Where tickles the priming the man with the PP-attachment? A cross-linguistic research into priming effects in Dutch.

Master Thesis Linguistics: Language & Communication

Faculty of Humanities

Student Ella Visser

Number 1398075

Supervisor dr. E. Bosma Second reader Prof. dr. L.L. Cheng

Date 14 February 2019

Where tickles the priming the man with the PP-attachment?

A cross-linguistic research into priming effects in Dutch.

Table of Contents

ACKNOWLEDGEMENTS ... 3

ABSTRACT ... 4

1.

INTRODUCTION ... 5

1.1

S

TRUCTURAL PRIMING

... 5

1.2

I

MPLICIT

L

EARNING

... 7

1.3

S

TRUCTURAL PRIMING IN

L1

COMPARED TO

L2 ... 8

1.4

T

ARGET STRUCTURES

... 10

1.5

P

RIMING AND TIME

... 11

1.6

T

HE PRESENT STUDY

... 12

2.

METHOD ... 15

2.1

P

ARTICIPANTS

... 15

2.3

M

ATERIALS

... 18

2.4

P

ROCEDURE

... 22

2.4

S

CORING

... 22

3.

RESULTS ... 23

3.1

F

IRST RESEARCH QUESTION

... 23

3.2

S

ECOND RESEARCH QUESTION

... 24

3.3

T

HIRD RESEARCH QUESTION

... 25

4.

DISCUSSION ... 27

4.1

H

YPOTHESES

... 27

4.2

L

IMITATIONS

... 29

4.3

I

MPLICATIONS

... 30

5.

REFERENCES ... 31

6.

APPENDICES ... 35

A

PPENDIX

I

–

C

ONSENT FORM

... 35

A

PPENDIX

II

–

Q

UESTIONNAIRE

... 37

A

PPENDIX

III

–

S

TRUCTURAL

P

RIMING

T

ASK

... 40

ACKNOWLEDGEMENTS

First, I would like to show my gratitude to my supervisor, dr. Evelyn Bosma. After a hectic period with two masters, I got enthusiastic about a topic for my thesis through one of her courses. As this area of linguistics was new to me, I am grateful for the guidance in the right direction during the process of this research. It has been difficult and hard work, but I have never learned more about

academic research and myself than during this process.

In addition, I would like to thank my fellow students from Leiden University and the VU for your never-ending patience and optimism, your support has helped me through the schedule struggles,

despair and general struggle that comes with writing a thesis.

Lastly, I would like to thank my friends and family, for their love and support. A special thanks is reserved for my boyfriend, who is probably looking forward to my next thesis as much as I am.

ABSTRACT

Within the linguistic field, structural or syntactic priming has been investigated for L1 and L2 speakers. Studies show mixed results for the role of the mother language and the duration of the priming effect. This thesis carried out an experiment with English and German L2 speakers of Dutch in comparison to native speakers of Dutch in a two-day structural priming task for relative clauses and prepositional phrases. The results show that priming works among English L2 speakers of Dutch, but not among the other two language groups. The priming effects for the English participants lasted at least one day, which indicates implicit learning as a result of long-term priming. A time-pressure component did not impact the effect of priming among the native speakers. The study adds to the existing body of knowledge of structural priming and confirms the existence of a difference in priming effects for fa-miliar or unfafa-miliar syntactic structures in the L1 of the participant.

1. INTRODUCTION

1.1 Structural priming

The definition of structural priming, also described as syntactic priming, is generally accepted as the phenomenon of speakers being more likely to repeat syntactic structures they have been exposed to, or have previously produced themselves (Bock, 1986). Repeating oneself is not uncommon in spoken language. Words, numbers, sound, mental models and structures can be repeated from the speech per-formed by others or perper-formed by the speaker himself (Kubovy, 1977; Dell, Burger & Svec, 1997; Garrod & Anderson, 1987; Garrod & Pickering, 2004; Ferreira & Bock, 2006). Previous research focused mainly on the repetition of syntactic elements which would cause a slip of the tongue. These linguistic errors implied the strong possibility of repetition in speech process.

As described by Ferreira and Bock (2006), speech production is generally viewed as a creative process, ‘but can also be surprisingly recapitulative’. Speakers tend to unconsciously repeat structures they have heard before, illustrated in the following example adapted from Ferreira and Bock (2006). Speakers who have been presented with a prepositional dative structure such as a are more likely to use this structure when describing a similar situation, for example b.

a. The mother made [some breakfast] [for her daughter]. b. The man gave [a new purse] [to his wife].

Whereas exposure to the initial sentence in a double-object structure such as c would lead to a produc-tion of a similar structure as can be seen in d.

c. The mother made [her daughter] [some breakfast]. d. The man gave [his wife] [a new purse].

Both of these sentences are correct and can be used in daily speech. This repetition can be caused by listening to speech (Bock, 1986), but also by reading. Additionally, it can be triggered through experi-ments. These effects can be seen among native speakers (L1) as well as second language (L2) learners (Kaan & Chun, 2018). Within priming L2 speakers, there are different possibilities; it is possible to prime someone in language A and test someone in language B, which is called cross-language priming (Loebell & Bock, 2003). It is also possible to prime and test someone in the same language, either the L1 or the L2 of the participant. This is called within-language priming. Priming effects can be seen among participants as young as four years old (Savage, Lieven, Theakston & Tomasello, 2006), posing questions about the nature of priming effects.

Structural priming as a tool for second language learning has developed into a popular research topic. Over the past decade, the subject has emerged on a large scale, investigating the possibilities of structural priming. Research from Hartsuiker and Bernolet (2015) shows the impact of the L1 of a speaker in the effects of priming. This has also been found by Jarvis (2011) and Odlin (2005). It is shown that the first language impacts the sensibility for priming, as the underlying syntactic structure is ei-ther already represented at the time of the priming, or a new structure that is acquired during the L2 learning. When the structure is already familiar for a speaker, the effects of priming appear to be dif-ferent than when the structure does not exist in the first language.

In addition to priming effects that are triggered in research settings, several researchers have shown in corpus studies that priming also arises in natural speech without any controlled input to enhance priming effects (Reiter & Moore, 2014; Thomas, 2016; Gries & Kootstra, 2017).

Taking the different possibilities for priming into account, priming syntactic structures pro-vides a valuable research method to investigate the process of second language learning. The role of the first language of a speaker can be investigated as participants can be primed in their first and in their second language. Additionally, it is possible to test participants with different first languages in their shared second language. This provides insight into the factors that influence language learning and the effects of priming.

The role of a speaker’s first language is subject to an ongoing debate. There appears to be a dichotomy between the first language being irrelevant and the first language playing a vital role. Some researchers have found negligible effects of the first language’s syntactic structure in the study (e.g. Clahsen & Felser, 2006), whereas several other researchers (e.g. Hartsuiker & Bernolet, 2015; Jarvis, 2011; Luk & Shirai, 2009; Odlin, 2005) have found that the first language can significantly impact sec-ond language acquisition of speakers.

