Sentence comprehension of the Zinnen en Anagrammen Test (ZAT) – An adaptation to German

- MASTER THESIS -

Sentence comprehension of the Zinnen en

Anagrammen Test (ZAT) – An adaptation to

German

University of Groningen Faculty of Arts

Master Linguistics (Neurolinguistics) LTS998M20 Student Lea Griemens Student ID: s3800733 Date: 29th of June 2020 Supervisor Dr. Dörte de Kok Second Assessor Srdjan Popov, PhD Word count 14,357

Preface

I am proud to present to you the findings of my Master Thesis ‘Sentence comprehension of the Zinnen en Anagrammen Test (ZAT) – An adaptation to German'. This thesis was written to complete the Master of Arts in Linguistics (Neurolinguistics) at the University of Groningen within the Faculty of Arts.

I would like to thank everyone who helped me with the realization of this thesis. I would like to thank Dr. D. A. de Kok as thesis coach for the intensive guidance and support in the past months. Her advice was always valuable and extensive, and her feedback was essential for the progress of this project. Furthermore, I would like to express my special thanks to all the people I have been able to integrate into my research phase. A word of thanks also goes to the second assessor of this thesis.

Last but not least I would like to thank my family and friends. They have always supported me morally during the writing process and have always found motivating words. I really appreciate their support.

I hope that with the final product, the German version of the ZAT, I can make a contribution for German-speaking countries to offer a more focused disorder-oriented treatment in terms of improving aphasia and the resulting problems. It would be a great result if I could contribute to improving their quality of life in this way.

Lea Griemens

Table of Contents

1 Introduction ... 1

2 Theoretical framework ... 3

2.1 Acquired brain injury ... 3

2.2 Aphasia ... 3

2.3 ICF – the basis for disorder-oriented treatment ... 4

2.4 Language comprehension ... 5

2.4.1 Sentence comprehension ... 6

2.4.2 Sentence comprehension in aphasia ... 10

2.5 Accuracy and reaction time ... 13

2.6 Psycholinguistic variables ... 13

2.7 Diagnostic tools ... 16

2.7.1 WEZT ... 16

2.7.2 action ... 17

2.7.3 WAT and ZAT ... 18

2.8 The present study ... 19

3 Method ... 21 3.1 Adaptation ... 21 3.1.1 Realization of adaptation ... 21 3.2 Online Test ... 22 3.2.1 Participants ... 22 3.2.2 Material ... 23 3.2.3 Procedure ... 26 3.2.4 Data analysis ... 27 3.2.4.1 Pre-study ... 27 3.2.4.2 Online study ... 27

3.2.4.3 Exploratory data analyses ... 28

4 Results ... 29

4.1 Results of pre-study... 29

4.2 Results of online study ... 31

4.2.1 Results of descriptive statistics ... 31

4.2.2 Results of hypotheses ... 33

4.2.3 Results of exploratory data analysis ... 35

5 Discussion ... 36

Conclusion ... 42

6 References ... 43

7 Appendix ... 51

A Zinnen en Anagrammen Test-Duits – subtest sentence comprehension ... 51

B Accuracy and mean RT per item ... 54

C Individual demographic information of each participant ... 55

Abstract

Introduction: People with a language disorder such as aphasia often have trouble comprehending spoken sentences. The number of language tests, which examine language comprehension and production on sentence level, is limited. Therefore, there is a need for a newly developed diagnostic tool that investigates the comprehension and production of verbs on word level and sentence level in aphasia. In German, there is currently no digital diagnostic tool that aims to determine the underlying disorder in sentence comprehension. In Dutch, however, there is such an instrument called the Zinnen en Anagrammen Test (ZAT).

Method: The study was divided into two research phases. Firstly, the subtest sentence comprehension of the ZAT was adapted to German using a shortened realization of the back-translation method. Secondly, this subtest was initially carried out with 30 healthy individuals. It was investigated first whether the items used in the ZAT-D (ZAT-German) are applicable which were examined based on the accuracy rating of the participants. Furthermore, it was investigated to what extent processing (measured by reaction time) is influenced by frequency and imageability of the verbs used as well as the influence of different sentence types.

Results: None of the items had to be excluded from the data analysis as all items were answered correctly by more than 80 % of the participants. It turns out that frequency and imageability of the verbs used in the subtest sentence comprehension of the ZAT-D do not significantly influence the reaction times of the participants. Regarding the different sentence types the reaction times within the regular active sentence structure did not significantly differ from other active sentence structures in different order. The reaction times in passive sentences, however, was higher compared to the reaction times in active sentence structures.

Discussion: The findings show the added value of an adapted German online version of the ZAT due to low inaccuracy ratings. However, these only relate to healthy individuals and, hence, do not imply the validity for people with aphasia (PWA). The insignificant findings are suspected to stem from the inability to show effects via stronger differences in range of frequency and imageability within verbs used and effect in conditions of sentence types. Future research is recommended to check the generalizability and representativeness for PWA. It is assumed that the completed version of the ZAT-D will provide an innovative digital diagnostic instrument that aims to determine the underlying disorder in sentence comprehension and production in PWA.

1

1

Introduction

Annually, there are about 270.000 people in Germany affected by a stroke (Stiftung Deutsche Schlaganfall-Hilfe, 2020). The annual number of incidences for newly developed aphasias requiring treatment after a stroke is estimated at around 25.000 (Rupp-Adelmann & Bauer, 2010). Aphasia is an acquired language disorder, caused by focal brain damage that arises after language has been acquired. It can affect both language production and language comprehension as well as spoken and written modality (Bastiaanse, 2010).

People with aphasia (PWA) often have trouble comprehending spoken sentences (ASHA, 2020). However, only a limited number of language tests are available to detect comprehension problems on sentence level. Until a while ago the Dutch Werkwoorden en Zinnen Test (WEZT; Bastiaanse, Maas, & Rispens, 2000) was used to examine the comprehension and production of verbs at word level and at different positions within a sentence in PWA. This is no longer the case due to its complexity and length regarding its interpretation and is, hence, soon to be replaced by two digital and more comprehensive versions. One of them is the Werkwoorden en Actie Test (WAT; De Kok, Wolthuis, & Bastiaanse, 2013). This test investigates the comprehension and production of verbs on word level and sentence level. The other one is the Zinnen en Anagrammen Test (ZAT; De Kok & Bastiaanse (in prep.-a)), which aims to determine the underlying disorder in sentence comprehension and production. With the use of those digital tests, the testing time could be shortened. Firstly, the patient can go through the test independently with only limited help of the researcher and secondly, results are immediately visible to the researcher since the software does all the calculations itself. Even though the content of those tests is based on the WEZT, the illustrations were renewed, and subtests have been adapted using new stimuli.

