Language-mediated visual attention in toddlers

Language plays an important role in driving visual attention. The aim of this thesis is to gain a better understanding of how language may mediate visual attention in toddlers, who begin to learn about the world through language communication. In Part I of the thesis, I examine the time course and order in which different types of stored linguistic representations are activated during spoken-word recognition. In Experiments 1–4, I demonstrate that the four-picture adult visual world paradigm (VWP) in combination with growth curve analysis and mixed effects modelling (Mirman, 2014), is a suitable tool for measuring language-mediated cognitive processes in toddlers. The results indicate that, like adults, toddlers extract phonological information faster than semantic information during spoken-word recognition, and that these differences hold irrespective of whether the semantic information is taxonomic or thematic in character. Moreover, phonological effects are related to the toddler’s overall vocabulary size, suggesting a role for emerging phonological organisation and/or word familiarity in the phonological decoding of words. The results of the study are compatible with an interactive view of language-mediated visual attention in toddlers whereby implicitly generated phonological codes and speech-driven phonological encoding work together to support referent identification. In Part II of the thesis, I investigate how language as an attentional set may guide toddlers' visual attention. Specifically, I examine the role of dis- tractor inhibition during language-mediated visual attention using a negative priming (NP) procedure. NP refers to the impaired response observed when a participant has to respond to a stimulus identity or location that they previously ignored (i.e., identity or location NP). Any observed NP effect would suggest that language-mediated selective attention involves inhibition of response towards the distractor. In Experiment 5 (six conditions), participants were shown trials each containing a prime phase, followed by a probe phase. In both phases, participants heard the target picture being named and were allowed to look freely at the pictures. In all conditions, participants were 'on-task' and showed greater than chance target preference after target label onset. No location NP effect was observed when the prime stimuli were repeated in the same locations in the probe phase, or when the probe stimuli were novel. However, location NP was observed when the prime stimuli were repeated in swapped locations in the probe phase. That is, there was significantly less target preference in the ignored location–attended identity condition. Rather than by distractor inhibition, these results could be better explained by the mismatching of feature–location–response bindings in the prime and probe phases. These findings indicate that language drives the encoding of temporary episodic representations of events, which are stored, retrieved and compared with later events, and may mediate visual attention. In sum, this thesis demonstrates that language-mediated visual attention involves the activation and retrieval of 'higher level' representations, which are then matched and integrated with the 'lower level' visual and auditory inputs. While Part I demonstrates how relatively long-term lexical-semantic knowledge mediates visual attention, Part II demonstrates how recently encountered events are encoded into episodic representations, retrieved and in turn, influence mediated visual attention.