From 2008 onwards, there has been an increasing focus on the comprehension of complex syntactical expressions, the implications of grammatical representations in first languages and the possibility to influence repetition or comprehension by exposing participants to syntactically similar structures (Pickering & Ferreira, 2008). Several studies have investigated the effect of short-term priming, which implicates that the priming has an effect that occurs directly after the treatment. It is found that this could be caused by residual activation of the represented syntactic structure or combi-natorial nodes (Jackson & Ruf, 2017). It is assumed that speakers use nodes while forming sentences; these nodes represent certain lexical items (e.g. verbs, nouns or a combination of words). These nodes are connected to the nodes that indicate how certain lexical items can be combined to form a sentence. An example of this, adapted from Jackson & Ruf (2017) is the following:

e. The man sends the boy a card

In this example, the double object dative is formed through the combination noun phrase + noun phrase, activated for the verb send. The link between this verb and the noun phrases is activated to form a sentence, after which the activation remains active for several seconds. This could prompt the use of a second double object dative in the next sentence. This approach can be described as the lexicalist ap-proach and is commonly used to explain short-term priming effect. The long-term priming effects that can last up to a week (Conroy & Antón-Méndez, 2015; Kaschak, Kutta & Coyle, 2014; Saffran & Martin, 1997) cannot be explained with the lexicalist approach; the activation of the links and nodes last for several seconds, which could not impact the production enough to last this long.

One of the explanations that have been found for long-term priming effects, is implicit learn-ing. Research from Chang, Dell & Bock (2006) describes how priming is an error-based learning pro-cess. They developed a model that illustrates how priming is accounted for in the development of lan-guage utterances. It can be explained as a preference model, which adds weight to a certain structure when it is used. Structures with more weight are more likely to be repeated by the speaker, causing more weight to be added.

As this process is circular, the system pushes certain structures which causes a cumulative change in the language production system (Jackson & Ruf, 2017). This change enhances language learning.

Priming, as mentioned, can be used for implicit learning, which is specifically interesting for second language acquisition. Several researchers have investigated the possible benefits of structural priming for L2 development. Independent from the type of priming, many studies have found that priming shows a beneficial effect to the learning of a second language (Gries & Kootstra, 2017; McDonough & Kim, 2009; McDonough & De Vleeschauwer, 2012). The effect has been found in Thai learners of English as their L2, who started producing more wh-questions after being primed in their L2. Conversely, a study from Bernolet, Hartsuiker and Pickering (2013) shows that priming effects show an increasing tendency when L2 proficiency improves.

According to Flett (2003), priming effects in a study into Spanish as the first and second lan-guage of participants showed a small but significant difference between the use of a passive construc-tion. The L2 participants were more receptive for priming than the L1 speakers of Spanish. The partic-ipants of this study were highly proficient speakers of Spanish. A more detailed explanation of implicit learning will be discussed in the next chapter.

1.2 Implicit Learning

Learning something new occurs from the start of life till the very end. Children gradually learn how to sit, walk, and eventually speak. A friend might learn how to perform a new trick on his skateboard and a grandmother might learn how to use her smartphone to make a videocall to her grandkids. All these types of skill acquiring happens through explicit help and instruction: someone else tells us if we are doing it right or helps us non-verbally by lending a hand. Although this form of teaching happens ex-plicitly, a child learns how to sit without learning ‘rules’ for sitting. No parent tells their child to use their abdominal muscles to keep upright while sitting. Still, by doing and over time, the skill is acquired. This same phenomenon occurs with many grammatical rules.

People learn how to comprehend and form grammatically correct sentences without being able to explicitly articulate the rules for doing so. Dienes & Berry (1997) have investigated the subject of implicit learning. They have distinguished three types of implicit knowledge; according to their re-search, knowledge might be called implicit because articulating the knowledge is difficult, but also be-cause it can be implicit according to an objective or subjective threshold. These thresholds can be iden-tified as a point that the speaker indicates as knowledge or not. It is argued that implicit learning occurs when the attention of the participant is focused on items and not the underlying rules that support the item (Dienes & Berry, 1997).

Shin and Christianson (2012) describe implicit learning as an automated process that handles the acquisition of a complex structure without the explicit intent to do so. The knowledge that results from it, is difficult to express explicitly. In Dutch, this often happens to people using articles of nouns. Most native speakers of Dutch know that the correct article for tree is ‘de boom’ and for house, ‘het huis’. If you were to ask for the specific rule and any exceptions to it, most of them would be unable to articulate this. Opposed to implicit learning, explicit learning occurs fully conscious and results in knowledge that can be expressed verbally.

In second language acquisition (SLA), implicit learning would account for the effects of priming on a long-term basis. It implies that learners are occupied with producing and comprehending words and sentences without consciously focussing on the syntactic structure underlying the sentence forming (Shin & Christianson, 2012). Learning a language via implicit learning is a time-consuming process, which is seen as less effective and efficient than explicit instructions for SLA (Norris & Ortega, 2000).

Research from Segalowitz and Hulstijn (2005), however, shows that L2 learners often struggle with difficulties in conveying ideas in their second language. It is argued that this is prompted by a lack of automaticity, which could indicate a lack of implicit knowledge for production of a second language. This, in turn, could indicate that implicit learning, however less efficient, is an important element in acquiring a second language. This is also supported by several researchers, implying that implicit learning is more effective in learning complex structures (Krashen, 1994; Reber, 1989)

1.3 Structural priming in L1 compared to L2

Research suggests that language use differs between native and second language speakers. It is be-lieved that native speakers easily adapt their language use to the situation presented, which can apply to speaking pace, articulation, referring to objects and syntactic structures (Kaan & Chun, 2018; Kraljic, Samuel & Brennan, 2008; Bock, 1986).

The adjusting that is observed can be explained by the theory of implicit learning; Chang, Dell and Bock (2006) argue that speakers predict what the next utterance will be and when this expectation is not met, the knowledge of the user is updated with a new possibility (Kaan & Chun, 2018). A thorough analysis from Pickering and Ferreira (2008) has shown that priming is observed in both written and spoken language production, for which both comprehension and production are influenced by priming.

The effect of priming differs in magnitude; it is found that the lexical boost effect accounts for this difference. This effect describes how short-term priming has a stronger effect when target struc-tures have the same verb as the prime strucstruc-tures. Target strucstruc-tures that do not have any lexical overlap with the priming structure, show smaller priming effects (Hartsuiker, Bernolet, Schoonbaert, Speybroeck, & Vanderelst, 2008). Additional research from Rowland, Chang, Ambridge, Pine, and Lieven (2012) confirms the lexical boost effect. They investigated three participant groups, varying from 3- to 4-year-olds to adults, and presented them with double object datives as in f and preposi-tional datives as in g:

f. Wendy gave Bob a dog. g. Wendy gave a dog to Bob.

Half of the participant group was presented with identical verbs in the prime and target sentence, for instance ‘gave’ for both sentences. Others received differing verbs, for instance ‘sent’ and ‘gave’. The adult participant group showed a significant increase in priming effects with the lexical overlap sen-tences. This adds to the body of research that acknowledges that the effect of priming is influenced by different factors.

Causes of syntactic priming have been described by Jaeger and Snider (2013), explaining that priming is dependent on the prediction error. Jaeger and Snider (2013) describe this prediction error as the deviation between what a speaker hears or observes and the expectations prior to the utterance.

