In this thesis, I investigate the semantic processing of the differential non-literalness in idioms. Specifically, the question is whether the common non-literal phenomena of metonymy and metaphor are processed in idioms, given that idioms are often metonymic or metaphoric (but can also be literal). So far, there is empirical research on the semantic processing of idioms on the one hand, and on the semantic processing of metaphors and metonymies on the other. However, the findings of these separate strands of research have not been brought together. In addition, the property of non-literalness and its impact on semantic processing has not been investigated in any complexity yet. To partially bridge this gap, I conducted one rating study to test the hypothesis that metonymic idioms are perceived as more literal than metaphoric idioms, and to explore how non-literalness is related to other idiom-specific properties. In the next step, I conducted several experiments to examine whether there is any difference in how metonymic and metaphoric idioms are semantically processed. Where possible, literal idioms are also included as non-literalness effects can be understood best if investigated in comparison to the absence of non-literalness. Chapters 2-4 outlined below are individually published or accepted for publication in scientific journals. In chapter 2, I present the large rating study which consists of four separate questionnaires for separate groups of participants. Each idiom was rated on its degree of non-literalness, familiarity, comprehensibility, and on how closely related its literal and its idiomatic meaning were. Most importantly, the rating study reveals that metonymic idioms are perceived as considerably more literal than metaphoric idioms. In chapter 3, I explore and discuss whether these differences in non-literalness also affect the difficulty of reading sentences which contain literal, metonymic, metaphoric, or no idioms. In this manner, highly non-literal (metaphoric) idioms can be compared to less non-literal (metonymic) idioms and completely literal idioms. The findings show that literal and metonymic idioms are read faster than metaphoric idioms, independent of whether the idioms were semantically expected from the sentence or not. Metaphoric idioms were only read faster than sentences without idioms if they were expected. Overall, all idioms are easier to read when expected than sentences without idioms, but only the idioms with higher literalness are also read faster when they are not expected. Thus, non-literalness has a partial effect: It can be harder to process depending on context. In chapter 4, I discuss whether the processing system automatically differentiates between (rather literal) metonymic and (rather non-literal) metaphoric idioms, that is, from the first milliseconds upon having heard them. A primed lexical decision experiment was conducted twice to this aim. Again, the reaction speed to stimuli is measured in this paradigm. If the reaction speed metaphoric idioms were slower, it would mean that higher non-literalness is more difficult even in automatic (i.e. initial) processing. However, no such effect could be found in either version of the experiment. In fact, there is actual evidence that different non-literalness does not or only minimally affect automatic processing. Thus, non-literalness effects are likely not essential to recognition and automatic processing but come into play at a later time frame of the processing course. In chapter 5, I compare the experimental findings and discuss them in the framework of selected (idiom) processing models. Findings fit with the Graded Salience Hypothesis and partly with the Configuration Hypothesis and the Hybrid Model. Moreover, it seems that depending on the experimental task, there is a trade-off between non-literalness and idiomaticity. Chapter 5 also discusses generalizability of the results and gives incentives of how to further pursue the question of how non-literalness affects the processing of idioms.
