Truth conditions of sentences ascribing non-doxastic propositional attitudes seem to depend on the information structure of the embedded clause. In this paper, we argue that this kind of sensitivity is a semantic phenomenon rather than a pragmatic one. We report four questionnaire studies which explore the impact of the information structure on the truth conditions of non-doxastic attitude ascriptions from different perspectives. The results of the first two studies show that the acceptability of those ascriptions can be affected by some structural modifications of the embedded clause, in particular, when we replace a simple form by an equivalent complex conjunctional form (‘p and q’). However, it is possible that different evaluations of such ascriptions have a pragmatic source, namely, the ascriptions with embedded conjunction imply that the agent’s attitude transfers to both conjuncts. In the second pair of studies, we further investigate the nature of this implication which can be classified as ‘Conjunction Elimination’ (CE) in the scope of an attitude verb. The results show that CE-inferences in the context of non-factive non-doxastic attitude ascriptions are not easily cancellable and hence of a semantic rather than pragmatic nature. The results are not conclusive when it comes to the factive non-doxastic attitudes. We conclude our findings by some considerations about a potential source of the observed difference between non-factive and factive attitude verbs and the significance of our general findings to the semantic theory of non-doxastic attitude ascriptions. This repository contains the files related to the studies reported in the article. * `Study1a` - stimuli, data, analyses concerning Study 1a * `analysis/analysis_study_1a_RPP.Rmd` - code used to generate the report * `analysis/analysis_study_1a_RPP.html` - complied report * `analysis/data_study3.csv` - experimental data * `analysis/plot_study_1a.{png|tiff}` - plots included in the article * `stimuli/study1a.xlsx` - stimuli used in the experiment * `Study1b` - stimuli, data, analyses concerning Study 1b * `analysis/analysis_study_1b_RPP.Rmd` - code used to generate the report * `analysis/analysis_study_1b_RPP.html` - complied report * `analysis/data_study3.csv` - experimental data * `analysis/plot_study_1b.{png|tiff}` - plots included in the article * `analysis/prolific_export_....csv` - demographic data * `stimuli/study1b.xlsx` - stimuli used in the experiment * `Study2a` - stimuli, data, analyses concerning Study 2a * `analysis/analysis_study_2a_RPP.Rmd` - code used to generate the report * `analysis/analysis_study_2a_RPP.html` - complied report * `analysis/results_all.csv` - experimental data * `analysis/plot_study2a.{png|tiff}` - plots included in the article * `analysis/prolific_export_....csv` - demographic data * `stimuli/study2a.xlsx` - stimuli used in the experiment * `Study2b` - stimuli, data, analyses concerning Study 2b * `analysis/analysis_study_2b_RPP.Rmd` - code used to generate the report * `analysis/analysis_study_2b_RPP.html` - complied report * `analysis/results_all.csv` - experimental data * `analysis/plot_study2b.{png|tiff}` - plots included in the article * `analysis/prolific_export_....csv` - demographic data * `stimuli/study2b.xlsx` - stimuli used in the experiment The analyses were performed using R and are contained in RMarkdown files (`Rmd`). Data used for analyses is included in each directory that also contains the code. Tables and figures that appear in the article are marked as such in the RMarkdown files.