ERP markers of semantic and grammatical statistical information processing in the absence of learning

ÁNGEL TABULLO; PAMELA LOPES DA CUNHA; ENRIQUE SEGURA; SILVANO ZANUTTO; ALEJANDRO WAINSELBOIM

Florianópolis

Congreso; Language and Neuroscience Conference; 2012

Universidade Federal de Santa Catarina

First language acquisition occurs by mere exposure to the linguistic context. During this process, infants must solve among other problems, two basic issues: i) infer word meanings by pairing specific aspects of their perceptual experience with particular segments of the concurrent linguistic input (semantics); and ii) acquire the combinatorial rules that allow to convey a coherent message within that language (grammar). The ability to solve these issues requires to statistically analyze cross-situational information, i.e. to associate information that is concurrent along multiple events or situations. Experimentally, it has been shown that adults are able to infer new noun meanings by cross-situational pairings of images and auditory non-words (Yu & Smith 2007). Similarly, we analyzed whether adults are also able to infer new verbal meanings by associating concurrent visual and formal information (Lopes da Cunha et al. 2010). In this experiment training consisted of 70 different visual scenes presented on a computer screen (two geometrical figures, one static and the other performing one of 6 possible movements). A sentence describing the scene in an artificial language was presented audio-visually in simultaneous. Participants had to learn which word denoted each movement. During testing 80 new scenes were shown, 40 of them presented a mismatch between the movement and the presented “verb”. Subjects decided on-line whether each sentence correctly described the scene. Simultaneous EEG recordings were obtained in this stage. Most subjects responded significantly above chance level during testing. In those who learned, the appearance of a mismatch between verb and referent elicited an N400-like wave. Thus, under controlled experimental conditions, subjects infer new verbal meanings by conscious cross-situational pairing of images and audiovisual non-words, if they are given the explicit instruction to learn the presented terms. In a second experiment we studied if new verbal meanings could be learned in the absence of intention to acquire regularities between linguistic and contextual information. 22 right-handed healthy adults volunteered for the experiment. The training protocol followed that of Lopes da Cunha et al 2010, but participants were told not to pay attention to the sentences and concentrate on classifying the movements they saw as horizontal or non-horizontal. During the test stage (as in Lopes da Cunha et al. 2010), subjects were told to decide on-line whether each sentence correctly described the scene. Simultaneous EEG recordings were obtained in this stage. Results showed that participants were completely unable to classify the sentences as correct or not. Nevertheless, the appearance of a mismatch between verb and referent elicited an ELAN-like potential. Thus, although no behavioral evidence could be found of new semantic knowledge, at the neurobiological level evidence was found that an association was achieved between movements and their corresponding terms. Finally, in a third experiment we eliminated the semantic (visual) referent and asked participants to learn the combinatorial rules of the artificial grammar. During testing, subjects were asked to classify as grammatical or ungrammatical 160 new sentences that followed or not  the trained combinatorial rules. Results showed that certain sequence types elicited a late positive component which was modulated by different factors in two distinct time windows. In an earlier window, the component was higher for sequences which had a low or null probability of occurrence during training, while in a later window, the component was higher for incorrect than correct sequences. Although the late window effect was absent in subjects that were unable to learn the trained rules, these subjects presented the earlier positive component that was sensitive to the probability of occurrence during training. Thus, even in the absence of knowledge, neurobiological evidence was found showing that statistical information was acquired of the possible combinations between items.