Journal of Linguistics and Language Teaching (JLLT)

Journal of Linguistics and Language Teaching

Volume 12 (2021) Issue 2, pp. 193-217

Omission or Generalization?

Subject-Verb Agreement among Young Norwegian Learners of English¹

Kristin Killie (Tromsø, Norway)

Abstract (English)

This paper investigates affixal subject-verb agreement errors among young Norwegian learners of English. The data are taken from the Corpus of Young Learner Language (CORYL), which consists of texts written by Norwegian learners aged 12–13 and 15–16 years (7th and 10th grade). The figures suggest that some Norwegian learners start out using the base form of the verb as the only present-tense form, then start to overuse the present tense suffix -s. The shift must have started before the age of 12–13 years old but appears to continue after this age. Another topic is to what extent the assignment of agreement features by Norwegian learners may be disrupted by factors such as subject complexity and non-contiguity between the subject head and the agreement-requiring verb. The conclusion is that this may well be the case in a number of instances but that quantitative data in this area may conceal individual learner strategies. It is argued that to properly understand learners’ problems with subject-verb agreement, we need data from numerous longitudinal case studies.

Keywords: Subject-verb agreement, 3^rd person singular -s, English, omission, overgeneralization, Norwegian learners of English

Abstract (Norsk)

Artikkelen undersøker samsvarsbøyning mellom subjekt og verbal blant innlærere av engelsk med norsk språkbakgrunn. Fokuset er på hvorvidt disse innlærerne bruker grunnformen (Ainfinitiv) eller 3. person entalls -s’en som generell presensform. Dataene er tatt fra the Corpus of Young Learner Language (CORYL), som består av tekster skrevet av norske innlærere i alderen 12–13 og 15–16 år (7. og 10. skoletrinn). Tallene tyder på at en del norske innlærere først generaliserer grunnformen av verbet i presens, for så å gå over til å generalisere -s-formen. Overgeneraliseringa av -s ser ut til å begynne før 12–13-års alderen for en del språkinnlærere, men den er tilsynelatende vanligere blant 15–16-åringene. Et annet tema er om subjekt-verb-kongruens påvirkes av faktorer som subjektets kompleksitet og avstanden mellom subjektets kjerne og det bøyde verbet. Dette kan være tilfellet for en del innlærere, men datagrunnlaget er for svakt til å konkludere på dette punktet. Det pekes imidlertid på at data fra kvantitative tverrsnittsstudier kan hindre individuelle læringsstrategier i å komme til syne, og at en for å virkelig forstå innlæreres problemer med samsvarbøyning, trenger longitudinelle studier av et stort antall språkinnlærere.

Søkeord: subjekt-verb-kongruens, 3. person, engelsk, overgeneraliseringa,  innlærere av engelsk med norsk språkbakgrunn

1 Introduction

Of all aspects of English grammar, subject-verb agreement is one of the most challenging points for second language learners (Slabakova 2016). This is also a topic which is highly complex and which we do not know enough about to be able to help learners properly. Hence, more research is required in this area.

The present paper focuses on affixal agreement among young Norwegian learners of English, i.e. it discusses the use and non-use of the 3^rd person singular suffix -s, as in like vs likes. According to White (2003: 183), learners of English typically omit the third person singular suffix -s and do not start using it until they have gained mastery over it (Prévost & White 2000). This claim seems to gain support from studies involving learners of various language backgrounds, such as Dutch and French learners (Housen 2002) and Russian learners (Ionin & Wexler (2002). However, in recent years, White’s claim has been challenged by research which demonstrates that overgeneralization errors are frequent in learner English. Some of this research is based on data from young Norwegian learners of English. Thus, the 199 Norwegian learners of English (15–16 years old) studied by Garshol (2018, 2019) make more overgeneralization errors than omission errors in their written texts. Grammaticality judgements made by Norwegian 12-19-year-olds also reveal a more pronounced tendency to accept overgeneralization errors as compared to omission errors (Jensen 2016, Jensen, Slabakova & Westergaard 2019, Jensen, Slabakova, Westergaard & Sundquist 2019).

This study addresses the following research questions:

Research question 1: Do Norwegian learners of English commonly change their general affixal agreement strategy, from omission to generalization of -s?

Research question 2: If Norwegian learners of English commonly change their general affixal agreement strategy, from omission to generalization of -s, when does this change occur?

Research question 3: Does learners’ affixal agreement strategy vary according to subject category?

All these questions concern the use of omission or generalization of -s as a general learner strategy.² Research Question 3 was formulated to shed light on the change from an omission to a generalization strategy, the assumption being that generalization may typically start with one type of subject, then spread to another.

A final topic of interest is the role played by subject complexity and by the non-contiguity of the subject head and the agreement-requiring verb. It has been argued that such factors may disrupt agreement assignment (e.g. Garshol 2018, Killie 2019, Son 2020). To shed light on this question, the present study also addresses the following research question:

Research question 4: Do non-contiguity and subject complexity commonly disrupt affixal agreement assignment?

These research questions will be addressed, based on data from the Corpus of Young Learner Language (CORYL), which consists of texts written by Norwegian learners aged 12-13 and 15-16 years (7th and 10th grade).

The paper is organized as follows: Section 2 discusses relevant research on the acquisition of subject-verb agreement and provides information on methodology and data. In Section 3, the data are presented, while Section 4 offers a summary and a general discussion of the data and their implications, along with some suggestions for future research.

2 Theoretical Background

2.1 Subject-Verb Agreement: What is to be learnt?

Subject-verb agreement is an aspect of English grammar which is known to be a learning challenge for Norwegian learners (Garshol 2018, 2019, Hauge & Angelsen 2018,; Killie 2019, 2021). Norwegian does not have subject-verb agreement; hence, Norwegian learners of English need to learn not only the various person- and number-marked verb forms in English and how to match them with various types of subjects. They also have to grasp the very notion of person and number marking on verbs.

However, it is not only Norwegian learners of English who struggle with this area of grammar. Subject-verb agreement represents a cross-linguistic learning challenge, for both first language learners and for second language learners of all levels. According to Slabakova (2016: 397), subject-verb agreement, along with past tense morphology, represents the most difficult morphological aspect of English to acquire, because it is part of the complex area of functional morphology. What makes functional morphology so complicated is that it involves a whole range of grammatical meanings which are distributed differently across the morphemes of the world’s languages. Several features may be bundled together in one and the same morpheme, but the composition of the bundles varies from language to language (Lardiere 2009, Slabakova 2013, 2016, 2019). According to Lardiere (2009), the acquisition of functional morphology therefore involves the selection, composition, and restructuring of features. As for subject-verb agreement morphology in English, the relevant morphemes encode tense / finitude, aspect, mood and person and number features. The Norwegian morphemes which correspond most closely with the agreement-marked English morphemes do not carry person and number features. Thus, there is only a partial semantic overlap between the relevant morphemes in the L1 and L2 even though they occur in the same positions. Norwegian learners of English must therefore undertake ‘feature reassembly’, which is an extremely challenging task (Lardiere 2009).

