2)
Year 2	irregular form (53.6)	vs overregularization (34.5)	vs infinitive (10)
Year 3	irregular form (27.3)	vs overregularization (31.8)	vs infinitive (31.8)

As Table 4 demonstrates, during the first two years, Rita’s correct usage and overregularization errors combined occupied at least 88% of her irregular verb past tense usage while infinitive error rates were 10% or less. It appears that the correct irregular form was winning the competition over overregularization for two years – this is the stage that directly precedes reaching native-like proficiency in the acquisition process. Then, in year 3, accuracy fell to 27% while the overregularization rate remained around 32%. Meanwhile, although overregularization exceeded use of the correct form, the rate of infinitive error jumped from under 10% to 36%, becoming the winner of the overall competition. It was mainly a two-way competition during the first two years because Rita’s proficiency was past the infinitive level then. However, by the third year, the situation had transformed into a three-way competition because Rita had reverted to the infinitive stage as attrition struck. More accurately put, it seems that all possible past tense forms in the child’s mind were competing every year. Even with regular verbs, Rita’s language use provides an example of the regular past tense form and the infinitive form competing to describe an incident that happened one weekend at her grandmother’s house.

(4) but I wanted to take pictures inside the…, I want to take my picture in the living room but my grandma just put it away and I wanted to do that! (Rita, round 11)

Rita’s interlanguage grammar reflects the prediction of the regression hypothesis in (2) above. At the beginning, Rita was headed toward native-like proficiency with a more than 35% gap between the correct usage (64.2%) and overregularization (29.1%): in-between (2a) and (2b). Then, the gap halved to about 19% in year 2. After the fierce competition in (2b), overregularization finally overtook the correct form in year 3, while the infinitive form became the majority (2d).

Rita’s pattern of irregular past tense forms over the three years cannot be accounted for by the blocking hypothesis for at least two reasons. First, errors such as overregularization or infinitive forms are occurring too frequently. Blocking predicted that the existence of the correct irregular form would ban other past tense forms from being produced and that errors would be the result of speech errors, and should therefore not be frequent. Second, the period of various past tense forms being produced is too long (three years). If blocking were at work, it is strange that it would be unable to perform appropriately for such a long time. If blocking had failed, we should at least be able to observe a stage in Hera and Rita’s language use where all correct forms were substituted by other incorrect forms. However, the correct irregular form was far from extinct in year 3.

Therefore, it appears that the competition hypothesis is able to account for the sisters’ data. All possible past forms in the children’s minds were constantly competing at each stage. At the beginning of the project, the older sister, Hera, was a highly proficient English speaker. However, competition was underway, although it was dominated by the licit form, which is why occasional errors such as overregularization, infinitives, or non-standard forms (see Table 2) are observed. This stage may be accounted for by the blocking hypothesis as well, since the correct form seems to be blocking (very few) illicit forms. However, the non-standard form brang dominated over the licit form brought as Hera reached the end of the project, which cannot be explained by the blocking hypothesis because blocking should inhibit the simultaneous appearance of licit and illicit forms. Although Hera may never resort to bare infinitives thanks to her general metalinguistic awareness level, competing past tense forms of bring might have signaled the beginning of the changing state of her interlanguage. Her younger sister, Rita, was a less proficient English speaker, which is why the competition during the first two years was fiercer than in Hera’s case (Table 4). When attrition eventually affected the competition in year 3, the illicit infinitive form outraced all other forms, including the correct one.

It therefore follows that each stage of attrition (and acquisition) is controlled by the dominant form that wins the competition. In acquisition, the first step is to use infinitives without any inflectional morpheme. In the second stage, where children use the correct irregular form, there is still likely to be no competition if we assume their correct usage is a result of mere imitation rather than a true understanding of the past tense form. Once they acquire the regular past tense morpheme, competition begins. Moreover, the situation gets more competitive after the children come to understand the irregular past tense forms until they settle the competition, reaching native proficiency. The irregular past tense forms in the acquisition process could be accounted for by either the blocking or competition hypothesis because the acquisition process passes by very quickly, leaving little room for researchers to make a judgment. However, in the attrition process, which was in this case not over even in the third year, opportunity to witness the unlearning process in slow motion was provided.