Digital tools in German that aim to investigate the comprehension and production of verbs on word level and sentence level are limited. A program called action (Bastiaanse, Bung, & Perk, 2004) is a German treatment program that incorporates a screening for PWA that have problems finding and embedding verbs within the sentence. This test only consists of a limited number of subtests that do not completely investigate the underlying disorder in the comprehension and production of verbs on word level and sentence level. Recently, a German version of the WAT, called Werkwoorden en Actie Test-Duits (WAT-D; De Kok & Bastiaanse (in prep.-b)), has been developed in form of an application for the iPad. However, the ZAT is not yet adapted to German.The aim of this study is to develop an adapted version of the subtest

2

sentence comprehension of the ZAT that is a valid German diagnostic tool to investigate sentence comprehension in PWA. The test shall be applicable for all patients irregardless of their demographic factors such as age, gender or level of education.

In the following chapter, a theoretical framework is given on acquired brain injuries, aphasia and disorder-oriented treatment. Furthermore, the language comprehension in healthy individuals and individuals with a language disorder will be discussed. Finally, accuracy and reaction time will be discussed as well as how psycholinguistic variables have influence on linguistic processing. Chapters 3 to 5 will discuss the method, results and discussion before finally concluding the findings of the present study.

3

2

Theoretical framework

2.1 Acquired brain injury

Acquired brain injury (ABI) refers to post-natal cerebral damage, rather than an insult that is hereditary, congenital, degenerative or induced by birth trauma (Ontario Brain Injury Association, 2020). ABI can be subdivided into traumatic and non-traumatic subtypes (Prins, Greco, Alexander, & Giza, 2013). Cranial brain trauma is a traumatic brain injury that occurs after violence (e.g. motor vehicle collisions, fights and falls) and is caused by an external force. Non-traumatic brain injuries include inflammatory disorders (brain infections), brain tumors (benign and malignant), hypoxia (oxygen deficiency in the brain) and cerebrovascular accidents (Shepherd Center, 2020).

The most common cause of aphasia is a cerebrovascular accident (CVA). CVA is also the second most common cause of death in Germany. Annually, around 270.000 CVA occur, of which 200.000 are first time CVA’s (Stiftung Deutsche Schlaganfall-Hilfe, 2020). In the first four weeks post-onset of a CVA, around 28 % of the patients die, more than 41% die within a year (Hoffman, 2018).

One-sided paralysis, sensory disturbances of arms and legs, swallowing, balance, perception and language disorders are the most common consequences of a CVA. As a result, half of the victims are permanently disabled and dependent on the help of others (Stiftung Deutsche Schlaganfall-Hilfe, 2020).

2.2 Aphasia

Etymologically, the word aphasia comes from the Greek and is an expression for “without words”, which does not necessarily mean a complete loss of language (Rupp-Adelmann & Bauer, 2010). The symptoms of aphasia differ per individual. They can vary depending on the pattern and severity of the disorder (Kolonko & Hunziker, 2013).

According to the DGN (Deutsche Gesellschaft für Neurologie), the course of aphasia after the development of an ABI is divided into three phases: acute phase (the first 2 – 6 weeks), post-acute phase (up to 6 months post-onset) and consolidation phase (after 6 months post-onset) (Diener & Weimar, 2012).

During the acute phase, the linguistic symptoms change significantly and cannot yet be classified. Spontaneous remission is possible. Characteristic symptoms of a fluent or non-fluent

4

aphasia become evident in the post-acute phase but progress can still be made. Although progress can rarely be made in the consolidation phase, it is still possible (IZA, 2019).

The German version of the ZAT is mainly designed for PWA in the post-acute or consolidation phase.

2.3 ICF – the basis for disorder-oriented treatment

In 2001, the World Health Organization (WHO) presented the International Classification of Functioning, Disability and Health (ICF). The German translation of the ICF was published in 2005 by the ‘Deutsches Institut für Medizinische Dokumentation und Information’ (DIMDI) (Deutsches Institut für Medizinische Dokumentation und Information, 2020). Figure 1 shows the ICF model.

Figure 1. ICF model (World Health Organization, 2001)

The ICF model is a classification with which it is possible to describe the functioning of people and the possible problems that people experience. Moreover, the factors that influence that functioning can also be recorded. In the ICF, the emphasis is on the dynamic interaction between disabilities, the aspects of activities and participation and contextual factors (environmental and personal). When a change occurs in one of the components, this will also affect the other components (Rosenbaum & Stewart, 2004).

5

In order to provide objective information about the type and severity of the disorder, diagnostic instruments, such as standardized test procedures, are necessary. In the acute phase of aphasia only short assessments are carried out due to the low capacity of the patient. In the post-acute phase and the consolidation phase, however, it is important that the assessment provides detailed results about the specific linguistic and communicative possibilities and limitations (Schneider, Wehmeyer, & Grötzbach, 2014) in order to establish the underlying disorder. For a thorough evaluation of the language problems it is necessary to test all language levels within the different modalities.

The main goal in aphasia therapy is to improve the communication possibilities in people’s everyday life with aphasia (Vandenborre, Visch-Brink, & Mariën, 2014). In functional- or communicative therapy the PWA are trained to use their remaining language skills combined with compensation strategies and supporting communication tools. Disorder-oriented treatment focuses on the underlying disorder of the communication problem and is aimed at restoring language functions (Visch-Brink & Wielaert, 2005). It is also developed on the basis of linguistic models of language processing (De Jong-Hagelstein, 2011).

2.4 Language comprehension

Language comprehension is one of the most automatic tasks a human being is able to perform. Yet, it is also one of the most complex ones as it requires simultaneous integration of many different types of information, such as knowledge about letters and their sounds, spelling, grammar and word meanings. Language comprehension is a generic term to describe the ability to derive meaning from written and oral language (Catts, Adlof, & Weismer, 2006; Hoover & Gough, 1990). For people with comprehension deficits assembling all the information mentioned above can be quite challenging. In the case of language comprehension difficulties, the meaning of words and sentences is not understood, although there is no hearing impairment (Huber, 2013).

The words of a language are stored in the so-called lexicon (Bastiaanse, 1993). The lexicon contains two components: a meaning component and a lexeme component. The meaning component consists of the so-called lemmas, which contain the semantic and syntactic information of the words. Hence, the lemma of, for example, the verb ‘umarmen’ in German (‘to hug’ in English) contains the information that it is a verb and that it needs a subject and an accusative object to form a correct sentence. Furthermore, the verb lemma determines the thematic roles that need to be filled. The entity that performs the action is technically known

6

as the agent and the entity that is involved in or affected by the action is called the theme or patient (Yule, 2016). Sometimes theme and patient are used to refer to the same thing. However, theme describes something that does not change its state (e.g. ‘I have two children.’) whereas patient describes something that changes its state (e.g. ‘I crushed the car.’). The lexeme component contains information about the associated phonological word forms. Information about the inflection of a verb is also stored here. Figure 2 is a graphical representation of the lexicon.

Figure 2. Schematic representation of the lexicon

Verbs play a central role in sentence comprehension. As the verb must be recognized as such, an access to the conjugated form is needed and the syntax has to be comprehensible. Thus, in addition to lexical aspects, morphological and syntactic characteristics also play a role in processing.