It is found that the prior and recent experience with the prediction error that occurred impacts the processing of a prime sentence, where a stronger priming effect is found for the items with the biggest prediction error. A big prediction error implies that the speaker did not expect the utterance and expe-rienced a large surprisal effect. This supports the finding that syntactic priming is linked to expectation adaption for native speakers (Jaeger & Snider, 2013).

A study from Kaan and Chun (2018), investigated native and second language speakers of Eng-lish to determine if SLA is based on similar mechanisms as native speakers use for language adaption. This study focussed on language production rather than language comprehension but is nonetheless an important contribution to the research into L1 and L2 learning mechanisms for this thesis.

Within adaptation to syntactical structures, an inverse frequency effect is commonly observed; the structures that the speaker expects least are structures that are infrequently used. These structures cause the biggest priming effect when compared to more common structures (Kaan & Chun, 2018; Hartsuiker & Westenberg, 2000; Jaeger & Snider, 2013). This difference is explained by the ‘surprisal effect’ as described by Jaeger and Snider (2008). They state that ‘syntactic persistence is sensitive to surprisal’. Their research shows how learning and persistence are both error-driven: the effect is stronger when the prime structure is less expected. This effect is longer lasting than the lexical boost that is prompted when lexical items overlap in the prime and target sentence (Hartsuiker et al., 2008).

The surprisal effect occurs with structures that are not expected. According to Jaeger and Snider (2008), surprisal is defined as a structure with a low probability of occurring. This low proba-bility could be caused by several elements in the structure and is dependent on the sensitivity of the speaker. For the purpose of this thesis, when speaking of the surprisal effect, the contra-intuitive in-terpretation of the target structure will be meant.

As native speakers are adaptive with their speech production, it is expected that they are re-ceptive to priming, with the biggest priming effect in structures that are relatively infrequent to the speakers (Kaan & Chun, 2018). According to research from Reitter and Moore (2011), L2 learners have similar implicit learning mechanisms as native speakers and would therefore also show a bigger prim-ing effect for infrequent structures.

Taking the surprisal effect into account, L2 speakers would show greater effects of priming than native speakers with structures that are unknown in their first language. Research from Conroy & Antón-Méndez (2015) shows that participants can be primed for structures that do not exist in their L1. Taking a structure that is completely unknown and presenting it to a participant could maximize the surprisal effect. The research of Kaan and Chun (2018) showed no significant differences between the priming effects on L1 and L2 speakers, which can be explained by the syntactic structure that was used. For both language groups of the study, the syntactic structure was known in their native lan-guage. It is expected that a significant difference would be visible if the syntactic structure that was used, did not exist in Korean. The American English speakers did however show a preference for the double object ditransitive alternate whereas the Korean English speakers show a strong preference for the prepositional object ditransitive. This can be explained because the prepositional object ditransi-tive is more commonly used in Korean than the double object. Both structures are, however, possible and familiar for the Korean L2 speakers of English.

In the current research, the relative clauses (RC) and prepositional phrases (PP) carry an ambiguous interpretation for Dutch speakers. Both structures have two possible interpretations: subject versus object in the RC structure and agent versus patient in de PP structure.

Research from Belletti and Contemori (2009) and Guasti and Cardinaletti (2003) shows that partici-pants show a strong preference for producing subject-focussed relative clauses. When primed for an object-focussed relative clause, most participants used a passive form. One theory that is used to ex-plain this phenomenon, is the relativized minimality. Research from Rizzi (1990) and Friedmann, Bel-letti and Rizzi (2008) explains that the production and comprehension of relative clauses is impacted by the following: ‘a local relation cannot hold between X and Y when Z intervenes, and Z is somehow a potential candidate for the local relation’. When analysing an ambiguous relative clause, the interven-ing subject prevents speakers from logically forminterven-ing the object-focused structure between X and Y. In the subject-focused structure, there is no intervening subject, as the two subjects X and Z are placed together.

It is thus expected that, when priming participants, the object structure will have greater im-pact than the subject structure, as this is the less common alternative. Additionally, it is expected that for the L2 speakers of Dutch, proficiency plays a role in the effectiveness of priming. Research from Hartsuiker and Bernolet (2015) shows that priming effects differ among low-proficient and high-pro-ficient L2 speakers of Dutch. It was found that priming from L1 to L2 is stronger when the participant is more proficient in the L2. This implies that syntactic sharing across the L1 and L2 is more likely if the proficiency improves. The findings for the low-proficiency speakers was mixed: for related meaning conditions, the priming appeared to be more explicit-memory based. However, with different meaning conditions, the priming effect appeared stronger. The underlying theory assumes that SLA occurs in stages, and in the primary stages the speaker does not have syntactic representations yet. If possible, syntactic structures from the L1 will be used to form sentences in L2. When the proficiency improves, the speaker is also more likely to develop syntactic structures in the L2. This would impact the effect of priming and thus indicate that more proficient speakers of Dutch react similarly to priming as the na-tive speakers.

1.4 Target structures

For the current research, a previous study has been used to make a comparison between priming effects among native and non-native speakers of Dutch. The non-native speakers of the research were L1 speakers of German and L1 speakers of English who have learned Dutch at a later stage in life. It is assumed that English speakers interpret relative clauses differently from German speakers, whereas the German speakers allow for the same interpretation as the Dutch speakers. An illustration for this interpretation is given below. According to research from Schriefer, Friederici and Kühn (1995), Eng-lish speakers show a preference towards interpreting relative clauses in a subject structure instead of in an object structure. An adapted example from Schriefers et al. (1995) is shown below:

h. The professor that attacked the visitor admitted the error. (Subject) i. The professor that the visitor attacked admitted the error. (Object)

These sentences are both grammatically correct and unambiguous, but it is found that English speak-ers are more comfortable with interpreting the sentences as in h that in i (Schriefspeak-ers et al., 1995).

When looking at the syntactic system of German and Dutch speakers, an ambiguous interpretation of sentence j is possible, as shown in Dutch below:

j. De professor die de man aanviel hield een promotiepraatje.

De professor, in this sentence, can be perceived as the attacker as well as the attacked. The intended interpretation can be distilled via either interpunction in writing, or intonation in speaking. Whenever no interpunction is placed, the reader needs to decide what the most logical interpretation is. According to Fox and Thompson (1990), the English syntactic structure does not allow for an ambiguous inter-pretation of relative clauses, which could impact the comprehension of the subject interinter-pretation in the second language Dutch. In a study of Schriefers et al. (1995), German L1 speakers were presented with a sentence that was either subject- or object structure. The sentence would be introduced with an element that contained an element that would steer the sentence semantically. There were three alter-natives that the participant could be presented with, a subject interpretation was the first option and the object interpretation the second option. The third alternative was a neutral introduction. The re-sults showed that German speakers had a tendency towards the subject interpretation when presented with a neutral sentence.

