Experimental design and rationale of the study.

Panels A and B illustrate the contrast between functor-initial and functor-final word order in Spanish and Basque, as well as its consequences on their prosodic structure. Panels C and D show the design of the experimental and control conditions, respectively. The structure of the design is the same in both conditions (ababab), with 30 s sequences of two tones alternating at fixed intervals and occasional omissions. In the experimental condition, tones alternate in duration but not frequency, whereas in the control condition tones alternate in frequency but not duration. Panels E and F depict the hypothesized error responses associated with the different types of omissions for experimental and control conditions, respectively. Round brackets above the tones reflect the grouping bias of the two languages, based on their word order constraints. Dotted lines reflect short-term predictions based on the transition probabilities of the previous stimuli. Solid lines reflect long-term predictions based on the phrasal chunking scheme of the two languages.

Sensor-level topography and time course of neural responses to omitted sounds across groups and conditions.

Panel A shows the temporal unfolding and topographical distribution of the overall effect of omission (omissions minus tones). Channels belonging to the significant cluster are highlighted. Panel B shows the ERF generated by omissions and tones in a representative channel. Panel C (left) shows the topography of the t distribution of the interaction effect between the language background of the participants (Spanish, Basque) and the type of omission MMN (short, long). Channels belonging to the significant interaction cluster are highlighted. The interaction effect was present only in the experimental condition. Panel C (right) shows the averaged MEG activity over the 0.100 – 0.250 s time window and channels belonging to the significant cluster for each group and condition separately. Panels D, E, F, G show the effect of language experience in modulating the amplitude of the omission MMN associated with each experimental and control contrast. Topographies (top) show the scalp distribution of the averaged activity over the 0.100 – 0.250 s time window. Channels belonging to the significant interaction cluster are highlighted. ERFs (middle) show the temporal unfolding of brain activity averaged over the channels belonging to the significant interaction cluster for each contrast and group. The shaded area indicates the time window of interest for the statistical analysis. Box-plots (down) show the mean MEG activity for each participant over the 0.100 – 0.250 s time window and the channels belonging to the significant interaction cluster. The center of the boxplot indicates the median, and the limits of the box define the interquartile range (IQR = middle 50% of the data). The notches indicate the 95% confidence interval around the median. Dots reflect individual subjects. In D–G, asterisks indicate statistical significance for each contrast using a one-sided, independent sample t-test with FDR correction for multiple comparisons [statistical significance: **p < 0.01, *p < 0.05, and ns p > 0.05, respectively.

Source activity underlying the omission network and long-term predictions.

Panel A shows brain maps representing source activity underlying the omission response network over the 0.100 – 0.250 s time window. Panels B, C and D show source activity associated with the omission MMN responses in the left auditory cortex (left Helsch’s gyrus and left superior temporal gyrus), right auditory cortex (right Helsch’s gyrus and right superior temporal gyrus), and the left inferior frontal gyrus (pars triangularis and pars opercularis), respectively. The shaded grey area shows the 0.100 – 0.250 s time window, which was selected for the statistical analysis of the sensor-level data.

Main effects of omission type and language background.

Panel A shows the ERFs and topographies reflecting the main effect of omission type, with omissions of long auditory events generating larger omission MMN than short events between groups. Panel B shows the ERFs and topographies reflecting the (lack of) main effect of language background, with overall no group differences in the amplitude of the omission MMN. Because no main effect of language background was detected, Panel B uses the same channels of Panel A as representative channels for plotting the ERF. The remaining conventions for the plot are the same as in Figure 2.