In addition to understanding the concept of subject-verb agreement, learners must be able to match the various number and person-marked forms with the correct number and person values. With respect to English, this mapping task not only involves learning the difference between the base form and the -s form of present-tense verbs (affixal agreement), which is the topic under discussion in this paper; it also involves matching the irregular present and past-tense forms of be with the correct number and person values onto irregular forms (suppletive agreement).

The choice of the correct verb form further necessitates knowledge of various types of nouns and subjects, such as uncountable nouns, plural-only nouns, collective nouns or singular nouns that end in -s. (Fisher 1985: 9-12, Quirk et al. 1985: 755-766, Hasselgård, Lysvåg & Johansson 2012: 265-274). In addition to general knowledge of these categories, each noun must be assigned to the correct category. In some cases, this is challenging as the category membership of lexemes with the same meaning may vary from language to language. For example, the Norwegian words for news, advice, and furniture (viz. nyhet, råd, and møbel) are not uncountable, as in English, but countable.

Correct agreement marking also requires knowledge about the number (singular or plural) and person properties of some other types of subjects, for example NPs which contain quantifiers (e.g. a large number of, majority, and percent), clauses, nominalized adjectives, relative pronouns or indefinite pronouns.

The points discussed above involve knowledge about the number of verb forms, nouns, and NPs. However, for learners to select the correct agreement-marked verb form, it is also necessary to understand that it is normally the head of the subject NP which agrees with the verb. It requires analytical skills to find the head of the subject NP. In addition, it takes the awareness that in some cases, the structure of the subject is irrelevant because it is the meaning of the subject NP as a whole that determines the form of the verb (‘notional’ agreement/concord, e.g. Brehm & Bock 2013, Hasselgård et al., 2012: 273-274). It is also necessary to know about the phenomenon referred to variously as agreement attraction (Strang 1966), attraction (through proximity) (e.g. Quirk et al. 1985: 757), and proximity agreement (Biber, Johansson, Leech, Conrad & Finegan 1999), i.e. cases in which the verb is made to agree with a noun in the post-modifier rather than with the head of the subject NP. While in most cases, agreement attraction leads to ungrammaticality, there are cases in which it does not (Killie 2019).

Finally, in their leisure time, Norwegian children and youths are exposed to massive non-standard English input which frequently differs from Standard English in regard to subject-verb agreement. A case in point is African American Vernacular English, which, as we assume, figures prominently in the extra-mural input of Norwegian youths, given the ubiquity of the popular culture of its speakers. In this variety, there is a very strong tendency not to use the 3^rd person singular -s (Wolfram 2004: 122). It is not surprising if Norwegian adolescents become confused when the input they receive in and outside of school is so conflicting.

In sum, a learner who wants to gain full mastery of subject-verb agreement, needs to acquire a large and complex body of general and detailed knowledge. Because of this complexity, subject-verb agreement requires attention in the L2 classroom. Obviously, focusing on all the aspects discussed above is neither wise nor possible. According to Slabakova (2016: 403), classroom activities need to be “smart” in order to help learners, i.e. we need to focus on the points where learners need help to progress rather than to spend time and energy on aspects that will fall into place by themselves. We should therefore seek to uncover those factors that cause the biggest problems in the acquisition of subject-verb agreement and make these the focus of instruction.

2.2 Relevant Research on the Acquisition of Subject-Verb Agreement

Previous research has shown that error rates are much higher when the head of the subject noun phrase is a noun than when it is a personal pronoun (Fisher 1985, Garshol 2019, Killie 2019, Slabakova & Gajdos 2008, Vaurula 2012). Killie (2019) explains this phenomenon with regards to the drilling of pronoun-verb sequences (with be, have and do) and the much higher frequencies of personal pronoun subjects as compared to noun subjects. Based on a corpus of texts from seventeen Norwegian students, Son (2020) argues that the difference in agreement accuracy with nominal and pronominal heads has become all but neutralized when Norwegian learners reach adulthood.

Another factor which is assumed to negatively affect accuracy is the non-contiguity of the head of the subject noun phrase and the verb requiring agreement marking. Son (2020) reports much lower accuracies in non-contiguous contexts than in contiguous ones (also Fisher 1985 and Garshol 2018). In (1)-(4) below (Son 2020: 49), non-contiguity is the result of subject complexity in the examples (1)-(3) and of VP complexity in example (4)³:

NP with phrasal (prepositional) post-modifiers:

(1a) The suitable age for the pupils are .. (Dec18_1_PM2)

(1b) The competence aims after year 10 fits.... (Dec18_1_PM1)

(1c) The reason for such a high age for these books are.. (Dec18_17_PM_9 )

NPs with clausal post-modifiers:

(2a) The groups that stopped at every new word is.. (Dec18_6_Clause_long3)

(2b) The characters I will be focusing on is.. (Mar19_16_Clause_long9)

Noun head of a relative clause:

(3a) the pupils who struggles.. (Dec18_7_Clause_short5)

(3b) topics that does not depend on.. (Dec18_12_Clause_short9)

VP coordination:

(4a) The teacher tells the story and not just read.... (Dec18_1_CorV1)

(4b) b. Her lines are without rhythm and rhymes and does not.. (Mar19_6_CorV10)

Garshol (2018) argues that also intervening adverbials may lead to incorrect agreement (also Quirk et al. 1985: 757). The examples in (5) are taken from Garshol (2018: 52):

(5a) The Democrats, on the other hand, wants to draw a much softer line. (14STA13_Apr15)

(5b) Most families actually grows on these disagreements. (15STV52_Oct15)

The cases that have been most discussed in the literature are post-modified subjects such as those in (1) and (2). There is general agreement that in such cases, accuracy is lower than it is with simplex subject NPs, even among advanced learners (Garshol 2018,; Jensen 2016, Jensen et al. 2017, 2019, Ocampo 2013, Son 2020, Strang 1966). It is also a widespread view that errors with this type of subjects are caused by agreement attraction (Biber et al. 1999). However, in recent years this view has been challenged by evidence that learners make errors in such contexts even when there is no number conflict between the head of the subject NP and the noun placed closest to the verb. Garshol (2018: 52) even concludes that among her 15-16-year-old Norwegian pupils, “proximity agreement can be involved only in marginal cases”. Instead, she explains errors in non-contiguous contexts in terms of the “general cognitive load” (ibid.) that such cases represent, arguing that learners “might get distracted by the intervening elements and not assign the number of the verb correctly” (Garshol 2018: 53). Fisher explains the relevant cognitive load as follows:

In contiguous as well as noncontiguous S-V constructions, the subject must be stored in short-term memory; retrieval is obviously facilitated when the verb occurs immediately after the head of the subject noun phrase, since disturbing factors such as the limited capacity of short-term memory and attraction from intrusive elements do not obtain. (Fisher 1985: 11)

The fact that processing difficulties may be a relevant factor is supported by the evidence that even adult native speakers make a considerable number of agreement errors in such contexts (Bock & Miller 1991, Eberhard, Cutting & Bock 2005, Fisher 1985, Hartsuiker & Barkhuysen 2006, Strang 1966, Levin 2001; Killie 2019).⁴ However, we cannot conclude from this that learners’ agreement errors in such contexts are always directly comparable to those made by adult native speakers and hence are caused exclusively by processing difficulties. A lack of competence may play a major role for many learners. Clahsen, Fehlser, Neubauer, Sato & Silva (2010: 23) argue that the syntactic representations of adult L2 learners are generally less detailed than those of native speakers and that consequently “the L2 grammar does not provide the kind of information required to process complex syntax in native-like ways” (for a discussion, Killie 2019). Nevertheless, it is likely that not only language competence but also language processing may be a challenge for learners with post-modified subject NPs. A similar cognitive load presumably exists in cases of noun head of a relative clause (as in example 3 above), VP coordination (example 4), and intervening adverbials (example 5).

