Word-order typology and the acquisition of case marking: A self-paced reading study in Latin as a second language

Abstract

This article reports the findings of a study in which we investigated the possible effects of word order on the acquisition of case marking. In linguistic typology (e.g. Greenberg, 1963) a very strong correlation has been shown between dominant SOV (subject object verb) word order and case marking. No such correlation exists for SVO (subject verb object) languages. It is possible then that the mind is more likely to expect case marking when confronted with a language with SOV word order but not necessarily so if the language has SVO word order. We tested this hypothesis with 54 naive learners of Latin with English as a first language (L1). The participants were divided into two groups. One received a 100-word input treatment in Latin that contained only simple SOV sentences, and the other received the same input treatment except that the word order of the treatment sentences was SVO. After the treatment, a surprise self-paced reading test that contained grammatical and ungrammatical case-marked sentences was administered. Participants read test items that matched the word order of the treatment they received (i.e. SOV learners read SOV sentences, and SVO learners read SOV sentences). Results showed a significant slowing down on ungrammatical sentences for the SOV group but not for the SVO group. However, on a test of basic sentence comprehension in which case marking was the cue to determine who did what to whom, we found no distinction between the groups. We discuss these findings in light of how typological universals work in languages and what they could mean for language acquisition.

Keywords

case marking Latin linguistic typology second language acquisition word order

I Introduction

One of the major assumptions of contemporary syntactic theory is that, despite the apparent diversity in the world’s languages, languages are organized according to a set of universals and a set of permitted options within those universals. The best-known version of this hypothesis is found in Government and Binding theory and in Mark Baker’s (1996, 2001) work. Alongside the work of theorists, typologists have done descriptive work that documents which features co-occur in the world’s languages. The idea here is that, if features co-occur with greater than chance likelihood, we can assume that those features are related to each other on some deep level, and this tells us something about the organization of language. These ideas were codified in a set of universals published in 1963 by Joseph Greenberg. Here is one example of a universal describing a co-occurrence of features.

Languages with dominant VSO order are always prepositional. (Greenberg, 1963, p. 78)

According to this universal, prepositions automatically co-occur with VSO order regardless of how closely related the languages are (e.g. both Irish and Hawaiian are VSO languages and have prepositions but are clearly unrelated to each other). Other universals are less categorical, such as Greenberg’s Universal 41:

If in a language the verb follows both the nominal subject and nominal object as the dominant order, the language almost always has a case system. (Greenberg, 1963, p. 96)

The claim here is that case marking (strongly) tends to co-occur with SOV (subject object verb) word order. We will review this universal in detail later but we point out here the use of ‘almost always’. This phrase stands in stark contrast to the simple ‘always’ of the previous universal. Thus, co-occurrences among features within the typological framework are either absolute universals or they represent (very) strong correlations.

The other major focus of typological research is in markedness relationships or implicational hierarchies. The most well-known of these is the Noun Phrase Accessibility Hierarchy (NPAH), which organizes the possible ways to form relative clauses (RCs) (e.g. Comrie, 2007; Keenan and Comrie, 1977):

(1) SU > DO > IO > OBL > GEN > OComp

Implicational hierarchies are organized so that the most frequently occurring type of structure (subject relative clauses, in this case) is on the left, and the least frequent type is on the right. The presence of one type of structure in a given language implies the presence of every structure to its left in that language, but makes no predictions about the presence or absence of structures to the right of it in the hierarchy. The NPAH states that, cross-linguistically, subject relative clauses (SU, here) are the least marked relative clause structures. Direct object relative clauses (DO, here) are more marked than SU RCs, but are still relatively unmarked. Genitive (GEN, here) relative clauses and Object of Comparison (OComp, here) relative clauses are the most marked, and therefore the least frequent, relative clause type cross-linguistically. English has all of these relative clause types; examples of GEN and OComp relative clauses are presented in (2) and (3).

(2) The woman whose name I don’t remember …

(3) The woman who Jill is busier than …

The NPAH predicts that if a relative clause type is instantiated in a particular language, any relative clause to the left of it on the hierarchy will also be instantiated in that language. So, if a language has IO relative clauses, it will also have SU and DO relative clauses. The opposite prediction does not hold: if a language has IO relative clauses, it will not necessarily have oblique (OBL), GEN, and OComp relative clauses. The implicational hierarchy also means that a language cannot ‘skip’ a place on the hierarchy. In this way, a language can have SU, DO, IO and OBL relative clauses but cannot have SU, IO, and OBL relative clauses without also having DO relative clauses. The NPAH is a fairly robust typological generalization, although it has been updated based on languages sampled (for discussion, see Eckman, 2007).

II Background

1 Typology and L2 acquisition

Several contemporary second language (L2) acquisition researchers (e.g. Eckman, 1977, 1984, 1985; Hyltenstam, 1984) have used the basic assumptions of language typology – markedness relationships and the assumption that co-occurrence of specific features is not an accident – to make predictions about how learners will acquire specific features in an L2. The most influential of these is Eckman (1977, 1985), who proposed the Markedness Differential Hypothesis (MDH). The MDH states that where a learner’s native language (L1) differs from the target language (TL), the structures in the TL that are relatively more marked than those in the L1 will be more difficult for the learner. If the TL and the L1 differ in structures where the L1 is more marked than the TL, the TL structures should not present much difficulty to the learner. Because typological markedness applies to phonology and syntax, the MDH makes predictions about both learners’ ability to acquire contrasts and segments in an L2 as well as their relative difficulty with syntactic structures in the L2.

Although the MDH has fallen out of popularity in contemporary research in second language acquisition (SLA), the work that has tested the hypothesis has shown that, in general, the MDH explains some of the L2 acquisition data, and that this is particularly true for the relative difficulty of target language structures. For instance, several researchers have investigated whether the MDH explains the developmental stages in the acquisition of relative clauses. Researchers are interested in whether developmental sequences conform to the order spelled out in the NPAH, as the MDH predicts. The NPAH (Comrie, 2007; Keenan and Comrie, 1977), is repeated here in (4):

(4) SU > DO > IO > OBL > GEN > OComp

Recall that English allows all of the relative clauses in this hierarchy. The MDH predicts that L1 English speakers who are learning languages that allow fewer options for relative clause types in the TL will not find any one of these RC structures more difficult than any other because English has the most marked setting for relative clauses. In contrast, the MDH predicts L2 English learners whose L1 lacks any relative clause structure that is lower on the implicational hierarchy than indirect objects (IO, here), will find GEN and OComp RCs comparatively more difficult than IO and DO RCs. The work on the acquisition of relative clauses tends to confirm this hypothesis (e.g. Gass, 1979; Hyltenstam, 1984; Pavesi, 1986; but, for critical discussion, see Shirai and Ozeki, 2007).