Additional research shows that the interpretation of syntactic structures happens primarily on an intuitive basis; the speakers discovers the sentence word for word, as soon as the string of words indicates a certain structure, this is recognized by the reader (Garrod & Sanford, 1985; Tanenhaus, Carlson & Seidenberg, 1985). This process is difficult when the sentence offers an ambiguous option for interpretation, for example in the Dutch sentence in k and l (adapted from Schütze, 1995):

k. De vrouw zag de buurman… (The woman saw the neighbour)

l. De vrouw zag de buurman met de verrekijker (The woman saw the neighbour with the binoculars)

Research from Frazier (1990) shows that the preference of the participant is always directed to the VP-attachment, which indicates that in the example, the woman uses the binoculars to spot the neighbour. Additional research from Frazier (1987) shows that Dutch and English share the same structure pro-cessing in prepositional phrases.

1.5 Priming and time

A critical aspect of priming studies is the possibility to influence the unconscious language processing of the participant. Additionally, participants analysing and weighing options rationally, might grow aware of the purpose of the study, thus impacting the results. Research from Janis (1983) showed that high stress-levels prompts participants to make decisions prematurely, without revising all possible alternatives. It is assumed that time pressure increases stress-levels and triggers premature closure of decision making (Kruglanski & Webster, 1996).

According to the overview from Edland and Svenson (1993), most studies show that time pres-sure has an effect on decision making and increase selectivity of input of information. Additional re-search from Dijksterhuis (2004) shows that participants that were put under time pressure to make a decision, made better decisions overall. The conclusion of the study was that unconscious thinkers are better at decision making, leading to more clear, polarized and integrated representations.

To analyse the effect of priming, it would be interesting to see if time-pressure impacts the decision making of participants. This could be done by either limiting the time to answer, or by instructing the participants to work as quickly as possible. For the purpose of this thesis, half of the participants were instructed to answer as quickly as possible, entering their first instinct. This limits overthinking or thoroughly analysing the possibilities, while still allowing the participants to take in all the infor-mation presented to them.

1.6 The present study

The goal of this thesis is to compare the effects of structural priming in L1 speakers to L2 speakers of the same language, Dutch. In this investigation, several possible influencers are being analysed to ex-amine the effect of priming and the underlying role it plays in implicit learning. Previous research has investigated the effects of priming in L1, L2 and the impact of priming on implicit learning. It has re-cently been researched if there is a difference in the effects when comparing L1 and L2 in the same experiment with identical primes. It appears this is dependent on the type of prime and the proficiency of the speakers. It is thus very interesting to see if the results of this study align with previous results.

It is examined whether structural priming impacts the comprehension of certain grammatical structures and if there is a long-term effect, implying the process of implicit learning. On account of the role of the first language, there is an ongoing debate. Some research has suggested the role of the first language in priming and second language acquisition is negligible. Multiple other studies how-ever, have shown that the first language greatly impacts the effects of priming and SLA (Hartsuiker & Bernolet, 2015; Jarvis, 2011; Odlin, 2005).

If the first language indeed impacts these effects, the shared and differing grammatical attributes be-tween Dutch, English and German could provide an insight (Fox & Thompson, 1990; Frazier, 1990; Schriefers, Friederici & Kühn, 1995). This study is based on the following research questions:

1. To what extent does priming impact the immediate comprehension of relative clauses during the (delayed) post-tests in L1 Dutch speakers and English and German L2 speakers of Dutch?

2. To what extent does priming impact the immediate comprehension of prepositional phrase attachments during the (delayed) post-tests in L1 Dutch speakers and English and German L2 speakers of Dutch?

3. To what extent does a time-instruction impact the effect of priming?

It is expected that speakers show a tendency to choose for one of the structures, with a significant dif-ference between de English speakers and the German and Dutch speakers. It is expected that the Eng-lish speakers will choose the object interpretation more often as the subject interpretation of the rela-tive clause does not exist in English and the Dutch and German speakers will choose the subject inter-pretation more often. This difference will be visible in the relative clauses prior to the treatment phase, as English differs from German and Dutch in the syntactic structure underlying relative clauses. Ac-cording to research from Jackson (2018), priming is less effective if the primed structure is aligned with the original preference of the participant.

This effect is also found in the research from Kaan and Chun (2018), who found that participants show a higher adaptability when confronted with structures that were previously unknown to them. Based on the current knowledge in the field of structural priming, the following hypotheses have been made:

1. In a pre-test prior to priming treatment, L1 Dutch speakers will show a preference towards the subject-structure of the relative clauses and the agent-structure of the prepositional phrases. German speakers will show similar preferences to Dutch, whereas the English speakers will show a preference towards the object-structure of the relative clauses. For the prepositional phrases, no difference in preference is expected from the Dutch speakers. 2. Immediately after the priming treatment, speakers will more frequently choose the

primed structure, additionally:

a. If the preference for the structure is coherent with the primed structure, the effect of priming will be less visible. This implies that there will be a smaller interaction-effect with the subject priming in relative clauses. The Dutch and German speakers will show a smaller priming effect on the relative clauses when primed for the sub-ject interpretation than the English speakers. For the prepositional phrases, it is expected that all three groups show a smaller priming effect when primed for the agent interpretation.

b. If the preference for the structure is not coherent with the primed structure, the effect of priming will be more strongly visible. This implies that there will be a greater interaction-effect with the object priming in relative clauses. The Dutch and German speakers will show a bigger priming effect on the relative clauses when primed for the object interpretation than the English speakers. For the prep-ositional phrases, it is expected that all three groups show a bigger priming effect when primed for the patient interpretation.

In the ongoing debate about the role of proficiency and first language in priming, the results from the Dutch participants will be compared to L2 speakers of Dutch. It is assumed that the English speakers will show greatest priming effects in the relative clauses, as they will be presented with a structure that is not allowed in their L1. Therefore, the surprisal effect will be maximized (Jaeger & Snider, 2008). It is assumed that German speakers will show less priming effect that the English speakers, as they are already familiar with both structures. However, the effects will be bigger for both Dutch and German speakers in the relative clauses than the English participants; the object-interpretation is assumed to be less preferred among speakers of Dutch and German (Friedmann, Belletti, & Rizzi, 2008). Therefore, the last hypothesis is as follows:

3. The effects of priming in L1 Dutch speakers will be less significant than the effect of prim-ing in L2 Dutch speakers with English as mother tongue. The effects of primprim-ing will be most visible in the relative clause priming, the prepositional phrases will show the least effect.

According to Lemhöfer and Broersma (2012), the LexTale is a valid tool for measuring proficiency in speakers of L1 and L2. Research from Hartsuiker and Bernolet (2015) shows that proficiency impacts the effect of priming, with a bigger priming effect in more proficient speakers. This could indicate a possible impact on the results of priming. Therefore, the following is expected from the results of the current study:

4. The LexTale is predictive for the effects of priming, a higher score will result in bigger priming effects.

As described in paragraph 1.5, research has shown that time-pressure impacts decision making and minimizes analytical thinking. Half of the current participant group received instructions to be as quick as possible, without specifically limiting the time they received for answering. It is assumed that this will impact the effect of priming, without prompting premature decision making (Dijksterhuis, 2004). The following is assumed:

5. A time-instruction will impact the response time of Dutch participants significantly and will impact the effect of priming positively, showing more priming effects if speakers were instructed to be quick.

