Syntactic Priming in Comprehension

An investigation into the comprehension of

Dutch passive sentences by means of event-related potentials

E

DWIGE

S

IJYENIYO

Supervisors:

D

R

. S

IMONE

A. S

PRENGER

and P

ROF

. D

R

. R

OBERT

J. H

ARTSUIKER

U

NIVERSITY OF

G

RONINGEN

A Master’s thesis submitted to the University of Gronin-gen in accordance with the requirements of the degree of MASTER’S OFARTS.

Word Count: 18,800

S

yntactic priming is a phenomenon that facilitates the processing of a sentence given exposure to another, similarly-structured sentence beforehand (Ledoux, Traxler, & Swaab, 2007). Enhanced priming effects have been reported when there is verb overlap between the prime and target sentence (i.e., lexical boost) (Pickering & Branigan, 1998). Furthermore, there is no consensus among theorists on the mechanism(s) behind syntactic priming (Pickering & Branigan, 1998; Chang, Dell, & Bock, 2006). Working memory capacity has been suggested as the system that might be responsible for main-taining the syntactic representation of the first sentence which facilitates the processing of a sentence that immediately follows it (Ledoux et al., 2007). With this in mind, the present study investigated (1) whether syntactic priming facilitates the comprehension of Dutch passive sentences and (2) whether additional effects of verb overlap can be observed. Lastly, we explored to (3) what extent this phenomenon relies on working memory.

In an event-related potential (ERP) experiment, subjects read passive sentences that were either preceded by an active or a passive sentence. A reduced P600 was expected to be elicited by a facilitation in syntactic processing in the passive target sentence compared to the passive prime sentence; a reduced N400 was expected to be elicited by verb repetition between the passive target and passive prime sentence. Against expectations, we did not observe syntactic priming effects in the comprehension of Dutch passive sentences (i.e., no reduction in the P600 amplitude for the passive target sentence). Moreover, no enhanced priming effects due to verb overlap were found (i.e., no reduction in the N400 amplitude for the repeated verb in the passive target sentence). Next to the ERP experiment, participants performed a non-linguistic task which tested general working memory capacity. We did not find a relation between the proposed dependency of syntactic priming on working memory. Though we are aware of the study’s limitations, our experimental study and results offer promising insights for future work on syntactic priming in sentence comprehension.

Key words: syntactic priming, comprehension, event-related potentials, passives, working memory

Now that I have finished this study, it only remains for me to thank everyone without whom this work would not have been possible.

Simone, thank you for the help you have offered me throughout this whole process. I will especially remember our first session when I came into your office because I did not understand the study’s design. After leaving your office, everything was crystal clear! Thank you for showing me how important it is to be a 100% thorough and that I should constantly remind myself what I want to study and most importantly why.

Rob, though you were in a different country, thank you for providing me with your expert advice on the topic. I would especially like to thank you for preventing me from conducting my initial research plan. I was way too ambitious and if it were not for you, I would probably still be testing participants right now (reminder: I wanted to also include 30 elderly participants...). Syntactic priming has really intrigued me and I am happy that I can continue working on this fascinating topic with you and Sarah during my PhD!

Amélie, I think I can call you my "adopted" third supervisor. When I first started this project, I might have underestimated how time-consuming and challenging EEG research is. I really admire the (difficult) work that you do and your optimism! I am eternally grateful for all the help you have offered me throughout this entire project. I could always ask you any question, from the practical side of doing EEG research to the analysis. Thank you. I hope we get to work in the future together, hopefully with an equal number of questions from both sides ;-).

Merel, thank you for introducing me to the world of syntactic priming! I am happy that I approached you about possible thesis topics, and syntactic priming directly caught my attention when we first started discussing it. Thank you for always responding to my questions that I had via e-mail. I am also looking forward to working with you in a

I would also like to thank all 28 participants who were so kind enough to volunteer and sit through 2.5 hours (or more) of testing. Thank you all for the nice conversations, your enthusiasm and interest in my project. You guys really made testing fun!

I would like to reserve the last words for my friends and family:

Mama en papa, dank jullie wel voor de steun! Jullie zijn altijd mijn nummer één sup-porters ook al was het soms wat moeilijk voor jullie om te begrijpen waar ik precies mee bezig was ;-). Ndagukunda cyane!

Also a big thank you to my closest friends, the F.A.B.S.. Thanks girls for the many needed fun meet-ups we had during these few months.

Artur_ª, the last thank you is for you. Thank you for your constant support throughout this entire process. Thank your for your help with coding, teaching me so much. Thank you for always wanting to listen and think with me for solutions.

I

declare that the work in this dissertation was carried out in accor-dance with the requirements of the University’s Regulations and Code of Practice for Research Degree Programmes and that it has not been submitted for any other academic award. Except where indicated by specific reference in the text, the work is the candidate’s own work. Work done in collaboration with, or with the assistance of, others, is indicated as such. Any views expressed in the dissertation are those of the author.

Page

List of Tables vii

List of Figures viii

1 Introduction 1

2 Background 5

2.1 Sentence comprehension: an overview . . . 5

2.2 Investigating syntactic representations through syntactic priming . . . 6

2.3 The residual activation model . . . 9

2.4 The implicit learning account . . . 10

2.5 Working memory and comprehension . . . 12

2.6 Implications for syntactic priming and its dependency on working memory 13 2.7 Using the ERP methodology to study syntactic priming in comprehension 15 2.8 The N400 . . . 15

2.9 The P600 . . . 17

2.10 ERP studies regarding syntactic priming in comprehension . . . 18

2.11 The proposed stimuli for this study: passive sentences in Dutch . . . 21

3 Research Questions and Hypotheses 24 4 Method 27 4.1 EEG experiment . . . 27

4.1.1 Participants . . . 27

4.1.2 Materials and design . . . 28

4.1.3 Procedure . . . 30

4.2 Operation digit span task . . . 31

4.4 ERP analysis . . . 33

5 Results 36 5.1 First time window: 200-300 milliseconds . . . 36

5.2 Second time window: 300-400 milliseconds . . . 39

5.3 Third time window: 400–500 milliseconds . . . 40

5.4 Fourth time window: 500–700 milliseconds . . . 40

5.5 Fifth time window: 700–900 milliseconds . . . 43

5.6 Sixth time-window: 900–1,000 milliseconds . . . 43

5.7 Individual differences . . . 44

6 Discussion 45 6.1 The N400 . . . 45

6.2 The P600 . . . 48

6.3 Working memory and syntactic priming . . . 50

6.4 Limitations and future work . . . 50

7 Conclusion 52 References 53 Appendix A 59 Appendix B 62 Appendix C 65 Appendix D 68 Appendix E 71 Appendix F 74 Appendix G 77

TABLE Page

4.1 Characteristics of the Participants . . . 27 4.2 Example of Four Versions (i.e., Combinations of Prime and Target) of an

Experimental Item . . . 28 5.1 Participants’ Score on the Operation Span Task . . . 44