It has further been shown that Norwegian learners run into problems when the subject consists of two or more conjoined NPs (Killie 2019, Son 2020), as shown in (6) and (7) below (from Killie 2019):

(6) John, Jack and Adam is building a house in a tree in Jacks garden. (CORYL, Primary Verb Corpus, 12–13-year-old learner)

(7) But my littlebrother and his friend doesn’t see it. (CORYL, Primary Verb Corpus, 12–13-year-old learner)

In Killie’s (2019) corpus, which consists of all agreement-requiring primary verbs (be, have, and do), accuracy with simplex noun subjects is 80.8% for 12-13-year-olds and 86.2% for 15-16-year-olds, whereas the corresponding figures for conjoined NP subjects are 51.3% and 63.9%, respectively. The difference in accuracy between the two types of subjects is rather similar among Norwegian university students. Thus, Son (2020) finds that out of ten different conditions, subject coordination is the most challenging one, with an accuracy of 63.6% with all verbs and 66.7% with primary verbs. By comparison, simplex noun subjects have an accuracy of 93.3% for all verbs and 96.4% with primary verbs, respectively. Killie (2019) argues that errors with this type of subject occur due to agreement attraction with the second NP conjoint.

Another common problem in subject-verb agreement is insufficient acquisition of the agreement-marked verb forms. Learners who have not mastered the various forms may resort to generalization strategies. In regard to suppletive agreement, Garshol (2018, 2019) and Killie (2021) show that a number of Norwegian learners generalize the form are in the present tense, presumably because of its phonetic and graphemic similarity with Norwegian er /æ:r/, which is the only present-tense form of the verb være (‘be’) that is used in the populous southeastern parts of the country. However, the biggest mapping challenge for Norwegian learners with the verb be is to map the correct person and number features onto the two past-tense forms was and were. Even 15-16 year-olds commonly struggle with this task, and the generalization of one of the two forms appears to be a frequent strategy (Garshol 2018, 2019, Killie 2021).

The present paper focuses on two other possible generalization strategies, which may potentially be resorted to in affixal agreement, viz. the generalization of the third person singular suffix -s, and the omission of the relevant suffix. The question is to what extent the two strategies are used by Norwegian learners aged 1213 and 15-16-years old. As explained above, White (2003: 183) asserts that omission is the most common strategy among learners of all language backgrounds, and that the suffix -s is generally not used by learners until they have gained mastery over it (also Prévost & White 2000). This claim seems to gain support from several studies, such as Housen’s (2002) spoken data from young Dutch and French learners and Ionin & Wexler’s (2002) spoken data from young Russian learners. By contrast, Garshol (2018: 77) found that these learners produced slightly more overgeneralization (51.85% - 983 tokens) than omission errors (48.15% - 913 tokens). Similarly, Jensen (2016), who studied the grammaticality judgements of sixty Norwegian learners aged 12-19, found that they were more prone to accept overgeneralization errors than omission errors (also Jensen et al. 2017 and 2019). In her corpus of texts produced by seventeen Norwegian university students, Son (2020) found that omission errors were slightly more frequent than overgeneralization errors (65.2% vs. 43.8%). However, she argues that

“[t]he higher number of omission errors might have been influenced by a few texts where there is an unusually high number of errors with proper nouns (e.g., Shakespeare express) which seem to be idiosyncratic errors produced by only a few students”. (Son 2020:58)

After the removal of the relevant tokens, overgeneralization errors were in a slight majority, with 51.9% as opposed to 48.1% omission errors. The number of omission and overgeneralization errors in Son’s (2020) corpus is rather low (in total 96 before the removal of the above-mentioned examples, and 81 after. Nevertheless, the relevant figures show that overgeneralization errors are common among Norwegian university students. In sum, these studies suggest that Norwegian learners of English use the suffix -s quite extensively without mastering it properly. There is some evidence that this is also the case with learners whose mother tongues typologically are very different from Norwegian, such as Japanese learners (Shibuya & Wakabayashi 2008) and Finnish learners (Vaurula 2012).

Attempts have also been made to study omission and overgeneralization errors across different subject categories. Garshol (2018, 2019) found that overgeneralization errors represented the majority among Norwegian 15-16-year-olds only when the head of the subject NP was a noun, as in The boys smiles. Similar evidence is provided for Swedish learners by Fisher (1985), for Finnish learners by Vaurula (2012), and for Japanese learners by Shibuya & Wakabayashi (2008). Fisher (1985: 45) refers to this as “perseveration” of the -s, arguing that the similarity of forms has caused these learners to establish a link between the present-tense suffix -s and the plural suffix -s on nouns, so that -s is attached to both verbs and nouns when the subject is a plural noun.

The data on Norwegian and Swedish learners will be studied more closely in Section 3, in which the CORYL data on omission vs overgeneralization errors are presented.

2.3 Database and Methodology

The data of the present study are drawn from the Corpus of Young Learner Language (CORYL). This corpus was compiled by the Research Group for Language Testing and Assessment at the University of Bergen (Norway) and consists of 129,421 words of learner language taken from the National Tests of English (writing) in the years 2004 and 2005. The texts were drawn randomly from pupils in 7^th and 10^th grade across Norway and contain samples from 272 learners, 136 from each of the two age groups 12-13 years (7^th grade) and 15-16 years (10^th grade). The corpus was coded for error type, gender, age group and proficiency level in relation to the Common European Framework of Reference for Languages (CEFR). (Hasselgren & Sundet 2017 for more detailed information.)

All sentences which are tagged for errors in subject-verb agreement so as to test the hypothesis that learners tend to make omission errors. Two searches were carried out, one for the age group of 12-13 year olds and one for the age group of 15-16 year olds. Clauses in which the finite verb was a form of be were removed from the data, given the suppletive morphology of this verb, which makes the terms omission and overgeneralization inapplicable. The terms would have been appropriate if learners in CORYL had used the forms be and bes as agreement-marked forms, but this is not the case.⁵

Collective nouns are nouns referring to a group of people, such as government, committee, and team. When the head of the subject NP is a collective noun in the singular, some varieties of English, and in particular British English, allow a plural verb (Levin 2001, 2006). CORYL contains some examples of subjects whose head is a collective noun in the singular. In such cases, the use of a plural verb is coded as an error:

8. … the fact that this band do everything for there fans (CORYL, 15–16-year-old)

9. It’s a football feeld where the best football team play. (CORYL, 15–16-year-old)

Some examples are given in (8) and (9) above. Examples like these have been excluded from the present study as we cannot know whether the use of a plural verb is an error, or whether the relevant pupils actually knew that plural verbs are allowed in such cases.