In everyday life people tend not to speak in single words but in whole sentences. Therefore, all aspects, lexical, morphological and syntactic, have a certain function when it comes to a sentence comprehension.

2.4.1 Sentence comprehension

Sentence comprehension refers to the subfield of psycholinguistics that requires the interpretation of written and oral language (e.g. an auditory stimulus) on sentence level. In psycholinguistics, one of the most widely used language models is by Levelt (1989). Although this model represents a milestone in the field of language research it has been adapted several times. The theory by Bastiaanse (1993) described below shows which aspects play a role in

7

sentence comprehension and the construction of a sentence. A translated version of the language processing model by Bastiaanse (1993) is represented in Figure 3.

Figure 3. Schematic representation of a translated language processing model by Bastiaanse (1993)

A few preliminary remarks on the notation may be in order. Notation for the representation of constructions in grammar are shown in Table 1 below.

Table 1

List of abbreviations used for representation of constructions in grammar

Constituent Abbreviation Noun N Subject SBJ Verb V Agent A Theme T Patient P

Accusative Object AccOBJ

Adverbial ADV

In order to comprehend a sentence, knowledge must be gained about the meaning of each word that occurs in the sentence. The verb has a central role in sentence comprehension, and it must

8

be identified first. The process is called grammatical decoding. Within a sentence it is necessary to determine certain roles of action (thematic roles). As described in section 2.4, the role that performs the action is called the agent (A) and the role that is involved in or affected by the action is called the theme (T) or patient (P) (Yule, 2016). The action itself is represented by a verb (V). The step that assigns thematic roles to grammatical functions is called mapping. The following sentence is used to explain the different processes:

(1) Der Mann umarmt die Frau. The man hugs the woman.

When an auditory message is sent to the receiver, this message must be coded semantically, which activates lemmas in the lexicon. Thus, the concept ‘to hug’ contains the information that person X is doing something with person Y. The lemma ‘to hug’ is retrieved from the lexicon. The role of the agent is assigned to person X and the role of the theme is assigned to person Y. The structure looks as follows:

(2) umarmen (Mann Frau)

hug (man woman)

V N N

A T

This structure must now be decoded syntactically. The verb ‘to hug’ contains syntactic information, namely that it needs a subject and an accusative object. The syntactic roles must be assigned to the semantic roles. This results in the following structure:

(3) Mann umarmen Frau

man hug woman

SBJ V AccOBJ

A T

If the sentence is changed to ‘The woman is hugged by the man.’, the adverbial must be assigned to the agent and the subject to the theme. In this case, the structure looks as follows:

(4) Frau umarmen Mann

woman hug man

SUBJ V ADV

9

Sentence comprehension can be influenced by several factors such as reversibility of semantic roles, word order and morphology. A sentence is semantically reversible when two people that are involved in an action can take over the role of the one who performs the action (A) and the one who receives the action (T). In a semantically reversible sentence (e.g. ‘The man hugs the woman.’), the two people (man and woman) could take over the role of the agent or the theme. This means that both, ‘the man hugs the woman’ and ‘the woman hugs the man’, form meaningful sentences. Nevertheless, it should be emphasized that the reversibility only refers to a theoretical interchangeability of the thematic roles. The statement of the sentence, however, only allows one interpretation. In irreversible sentences the reversal of the roles is not possible. A sentence like ‘the father eats an apple’ contains a person (the father) and an object (the apple), but only the father can take on the role of the agent. Therefore, irreversible sentences only allow one interpretation as well, but they also do not allow a theoretical interchangeability of the thematic roles (Burchert, 2011).

Furthermore, there are additional properties that can increase the complexity of semantically reversible sentences in sentence comprehension. Even in healthy individuals there is a dissociation between sentences that are in regular structure (e.g. active sentences) and sentences that are in derived order (e.g. passive sentences). Several studies have reported that passive sentences take longer to comprehend than active sentences (Clark, 2019; Forster & Olbrei, 1973; Slobin, 1966).

The order of the words within a sentence is also important. In German, a change in word order often results in a change of orientation (Meyer, 2005). The words appearing at the beginning of a sentence have the ability to emphasize the constituent in question and give the sentence a certain direction. The German sentence ‘Ich kaufe Blumen für meine Oma.’ wants to emphasize the agent of the sentence meaning I am the one who is buying flowers for my grandmother and not my brother. By changing the word order in the German sentence one can give the sentence another orientation: ‘Blumen kaufe ich für meine Oma.’. This sentence emphasizes the patient meaning I want to buy flowers and no chocolate. The sentence ‘Für meine Oma kaufe ich Blumen.’ emphasizes the theme meaning I want to buy the flowers for my grandmother and not for my grandfather. In English it is not possible to change the word order of this sentence and still have a grammatically correct sentence:

10

(5) Ich kaufe Blumen für meine Oma.

I buy flowers for my grandmother.

A

Blumen kaufe ich für meine Oma.

* Flowers buy I for my grandmother.

P

Für meine Oma kaufe ich Blumen.

* For my grandmother buy I flowers.

T

In synchronic linguistics the asterisk (*) is used to mark expressions or sentences that are not well-formed according to the rules of the language in question (Meyer, 2005).

Depending on the language, morphological markings in nouns and verbs can be clear indicators of the grammatical function of the noun within a sentence (subject or object). The German language has an extensive system of word forms, such as case endings (nominative, accusative, dative, genitive) and number markings (singular, plural). Those markings are supportive when it comes to sentence comprehension, especially for people without any language disorder (Burchert, De Bleser, & Sonntag, 2003).

2.4.2 Sentence comprehension in aphasia

PWA often have problems with sentence comprehension (ASHA, 2020). However, the syntactic problems in sentence comprehension are often restricted to certain sentence types (Burchert, 2011). The use of morphology and syntax is also disturbed, which greatly shortens sentence structure and grammar (agrammatism). Regarding agrammatic sentence comprehension, a typical performance pattern has been described, which is characterized by a dissociation between semantically reversible active and passive sentences. Several studies have reported that sentence comprehension in agrammatic aphasia is impaired. In particular, semantically reversible sentences that are illustrated in their derived and not in their base position (e.g. passive sentences), are shown to be less often comprehend correctly than their counterparts (e.g. active sentences). In many cases, patients show an accuracy for active sentences that is significantly above chance level, while for passive sentences participants

11

perform on chance level (Burchert & De Bleser, 2004; Luzzatti et al., 2001; Meyer, Mack, & Thompson, 2012).

People with agrammatism have great difficulty understanding grammatically complex constructions such as passive sentences where elements appear in derived order compared to active sentences (Dick et al., 2001). As illustrated below, in the passive sentence the noun phrase (NP; ‘the woman’) moved from its original position to the first position in the sentence, but it does not function as the agent.

Active sentence:

(6) The man hugs the woman. Der Mann umarmt die Frau.

A T

Passive sentence:

(7) The woman is hugged by the man.

Die Frau wird von dem Mann umarmt.