FIGURE Page

2.1 Residual Activation Model as proposed by Pickering & Branigan, 1998 . . . . 9 4.1 The operation digit span task, constructed by La Roi, n.d. . . 32 5.1 The significant interaction between Anteriority and Condition in the first

time window (200-300 milliseconds). Note that the figure only visualizes the anterior electrodes as there was no such significant interaction in the posterior electrodes. The smaller plot on the top-right is a zoomed-in visualization of the specific time-window. The dotted lines represent the boundaries of the time window and the colored lines represent the different conditions. Negative voltages are plotted upwards. . . 37 5.2 The significant interaction between Anteriority, Condition and Hemisphere in

the 200-300 milliseconds time window. The dotted lines represent the time window boundaries and the colored lines the different conditions. Negative voltages are plotted upwards. . . 38 5.3 The differences in mean voltages in anterior and posterior electrodes in the

time window 500 – 700 milliseconds. Note that Condition did not prove to be a significant interaction term for these differences. The dotted lines indicate the time window boundary. Negative voltages are plotted upwards. . . 41 5.4 The significant differences in mean voltages in the left and right hemisphere

in the time window 500 – 700 milliseconds. Note that Condition did not prove to be a significant interaction term for these differences. The dotted lines indicate the time window boundary. Negative voltages are plotted upwards. . 42

C

H A P

1

I

NTRODUCTION

H

uman behavior is often repetitive (Pickering & Ferreira, 2008). For instance, infants tend to repeat actions of oneself or others. In language, repetition can be observed in the production of a syntactic structure after having been exposed to a similar structure. Previous exposure to a syntactic structure has been claimed to facilitate the following production, and this facilitation has been investigated extensively in syntactic priming studies. Importantly, this study will particularly focus on syntactic priming in comprehension.

One of the first ground breaking studies on syntactic priming in production was conducted by Bock (1986), who reported that speakers were more likely to describe a target picture with an active sentence (e.g., Lightning is striking the church) after they had been exposed to this structure in a preceding trial (e.g., One of the fans hit the referee). In addition, speakers showed a similar tendency with passive sentences (e.g., The church is being struck by the lightning) following (e.g., The referee was punched by one of the fans). The same effect was observed for datives with either a prepositional-object (PO; e.g., A rock star sold some cocaine to an undercover agent) or a double-object (DO; e.g., A rock star sold an undercover agent some cocaine). Since then, syntactic priming in production has been observed for different structures, including noun-phrase structures (Cleland & Pickering, 2003) and particle placement (Konopka & Bock, 2005; see also Griffin & Weinstein-Tull, 2003; Ferreira, 2003 and Branigan, Pickering, McLean, & Stewart, 2006 for other structures). Moreover, apart from English, syntactic priming has also been observed in other languages, such as Dutch (Hartsuiker & Kolk, 1998b)

and German (e.g., Scheepers, 2003; see Mahowald, James, Futrell, & Gibson, 2016 for a meta-analytic review paper). Furthermore, syntactic priming in production has been shown in bilinguals. Also, it has been reported to be present in the speech of young children and aphasic speakers (see Pickering & Ferreira, 2008 for an extended review on syntactic priming in special populations). Taken together, the tendency to repeat syntactic structures employed by either oneself or others, and the resulting facilitation is a phenomenon that manifests itself in different settings, languages and syntactic structures.

Throughout the years, two major theories have been proposed in an effort to explain syntactic priming — namely the residual activation model and the implicit learning account. The residual activation model (Pickering & Branigan, 1998) predicts that syntactic priming is short lived — that is, syntactic priming effects decay when the prime and target sentence are separated by other sentences. In contrast, the implicit learning account proposes that syntactic priming persists over several separated sentences (Chang et al., 2006). The main difference between these two accounts is that the first model explains the lexical boost (i.e., stronger priming effects due to verb overlap between the prime and target sentence), whereas the second model only accounts for abstract syntactic priming (i.e., when there is no verb overlap between the prime and target sentence). The repeated verb that is presented in the target sentence is a cue for the speaker to retrieve the memory of the prime structure more easily which results in the lexical boost (Bernolet, Collina, & Hartsuiker, 2016). These enhanced priming effects due to verb overlap are claimed to be short-lived since they have been found to decay when the prime and target sentence are separated by other sentences (Hartsuiker, Bernolet, Schoonbaert, Speybroeck, & Vanderelst, 2008). In contrast, the implicit learning account cannot account for the short-lived lexical boost. Instead, it proposes that the lexical boost specifically relies on the explicit memory of the lexical head (Chang et al., 2006). However — crucially — recent evidence from Bernolet et al. (2016) has nuanced these latter findings, demonstrating that, even when the prime and target sentence do not share the same verb, the explicit memory of the prime structure boosts syntactic priming when the prime and target sentence immediately follow each other. This entails that stronger priming effects are not exclusive to the repetition of the lexical verb but rather that any similarity (e.g., the syntactic structure) between the prime and target sentence may boost syntactic priming.

Apart from the theoretical mechanisms that have been suggested to account for syntactic priming, this phenomenon has been proposed to depend on working memory.

So far, only a few researchers have investigated individual differences such as working memory and its role in syntactic priming. As suggested by McDonough and Kim (2016), working memory is an important variable to consider in the debate of syntactic priming, especially in conditions where there are no other sentences separating the prime and target sentence, as it is alleged to be involved in activating and retrieving the explicit memory of the prime structure. In their study, they concluded that working memory supposedly contributes to the retrieval and short maintenance of the prime structure (see Chapter 2 for an extensive discussion of the study of McDonough & Kim, 2016).

As syntactic priming has successfully been demonstrated in production, with working memory allegedly mediating in conditions where the prime and target sentence immedi-ately follow each other, similar effects should be observed in comprehension. However, using online measures (e.g., eye-tracking, Arai, Van Gompel, & Scheepers, 2007 and event-related potentials (ERPs), Ledoux et al., 2007), researchers have only recently begun to investigate syntactic priming in comprehension in order to understand whether production and comprehension share a common processing mechanism. Moreover, in investigating syntactic priming in comprehension, researchers have begun to examine whether syntactic representations are built in a similar way in both modalities (e.g., Pickering & Traxler, 2004). One of the main differences between studies that have investigated syntactic priming in production vs. comprehension is that priming effects in comprehension are typically observed when the prime and target sentence share the same verb (e.g., Tooley, Traxler, & Swaab, 2009), whereas verb overlap is not necessary for observing priming effects in production. Interestingly, there are some studies that found syntactic priming in comprehension without verb overlap (e.g., Traxler, 2008).

Another difference between syntactic priming in the two modalities is that the sentences used in comprehension studies are typically more complex (e.g., reduced relative clauses: The defendant examined by the lawyer was guilty, in Ledoux et al., 2007) than the ones used in production studies (e.g., datives (DO/PO) in Hartsuiker et al., 2008). Complex sentences such as reduced-relative clauses are usually used in comprehension studies investigating syntactic priming because they require more processing costs (see e.g., Ledoux et al., 2007). Consequently, syntactic priming can be used as a tool for investigating whether previous exposure to a particular sentence decreases these costs (Pickering & Ferreira, 2008). However, the use of different structures to examine priming in comprehension and production makes it difficult to compare priming effects across modalities. A straightforward solution to this would be to test syntactic priming in comprehension with similar structures that have been used in production studies:

passives, for example (see Arai et al., 2007 for a similar reasoning).