2. METHOD

For this thesis, two experiments have been conducted and analysed. The first experiment was done by the research group ‘Nederlands als Tweede Taal’ (Dutch as Second Language) with English and German L2 speakers of Dutch. This experiment has been repeated with Dutch speakers to examine the effects of the first language in priming. For comparative reasons, only a few changes have been made in the repetitive study.

2.1 Participants

For this thesis, a group of 24 participants has been tested. The research group consisted of 50% females (n=12) and 50% males. All participants were first language speakers of Dutch and currently living in The Netherlands. For the sampling of the participants, a non-probability sampling method was used, the participant group consisted of close relatives. This convenience sampling impacts the representa-tiveness of the results for the study (Etikan, Musa & Alkassim, 2016). However, this approach was most appropriate for the design of this study. The original experiment consisted mainly of students and highly educated adults. Focusing on the importance of a relevant comparison in results implied creat-ing a comparable participant group, which could only be ensured if samplcreat-ing was not done randomly. Additionally, there was a limited timeframe and no monetary incentive for the participants in the first experiment. Close relatives were more willing to spend their time on a voluntary basis. Partic-ipants were approached on a personal basis and consisted mostly of a network of students from Dutch Universities. The range of age differed between 18 and 65 years old (M=35.5, SD=15.8). Participants were requested to enter their demographic information in a questionnaire. One of the questions regarded the highest educational level, resulting in the following scale: University Master (63%), HBO (29%) and MBO (8%).

Participants were asked to report their self-rated proficiency of Dutch in speaking, writing, reading and listening, the age they started learning Dutch at school and any other languages they ac-quired besides Dutch. Additionally, they were asked to indicate which activities they performed more often than twice a week regarding reading, listening and writing. Participants rated their Dutch profi-ciency in reading between 4 and 7 (M=6.3, SD=0.7.) on a scale of 1 to 7. The question about the Dutch education was deleted before analysis, as most participants did not understand this question and indi-cated not to have knowledge of this. At least 83% of the participants indiindi-cated to write professionally and to friends and family more than twice a week. Reading for work was also done more than twice weekly by 83%. Additionally, 96% of the participants indicated to watch television or Netflixmore of-ten than twice a week.

For comparative research, the Dutch participant group was compared to the research partici-pants from a previous study into Dutch as a second language. These participartici-pants were second language speakers of Dutch, minimal proficiency at A2 level. For these participants, first language was German or English. The German language has a similar grammatical structure as Dutch, whereas English does not allow for an ambiguous reading of the relative clause Waar is de tijger die het schaap kust? (Where is the tiger that the sheep kisses?) as illustrated in figure 1.

The second participant group consisted of 16 English and 14 German speakers with 79% female (n=22) and 21% male (n=6), collected by the research group NT2.

Participants were gathered through active recruitment via flyers, social media posts and promotion in Dutch classes. Experiment 2 was conducted before experiment 1 and contained the same elements as experiment 1. One distinction between the two experiments was made by the time instructions. No par-ticipants in experiment 2 received any instructions regarding response time in the Structural Priming Task. Additionally, the participants from experiment 2 received €10 for their participation.

Figure 1. German and English interpretation of the Dutch relative clause.

All participants indicated to have University as educational level, of which 64% had obtained a degree at the time of participating. The average number of years that participants spoke Dutch was 8.7 years on average for the entire group (SD=10.3, min.=1, max.=36). More specific had German L1 speakers 11.2 years on average (SD=12.4, min.=1, max.=36) and the English L1 speakers 6.5 years on average (SD=8.1, min.=1, max.=33). All participants were presented with a self-rating question in the questionnaire on their L2 Dutch skills, using a 7-point scale. For the research, only reading held relevant information. The analysis revealed that German speakers (M=6.0, SD=0.7) considered their proficiency significantly higher than the English speakers (M=4.7, SD=1.2). This difference was in line with the results of the proficiency test that was used. Analysing the LexTale results, German speakers on average scored sig-nificantly higher (M=45.7, SD=5.7) than the English speakers (M=41.9, SD=4.0) with a maximum score of 60 correct answers; t(26)=2.21, p=.036. A table with the results and additional demographic infor-mation is added below.

Table 1. Demographic information participants EN G LI SH P-va lu e .5 6 2 .1 4 7 .2 4 0 .6 4 0 .0 0 0 * .0 0 0 * .0 0 0 * Ra n ge 20 -63 ₀-6 1-33 19 -32 34 -49 56 .7% -81 .7 % 3-6 M (S D ) 31 .4 ( 10 .7) 9 1. 73 (1. 4 4 ) 6. 5 (8. 1) 24 .9 ( 3. 6 ) 41 .8 (4. 1) 69. 7( 6. 8) 4. 7 (1. 2) GE R MA N Ra n ge 19 -65 _1-5 1-36 17 -43 39 -57 65 % -95 % 4 -7 M (S D ) 37. 1 (1 5. 5) 9 2. 3 (1 .20 ) 11. 2 (12 .4 ) 25. 8 (6 .1 ) 45 .7 ( 5. 7) 76 .2% ( 9 .2) 6. 0 (0. 7) DUT CH Ra n ge 18 -65 _1-3 49 -59 81 .7 % -98. 3% 4 -7 M (S D ) 35. 5 (1 5. 8) 15 1. 5 (0 .7) - - 55. 3 (2. 5) 92 .2 % ( 4. 2) 6. 3 (0. 7)

A ge Ob ta in ed d eg re e Nu m be r of L 2 Nu m be r of y ea rs D ut ch Ag e st ar t le ar ni ng Du tc h Av er ag e ab so lu te L ex Ta le s co re Av er ag e pe rc ent ag e Le xTa le s co re Se lf -ra ti n g pro fi ci en cy re ad in g D ut ch 1 -7 * A ve ra ge d iffe re nc e si gn ifi ca n t at .0 5 le ve l

2.3 Materials

The study consisted of two consecutive days, during which participants were tested for 30 minutes per day. The first day, the participants were presented with a Structural Priming Task (SPT) and a Visual Digit Span Task (VDST). The second day started with a SPT, followed by a Flanker Test and a LexTale. Participants were asked to fill in a brief survey for demographic information after their participation. The Visual Digit Span Task and Flanker Test were tested as part of a bigger research. The results of these tasks were only used as control variables and did not lead to exclusion of participants.

The main focus of the research lies within the SPT, which consists of a pre-test, a treatment phase, a post-test and a delayed post-test. The delayed post-test was performed on the second day, thus addressing the possibility of any lingering priming effect. All participants were randomly assigned to one of four SPT versions and time-instructions were randomly divided across the participant group. The structural priming task was made with the program E-Studio 2.0 and was administered on a laptop from Leiden University. The primary experiment was written and designed by dr. Bosma, pictures for the relative clauses were provided by Kidd, Chan and Chiu (2015). Additionally, pictures for the prepo-sitional phrases were used in a previous study about priming, executed by Kootstra (in prep.). A com-plete set of the used images, 380 pictures, is added as Appendix III. The first-day experiment consisted of a pre-test, a treatment phase and a post-test. All three phases consisted of relative clauses, prepo-sitional phrases and fillers and will be described below.