T A

Numerous approaches and hypotheses have been developed in recent years to explain the deficits in agrammatic speakers. However, over the years there has been much controversy among researchers in characterizing and accounting those deficits. It even has been argued that the comprehension of each sentence structure is disrupted according to the complexity of the structure and the severity of someone’s aphasia and not because of a deficit of a certain linguistic aspect (Clark, 2016). Several linguistic theories have been proposed which can be grouped into two categories: representational deficit accounts and processing deficit accounts. One of each category will be presented.

The Trace Deletion Hypothesis (TDH; Grodzinsky, 1984) belongs to the representational category. The TDH is one theoretical approach, which tries to explain the underlying deficit of sentence comprehension in agrammatic speakers. In this theory, it is assumed that in patients with agrammatic sentence comprehension, the traces that remain in the base position when NPs are moved are deleted from the surface structure. However, thematic roles are only assigned to the corresponding base position and if necessary, throughout the traces. This means that moved NPs cannot be assigned a thematic role if the corresponding traces are no longer represented on the surface structure.

12

Grodzinsky argues that PWA follow a heuristic strategy (problem-solving strategy that ease the cognitive load of making a decision) in which they always assign the agent to the first NP in the sentence. As illustrated in the example above, the NP (‘the man’) is given the thematic role of the agent. Consequently, in the aphasic representation, the moved NP (‘the women’) remains in the passive sentence without a thematic role assigned as it is not in base position and not linked to one by a trace. Grodzinsky assumes that in order to produce a thematic interpretation, the affected patients use a strategy by which the first NP in the sentence to which no thematic role has yet been assigned is assigned the agent role. Therefore, in addition to the NP (‘the man’), the moved NP (‘the women’) also receives the thematic role of the agent. Thus, within the framework of the TDH for patients with agrammatic sentence understanding, passive sentences receive a double agent representation. Therefore, patients are forced to guess which one of the two NPs is the agent potentially leading to random performance.

The TDH, however, cannot account for all comprehension deficits in agrammatic speakers. Several researchers even have disputed the TDH. Berndt, Mitchum and Haendiges (1996) did research on the comprehension of passive and active constructions and claimed that there is no homogeneous pattern in agrammatic speakers with comprehension deficits. In 1998, Beretta and Munn showed that agrammatic representations do not involve double agent representations in passive sentences, as earlier suggested by Grodzinsky.

Revised has the TDH been by Drai and Grodzinsky (2006) including data from Dutch and German agrammatic speakers. This hypothesis also has been disputed through another hypothesis called the Derived Order Problem-Hypothesis (DOP-H; Bastiaanse, & Van Zonneveld, 2005; 2006). This theory belongs to the processing category. Unlike the TDH which is restricted to comprehension deficits, the DOP-H includes both comprehension and production deficits in agrammatic speakers. According to the DOP-H all languages have a base order in which the constituents of a sentence are represented. However, this base order is not the same for all languages. Word orders that differ from the base position are known as derived orders. The DOP-H claims that agrammatic speakers have difficulty comprehending and producing sentences with derived word orders while sentences in base order are relatively spared.

Whereas the TDH is a representational account that predicts agrammatic speakers’ deficits only with comprehension of sentences, the DOP-H is a processing account that makes similar predictions but for both production and comprehension. Until now there is no overarching

13

theory that covers all underlying deficits in agrammatic speakers that applies to any language. Nevertheless, both theories, TDH and DOP-H, have in common that sentences in a derived order are more difficult to comprehend for agrammatic speakers than those in base order.

2.5 Accuracy and reaction time

Accuracy and reaction time (RT) are commonly used measurements in test situations. In language comprehension tasks, accuracy refers to how correct someone uses their language system including vocabulary and grammar (Hinkel, 2003). Reaction time refers to the amount of time it takes between perceiving and responding to something. It is the ability to detect, process and respond to a stimulus (CogniFit, 2019). Therefore, RT depends on various factors such as perception, processing and response.

Perception: To have a good RT, it is important to be able to see, hear and feel a stimulus well. For example, when a signal sounds at the beginning of a task, the sound is received by the ears of the participant (perceive the stimulus).

Processing: To have a good RT, it is necessary to be concentrated and to understand the information well. According to the previous example, after hearing the signal, the participant is able to distinguish the sound from other background noises and knows that it is time to start the task (process the stimulus).

Response: To have a good reaction time, mobility is necessary to be able to respond. When the signal is perceived and correctly processed, a response is started (response to the stimulus).

One of the fundamental issues that researchers in linguistics should realize about accuracy and RT measurements is that they are extremely delicate. In any experiment involving accuracy and RT there are issues to consider. Not only individual-related factors, such as age (Myerson, Robertson, & Hale, 2007), gender (Der & Deary, 2006), physical (Welford, 1980) or mental state as well as level of education (Iverson, 2006) but also external factors can have influence that contribute (milli-)seconds to the measured RT. External factors could be slow connection to the server when conducting an online test or extraneous noises (Racine, 2013). All these different factors can have influence on the resulting data.

2.6 Psycholinguistic variables

One possible way to characterize the nature of linguistic processing is by using data from psycholinguistic variables. The ease and speed with which language is retrieved and processed is influenced by several linguistic factors, both in healthy individuals and PWA.

14

Psycholinguistic variables can affect various levels in linguistic processing. The factors that are relevant for the present study are frequency and imageability.

Word frequency

The difficulty of retrieving word forms from the lexicon is partly determined by the word frequency. Regarding the language processing model by Bastiaanse (1993) that was described in section 2.4.1, word frequency is considered to be a lexical variable that plays a role at the lexeme level (Jescheniak & Levelt, 1994). The more often a verb appears in a language the easier it is to retrieve because the activation threshold is lower. However, word frequency depends on the size of the corpus. When comparing between corpora, the main standardized measure has been frequency per million words (fpm). Low-frequency words are typically defined as having less than five fpm (e.g. ‘strait’ or ‘frenzy’) and high-frequency words are defined as having more than 100 fpm (e.g. ‘and’, ‘make’) (Brysbaert, Mandera, & Keuleers, 2018). However, this standardized measure lead to wrong intuitive understanding of the word frequency effect. Therefore, another standardized scale for word frequency was necessary. Today, a quite common measure is Log10 (frequency per billion words). It solves the problem of putting differences between items in perspective as it uses a logarithmic rather than a linear scale. Furthermore, it looks like a typical Likert scale (e.g. from 1 to 5), so that the values are easy to interpret and finally, the middle of the scale separates the low-frequency words from the high-frequency words (Van Heuven, Mandera, Keuleers, & Brysbaert, 2014). Of all the variables that have been studied, the effect of word frequency appears to be the most stable and replicable regarding results of PWA (Nickels, 2014).

