Search

The study has two goals. First, it tests whether the usage of gapping is subject to the information-theoretic processing constraint of Uniform Information Density (UID, Levy and Jaeger, 2007). UID predicts that gapping is more strongly preferred and easier to process the more predictable the potentially omitted verb is. Second, it serves the purpose of norming stimuli for follow-up acceptability rating and self-paced reading experiments. We measure and aim at manipulating the surprisal (-log 2 p(verb|context)) of the verb in the second conjunct of clausal coordinations like (1) given its subject and the first conjunct of the coordination. (1) The boxer beat up an opponent and the (hooligan|hippie)______. (Note that in German, unlike in English, the example is not ambiguous due to the case marking at the subject) We manipulate the surprisal of the target verb through the typicality of the second conjunct's subject as a subject of this verb (it is more typical for a hooligan to beat up people than for a hippie). In our study, we present the participants with German clausal coordinations equivalent to (1) that are cut off after the subject of the second conjunct. In a one-factor, two-level design, we will manipulate the typicality of the second conjunct's subject (prototypical/non-prototypical) in relation to the verb of the first conjunct. We expect that participants produce the same verb as the one in the first conjunct more often when the subject of the second conjunct is prototypical with regard to the verb. We also expect that the ratio of gapping produced by participants increases as a function of the verb's predictability. The typicality manipulation is based on our intuitions and the experiment allows us to investigate whether these intuitions are empirically supported.

The goal of the experiment is to explore whether the information-theoretic concept of Uniform Information Density (UID, Levy and Jaeger, 2007) explains the acceptability of gapping. According to the Uniform Information Density (UID) principle, speakers tend to spread information as evenly as possible across the linguistic signal. As a result, speakers try to avoid regions of low information by omitting predictable material from the utterance, i.e. by using an ellipsis. With respect to gapping, this predicts that listeners particularly prefer a gapped coordination when the verb of the second conjunct is predictable. We will test this prediction with an acceptability rating study. In a 2x2 design, we will cross the surprisal of the verb (high/low) with the form of the sentence (nonelliptical/gapping). We will manipulate the surprisal of the verb given the subject noun phrase though world knowledge, more specifically - through the typicallity of the subject. We expect that gapping is rated as relatively more acceptable (compared to the nonelliptical form) when the verb of the second conjunct is predictable. A production study ("Verb predictability effects on the usage of gapping (production study)", preregistered separately on OSF) will be used to measure the surprisal of the verb given the preceding context, which is used as a predictor in our analyses.

This is a follow-up acceptability rating study to "Coordination ellipsis, modality and memory (acceptability rating) - Exp. 1" (also preregistered). Experiment 1 investigates the hypothesis that the relatively stronger preference for coordination ellipsis (and particularly right node raising) in written corpus data as compared to spoken corpora (Harbusch 2011) is explained by the possibility of backtracking in written language. If this is expectation is confirmed, we will investigate in experiment 2 whether processing coordination ellipsis is indeed inhibited by memory limitations, which are counterbalanced by backtracking in the written modality. To this end, we will explicitly manipulate the memory load of our participants in a replication of the oral conditions of experiment 1: Half of our participants will be asked to perform mathematical calculations between the trials and to memorize the result in order to later (after 1 to 3 trials) enter it into a text field. We expect that participants whose memory is partly used for this additional task will prefer more explicit, i.e. non-elliptical, structures more strongly than participants who do not need to memorize the results. We expect that this effect is particularly strong for right node raising, since this construction requires comprehenders to keep the incomplete first conjunct in their working memory until the expression filling the gap is encountered.

The experiment investigates the interpretation of German plural masculine nouns referring to a group of people (e.g. the actors, the teachers) in an experimental paradigm based on Gygax et al. (2008) and Körner et al. (2022). Based on binary sentence continuation judgment tasks these studies conclude that such masculine nouns are more often interpreted as referring only to male referents than generically, that is, possibly including females. Our experiment addresses a possible design flaw of the studies by Gygax et al. (2008) and Körner et al. (2022) which consists in a purely written presentation of the stimuli which does not control for prosody. We present similar stimuli auditorily in two conditions, one of which (unmarked intonation pattern) suggests that the group in the target utterance is only male or female, whereas we expect that the other one (contrastive topic intonation) indicates a gender-mixed interpretation of the masculine nouns. This will reveal to what extent previous findings depend on the unmarked intonation pattern, which is assumed to be the default prosodic structure in silent reading (Féry 2005). The main experiment will be followed by a questionnaire to determine the participants' attitude toward gender-neutral language and their explicit interpretation of the generic masculine forms.

