Department of English
Master Degree Project English Linguistics
Spring 2021
Grammar “bores the crap out of me!”:
A mixed-method study on the XTYOFZ construction and its usage by ESL and ENL speakers
Nok Chin Lydia Chan
Grammar “bores the crap out of me!”:
A mixed-method study on the XTYOFZ construction and its usage by ESL and ENL speakers
Abstract
Different from Generative Grammar which sees grammar as a formal system of how words are put together to form sentences, Construction Grammar suggests that grammar is more than just rules and surface forms; instead, grammar includes many form-and- meaning pairings which are called constructions. For years, Construction Grammarians have been investigating constructions with various approaches, including corpus- linguistics, pedagogical, second language acquisition and so on, yet there is still room for exploration. The present paper aims to further investigate the [V the N
taboo-wordout of]- construction (Hoeksema & Napoli, 2008; Haïk, 2012; Perek, 2016; Hoffmann, 2020) (e.g., I kick the hell out of him.) and propose a new umbrella construction, “X the Y out of Z” (XTYOFZ) construction, for it. Another aim is to examine the usage and comprehension of the XTYOFZ construction by English as a Second Language (ESL) and English as Native Language (ENL) speakers. The usage context, syntactic and semantic characteristics of the XTYOFZ construction were examined through corpus linguistic methodology. Furthermore, processing and understanding of the construction by ESL and ENL speakers were tested via an online timed Lexical Decision Task as well as an online follow-up survey consisting of questions on English acquisition and usage, and a short comprehension task on the XTYOFZ construction. Corpus data shows that in general, the combination of non-motion action verbs (e.g., scare, beat) as X and taboo terms (e.g., shit, hell) as Y was the most common. Also, it was found that the construction occurs mostly in non-academic contexts such as websites and TV/movies. On the other hand, results from the Lexical Decision Task show that ESL speakers access constructional meaning slightly more slowly than ENL speakers. The follow-up survey also reflects that ESL speakers seem to have a harder time to produce and comprehend the construction compared to ENL speakers. By investigating the features of a relatively less-discussed construction and its usage by ESL speakers, this study hopes to increase the knowledge base of Construction Grammar and ESL construction comprehension and usage, particularly on the constructions that are mainly used in more casual settings.
Keywords
construction grammar, constructions, corpus linguistics, ESL, processing, priming,
Jabberwocky sentences.
Contents
1. Introduction ... 1
2. Literature Review ... 3
2.1 Constructions ... 3
2.1.1 Identifying constructions ...4
2.1.2 Relevant constructions to the XTYOFZ construction ...6
2.2 Applied Construction Grammar ... 8
2.2.1 Effects of ESL and EFL context on constructions usage ...8
2.2.2 ESL learners’ production of the verb-argument constructions ...9
2.3 Priming with constructions ... 10
2.3.1 Automatic accessing of constructional meaning by ENL speakers ... 10
2.3.2 Constructional priming in ESL learners ... 12
3. Methodology ... 13
3.1 Corpus Linguistics ... 13
3.2 Timed Lexical Decision Task ... 14
3.2.1 Pilot Study ... 14
3.2.2 Participants ... 14
3.2.3 Materials ... 15
3.2.4 Procedure ... 17
3.3 Follow-up survey on construction usage comprehension ... 20
3.3.1 Participants ... 20
3.3.2 Materials and Procedure ... 20
4. Results ... 21
4.1 Corpus study on the XTYOFZ construction ... 21
4.1.1 Search 1: Frequencies of the words used as X, Y and Z ... 21
4.1.2 Search 2: Frequency of XTYOFZ construction across genres ... 26
4.2 Timed Lexical Decision Task ... 27
4.2.1 General performance of the participants ... 27
4.2.2 RTs for different groups of target words ... 28
4.3 Follow-up Survey on construction usage and comprehension ... 30
4.3.1 Usage/ Productivity... 31
4.3.2 Comprehension ... 34
5. Discussion ... 35
5.1 Corpus study on the XTYOFZ construction ... 35
5.1.1 Syntactic characteristics ... 35
5.1.2 Semantic characteristics ... 36
5.1.3 Usage Context ... 37
5.2 Reasons for why XTYOFZ is a construction ... 38
5.2.1 Canonical pattern ... 38
5.2.2 Non-compositional meaning ... 39
5.2.3 Idiosyncratic constraints ... 40
5.2.4 XTYOFZ construction as an “umbrella” construction ... 41
5.3 Timed Lexical Decision Task ... 43
5.4 Follow-up survey on construction usage and comprehension ... 45
6. Conclusion ... 46
References ... 49
Appendix A ... 51
Appendix B ... 52
Appendix C ... 55
Appendix D ... 59
Appendix E ... 60
1. Introduction
“Grammar bores the crap out of me!” might sound familiar to many English teachers as it might be true for many English learners who learn English grammar in a traditional way. For instance, in many Asian countries where English is spoken as a second language or a foreign language, grammar is taught and learnt through repetition of exercises, memorizing rules and forms, drawing tree-diagrams, and so forth. Grammar has always been a vital part of language teaching and learning, and it is particularly emphasized in traditional English as second language (ESL) and English as foreign language (EFL) teaching. To ESL and EFL learners, target language grammar is probably something complicated yet sometimes meaningless. Very often grammar is presented “through mechanical exercises, the learning of irregularities, and the memorization of seemingly endless paradigms” (Langacker, 2013, p. 3). In fact, from the Generative Grammar point of view, “grammar (syntax in particular) is an independent aspect of linguistic structure distinct from both lexicon and semantics” (Langacker, 2013, p. 1). In this view, grammar is referred to as a formal system of how words are put together in a language to form sentences. However, Construction Grammar allows us to abandon this old, “unproductive mathematical metaphor for grammar” (Ibbotson & Tomasello, 2009, p. 60). Instead of viewing grammar as a formal system of a language, Construction grammarians see
“grammar as a mental network of constructions” (Hoffmann & Trousdale, 2013, p. 3).
As Hoffmann and Trousdale (2013) explain, Construction Grammar has its roots in the Saussurean notion of the linguistic sign as an arbitrary and conventional pairing of form and meaning. Under the Saussurean notion, even though the Swedish sign äpple and its Spanish equivalent manzana have different associated conventional forms, they have the same underlying meaning ‘apple’. Nevertheless, it is suggested that “the arbitrary form- meaning pairings might not only be a useful concept for describing words or morphemes but that perhaps all levels of grammatical description involve such conventionalized form-meaning parings” (Hoffmann & Trousdale, 2013, p.1). This extended notion is known as a ‘construction’, and the linguistic approaches investigating this idea are
‘Construction Grammar’ (Hoffmann & Trousdale, 2013).