Several studies have reported that PWA show effects of word frequency which means having more difficulty processing low-frequency words than high-frequency words. Newcombe, Oldfield, and Wingfield (1965) found a linear relationship between the latency of correct naming and the logarithm of the picture name’s frequency for PWA compared to healthy controls. This statement has been confirmed throughout the years by a number of studies which have found a relationship between naming accuracy and word frequency for PWA (Bastiaanse, Wieling, & Wolthuis, 2016; Butterworth, Howard, & Mcloughlin, 1984). According to DeDe (2012) sentence comprehension in PWA is also influenced by word frequency. Regarding the accuracy data in that study, PWA had more difficulty in sentence comprehension containing low-frequency words.

15

One study reported measures for accuracy and RT of PWA during an auditory comprehension task (Schulte, 1986). The response accuracy of the PWA was significantly affected by the complexity of the stimulus. Furthermore, RT measures indicated that PWA were slower in responding compared to the control group. Frequency is a well-studied variable that affects lexical access time. It was claimed that measuring RTs are even more sensitive for comprehension deficits in PWA than measuring accuracy. According to DeDe (2012), PWA are not only sensitive to word frequency but also RTs provide a more sensitive measure of word frequency effects than accuracy. In summary, it can be concluded that word frequency effects in PWA affects both accuracy and lexical access time which in turn leads to longer RTs in several tasks.

In healthy individuals, sentence processing also depends on lexical characteristics of read words in a sentence, such as word frequency (Van Petten and Kutas, 1990). Since high-frequency words are known to more people and are processed faster than low-high-frequency words, the word frequency effect also affects the processing efficiency in healthy individuals (Monsell, Doyle, & Haggard, 1989). Multiple studies using various tasks established that non-brain damaged individuals also require more time to process low-frequency words than high-frequency words (Brysbaert, Lange, & Wijendaele, 2000; Ferreira, Henderson, Anes, Weeks, & McFarlane, 1996; Turner, Valentine, & Ellis, 1998).

Imageability

Imageability is a semantic variable that plays a role at the level of the concept and the lemma. It is used to indicate how well a mental image or sensory experience can be evoked by a particular word (Bird, Howard, & Franklin, 2000). It is assumed that imageability is an important aspect of the nonverbal code and that the images elicited by concepts serve as mediators in associations (Paivio, 1971). Imageability ratings are typically collected through questionnaires where the imageability of words is scored on a Likert scale by healthy individuals on the basis of relativity. In some studies a higher score means that a word is more concrete/more imaginable (Rofes et al., 2018) while in other studies a higher score stands for how difficult a word is to imagine (De Kok, Wolthuis, & Bastiaanse, 2013).

Psycholinguistic studies have shown that words that are rated high in imageability are typically processed faster and more accurately than low-imageability words (Bird, Franklin, & Howard, 2001; Dubé, Monetta, Martínez-Cuitiño, & Wilson, 2014). For example, the verb ‘to sleep’ typically rates high in imageability, whereas ‘to happen’ rates low in imageability, as it is more

16

abstract. The fact that processing is more accurate and efficient for high-imaginable words than low-imaginable words is true for both healthy individuals and PWA (Alyahya, Halai, Conroy, & Ralph, 2018).

Various studies have shown that high-imaginable verbs are easier to retrieve for PWA than low-imaginable verbs (Howard & Gatehouse, 2006; Luzzatti et al., 2002; Park, Goral, Verkuilen, & Kempler, 2013). In comprehension tasks, PWA showed lower accuracy for abstract/low imageability words than concrete/high imageability words (Franklin, 1989). This effect was also found in PWA studied by Nickels & Howard (1995).

It is important to note that effects of word frequency and imageability are commonly found in psycholinguistic research. Even people without language impairments show differences in the ability to retrieve words relative to those variables (Faust, 2015). This difference, however, is only captured in a sensitive measure such as RT. For PWA on the other hand, frequency and imageability effects can be quite strong depending on the level of processing that is impaired (Papathanasiou, Coppens, & Potagas, 2016).

2.7 Diagnostic tools

The number of language tests, which examine language comprehension and language production, is limited. Therefore, there is a need for a newly developed diagnostic tool that can be used to investigate the comprehension and production of verbs on word level and sentence level in aphasia.

2.7.1 WEZT

The Dutch diagnostic instrument, WEZT (Bastiaanse et al., 2000) is a very comprehensive test that focuses on determining the underlying disorder based on the sentence processing model (Levelt, 1989) and examines both the comprehension and production of verbs at word and sentence level.

The test consists of the following eight subtests: Language production:

1. Naming actions: The patient has to say in one word what the person in the illustration is doing. Each illustration contains either a transitive or intransitive action verb. 2. Filling in a verb in a sentence: The patient is offered sentences in which the verb is

replaced by dots. The patient has to fill in correctly either the conjugated verb or the infinitive.

17

3. Producing a sentence: The patient has to produce sentences with transitive or intransitive verbs with an illustration. Half of the transitive sentences are semantically reversible and half semantically irreversible.

4. Arranging sentences out of three anagrams (ordering the words): Half of the items are presented with illustrations and half of the items are presented without illustrations; the task consists of active and passive sentences, half of which are semantically reversible and half semantically irreversible.

5. Arranging wh-questions out of anagrams: Again, the patient is offered an illustration and additionally cards with separate words. The patient is asked to construct a question that matches the illustration (two types of questions: argumentative- and adjunction-questions).

Language comprehension:

6. Grammatical assessment test: The patient is offered 50 sentences (right/wrong judgment, the accuracy of the sentence always relates to the thematic roles).

7. Verb comprehension test: The test consists of 60 items, in which the patient has to indicate the heard verb out of four illustrations depicting different actions (transitive and intransitive verbs).

8. Sentence comprehension test: This test examines whether the patient can assign the right thematic roles to the different constituents in the sentence. Again, the patient has to choose from four illustrations (semantically reversible sentences with differences in the order of the thematic roles, such as active and passive sentences).

Even though the WEZT focuses on determining the underlying disorder it appeared to be scarcely used in practice. This is likely due to its complexity in application and interpretation. Firstly, the WEZT is a comprehensive paper test and practice shows that paper tests can be laborious and time consuming. Secondly, the interpretation of the results is based on language models which requires the investigator to understand the models in order to be able to interpret the results correctly. This, however, is rather subjective. After all, this means that not only carrying out the test does take a lot of time but also scoring and interpreting the results due to their complexity. This may be the reason that the demand for this test has declined over time and the test is no longer available.

2.7.2 action

In 2004, Bastiaanse et al. developed a German treatment program called action. This program was primarily developed for PWA who have problems finding or embedding verbs within the sentence. The treatment program was accompanied by a short test to assess the level of impairment.

18

This test consists of the following four levels:

1. Lexical level: The patient has to name the action the person is doing in the illustration. 2. Syntactic level: The patient is offered sentences in which the infinitive verb is missing.

The patient has to complete the sentence with the correct infinitive verb.

3. Morphological level: The patient is offered in which the inflected verb is missing. The patient has to insert the correct inflected verb into the sentence.