Although markedness relationships have been the primary way in which typology has been applied to L2 research, there may also be a role for typological generalizations that do not involve markedness relationships. As noted earlier, several Greenbergian universals state that linguistic feature X and linguistic feature Y co-occur with greater than chance likelihood. If these universals do reflect something basic about the organization of language, they might also help us predict – and, ideally, explain – aspects of the language acquisition process. If linguistic features are not related by chance, then the processing mechanisms responsible for language acquisition may be sensitive to those relationships, and use the presence of one feature – such as word order – to hypothesize the presence of another feature. For example, one prediction related to the universal regarding VSO word order cited above is that once exposed to simple three-word sentences in Irish or Hawaiian, a language acquirer may ‘assume’ that the language has prepositions even before encountering any.¹ These typological relationships, in other words, may help constrain the hypothesis space and help the learning mechanisms be more efficient.

Indeed, research on typological universals in second language acquisition has been fruitful in suggesting that development over time is influenced by both typological universals and markedness (e.g. Eckman et al., 1989). A major finding is that unmarked or lesser-marked structures are typically acquired before marked or more marked structures. Moreover, learners may generalize from more marked structures to less marked structures but not vice versa (e.g. Gass, 1979). For example, there is evidence that second language (L2) English learners who encounter indirect object relative clauses in English ‘infer’ that English also has subject and direct object relative clauses, which is in keeping with the predictions of the NPAH. However, if they encounter only subject relative clauses, no inference can be made about other relative clauses (for critical discussion, see Shirai and Ozeki, 2007, as well as Eckman, 2007). In addition, typological research has also suggested that markedness can affect the role of the L1 in the L2 context (e.g. Eckman, 1977). For example, learners have demonstrated greater ease in acquiring unmarked structures different from their L1 than marked structures different from their L1.

In the present study, we use linguistic typology to look at second language acquisition from a different viewpoint and use a methodology not yet used in such research: self-paced reading. Our question concerns the role of basic word order (i.e. SVO vs. SOV) on the acquisition of case marking on nouns using naive learners of Latin as a second language. The article is organized in the following fashion. First, we discuss the role of word order and case marking from a typological perspective. We then review a study we conducted in Japanese (VanPatten and Smith, 2015) that led us to ponder how word order might affect the acquisition of case marking. We then present our study and its results, followed by a discussion and conclusion.

2 Word order and case marking

The relationship between word order and case marking has been a focus of discussion since Greenberg (1963) published his seminal work on universals of grammar. Of relevance here is Greenberg’s Universal 41 (repeated from above).

If in a language the verb follows both the nominal subject and nominal object as the dominant order, the language almost always has a case system. (Greenberg, 1963: 96)

Greenberg’s original survey relied on a small sample of languages, and subsequent work, which expanded the sample size, suggests that ‘almost always’ is not quite accurate. Nevertheless, there is a strong correlation between word order and case marking. Dryer (2002), for example, offers observations on a database containing 502 languages. He notes that SOV languages are much more likely to have case marking (what he calls ‘case distinction’) than SVO languages. In his database, 181 out of 253 (72%) SOV languages marked case, while only 26 out of 190 (14%) SVO languages marked case.²

Although 72% vs. 14% case marking does not indicate a clear-cut (‘almost always’) universal, the difference is strongly suggestive and raises the question of why there is such a strong correlation between word order and case marking. If word order and case marking are unrelated to each other, we might expect to see roughly half of the world’s languages marking case, and that SOV and SVO languages would be equally represented as case marking languages. That SOV languages mark case much more frequently suggests that this correlation is not accidental. There are both functional and processing reasons to think that case marking and word order would be related. From a functional perspective, and following Gilligan (1987), Dryer argues that when languages permit subject and/or object drop, it is easy to tell what the function of an NP is in an SVO language: NP-V (object drop) would mean the NP is probably the subject whereas V-NP (subject drop) would probably indicate an object. However, with SOV languages, NP-V could equally represent SV (object drop) or OV (subject drop). In short, the role of the NP in such cases is ambiguous. Case marking appears more often in SOV languages, then, to avoid such ambiguity.

What is more, as several researchers who work on Japanese psycholinguistics have argued, case marking becomes a valuable tool for the parser during sentence comprehension. Unlike SVO languages, in which the verb comes early and the agent and patient are distinguished by their relative position to the verb, the verb comes late in SOV languages. Case particles provide the parser with information about a noun’s thematic role, so that each element can be incorporated into the sentence as the structure is built (see, amongst others, Miyamoto, 2002; Ueno and Polinsky, 2009; Yamashita, 1999, 2000). That is, with two NPs preceding the verb that are both capable of being either subjects and objects (e.g. ‘Mary-John-kisses’ vs. ‘Mary-apple-eats’), the processor/parser does not have to wait until the verb to use lexical semantics to disambiguate the functional roles of the NP. It is true that, in many cases, lexical semantics distinguishes the agent and the patient. So, with the ‘eat’ example, the parser may be attuned to animates being agents more often than inanimates, but there are other cases in which lexical semantics or animacy will not be enough to disambiguate arguments. With the ‘kiss’ example, either Mary or John could be kisser or kissee. Similarly, compare the sentence ‘Mary-apples-eats’ with the sentence ‘Mary-apple-poisons.’ Is Mary putting poison in the apple or is the apple poisoning Mary? In cases like these, case marking provides structural and thematic information so that parsing is not delayed. So if a language has both ‘Mary-nom John-acc kisses’ and ‘Mary-acc John-nom kisses’ (as do languages such as Japanese and Turkish), the parser can project sentence structure as soon as the first case marked noun is encountered. As Ueno and Polinsky (2009) point out, for native speakers, processing an SOV sentence does not take any longer than processing an SVO sentence, which suggests that case marking helps the parser avoid any delay in structure building. Because case particles provide the parser with syntactic cues, case marking languages also tend to allow scrambling. SVO languages tend to have fixed word orders, but case-marking SOV languages almost always allow both SOV and OSV word orders, and some languages, like Russian, allow all six possible word orders.

In short, there is strong typological evidence for a correlation between case marking and SOV word order that contrasts with a lack of case marking in SVO languages. Functional and processing constraints also provide suggestive evidence that this relationship is not accidental. If learners (or the language processing mechanisms) are sensitive to typological relationships, it is possible that learners of an SOV language are more likely to ‘search’ for case marking in the input to which they are exposed. But when exposed to an SVO language, the parser may not incorporate case marking as quickly as it does in SOV languages. In Section II.3, we describe the results of a previous experiment on head-directionality that caused us to consider this prediction.

3 An experiment in Japanese L2

In VanPatten and Smith (2015), we tested whether exposure to simple SOV structures allows learners to project beyond the input data (a kind of poverty of the stimulus situation for beginning learners). Our focus was on whether naive second language learners of Japanese would ‘determine’ that, after exposure to simple sentences like ‘Taro-apple-eats’, Japanese is head-final in all domains. That is, would learners ‘infer’ that all phrase structure (e.g. CP) was head-final and not just VP?