Unlike Generative Grammar which argues that there is a clear-cut division of lexicon and syntax, Construction Grammar suggests that all constructions are “part of a lexicon- syntax continuum” (Hoffman & Trousdale, 2013, p.1). From the Construction Grammar point of view, grammar is embedded with meanings. Grammar consists of more than just the surface forms, but rather is “conventionalized parings of complex forms with complex semantic/ pragmatic functions” (Ibbotson & Tomasello, 2009, p. 59). In other words, constructions are form-meaning pairs that “carry meaning, independently of the words in the sentence” (Goldberg, 1995, p. 1), and they are the basic units of language. Therefore, in the world of Construction Grammar, constructions are the only thing that speakers must know in order to know a language (Hilpert, 2014).
According to Goldberg (2006), “all levels of grammatical analysis involve constructions”
(p. 5) which vary in size and complexity. Starting from morpheme or words, idioms, to
partially lexically filled and fully general phrasal patterns, constructions can be of various
types. Some are entirely schematic, whilst some have fixed lexical elements in them. For
instance, the morpheme pre- is a construction, the idiom kick the bucket is also a construction, even the unusual linguistic pattern What’s X doing Y? such as What’s that fly doing in my soup?! (Kay & Fillmore, 1999) is also considered to be a construction.
Other examples of constructions would be the passive construction (e.g., The fee was paid.), argument structure constructions (e.g., She gave him an apple. / He threw the ball over the fence.), N P N construction (e.g., week after week, letter by letter), and so forth.
Since Construction Grammar “offers a psychologically plausible, generative theory of human language” (Hoffmann & Trousdale, 2013, p. 3), it allows many new insights in grammar and much room for exploration. Scholars have been studying different aspects of Construction Grammar with various approaches, not only the constructions themselves (e.g., Goldberg, 1995; Kay & Fillmore, 1999; Hoeksema & Napoli, 2008; Hoffmann, 2013; Hoffmann, 2020), but also areas such as the accessing of constructional meanings (e.g., Bock & Loebell, 1990; Hare & Goldberg, 1999; Johnson & Goldberg, 2013;
Baicchi, 2016), L1/L2 constructions acquisition and command (e.g., Ibbotson &
Tomasello, 2009; Gilquin, 2016; Römer, Skalicky, & Elli, 2020), as well as pedagogical construction grammar (e.g., Herbst, 2016; Ruiz de Mendoza & Agustín Llach, 2016; Sung
& Yang, 2016).
Doubtlessly, the studies mentioned above all provided much convincing evidence regarding their own research questions and brought useful insights to the field of Construction Grammar. Nevertheless, many of the previous studies focus mainly on argument structure constructions, which is a rather broad group of constructions that have received quite a lot of researchers’ attentions. There are many other less frequent constructions which are worth investigating and it is possible that there are still some undiscovered constructions. Furthermore, the participants of many of the previous studies had relatively narrow backgrounds in the way that many of the previous studies focus on either native English speakers or ESL/ EFL learners with a specific L1 background. For instance, Johnson and Goldberg (2013) focused only on native English speakers who were all undergraduate students; as for Ibbotson and Tomasello (2009), the focus fell on language acquisition of monolingual children solely; Sung and Yang (2016) focused on Korean EFL students only. There is not a lot of previous research with a particular focus on both native English speakers and ESL learners with diverse L1 backgrounds. Neither are there many existing studies investigating both the accessing of constructional meaning and the usage of constructions.
The present thesis is a mixed-method study on the “X the Y out of Z” (XTYOFZ)
construction as well as its comprehension and usage by ESL and ENL speakers. In this
study, a relatively less discussed and infrequent construction, the XTYOFZ construction
will be discussed. The XTYOFZ construction is an umbrella construction which embeds
various sub-constructions that share the same syntactic form, non-compositional meaning
and idiosyncratic constraints, but have a slight difference in one of their semantic
characteristics. One of the sub-constructions under this umbrella is B-construction
(Hoeksema & Napoli, 2008)/ Construction B (Haïk, 2012)/ “V the hell out of NP”
construction (Perek, 2016)/ the [V the N
taboo-wordout of]-construction (Hoffmann, 2020)
1. On the other hand, due to the limited scale of the study, only ESL speakers but not EFL learners were selected as subjects. Also, it was found that ESL speakers do have a better, more native-like command of some constructions (e.g., Gilquin, 2016; Römer et al., 2020) and it would be worth investigating whether that also applies to the XTYOFZ construction.
This thesis aims to fulfil the research gap mentioned and further investigate Construction Grammar with a mix-method approach. There are two main research aims. The first aim is to propose an umbrella construction, XTYOFZ construction, and examine its features.
The second aim is to examine how ENL speakers and ESL speakers of different L1 backgrounds comprehend and use the XTYOFZ construction. The current study will try and attain these objectives by answering the following research questions:
1. What are the syntactic and semantic properties, and the usage contexts of the XTYOFZ construction?
2. Why should XTYOFZ construction be considered as a construction?
3. How do ESL speakers comprehend and use the XTYOFZ construction compared to ENL speakers?
To answer research questions 1 and 2 (i.e., the first research aim), a corpus-based study with the Corpus of Contemporary American English (COCA) was performed. As to answer research question 3 (i.e., the second research aim), a timed Lexical Decision Task and a follow-up survey were conducted.
This thesis has the following structure. In Section 2, a literature review will be included to provide more information on the concept of constructions, including previous research relevant to the XTYOFZ construction, as well as on the usage and accessing of constructions by both ESL and ENL speakers. Section 3 will discuss the methodology and give a description of the data collection. Sections 4 and 5 will include the analysis and discussion of data. Finally, Section 6 will provide a final conclusion and insights for future research.
2. Literature Review
2.1 Constructions
As mentioned above, Construction Grammar does not uphold a strict lexicon-syntax distinction which is assumed by most other syntactic theories (Hoffmann, 2016). Instead, Construction Grammar emphasizes the lexicon-syntax continuum and "the role of
1 The B-construction (Hoeksema & Napoli, 2008)/ Construction B (Haïk, 2012)/ “V the hell out of NP”
construction (Perek, 2016)/ the [V the Ntaboo-word out of]-construction (Hoffmann, 2020) are different names for the same construction. This will be discussed further in Sections 2.1.2 and 5.2.4.
grammatical constructions [, i.e.] conventionalized pairings of form and function”
(Goldberg, 2006, p. 3). In other words, constructions are form-meaning pairings which constitute the basic units of grammar in the constructionist sense.
A fundamental claim of constructionist approaches is that “knowledge of language consists of a large network of constructions, and nothing else in addition” (Hilpert, 2014, p. 2). The large network of constructions is called ‘construct-i-con’, and it “contains everything that would be contained in a lexicon, but in addition to that a large number of symbolic units that are larger in size than single words” (Hilpert, 2014, p. 2). This means that everything a speaker knows “about words, about syntactic patterns, about prefixes and suffixes, about idioms, and about the intricacies of what is said and what is meant…
is to be recast as knowledge of constructions” (Hilpert, 2014, p. 2).
