The Role of Simple Semantics in the Process of Artificial Grammar Learning

Abstract

This study investigated the effect of semantic information on artificial grammar learning (AGL). Recursive grammars of different complexity levels (regular language, mirror language, copy language) were investigated in a series of AGL experiments. In the with-semantics condition, participants acquired semantic information prior to the AGL experiment; in the without-semantics control condition, participants did not receive semantic information. It was hypothesized that semantics would generally facilitate grammar acquisition and that the learning benefit in the with-semantics conditions would increase with increasing grammar complexity. Experiment 1 showed learning effects for all grammars but no performance difference between conditions. Experiment 2 replicated the absence of a semantic benefit for all grammars even though semantic information was more prominent during grammar acquisition as compared to Experiment 1. Thus, we did not find evidence for the idea that semantics facilitates grammar acquisition, which seems to support the view of an independent syntactic processing component.

Formal language hierarchy reflects different levels of cognitive complexity.

“…Strengthening this possibility, de Vries et al (2012) found advantages in reaction times for nested grammars in German-speaking (nested-based language) participants when compared to Dutch ones. Further studies with German participants (Öttl et al, 2015, 2017) showed a different pattern in that nested and crossed dependencies were equally difficult to learn and process. However, Öttl and colleagues did not use an additional baseline classification test before training, thus leaving open a complete characterization of learning outcomes and subsequent considerations on the role of linguistic experience.…”

“…Overcoming the limitations of studies with Dutch participants, we recruited Portuguese-speaking participants, who have little or no experience with crossed dependencies between different words or constituents (but please see Martins, 2006, and Piechnik, 2015, for reduplication phenomena in Portuguese). Overcoming the limitations of Öttl and colleagues’ (2015, 2017) studies with German participants (also unexperienced with crossed dependencies), we used a learning approach that included pre- and posttraining preference classification tests. If nested dependencies proved to be easier to learn than crossed ones, this would strengthen the idea that formal language hierarchy is compatible with the cognitive architecture of humans.…”

“…In humans, evidence that adjacent and nonadjacent dependencies recruit different brain substrates has gained strength (Calmus et al, 2020;Friederici et al, 2006;Uddén et al, 2020). Nonadjacency has been shown to be acquired later than adjacency in development (Gómez & Maye, 2005;Teinonen et al, 2009), and direct comparisons between adjacent and nonadjacent dependencies showed increased learning for adjacent ones (Friederici et al, 2006;Öttl et al, 2017). One detour from this scenario was made by Uddén and colleagues (Uddén et al, 2009(Uddén et al, , 2012, who showed that multiple nonadjacent dependencies could be learned as efficiently as adjacent dependencies, provided that participants were given more time to learn (9 days instead of 5; see also Uddén et al, 2017).…”