In our study, learners with no knowledge of Japanese or any SOV language (and all with English as their first language) participated in a laboratory study in which they were exposed to 100 sentences in Japanese. Fifty of these sentences were basic SOV sentences and another 50 were SOV sentences with locative phrases. In Japanese, the equivalents of locatives such as in are postpositional, in line with the rigid head-final nature of Japanese phrase structure (e.g. ‘Taro kitchen-in apple eats’). The treatment consisted of hearing Japanese sentences matched to pictures and then periodically testing the learners on vocabulary (i.e. the learners would hear a sentence such as ‘Taro-apple-eats’ and then select from two pictures in which Taro is eating an apple or drawing an apple). After the 100-sentence treatment, learners were given a surprise self-paced reading test in which they read Japanese sentences (in Roman script) and had to respond to a comprehension questions about the content. Fifty percent of the sentences followed head-final word order and the other 50% followed head-initial word order (i.e. the word order of their L1, English). Part of the surprise in this reading test was that we embedded novel structures they had not been exposed to in order to see if they projected beyond the input data of the treatment. Those structures were questions (i.e. the particle ka, which marks polar questions in Japanese, is the head of CP and is in final position) and embedded clauses (i.e. the complementizer toh, which is the equivalent of that in English, is also the head and in CP final position).

As distractors, we included sentences with anomalous case marking such that ‘Taro-apple-eats’ would appear at ‘Taro-acc apple-nom eats.’ Although we found some evidence for projection beyond the data they were exposed to regarding phrase structure, what we incidentally found was that every single participant showed sensitivity to case marking in the distractors. That is, every single participant slowed down on the SOV sentences with anomalous case marking compared to the SOV sentences with correct case marking. (In self-paced reading, slowing down is an indication that the reader ‘detects’ something is amiss even though the reader is focused on meaning.) Case marking was not the focus of our study but we were intrigued that 100% of the participants seemed to show evidence of being sensitive to case marking after such a short initial exposure period.

These incidental data led us to the idea that perhaps SOV order in the input taps a linguistic predisposition to look for case marking, as suggested in Section II.2. However, we could not be sure this was the case because Japanese is rigidly verb-final and there was no way to directly compare exposure to SOV and exposure to SVO. What is more, most analyses of Japanese posit a KP (case phrase) for which the bound morphemes that mark case are heads (e.g. ‘Taro-ga [nom]’ and ‘Taro-o [acc]’) (e.g. Hinds, 1982; Koopman, 2005). These phrasal heads may be phonologically more salient than, say, inflections on nouns. For these reasons, we decided to conduct a study using Latin because, although the canonical sentence structure of Latin is SOV, it permits almost all word orders and has an inflectional case marking system. In other words, Latin case marking is a reflex of a functional feature, and not a separate projection in its own right.

4 Latin

Latin is an SOV language with rich case marking. Here we focus on nominative and accusative case marking only. Although Latin has what are classically called five ‘declensions’ in which nominative case varies between noun-final -s, noun-final -r, or a bare vowel, all accusative inflections end in a vowel plus -m. For example, puer (boy-nom), puerum (boy-acc), aqua (water-nom), aquam (water-acc), volpis (fox-nom), volpem (fox-acc), caballus (horse-nom), and caballum (horse-acc). In addition, Latin allows scrambling. Of interest here is SOV and SVO as exemplified below.³

(5) Volpis caballum videt. [SOV]

‘The fox sees the horse.’

(6) Volpis videt caballum. [SVO]

‘The fox sees the horse.’

(7) Caballus volpem videt. [SOV]

‘The horse sees the fox.’

(8) Caballus videt volpem. [SVO]

‘The horse sees the fox.’

Because case marking is obligatory in Latin, two nouns that are both marked with either nominative or accusative case in simple sentences are ungrammatical, as in the following examples.

(9) * Volpis caballus videt. [SSV]

‘The fox sees the horse.’

(10) * Volpis videt caballus. [SVS]

‘The fox sees the horse.’

(11) * Caballum volpem videt. [OOV]

‘The horse sees the fox.’

(12) * Caballum videt volpem. [OVO]

‘The horse sees the fox.’

These features of Latin allow us to directly test whether there is a predisposition to search for case marking in the input if the learner consistently encounters SOV sentences as opposed to consistently encountering SVO sentences. That is, with Latin we could arrange two separate treatments for naive learners: one with exclusively SOV input sentences and another with exclusively SVO input sentences.

5 Self-paced reading

Over the years, researchers in L2 acquisition have become increasingly concerned with the use of what are called off-line measures of learner knowledge. These off-line measures are those in which knowledge is tested after learners read or hear something in the L2. Such measures include grammaticality judgments, truth-value tasks, and other kinds of ‘paper-and-pencil’ tests, even when they are delivered via computer. As argued by Hulstijn (2005) and Ellis (2005), the problem with such measures is that they may invite the use of explicit knowledge without tapping underlying and developing implicit knowledge or mental representation. They may invite introspection such that research participants simply think too much about what they are responding to rather than providing more intuitive and immediate reactions.

For a number of years now, research in L2 performance (e.g. sentence processing) has borrowed on-line methods from L1 psychological research. These on-line methods include self-paced reading, eye-tracking, event-related potentials, and others (for an overview of such methods, see Jegerski and VanPatten, 2014). Typically used to track problems in sentence computation during real-time comprehension (such as ambiguity resolution, garden paths, and filler gap dependencies) such methods have been successfully used to examine learner responses to ungrammaticalities during real-time comprehension (e.g. Foote, 2011; VanPatten et al., 2012).

Self-paced reading is a computer-mediated research tool used to measure participants’ reading times during sentence comprehension (e.g. Jegerski, 2014; Mitchell, 2004). Typically, participants are seated in front of a computer screen and after a fixation point on the screen directs their attention to where a sentence will appear, they press a button to reveal a series of dashes that stand for words or phrases. For example, the ungrammatical Latin sentence *Volpis caballus videt would be replaced with the following on the screen: –— –– ––. The participant would then press a designated button to reveal the first word: Volpis –– –– The participant would then press a designated button to reveal the second word while the first word disappears: –– caballus ––. A final button press would reveal the third word by itself: –– –– videt. The next button press would reveal a question about the sentence related to general sentence meaning, words used, or other meaning-based content but not about the grammaticality of the sentence. In this way, the participant is ‘forced’ to read for meaning in order to answer the question that follows.

In the example just offered, the technique is known as non-cumulative or moving window. This is because the participant reads one word or one phrase at a time, each one disappearing as the participant moves on to the next segment in the sentence. This technique pushes the participant to keep what was just read in working memory as sentence structure is built during reading. What is normally the case in non-cumulative self-paced reading is that particular segments are targeted for places where anomalies or ungrammaticalities occur. These are called regions of interest. The idea is that at a particular point, the reader slows down as something anomalous is detected (Mitchell and Green, 1978). Researchers then compare the reading times on good vs. bad sentences and if such anomalies are detected during reading, the researcher can infer that the participant was responding to the ungrammaticality.