In addition, constructionist approaches also emphasize that “languages are learned — that they are constructed on the basis of the input together with general cognitive, pragmatic, and processing constraints” (Goldberg, 2006, p. 3). Therefore, one can say that “a construction is a generalization that speakers make across a number of encounters with linguistic forms” (Hilpert, 2014, p. 9). For instance, one would probably agree that following examples are ‘the same’:
(1) the more, the merrier (2) the older, the wiser
(3) The redder the apple, the sweeter the taste.
(4) The more you revise, the better you will do in the exam.
(5) The smaller the apartment, the less you need to clean.
The above examples are indeed ‘the same’ in the sense that they all belong to the same construction family, which is the comparative correlative construction. Through experience with language, speakers “abstract away from individual differences between sentences... [and] form a generalization” (Hilpert, 2014, p. 9). It is this generalization that Construction Grammarians focus on when they talk about constructions.
2.1.1 Identifying constructions
Many Construction Grammarians “enjoy nothing more than finding a construction with peculiar non-predictable characteristics, preferably one that no one has investigated before” (Hilpert, 2014, p. 14). However, finding constructions is certainly not an easy activity. To ease the process of searching for constructions, Hilpert (2014) suggests that there are a number of strategies for detecting and identifying constructions. In what follows, three of the strategies which will be referenced in the corpus-based study of the present thesis will be discussed.
The first strategy is to consider whether the linguistic expression deviates from canonical patterns, which relates to Goldberg’s criterion of non-predictability (Hilpert, 2014, p. 14).
According to Goldberg (2006), “knowledge about language must be learned and stored
as such whenever it is not predictable from other facts” (p. 64). Therefore, if an expression
is not strictly predictable in which it “exhibits formal characteristics that deviate from
more canonical grammatical patterns” (Hilpert, 2014, p. 14), then the expression can be argued to be a construction. One of the examples that Hilpert (2014) used for illustration is the expression by and large. This expression is a phrase that consists firstly of a preposition, then a conjunction, and lastly an adjective. The order of these parts of speech is unique in English grammar. Also, the expression would become unintelligible if the adjective large is replaced by its synonym big. Therefore, one can conclude that there is
“no broader generalisation that would allow speakers to produce or comprehend the expression by and large” (Hilpert, 2014, p. 15) and this provides evidence for arguing that by and large is a construction. Even though it is not required for constructions to have formal non-predictability, “finding formal idiosyncrasies is an excellent source of evidence for calling something a construction” (Hilpert, 2014, p. 15).
Another strategy that helps the detection and identification of constructions is to consider if the expression carries non-compositional meaning (Hilpert, 2014, p. 15). Hilpert (2014) suggests that “if the meaning of an expression is ‘more than the sum of its parts’” (p. 16), one can argue that the expression is a construction. What it means is that constructions
“may override word meanings, creating non-compositional constructional meanings in the process” (Hilpert, 2014, p. 17). This effect of constructions changing or supressing semantic characteristics of lexical items is called coercion effects (Hilpert, 2014, p. 17).
An example that Hilpert (2014) mentions is the expression sauced the pizza. In this expression, the noun sauce is used as a verb and it conveys the meaning of that somebody applied sauce on the pizza. It is not possible to derive this meaning from the individual words and so ESL and EFL learners might have different interpretations of the expression (Hilpert, 2014, p. 17). As for advanced speakers of English, they can arrive at the intended interpretation “because their linguistic knowledge includes a subpattern of the [transitive]
construction with denominal verbs” (Hilpert, 2014, p. 17). In expressions such as pepper the steak, butter the toast and sauced the pizza, the construction “coerces the lexical meanings of pepper, butter, and sauce into the meaning ‘apply to a surface’, which is a substantial semantic enrichment” (Hilpert, 2014, p. 17). Since coercion effects provide solid evidence of constructions being symbolic units embedded with meanings, it is useful to look for non-compositional meanings when finding constructions (Hilpert, 2014).
In addition, checking whether the expression exhibits idiosyncratic constraints also facilitates the detection and identification of constructions, and this strategy relates to both the form and meaning of a construction (Hilpert, 2014, p. 18). Idiosyncratic constraints in constructions must be learned and so such constraints are “something that constitutes knowledge of language, and hence it needs to be included in the construct-i- con" (Hilpert, 2014, p. 19). Consider these two sentences: I bought you a bouquet of flowers and I bought the vase a bouquet of flowers. One would agree that the former makes perfect sense, but the latter simply does not work. In this kind of expressions, “the referent of the recipient argument [must] be animate when actual transfers are at issue”
(Hilpert, 2014, p. 19). As there is this idiosyncratic constraint, this type of expression is
considered as a construction, which is called the ditransitive construction. Nevertheless,
discovering idiosyncratic constraints is not as straightforward as looking for formal non-
predictability or non-compositional meanings (Hilpert, 2014). It is doubtless that
linguistic intuitions are necessary for identifying idiosyncratic constraints, but it is
certainly not enough. Hilpert (2014) thus recommends analysing constructions and their
constraints, first by using “intuition to construct examples and [then checking] those examples against a large database” (Hilpert, 2014, p. 20) such as corpora.
Despite how complex finding constructions seems to be, for years Construction Grammarians have been proposing and examining various types of constructions, such as argument structure constructions, causative constructions, and N P N construction. In the following section, two constructions related to the XTYOFZ construction proposed in this thesis, will be discussed.
2.1.2 Related constructions to the XTYOFZ construction
Hoeksema and Napoli (2008) compared two taboo-term constructions, which they called GET-THE-HELL-OUT, or G-construction and BEAT-THE-HELL-UP
2, or B- construction. Through a corpus investigation, they analysed the use of taboo terms in these two constructions. Hoeksema and Napoli (2008) found that “B-construction started as a literal exorcism... and semantic bleaching ultimately made the literal sense give way to simple emphasis, with any taboo term jumping in” (p. 347). The G-construction may have developed at the same time, or possibly later, on analogy with B-construction (Hoeksema & Napoli, 2008). In fact, Hoeksema and Napoli (2008) suggest that “[if] one construction derived from the other, [it would be] that G-construction was derived from the B-construction” (p. 374).
Both types of constructions have the identical surface strings: verb + taboo term + preposition (Hoeksema & Napoli, 2008). G-construction and B-construction are exemplified as below in 6-10 and 11 respectively (Hoeksema & Napoli, 2008, p. 348):
(6) Let’s get the hell out of here.
(7) Get the hell off my property!
(8) Shut the hell up.