Taking all the above into consideration, the current ERP-experiment aims to in-vestigate whether syntactic priming can be observed in the comprehension of passive sentences. Moreover, we also explore the extent to which this phenomenon relies on working memory. Passive sentences were chosen mainly due to their widespread use in studies concerning syntactic priming in production (e.g., Hartsuiker & Kolk, 1998b). Moreover, passives are less frequent than active sentences and therefore may benefit more from priming (Hartsuiker & Kolk, 1998b). Lastly, the comprehension of passives compared to active sentences has been alleged to be related to higher processing costs for working memory (see e.g., Colman, Koerts, Stowe, Leenders, & Bastiaanse, 2011). Thus, we investigate whether these processing cost are strongly diminished when participants are primed with the same syntactic structure.

ERPs were used to investigate syntactic priming in comprehension of passives be-cause it is a method that is highly sensitive to changes in the time-course of comprehen-sion (Luck, 2014). With regard to syntactic priming, ERPs are particularly helpful as they make it possible to study when the cognitive system recognizes a syntactic rela-tionship between two sentences that may result in priming effects (Branigan, Pickering, Liversedge, Stewart, & Urbach, 1995).

The following section provides a theoretical background of syntactic priming in production and comprehension and what the main difference is between both modalities. This is followed by a discussion of the different mechanisms that have been proposed to account for syntactic priming. The third section discusses implications for working memory and syntactic priming. Subsequently, relevant ERP components for language processing will be discussed which will then be related to studies that have employed this technique towards investigating syntactic priming in comprehension. The last section elaborates on why passive sentences were used to investigate syntactic priming in comprehension in the present study. Finally, the research questions and hypotheses are formulated and elaborated based on theory and previous findings.

C

H A P

2

B

ACKGROUND

2.1

Sentence comprehension: an overview

Sentence comprehension has been fascinating psycholinguists for decades, as an under-standing of this process may provide us insights into the cognitive mechanisms that enable comprehension (Gordon & Lowder, 2012). The retrieval of the meaning of words (i.e., the lexical analysis) and syntactic analysis (i.e., parsing the words into a structure) have frequently been noted as important steps in understanding how comprehension operates (see e.g., Traxler & Tooley, 2008; Giavazzi et al., 2018). Some researchers have stressed the importance of syntax in enabling us to understand what a sentence means (see e.g., Fodor & Garrett, 1967). According to Fodor and Garrett (1967), sentences are syntactically parsed and the syntactic category of each word determines how they relate to each other. This results in the meaning of a sentence. For instance, in a passive sentence, the object is first encountered and is then followed by the lexical head. In such cases, the agent is the last noun in the sentence (e.g., Hanna was bitten by the dog). These separate words are related to each other such that at the end, one can infer the meaning of the sentence by determining ‘who did what to whom’.

Different accounts for parsing have been proposed and how syntactic representations contribute to the comprehension of a sentence. For instance, the lexicalist account of syntax puts emphasis on the importance of the information linked to words that are essential in gaining access to the structural representation of a sentence (see e.g., Ford, Bresnan, & Kaplan, 1982; MacDonald, Pearlmutter, & Seidenberg, 1994). The claim is

that the lexicon and syntax are tightly linked and the information that is necessary to construct a syntactic representation is stored in lexical items. The lexical representation includes information such as semantics, phonology, orthography and the argument structure. MacDonald et al. (1994) illustrated this with the word watch which can have a ‘noun meaning’ (something that denotes the time) or a verb meaning (to observe someone). Moreover, they highlighted the importance of lexical verbs in creating syntactic ambiguity, for example, Peter watched Hanna or Hanna watched intently. Here, watch is transitive in the first sentence and intransitive in the second. These examples demonstrate that the verb plays an important role in comprehension due to its syntactic relations to the other words in the sentence. Taken together, the lexicalist account proposes that syntactic representations are closely connected to the information in lexical items.

In contrast, the autonomous syntax account proposes that syntactic representations are constructed without any reference to specific words, semantic properties, or other non-grammatical information (see e.g., Frazier, 1979, 1987). This entails that syntactic representations are built on “generic members of a category such as verbs, nouns, prepositions etc.” (Traxler & Tooley, 2008, p. 610) without the specific identify of words (e.g., such as in the previous example with the verb watch). For instance, the input of the parser could be determiner – noun – verb – preposition – determiner – noun, and consequently words are recognized as belonging to one of the general word categories.

Taken together, the lexicalist account of syntax focuses on the importance individual words which contributes to the mental representation of syntactic structures and the comprehension of a sentence. In contrast, the autonomous syntax account suggests that knowledge of a more generalized syntactic structure of a language aids in constructing syntactic representations.

2.2

Investigating syntactic representations through

syntactic priming

The different accounts described above have provided possible explanations for parsing and comprehension. However, they differ with respect to the role of lexical items and the role of general word categories. Syntactic priming has proven to be a method to investigate the different accounts of parsing (Pickering & Ferreira, 2008). Moreover, it has also been employed as a tool for examining how people construct mental representations of syntax in production and comprehension (Branigan & Pickering, 2017). Syntactic priming occurs when the act of processing a stimulus affects the processing of another

stimulus with the same (or related) syntactic structure (Branigan et al., 1995). Processing the target sentence is facilitates because the prime sentence influences the activation of its syntax and this activation persists for a short time making it less demanding to comprehend a sentence with the same structure.

Most studies on syntactic priming have been conducted on language production. As mentioned before, Bock (1986) concluded that participants were more likely to produce a passive sentence after they were exposed to a preceding passive sentence. Using prime and target sentences without verb overlap, Bock (1986) concluded that the observed priming effects could only be accounted for by a shared syntactic structure. This finding was groundbreaking as it observed that the representation of a syntactic structure may not necessarily be dependent on the lexical head.

However, more recent studies reported stronger priming effects in production when the prime and target sentence share the same verb as well as the same syntactic structure (lexical boost, Pickering & Branigan, 1998). The authors used a written completion task where participants were presented with incomplete sentences with a verb that could take a preposition object (PO) or a double object (DO) structure. For instance, the incomplete prime sentence The racing driver showed . . . could ei-ther be completed with a PO structure which can result in The racing driver showed the torn overall to the team manager (p. 637) or with a DO structure: The racing driver showed the helpful mechanic the damaged tire. After reading and completing the prime sentence, participants read a target sentence, which they were also asked to complete (e.g., The patient showed. . . ). If participants had previously completed the prime sentence with, for instance, a PO structure, they were also more likely to complete the target sen-tence with the same structure. Importantly, even though priming effects were observed in the condition where the prime and target did not share the same verb, priming effects were stronger when the prime and target sentence did share the same verb. This finding provided counter evidence for the previous claim that syntactic representations do not rely on specific words. Altogether, syntactic priming in production occurs when the prime and target sentence do not share the same verb. However, these effects are stronger when there is lexical similarity between the prime and target sentence.