Research in L2 acquisition from a typological framework has tended to rely on either production data or off-line measures. In the present study, we have adopted the on-line measure of self-paced reading. First, because we tested ab-initio learners of Latin who spent less than an hour in our lab, they cannot produce language in a spontaneous manner, the preferred method for gathering oral data in language acquisition. Second, as noted above, off-line measures invite the criticism of not tapping underlying knowledge. Because of its focus on reading for meaning and because it has been used successfully in previous L2 research, self-paced reading allows the researcher more confidence in inferring that underlying knowledge is tapped during comprehension (again, for discussion, see Mitchell, 2004, as well as Jegerski, 2014).

6 Sentence comprehension

Research in the L2 context has shown a strong tendency for learners in the early phases of acquisition to assume the first noun or pronoun encountered is the subject/agent of a sentence (e.g. VanPatten, 1996, 2015; VanPatten et al., 2013). In the case of Latin, which allows sentence scrambling, this is not a problem for SVO and SOV sentences but is a problem for OSV sentences, for example. Thus, if learners are applying a ‘first-noun strategy’ when beginning to learn Latin, they might miscomprehend something like Ursum tigris amat (bear-acc tiger-nom loves) as ‘The bear loves the tiger.’ For this reason, it is not enough to only measure reading times of grammatical and ungrammatical sentences to see if learners develop sensitivity to case marking. It is also prudent to determine whether learners can make use of case marking during sentence comprehension. That is, can they correctly interpret something like Ursum tigris amat as ‘The tiger loves the bear’? If learners demonstrate both sensitivity to case marking violations while reading and also demonstrate the ability to use case marking for the comprehension of non-canonical sentences, then we would have additional (and stronger evidence) for any effects of word order on the acquisition of case marking.

III Research questions

Given the previous discussion, we designed a study to address the following questions:

Do learners who are exposed to only SOV sentences in Latin demonstrate sensitivity to violations of nominative and accusative case marking on a self-paced reading test?

Do learners who are exposed to only SVO sentences in Latin demonstrate sensitivity to violations of nominative and accusative case marking on a self-paced reading test?

Do learners who are exposed to only SOV sentences in Latin demonstrate the ability to use case marking to comprehend non-canonical sentences (i.e. who did what to whom)?

Do learners who are exposed to only SVO sentences in Latin demonstrate the ability to use case marking to comprehend non-canonical sentences (i.e. who did what to whom)?

Based on the research on typology of languages as well as the incidental results of the VanPatten and Smith study, we hypothesized that learners in the SOV group would show evidence of sensitivity to case marking by slowing down on ungrammatically case-marked sentences such as (5)–(8) above compared with correctly case-marked sentences as in (1)–(4). We further hypothesized that learners in the SVO group would not show evidence of sensitivity to case marking and there would be no reading time differences in their grammatical and ungrammatical sentences. For sentence comprehension, we hypothesized that learners in the SOV group would be able to use case marking to comprehend sentences regardless of word order. However, the SVO group should not be able to use case marking to comprehend sentences regardless of word order. In short, we expected to see sensitivity to case violations and ability to comprehend various word orders in the SOV group whereas we expected null results for the SVO group.

IV Method and procedure

1 Participants

A total of 54 people participated in this study. Participants were recruited from beginning level Spanish language classes at Michigan State University. All participants were monolingually raised, native English speakers. Participants had had no previous exposure to an SOV language, and no participants were native speakers of or students of a morphological case-marking language.

2 Materials

The researchers created all of the materials for this study. The study consisted of two phases: the treatment phase and the test phase. Materials for each phase are discussed in turn.

a Treatment phase

The treatment phase was designed to provide participants with input in Latin in which word order was manipulated as a condition. As discussed above, although Latin was a canonically SOV language, it also allowed SVO word orders and it marked case on nouns. Because we were interested in the hypothesized relationship between case marking and word order, one treatment group was exposed to SOV sentences exclusively (n = 29); the other was exposed to SVO exclusively (n = 25). In order not to overwhelm our participants with input on a lot of different structures, each input treatment section contained 100 sentences. Each sentence contained a subject, an object, and a verb. Nouns and verbs were chosen so that they were plausibly reversible; in other words, they were chosen so that either of the two nouns could plausibly be either the agent or the patient of the verb. The master list of materials contained a set of verbs – including verbs like see, hit, hate, hear, and greet – and a set of nouns. All of the nouns were names of animals, such as bee, bear, bull, dog, and donkey. Half of the nouns were masculine and the other half were feminine.

These nouns and verbs were combined to form two master lists of 100 sentences each for the treatment section. The first list consisted of 100 SOV sentences and the second list consisted of 100 SVO sentences. The lists were identical except for word order as illustrated below. Sentences were grouped in sets of five, and each set of five sentences was constructed to control how much new vocabulary participants were exposed to in each set. For example, the first set of five sentences for the SOV introduced five masculine nouns and one verb, as in (9).

(13) a. Ursus taurum amat hic.

‘The bear loves the bull.’

b. Taurus camelum amat hic.

‘The bull loves the camel.’

c. Camelus crocodilum amat hic.

‘The camel loves the crocodile.’

d. Crocodilus asellum amat hic.

‘The crocodile loves the donkey.’

e. Asellus ursum amat hic.

‘The donkey loves the bear.’

For the SVO group, the first set was identical except for word order:

(14) a. Ursus amat taurum hic.

‘The bear loves the bull.’

b. Taurus amat camelum hic.

‘The bull loves the camel.’

c. Camelus amat crocodilum hic.

‘The camel loves the crocodile.’

d. Crocodilus amat asellum hic.

‘The crocodile loves the donkey.’

e. Asellus amat ursum hic.

‘The donkey loves the bear.’

The second set introduced five feminine nouns and a second verb. The third set recycled the masculine and feminine nouns from the first two sets of sentences and introduced a third verb. The first noun (the subject) was always marked with nominative case, and the second noun (the object) was always marked with accusative case. We note that each sentence ended in the word hic ‘here’ meaning ‘in this drawing’. This was done because in another experiment with Arabic in which we varied word order as SVO and VSO, sentences required comparability across languages. For the second noun in Arabic to have case marking, a word needed to follow. We settled on ‘here.’ (We will be reporting on our work on Arabic in a future article.)

Each sentence was accompanied by a black and white drawing illustrating the scene. All sentences were audio recorded by an adult native English speaker who had studied Latin and participants heard each sentence as they simultaneously saw each drawing. In addition, each drawing was accompanied by a written caption of the sentence. Thus, learners had simultaneous aural and written input for each treatment sentence and drawing.