(9) Leave her the hell alone.
(10) Why don’t you go the hell away.
(11) I {beat/kicked/annoyed/punched/surprised/irritated} the hell out of him.
However, the two constructions differ in both syntax and semantics. According to Hoeksema and Napoli (2008), one of the main differences between the G-construction and the B-construction is the object status of the taboo term. In B-construction, the taboo terms are direct objects, whereas in G-construction, it is not the case. For instance, in example 11 above, the taboo term hell is the direct object in the sentence, whereas in examples 6-10, hell is not a direct object. Also, Hoeksema and Napoli (2008) suggest that the taboo term cannot be omitted from the B-construction, but it is possible to do so with the G-construction. Consider examples 6 and 11. It is fine to omit the hell in example 6 and simply say Let’s get out of here. However, it would be ungrammatical to omit the hell in example 11 and say I beat out of him. Furthermore, Hoeksema and Napoli (2008)
2 In Hoeksema and Napoli (2008), the authors named the construction BEAT-THE-HELL-UP (B-construction).
However, up was never a preposition that occurs in this construction, as Hoeksema and Napoli (2008) also mentioned in their study.
note that the coda of the B-construction “always involves out of or its phonological variant outta” (p. 365); whereas the coda of the G-construction can be preposition, prepositional particle, adverb, or adjective. Another difference the authors mentioned is directiveness.
Hoeksema and Napoli (2008) found that most of the occurrences of the G-construction are directive. Commands such as example 7 are the most common, but adhortative sentences such as example 6 are also frequent. There is also embedded directive, which
“is the complement of a predicate such as order or tell” (p. 362), such as The police told us to get the hell out of here. Therefore, Hoeksema and Napoli (2008) suggest that the ruder terms are more frequently used in the G-construction with its directiveness and “the attitude of the speaker to the listener [appears] to be relevant” (p. 363). Whilst with the B-construction, the taboo term is simply emphasis and “the expletive is not related to the attitude of the speaker toward the listener” (Hoeksema & Napoli, 2008, p. 363).
Although B- and G-constructions have similar surface strings, only B-construction is under the XTYOFZ construction umbrella as G-construction differs from the XTYOFZ construction both syntactically and semantically. The overall pattern of the XTYOFZ construction is “X the Y out of Z”, where X is usually a non-motion action verb such as beat and kick, or a verb related to feelings or emotions such as bore, scare and annoy; Y is a noun, and for many cases, it is a taboo term such as hell, crap and shit, whereas Z is a Noun Phrase (NP) which can be simple nouns, proper nouns, personal pronouns, and so on.
Hoeksema and Napoli (2008) provided a detailed analysis of both the G-construction and B-construction, from the history of the constructions, their syntactic and semantic differences, to their occurrences in the corpus. Apart from Hoeksema and Napoli (2008), other researchers (e.g., Haïk, 2012; Perek, 2016; Hoffmann, 2020) have also investigated various aspects of the B-construction, but all with different names including Construction B (Haïk, 2012), “V the hell out of NP” construction (Perek, 2016), and the [V the N
taboo- termout of]-construction (Hoffmann, 2020). Hereafter in this thesis, it will be referred to as the [V the N
taboo-termout of]-construction. Both Perek (2016) and Hoffmann (2020) examined the syntactic productivity of the construction, whilst Haïk (2012) investigated the use of the hell in English grammar, in which one of the constructions discussed was the [V the N
taboo-termout of]-construction.
Nevertheless, Hoeksema and Napoli (2008) mainly focused on the use of taboo terms and
comparison between the G- and B-constructions, while Perek (2016) and Hoffmann
(2020) put their attention solely on the syntactic productivity of the construction, and
Haïk (2012) exclusively examined the usage of the taboo term the hell. The above studies
did not really provide a detailed discussion on the other properties of the construction
aside from the productivity of verb-slot or the use of taboo terms. Therefore, the present
thesis intends to further investigate the [V the N
taboo-termout of]-construction in a broader
sense. In other words, this study aims to propose and examine the umbrella construction
of the [V the N
taboo-termout of]-construction, namely the XTYOFZ construction and
discuss particularly its usage context, syntactic and semantic characteristics.
2.2 Applied Construction Grammar
Most of the current research within the framework of Construction Grammar adopts a theoretical or descriptive approach, whereas the applied perspective such as ESL and EFL acquisition and teaching has been relatively neglected. Even though there is research contributing to the issues related to L2 applications of Construction Grammar, the number of publications and the findings are rather limited. Therefore, Gilquin and De Knop (2016) proposed that there is a need for “a continued and collaborative effort to pursue research in this field” (p. 8), which they referred to as Applied Construction Grammar. In Gilquin and De Knop’s (2016) volume, the authors addressed the main issues related to acquisitional and pedagogical implications of Construction Grammar in ESL and EFL context by bringing together different relevant contributions. In the following sections, one of the chapters in their volume which focuses on the effects of ESL and EFL context on constructions acquisition and command will be discussed, along with Römer et al.
(2020) which examines ESL learners’ production of verb-argument constructions.
2.2.1 Effects of ESL and EFL context on constructions usage
Adopting a constructionist corpus-based approach, Gilquin (2016) investigates how acquisition context affects the learning of constructions by ESL and EFL learners. Gilquin (2016) focused on periphrastic causative constructions with cause, get, have and make in English. Periphrastic causative constructions are “made up of a causative verb (like cause or make) controlling a non-finite complement clause, whose combination expresses the idea that a CAUSER acts upon a CAUSEE to bring about an EFFECT” (p. 116). An example of causative construction would be as followed:
(12) Grammar (CAUSER) makes me (CAUSEE) cry (EFFECT).
Gilquin (2016) hypothesized that ESL learners who acquire English in a natural setting receive more exposure to English and thus “have a better [, more native-like] command of constructions than [EFL learners], who rely almost exclusively on formal instruction”
(p. 115). To test the hypothesis, data was extracted from three types of corpora. For EFL data, 3,621,892 words were collected from the International Corpus of Learner English (ICLE) which consists of essays produced by EFL students from 16 different L1 backgrounds. Only data from learners whose native language was the main language they spoke at home and who lived in a country where that language was dominant was selected (Gilquin, 2016, p. 123). As for ESL data, 1,435,187 words were extracted from the non- professional writing sections of the International Corpus of English (ICE) and the whole NUS Corpus of Learner English (NUCLE). The ICE corpus contains both student essays and examination scripts from eight Asian and African countries, whereas the NUSCLE corpus consists of student essays from the National University of Singapore (Gilquin, 2016). In order to compare the EFL and ESL data, five million words was obtained as ENL sample from the academic writing sections of the British National Corpus (BNC).
