EEG signatures of elementary composition: Disentangling genuine composition and expectancy processes

•We study linguistic composition in Spanish adapting the simple composition paradigm of Bemis and Pylkkanen in 2011.•Using EEG and applying an adapted cluster permutation procedure, we reproduce the main results obtained using MEG.•Crucially, we found that the effect changes systematically from the first to the last trial pointing to a learning process.•We tested the alternative hypothesis that the EEG activity is related to general expectancy and not to composition.•This shows that in order to reveal a composition related activity, general expectancy effects have to be eliminated. We adapted Bemis & Pylkkänen’s (2011) paradigm to study elementary composition in Spanish using electroencephalography, to determine if EEG is sensitive enough to detect a composition-related activity and analyze whether the expectancy of participants to compose contributes to this signal. We found relevant activity at the expected channels and times, and a putative composition-related activity before the second word onset. Using threshold-free cluster permutation analysis and linear models we show a task-progression effect for the composition task that is not present for the list task. In a second experiment we evaluate two-word composition incorporating all conditions in a single task. In this case, we failed to find any significant composition-related activity suggesting that the activity measured with EEG may be in part carried by expectancy processes arising from the block design of the experiment, which can be prevented by using a non-blocked design and data-driven techniques to analyze the data.