Each set of sentences was followed by a three-item ‘quiz’ to ensure that participants were learning vocabulary and paying attention to the meanings of the sentences. These quizzes consisted of an audio recorded sentence presented with two drawings. Participants were asked to pick the picture that matched the recording; for a sample set and follow up ‘quiz’, see Appendix 1. These ‘quizzes’ did not focus on grammar or case endings. In order to be successful on each ‘quiz’, participants only needed to focus on vocabulary.

b Testing phase: Self-paced reading

The treatment task was immediately followed by a self-paced reading task designed to test participants’ sensitivity to case marking. Self-paced reading was used because it is considered less metalinguistic than offline measures like grammaticality judgment tasks. In a self-paced reading task, participants are pushed to read for meaning because stimulus sentences are followed by some type of comprehension question related to the content of what was just read (Jegerski, 2014).

Sentences for the self-paced reading task were constructed in the following way. First, we decided to bias each group for the type of sentence it was exposed to; that is, the SOV group would read test sentences that were all SOV and the SVO group would read test sentences that were all SVO. For the SOV group, twelve SOV sentences using the vocabulary learned during the treatment were developed with six sentences being grammatical (see example 15a below) and six sentences being ungrammatical. Of the six ungrammatical sentences, three were ungrammatical for having both nouns marked with nominative case (see 15b below) and the other three having both nouns marked with accusative case (see 15c below). For the SVO group, the same sentences were used as for the SOV group but all were SVO word order; see (16a)–(16c) below.

(15) a. Ursus tarum amat.

‘The bear loves the bull.’; SOV grammatical

b. * Ursus taurus amat.

Two nominative NPs; SOV ungrammatical

c. * Ursum taurum amat.

Two accusative NPs; SOV ungrammatical

(16) a. Ursus amat taurum.

‘The bear loves the bull.’; SOV grammatical

b. * Ursus amat taurus.

Two nominative NPs; SOV ungrammatical

c. * Ursum amat taurum.

Two accusative NPs; SOV ungrammatical

Sentences were blocked and counterbalanced across three lists so that each participant only saw one version of the sentence. Each sentence was followed by a question in English that asked if a particular animal was named. For example, for Taurus tigrem audit hic ‘The bull hears the tiger’ the comprehension question was ‘Was this animal just named in the sentence? Tiger.’ The participant then would press one of two labeled buttons to indicate and answer: YES NO. The questions were balanced so that 50% had YES answers and 50% had NO answers. Although it is typical in self-paced reading to ask questions aimed more at global meaning of the sentences, given the short nature of the sentences (two nouns and a verb), such comprehension questions were not possible. We instead opted for a question that asked for the recall of a particular entity (noun) in the sentence. Had we asked for participants to indicate, say, who did what to whom as a comprehension question (e.g. Is this what you read? ‘The bear loves the bull’ YES NO), we might have inadvertently pushed participants to focus on case marking.

Test sentences were presented one word at a time, such that participants had to press a button to move through the sentence. For example, Taurus tigrem audit hic was rendered as –— –— –— –— and each button press revealed one word while the previous word disappeared. (See the description above of non-cumulative self-paced reading.) All reading times were recorded by SuperLab.

c Testing phase: Sentence comprehension

After the self-paced reading test, participants took a sentence comprehension test that consisted of hearing a sentence in Latin of varying word order: SOV, SVO, OVS. As an example, learners might hear one of the three sentences below, all meaning the tiger loves the bear:

(17) SOV: Tigris ursum amat.

(18) SVO: Tigris amat ursum.

(19) OSV: Ursum amat tigris.

Simultaneously, the participant would see two pictures, one in which the tiger is shown loving the bear and one in which the bear is loving the tiger (for sample drawing, see Appendix 1). One picture was labeled ‘A’ and the other ‘B’. The participant pressed an A or B button on the response pad to indicate the picture that matched what was heard.

Three blocks of 18 sentences were constructed for the comprehension test: six SOV sentences, six SVO sentences, and six OSV sentences. All of the nouns and verbs used were the same as those used in the treatment. The blocks did not repeat the sentences from another block and within each block, no sentence repeated the same two animals doing the same action (e.g. a participant heard only one of the sentences in 13–15 above in a block). Sentences were randomized in each block and the blocks were randomized among the various participants for counterbalancing.

It is important to note that all participants heard all word orders in this test. The real comparison, however, is between the canonical sentences of the treatment and the non-canonical OSV that neither group was exposed to. In short, SOV vs. OSV for the SOV group and SVO vs. OVS for the SVO group.

3 Procedure

Participants were tested individually in the research laboratory. Participants read an information sheet and proceeded to the first phase of the experiment. Both the input treatment section and the test section were presented using SuperLab 5.0. Participants completed the input treatment section first. They moved through the input treatment section at their own pace, pressing the middle button on a five button array on a response pad to move through each sentence, and pressing a green A (for drawing labeled A) or B (for drawing labeled B) button to answer ‘quiz’ questions. On average, participants took 30 minutes to complete the treatment section.

The assessment portion followed the treatment. Again, participants worked at their own pace. They read a set of instructions that explained the self-paced reading procedure, and then completed the test. They read the test sentence on the screen, pressed any button to advance to the comprehension question, and again pressed a green A (YES) or B (NO) button to answer each question. For the comprehension test, the participant would see two pictures (e.g. one in which the tiger is shown loving the bear and one in which the bear is loving the tiger). One picture was labeled ‘A’ and the other ‘B’. The participant pressed an A or B button on the response pad to indicate the picture that matched what was heard. On average, the test section took about 15 minutes. All participants completed the entire experiment (treatment and test) within an hour. Participants were paid $20 for their time.

4 Scoring and analysis

The ‘quizzes’ from the treatment section were scored first to make sure that all participants were learning vocabulary. The average score on these quizzes was 54.5/60 (91%), indicating that participants were paying attention during the learning phase and had learned the vocabulary they were exposed to.

The results of the assessment were tabulated next. A participant’s reading time for each region of a sentence was extracted from the SuperLab files. Reading times on individual regions of each sentence were then summed to get a whole sentence reading time. Because there were 12 grammatical versions for each type of sentence (i.e. SOV, SVO) we combined the scores for the six ungrammatical sentences in which two nominative case endings occurred and the six ungrammatical sentences in which two accusative case endings occurred. This procedure allowed for a balanced comparison across grammatical and ungrammatical sentences: 12 each. We used whole sentence reading times because of the difference between SOV and SVO word order test sentences. That is, we could not compare reading times on, say, the O of the SOV sentences with the reading times of O on the SVO sentences because these occurred in different locations. In particular, the O of SVO sentences appeared toward the end of each sentence. Self-paced reading research avoids putting specific targets and spillover regions at the end of a sentence due to general sentence wrap-up effects (Jegerski, 2014; Mitchell and Green, 1978), a finding corroborated by research using other kinds of online processing measures such as eye-tracking (Just and Carpenter, 1980).