In this context, it needs to be mentioned that the removal of be from the data dramatically reduced the number of corpus tokens. 342 out of 473 agreement errors made by the 12-13-year-olds (i.e. 72.3 %) involve be, while the corresponding number for the 15-16-year-olds is 157 out of 334 errors (47.0 %). With all these examples, as well as the examples involving collective nouns, removed from the data, the corpus stands at 308 tokens, of which 131 were produced by 12-13-year-olds and 177 by 15-16-year-olds. Evidently, this is not a very large number of tokens, and it naturally cannot serve as the basis for any definitive conclusions. The claims proposed on the basis of the CORYL data are, however, supported by other studies, as we will see below.

The quantitative data are complemented by a closer look at the total language production of a few individual learners (which can be accessed through the “Show context” option in the concordance). An advantage of looking more closely at the production of individual learners is that this additional step may help reveal individual problems and strategies, and combinations of these. Thus, one learner may consistently mark present-tense verb forms by the suffix -s, another may consistently omit the suffix -s while for a third learner, strategies may vary depending on the linguistic context. Such individual problems and strategies are not visible in aggregate data, but they are well worth looking into.

3 Presentation of the Data

In this section, the CORYL data are presented and discussed in relation to the research questions. The data are also compared to those on Scandinavian learners presented in section 2.2. Section 3.1 addresses the Research Questions 1 to 3, which concern the use of generalization strategies. Section 3.2 discusses factors which possibly disrupt agreement assignment, before, in Section 3, it is argued that aggregate data on disruptive factors may, to some extent, obscure learner problems and strategies.

3.1 Generalization Strategies of Norwegian Learners

In the context of the research questions 1 and 2, Table 1 and Figure 1 show the number and proportion of omission and overgeneralization errors in the CORYL corpus:

	12–13-year-olds	15–16-year-olds
Omission	89 (67.9%)	78 (44.1%)
Overgeneralization	42 (32.1%)	99 (55.9%)
Total	131	177

Table 1: Omission and Overgeneralization Errors in the CORYL Corpus

Figure 1: Omission and Overgeneralization Errors in the CORYL Corpus

The figures show an interesting contrast between the 12-13-year-olds and the 15-16-year-olds: while more than two thirds of the errors made by the former group involve omission, well over half of the errors produced by the latter group involve overgeneralization. Thus, White’s (2003: 183) hypothesis that omission is the primary strategy among learners of English seems to also hold for these Norwegian learners, but only for the younger group, and Garshol’s (2018: 77) claim that overgeneralization errors are more common than omission errors among 15-16-year-olds is also supported. The figures suggest that for a number of Norwegian learners of English, there is a shift in strategy from omission to generalization between grades 7 and 10. The data also indicate that for many learners, this shift starts before the age of 12 or 13. The number of tokens is rather low; however, the study on omission and overgeneralization errors presented in Section 2.2 supports this scenario.

With regards to our third research question, we will first study relevant data presented by Garshol (2018, 2019). Table 2 below presents her data on omission and overgeneralization errors broken down across different subject types:

	PersPro	IndefPro	DemPro	Expletive	Clause	Noun	Total
Omission	307 (66.3%)	71 (81.6%)	27 (67.5%)	6 (37.5%)	19 (100%)	483 (38%)	913 (48.2%)
Overgeneral- ization	156 (33.7%)	16 (18.4%)	13 (32.5%)	10 (62.5%)	0	788 (62%)	983 (51.8%)
	463	87	40	16	19	1271	1896

Table 2: Distribution of Affixal Agreement Error Types across Subject Types (adapted from Garshol 2018: 49, her Table 7; percentages added)

As can be seen, Garshol finds that her learners primarily generalize the third person singular -s with noun subjects: about 60% of the errors with this type of subject are overgeneralization errors.⁶ For the other subject types, omission errors are in a clear majority.

Given the common ancestry of Norwegian and Swedish, Garshol (2018: 79) compares her data with Fisher’s (1985) written data from Swedish university students of English. The relevant figures are presented in Table 3, in which, due to Fisher’s classification strategies, noun subjects are not compared with personal pronoun subjects but with pronominal subjects as a group (This is not quite clear). Only lexical verbs are included:

	Number of Errors – Swedish Learners		Number of Errors – Norwegian Learners
	Nominal head	Pronominal head	Nominal head	Pronominal head
Overgenerali- zation	82 (52.90%)	9 (11.84%)	788 (62.00%)	195 (32.18%
Omission	73 (47.10%)	67 (88.16%)	483 (38.00%	411 (67.82%)
Total	155 (100%)	76 (100%	1,271 (100%)	606 (100%)

Table 3: Summary of Error Counts from Swedish and Norwegian Data Divided by Subject Type

The above figures show that the Swedish students, just like the learners whose performance was documented in the studies by Housen (2002) and Ionin & Wexler (2002), produce more omission errors than overgeneralization errors. In this respect, they differ from Garshol’s Norwegian learners. However, the similarities between Swedish and Norwegian learners emerge when the data are viewed across different subject types:

The Swedish learners show the same distribution of overgeneralization and omission errors across these two subject categories although the difference in frequencies with NP subjects is smaller and the difference with pronominal subjects is larger than in the Norwegian data. It is thus clear that even though the Swedish learners produce more omission errors than overgeneralization errors in total […], when the erroneous clauses are divided based on the type of subject, the same overgeneralization pattern as in the Norwegian data emerges (Garshol 2018: 79)

Thus, both Garshol’s (2018) data on Norwegian 15-16-year-olds and Fisher’s (1985) data on Swedish university students show that overgeneralization errors are much more common with noun subjects than with pronoun subjects. This is in line with the studies on Finnish learners (Vaurula 2012) and Japanese learners (Shibuya & Wakabayashi 2008) discussed in Section 2.2.

What happens if we break down the CORYL data in Table 1 across subject types is shown in Table 4:

	12-13-year-olds		15-16-year-olds
	Omission	Overgeneralization	Omission	Overgeneralization
Noun	48 (84.2%)	9 (15.8%)	29 (40.3%)	43 (59.7%)
PersPro	24 (43.6%)	31 (56.4%)	19 (44.2%)	24 (55.8%)
RelPro	12 (85.7%)	2 (14.3%)	11 (28.2%)	28 (71.8%)
Other	5 (100%)	---	19 (82.6%)	4 (17.4%)
Total	89 (67.9%)	42 (32.1%)	78 (44.1%)	99 (55.9%)

Table 4: Distribution of Affixal Agreement Error Types across Subject Types in the CORYL Corpus

We see that the CORYL data resemble the figures presented by Garshol (Table 2) in that about 60% of the errors produced by 15-16-year-olds with noun subjects are overgeneralization errors (example 10). However, there are differences between the two datasets as well. Thus, the 15–16-year-olds in CORYL also overgeneralize the -s with personal pronoun subjects (example 11) and with relative pronoun subjects (example 24).⁷ Furthermore, 12-13-year-olds overgeneralize the -s with personal pronoun subjects, as illustrated in examples (13) and (14):

    (10) However, many young people acts before they use their brain. (CORYL, 15-16-year-old)

   (11) When you comes over here I’ll take you to the Kontiki Museum. (CORYL, 15-16-year-old)