The corpus data shows that both EFL and ESL writers “heavily overuse the construction
in comparison with [ENL writers]” (p.125), although EFL students do have a stronger
tendency than ESL students. The findings partly support the frequency-based
interpretation hypothesis as the ESL students do “appear to better approximate to the native frequency of the construction than EFL students” (Gilquin, 2016, p. 125).
However, the results at the same time “lead to a refinement of the usage-based model of L2 acquisition” (Gilquin, 2016, p. 115). The usage-based view of language predicts that among the three groups, ENL writers should use the construction the most frequently, followed by ESL and then EFL writers. Yet, the results show the exact opposite in which the highest frequency of the causative constructions was found in EFL corpus, followed by ESL and the ENL the least (Gilquin, 2016). On the other hand, differences between ESL and EFL learners were revealed. Gilquin (2016) found that ESL students “generalise at a higher level of abstraction than EFL students” (p. 144) due to the natural input they get, whereas EFL students “receive arguably focused attention on more concrete, low- level constructions” (Gilquin, 2016, p. 144) as the instruction they have is explicit.
Gilquin (2016) did provide insightful findings regarding acquisition of constructions in EFL and ESL contexts. Nevertheless, Gilquin (2016) focused only on the usage and acquisition of the English periphrastic causative constructions, which is one of the argument structure constructions. Gilquin (2016) also specifically fixated on causative constructions in academic written context. It is worth to also look at different constructions which are more common in spoken context to test the hypothesis of input- dependent ESL and/ or EFL acquisition. Thus, the present thesis addresses ESL learners’
acquisition and comprehension of the XTYOFZ construction, which is a construction that is more common in non-academic and more relaxed contexts (see Section 4.1.2).
2.2.2 ESL learners’ production of the verb-argument constructions
Based on data retrieved from learner and native-speaker corpora as well as verbal fluency tasks, Römer et al. (2020) provided insights into ESL learner knowledge of the English verb-argument constructions (VACs). The authors focused on 34 different VACs (Römer et al., 2020), including the ‘V with n’ construction (e.g., she dealt with the problem), ‘V about n’ construction (e.g., I thought about the issue), the ‘V reflexive pronoun’
construction (e.g., he hurt himself), and others.
As mentioned, Römer et al. (2020) used two types of data in their study. First, they collected data from four different corpora (Römer et al., 2020), including the International Corpus of Learner English (ICLE), the Louvain International Database of Spoken English Interlanguage (LINDSEI), the Louvain Corpus of Native English Essays (LOCNESS) and the Louvain Corpus of Native English Conversations (LOCNEC). From ICLE and LINDSEI, data of German and Spanish learners of English was selected, which includes argumentative essays as well as transcripts of informal interviews (Römer et al., 2020).
From LINDSEL and LOCNESS, VAC data of native English speakers was collected,
which also includes argumentative essays and transcripts of interviews (Römer et al.,
2020). The corpus data was then analysed with the concordance tool AntConc (Anthony,
2014). Second, Römer et al. (2020) draw upon data from verbal fluency tasks with three
groups of university students in their twenties, including native English speakers, German
and Spanish learners of English. Participants were asked to “generate as many words as
come to mind within a given time [when presented] with bare sentence frames based on
40 different VACs” (Römer et al., 2020, p. 309).
The results show that “VACs are strongly entrenched in the minds of advanced L2 English learners of [German and Spanish L1 backgrounds]” (Römer et al., 2020, p. 325).
Römer et al. (2020) found that there is “considerable overlap between the verb-VAC associations of learners and native speakers” (Römer et al., 2020, p. 325), especially with verbs that are strongly associated with the constructions. This indicates that the ESL learners in their study had a rather native-like command of VACs, which echoes Gilquin (2016) in which it was found that to a certain extent ESL learners have a native-like command of the English periphrastic causative constructions. However, differences in the performance of the native speakers and ESL learners were also found. Römer et al. (2020) suggests that those differences are “sometimes the results of cross-linguistic transfer from the learner’s L1” (p. 325).
Römer et al. (2020) provided an all-round investigation on ESL learners’ production of constructions by combining corpus and cognitive approaches and their study shows how a combined ‘corpus plus experimental data’ approach could be useful to Construction Grammar research. Römer et al. (2020) certainly gave useful insights into ESL learners’
production of constructions. Nonetheless, their focus fell on VACs, which is a rather frequently discussed and common construction, particularly in academic contexts. As mentioned, it is worth to also look at less-discussed, infrequent constructions which are more common in spoken and relaxed contexts. Thus, also taking on a combined ‘corpus plus experimental data’ approach, the present thesis aims to look at a less formal and non- academic construction, i.e., the XTYOFZ construction, and the production of this construction by ESL and ENL speakers.
2.3 Priming with constructions
Apart from using corpus-based studies, researchers have also been using priming tasks to investigate the acquisitional and pedagogical aspects of Construction Grammar. Priming tasks aim to find out “how quickly and/ or accurately participants respond to a stimulus (the probe) that has been preceded by another stimulus (the prime)” (Warren, 2013, p.
101). There are various types of priming, such as identity priming, semantic priming, and associative priming. In the following sections, two studies which examine Construction Grammar using priming tasks will be discussed.
2.3.1 Automatic accessing of constructional meaning by ENL speakers
According to Johnson and Goldberg (2013), although it has been shown that “phrasal constructions are readily associated with specific meanings, it remains unclear whether this meaning is accessed automatically” (p. 1399). Therefore, using a Lexical Decision Task with both associative and semantic priming, Johnson and Goldberg (2013) tested whether the accessing of constructional meaning is automatic. In the study, participants were asked to “[perform] a Lexical Decision Task on individual target words which were preceded by abstract skeletal constructions” (Johnson & Goldberg, 2013, p. 1399).
Four abstract argument structure constructions, including ditransitive, resultative, caused-
motion and removal constructions were selected as primes. Jabberwocky sentences with
the four selected constructional frames were used as construction primes to create stimuli.
Jabberwocky sentences refer to sentences that are syntactically well-formed but “contain no meaningful open-class items” (Johnson & Goldberg, 2013, p. 1399). For instance, the sentence in example 14, She jorped it miggy, seems acceptable syntactic-wise, but the nonsense words jorped and miggy do not have any meaning and thus making the sentence a Jabberwocky sentence. In other words, sentences in argument structure constructional frames with nonsense open-class items were created. Examples of Jabberwocky sentences used in Johnson and Goldberg (2013, p. 1399) are as illustrated below:
(13) He daxed her the norp. (Ditransitive) (14) She jorped it miggy. (Resultative)
(15) He lorped it on the molp. (Caused-motion) (16) She vakoed it from her. (Removal)
On the other hand, lexical targets were selected based on the data from COCA. There were three groups of target words in the experiment, including "high [and] low associates of one of the constructions, [as well as] words that were semantically related to one of the constructions” (Johnson & Goldberg, 2013, p. 1445).