Despite an abundance of studies on syntactic priming in production, few studies have focused on this phenomenon in comprehension. However, the phenomenon has gained interest (e.g., Pickering & Traxler, 2004; Arai et al., 2007; Ledoux et al., 2007; Traxler & Tooley, 2008; Weber & Indefrey, 2009). These studies show that unlike syntactic priming in production, which typically does not require lexical overlap, priming in comprehension

is often restricted to prime-target pairs that share the same verb. Priming effects due to verb overlap between the prime and target sentence has been found for reduced relative clauses (e.g., The defendant examined by the lawyer; Ledoux et al., 2007; Tooley et al., 2009), PO and DO structures (e.g., Arai et al., 2007; Carminati, van Gompel, Scheepers, & Arai, 2008), and high-low attachment ambiguities (e.g., The girl hit the boy with the paddle earlier today, Boudewyn, Zirnstein, Swaab, & Traxler, 2014). The alleged lexical dependency of syntactic priming in comprehension implies that syntactic representations in comprehension may be more sensitive to the content of specific words (Tooley & Traxler, 2010). The authors argued that when words are encountered during reading, their order constrains the possible choices for syntactic structure. This is especially true for verbs, as they strongly constrain syntactic and thematic processing. For this reason, Tooley and Traxler conclude that it is not surprising that syntactic priming in comprehension is more often observed when the prime and target sentence share the same verb. Moreover, Giavazzi et al. (2018), claimed that comprehension relies more on lexical overlap. This is because it can be successful without complete syntactic parsing, as “the word order and the sole analysis of the lexical content are often sufficient to derive meaning from the sentence” (p. 2).

However, there are studies that have reported syntactic priming in comprehension without verb overlap. For instance, using eye-tracking, Traxler (2008) observed syntactic priming in sentences with modifier-goal ambiguities (e.g., The vendor tossed the peanuts in the box into the crowd during the game) even though the prime and target sentence did not share the same verb. Moreover, Thothathiri and Snedeker (2008) demonstrated syntactic priming in comprehension in prepositional object (e.g., Give the ball to the lion) and double-object (e.g., Give the lion the ball) datives when the prime and target sentence did not share the same verb. Importantly, the prime and target sentence in Traxler (2008) and Thothathiri and Snedeker (2008) were not separated by other sentences.

Taken together, there is inconsistency in the results of aforementioned studies: while lexical repetition seems to be crucial in observing syntactic priming in comprehension in some studies, others have reported syntactic priming in comprehension when the prime and target sentence do not share a similar verb. This implies that there may be other factors other than verb overlap that contribute to priming effects in comprehension. For instance, it could be the case that these effects can also be boosted by the explicit memory of the prime syntactic structure (or any other similarity between the prime and target sentence) if the prime and target sentence immediately follow each other and share no verb overlap (Bernolet et al., 2016). The following section discusses mechanistic

accounts that have been proposed to explain syntactic priming. One of the discussed mechanisms focuses on the explicit memory of the prime structure which may explain syntactic priming without verb overlap when the prime and target sentence immediately follow each other.

2.3

The residual activation model

In the past two decades, different mechanisms have been proposed to explain syntactic priming. Pickering and Branigan (1998) suggested that the residual activation model could account for syntactic priming. Although this model was based on production processes, some researchers (e.g., Traxler & Tooley, 2012) have suggested that this mechanistic account of syntactic priming might apply to comprehension too. In their model, Pickering and Branigan (1998) claimed that syntactic priming is caused by short-term activation. They argued that activation of a specific verb and its grammatical features (e.g., the verb “give”) is combined with the activation of the combinatorial nodes — (i.e., possible structures in which a verb can occur. For example, “give” has two possibilities: The woman gives the girl a book, which is a noun phrase-noun phrase structure and The boy gives a bone to the dog, which is a noun phrase-prepositional phrase structure) (see Figure 2.1).

Figure 2.1. Residual Activation Model as proposed by Pickering & Branigan, 1998 Activation in the combinatorial nodes is maintained for a short time. When a subse-quent verb is presented that can take one of the structures in the combinatorial nodes,

this syntactic structure is facilitated (Tooley & Traxler, 2010). Moreover, the model predicts that there is also residual activation between the link of the verbs and the activated combinatorial node (Pickering & Branigan, 1998). This entails that when a target sentence repeats the same verb as in its prime sentence, there is a summation of the residual activation at the combinatorial nodes and at the link between the verb and the combinatorial node. This process diminishes processing efforts even more (Pickering & Branigan, 1998; Tooley & Traxler, 2010). This model provides an explanation not only for priming that occurs due to a similar syntactic structure, but also for larger priming effects where there is a repetition of the verb in the prime and target sentence.

As mentioned by Tooley and Traxler (2010), the model of Pickering and Branigan (1998) implies that similar short-lived effects should not be observed when the prime and target sentence are separated by other sentences since the residual activation of the prime structure is only shortly activated. Yet, Bock and Griffin (2000) demonstrated that syntactic priming persists across pairs of prime and target sentences even when the pairs are separated by filler sentences. Bock and Griffin (2000) reported long-lived syntactic priming effects when the prime and target sentence did not share a similar verb and short-lived priming effects due to the explicit memory of the repeated verb. In addition, Hartsuiker et al. (2008) observed similar long-lived abstract syntactic priming effects and short-lived priming effects when the prime and target sentence shared a similar verb. This entails that the lexical boost decays, whereas abstract syntactic priming persists across sentences. According to Tooley and Traxler (2010), the findings of Bock and Griffin (2000) and Hartsuiker et al. (2008) are two of the many studies that have illustrated that syntactic priming cannot be fully explained on the basis of the residual activation model (see e.g., Chang, Dell, Bock, & Griffin, 2000; Kaschak, Kutta, & Jones, 2011, for similar conclusions). That is, since the residual activation model does not explain the time course of syntactic priming (i.e., long-lived vs. short-lived), it is not a complete model of the phenomenon. The following section discusses an alternative theory for syntactic priming, which can better explain long-lived syntactic priming.

2.4

The implicit learning account

The implicit learning account was proposed as an explanation for syntactic priming in production (Bock & Griffin, 2000; Chang et al., 2006) and comprehension (Chang et al., 2006). Implicit learning refers to subconsciously acquiring abstract information such as syntactic structures over a longer period of time (Tooley & Traxler, 2010). Chang et al.

(2006) illustrate their perspective with evidence from a connectionist model that utilizes error-based learning to establish syntactic regularities. More specifically, their model uses the difference between the predicted and actual output (i.e, the predicted output is based on the input), and then adjusts the connections that are responsible for the initial prediction. In language, this means that people have predictions about how a syntactic structure should be constructed and, over time, they adapt their predictions according to the correct output based on surrounding speech (Chang et al., 2006). Without being conscious of the process, people learn to modify their predictions and the connections between syntactic categories that form a correct syntactic structure are strengthened. As a result, syntactic regularities are formed through learning. With this in mind, Chang et al. (2006) suggested that syntactic priming is a form of “error-based implicit learning” (p. 245). This entails that, for example, when one is repeatedly exposed to the same syntactic structure, one is less likely to make prediction errors about the syntactic representation of this structure. For instance, repeated exposure to passive sentences will result in learning that the first noun that is encountered is the grammatical subject but not the agent of the sentence. The prediction error (i.e., assuming that the first noun in a sentence is always related to the agent) is minimized due to previous exposure.