   (12). It is not just the young people who likes to be outside. (CORYL, 15-16-year-old)

  (13) And I thanks God I never will eader. (CORYL, 12-13-year-old

  (14) Every day they goes to the house! (CORYL, 12-13-year-old)

The facts described above are unexpected: since Garshol’s figures show overgeneralization only with noun subjects among her 15-16-year-olds, it could be expected that learners start generalizing the third-person-singular -s with this category of subjects. The proportion of overgeneralization errors was therefore expected to be highest with noun subjects among 12-13-year-olds as well. However, as mentioned above, the CORYL error data must be analyzed with caution given the modest number of tokens, and especially in cases like these, where there are numerous subcategories and comparatively few tokens in each category. In this case, it is not impossible that many of the overgeneralization errors with personal pronoun subjects are produced by learners who do not only use -s with personal pronouns, but who use it with all subjects, and whose noun subjects happen to be in the singular so that the relevant learners’ generalization of -s does not result in a large number of overgeneralization errors with noun subjects. Despite this weakness, the CORYL data are useful in showing that both age groups make both types of errors and that both types of errors are relatively common among both age groups.

One explanation for the variation in our data could be that other factors interfere with learners’ normal agreement strategies.

In what follows, we will discuss some potentially disturbing factors.

3.2 Subject Complexity and Non-Contiguity as Disruptors of Affixal Agreement

3.2.1 Subject Complexity and Non-Contiguity: Quantitative Data

As mentioned above, Killie (2019) and Son (2020) found that it is hard for Norwegian learners to come up with correct agreement marking when the subject is realized by two or more conjoined NPs. In the CORYL error data, there are ten instances of conjoined subjects, seven among younger and three among older learners. This is not a large number, of course, but the relevant figures sheds new light on the data in Table 4. Specifically, given that seven out of nine instances of overgeneralization among 12-13-year-olds involve a conjoined noun subject, we may hypothesize that conjoined NP subjects may cause overgeneralization to occur, but that apart from this, overgeneralization is confined to pronominal subjects among younger pupils. Some corpus examples of overgeneralization with conjoined NP subjects are provided in (15) and (16) below:

    (15) Now I have to do my homework before my mum and my dad come’s home. (CORYL, 12–13-year-old)

    (16) Me, Tommy and Frank has just build a little house up in the tree. (CORYL, 12–13-year-old)

Our discussion of conjoined subject NPs leads us to the question to what extent other types of complexity may have given rise to omission or overgeneralization errors in CORYL, more specifically, to what extent elements that intervene between the head of the subject NP and the agreement-marked verb may have disrupted the assignment of person and number marking to the verb, as explained in Section 2.2. There are twelve cases of post-modified subject NPs in the error corpus, all of which are produced by 15-16-year-olds.⁸ Some representative corpus examples are given in (17)-(20) below:

    (17) Young people today cares less about the environment than adults. (CORYL, 15–16-year-old)

    (18) … a old lady with a broken finger have to use the bed now, the nurse said. (CORYL, 15–16-year-old

    (19) … and you can se åssen people in this city lives (CORYL, 15–16-year-old, where åssen is dialectal Norwegian for ‘how’, KK))

    (20) The sound of them make me sick. (CORYL, 15–16-year-old)

Although twelve of these errors contain post-modifiers, their complexity is generally limited. Three of the post-modifiers do not even contain a nominal element, as is the case in (19). In the remaining nine cases, there is a post-modifier with a noun or personal pronoun, but in seven cases, the head and the local nominal element share the same number marking, as in example (20). Hence, ten out of the twelve errors involving post-modification cannot be explained in terms of agreement attraction. This finding lends support to the claim that with post-modified subjects, general processing problems may be a relevant explanation of agreement errors. The two cases in which agreement attraction is a possible explanatory factor are documented in (19) and (20) above.

As argued in Section 2.2, the assignment of agreement features may probably also be disrupted by adverbials which intervene between the subject and the agreement-requiring verb. The CORYL corpus contains 22 potential cases, two among younger learners and twenty among older ones. Some of these cases (two in the corpus of 12-13-year-olds and five in that of 15-16-year olds) contain other potentially disruptive elements as well, such as an intervening relative pronoun. The adverbials that intervene between the subject and the verb mostly take the form of a single adverb (e.g. always or just), as in (21) - (23) below.

    (21) … he is 12 years old and loves food, he always have food with him. (CORYL, 12–13-year-old)

    (22) …my brother alvays have the books he need for school (CORYL, 12–13-year-old)

    (23) I always gets out of the bed on the right, but not today. If we goes ten min. back, I’ll tell you why. (CORYL, 15–16-year-old)

It is not clear to what extent such short intervening elements really disrupt the assignment of person and number features to the verb, and this question should be tested experimentally. In (21), the adverb could well be the problem since this learner first produced two correct verb forms with he and then failed to produce a third correct form with he after always. On the other hand, the learner was able to get the second verb form right even though it was not adjacent to the subject. It appears that the processing of subject-verb agreement in non-contiguous contexts may be very unstable and that any generalizations and predictions on this point are difficult to make.

The figures in Table 4 above suggest that relative pronoun subjects represent a challenge for learners. As many as 39 of the 177 errors made by the older learners (viz. 22.0%) involve this type of subject. This is a large figure, in view of the fact that relative pronoun subjects are rather infrequent in CORYL: in Killie’s (2019) corpus of primary verbs, relative pronouns make up only 4-5% of the subjects in both age groups. Examples of incorrect agreement with relative pronouns from CORYL are documented in (24) and (25 below):

    (24) In his lyrics he talks about many things that affects the world. (CORYL, 15–16-year-old)

    (25) Ofcours you can find teenagers who doesen’t like flowers, but you can also find adults who hates flower. (CORYL, 15–16-year-old)

As explained in Section 2.2, the assignment of agreement features may also be disrupted when the verb phrase consists of two or more conjoined verbs, in which case only the first verb is adjacent to the subject. Example (23) above is a case in point, in which, however, no disruption occurs. Two examples in which the relevant structure may have caused such adisruption are provided in (26) and (27):

    (26) I play soccer very much and eats a lot of icecream. (CORYL, 12–13-year-old)

    (27) Oslo is the captial of Norway and have a lot to offer. (CORYL, 15–16-year-old)

Although 25 of the agreement errors in CORYL occur in clauses that contain conjoined verbs, this structural property cannot explain this type of errors in all these cases. In as many as 11 cases (four among the 12-13-year-olds and seven among the 15-16-year-olds), it is also the verb that immediately follows the subject which is incorrect, which presumably shows that the conjoined verb phrase is not the problem.