Forty undergraduate students who were all native English speakers from Princeton University participated in the experiment. In order to compare the priming effects for congruent and incongruent target words, the participants were divided into two groups, including one group that saw two experimental words of each type in a congruent context and the other group in an incongruent context (Johnson & Goldberg, 2013). For instance, a congruent context would be if the target word was gave and ditransitive construction was presented. Whereas if the target word was gave and removal construction was presented instead, it would be counted as an incongruent context. Both groups were first presented with a fixation (3000 ms), then followed by a constructional prime (1000 ms), then after a 300 ms fixation, the target word was shown. The participants were instructed to read aloud the constructional prime and respond if the target word is a real word or a non-word within 1000 ms. In the process, the participants’ reaction times were recorded.
Johnson and Goldberg (2013) found positive evidence of priming, which indicates that
“phrasal abstract constructions are associated with semantics even when they contain no open-class lexical items, and that the meaning is accessed quickly and without explicit instruction” (Johnson and Goldberg, 2013, p. 1451). Particularly, their results show that there is indeed difference between the priming effects for congruent and incongruent target words. The data from the experiment shows “significant priming effects for congruent over incongruent target words, both for associated targets... and to a lesser extent, for target words that are semantically related to the construction but which rarely occur in the construction” (Johnson & Goldberg, 2013, p. 1339).
The Lexical Decision Task by Johnson and Goldberg (2013) provided undeniable evidence proving automatic accessing of constructional meaning. However, all the participants were undergraduate students with English as their native language.
Therefore, the evidence found in this study applies to only native English speakers. Also,
the focus of Johnson and Goldberg’s (2013) study fell on argument structure
constructions, like many other previous studies. Therefore, this thesis attempts to examine
and discuss the accessing of constructional meaning of the XTYOFZ construction using
a Lexical Decision Task with reference to Johnson and Goldberg (2013), but, in addition to ENL speakers, with ESL speakers who come from a more varied background.
2.3.2 Constructional priming in ESL learners
With reference to two previous studies (Bock & Loebell, 1990; Hare & Goldberg, 1999) on constructional priming with native speakers of English, Baicchi (2016) conducted a constructional priming experiment in an L2 acquisition context. Using a sentence- elicitation task with argument structure constructions, Baicchi (2016) aimed to “ascertain whether, and to what extent, constructional priming plays a part in the learning process”
(p. 211) of Italian ESL learners.
Thirty-two Italian university students whose mean age was 21 volunteered to participate in the experiment. The participants were divided into two groups based on their English proficiency according to the Common European Framework of Reference for Languages (CEFR, Council of Europe 2001), namely a B1 group and a B2 group. In the experiment, a priming sentence was first played to the participants and they were instructed to repeat the sentence aloud. Then, a picture of a dative event matching the priming sentence was presented and the participants were asked to describe the event depicted without using pronouns (Baicchi, 2016). The responses of the participants were recorded and scored. In total, there were 10 pictures matching 10 sets of priming sentences, representing actions that involve “an agent and a receiver of the action, and a non-human entity undergoing the action” (Baicchi, 2016, p, 225). The priming sentence set was composed of the materials from Hare and Goldberg (1999), which include prepositional dative, double- object and fulfilling sentences as the three primes, as well as an intransitive sentence serving as a filler (Baicchi, 2016).
The results demonstrate that B1 level learners produced prepositional dative construction the most as it “represents the prototypical Italian constructional for verbalization of the transfer schema” (Baicchi, 2016, p. 226). As for double-object and fulfilling constructions which are absent from the Italian construct-i-con, B1 students scored significantly lower.
Comparatively, B2 level students produced much less dative construction and more of the other two constructions. This suggests that “the higher the proficiency level, the more the learners have recourse to English-specific constructions” (Baicchi, 2016, p. 228).
Nevertheless, the double-object and fulfilling constructions were not easy to prime for both groups of learners and “the fulfilling prime prompted double-object responses because of the semantic similarity” (Baicchi, 2016, p. 228). This means that “conceptual similarity is more salient than structural similarity” (Baicchi, 2016, p. 228) and that so there is indeed mapping between syntax and semantics in the mind of ESL learners.
Through replicating two previous studies, Baicchi (2016) successfully demonstrated that
constructional priming has a role in the learning process of not only native speakers but
also ESL learners. Nonetheless, similar to many previous studies, Baicchi’s (2016) study
focused on argument structure constructions. Furthermore, although Baicchi (2016)
included participants with two different levels of English proficiency, they were all
undergraduate students and came from Italian L1 background. With the inspiration from
Baicchi (2016), the present study also aims to investigate constructional priming in ESL
speakers, but with ESL participants from varied L1 and educational backgrounds.
3. Methodology
In the present thesis, corpus linguistic methodology was used to examine the XTYOFZ construction. Comparison to the [V the N
taboo-termout of]-construction (Hoeksema &
Napoli, 2008; Haïk, 2012; Perek 2016; Hoffmann, 2020) were also made. The Corpus of Contemporary American English (COCA) was selected for data collection on the construction. Apart from the corpus-based approach, a Lexical Decision Task with Jabberwocky sentences and a follow-up survey were also used to investigate the processing and usage of the construction by ESL learners and ENL speakers.
3.1 Corpus Linguistics
To discuss what corpus linguistics entails, it is necessary to first clarify what a corpus is.
According to Tognini-Bonelli (2001), a corpus can be defined as “a collection of texts assumed to be representative of a given language put together so that it can be used for linguistic analysis” (p. 2). In general, it is assumed that the language stored in a corpus is natural and authentic (Tognini-Bonelli, 2001). Thus, given that a corpus is a collection of many different naturally occurring texts, corpus linguistics allows researchers to investigate language in use with the patterns of repetition shown in the corpus. In the following section, the corpus that was used in this study and how the data was selected for analysis will be discussed.
The Corpus of Contemporary American English (COCA) was released in 2008 and it is
“the first large and diverse corpus of American English” (Davies, 2009, p. 159). The corpus contains more than one billion words of texts, which is about 25 million words each year between the years 1990-2019
3, including both spoken and written components.
The words in COCA come from eight genres, including spoken, fiction, popular magazines, newspapers, academic texts, TV and movies subtitles, blogs and other web pages.
Given that the scale of the present study is relatively small, only instances of XTYOFZ construction in the context where personal pronoun is the coda were considered. The pattern VERB the NOUN out of _pp was searched in COCA grouped by lemmas. VERB, NOUN and _pp are the input tags for parts of speech in COCA. With the input tags VERB and NOUN, it will add the parts of speech, that is verbs and nouns, as a “full word” in the search. Whilst with the input tag _pp, it limits the word searched to personal pronouns.
