Twenty-five healthy subjects participated in Experiment 2a (22 female, mean age 23.96 years, range 19–38) who were either students from University Marburg, Germany, or people from Marburg area. All of them were monolingually raised, native speakers of German and had normal or corrected-to-normal vision. All participants were naïve con- cerning the purpose of the experiment and received about 10€ for their participation. 6.2.1.2 Combined fMRI / eye-tracking experiment (Experiment 2b)
Twenty-two healthy students (13 female, mean age 23.63 years, range 19–31) of the Uni- versity of Marburg, Germany participated in Experiment 2b. One additional subject was excluded due to unusably long reading times that did not allow for the experiment to be completed. All participants were right-handed (indicated by an adapted German version of the Edinburgh Handedness Inventory;Oldfield, 1971), monolingually raised native speakers of German and reported neither impairments with regard to reading and writ- ing ability nor neurological, psychiatric or other disorders. All participants met the safety and participation requirements and gave written informed consent prior to participa- tion. They were naïve concerning the purpose of the experiment and received 20€ for their participation. The study procedure conformed to the Declaration of Helsinki and was approved by the ethics committee of the Medical Faculty at the University of Mar- burg. None of the subjects had participated in the single eye-tracking experiment.
6.2.2 Materials
The sentence material used here was the same for both Experiments 2a and 2b. Since we were (besides the neural underpinnings of regressive vs. progressive saccades) also in- terested in the interaction of syntax and semantic processing during sentence reading,
123 we created German-counterparts of the English SRA sentences used in Experiment 1. This was done to allow for a direct comparison of Experiment 1 and 2 in two different languages and to have an additional testing tool, especially in view of the reliability of the combined fMRT / eye-tracking experiment.
Table 7: Example stimuli of the German SRA sentences used in Experiment 2. Sentence regions for the region- based eye-tracking analysis are marked by the vertical strokes.
Semantic Reversal Anomalies (SRA) – German
a) Der Lehrer | korrigiert | den Fehler | nach zwanzig Berufsjahren schon | fast automatisch. The teacher(NOM) corrects the mistake(ACC) after twenty years of experience almost automatically.
non anomalous, high associated b) Der Lehrer | errät | den Fehler | nach zwanzig Berufsjahren schon | fast automatisch.
The teacher(NOM) guesses the mistake(ACC) after twenty years of experience almost automatically.
non anomalous, low associated c) Der Fehler | korrigiert | den Lehrer | nach zwanzig Berufsjahren schon | fast automatisch.
The mistake(NOM) corrects the teacher(ACC) after twenty years of experience almost automatically.
anomalous, high associated d) Der Lehrer | errät | den Fehler | nach zwanzig Berufsjahren schon | fast automatisch.
The mistake(NOM) guesses the teacher(ACC) after twenty years of experience almost automatically.
anomalous, low associated
An example of the SRA sentences employed is given in Table 7. Because it is not possible in German that two constituents precede the main verb in standard declarative sentences (in contrast to the English sentence structures), in all of our SRA sentences only one argument (which always was the subject) preceded the verb. Although an ob- ject-first structure is possible in German, all arguments were unambiguously case- marked on the masculine determiner “der” (nominative) or “den” (accusative), respec- tively, so that they could be unequivocally identified as the subject or object, respectively.
As in the English counterparts, one of the two arguments was animate (e.g. “Leh- rer” [teacher]) and the other was inanimate (e.g., “Fehler” [mistake]). When the inani- mate argument was used as the subject, this resulted in implausibility (“Der Fehler kor- rigiert den Lehrer” [the mistake corrects the teacher]). In addition, we manipulated the degree of association between the verb and the inanimate object (high: “Fehler – korri- gieren” [mistake – to correct] vs. low: “Fehler – erraten” [mistake – to guess]), leading to a 2 x 2 factorial design (ANOMALY x ASSOCIATION). In total, we created 72 experimental stimulus sets of four stimulus versions each (all stimulus sentences can be found in the Appendix).
In an online pre-test, we asked a separate group of participants to rate the asso- ciation between pairs of the inanimate noun and verb on a 5-point scale (1 – very high
124 associated, 5 – not associated) and the acceptability of the sentences, again on a 5-point scale (1 – not acceptable, 5 – very acceptable). The word pairs and sentences were as- signed to four lists separately and each list was rated by 18 people (for the association judgement) and 20 people (for the acceptability judgement). The analysis revealed that the word pairs in the low associated condition were significantly lower associated than in the high associated condition (mean verbs low: 3.70, mean verbs high: 1.44; t(74.40) = -24.93, p<.001). Also, the non-anomalous sentences were significantly more accept- able than their anomalous counterparts (mean anomalous sentences: 1.27, mean non- anomalous sentences: 4.49, t(202) = 57.93, p<.001).
There were no differences between the word length of the high and low associ- ated verbs (mean verbs high: 6.49 characters, mean verbs low: 6.71 characters; t(141.76) = -0.71, p = .48) or between the animate and inanimate nouns (mean animate nouns: 6.83 characters, mean inanimate nouns: 6.42 characters; t(136.77)=1.20, p = .23), neither were there differences between the frequency of the two verb types (fre- quency based on the frequency classes provided by the Wortschatz Leipzig; mean verbs high: 11.31, mean verbs low: 11.21; t(140.96) = 0.21, p = .83) or noun types (mean an- imate nouns: 11.72, mean inanimate nouns: 11.29; t(140.55) = 1.16, p = .25).
The sentences of the 72 stimulus sets were allocated to 4 experimental lists ac- cording to a Latin Square design so that each subject read 72 SRA sentences. In addition, we created 144 filler sentences, containing anomalous and non-anomalous sentence ma- terial, that were used to prevent the readers from developing certain reading strategies. Additionally, some of these sentences were constructed for another experimental ma- nipulation which is of no interest for the present purpose. 40 of the filler sentences were followed by a simple YES / NO comprehension question to ensure that the subject read the sentences carefully. The 72 stimulus sentences were randomly intermixed with the 144 filler sentences, so that each subject read 216 sentences in total (60 anomalous). All sentences were syntactically well-formed German sentences.
6.2.3 Procedure
6.2.3.1 Single eye-tracking experiment (Experiment 2a)
After reading the instructions and giving written consent, the participants were seated in front of a computer screen. An SR Research EyeLink 1000 tower mount tracker with a sampling rate of 1000Hz was used to record the participants’ eye movements during
125 reading. Viewing was binocular but only the right eye was tracked. All sentences were displayed on one line in 20 point Monaco font on a 22’’ monitor with a refreshing rate of 60Hz that was positioned 80 cm from the participant.
Participants were instructed to read as normally as possible, “as they would read the newspaper” and were also told that some of the items might be “a little weird”. Com- prehension questions were presented on the screen and participants responded to them by pressing one of two buttons on a gamepad. Before the experiment began, a 9 point calibration and a short practice session of 6 practice trials was performed. The experi- ment took about 40 minutes in total.