Inversion is a final structural factor that may disrupt subject-verb agreement. Presumably, it is difficult for a learner to come up with a verb before the subject has been produced. There are eight incidences of inversion in the error corpus - six among younger and two among older learners - , examples of which are given in (28) and (29):

    (28) One hollyday when I come home from my grandmother and my grandfather. have my friend paint “ the clubb house “. (CORYL, 12–13-year-old)

    (29) who are you, asks I (CORYL, 12–13-year-old)

Table 5 shows that quite a few of the sentences taken from the CORYL contain one or more of the structural features that possibly disrupt agreement assignment. These figures support the hypothesis that the relative pronoun represents the biggest stumbling block:

	12–13-year-olds	15–16-year-olds
Conjoined subject NPs	7	3
Subject NPs with post-modifiers	---	12
Intervening adverbial	2	20
Relative pronoun subject	14	39
Conjoined VPs	8	17
Inversion	6	2
Total	37 of 131 (28.2%)	93 of 177 (52.5%)
Number of sentences with one or more of the structural properties	36 of 131 (27.5%)	83 of 177 (46.9%)

Table 5: Number of Agreement Errors in CORYL involving Structural Properties possibly disrupting Subject-Verb Agreement

3.2.2 Subject Complexity, Non-Contiguity and Learner Strategies

Although the structural features discussed above occur in a considerable proportion of the clauses, we cannot be sure that they all play a major role. In about half of the sentences that contain a conjoined verb phrase, it is not only the verb phrase that is farthest away from the subject which is incorrect, but also the one that is adjacent to the subject. And with regard to post-modified subjects, only two of the twelve cases have a noun in the modifier which disagrees with the head noun in number. However, even if the structural properties discussed here may not in all cases have had a direct effect on the choice of the verb form, we cannot rule out the possibility that the use of such complex structures in itself may have increased the cognitive load on the learner and indirectly disrupted the assignment of agreement features, as explained in Section 2.2.

There is, however, another possibility that needs to be considered, i.e. whether some of the errors which were hypothesized to be the result of structural properties might instead be the result of a generalization strategy. For example, the conjugated verb helps, which is given in boldface in the extract below, is found in the statistics of errors, which may be due to non-contiguity caused by a conjoined VP, but if we study the other affixal verb forms in the extract (underlined), the relevant learner has apparently generalized the -s suffix, so helps could very well be the result of this generalization strategy instead:⁹

    (30) am [I?] like to play in the garden vid my frends Sarah and Adam. we are best frends and always helps icader. But too day we ar going to make a three-house. we have hammer planks and a three in the garden to sarah its a little bit dificolt to climb up in the three but wid some help of Sarahs Dad Wi did it and Den we begins to Bild it after ten minits, we gets the flor redy now we Bigins vid the valls Byt not gets sov long Befor Sarahs mum coms vid caks and milk. after a kvarter vid caks and milk we Biging [begin?] vid the valls agen after one houer Its redy den We Bigins too shin it upp the coler is green vid a flag on the top. inn il is the coler blu We got a taple and som chers. Wolh its redy. sarahs mum and dad cam [come? came?] to see it they think its good and we gets more cake and milk. bye bye the end. (CORYL, 12–13-year-old)

In connection with example (24) above, we argued that the incorrect verb form affects could be the result of non-contiguity caused by a relative pronoun. In (24b) below, the relevant example is given again, but this time with more context, and this context suggests that the cause of error is not the relative pronoun but a generalization strategy. The learner appears to have mastered agreement with personal pronouns but probably generalizes -s with other subject types. Hence, in (24b), neither the verb form affects nor the other erroneous verb forms in bold can be straightforwardly attributed to the non-contiguity caused by the relative pronoun:

(24b) I think all the people in this world can make a difference. I know many people who has made a difference to me. Good and bad ones. I know about many famous people that tries to make a difference to the world. " Band Aid ", for example. Famous singers gets together, sings a song, and the money from the concert goes to people who need it. Ones they did it to give food to Africa, and now they're gonna do it to help the houseless people after the tsunami in south-east -Asia. I also know that rich singers and actors/actresses gives money to poor people. Its hard to pick one, because many people does a great job. But if I don't think of the people that gives money, and helps other people, I know a singer that has made a difference. His name is Jhonny Bravo. Jhonny Bravo has made a difference to the world through his lyrics and melodies. His music brings joy to the people, and it also makes us think about the world we live in. In his lyrics he talks about many things that affects the world. (CORYL, 15–16-year-old)

In conclusion, complex subjects and non-contiguity are factors that may well disrupt the assignment of agreement features. However, to settle this question, we need further investigations which study individual learner data more closely, not only focusing on general quantitative figures.

4 Discussion and Future Prospects

4.1 Discussion

The hypotheses, results, and claims presented in this paper raise a number of issues that should be addressed. For example, it should be pointed out that learners’ omission of the suffix -s does not necessarily involve a lack of awareness that the suffix exists and should be used in some present-tense contexts. Many learners may have noticed that verbs have a present tense -s form, but may avoid using it for some time because they have not yet been able to figure out when it should be used. As explained by Gass, Behney & Plonsky (2013), avoidance of structures which are perceived as difficult is not uncommon in second language acquisition.

As for the increased use of the suffix -s, we argued that Norwegian learners of English initially stick to the base form of the verb and then commonly start to overuse the suffix -s. We should, however, ask ourselves what this shift really consists in. In Section 3.1, we – implicitly – assumed that the suffix -s spreads in a uniform way from one specific subject type to another. However, the possibility that learners may come up with different hypotheses and that as a result, learner grammars may develop along different routes, should be considered. For example, between the ages of 12-13 and 15-16, some Norwegian learners start producing subject–verb combinations such as The boys smiles. We hypothesized that these learners use the suffix -s with plural noun subjects, as is commonly assumed (e.g. Fisher 1985). However, as suggested in Section 3.2.2, we cannot exclude the possibility that some learners use -s with all plural subjects (because of their association of nominal -s with the plural), with all noun subjects regardless of number (because -s is associated with nouns), or with all subjects except personal pronoun subjects.¹⁰ Without proper data, all these scenarios remain mere speculation.

We should also ask ourselves why Norwegian learners commonly generalize the present tense suffix -s. This phenomenon may be a teaching effect, as affixal subject-verb agreement is typically taught explicitly in lower secondary school, i.e. from 8^th grade to 10^th grade. As we know, extensive focus on a given form may lead to overgeneralization of that form (Lightbown 1983, Pica 1985).

4.2 Future Prospects

In investigating the effect of potentially disruptive factors, we observed that it is difficult to ascertain to what extent such factors are the cause of error, as in some cases, the error could instead be the result of a general omission or a generalization strategy. It is important to note that such strategies cannot be adequately studied within an obligatory context study, either, since such studies do not include contexts in which generalization strategies lead to overgeneralization, i.e. non-obligatory contexts, as pointed out by Ellis (1994: 75).

Accurately explaining why agreement errors occur is clearly not easy, and factors which seem to play a role at aggregate level need not play a role for every individual. To fully understand the problems that learners experience, we therefore need to look beyond aggregate figures and study individual learners’ language production more closely. We need investigations which involve a considerable number of case studies and where the textbase for each learner is a solid one. Such studies should be of longitudinal orientation. This orientation will allow us to study learner competence, learner problems and strategies of individual learners at various points, and, in a next step, generalisations of such factors may become possible. Below are some questions that need to be addressed for each learner:

• Does the learner generally master subject-verb agreement?

• Is there a difference in accuracy depending on the category of the head of the subject, e.g. noun, personal pronoun, indefinite pronoun etc.?

• Is there a difference in accuracy between suppletive and affixal agreement marking?

