The semantic relation between a verb and its argument rapidly impacts language comprehension (Altmann & Kamide, 1999; McRae et al., 1998) and so does world knowledge and linguistic context (e.g., Altmann & Steedmann, 1988; Kutas & Hillyard, 1984). Situation formality represents another facet of world knowledge that could incrementally modulate comprehension, and interact (or not) with a verb’s thematic requirements. In two self-paced reading experiments with cumulative, phrase-by-phrase presentation, we examined the interplay of register-context (mis)matches, with (mis)matches in verb-argument relations. Additionally, a secondary picture-selection task was used in order to explore to which extent comprehenders may be able to differentiate between depicted register variants. We further assessed whether register effects emerge when situation formality varied from one trial to the next or only when it was blocked.

Design: Both experiments employed a 2x2 design (register-context congruence x verb- argument relation congruence). The register-context factor was realized via a sentence introducing a situation-formality context that (mis)matched the register of the target sentence. The verb-argument relation manipulation was embedded in the target sentence (match: tie shoes vs. mismatch: tie clothes, Table 1). While in Exp. 1 the items were blocked by formality level, in Exp. 2 they were fully pseudo-randomized. We expected a main effect of verb-argument (in)congruence, with increased reading times for incongruent(vs. congruent) verb-argument relations in the critical or spillover regions, or for the entire sentence, longer reaction times and decreased accuracy. Additionally, a main effect of register-context (in)congruence was expected (increased reading and reaction times, as well as decreased picture selection accuracy for register-context mismatches than matches). An interaction of the two factors was also expected, whereby (in)congruent verb-argument relations could impact the comprehension patterns of (mis)matching register-context relations.

Analysis: Log transformed reading times for the spillover and critical regions, total sentence reading times and reaction times were analyzed using linear mixed models. Accuracy resulting from picture selections was analyzed with binomial generalized mixed-effects models.

Results: Verb-argument relation (in)congruency significantly impacted sentence reading in Exp.1 ( = -0.018, SE = 0.007, df = 7655.236, t = -2.502, p < 0.05) and Exp 2. (= −0.029, SE = 0.011, df = 39.176, t = −2.591, p < 0.05). The same pattern was observed for the spillover region (Exp. 1: = -0.130, SE = 0.030, df = 552.306, t = -4.311, p < 0.001, Exp. 2: = -0.144, SE = 0.021, df = 44.477, t = -6.845, p < 0.0001). Moreover, the verb-argument effect was significant for picture selection latencies and accuracy. No significant main effect of register emerged, but participants were less accurate (n.s.) in choosing correct target pictures for register-context mismatches vs. matches (Exp. 1 mean: 42 % vs. 64%, Exp. 2 mean: 38% vs. 62%). Analyses further revealed a significant interaction of register-context congruence with verb-argument relation congruence: Verb-argument mismatches elicited longer reading times than matches in register-context matches, but not mismatches (in the blocked version only).

Discussion: These analyses revealed effects of verb-argument relation congruence. While register-context congruence effects were absent in reading times, the accuracy results (n.s.) point towards the fact that the processing of social-contextual information might impact late sentence processing. The interaction of the two factors in reading times suggested that the match of a context and target sentence in formality may have served as a “filter” for further grammatical processing of verb-argument relation congruence, i.e., when participants were habituated to the formality of a context (blocked presentation), they were more sensitive to verb-argument relation mismatches.