Although the model of Chang et al. (2006) supports findings on long-lived abstract syntactic priming, it nonetheless cannot account for the lexical boost. Similar to Bock and Griffin (2000), Chang et al. (2006) propose that larger priming effects related to verb overlap between prime and target sentence are due to the explicit memory of the verb. This view is supported by previous findings from Konopka and Bock (2005), who investigated syntactic priming in production for verb-particle structures (e.g., lace the boot up vs. lace up the boot). Prime sentences had either similar or different verbs than the target sentences and they were separated by 0, 1, 2 or 3 sentences. When there was a lag of three sentences, the lexical boost disappeared but abstract syntactic priming remained. Taken together, the model of Chang et al. (2006) accounts for abstract syntactic priming as a form of implicit learning. However, enhancement of syntactic priming due to verb overlap is suggested to be due to the explicit memory of the lexical head.

Interestingly, more recent findings have demonstrated that abstract syntactic priming also involves some explicit memory when the prime and target immediately follow each other. Bernolet et al. (2016) employed transitives (active and passives), datives (double object and prepositional object datives) and the word order in the verb phrase of relative clauses (e.g., De chauffeur zei dat de weg was geblokkeerd; the driver said that the road was blocked vs. De chauffeur zei dat de weg geblokkeerd was; the driver said that the

road blocked was) to investigate whether abstract syntactic priming may also rely on the explicit memory of prime sentences. To examine this, participants were instructed to read prime sentences and then provide descriptions (target sentences) in a picture matching task. Importantly, the prime and target sentences did not share the same verb. The prime and related target sentence either directly followed each other or they were separated by two or six filler sentences. Moreover, participants were explicitly asked to remember the syntax of the prime sentence and to use it to formulate the target sentence. For the three syntactic structures, the researchers reported stronger priming effects when the prime and target immediately followed each other, and a decay when the filler sentences increased between the pair.

Bernolet et al. (2016) argued that even in abstract syntactic priming, any similarity (e.g., the syntactic structure) may be a cue to retrieve the explicit memory of the prime structure when the prime and target sentence are not separated by filler sentences. This finding has important implications, because it proposes that abstract syntactic priming stems from different sources: long-lived abstract priming is related to implicit learning whereas short-lived abstract priming is associated with the explicit memory of the prime structure. Even though the study of Bernolet et al. (2016) focused on production, their findings may provide an explanation for findings from Traxler (2008) and Thothathiri and Snedeker (2008), who observed syntactic priming in comprehension without verb overlap. One could claim that the explicit memory of the prime structure boosted syntactic priming since there was no verb overlap between the prime and target sentence that immediately followed each other.

If syntactic priming partially underpins the explicit memory of the prime structure and lexical head in successive conditions, there should be a cognitive system that is involved in supporting the retrieval of the explicit memory of the prime sentence or the lexical head. The following section discusses working memory and its role in sentence comprehension. Subsequently, the alleged relation between working memory and syntac-tic priming will be discussed since syntacsyntac-tic priming has been suggested to rely on this cognitive system (see e.g., McDonough & Kim, 2016).

2.5

Working memory and comprehension

In 1980, Daneman and Carpenter suggested that working memory is an important individual variable in language comprehension (see also e.g., Carriedo, Elosúa, & García-Madruga, 2011). Working memory is defined as a system that executes higher cognitive

processes (e.g., reading), while storing the products of these processes (Baddeley & Hitch, 1974; Daneman & Carpenter, 1980). For example, during reading, people are exposed to syntactic, lexical, semantic, pragmatic and phonological information. Subsequently, they have to shortly store this information while also using it to adequately parse and process the input to arrive at an understanding of a sentence. Daneman and Carpenter (1980) proposed that working memory capacity is an important factor for successful comprehension. The claim is that individuals with less working memory capacity are less likely to maintain the shortly-activated information in memory while also computing a meaningful representation of a sentence (Daneman & Merikle, 1996). This hypothesis was previously tested in the study of Daneman and Carpenter (1980), where partici-pants listened to or read a number of unrelated sentences and were then asked to recall the final word of each sentence. As the number of sentences increased, participants’ working memory was measured by the number of words they successfully recalled. The researchers correlated the participants’ reading span score to their language comprehen-sion abilities, which were assessed with the Verbal Scholastic Aptitude test (VSAT; Lord, 1968). Ultimately, Daneman and Carpenter (1980) concluded that less working memory capacity entails less language comprehension abilities.

Taken together, working memory is an important cognitive system that is involved in higher processes such as comprehension. One could imagine that the processing costs for working memory can be diminished if similar information is accessed and shortly maintained due to previous exposure. In the following section, working memory is discussed in relation to syntactic priming.

2.6

Implications for syntactic priming and its

dependency on working memory

As studies on syntactic priming in comprehension have suggested that priming could facilitate sentence comprehension (e.g., resulting in faster reading times for the target sentence), one could assume that priming might therefore reduce the load for working memory. At the same time, syntactic priming may rely on working memory to shortly retain syntactic representations which aid in processing the subsequent sentence.

Until now, the relation between working memory and syntactic priming in compre-hension (and production) has rarely been investigated. Several researchers have alluded to the relation between working memory and syntactic priming. For instance, Chipere (2003) proposed that syntactic priming might occur because the syntactic parser uses

representations that are already available in working memory and could therefore be reused to process a similar subsequent sentence. In addition, Fink, Schwartz, and Myers (1998) suggested that syntactic priming could reduce the mental resources needed to retrieve syntactic representations in aphasic speakers as they have been proposed to have reduced working memory and slowed access to syntactic structures (see also Cas-pari, Parkinson, LaPointe, & Katz, 1998 and Hartsuiker & Kolk, 1998a). More recently, Yan, Martin, and Slevc (2018) argued that if syntactic priming has an explicit memory component which accounts for priming effects in conditions where the prime and target sentence immediately follow each other, then a short-term memory system should sup-port the short-term retention of the syntactic information in the prime sentence (e.g., such as the explicit memory of the prime structure).

One of the very few studies regarding the possible relation between working memory capacity and syntactic priming was conducted by McDonough and Kim (2016). More specifically, they investigated the relationship between primed production and working memory. In their study, participants were instructed to listen to sentences and then describe pictures. The participants were told to describe the pictures with a transitive verb in either a passive or active sentence structure. Working memory was tested with the running memory span task (Broadway & Engle, 2010). In this task, participants were exposed to a series of an unpredictable number letters and were instructed to recall letters that were presented towards the end of the series. Their results revealed a positive correlation between a higher working memory and the subsequent production of a passive sentence, but only when the prime and target sentence directly followed each other. That is, participants with a higher working memory score, significantly produced more passives than individuals with a lower working memory score. The researchers concluded that individual differences such as working memory are crucial in the debate of syntactic priming as it presumably contributes to the retrieval and short maintenance of the prime structure.

Apart from the main goal of the present study (i.e., investigating syntactic priming in comprehension using ERPs), we therefore also aim to explore individual differences, such as working memory and its alleged relation to syntactic priming. The following section elaborates on why ERPs can be used to investigate syntactic priming in comprehension. This is followed by a discussion of the two main ERP components relevant to language processing.