(5) Among the competition of irregular vs overregularization vs infinitive vs non-standard forms during the attrition process…

a. Irregular form: held (irregular form is the winner)

b. Mixing irregular and overregularized forms: held and holded (irregular andoverregularization forms take turns being the winner)

c. Irregular form: held

d. No inflection: hold (infinitive is the winner)

Thus, if the attrition process predicted by the regression hypothesis in (2) is specified here as (5), we can easily understand why other forms are not difficult to find, even when one form is dominant at each stage. In fact, if the non-standard form wins the competition, the result is something akin to Hera’s year 3, when there are more tokens of non-standard form brang than any other forms for the past tense of bring. Although the stage where overregularization becomes the winner of the competition was not documented here, it may have appeared at some time between year 2 and 3 in Rita’s data.

Conclusion

Just as language acquisition is known to have developmental stages, attrition seems to follow its own path. The course of two Korean girls losing English irregular past tense forms revealed several interesting findings regarding the course of attrition and hypotheses of irregular past tense acquisition. As predicted by the regression hypothesis, the course of irregular past tense attrition followed the reversed order of irregular past tense acquisition. Then, the prolonged attrition period provided evidence for the competition hypothesis (rather than the blocking hypothesis) of irregular past tense acquisition, which was difficult to find during the acquisition process, which happens quickly; various past tense forms in competition can produce a different winner over different periods of time. This research also provides rare empirical evidence to refute the blocking hypothesis. Although results from a single longitudinal study with only two children may not be able to provide the full set of data necessary to answer controversial questions related to the issue of language acquisition and attrition, this study makes an invaluable contribution for the development of language attrition research.

Footnotes

Funding

This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

Notes

Author biography

Sang-Gu Kang received his PhD in Linguistics from the University of Hawaii at Manoa in 2011. He is currently a Professor in the Department of English Language and Literature at Cheongju University, Korea. His research focuses on English attrition in Korean children and morpho-syntactic aspects of L2 acquisition.

References

Anderson

(2001). Lexical morphology and verb use in child first language loss: A preliminary case study. International Journal of Bilingualism, 5, 377–401.

Bardovi-Harlig

Stringer

(2010). Variables in second language attrition. Studies in Second Language Acquisition, 32, 1–45.

Berko

(1958). The child’s learning of English morphology. Word, 14, 150–177.

Berman

Olshtain

(1983). Features of first language transfer in second language attrition. Applied Linguistics, 4(3), 222–234.

Brown

(1973). A first language: The early stages. Cambridge, MA: Harvard University Press.

Cohen

(1975). Forgetting a second language. Language Learning, 25, 127–138.

Dulay

Burt

(1974a). Natural sequences in child second language acquisition. Language Learning, 24, 37–53.

Dulay

Burt

(1974b). You can’t learn without goofing. In Richards

(Ed.), Error analysis: Perspectives on second language acquisition (pp. 95–123). London, UK: Longman.

Dulay

Burt

(1975). Creative construction in second language learning and teaching. In Burt

Dulay

(Eds), On TESOL ’75: New directions in second language learning, teaching and bilingual education (pp. 21–32). Washington, DC: Teachers of English to Speakers of Other Languages.

10.

Ecke

(2004). Language attrition and theories of forgetting: A cross-disciplinary review. International Journal of Bilingualism, 8(3), 321–354.

11.

Foster-Cohen

(1999). An introduction to child language development. London, UK: Longman.

12.

Francis

(2011). Imbalances in bilingual development: A key to understanding the faculty of language. Language Sciences, 33, 76–89.

13.

Gass

Behney

Plonsky

(2013). Second language acquisition. New York, NY: Routledge.

14.

Gürel

(2004). Selectivity in L2-induced L1 attrition: A psycholinguistic account. Journal of Neurolinguistics, 17, 53–78.

15.

Jakobson

(1941). Kindersprache, Ahpasie und allgemeine Lautgesetze. Uppsala, Sweden: Almqvist and Wiksell.

16.

Kang

S.-G.

(2011). English attrition in Korean-English bilingual children. Unpublished doctoral dissertation, University of Hawaii at Manoa, USA.