4. Construction of sentences: The patient has to arrange single words to a sentence.

Furthermore, it is checked whether the psycholinguistic variables, frequency of the verb and transitivity, influence the results. Whereas frequency measures how often a verb appears in a certain language, transitivity describes whether a verb requires a direct object. After the test has been taken by the patient, therapy is started at the level with which the patient struggles the most.

2.7.3 WAT and ZAT

Since the WEZT is no longer available, it will soon be replaced by two Dutch digital tests. The one is called the WAT (De Kok, Wolthuis, & Bastiaanse, 2013) and the other is called the ZAT (De Kok & Bastiaanse (in prep.-a)). Both are based on the WEZT but are developed for the iPad in form of an application (app). The content of each test is listed below.

The WAT consists of the following six subtests: Language production:

1. Naming actions on word level: The patient is asked to name one word what the person in the illustration is doing. Half of the items consisted of transitive actions and half of intransitive actions.

2. Naming objects on word level: The patient is asked to name the object seen on the illustration in one word.

3. Infinitive completion on sentence level: The patient is offered sentences in which the infinitive verb is missing. The patient has to complete the sentence with the correct infinitive verb.

4. Filling in inflected form on sentence level: The patient is offered sentences in which the inflected verb is missing. The patient has to insert the correct inflected verb into the sentence.

Language comprehension:

5. Semantic association for actions: The patient sees four illustrations in total. The illustration on top is the stimulus that can be associated with one of the three

19

illustrations below. The patient has to identify the illustration that has the strongest relation to the stimulus on top.

6. Verb comprehension on word level: The patient is told a verb on which the corresponding action has to be indicated.

The ZAT consists of the following six subtests:

1. Sentence comprehension: The patient is offered auditory sentences. Out of four illustrations the patient has to identify the correct one.

2. Plausibility assessment: The patient is offered sentences of which it must be identified whether the sentence was correct or simply not possible in terms of content.

3. Sentence anagrams with illustration: The patient is offered an illustration and additionally cards with separate words. The patient is asked to construct a sentence that matches the illustration.

4. Sentence anagrams without illustration: The patient is offered separate words. The patient is asked to construct a sentence without the help of an illustration.

5. Wh- anagrams (who, what, when, …): The patient is asked to construct a question with the help of the offered anagrams.

6. Sentence construction: The patient is asked to construct a sentence.

Recently, a German version of the WAT has been developed, called the WAT-D (De Kok & Bastiaanse (in prep.-b)). A number of linguistic factors have been included in the WAT-D to investigate which of these factors play a role in a possible disorder of verb processing. Frequency, age of acquisition, imageability, visual complexity, word length, instrumentality, name-relation, transitivity and animacy are the factors involved in the analysis of the sub-tests. As can be seen above, the ZAT contains subtests that are not part of the action. However, until now, there is no German version of the ZAT available.

2.8 The present study

From the above it becomes clear that there is a need for a German digital version of the ZAT, which examines verb and sentence processing in detail. In the context of this master thesis, the subtest sentence comprehension of the Dutch version of the ZAT will be adapted to German. The aim is to create a diagnostic instrument for German-speaking countries with which sentence comprehension of PWA can be measured in an effective and reliable manner. The completed test will provide a digital tool for working out a specific therapy plan, which benefits both the therapist and the client. This is the foundation to offer a more focused disorder-oriented

20

treatment in terms of improving the aphasia and the resulting problems. The adapted test is coined Zinnen en Anagrammen Test-Duits (ZAT-D).

Initially the test will be carried out with individuals without a language disorder. First, it will be investigated whether the used sentences are applicable in the test. This will be examined based on the accuracy ratings of the participants. Furthermore, it will be measured whether the different types of sentences and psycholinguistic variables of the stimuli influence the processing of the sentences. This will be measured by the reaction time of the individuals participating in the present study. This results in the research question to what extent is processing influenced by frequency and imageability of the verbs used and the characteristics of the sentence types. Based on the findings from previous research, the following hypotheses were formulated (see also Figure 4):

Hypothesis 1a: The higher the word frequency of the verb the lower the reaction time. Hypothesis 1b: The higher the imageability of the verb the lower the reaction time. Hypothesis 2a: Reaction time (RT) is lower in case of regular active sentence structure (agent (A), verb (V), theme (T)) than in any other active sentence structure (ATV and VAT). Hence, RT is lower in AVT than in ATV as well as RT is lower in AVT than in VAT.

Hypothesis 2b: Reaction time (RT) is higher in case of passive sentence structure (TAV) than in active sentence structures (AVT, ATV and VAT). Hence, RT is higher in TAV than in AVT, RT is higher in TAV than in ATV as well as RT is higher in TAV than in VAT.

21

3

Method

3.1 Adaptation

This section describes the first research phase in which the Dutch version of the ZAT was adapted to German. In the translation process, a shortened realization of the back-translation method was used. The method was used to compare the adapted version with the original for accuracy and quality (Racoma, 2016). The back-translation method prevents a test from losing its reliability and validity (WHO, 2018).

It is a common technique to detect a bias of an item. An independent back-translation means that a translator translates items from the original version of the instrument into a second language and a second translator – someone unfamiliar with the instrument – translates the instrument back into the original language (Lin, Chen, & Chiu, 2005; One Hour Translation, 2015).

However, the use of a translated instrument does not guarantee that it will also measure the same constructions as the original version due to lingual differences (Lin et al., 2005). It is therefore possible that the back-translation is not 100% the same as the original version (Racoma, 2016). Aspects of dealing with possible lingual differences are described next.

3.1.1 Realization of adaptation

The Dutch version of the subtest sentence comprehension of the ZAT was adapted to German by the author of this thesis. Regarding the back-translation method, the next step was to translate the translated version back into Dutch. As mentioned before, only a shortened realization of the back-translation method was used. This means that the translated version was checked by the supervisor, who is bilingual (German and Dutch). Additionally, the German version of the ZAT was checked in terms of correctness (correct use of words and word order) by multiple German native speakers.

Languages can differ in many ways. When translating between two languages, translation errors can occur due to gaps in knowledge or interference (the impact of one language on another). Differences between source text and target text are often caused by linguistic, cultural and situational differences. Some words cannot be translated literally, and a word can have several meanings, even in the same language (Table 2).

22 Table 2

Examples of language differences

Dutch Literal translation to German Other possible translation in German

duwen drücken stoßen, schubsen, vorschieben

bespioneren bespitzeln nachspionieren, hinterherspionieren

Therefore, language differences between the two versions were detected and analyzed. Especially when translating the verbs, the illustrations were helpful to avoid linguistic differences. As a last step, it was determined which items needed to be adjusted. After adjusting the items, the adaptation process was completed. The German version is now called Zinnen en Anagrammen Test-Duits (ZAT-D; see Appendix A).

3.2 Online Test 3.2.1 Participants

German-speaking individuals were approached via e-mail to participate in the study. General information about the study was given and the link to the online test was sent. In order to be included in the study each participant had to be a native speaker of German and between the age of 18 – 80 years. Further, everyone who was diagnosed with a language disorder, has hearing or vision problems and any kind of neurological impairment were excluded from the results. Table 3 shows the in- and exclusion criteria of this study.