Die Partizipation von Interessengruppen in Gesetzgebungsprozessen der Europäischen Union (EU) ist von gegenseitiger Abhängigkeit im Sinne eines politischen Tauschs geprägt: Die gesetz-gebenden Institutionen benötigen Fachinformationen zur politischen Entscheidungsfindung. Inte-ressengruppen liefern Informationen und erhoffen sich dadurch Einfluss auf das Politikergebnis. Angesichts der Flut an konkurrierenden Akteuren, die auf diesen diskursiven Tausch einwirken, beschäftigt sich diese Arbeit mit der Frage, welche Faktoren politische Einflussnahme begünsti-gen. Dabei liegt die Annahme zugrunde, dass Interessengruppen versuchen, vor allem diskursi-ve Macht zu mobilisieren, um gesetzgebende Akteure zu überzeugen. Ihr Erfolg variiert je nach-dem, ob sie in der Lage sind, Koalitionen zu schmieden, politische Ansprechpartner zu finden und Informationen bereitzustellen. Um ihren Botschaften dabei mehr politisches Profil zu geben, nutzen sie Framing. Am Beispiel der Revision der EU-Tabakrichtlinie von 2012-2014 befasst sich diese Arbeit mit dem analytischen Potenzial der Erfolgsfaktoren politischer Einflussnahme in einem dynamischen Setting. Dazu wurde der Gesetzgebungsprozess in vier Analysephasen unterteilt und metho-disch mithilfe einer Kombination aus Prozessanalyse und dem Abgleich von Framing zwischen Interessengruppen und Institutionen, kurz Framing-Analyse, aufbereitet. Die Tabakrichtlinie polarisierte den politischen Diskurs und gilt heute als exemplarischer Kasus für intensives Lobbying, bei dem die Interessen von Tabakindustrie und Verbraucherschutz-Gruppen aufeinanderprallten. Die ausgezeichnete Quellenlage bereitete die Basis für die Anwen-dung der Methodenkombination, die zu teils überraschenden Erkenntnissen kommt. Erstens erwiesen sich formelle Koalitionen für erfolgreiche Interessenvertretung im vorliegenden Fall als weniger bedeutsam als angenommen. Wichtiger sind einheitliche Botschaften, die von einer möglichst heterogenen Gruppe aus Akteuren reproduziert werden. Zweitens sind möglichst exklusive Kontakte zu politischen Schlüsselfiguren von großem Vorteil. Allerdings versuchen Interessenvertreter auch über die Ansprache fachfremder politischer Akteure innerhalb der Insti-tutionen Einfluss auszuüben und innerinstitutionelle Opposition zu generieren. Diese Strategie führte im untersuchten Fall zum Erfolg – ausgerechnet das Europäische Parlament schwächte den Gesetzentwurf der Kommission an entscheidenden Stellen ab. Drittens begünstigt die Fä-higkeit, relevante Informationen bereitzustellen, die Chancen auf Einflussnahme. Interessen-gruppen nutzen Framing dabei gezielt zur Zuspitzung und Legitimierung ihrer Forderungen sowie zur Abgrenzung von konkurrierenden Interessen. Die Kombination aus Prozess- und Framing-Analyse erlaubt im Gegensatz zu vielen Large-N Studien eine Fokussierung auf mehrere Variablen, die als Erklärungsfaktoren politischer Ein-flussnahme dienen, und trägt gleichzeitig dazu bei, eine Triangulation der Befunde zu gewährleis-ten. Angesichts der zunehmenden Partizipation organisierter Interessen an EU-Gesetzgebungsprozessen, leistet diese Arbeit einen relevanten Beitrag zu einem dynamischen Verständnis von Interessenvertretung und bietet eine neue methodische Perspektive zur weite-ren Erforschung von Interessenvertretung in der EU.