17.

Kang

S.-G.

(2012). Stages in the attrition of English passives. Korean Journal of Linguistics, 37(4), 783–805.

18.

Kang

S.-G.

(2013). The role of language dominance in cross-linguistic syntactic influence: A Korean child’s use of null subjects in attriting English. Bilingualism: Language and Cognition, 16(1), 219–30.

19.

Kang

S.-G.

Lee

(2013). Lexical diversity and functional categories in English attrition of Korean returnee children. Korean Journal of Linguistics, 38(2), 239–58.

20.

Kang

S.-G.

(2014). Fluency and accuracy in English attrition: A case study of two Korean sisters. Language and Linguistics, 62, 25–56.

21.

Keijzer

(2010). The regression hypothesis as a framework for first language attrition. Bilingualism: Language and Cognition, 13, 9–18.

22.

Kuhberg

(1992). Longitudinal L2-attrition versus L2-acquisition in three Turkish children – empirical findings. Second Language Research, 8(2), 138–154.

23.

Lardiere

(1998). Case and tense in the ‘fossilized’ steady state. Second Language Research, 14, 1–26.

24.

Maratsos

(2000). More overregularizations after all: New data and discussion on Marcus, Pinker, Ullman, Hollander, Rosen & Xu. Journal of Child Language, 27, 183–212.

25.

Marcus

Pinker

Ullman

Hollander

Rosen

(1992). Overregularization in language acquisition. Monographs of the Society for Research in Child Development, 57, 1–178.

26.

McClelland

Patterson

(2002a). ‘Words or rules’ cannot exploit the regularity in exceptions. Trends in Cognitive Sciences, 6, 464–465.

27.

McClelland

Patterson

(2002b). Rules or connections in past-tense inflections: What does the evidence rule out? Trends in Cognitive Sciences, 6, 465–472.

28.

McLaughlin

(1978). Second-language acquisition in childhood. Hillsdale, NJ: Lawrence Erlbaum Associates.

29.

Olshtain

(1989). Is second language attrition the reversal of second language acquisition? Studies in Second Language Acquisition, 11, 151–166.

30.

Pica

(1983). Methods of morpheme quantification. Studies in Second Language Acquisition, 6, 69–78.

31.

Pinker

Ullman

(2002). The past and future of the past tense. Trends in Cognitive Sciences, 6, 456–463.

32.

Prévost

White

(2000). Missing surface inflection or impairment in second language acquisition? Evidence from tense and agreement. Second Language Research, 16, 103–134.

33.

Reetz-Kurashige

(1999). Japanese returnees’ retention of English-speaking skills: Changes in verb usage over time. In Hansen

(Ed.), Second language attrition in Japanese contexts (pp. 21–58). New York, NY: Oxford University Press.

34.

Seliger

(1991). Language attrition, reduced redundancy, and creativity. In Seliger

Vago

(Eds), First language attrition (pp. 227–240). Cambridge, UK: Cambridge University Press.

35.

Sorace

(2005). Selective optionality in language development. In Cornips

Corrigan

(Eds.), Syntax and variation: Reconciling the biological and the social (pp. 46–111). Amsterdam, Netherlands: John Benjamins.

36.

Tomiyama

(1999). The first stage of second language attrition: A case study of a Japanese returnee. In Hansen

(Ed.), Second language attrition in Japanese contexts (pp. 59–79). New York, NY: Oxford University Press.

37.

Tsimpli

Sorace

Heycock

Filiaci

(2004). First language attrition and syntactic subjects: A study of Greek and Italian near-native speakers of English. International Journal of Bilingualism, 8(3), 257–277.

38.

Ullman

(1999). Acceptability ratings of regular and irregular past-tense forms: Evidence for a dual-system model of language from word frequency and phonological neighboring effects. Language and Cognitive Processes, 14, 47–67.

39.

Pinker

(1995). Weird past tense forms. Journal of Child Language, 22, 531–556.

40.

Yoshitomi

(1999). On the loss of English as a second language by Japanese returnee children. In Hansen

(Ed.), Second language attrition in Japanese contexts (pp. 80–111). New York, NY: Oxford University Press.