It is common in research using self-paced reading to trim data for outliers or extreme scores. Although this practice has come under increased critical scrutiny and it is not clear that trimming the data for outliers actually results in different outcomes or may actually compromise data (e.g. Bakker and Wicherts, 2014), we decided to trim outliers in as per standard procedure in the field of L2 research using on-line measures. However, data trimming in L2 research is not uniform and varies across studies. Some researchers replace outlying means with group means (e.g. Hopp, 2006). Other researchers remove outlying scores altogether (e.g. Jackson and Dussias, 2009). For the present study, we chose to remove outlying scores. What is more, L2 researchers vary as to the measurement they use for dealing with outliers: +/–2 SD and +/–2.5 SD. Following the suggestions in Baayen and Milin (2010) and used in Hopp (2006), we used the most conservative method and trimmed the data by removing those participants whose mean scores were +/–2.5 SD from the mean. The result was that for the SOV group, 1.3% of the scores were removed. For the SVO group, data trimming resulted in the same amount of scores removed: 1.3%.

The comprehension test was scored using a 1 or 0 system. If the learner selected the correct picture that matched what was heard, the score was 1. If the selection was incorrect, the score was 0. The range for each set of sentences (i.e. SOV, SVO, OSV) was 0–6. Thus each participant received three different comprehension scores, one for each word order.

V Results

1 Reading times

Mean cumulative reading times and standard deviations for the SOV and SVO groups appear in Table 1. For the SOV groups, there is a mean difference of just over 575 ms between the grammatical and ungrammatical sentences, suggesting the ungrammatical sentences took the participants longer to read. For the SVO group, there was a mean difference of about 389 ms, also suggesting it took these participants longer to read the ungrammatical sentences.

To address our two research questions, we submitted the results to paired samples t-tests.⁴ The results were as follows:

for the SOV, group the difference between means was significant, t(28) = −3.215, p = .003

for the SVO group, the difference between means was not significant, t(24) = −1.150, p = .26

Based on the results just presented, the answers to our first research question is yes: overall, learners in the SVO group showed sensitivity to violations of case marking on the self-paced reading task. As a group, they took longer to read the ungrammatical sentences compared to the ungrammatical sentences. For the second question, the answer is no: overall, the SVO group did not show sensitivity to violations of case marking.

Table 1.

Trimmed mean and standard deviations for cumulative reading times for SOV and SVO groups in milliseconds.

Group	Grammaticality	Mean reading time	Standard deviation
SOV	Grammatical	4451.19	1132.24
	Ungrammatical	5026.24	1484.16
SVO	Grammatical	4448.25	2094.86
	Ungrammatical	4837.33	1665.52

2 Sentence comprehension

The results for sentence comprehension (i.e. matching what is heard to a picture to indicate who did what to whom) appear in Table 2. Paired samples t-tests were performed on the canonical vs. non-canonical sentences for each group. Results for the SOV group were as follows: SOV vs. OSV sentences, t(28) = 10.22, p < .001. These results demonstrate that the SOV group comprehended SOV sentences better than OSV sentences. Given the very low scores for OSV sentences, the results suggest that the participants in this group tended not to use case morphology on nouns to determine who did what to whom. Thus, the answer to our third research question is ‘no’: overall, learners in the SOV group did not show evidence of using case marking for comprehension of non-canonical word order.

Table 2.

Mean scores and standard deviations for sentence comprehension test.

Group	Sentence type	M	S
SOV	SOV	4.28	1.03
	OSV	1.45	0.95
SVO	SVO	5.44	1.08
	OSV	1.80	1.61

For the SVO group, the results were as follows: SVO vs. OSV sentences, t(24) = 7.14, p < .001. Similar to what was just presented, these results demonstrate that for this SVO group, the participants comprehended SVO sentences better than OSV sentences. Once again, given the very low scores for OSV sentences, the results suggest that the participants in this group tended not to use case morphology on nouns to determine who did what to whom in. Thus, the answer to our fourth research question is ‘no’: overall, learners in the SVO group did not show evidence of using case marking for comprehension of non-canonical word order.

VI Discussion and conclusions

The results presented in Section V offer what look to be contradictory results. On the one hand, the test of on-line sensitivity to violations of case marking among Latin ab-initio learners suggests that only the SOV group, as a whole, demonstrated such sensitivity. The SVO group did not. Sensitivity to case marking is inferred based on whether or not readers slowed down on ungrammatical sentences. Because case marking was the only thing manipulated in the self-paced reading task (e.g. Tigris ursum amat hic vs. Tigris ursus amat hic) the conclusion is that reading times are a reflection of sensitivity to case marking.

On the other hand, the sentence comprehension data suggests that neither group was able to use case-marking morphology to comprehend non-canonical sentences; in this case, OSV sentences. Thus, word order within the confines of this study does not seem to significantly affect early-stage learners’ ability to use case marking morphology to comprehend sentences regardless of their sensitivity to morphological violations regarding case. Instead, it appears that the participants in this study, regardless of group, tended to rely on the First-noun Principle (VanPatten, 1996, 2015) by which learners tend to interpret the first noun (or pronoun) as the subject/agent of the sentence. The reader will recall that the sentences that tested ability to deploy case in real-time comprehension were OSV sentences. In such sentences, the object has moved from its canonical spot to occupy a higher position, making it the first NP that a learner encounters. The First-noun Principle as an interpretation strategy may require much more input to be overcome. That is, the First-noun Principle is a very strong principle and takes much more than just 25 minutes or 100 sentences to be weakened such that we see a shift toward deployment of morphology over word order in comprehension.

In spite of the difference between groups on the reading test and the non-difference on the sentence comprehension test, the results may not be contradictory after all. We have seen in other kinds of research, namely generative, that learners may demonstrate underlying knowledge of a morphosyntactic feature of a language but be unable to deploy that knowledge during real-time communication (both comprehension and production). Known as the Missing Surface Inflection Hypothesis (Prévost and White, 2000), the idea is that new functional features or projections may be in place but learners have difficulty with mapping between the syntax and the functional mental lexicon. In the present study, the suggestion would be that even after just 100 sentences, learners in the SOV group may have formulated some kind of new agreement category for case and thus a possible new functional projection (e.g. a KP). They have thus become sensitive to this new projection needing to have features checked. Alternatively, the majority of learners in the SOV group could have become sensitive to nominal inflections such that their internal systems were noting that nouns ended in -s and -m and that at no time in a given sentence did both nouns end in the same ending. That is, two -s and two -m in the same sentence simply seemed ‘weird’. The SOV group’s reading time differences, then, could be due to having picked up on this and not on any new functional feature. However, the two explanations may not be unrelated. A first step in getting case in a language like Latin may involve the internal system noting the distribution of noun endings followed by a step in which new functional features are projected into the syntax.

