Background. Interpretation is the process through which humans attribute meanings to every input they grasp from their natural or social environment. Formulation and exchange of meanings through natural language are basic aspects of human behaviour and important neuroscience subjects; from long ago, they are the object of dedicated scientific research. Two main theoretical positions (cognitivism and embodied cognition) are at present confronting each other; however, available data is not conclusive and scientific knowledge of the interpretation process is still unsatisfactory. Our work proposes some contributions aimed to improve it. Methodology. Our field research involved a random sample of 102 adults. We presented them a real world-like case of written communication using unabridged message texts. We collected data (written accounts by participants about their interpretations) in controlled conditions through a specially designed questionnaire (closed and opened answers). Finally, we carried out qualitative and quantitative analyses through some fundamental statistics. Principal Findings. While readers are expected to concentrate on the text’s content, they rather report focusing on the most varied and unpredictable components: certain physical features of the message (e.g. the message’s period lengths) as well as meta-information like the position of a statement or even the lack of some content. Just about 12% of the participants' indications point directly at the text's content. Our data converge on the hypothesis that the components of a message work at first like physical stimuli, causing readers' automatic (body level) reactions independent of the conscious attribution of meaning. So, interpretation would be a (learned) stimulus-reaction mechanism, before switching to information processing, and the basis of meaning could be perceptual/analogical, before propositional/digital. We carried out a first check of our hypothesis: the employed case contained the emerging of a conflict and two versions (“H” and “S”, same content, different forms) of a reply to be sent at a crucial point. We collected the participants’ (independent) interpretations of the two versions; then, we asked them to choose which one could solve the conflict; finally, we assessed the coherence between interpretations and choice on a 4-level scale. The analysis of the coherence levels' distribution returned that, with regards to our expectations, incoherence levels are over-represented; such imbalance is totally ascribable to “H” choosers. “H” and “S” choosers show significant differences (p<<0.01) in the distributions of coherence levels, what is inconsistent with the traditional hypothesis of a linear information processing resulting in the final choice. In the end, with respect to the currently opposing theories, we found out that our hypothesis has either important convergences or at least one critical divergence, joined with the capacity to encompass they both.
Collection of Meta Information with User-Generated Question Answer Pairs and its Reflection for Improving Expressibility in Response Generation
