Slide Slam S7 Sandbox Series
Proto-Language as a Structurer and Enhancer of Perception
James Scott1, Robert Foley1, Mirjana Bozic1; 1University of Cambridge
The evolution of the capacity for language is hotly debated. To date, much of the investigation has focused on the communicative functions of language. By contrast we examine the cognitive structuring consequences of language capabilities, which may have acted as a potential evolutionary route by providing an immediate adaptive advantage over ape communication. Language has been shown to enhance a variety of cognitive processes both on- and offline, including low level visual perception, and categorisation. Notably, these effects pertain to semantics without syntax. There has also been a resurgent interest in the concept of the non-arbitrariness of elements of language (iconicity). Iconicity may have provided an important foothold on the path to full-blown language by providing an immediate link between sound and meaning. Iconicity therefore renders the problem of language evolution less intractable, while links to embodiment and the cognition of other primates provide plausibility. The present study seeks to investigate language evolution by linking iconicity to the non-communicative functions of semantics. We hypothesise that early proto-words were iconic, and significantly structured cognition. This evolutionary route does not necessarily exclude communication as an additional selective pressure, and proximate function of language. To test this hypothesis, data were collected via an online game. First, new conceptual categories were created, consisting of animal-like visual stimuli which varied along four dimensions of shape and colour. This 4D tensor was then bisected to create two distinct categories, each containing several hundred exemplars which were more or less representative of the category. 141 participants were first trained on the new semantic categories using a protocol adapted from1. Participants were then tested using match to sample (MTS) and a novel ‘camouflage’ task. The former attempted to assess how the newly learnt categories affected visual recognition, including how participants responded to borderline stimuli and if the perceptual magnet applied here. The camouflage test used filters to distort stimuli, in an attempt to access lower levels of the visual cognition hierarchy, which have previously been shown to be affected by top-down semantic influence. Inter-participant conditions varied by the nature of an auditorily presented verbal label, in both training and testing phases of the experiment. In training, stimuli were paired with either an iconic pseudoword label; a non-iconic pseudoword label; or no label. The same conditions were used in the test phase, though half participants from labelled training conditions received no label in testing. This was to distinguish between the online and offline effects of labels. We predicted that both learning and testing performance would be significantly enhanced by iconic labels, both on- and offline. Hence, participants both trained and tested on iconic labels would perform best. Participants trained and tested on non-iconic labels were expected to do better than those in no label conditions. Pilot data tentatively support these predictions; full analyses still are ongoing. 1. Lupyan, G. & Casasanto, D. Meaningless words promote meaningful categorization. Lang. Cogn. 7, 167–193 (2014).