• Is accuracy the same with be, have, and do when used as auxiliaries and as lexical verbs?

• Is accuracy higher with frequent verbs such as have and do than with less frequent verbs?

• Does the learner have a suppletive generalization strategy, i.e. does he or she generalize a specific form in the present tense and / or a specific form in the past tense?

• Does the learner have an affixal generalization strategy, viz. omission or generalization of the suffix -s?

• Does the learner generally use certain (frequent) collocations correctly, such as I think, I have and it has, but in other cases resorts to a default generalization strategy?

• Do structural features, such as complex subjects, intervening adverbials, disrupt the learner’s agreement assignment?

• Does the learner’s accuracy vary with text type, register or style, e.g. spoken vs. written language / situations, formal vs. informal language?

• Is there a degree of randomness in the learner’s choices?

These are important questions to be addressed in the future. Knowledge about generalization strategies and construction types which complicate agreement assignment will enable language teachers to create ‘smart’ teaching activities, in the words of Slabakova (2016: 403), and hence help them acquire this very complex area of grammar.

References

Berko, J. (1958). The child’s learning of English morphology. In: Word 14, 150-177.

Biber, D., S. Johansson, G. Leech, S. Conrad & E. Finegan (1999). Longman Grammar of Spoken and Written English. Harlow: Longman.

Brehm, L. & K. Bock (2013). What counts in grammatical number agreement? In: Cognition, 128 (2), 149-169.

Breiteneder, A. (2005). The naturalness of English as a European lingua franca: The case of the ‘third person -s’. In: Vienna English Working Papers, 14 (2), 3-26.

Bock, K. & C. A. Miller (1991). Broken agreement. In: Cognitive Psychology, 23 (1), 45-93.

Brown, R. (1973). A First Language: The Early Stages. London: George Allen and Unwin.

Cheshire, J. (1999). Spoken Standard English. In Bex, T. & Watts, R. J. (eds.) (1999). Standard English: The Widening Debate. New York: Routledge, 129-148.

Clahsen, H., C. Felser, K. Neubauer, M. Stato & P. Silva (2010). Morphological structure in native and non-native language processing. In: Language Learning 60, 21-43.

Dröschel, Y. (2011). Lingua Franca English: The Role of Simplification and Transfer. Bern: Peter Lang.

Dulay, H. C. & M. K. Burt (1974). Natural sequences in child second language acquisition. In: Language Learning 24, 37-53.

Eberhard, K.M, J. C. Cutting & K. Bock (2005). Making sense of syntax: Number agreement in sentence production. In: Psychological Review 112, 531-559.

Ellis, R. . 1994. The study of second language acquisition. Oxford: Oxford University Press.

Fayol, M., M. Hupet & P. Largy (1999). The acquisition of subject-verb agreement in written French: From novices to experts’ errors. In: Reading and Writing: An Interdisciplinary Journal 11, 153-174.

Fisher, U. T. (1985). The sweet sound of concord: A study of Swedish learners’ concord problems in English. Lund: CWK Gleerup.

Garshol, L. (2019). We agreed to disagree: Agreement in L2 English written by young Norwegian learners. In: Abel, A., Glaznieks, A., Lyding, V. & Nicolas, L. (eds.) (2019). Widening the Scope of Learner Corpus Research: Selected Papers from the Fourth Learner Corpus Research Conference. Louvain: Presses universitaires de Louvain, 237-252.

Garshol, L. (2018). I just doesn’t know: Agreement errors in English texts by L2 Norwegian learners: Causes and remedies (PhD thesis). Kristiansand, Norway: University of Agder. https://uia.brage.unit.no/uia-xmlui/handle/11250/2589044

Gass, S., Behney, J. & L. Plonsky (2013). Second language acquisition: an introductory course. New York: Routledge.

Hartsuiker, R. J. & P. N. Barkhuysen (2006). Language production and working memory: The case of subject-verb agreement. In: Language and Cognitive Processes 21 (1-3), 181-204).

Hasselgren, A. & K. T. Sundet (2017). Introducing the CORYL Corpus: What it Is and How We Can Use it to Shed Light on Learner Language. In: Bergen Language and Linguistics Studies, 7. (https://doi.org/10.15845/bells.v7i0.1107; 15-10-2018)

Hasselgård, H., P. Lysvåg & S. Johansson (2012). English grammar: Theory and use. Oslo: Universitetsforlaget.

Hauge, H. & A. Angelsen (2018). Developing language awareness. In: Bøhn, H., Dypedahl, M. & Myklevold, G.-A. (eds.) (2018). Teaching and Learning English. Oslo: Cappelen Damm Akademisk, 266-285.

Haznedar, B. (2001). The Acquisition of the IP system in child L2 English. In: Studies in Second Language Acquisition, 23, 1-39.

Housen, A. (2002). A corpus-based study of the L2-acquisition of the English verb system. In Granger, S., Hung, J. & Petch-Tyson, S. (eds.) (2002). Computer learner corpora, second language acquisition, and foreign language teaching. Amsterdam/Philadelphia: John Benjamins, 77-116.

Ionin, T. & K. Wexler (2002). Why is ‘is’ easier than ‘-s’?: Acquisition of tense / agreement morphology by child second language learners of English. In: Second Language Research, 18 (2), 95-136.

Jensen, I. N. (2016). The Bottleneck Hypothesis in L2 acquisition: Norwegian L1 speakers’ knowledge of syntax and morphology in English L2 (Master’s thesis). Tromsø, Norway: UiT The Arctic University of Norway.

Jensen, I. N., R. Slabakova, & M. Westergaard (2017). The Bottleneck Hypothesis in L2 acquisition: Norwegian L1 learners’ knowledge of syntax and morphology in English L2. In: LaMendola, M. & Scott, J. (eds.) (2017). Proceedings of the 41^st annual Boston University Conference on Language Development. Somerville, MA: Cascadilla Press, 333-346.

Jensen, I. N., R. Slabakova, M. Westergaard & B. Sundquist (2019). The Bottleneck Hypothesis in L2 acquisition: L1 Norwegian learners’ knowledge of syntax and morphology in L2 English. In: Second Language Research, 36-1, 3-29.

Keeney, T. J. & J. Wolfe (1972). The acquisition of agreement in English. In: Journal of Verbal Learning and Verbal Behavior, 11, 698-705.

Killie, K. (2019). The acquisition of subject-verb agreement among Norwegian (teenage) learners of English: focus on the subject. In: Journal of Linguistics and Language Teaching, 10-2.

https://linguisticsandlanguageteaching.blogspot.com/2019/12/blog-post_25.html

Killie, K. (2021). ‘I are in New York’: om tilegnelsen av samsvarsbøyning i engelsk blant innlærere med norskspråklig bakgrunn. [’I are in New York’: on the acquisition of subject-verb agreement in English by learners with a Norwegian language background.] In: Norsk lingvistisk tidsskrift, 39 (1).

Lardiere, D. (2009). Some thoughts on a contrastive analysis of features in second language acquisition. In: Second Language Research, 25 (2), 173-227.

Levin, M. (2001). Agreement with Collective Nouns in English. Lund: Lund University Press.