2.3.1. SPT - The pre-test, treatment and post-test

All three phases consisted of identical elements with identical duration, which are shown in table 2. These phases were represented in all four versions of the test, which were dependent on the primed target. The four version were AO (Agent and Object primed), AS (Agent and Subject primed), PO (Patient and Object primed) and PS (Patient and Subject primed). All participants were randomly assigned to one of the versions, all versions were equally represented in the participant group. Additionally, half of the group with Dutch participants received time-instructions. For half of the group, a quick response was requested, urging to answer on first instinct without thinking. The other half of the group received no time-instructions. After this instruction, the participant was installed with the laptop.

The introduction screen was shown at the beginning of the test, explaining what was expected of the participant. The participant read the introduction and was instructed to start the test by pressing the spacebar. After that, a sentence display was shown with an infinite duration. This display contained a Dutch sentence describing a situation and was formulated in a question (e.g. the relative clause: Waar is het schaap dat de beer kust? - Where is the sheep that the bear kisses?). The participant could read the sentence for as long as s/he needed.

When the participant pressed the spacebar to continue, a blank screen was shown for 1000 milliseconds, after which an image display with two pictures was shown. The participant could choose either one of the pictures by pressing z (left picture) or / (right picture). The duration was set to infinite, the participant could take as long as s/he needed. After choosing, a blank screen was shown again for 1000 milliseconds, after which the next sentence display followed. All elements of the task consisted of this structure, which is illustrated in figure 1.

The pre-test consisted of 8 target structures for the relative clauses, 8 target structures for the prep-ositional phrase attachments and 8 fillers. The treatment phase consisted of 24 fillers, 24 prepprep-ositional phrase attachments and 24 relative clauses. The post-test and the delayed post-test were designed like the pre-test, with 8 sentences of each. All elements of the task were pseudo-randomized in the exper-iment to ensure a good balance in the order of presenting. The pseudo-randomization presented par-ticipants with a fixed order of a relative clause, a prepositional phrase and a filler.

Table 2. Representation of elements in SPT

Relative clause Prepositional phrase attachment Filler Pre-test 8 target structures 8 target structures 8

Treatment 24 24 24

Post-test 8 8 8

Delayed post-tests 8 8 8

The filler structures were added to ensure participants would not become consciously aware of the priming stimuli. An example of a filler was de sentence ‘Waar is de blije kikker?’ (‘Where is the happy frog?’) showing a picture of a crying frog and a picture of a smiling frog. Additionally, these served as control factor to check if participants understood the task they were presented with. A second control factor was built in the treatment phase, of which participants needed to score at least 60% correct. This treatment phase consisted of a sentence with two pictures, of which one was clearly incorrect. E.g. ‘Waar is de aap die de giraffe duwt?’ (‘Where is the monkey that the giraffe pushes?’), followed by a picture of a giraffe and a monkey, and a picture of a giraffe and a cow. Which picture was replaced with an impossible alternative was dependent on the version of the test the participant received.

After the last sentence and picture display, the participant was presented with a screen ex-plaining the first part of the experiment ended. Any questions or remarks the participants made were redirected and answered after the experiment on the second day. After this, the second task was started for the participant.

Sentence Sentence Infinite Infinite Infinite Infinite 1000ms Figure 2. STP Structure

2.3.2. Visual Digit Span

After completing the SPT, all participants were presented with a Visual Digit Span Task. A Visual Digit Span Task (VDST) is a tool for measuring intelligence and is assumed to reflect short-term memory output (Wechsler, 1941).

For the purpose of this thesis, the VDST is used as control variable for minimal performance. Addition-ally, it will be used to analyse if scores on the VDST correlate with the effect of priming in the SPT. Starting the VDST, the participant was presented with an introduction screen. If no questions remained after reading the introduction, the participant started the practice phase. This consisted of 3 practice runs, adaptive to the occurrence of mistakes. Participants were provided with feedback after each prac-tice run (e.g. ‘Well done’, ‘Too bad, try again’). After the pracprac-tice phase, the participant pressed the spacebar to start the test.

The first half of the test, participants were requested to remember and reproduce the numbers that were showed on screen. After the last number, a screen was showed where the answer had to be filled in. The amount of consecutive numbers that were shown to the participant started at 1 and were adaptive to the number of mistakes, with a maximum of eight consecutive numbers if the participant did not make more than 3 mistakes consecutively. These eight blocks consisted of six trials per block. If the participant filled in the first four trials flawless, the fifth and sixth trial would be skipped, as-sumed these would be answered correctly as well. If the participant made one mistake in the first four trials, only the sixth trial would be skipped. After the first half, the participant was requested to fulfil the same task, reproducing the numbers backwards. This part also started with a practice phase and was adaptive to the number of mistakes, with a maximum of seven consecutive numbers. Afterwards, the scores were revised. For every participant that had trial 5 and 6 skipped, two points were added. For the backwards block, six points were added to all participants, as the first block could not be tested: only one number cannot be remembered backwards.

2.3.3. SPT - Delayed post-test

On the second day, participants were presented with a shorter version of the SPT they received the day before. This version did not contain a treatment phase and was thus much shorter with 8 target struc-tures for the relative clauses, 8 target strucstruc-tures for the prepositional phrases and 8 fillers. Participants received the same version of the delayed post-test as the version they were presented with during the first day. After completing the SPT, an instruction screen was shown indicating that this part of the experiment was finished, after which the Flanker test was set up.

2.3.4. Flanker test

The Flanker test (Eriksen & Eriksen, 1974) consisted of 70 tasks asking the participant to indicate the direction of an arrow. The arrow they had to focus on was the middle of five arrows. The other four arrows served as distractors. Participants could indicate the direction of the arrow by pressing either the left or right arrow on the laptop. The test consisted of congruent and incongruent conditions. In the congruent condition, all arrows pointed to the same direction, in the incongruent condition, the target arrow pointed the other way (illustrated in figure 2, adapted from Lee, Han, Lee & Choi, 2010).

The participant was presented with two instruction screens and could enter the practice phase by pressing the spacebar. Participants were presented with feedback during this practice phase (e.g. ‘A little faster’, ’Good job’, ‘Too bad, try again’) after which they could ask any remaining questions. The official task started with a blank screen, shown for 1000ms after which a fixation plus-sign was shown for 500ms. The arrows were displayed for 1000ms, after which the participants had 2000ms to answer. The Flanker test can be used to predict information processing and has been used to study memory processes (Eriksen, 1995). For the purpose of this thesis, the Flanker test is used as control variable to see if participants were able to make a distinction between left and right, which could impact their results in the SPT. Additionally, the Flanker test can be used to predict inhibitory control. It is assumed that congruent Flanker pictures are answered more rapidly than incongruent pictures. Par-ticipants who are quick on the incongruent pictures as well, score a smaller Flanker effect than partic-ipants who need a little more time for incongruent pictures. A small Flanker effect indicates that par-ticipants are relatively good at suppressing irrelevant information. The study of Fuentes, Vivas and Humphreys (1999) has found that it is possible to draw attention away from familiar information to new information, which would improve the results of priming. Participants who have a small Flanker effect might thus be better at suppressing irrelevant information and would be less impacted by prim-ing.