In total there were 2111 types (total frequency 5650 tokens), and the 500 most frequent types were taken for analysis. Some of the types were eliminated as they belong to G- construction in which the nouns (Ys) could be omitted (Hoeksema & Napoli, 2008), and some were deleted as there was no verb (X) or noun (Y) indicated. Therefore, 486 types (3984 tokens) were actually used for analysis. The frequency of occurrences of the verbs, nouns, personal pronouns as X, Y and Z respectively were then grouped and counted.
Frequency of occurrences of the linguistic pattern was also searched and compared in each section of the corpus, including TV/movies, blog, websites, spoken, fiction,
3 Note that texts from the sections BLOG and WEB are not assigned years from which they are.
magazine, newspaper, and academic. The same pattern VERB the NOUN out of _pp was searched using the CHART option which shows the total frequency in each section (including genres, years and sub-genres).
Based on the corpus search of the pattern VERB the NOUN out of _pp, the most frequent word choices for X, Y and Z in the construction were chosen for analysis in order to see what forms the construction most commonly occurs in. Furthermore, the frequency of occurrences of the construction in different genres was compared and will be discussed in the thesis. The syntactic and semantic features, as well as the usage context (both genre- wise context and immediate context) of the construction were also analysed based on the corpus data. Apart from analysing the different features of the construction, the corpus data was also used to support the argument that XTYOFZ construction should be considered as a construction. With the support of the corpus data, three aspects of identifying constructions suggested by Hilpert (2014) were analysed, including canonical patterns, non-compositional meaning and idiosyncratic constraints of the XTYOFZ construction.
3.2 Timed Lexical Decision Task
To test whether ESL speakers and ENL speakers access constructional meanings automatically, a Lexical Decision Task with Jabberwocky sentences based on the previous study by Johnson and Goldberg (2013) was conducted online.
3.2.1 Pilot Study
Before launching the Lexical Decision Task, a pilot study was conducted. The participants in the pilot study were given the link to the experiment and they were asked to complete both the Lexical Decision Task and the follow-up survey. Then they were asked to provide feedbacks on how the experiment works, the amount and clearness of the instructions and anything else that they noticed. Most of the feedback received was regarding the time for reading the target words and responding in the Lexical Decision Task, and some of them were about the technical issues in the follow-up survey.
According to the feedback gathered, both the Lexical Decision Task and the follow-up survey were amended and fine-tuned before the actual data collection.
3.2.2 Participants
Two groups of participants were recruited, including an ESL group and an ENL group.
Each group consists of 21 participants aged between 20-30. For the ESL group, the participants came from different L1 backgrounds including Cantonese, Mandarin, French, Latvian, Russian, Thamazight (Berber), Polish, Finnish and Luganda
4. In the present
4 It should be noted that different L1 backgrounds might possibly influence the participants’ performance as the construction might or might not exist in their L1s. Also, participants’ knowledge of different languages might also contribute to their command of constructions. Nevertheless, the sample size of the present study is relatively
study, there was no exact measure of the ESL participants’ English proficiency, but in the follow-up survey (see Part I: Background Information on English Acquisition and Usage in Appendix C) there are questions on their English acquisition and usage, such as how old they started learning English, what impacts their English acquisition the most, how often they use English, and so on. It was found that all the ESL participants use English quite often, mostly in the form of text such as reading and writing. All the ESL participants had at least secondary school background. In fact, 20 of them had Bachelor’s or higher education background. This indicates that the English proficiency of ESL participants were at least intermediate. For the ENL group, the participants either speak British English or American English. Interestingly, all of the ESL participants speak more than two languages but only half of the ENL participants speak three or more languages.
In fact, seven of the ENL participants only speak English. Same as the ESL group, all the ENL participants had at least secondary school education background, in which most of them had Bachelor’s or higher education.
3.2.3 Materials
For the programming and design of the Lexical Decision Task, PsyToolkit (Stoet, 2010, 2017) was used. PsyToolkit (Stoet, 2010, 2017) is “a free web-based service designed for setting up, running, and analysing online questionnaires and reaction-time (RT) experiments” (p. 24). Based on the tutorials and the existing Lexical Decision Task coding from the PsyToolkit library (Stoet, 2010, 2017), the Lexical Decision Task used in the present study was created.
3.2.3.1 Construction primes
The XTYOFZ construction was chosen as the construction frame for priming. The construction primes were all made up of nonsense open-class words and so they were abstract. Sixty nonsense words were used in total, in which 30 were in the form of verbs (i.e., the X slot in the construction) and 30 in the form of nouns (i.e., the Y slot). Most of the nonsense words were created tailored to the present experiment, while a few of them were borrowed from Johnson and Goldberg (2013). Each construction prime consists of a different nonsense verb and a different nonsense noun. None of the nonsense words was repeated in the experiment. Example sentences are given in Table 1.
Table 1. Constructional frame used as primes.
Abstract construction Constructional frame Examples
XTYOFZ
Personal Pronoun nonsense-ed the nonsense out of Personal Pronoun.
She skeuped the sholp out of me.
He bort the colp out of him.
I ensuked the fulk out of myself.
They seited the byliets out of us.
small and therefore it was not sufficient to support the argument of L1 transfer or effect of language knowledge on ESL construction usage or comprehension.
3.2.3.2 Lexical targets
With reference to Johnson and Goldberg (2013), high frequency (HF) associates and low frequency (LF) associates were selected as the target words based on a search from COCA
5. Words that occur frequently as X (verb) and Y (noun) in the construction in COCA were selected as the HF associates, whereas the ones with a relatively low frequency of occurrences were chosen as the LF associates. For the HF group, the ones with the relatively higher frequencies were chosen as the target words, whereas those with less occurrences were picked as the basis of the nonsense target words, with hell and shit as the exception as their frequencies were rather similar. As for LF group, the ones with relatively few occurrences were chosen as the target words; whereas those that occurred comparatively more frequently were picked as the basis of the nonsense target words, except that tar was not used as a target word but the basis of a nonsense word as it is possibly a rather uncommon word for ESL speakers. The frequencies of each of the target words are illustrated in Table 2. All the nonsense target words were included in the brackets next to their real word basis.
Table 2. Raw frequencies of the target words (all tenses included) based on a search from COCA.
High Frequency Group Low Frequency Group
X (verb)
Y (noun)
scared
kicked (kacked) annoyed (annoped)
shit hell (holl) heck (huck)
1451 112 75 956 1093 112
sued
surprised (sorprosed) enjoyed (enjaked)
daylights tar (tir) fuck (fack)
8 38 31 37 18 39
Including both the practice phase and the experimental block (HF and LF groups) (see
Section 3.2.4), there were 24 target words in total. Half of the target words were realwords and half were nonsense words. The eight nonsense words used as target words in the HF group and LF group were created phonetically similar to the real HF and LF associates found in COCA as shown in Table 2. Among the real target words, two groups were divided according to whether the words are congruent or incongruent to the XTYOFZ construction. In other words, the target words that occur commonly in the XTYOFZ construction were grouped together as ‘congruent’; and the words that do not ever occur in the construction were grouped together as ‘incongruent’. All the target words used in the experiment are listed in Table 3 below.