Based on the self-paced reading test, we have at least tentative evidence for typological word order and its correlation with case marking that aids acquisition in SOV languages, especially if we compare the present study with that found in Japanese early-stage learners in VanPatten and Smith (2015). However, the key word is ‘tentative’. Aside from the fact that research such as the present study needs to be replicated under other conditions, it is also the case that at least some learners in the SVO group demonstrated sensitivity to case marking violations in the self-paced reading part of the study. The reader may remember that there was a mean increase in reading time of 389 ms from correctly case-marked sentences to incorrectly case-marked sentences for the SVO group. Some of the participants in this group may have had ‘significantly’ longer reading times on the ungrammatical sentences. The problem is that this number of participants was not large enough and/or the reading times themselves were not large enough for the analyses to demonstrate a significant effect for case marking under the SVO condition.

Before continuing we would like to address one comment made by an anonymous reviewer. That reviewer wondered whether our results really speak to the question of whether learners acquired case marking, or whether they just reflect simple processing behavior. First, in most L2 frameworks, processing of input is seen as an integral part of acquisition or it is assumed to happen in one way or another. Learners must process input in order to get data into the internal system. That we used a processing measure to see if case marking was making its way into the system should not be a problem because processing measures are increasingly being used to determine what’s in the learner’s internal system (see our previous discussion prior to presentation of the study itself). Second, because we have two measures for what learners know and can do (i.e. the self-paced reading task and the sentence comprehension task) we have a better picture of what they might have actually acquired. For this reason, as explained in the first paragraphs of this section, we suggest that learners are developing sensitivity to case marking and might even have ‘penciled in’ the difference between -s and -m in their grammars but have difficulty deploying case marking for comprehension when confronted with non-canonical sentences. In short, our claim is not that learners have acquired case marking but that SOV word order gives learners an edge compared to SVO order as they begin the acquisition of case marking.

Another issue to address is L1 transfer. In particular, could it be that the SVO group lagged behind because their input sentences matched English canonical word order? That is, when confronted with only SVO sentences, participants in the SVO group simply transferred their L1 processing routines to the new input. Because English is SVO and does not mark NPs for case, the result would be a tendency to overlook case marking on nouns in Latin SVO sentences. Even if L1 transfer is part of the explanation for what happened in the SVO group, we note here that the SOV group seemed to move beyond the transfer of L1 processing routines. This result only reinforces the idea that word order typology plays a role in something like picking up case marking. In the present study, SOV word order seemed to quickly push learners in the SOV group toward sensitivity to case marking.

Overall, we think our results suggest that at least some of typological generalizations discussed in the background section have implications for SLA. These results suggest that when quantity and quality of input is held constant – the only difference in the input between the two groups was the basic word order – word order influences the extent to which participants become sensitive early on to case marking in the input they receive. Participants who are exposed to SOV sentences are more likely to be sensitive to ungrammatical case marking than participants who are exposed to SVO input are. The question of course is why. As suggested in the background section of this article, the human language processor assigns syntactic roles to nouns (noun phrases) as soon as possible (see also Pritchett, 1992). If the processor has to wait until the end of the sentence to encounter the verb and then the verb offers no lexical semantic restrictions on which noun could be the subject and which could be the object (as was the case in the present study), the processor may begin to look for surface clues to assign syntactic roles. Thus, the greater sensitivity to case marking in the SOV group in this study could be a result of ‘processing pressure’ and the need to compute sentence structure as efficiently as possible. This discussion is, of course, speculative and the present study cannot directly say that processing pressure is the driving force in accounting for the differential results obtained in our study; again, if we consider that at least some of the participants in the SVO group demonstrated sensitivity to case violations during reading. However, because of the previous literature on SOV languages and processing (e.g. see the discussion in Ueno and Polinsky, 2009), we think this is the most likely case.

At the same time, the differential results in our study do not mean that participants who were exposed to SVO input cannot acquire case marking; it means that they might take longer to do so. This may be particularly true of case marking in a language like Latin compared to Japanese, something we alluded to in Section II. In our previous work on initial stage Japanese (VanPatten and Smith, 2015), we found 100% of the participants demonstrated sensitivity to case marking. However, as we pointed out early in the present article, Japanese case marking is syllabic and consists of non-inflected forms; heads of case marking phrases. Latin case marking, on the other hand, is non-syllabic (at least as far as nominative and accusative are concerned), is inflectional, and varies depending on declension. Thus the learning burden may be considered greater in Latin SOV compared to Japanese SOV and this would explain the 15% difference in total number of participants who showed sensitivity to case marking between the two studies (i.e. VanPatten and Smith, 2015 = 100%; the present study = 85%, when focusing just on SOV).

Before concluding we would like to draw the reader’s attention to previous work in instructed SLA with ab-initio learners of Latin, namely the work by Cristina Sanz and her students and colleagues (e.g. Lenet et al., 2011; Stafford et al., 2011). In this line of research, naive learners of Latin receive a treatment that directs their input processing to case endings based on VanPatten’s processing instruction and the First-noun Principle (e.g. VanPatten, 2004, 2015; VanPatten and Cadierno, 1993). The focus of this line of research has been on the roles of explicit learning/training and feedback type as well as factors such as age and working memory in an instructed situation. This research stands in contrast to the present research in that in our study, no explicit attempt to push learners to attend to case marking in the treatment occurred and our focus is not on explicit (or less explicit/implicit) learning and teaching, feedback, or any individual differences that might interact with these instructional factors. One conclusion from our research (both here and in our 2015 study on ab-initio learners of Japanese L2) is that consistent and constant exposure to simple SOV sentences likely pushes naive learners of any languages like Latin (e.g. Japanese, Turkish) to become sensitive to case. It could be that processing instruction speeds up the process but it seems that something in the mind/brain of humans may naturally push learners toward ‘looking for case marking’ in an SOV language without instructional intervention.

So far, typological work in SLA research has been confined to typological markedness. Our results indicate that SLA researchers might want to consider whether typological generalizations could be a starting place for investigating the ways in which the mind filters or processes input data to create intake. Our results suggest that one property of the input – word order, in this case – may influence how quickly learners attend to other properties of language. That typological generalizations are not always universals in the strict sense of the term should not be reason to overlook investigation into how they might influence SLA. It’s possible that these generalizations are reflections of the ways the mind tends to organize language. And that is certainly of interest to SLA research.

Footnotes

Appendix 1

Sample set of learning SOV items [Note: SVO items were identical except for word order]

“Quiz” that followed first set of learning items in SOV treatment [Note: SVO items were identical except for word order]

Acknowledgements

We thank the College of Liberal Arts and Sciences for the Humanities and Arts Research Program (HARP) grant awarded to Bill VanPatten to conduct this research. We thank Jeffrey Maloney and Charles Moulding who played critical roles in data gathering. We thank the anonymous reviewers and the editors for their comments and suggestions. All errors remain with the authors.