2.7

Using the ERP methodology to study syntactic

priming in comprehension

Language comprehension is a rapid process and event-related potentials (ERPs) have been found to be a solution to investigate comprehension due to its fine-grained temporal resolution (Luck, 2014). Event-related potentials are measured brain responses “to discrete stimuli associated with the recognition of familiar or novel objects” (Bell & Cuevas, 2012, p.2). The ERP methodology has several advantages that make it a suitable technique to investigate the successive steps that are involved in language processing. For instance, (1) participants are not required to perform any other task other than reading the words that appear on the screen, (2) data are collected continuously and (3) exact time-measurement of when the stimuli is processed are associated with specific cognitive processes (Fonteneau, Frauenfelder, & Rizzi, 1998; Bell & Cuevas, 2012).

In terms of syntactic priming, ERPs can be employed since syntactic priming taps into the mental representation of syntax. With the use of ERPs, it becomes possible to study when the cognitive system recognizes a relationship between two mental representations of a particular syntactic structure and whether processing is then facilitated due to this recognition (Branigan et al., 1995). Moreover, since syntactic priming in comprehension is typically reported when the prime and target share the same verb, ERPs can shed light on whether this lexical repetition indeed reduces the processing of a subsequent syntactic structure. By time-locking the EEG signal to the repeated verb in the prime and target sentence, it becomes possible to extract ERPs from the continuous EEG signal (see e.g., Ledoux et al., 2007). Moreover, ERPs might reveal whether effortful syntactic processing is diminished (i.e., a reduced P600) when one is repeatedly exposed to similar syntactic structures (for example, in passives).

The following two sections elaborate on the two main language components relevant to language processing, the N400 and the P600, in order to make clear why these two components are also relevant for examining syntactic priming in comprehension.

2.8

The N400

The N400 was first described by Kutas and Hillyard (1980). In their study, they discovered that words that violate semantic expectancy show different ERPs in comparison to words that semantically adhere to their context. In other words, the N400 has been frequently reported to be due to difficulty in semantic integration (e.g., John eats his soup with a

sock – here ‘a spoon’ is expected). The difference in ERPs is observable as a negative peak around 400 milliseconds post-stimulus. Since the study of Kutas and Hillyard (1980), many other studies have been conducted reporting on the N400 in language processing as well as non-linguistic experiments, such as gesture processing, mathematical cognition and recognition memory among others (see Kutas & Federmeier, 2011, for an extensive review).

Several findings of the N400 are relevant to syntactic priming in comprehension. For instance, Besson, Kutas, and Petten (1992) reported that repetition of words reduces and shortens the duration of the N400. In the same year, Paller and Kutas (1992) demonstrated that words that had been presented previously were processed more easily, which was reflected in an N400 effect. According to Van Petten and Senkfor (1996), the reduced amplitude for repeated words may also be related to the claim that the N400 is sensitive to semantic context. For this reason, repetition of a particular word may reflect a strong semantic relationship. For example, if readers have been primed with Dutch passive sentences (e.g., Het hondje met de kwispelende staart wordt geslagen door de boer – The little dog who is wagging his tail is being beaten by the farmer), they may be more prone to pre-activate semantic (as well as syntactic) features of the auxiliary verb wordt. Once they encounter this word, they might infer that they have not yet encountered the agent of the sentence because this is revealed at the end of the sentence.

Related to this reasoning is the Retrieval-Integration account proposed by Brouwer, Fitz, and Hoeks (2012). In this account, Brouwer et al. (2012) reconsidered the ‘conven-tional’ understanding of what the N400 and the P600 reveal about language processing. For example, they suggested that the N400 does not reflect semantic integration (see Ku-tas & Federmeier, 2011) but it is related to the memory retrieval of all information (i.e., syntactic, semantic and pragmatic) linked to a word from long term memory (Brouwer et al., 2012). Focusing on syntactic priming, this entails that, for example, when one is primed with a passive structure, he or she recognizes the explicit form of the repeated word wordt. At the same time the lexical information of wordt is retrieved from long term memory. This implies that one may co-activate features of a passive syntactic structure because he or she has inferred that the repetition of wordt is a cue for a passive struc-tures according to previous observations. In this case, facilitated retrieval of syntactic, semantic and pragmatic information related to the repeated word is expected. Therefore, this might be reflected in the amplitude of the N400.

2.9

The P600

The P600 is another ERP component that is relevant to language processing. The P600 reflects a positive peak around 600 milliseconds post-stimulus. Initially, the P600 was considered to reflect syntactic reanalysis of complex sentences such as garden-paths (e.g., Osterhout, Holcomb, & Swinney, 1994) or ungrammatical sentences (e.g., Hagoort, Brown, & Groothusen, 1993). However, as mentioned beforehand, Brouwer et al. (2012) proposed a different explanation for the N400 and the P600 in their Retrieval-Integration account. Focusing on the P600, Brouwer et al. (2012) suggested that the P600 reveals processes that are related to integration rather than syntactic reanalysis. Syntactic reanalysis occurs when a sentence is rendered as ungrammatical due to e.g., number agreement (*Het verwende kind gooien - The spoiled child throw (Hagoort et al., 1993). In contrasts, Brouwer et al. (2012) argued that the P600 reflects the mental representation of what is being communicated (the MRC hypothesis). The MRC hypothesis entails that the P600 reflects processing efforts that are necessary for “the construction, revision, or updating of a mental representation of what is being communicated” (p. 137). Their view on the reconsideration of the P600 supports findings of, for example, Kaan, Harris, Gibson, and Holcomb (2000) who found P600 effects for long-distance wh-dependencies (e.g., Jane wonders who the dancers in the ballroom imitate. . . ) compared to sentences that did not contain such wh-dependencies (e.g., Jane wonders whether the dancers in the ballroom imitate.... Importantly, the long-distance wh-dependencies were well-formed so any effects of the P600 could not be attributed to syntactic reanalysis. Rather, readers were expected to infer the difficulty related to linking the verb imitate to the wh-pronoun and the following noun phrase whereas there is no such dependency in the other sentence. Brouwer et al. (2012) utilized the study of Kaan et al. (2000) to point out that this is an evident example of the P600 reflecting syntactic integration.

In terms of comprehending passive sentences, the Retrieval-Integration account supports the findings of Jackson, Lorimor, and van Hell (n.d.) who found a P600 effect for passive sentences in comparison to active sentences. Using grammatically correct passives sentences, Jackson et al. (n.d.) concluded that “passives sentences require revision of previous expectations and predictions rather than syntactic reanalysis per se” (no page number). This means that previous expectations of encountering the agent as the first noun in a sentence need to be revised and updated since the agent is not the grammatical subject in a passive sentence but the grammatical object.

P600 may be smaller if effortful syntactic processing is reduced because of previous exposure to a similar structure. In this case, effortful syntactic processing may consist of construction, revision and updating a mental representation of the sentence.

Altogether, based on the Retrieval-Integration account, the P600 may account for any revision efforts that may be observed for passive sentences. Moreover, a decline in the P600 could be observed due to continuous priming of passives structures. The next section elaborates on previous studies that have used ERPs to examine syntactic priming in comprehension.