2.3.5. LexTale

As a final task, participants were asked to make the LexTale to assess their Dutch proficiency. The LexTale is a digital test that is developed by Lemhöfer and Broersma (2012) to provide a brief and valid proficiency test. Participants can take the test online, after which their score is shown. In this experi-ment, the participants took the LexTale on the laptop in an excel file. They were presented with 60 words and were asked to indicate whether these words exist in the Dutch language. From this list, 33% (20 words) were made up and should be recognized by the participant as non-existent. They could take as long as needed. Participants entered their answers with 0 (this word does not exist) and 1 (this word exists).

Afterwards, these scores were recoded; participants would receive one point for a correct an-swer, which could have been either 0 or 1. All points for correct answers were added and divided by the maximum score of 60 to calculate the percentage. The lowest score for native speakers was 92.2%, for the German speakers 76.2% and for the English speakers the lowest score was 69.7%. According to Lemhöfer and Broersma (2012), all scores between 60-80% indicate a proficiency level of B2. All L2 speakers in this study were found within this range, thus indicating that all speakers had a reading proficiency of B2 in their L2 Dutch.

2.4 Procedure

Participants were asked to sign an informed consent sheet before starting the experiment. Any ques-tions could be asked beforehand or during the experiment. Specific information about the purpose of the experiment was discussed after the second day tasks. All participants were tested individually in a recording booth at Leiden University or in the privacy of their own homes over the course of two con-secutive days. This ensured a quiet space with a computer or laptop to run the experiment on. The Structural Priming Task of the first day took participants with a time-instruction about 15 minutes to complete. Participants without time-instruction took a bit longer, varying between 20 and 25 minutes. After the first task, another average of 15 minutes was spent on the Visual Digit Span Task. If partici-pants made a lot of mistakes, the task was shorter.

The second day, the SPT took about 10 minutes for all participants, after which the Flanker task was set up. If the participant made mistakes during the phase, e.g. responding too slow, the test-phase would last a little longer. After this, they spent about 10 minutes on the Flanker test and another 10 minutes on the LexTale. The Qualtrics questionnaire took about 3 minutes to complete. After the tasks, the participants were asked about their idea of the research. All answers were logged. After this, all participants received a debriefing about the purpose of the study.

2.4 Scoring

In the experiment, the SPT collected scores from participants dependent on the version of the experi-ment they received. In E-Studio 2.0, all answers had to be marked right or wrong. Whenever the par-ticipant opted the picture that was in line with the prime stimuli they received, the answer was marked correct and one point was added. A participant with the AO (Agent-Object) version would receive one point for the picture with the Agent interpretation and one point for every picture with the Object in-terpretation. Whenever the participant had chosen the Subject or Patient interpretation, zero points would be added. Before the analysis, the scores were recoded; all subject and agent interpretations were marked with one point, regardless of the test version. The final analysis was run with data for all the subject and agent interpretations. During the treatment phase, only one answer was logically correct. The answers for this phase were not recoded; a correct answer to the question would result in one point, a wrong answer would result in zero points.

3. RESULTS

For this thesis, three groups of participants were tested using Dutch relative clauses and prepositional phrases. The participant groups were based on mother language, for which German, English and native Dutch were tested. Three research questions were posed regarding the effect of priming. It is investi-gated what the impact of the first language is on priming, and the possible effect of time-pressure on effectiveness of priming. In this chapter the results will be presented, which will be analysed in the discussion.

3.1 First research question

The first research question of this thesis was to what extent priming impacts the immediate compre-hension of relative clauses during the (delayed) post-test in L1 Dutch speakers, L1 English and L1 Ger-man speakers of Dutch. In a pre-test prior to priming treatment, all participants were presented with relative clause sentences and two pictures that represented the subject- and object interpretation. The results were analysed using a one-way ANOVA. This showed that there is a significant difference in preference during the pre-test between the subject and object interpretation of relative clauses be-tween the Dutch, German and English speakers, F(2,166)=80.729, p<.001. Dutch and German speakers showed a preference for the subject interpretation, respectively M=7.58, SD=0.88 and M=7.14, SD=1.83. The English speakers preferred the object interpretation, resulting in a low score for the subject inter-pretation (M=1.87, SD=1.73). An independent t-test was performed for the pre-test on the relative clauses between all language groups, comparing Dutch to German, Dutch to English and German to English with language as between-subject variable. These results showed no significant difference in the interpretation of the relative clauses for the Dutch (M=7.58, SD=0.88) and the German speakers (M=7.14, SD=1.83); t(36)=1.00, p=.28. There were, however, significant differences between the Dutch (M=7.58, SD=0.88) and English speakers (M=1.87, SD=1.73); t(37)=13.69, p<.05, where the English speakers choose the subject interpretation significantly less often. The German and English speakers subsequently showed a significant difference, where the German speakers (M=7.14, SD=1.83) choose the subject interpretation significantly more often than the English speakers (M=1.87, SD=1.73); t(26)=11.56, p<.01.

A repeated measures ANOVA was done for the relative clauses with time as within-subject var-iable (pre-, post- and delayed post-test) and mother language and type of priming as between-subject variables. A Helmert contrast was used to compare the pre-test to the post- and delayed post-test, and the post-test to the delayed post-test. The post-test and delayed post-test will be indicated as ‘later’ in this thesis. An interaction effect was visible in the relative clause data between time and type of prime in the comparison of pre-test to later (p<.001, h2_{p=.28). There was no significant interaction effect} vis-ible in the post-test compared to the delayed post-test (p=.34, h2_p=.02).

The pairwise comparisons show that during the pre-test, no significant interaction effect is visible between the subject and object type of prime (p=.09). However, during the post-test and the delayed post-test, there is a significant interaction effect visible between the subject and the object prime (p=.001, p<.05).

Additionally, a significant interaction effect was found between time, type of priming and language. This interaction effect was found between the pre-test and later (p<.05, h2

p=.13). There was no significant interaction effect between the post-test and the delayed post-test (p=.32, h2_{p=.05). The pairwise} com-parisons show that a significant effect occurs with the English speakers of Dutch, comparing the pre-test to later and post-pre-test to the delayed post-pre-test in the subject priming (p<.001, p<.05). The object priming showed no significant interaction effects, with no effects in the pre-test vs. later and neither in the post-test compared to the delayed post-test (p=.07, p=.21, p=.85).

The Dutch and German speakers showed no significant interaction effects for the subject prime, analysing pre-test vs. later (p=.64, p=.84) and post-test versus delayed post-test (p=.72, p=.15). The object priming also showed no significant effect (p=.09, p=.11, p=.10, p=.10). Additionally, it was investigated whether proficiency impacts the effect of priming among speakers of Dutch. A repeated measures ANOVA with within subject variable absolute LexTale score and a Helmert contrast for the var-iable time shows no significant interaction effect between the pre-test and later (p=.07) for the relative clauses.