Declaration of Conflicting Interest

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Notes

References

Baayen

Milin

(2010) Analyzing reaction times. The International Journal of Psychological Research 3: 12–28.

Baker

(1996) The polysynthesis parameter. Oxford: Oxford University Press.

Baker

(2001) The atoms of language: The mind’s hidden rules of grammar. New York: Baker Books.

Bakker

Wicherts

(2014) Outlier removal and the relation with reporting errors and quality of psychological research. PLOS One. Available at: http://doi.org/10.1371/journal.pone.0103360 (accessed June 2018).

Comrie

(2007) The acquisition of relative clauses in relation to language typology. Studies in Second Language Acquisition 29: 301–09.

Dryer

(2002) Case distinctions, rich verb agreement, and word order type. Theoretical Linguistics 28: 151–57.

Eckman

(1977) Markedness and the contrastive analysis hypothesis. Language Learning 27: 315–30.

Eckman

(1984) Universals, typologies and interlanguages. In: Rutherford

(ed.) Language universals and second language acquisition. Philadelphia, PA: John Benjamins, pp. 79–105.

Eckman

(1985) Some theoretical and pedagogical implications of the Markedness Differential Hypothesis. Studies in Second Language Acquisition 7: 289–307.

10.

Eckman

Moravcsi

Wirth

(1989) Implicational universals and interrogative structures in the interlanguage of ESL learners. Language Learning 39: 173–205.

11.

Ellis

(2005) Measuring implicit and explicit knowledge of a second language: A psychometric study. Studies in Second Language Acquisition 27: 141–72.

12.

Foote

(2011) Integrated knowledge of agreement in early and late English-Spanish bilinguals. Applied Psycholinguistics 32: 187–220.

13.

Gass

(1979) Language transfer and universal grammatical relations. Language Learning 29: 327–44.

14.

Gilligan

(1987) A cross-linguistic approach to the pro-drop parameter. Unpublished PhD thesis, University of Southern California, Los Angeles, CA, USA.

15.

Greenberg

(1963) Some universals of grammar with particular reference to the order of meaningful elements. In: Greenberg

(ed.) Universals of Language. Cambridge, MA: MIT Press, pp. 73–113.

16.

Hinds

(1982) Case marking in Japanese. Linguistics 20: 541–57.

17.

Hopp

(2006) Syntactic features and reanalysis in near-native processing. Second Language Research 22: 369–97.

18.

Hulstijn

(2005) Theoretical and empirical issues in the study of implicit and explicit language learning. Studies in Second Language Acquisition 27: 129–40.

19.

Hyltenstam

(1984) The use of typological markedness conditions as predictors in second language acquisition: the case of pronominal copies in relative clauses. In: Andersen

(ed.) Second Languages: A cross-linguistic perspective. Rowley, MA: Newbury House, pp. 39–58.

20.

Jackson

Dussias

(2009) Cross-linguistic differences and their impact on L2 sentence processing. Bilingualism: Language and Cognition 12: 65–82.

21.

Jegerski

(2014) Self paced reading. In: Jegerski

VanPatten

(eds) Research methods in second language psycholinguistics. New York: Routledge, pp. 20–49.

22.

Jegerski

VanPatten

(eds) (2014) Research methods in second language psycholinguistics. New York: Routledge.

23.

Just

Carpenter

(1980) A theory of reading: From eye-fixations to comprehension. Psychological Review 87: 329–354.

24.

Keenan

Comrie

(1977) Noun phrase accessibility and Universal Grammar. Linguistic Inquiry 8: 63–99.

25.

Koopman

(2005) Korean (and Japanese) morphology from a syntactic perspective. Linguistic Inquiry 36: 601–33.

26.

Lenet

Sanz

Lado

Howard

(2011) Aging, pedagogical conditions, and differential success in SLA: An empirical study. In: Sanz

Leow

(eds) Implicit and explicit language learning. Washington, DC: Georgetown University Press, pp. 73–84.

27.

Mitchell

(2004) On-line methods in language processing: Introduction and historical review. In: Carreiras

Clifton

(eds) The on-line study of sentence comprehension: Eyetracking, ERPs and beyond. New York: Psychology Press, pp. 15–32.

28.

Mitchell

Green

(1978) The effects of context and content on immediate processing in reading. Quarterly Journal of Experimental Psychology 30: 609–36.

29.

Miyamoto

(2002) Case markers as clause boundary introducers in Japanese. Journal of Psycholinguistic Research 31: 307–47.

30.

Pavesi

(1986) Markedness, discoursal models, and relative clause formation in a formal and an informal context. Studies in Second Language Acquisition 8: 38–55.

31.

Prevost

White

(2000) Missing Surface Inflection or Impairment in second language acquisition? Evidence from tense and agreement. Second Langauge Research 16: 103–133.

32.

Pritchett

(1992) Grammatical competence and parsing performance. Chicago, IL: University of Chicago Press.

33.

Shirai

Ozeki

(2007) Introduction. Studies in Second Language Acquisition 29: 155–67.

34.

Stafford

Wood Bowden

Sanz

(2011) Optimizing language instruction: Matters of explicitness, practice, and cue learning. Language Learning 62: 741–68.

35.

Ueno

Polinsky

(2009) Does headedness affect processing? Another look at the VO–OV contrast. The Journal of Linguistics 45: 675–710.

36.

VanPatten

(1996) Input processing and grammar instruction: Theory and research. Norwood, NJ: Ablex.

37.

VanPatten

(2004) Processing instruction: Theory, research and commentary. Mahwah NJ: Lawrence Erlbaum.

38.

VanPatten

(2015) Input processing in adult SLA. In: VanPatten

Williams

(eds) Theories in second language acquisition. 2nd edition. New York: Routledge, pp. 113–34.

39.

VanPatten

Cadierno

(1993) Explicit instruction and input processing. Studies in Second Language Acquisition 12: 225–43.

40.

VanPatten

Smith

(2015) Aptitude as grammatical sensitivity and the initial stages of learning Japanese as an L2: Parametric variation and case marking. Studies in Second Language Acquisition 37: 135–65.

41.

VanPatten

Keating

Lesser

(2012) Missing verbal inflections as a representational issue: Evidence from on-line methodology. Linguistic Approaches to Bilingualism 2: 109–40.

42.

VanPatten

Borst

Collopy

Qualin

Price

(2013) Explicit information, grammatical sensitivity, and the First-noun Principle: A cross-linguistic study in processing instruction. The Modern Language Journal 92: 506–27.

43.

Yamashita

(1999) The effects of word-order and case marking information on the processing of Japanese. Journal of Psycholinguistic Research 26: 163–88.

44.

Yamashita

(2000) Structural computation and the role of morphological markings in the processing of Japanese. Language and Speech 43: 429–59.