Using fMRI to Decode Attention to Semantic versus Perceptual Features over Development

• In Event: Poster Session 10
  In Poster Session: PS 10 Topic - Attention, Learning, Memory

Abstract

Prior work suggests adults and children remember different information from the same experience, with children remembering visually salient details more than adults (Fisher & Sloutsky, 2005; Rabi et al., 2015). A potential explanation for this effect is differential attention during learning. Specifically, children may attend to perceptual details more than adults and consequently form more perceptually detailed memories (Brainerd & Reyna, 2014; Vendetti et al., 2015). Past research in adults has shown different neural signatures are recruited when processing semantic versus perceptual information (Schott et al., 2013; Sheldon et al., 2016). However, it is unclear whether developmental differences in engagement of semantic and perceptual neural states during memory formation might account for changes to memory quality.

Here, we will use a neural decoding approach to ask whether the adult semantic and perceptual states defined during a selective attention task 1) can be used to decode attentional biases during memory retrieval, and 2) are consistent across late childhood and adolescence.

In the selective attention task, we cued participants to attend to either semantic or perceptual dimensions of complex visual (storybook illustration) stimuli by asking them to detect repeats along the cued dimension during functional magnetic resonance imaging (fMRI; Figure 1A). We then trained a machine learning classifier to discriminate between semantic and perceptual attentional states on the basis of whole-brain fMRI activation patterns (Figure 1B). Our classifier was able to decode attentional state well above chance (p<0.0001)—even across participants—suggesting that semantic and perceptual attention evoke distinct states that are shared across individuals.

Next, we will examine whether the same semantic and perceptual states defined in our selective attention task are engaged when attention is directed to general semantic or specific episodic aspects of a retrieved memory. We will apply our trained classifier to a public fMRI dataset in which participants performed an autobiographical retrieval task (Fynes-Clinton, Marstaller, & Burianová, 2019; Figure 1C). Participants were cued to retrieve autobiographical memories, semantic facts, and episodic details for target events. We will apply our trained classifier to this new dataset to quantify the prevalence of semantic and perceptual attention states during the cued semantic and episodic retrieval tasks. We predict that all ages will show greater classifier evidence (log odds of classifier probabilities) for semantic attention during the semantic retrieval task and perceptual attention during the episodic retrieval task (Figure 2A), suggesting that our observed differences in attention states capture relevant distinctions during retrieval. We will interrogate the autobiographical retrieval task to ask whether children show a perceptual bias and adults a semantic bias or mix (Leyhe et al., 2009; Sheldon et al., 2016) of attention in this uncued state (Figure 2B).

These results will provide a foundation for future work relating developmental differences in attention to memory and inform the tailoring of instructional strategies to different age groups.

All data for this task has been collected and we have prior experience performing analyses similar to the ones proposed. As such, completing this work before the meeting is highly feasible.

Authors

• Sagana Vijayarajah, University of Toronto
  Presenting Author

• Meg Schlichting, University of Toronto
  Non-Presenting Author