3.2 Second research question

For the second research question of this thesis, the results of the prepositional phrases were analysed for the Dutch, German and English speakers as well. The results were analysed using a one-way ANOVA. This showed that there is no significant difference in preference during the pre-test between the agent and patient interpretation of prepositional phrases between the Dutch, German and English speakers, F(2,12)=2.113, p=.132. An independent t-test was performed for the pre-test on the prepositional phrases between all language groups, comparing Dutch to German, Dutch to English and German to English with language as between-subject variable. These results showed no significant difference in the interpretation of the relative clauses for the Dutch (M=3.50, SD=2.30) and the German speakers (M=5.00, SD=2.18); t(36)=-1.97, p=.63. There was also no significant difference between the Dutch (M=3.50, SD=2.30) and English speakers (M=3.40, SD=2.75); t(37)=.12, p=.30. The German and English speakers subsequently showed no significant difference, with the German speakers (M=5.00, SD=2.18) and the English speakers (M=3.40, SD=2.75); t(27)=1.73, p=.19.

A repeated measures ANOVA with Helmert contrast was performed on the data from the prep-ositional phrases with time as within-subject variable (pre-, post- and delayed post-test) and mother language and type of priming as between-subject variables. The pre-test was again compared to the post-test and the delayed post-test, the latter two indicated as ‘later’. This analysis showed a signifi-cant interaction effect between time and type of prime in the comparison of the pre-test with later (p=.001, h2_{p=.21). There was no significant interaction effect between the post-test and the delayed} post-test (p=.06, h2_{p=.07). The results showed no significant interaction effect between time, language} and type of priming (p=.13, h2

p=.08).

The pairwise comparisons show that during the pre-test, no significant interaction effect is visible between the agent and patient type of prime (p=.27). However, during the post-test there is a significant interaction effect visible between the subject and the object prime (p<.001). The delayed post-test showed no significant interaction effect (p=.06). Additionally, it was investigated whether proficiency impacts the effect of priming among speakers of Dutch.

A repeated measures ANOVA with within subject variable absolute LexTale score and a Helmert contrast for the variable time shows no significant interaction effect between the pre-test and later (p=.10) for the prepositional phrases.

Table 3. Results SPT per language group

Dutch German English

RC PP RC PP RC PP

Subject

Prime Object Prime Agent Prime Patient Prime Subject Prime Object Prime Agent Prime Patient Prime Subject Prime Object Prime Agent Prime Patient Prime

Pre-test 7.33(1.2) 7.83(0.4) 3.17(2.1) 3.83(2.5) 6.71(2.6) 7.57(0.5) 5.33(1.4) 4.75(2.7) 1.50(1.1) 2.29(2.3) 2.38(2.4) 4.57(2.8)

Posttest 7.58(0.8) 6.92(2.0) 3.5(2.4) 2.58(2.1) 6.86(3.0) 6.43(2.1) 5.67(2.4) 2.38(2.1) 5.38(1.6) 1.00(1.5) 6.13(2.6) 2.29(2.6)

Delayed

posttest 7.58(0.8) 6.67(2.4) 3.17(2.9) 2.50(2.8) 8.0(0.0) 7.7(0.5) 6.00(1.7) 4.38(2.7) 3.75(3.3) 1.14(1.5) 6.25(1.6) 4.29(3.0)

3.3 Third research question

The third research question is focussed on the data from the Dutch participants. For this question, it is analysed if a time instruction impacts the effect of priming in relative clauses and prepositional phrases. For this question, half of the group received instructions prior to the structural priming task to answer as quickly as possible. The other half of the group did not receive any time instructions. The results were analysed using an independent t-test. The results were mixed. It showed no significant difference in the pre-test for the relative clauses where the participants without time instruction (M=7.50, SD=1.17) and the participant with a time instruction (M=7.67, SD=0.49) chose the subject terpretation. The post-test did show a significant difference, where the participants with a time in-struction (M=7.58, SD=0.67) chose the subject interpretation more often than the participants that did not receive a time instruction (M=6.92, SD=2.07). The delayed post-test did not show a significant dif-ference. The results can be found in table 4. For the prepositional phrases, only the pre-test showed a significant difference in interpretation, the participants with a time instruction chose the agent inter-pretation more often (M= 4.08, SD=2.75) than the participants without a time instruction (M= 2.92, SD=1.68).

Table 4. Results time instruction SPT

Time instruction N Mean SD p-value

RC Pre-test 0 12 7.50 1.17 .243 1 12 7.67 0.49 RC Post-test 0 12 6.92 2.07 .045* 1 12 7.58 0.67 RC Delayed 0 12 7.00 2.30 .434 1 12 7.25 1.22 PP Pre-test 0 12 2.92 1.68 .089* 1 12 4.08 2.75 PP Post-test 0 12 3.25 2.53 .295 1 12 2.83 1.99 PP Delayed 0 12 3.00 2.89 .916 1 12 2.67 2.87

4. DISCUSSION

For this research, the effect of priming has been investigated among native speakers of Dutch and Ger-man and English L2 speakers of Dutch. Different aspects that might influence the effect of priming have been analysed with a focus on the presence of a syntactic structure in the L1 when priming in the L2, the difference between native and non-native speakers and the effect of time-pressure. The Struc-tural Priming Task has been used for this and the results will be discussed below.

4.1 Hypotheses

Before conducting the experiment, several hypotheses have been formulated, which will be discussed with regard to the results that have been found. The first hypothesis was as follows:

1. In a pre-test prior to priming treatment, L1 Dutch speakers will show a preference towards the subject-structure of the relative clauses and the agent-structure of the prepositional phrases. German speakers will show similar preferences to Dutch, whereas the English speakers will show a preference towards the object-structure of the relative clauses. For the prepositional phrases, no difference in preference is expected from the Dutch speakers. The results show that there is a significant difference in preference during the pre-test of the partici-pants. It was found that Dutch and German speakers have a strong preference for the subject interpre-tation of the relative clause, whereas the English speakers show a strong preference for the object in-terpretation of the relative clause. This was expected, as the subject inin-terpretation is not grammatically acceptable in English.

For the prepositional phrases, no significant difference in preference was found for the Dutch, German and English speakers. It was also found that there was no preference for either the patient or the agent interpretation visible within the language groups. This was expected as the prepositional phrases allow for ambiguous interpretation in all three languages (Frazier, 1990). Therefore, hypoth-esis 1 can be confirmed. For the second hypothhypoth-esis, the effect of priming in the and delayed post-test were analysed. These results were compared with the initial preferences that were analysed in the pre-test. Hypothesis 2 and 3 were formulated as follows:

2. Immediately after the priming treatment, speakers will more frequently choose the primed structure, additionally:

a. If the preference for the structure is coherent with the primed structure, the effect of priming will be less visible. This implies that there will be a smaller interaction-effect with the subject priming in relative clauses. The Dutch and German speakers will show a smaller priming effect on the relative clauses when primed for the sub-ject interpretation than the English speakers. For the prepositional phrases, it is expected that all three groups show a smaller priming effect when primed for the agent interpretation.

b. If the preference for the structure is not coherent with the primed structure, the effect of priming will be more strongly visible. This implies that there will be a greater interaction-effect with the object priming in relative clauses. The Dutch and German speakers will show a bigger priming effect on the relative clauses