2.10

ERP studies regarding syntactic priming in

comprehension

With the rising interest in syntactic priming in comprehension, some studies have employed the ERP methodology to examine this phenomenon. For instance, Ledoux et al. (2007) investigated syntactic priming in comprehension under EEG recording and used reduced-relative (RR) clauses as their stimuli (e.g., The defendant examined by the lawyer was unreliable) which are more difficult to comprehend than main clauses (MC; e.g., The defendant examined the evidence). The researchers’ assumption was that readers would initially interpret the defendant as the subject of the sentence in the RR condition (i.e., readers are led down a garden-path). However, readers had to revise their interpretation once they encountered the disambiguating phrase by the lawyer which forced the interpretation that the lawyer was the subject instead of the defendant.

In the study of Ledoux et al. (2007), all target sentences were RR clauses while the prime sentences could either be RR or MC sentences. In the prime sentences, they found a greater positivity at the post-verbal critical disambiguating noun in the RR clauses (e.g., The speaker proposed by the group. . . ) compared to the MC sentences (e.g., ’The speaker proposed the solution. . . ). The researchers interpreted this finding according to the conventional view of the P600 which indicates syntactic reanalysis of a garden-path sentence. Importantly, they also observed a processing difference which elicited a reduced P600. This difference was present at the disambiguating noun if the target (i.e., RR clause) was preceded by an RR prime compared to when it was preceded by an MC prime. Subsequently, this indicated that processing the target sentence was less effortful when the reader had been previously exposed to a sentence with the same structure.

Next to syntactic priming, Ledoux et al. (2007) investigated whether lexical repetition would elicit an additional facilitating effect in processing the difficult RR target sentences

(e.g., The manager proposed by the director. . . ). The researchers found a reduction of the N400 component in the second presentation of the repeated verb (i.e. proposed) in the target sentence when participants were previously exposed to the same verb in an MC prime. Unexpectedly, the same result was not observed when participants were first presented with the repeated verb in an RR prime followed by an RR target. Instead, the researchers observed a reduced positivity (P600) at the second presentation of the repeated verb in the target RR sentence. Ledoux et al. (2007) discussed this surprising finding stating that since RR sentences are more cognitively demanding to process, lexical overlap between the RR prime and RR target alone did not result in additional facilitating effects. They suggested that the syntactic properties of the repeated verb and its use in syntactic structures may have remained in memory, resulting in a facilitating effect. For this reason, their findings corroborate the lexicalist account of parsing, which states that syntactic information is linked to specific words.

In 2009, Tooley et al. also investigated syntactic priming in comprehension with ERPs in addition to eye-tracking. This section will only discuss their results concerning ERPs. In their study, they investigated to what extent lexical overlap plays a role in syntactic priming in comprehension. More specifically, they investigated whether repetition of identical verbs alone result in facilitating effects or whether these effects could also be observed in synonyms that have substantial semantic overlap and can occur in the same syntactic structure (e.g., repetition prime: The man watched by the woman. . . .; synonym prime: The man observed by the woman. . . .; target: The child watched by the parent. . . .). Unlike the experiment of Ledoux et al. (2007), the study of Tooley et al. (2009) consisted of primes with only RR structures. Half of these primes contained repeated verbs and the other half contained synonyms (see aforementioned example). The researchers proposed that if structural information is indeed only linked to specific verbs — as proposed by the lexicalist account — then facilitating effects should only be observed in the repeated verbs between the prime and target sentences. This effect was expected to be reflected as a reduced P600 at the word by. . . . However, if structural information is also related to semantically similar verbs that can occur in the same syntactic structure, then similar facilitating effects should also be noticed when the prime contains a synonym of the verb appearing in the target sentence (Tooley et al., 2009). The latter suggestion relates to the autonomous syntax account since this theory suggests that words are treated interchangeably and therefore syntactic priming may occur regardless of whether words are repeated across prime and target sentences or not. Moreover, the researchers also predicted an N400 effect only when the verb was repeated in the prime and target

sentence and not when the prime had a synonym of the verb appearing in the target sentence.

The results of Tooley et al. (2009) revealed a reduced P600 at the critical word by. . . only when the verb was repeated in the prime and target sentence, and no similar result was present in the synonym condition. This entails that, upon presentation of an RR target that was preceded by a synonym RR prime, readers still experienced difficulty in syntactic processing whereas this difficulty was diminished when they were presented with a repetition RR prime. In addition, the researchers observed a reduced N400 in the repetition prime condition while no such effect was observed for the synonym prime condition. Taken together, their findings support the lexicalist account since their results demonstrated that the activation of syntactic information is specifically tied to repeated verbs and not to synonyms. This strengthened claims that syntactic priming in comprehension is dependent on verb repetition, at least in syntactic ambiguous sentences such as reduced-relative clauses.

More recently, Chen, Xu, Tan, Zhang, and Zhong (2013) investigated syntactic priming in Chinese sentence comprehension by means of ERPs. Chen et al. (2013) examined a similar hypothesis as in Tooley et al. (2009): if priming effects can only be observed in sentences with lexical overlap between the prime and target sentence, then these effects should not be noticeable in sentences that share verbs that are closely related in meaning and can occur in the same syntactic structures. Similar to Tooley et al. (2009), they found a reduction in the P600 when the prime and target sentence shared identical verbs. This finding was not observed for the synonym prime condition. For this reason, Chen et al. (2013) related their finding to the lexicalist account since a reduction of the P600 was only detected when the prime and target sentence shared the same verb and not when the prime sentence contained a synonym to the verb in the target sentence.

To summarize, even though syntactic priming in comprehension has mostly been investigated in complex sentences such as reduce-relative clauses, the studies discussed suggest that priming effects in comprehension are dependent on verb overlap. However, it is still not known whether similar priming effects in comprehension may be observed for other syntactic structures and in different language than English. Therefore, the next section will elaborate on the syntactic structure that is going to be employed in the investigation of syntactic priming in comprehension in the present study — namely passive sentences in Dutch.

2.11

The proposed stimuli for this study: passive

sentences in Dutch

As mentioned beforehand, the studies on syntactic priming and sentence comprehension, typically use complex sentences such as reduced-relative sentences. However, production studies have used either passive and active sentences or datives (DO/PO) as their stimuli. This difference in stimuli makes it difficult to compare syntactic priming in compre-hension and production. Syntactically complex sentences such as reduced-relatives are more often used in studies investigating syntactic priming in comprehension because these sentences require more processing costs than the relatively easier sentences used in production studies. It is assumed that the advantages of priming in these sentences would be more evident as it may reveal whether syntactic priming reduces processing costs (Pickering & Ferreira, 2008). However, as suggested by Tooley and Traxler (2010), ideally, direct comparison between syntactic priming in comprehension and production requires the same sentence types. Using similar syntactic structures —as the ones used in production studies — for comprehension studies can shed light on whether priming ef-fects in comprehension may be only limited to complex sentences (e.g., reduced-relatives). Moreover, direct comparison between production and comprehension by means of inves-tigating the same sentence type could clarify whether the same processes are involved that contribute to syntactic priming effects (Tooley & Traxler, 2010).