5 This search from COCA was done preliminarily before the two searches for the corpus study, which was grouped by words on COCA and then by lemma manually. Therefore, the frequencies of the target words vary a bit from what the data for the corpus study presented in Section 4.1.1.
Table 3. Target words in the experiment.
Practice Group High Frequency Group Low Frequency Group
Congruent real words X: frightened
Y: sense
Incongruent real words X: showed
Y: wand
Nonsense words X: kaxed, bolped Y: miggy, krap
Congruent real words X: scared
Y: shit
Incongruent real words X: told
Y: book
Nonsense words X: kacked, annoped Y: holl, huck
Congruent real words X: sued
Y: daylights
Incongruent real words X: danced
Y: pants
Nonsense words X: sorprosed, enjaked Y: tir, fack
3.2.4 Procedure
Ideally, the task would be face-to-face where the researcher meets with each participant in person and records their reaction times. However, given the present circumstance with global pandemic, it was unlikely to conduct a face-to-face experiment. Thus, the Lexical Decision Task was conducted online. Participants were given an information sheet (see
Appendix D) and opportunities to ask questions about the experiment. They were askedto sign a consent form (see Appendix E) indicating that they agreed to participate. The participants were also given the option to choose whether they prefer to conduct the experiment via a Zoom meeting with the researcher present or to complete the task by themselves but with their computer screen recorded with audio. This is to ensure that the participants actually read the construction primes out loud so that the primes are in effect.
Before the Lexical Decision Task, the participants were asked to answer some questions regarding their background information such as gender, age, first and second languages (see Part 1: General Background Information in
Appendix B). After they had answeredthe background questions, they were presented an instruction video on how to do the Lexical Decision Task. Then a short trial was given for the participants to see how the task will be like. Afterwards, they were asked to proceed to the Lexical Decision Task and another more detailed instruction (see Figure 2) was given right before they start.
The Lexical Decision Task was divided into two parts, including a practice phase and the experimental phase which was further divided into HF and LF blocks. Between the HF and LF blocks, a 30-second break was given. In both the practice phase and the experimental phase, there were eight ‘priming sentence + target word’ combinations (2 congruent real words, 2 incongruent real words, and 4 nonsense words). The general structure of the Lexical Decision Task is as illustrated below and in Figure 4.
1. Practice phase (8: 4 real target words and 4 nonsense words) -10-sec intersection-
2. Experimental phase
2.1. HF group (8) -30-sec break- 2.2. LF group (8)
The general procedure was similar to Johnson and Goldberg (2013), but with a few modifications. In Johnson and Goldberg (2013), the participants were asked to “[perform]
lexical decisions following sentential primes” (p. 1444), specifically with argument structure constructions. The participants were instructed to read a sentence written in black which was followed by a word in green. Then they were asked to respond as fast as possible, “pressing ‘1’ if the green ‘word’ was a real word and ‘2’ if it was a nonword”
(Johnson and Goldberg, 2013, p. 1445). Same as Johnson and Goldberg (2013), in the present study participants were first presented with a fixation cross, then the construction prime written in black which they were asked to read aloud. As mentioned above, in the present study the sentential primes with XTYOFZ construction were used instead of argument structure constructions. Then a fixation cross was shown again briefly, and after that the target word written in green was presented which the participants were asked to respond whether it is a real English word or a nonsense word. However, instead of pressing ‘1’ and ‘2’, participants in the current study were asked to press key ‘A’ and key
‘L’
6on their own keyboards to indicate if the green word is a real English word or a nonsense word respectively.
In addition, the reading and reaction time were also altered from Johnson and Goldberg (2013). In Johnson and Goldberg (2013), the participants were only given 1000 ms to read the construction phrase and they were to respond whether the target word was a real word or a nonsense word within 1000 ms. In the present study, the reading time was increased to 2500 ms and the response time to 2000 ms. Generally speaking, ESL speakers would have a slower reading rate than ENL speakers. According to Fraser (2007), a widely acknowledged problem of ESL readers is that “they read slowly, perhaps too slowly in English, their second language” (p. 372). Existing research (e.g., Cushing- Weigle & Jensen, 1996; Segalowitz, Poulsen, & Komoda, 1991) has shown that there is
“a considerable gap in first (L1) and L2 reading rates for the majority of L2 readers”
(Fraser, 2007, p. 373) in which ESL readers read more slowly in English than ENL readers. Pilot study also shows that 1000 ms was too little time for ESL speakers to respond and even 2000 ms was too short for them to read the sentential primes. Also, a previous study on L2 English lexical processing by Chen, Dong, & Yu (2018) supports the fact that ESL speakers need more time to read and respond. In Chen et al. (2018), 77 Chinese ESL speakers were asked to perform a Lexical Decision Task with word stimulus. Even though in their task the stimulus were all single words, the participants were given up to 2500 ms to read and respond. Therefore, to ensure that the ESL participants have enough time to read the primes and react, the reading time and reaction
6 In the existing sample codes of Lexical Decision Task in PsyToolkit (Stoet, 2010, 2017), the default keys are
‘A’ and ‘L’. Also, different from Johnson and Goldberg (2013), in which the experiment was conducted face-to- face with specialized equipment (e.g., keyboard with only ‘1’ and ‘2’) in laboratory, the present experiment was conducted online with participants’ own keyboards in which keys ‘A’ and ‘L’ are in the middle row and the furthest from each other. Thus, keys ‘A’ and ‘L’ were chosen as the responding keys in the present experiment.
time were increased. Figure 1 represents the basic structure for a priming trial in the present study. Figure 3 illustrates how the process of a priming trial looks like with both correct and wrong responses. The cross + symbolises the fixation cross shown in the task.
Figure 1. Basic structure for a priming trial in the present study.
Figure 2. Instructions before the Lexical Decision Task.
Figure 3. Example of a priming trial in the Lexical Decision Task (correct or wrong responses).
Figure 4. Basic procedure of the Lexical Decision Task.
3.3 Follow-up survey on construction usage comprehension
In addition to the corpus-based study and the Lexical Decision Task, the present thesis also investigates the usage and comprehension of the XTYOFZ construction by ESL and ENL speakers through a follow-up survey which was conducted online.
3.3.1 Participants
The same participants from both the ESL group and the ENL group were asked to continue to finish a follow-up survey after they had completed the Lexical Decision Task.
3.3.2 Materials and Procedure