Table 3

In- and exclusion criteria

Inclusion criteria Exclusion criteria

Native speaker of German Diagnosed with a language disorder Between the age of 20 – 80 Hearing and vision problems

Neurological impairments

In this study, N = 30 native speakers of German participated. All of them are people without a language disorder. The group consists of n = 15 females and n =15 males with a mean age of M = 39,03 years (SD = 16,782, Min = 21, Max = 75). Most of the participants are right-handed (n = 27), the other participants are either left-handed (n = 2) or ambidextrous (n = 1). An illustration of the level of education (highest obtained degree) of the participants can be seen

23

in Figure 5 below. The illustration shows that most of the participants hold a university/college degree. Individual demographic information of each participant can be found in Appendix C.

Figure 5. Highest obtained degree of participants

3.2.2 Material

Based on the WEZT (Bastiaanse et al., 2000), in the last years the ZAT has been developed by the Neurolinguistics group of the University of Groningen. In this research the only recently developed German subtest of the ZAT (ZAT-D, see Appendix A) will be used in form of an online test. The subtest sentence comprehension of the ZAT-D consists of 40 items and two practice items. Each sentence is represented in one of the following four conditions:

Order of elements

Active sentence structure:

a. A V T

Der Mann umarmt die Frau. The man hugs the woman.

Passive sentence structure:

b. T A V

Die Frau wird von dem Mann umarmt. The woman is hugged by the man.

Active sentence structure:

60% 20%

10%

7% 3%

Highest obtained degree

university/college degree professional degree high school diploma middle school lower secondary education

24

c. A T V

Zeige, auf welchem Bild der Mann die Frau umarmt. Show in which illustration the man hugs the woman.

Active sentence structure:

d. V A T

Auf welchem Bild umarmt der Mann die Frau. In which illustration the man hugs the woman.

Each target sentence appears in a selection of four illustrations. Within the four illustrations there is only one target picture. The other three illustrations are similar but not identical. A semantic distractor and illustrations with reversed agent and theme make the correct identification more difficult. An example is illustrated below in Figure 6:

Figure 6. Example of an item of the sentence comprehension task

As can be seen above, all actions were represented in black and white illustrations. Those were shown to the participants in a pre-programmed order (non-randomized). A complete overview of the items and the respective conditions can be found in Appendix A. An overview of the verbs and distractors can be found in Appendix D.

Before working with the data of the online study, a pre-study was done where the frequency data and the imageability data was compared statistically between the four conditions to ensure that there was no difference in these parameters between the sentence types.

Within the online study the verbs and distractors were balanced for frequency and imageability. The word frequency data was obtained with the help of the Datenbank für gesprochenenes Deutsch (DGD; Institut für Deutsche Sprache, 2020). This system allows access to selected corpora of the German spoken language. The same corpora that was used for the WAT-D (De Kok & Bastiaanse (in prep.-b)) were also used for the present study:

25

- Dialogstrukturen (DS)

- Forschungs- und Lehrkorpus gesprochenes Deutsch (FOLK) - Grundstrukturen: Freiburger Korpus (FR)

- Gesprochene Wissenschaftssprache Kontrastiv (GWSS) - Deutsche Hochlautung (HL)

- Deutsche Umgangssprachen: Pfeffer-Korpus (PF)

The data of the present study is measured in Log10. In the present study the range of word frequency was Min = 0.845 and Max = 3.009 (ordered from smallest low-frequency word to largest high-frequency word).

The imageability data was obtained from the Dutch version of the ZAT (De Kok & Bastiaanse (in prep.-a)). In the Dutch version of the ZAT imageability was rated on a 5-point Likert scale (1 = very easy to imagine, 5 = very difficult to imagine). However, there are some missing values within the imageability data. These data were not available and are therefore not in-cluded in the calculations. In the present study the range of word imageability was 1.27 – 2.9 (ordered from easy to imagine to difficult to imagine). The values of word frequency and im-ageability of each verb can be seen in Appendix D.

As described above, each item is presented in one of the four conditions (a, b, c, d). Each condition occurs ten times with ten different verbs. There are ten verb pairs which each occur four times. Each verb pair is matched on frequency meaning the difference in frequency be-tween the two verbs is < 1.0. Thereupon, some changes had to be made regarding the adaptation of the subtest (see Table 4).

26 Table 4

Adaptation of verb pairs based on frequency of the verbs

Version of test Verb pairs Frequency

Original Dutch version helpen duwen

3.19 2.54

Difference in frequency: < 0.1 1st German version helfen

schubsen

2.868 1.041

Difference in frequency: > 0.1 Adapted German version helfen

verfolgen

2.868 2.365

Difference in frequency: < 0.1

3.2.3 Procedure

The online study was created using PsychoPy (Peirde et al., 2019) an open source software program for science to design and run experiments online and carried out on a platform called Pavlovia (Pavlovia, 2019). It was made publicly available as a study entitled ‘Test for the Examination of Sentence Comprehension’.

The participants were able to take the test irregardless of location or time. Each participant was asked to take the online test on their own. First of all, demographic data such as age, gender, handedness and highest obtained degree were collected. The collection of demographic data was used for descriptive data of the sample. After filling in the demographic data, the procedure of the study was described. Further, the participants were informed that their data would be treated confidentially and anonymously. After the participants gave their consent to participate in the study, instructions for the sentence comprehension task followed. They were also reminded that the speakers of the laptop or computer must be turned on since the task required the participants to assign a heard sentence to an illustration. Furthermore, the participants were informed that mobile participation was not possible as some elements of the test would otherwise not be correctly illustrated. Each sentence appeared in a selection of four illustrations (multiple choice). Only one illustration could be selected by clicking on it. The participants were asked to answer as accurately and quickly as possible.

27

The task started with two practice items. After the practice items the participants received feedback whether the answer was correct or wrong. No feedback was given on the remaining items. Each participant was presented 20 items in two blocks. Between the two blocks the participant had the possibility to take a break and continue whenever the space key was pressed. The participants were able to stop the study at any time and without giving reasons. However, if someone discontinued the study within one of the blocks, the data was not collected, and neither was the participant counted. After the set of 40 items was completed, the participants were thanked for participation. The results were saved after the participants pressed the space key. Lastly, the participants were asked to close the browser only after a confirmation window had appeared.

3.2.4 Data analysis

The statistical analysis of the present study was carried out with a statistical data program called SPSS 26 (Statistical Package of the Social Science; Nie, Bent, & Hull, 1970).