Lightbown, P. M. (1983). Exploring the relationships between developmental and instructional sequences in L2 acquisition. In Seliger, H. W. & Long, M. H. (eds.) (1983). Classroom oriented research in second language acquisition. Rowley, MA: Newbury House, 217-245.

Neff, J., F. Ballesteros, E. Dafouz, F. Martínez, J.-P. Rica, M. Díez & R. Prieto (2007). A contrastive functional analysis of errors in Spanish EFL university writers’ argumentative texts: Corpus-based study. In: Fitzpatrick, E. (ed.) (2007). Corpus linguistics beyond the word: Corpus research from phrase to discourse. Amsterdam: Rodopi, 203-225.

Negro, I. & L. Chanquoy (2000). Subject-verb agreement with present and imperfect tenses: A developmental study from 2nd to 7th grade. In: European Journal of Psychology of Education, 15 (2), 113-133.

Ocampo, S. (2013). Morphological variability in long-distance subject-verb agreement: A study of native and nonnative processing (Doctoral dissertation). Kansas, USA: University of Kansas.

Pica, T. (1985). Linguistic simplicity and learnability: Implications for language syllabus design. In Hyltenstam, K. & M. Pienemann (eds.) (1985), Modelling and assessing second language acquisition. Philadelphia: Multilingual Matters, 137-151.

Ponmani, M. & S. Mekala (2016). An analysis of concord errors in ESL learners’ writing. In: The Global Journal of English Studies, II(I). (http://thegaes.org/files/documents/Kongu-23-M-Ponmani-and-Dr-S-Mekala.pdf; 15-10-2018)

Prévost, P. & L. White (2000). Missing Surface Inflection or Impairment in second language acquisition? Evidence from tense and agreement. In: Second Language Research, 16(2), 103-133.

Quirk, R., S. Greenbaum, G. Leech & J. Svartvik (1985). A Comprehensive Grammar of the English Language. London: Longman.

Radford, A. (1990). Syntactic Theory and the Acquisition of English Syntax: The Nature of Early Child Grammars of English. Oxford: Blackwell.

Reifegerste, J., Hauer, F. & Felser, C. (2017). Agreement processing and attraction errors in aging: evidence from subject-verb agreement in German. In: Aging, Neuropsychology, and Cognition, 24 (6), 672-702.

Shibuya, M. & S. Wakabayashi (2008). Why are L2 learners not always sensitive to subject-verb agreement? In Roberts, L., Myles, F. & Davis, A. (eds.) (2008). EUROSLA Yearbook 8. John Benjamins, 235-258.

Slabakova, R. (2013). What is easy and what is hard in second language acquisition: A generative perspective. In: Mayo, M. P. G., M. J. G. Mangado & M. M. Adrián (eds.) (2013). Contemporary Approaches to Second Language Acquisition. Amsterdam: John Benjamins, 5-28.

Slabakova, R. (2016). Second language acquisition. Oxford: Oxford University Press.

Slabakova, R. (2019). The Bottleneck Hypothesis Updated. In: Ionin, T. & Rispoli, M. (eds.) (2019). Three streams of generative language acquisition research : selected papers from the 7^th meeting of Generative Approaches to Language Acquisition – North America, University of Illinois at Urbana-Champaign. Amsterdam: John Benjamins. 319–345.

Slabakova, R. & J. Gajdos (2008). The Combinatorial Variability Hypothesis in the Second Language. In: Bowles, M., Foote, R., Perpinján, S. & Bhatt, R. (eds.) (2008). Selected Proceedings of the 2007 Second Language Research Forum. Somerville, MA: Cascadilla Proceedings Project, 35-43.

Son, M. (2020). Subject-verb agreement in written English by L1 Norwegian university students: Error patterns, causes, and implication for teaching (MA thesis). Tromsø, Norway: UiT The Arctic University of Norway.

Strang, B. M. H. (1966). Some features of S-V concord in present-day English. In Cellini, I. & G. Melchiori (eds.) (1966). English studies today: Fourth series. Rome, Italy: Edizioni di Storia e Letteratura, 73-87.

Vaurula, N. (2012). Subject-Verb Agreement of Finite Verbs in the Present Tense in Written Productions of Finnish Secondary School Learners of English as a Second Language (Master's Thesis). The University of Jyväskylä.

Wee, R., J. Sim & K. Jusoff (2009). Enhancing the Quality of EAP Writing through Overt Teaching. In: English Language Teaching, 2 (3), 58-68.

White, L. (2003). Second language acquisition and Universal Grammar. Cambridge: Cambridge University Press.

Wolfram, W. (2004). The grammar of urban African American Vernacular English. In: Kortmann, B. & E. W. Schneider (eds.) (2004). Handbook of Varieties of English. Berlin: Mouton de Gruyter, 111-132.

Author:

Professor Kristin Killie

Department of Education

UiT The Arctic University of Norway

N-9037 Tromsø

Norway

______________________

1 I wish to thank the editor of JLLT and the reviewers for valuable help and feedback. My thanks also go to the Research Group for Language Testing and Assessment at the University of Bergen for creating the Corpus of Young Learner Language and making it freely available.

2 Note that while the term omission error refers to the non-use of the present tense -s suffix, strictly speaking, it is only correct in cases where the learner knows that there is something to omit. A term like ‘consistent/general use of the base form’ would be more neutral here. For the sake of simplicity, the term omission is used in the present paper.

3 The examples and categorizations are from Son 2020: 49.

4 This type of error also occurs among native speakers in other languages that have subject-verb agreement. Fayol, Hupet & Largy (1999) and Negro & Chanquoy (2000) discuss such errors for French speakers and explain them in terms of processing problems. The same conclusion is drawn by Reifegerste, Hauer & Felser (2017) based on a study of older and younger speakers of German.

5 Bes is unattested in CORYL, while be is used only once in a present-tense context:

Now Peter be very mad and so he scream: - Come up here with your fat ass and help me you little fat ass (CORYL, 12-13-year-old)

6 The distribution is very similar with expletive subjects, but here the token rates are so low that they must be interpreted with caution.

7 Garshol (2018, 2019) does not provide separate figures for relative pronoun subjects. These are presumably categorized according to the antecedent, which, in most such cases, is a noun. Merging the categories noun and relative pronoun for 15-16-year-olds in Table 4 into a larger noun category brings the proportion of overgeneralization cases for nouns up to 64% (71 of 111 tokens), which is very close to Garshol’s proportion (viz. 62%).

8 As shown in Killie (2019), the reason why the younger learners in CORYL do not make such mistakes is simply that they generally do not produce post-modified subject NPs.

9 Note that even this learner, who clearly has a number of issues when it comes to his or her English, including both grammar and spelling, manages to use the correct form of be. This is also the case with the learner who produced the text documented in (21b). This is in agreement with the claim that the present-tense forms of the verb be are acquired before any affixal agreement is (Killie 2019).

10 If it really is the case, as the data in Table 4 may suggest, that some learners first hypothesize that -s is used with personal pronouns, then this hypothesis could be the result of frequent exposure to combinations of present-tense verb forms and the pronoun it. According to Biber et al. (1999: 334), it is the only personal pronoun which is frequent in both formal and informal genres, and the ubiquity of this pronoun may make both the pronoun – and the suffix -s – highly salient.