According to Mack, Meltzer-Asscher, Barbieri, and Thompson (2013), passives are non-canonical and therefore more difficult to process than active sentences. In passives, the theme is the subject and the agent is mapped into the adjunct position and serves as the object of the sentence. Mack et al. (2013) stated that passive sentences require revision of the position of the nouns (i.e., noun phrase-movement). Moreover, due to the movement of the noun phrase, thematic reanalysis also occurs because people tend to interpret the first noun as the agent of the sentence (‘first-agent bias’, see Mack et al., 2013). However, when one has come to the end of the sentence, the thematic role of the first noun must be revised from agent to theme (Mack et al., 2013). With regards to processing passive versus active sentences, a recent study has demonstrated that processing differences between these two sentence structures are observable in continuous EEG signal. Jackson et al. (n.d.) used ERPs to investigate whether passive and active sentence comprehension elicit differences in ERPs. In their study, they found a frontal positivity for passives and this was related to a larger P600 effect. The researchers speculated that this difference could be due to revision of passive sentences while reading.

Only few studies have investigated syntactic priming in actives and passives in Dutch (e.g., Hartsuiker & Kolk, 1998b; Segaert, Kempen, Petersson, & Hagoort, 2013). Hartsuiker and Kolk (1998b) investigated syntactic priming in production and examined whether similar effects for Dutch passives could be observed as reported in English (see e.g., Bock, 1986). In comparison to English, Dutch allows for more word orders than English in the construction of passives with transitive verbs. The more common word order in passives are clauses with a final by-phrase, similar to English (see Hartsuiker & Kolk, 1998b: De wandelaar wordt bevuild door de modder – The walker is dirtied by the mud). This structure had already been tested in Bock (1986), however, Hartsuiker and Kolk (1998b) also examined whether priming effects would be observed in structures that had not been tested before. That is, they also tested passives with a sentence-final passive participle (e.g., De wandelaar wordt door de modder bevuild – *The walker is by the mud dirtied). The verb-final passive is regarded as a variation to the more standard word order of passives in Dutch. They also included actives in their experiment. However, their results only showed priming effects for both passive structures, but no effect was found for actives, even though Bock (1986) did find priming effects for English actives. One of the proposed explanations for their finding was that passive structures are less frequent than active sentences. Therefore, these structures might benefit more from syntactic priming than more frequent structures, such as active sentences (Hartsuiker & Kolk, 1998b).

Segaert et al. (2013) employed functional magnetic resonance imaging (fMRI) to investigate syntactic priming in production as well as in comprehension using Dutch passive and active sentences. The researchers measured the magnitude of brain activity during the presentation of sentences in active or passive voice. The sentences either contained verb repetition or not. In the production task, participants were instructed to produce a sentence describing a picture which illustrated an agent and an object. The picture had to be described with a transitive verb. Prior to describing the picture, participants were presented with the infinitive form of the transitive verb that they were instructed to use in their description of the picture. In the comprehension task, participants were shown a picture of an action being performed and they simultaneously heard a sentence describing the picture in either a passive or an active voice. Participants were also given the opportunity to press a button if they believed that there was a mismatch between the spoken description and the picture. The results showed brain activity for active sentences only when the prime and target shared the same verb. No similar effects were found when the active prime and target sentence did not share a

similar verb. In contrast, the result showed brain activity for passive sentences in both conditions: when the verb between the prime and target sentence was repeated and when it was not. They explained their findings based on the assumption that actives have a higher frequency than passives and therefore repeating active sentences with novel verbs did not reduce neural activity. In contrast, passives are less frequent than actives and therefore participants showed priming effects in both the verb repetition and novel verb condition and benefited more from priming. Importantly, their findings were not affected by the modality (production vs. comprehension) in which the task was performed. Consequently, the researchers suggested that syntactic priming may be driven by the same or similar mechanisms in both modalities.

Altogether, the need for comparison between syntactic priming in production and in comprehension motivated the present study to employ passive sentences to investigate whether priming effects can also be observed for these structures in comprehension. Therefore, we can now formulate our research questions and the accompanying hypothe-ses.

C

H

A

P

3

R

ESEARCH

Q

UESTIONS AND

H

YPOTHESES

T

o investigate syntactic priming in comprehension, we base our study in the four following research questions:

1. Does syntactic priming facilitate the comprehension of passive sentences? 2. If any, how are these priming effects reflected electrophysiologically? 3. Are there additional facilitating effects of verb repetition?

4. To what extent does syntactic priming rely on WM?

Based on studies concerning passive sentences, these structures are relatively in-frequent. Due to their infrequent nature in comparison to active sentences, passive sentences are also more difficult to process. As mentioned before, people have the ten-dency to process the first noun that they encounter as the agent of the sentence. When one encounters a passive sentence and realizes that he or she has to revise their syntactic expectations, this might result in effortful processing. For this reason, repeated exposure to passive sentences may reduce this effortful processing. Therefore, it is expected

that syntactic priming facilitates the comprehension of passive sentences. In

the current study, a number of methodological manipulations were chosen to enhance possible syntactic priming effects. Firstly, possible priming effects were expected to be enhanced by the repeated use of complex noun phrases. This was employed since Cleland

and Pickering (2003) observed that syntactic overlap in noun phrases can also result in syntactic priming effects (e.g., De charmante man met de brede lach wordt gevolgd door de vrouw – The charming man with the big smile is being chased by the woman). Moreover, possible priming effects were expected to be further enhanced by the fact that the prime and target sentence always immediately follow each other. Presenting the passive sentences in immediate conditions would enhance the short maintenance and activation in memory (see e.g., Bernolet et al., 2016). Importantly, possible confounds due to the added methodological manipulations were resolved by means of filler sentences (see Chapter 4 for a discussion of the filler sentences).

Due to the continuous presentation of passive sentences, it is expected that

pos-sible syntactic priming effects might be reflected in the P600. As readers are

expected to revise and update their mental representation of the passive sentences, the P600 is hypothesized to reflect this process (see also Retrieval-Integration account). With regard to additional facilitating effects due to verb repetition (research question 3), this may be observed when time-locking the continuous EEG data to the word wordt — to be/become. According to Ledoux et al. (2007), who found a decreased N400 effect that reflected lexical repetition of the main verb in the target sentence, lexical repetition ef-fects may be also found for wordt even though it is an auxiliary verb. This entails that

the presentation of wordt in the prime sentence, followed by a repetition of the same word in the target sentence is expected to elicit repetition priming revealed in the N400.

The final research question is an additional exploratory part of the present study that examines whether cognitive abilities that are linked to working memory could explain differences in processing a prime passive sentence compared to a target passive sentence. As mentioned by Bernolet et al. (2016), not only verb overlap between prime and target sentence has been alleged to result in larger priming effects, but also the activation of the explicit memory of the prime sentence surface structure. If it is indeed the case that the explicit memory of the prime sentence may result in a short-lived syntactic priming effect, then there should be a memory system that supports this retention. The retention of the explicit memory of the prime sentence then functions as a cue for readers once they encounter a similar structure. Ledoux et al. (2007) already alluded to the possible role of working memory in the aforementioned process. As working memory is a system that is involved in the maintenance of task-relevant information (e.g., the storage of a syntactic structure of a sentence) during the performance of a cognitive task (e.g., sentence processing; see also Miyake & Shah, 1999), it is expected that electrophysiological

differences between the prime and the target sentence may be explained by adding working memory score as an additional predictor to the model of the mean voltage in the ERP signal.