3.2.4.1 Pre-study

Within the pre-study, the frequency data and imageability data of the used verbs were compared between the four conditions to ensure that there was no significant difference of these variables between the four sentence types. For the frequency data, a one-way ANOVA was performed. Frequency was defined as the dependent variable, whereby the four conditions (a, b, c, d) re-present the four-level factor of the ANOVA. In order to calculate the direction of the signifi-cant effects of ANOVA, corrected post-hoc tests were performed according to Bonferroni. Be-fore performing the analysis of variance, the normal distribution assumption of the dependent variable was tested using the Kolmogorov-Smirnov test. Each test was carried out with an α-value of α = .05. According to the recommendations of Field (2013), the normal distribution assumption was also assessed by descriptive observations of quantile-quantile plots. The Levene test is used to check whether the variances between the four conditions do not differ (= are homogeneous). This is required in order to run an ANOVA test. The calculation on image-ability was done with a Kruskal-Wallis test as the imageimage-ability data have not fulfilled the re-quirement to calculate an ANOVA.

3.2.4.2 Online study

First of all, it was checked whether there were items that were answered incorrectly by more than 80% of the participants. Furthermore, the RTs that were assigned to an incorrect answer were excluded from the data analysis.

28

For statistical analysis of hypothesis 1a the Pearson correlation between word frequency and the mean reaction time across all participants and items was calculated. Then the correlation coefficient was checked for statistical significance, that means it was tested whether the correlation coefficient in the sample significantly differs from r = 0.

For statistical analysis of hypothesis 1b the Pearson correlation between word imageability and the mean RT across all participants and items was calculated. Then the correlation coefficient was checked for statistical significance, that means it was tested whether the correlation coefficient in the sample significantly differs from r = 0.

To test hypotheses 2a and 2b a single factor ANOVA with repeated measurements was performed. The four conditions were defined as dependent factor, consisting of four levels (condition a, b, c and d). Correspondingly, RT was assigned as dependent variable. In the first step, the assumptions of the ANOVA with repeated measures were checked. Consequently, it was checked whether the dependent variable was normally distributed for each level of the within-subject factor. This was tested using the Kolmogorov-Smirnov test. In addition, it was tested whether the assumption of sphericity was met. This was checked with the Mauchly test. Subsequently, to determine between which conditions significant differences in mean values are present, Bonferroni post-hoc tests were performed.

3.2.4.3 Exploratory data analyses

In further exploratory data analyses, the relation between demographic factors (age, gender and highest obtained degree) and mean RT across all items were explored and tested for statistical significance. Age and RT were correlated with the help of the Pearson correlation coefficient and assessed for statistical significance. An independent samples t-test was conducted to compare whether there was a significant difference in the average RT between the two genders, male and female. Lastly, a one-way ANOVA was performed in order to see whether there were significant differences in RT for the highest degree obtained.

29

4

Results

4.1 Results of pre-study

The descriptive statistics of the frequency values can be seen in Table 5. The consideration of the quantile-quantile plot did not indicate any deviations from the normal distribution (Figure 7). The verification of the normal distribution of the frequency data using the Kolmogorov-Smirnov test yielded a non-significant result with p = .200 so that the assumption of a normal distribution can be assumed.

Figure 7. Quantile-quantile plot of frequency values

Table 5

Descriptive statistics of frequency values

Condition Number of items M SD Min Max

a 10 1.987 0.608 1.301 3.009

b 10 2.027 0.599 1.146 3.009

c 10 2.001 0.682 0.845 2.727

d 10 1.961 0.689 0.845 2.868

Total 40 1.994 0.621 0.845 3.009

30

The Levene test yielded a non-significant result with p = .962. This means that there is no significant difference in the variance between the four conditions. The homogeneity of variances can therefore be assumed. The one-way ANOVA showed no significant differences in frequency between the four conditions with F(3,36) = 0.018, p = .997.This means that the four groups do not differ in frequency on average.

The descriptive statistics of the imageability values can be seen in Table 6. The consideration of the quantile-quantile plot did indicate some deviations from the normal distribution (Figure 8). The verification of the normal distribution of the imageability data using the Kolmogorov-Smirnov test yielded a significant result with p < .001 so that the assumption of a normal distribution must be rejected.

Figure 8. Quantile-quantile plot of imageability values

Table 6

Descriptive statistics of imageability values

Condition Number of items M SD Min Max

a 10 1.896 0.542 1.27 2.90

b 9 1.768 0.445 1.27 32.50

c 9 1.968 0.549 1.27 2.82

d 10 1.976 0.582 1.36 2.90

Total 38 1.930 0.524 1.27 2.90

31

The Levene test yielded a non-significant result with p = .702. This means that there is no significant difference in the variance between the four conditions. Due to the results of the Kolmogorov-Smirnov test a Kruskal-Wallis test was computed, which showed no significant differences in imageability between the four conditions with 2_{(3) = 1.374, p = .712.Thus, the}

four groups do not differ from each other in terms of the imageability of the words.

4.2 Results of online study 4.2.1 Results of descriptive statistics

The RT was always measured in seconds. This applies to all further mentioned values of RT. The accuracy per participant, as well as the mean values of the RT per participant can be seen in Table 7 below.

Overall, there were 17 wrong answers out of 1200 answers meaning only approximately 1 % of all answers were wrong. As can be seen in Table 7, no item had to be excluded from the data analysis as all items were answered correctly by more than 80 % of the participants. The item with the most incorrect answers (Item 27: Dem Jungen wird von dem Mädchen gratuliert./The boy is congratulated by the girl.) was answered incorrectly by three participants (10 %).

32 Table 7

Descriptive statistics of accuracy and RT per participant

Participant Accuracy in % Mean RT in seconds SD Range

1 100 1.22 0.92 0.082 – 4.471 2 100 0.76 0.48 0.067 – 2.603 3 97.5 0.57 0.38 0.115 – 1.785 4 100 0.41 0.31 0.016 – 1.383 5 97.5 0.68 0.60 0.017 – 2.617 6 100 0.92 0.94 0.041 – 5.562 7 100 1.94 1.26 0.035 – 5.901 8 100 0.84 0.65 0.035 – 2.416 9 100 1.12 0.81 0.054 – 3.489 10 97.5 1.33 0.91 0.035 – 3.550 11 100 0.67 0.46 0.117 – 1.983 12 97.5 0.80 0.74 0.017 – 3.250 13 100 1.85 1.34 0.067 – 7.049 14 100 1.05 0.70 0.083 – 2.712 15 100 0.57 0.46 0.049 – 2.013 16 95 0.97 0.51 0.133 – 2.330 17 100 1.28 0.71 0.116 – 2.878 18 92.5 1.56 1.15 0.133 – 6.100 19 97.5 0.51 0.33 0.017 – 1.333 20 92.5 2.23 0.92 0.400 – 4.398 21 97.5 0.70 0.50 0.033 – 2.400 22 95 0.96 0.79 0.016 – 3.168 23 97.5 0.60 0.47 0.031 – 1.979 24 100 0.83 0.80 0.044 – 3.870 25 97.5 0.56 0.34 0.084 – 1.536 26 100 0.77 0.80 0.022 – 3.781 27 100 0.93 1.07 0.167 – 6.272 28 95 1.50 0.71 0.139 – 3.020 29 100 1.36 1.14 0.067 – 6.416 30 97.5 0.78 0.49 0.017 – 1.950