We will conduct an acceptability rating study to test whether ellipsis, in particular right node raising, but also gapping, is rated as more acceptable (as compared to the nonelliptical utterance, in what follows "full form", as baseline) in written than in oral modality due to memory considerations. The idea is that while in the written modality the addressee can reread parts of the utterance, i.e., perform backtracking, this is not possible in the oral modality where the addressee hears an utterance exactly once. Therefore, more explicit full forms might be more strongly preferred over ellipsis in oral modality. This should be especially true for RNR where hearers have to keep the incomplete first conjunct in memory until encountering the missing material and completing its processing. We will compare not only oral and written modality, but also include a written, but timed word-by-word presentation (timed written). This will show whether the expected difference between written and oral modality is due to the availability of backtracking (as we expect) or to other differences between the modalities.

Multi-class datasets often comprise samples of various classes united under a common research context. Based on this limited amount of data one searches for concepts that employ specific data characteristics to define general rules that can then be applied to new data. One concept of interest is the relation between classes. Hereby, the classes correspond to a grouping and labelling of samples, and the samples are given by the measurements. These relations might be of interest in basic research to analyse development or relationships between organisms or phenotypes. But they are also relevant in medical diagnostics as they might guide treatment options or serve as basis for knowledge transfer. If solely one feature is measured, an ordinal relation might be concluded directly. For example, in histological examinations of pre-cancer stage tissues only the grade of observable tissue disruption with respect to the normal stage is measured, and a semantic order on pre-cancer stages is proposed based on this measure. If in the following a molecular profile is gathered additionally, an order between classes cannot easily be defined due to the quantity of different features but the previously proposed order might be assumed to be reflected in the molecular profile. This assumption, however, has not been proven and requires further analysis. Transferring semantic ordinal relations between measurements or searching for them in high-dimensional data goes along with the following questions: • How to check whether a proposed linear order is reflected in the given measurements, how to reveal linear orders from them and how to combine several linear ordinal relations to one order structure? • Which data structures should be considered as ordinal in high-dimensional space? • How can the revealed ordinal relations be interpreted? In this thesis, I target the first and second question by the functionality of the ordinal classifier cascade. The third question is targeted by the application of ordinal classifier cascades to various real data examples, among them pre-cancer stages of different cancer types.

Background: A growing body of research demonstrated impaired executive functions in individuals with severe obesity, including increased sensitivity to reward and impulsive decision making under risk conditions. For the assessment of decision making in patients with severe obesity, studies widely used the Iowa Gambling Task (IGT) or the Delay Discounting Task (DDT), which cover short-term or long-term consequences of decisions only. A further development originating from the field of addiction research is the Cards and Lottery Task (CLT), in which each decision made has conflicting immediate and long-term consequences at the same time. The present study aimed to validate the CLT in individuals with severe obesity. Methods: Patients with severe obesity (N = 78, 67% women, 42.9 ± 10.4 years old, body mass index of 48.1 ± 8.3 kg/m²) were included. Convergent validity was evaluated using the computerized Delay Discounting Task and well-established self-report questionnaires assessing different aspects of impulsivity. For discriminant validity, CLT performance was compared between symptom groups characterized by high versus low impulsivity. The task’s clinical validity was evaluated based on associations with general and eating disorder psychopathology, and body mass index. Test-retest reliability was determined by administering the CLT in n = 31 participants without weight-loss treatment one year later. The task’s sensitivity to change due to weight loss was evaluated by retesting n = 32 patients one year after receiving obesity surgery. Results: The number of advantageous decisions in the CLT was significantly positively associated with delay discounting and effortful control, and significantly negatively correlated with behavioral impulsivity. CLT performance differed significantly between individuals with and without symptoms of attention-deficit/hyperactivity disorder and between samples with severe obesity and healthy controls. Clinically, CLT performance was significantly associated with general, but not eating disorder psychopathology. The CLT showed moderate test-retest reliability after one year in weight-stable individuals and was sensitive to change in those undergoing obesity surgery. Conclusions: This study identified the CLT to be a highly promising, new complex measure of short- and long-term decision making with good reliability and validity in individuals with severe obesity. Future studies should assess its association with the IGT and predictive value for real-life health behavior. CA extern