Decoding Digits and Dice with Magnetoencephalography: Evidence for a Shared Representation of Magnitude

Abstract

Numerical format describes the way magnitude is conveyed, for example, as a digit ("3") or Roman numeral ("III"). In the field of numerical cognition, there is an ongoing debate of whether magnitude representation is independent of numerical format. Here, we examine the time course of magnitude processing when using different symbolic formats. We presented participants with a series of digits and dice patterns corresponding to the magnitudes of 1 to 6 while performing a 1-back task on magnitude. Magnetoencephalography offers an opportunity to record brain activity with high temporal resolution. Multivariate pattern analysis applied to magnetoencephalographic data allows us to draw conclusions about brain activation patterns underlying information processing over time. The results show that we can cross-decode magnitude when training the classifier on magnitude presented in one symbolic format and testing the classifier on the other symbolic format. This suggests a similar representation of these numerical symbols. In addition, results from a time generalization analysis show that digits were accessed slightly earlier than dice, demonstrating temporal asynchronies in their shared representation of magnitude. Together, our methods allow a distinction between format-specific signals and format-independent representations of magnitude showing evidence that there is a shared representation of magnitude accessed via different symbols.

Neural structure mapping in human probabilistic reward learning

“…Using representational similarity analysis (RSA) (Kriegeskorte and Kievit, 2013), we replicated the previous finding (Spitzer et al, 2017; Teichmann et al, 2018) that patterns of neural activity across the scalp from ~100 ms onwards were increasingly dissimilar for numbers with more divergent magnitude, that is codes for ‘3’ and ‘5’ were more dissimilar than those for ‘3’ and ‘4’ (Figure 1B, green line). This occurred irrespective of task framing (report higher vs. lower average) and category (orange vs. blue numbers), suggesting that neural signals encoded an abstract representation of magnitude and not solely a decision-related quantity such as choice certainty (Spitzer et al, 2017).…”

“…In scalp M/EEG signals, neural patterns evoked by Arabic digits vary continuously with numerical distance, such that multivariate signals for ‘3’ are more similar to those for ‘4’ than ‘5’. (Spitzer et al, 2017; Teichmann et al, 2018). In scalp M/EEG signals, neural patterns evoked by Arabic digits vary continuously with numerical distance, such that multivariate signals for ‘3’ are more similar to those for ‘4’ than ‘5’.…”

Untangling featural and conceptual object representations

“…Some low-level visual information may have been lost in the scrambling process, such as shape and curvature information, which is a strong cue for animacy (Levin, Takarae, Miner, & Keil, 2001;Schmidt, Hegele, & Fleming, 2017;Zachariou, Giacco, Ungerleider, & Yue, 2018). In MEG and EEG decoding studies, classification can be strongly driven by differences in object shape (Proklova et al, 2019), and silhouette similarity is often a strong predictor of the similarities between the earliest neural responses (Carlson et al, 2013;Grootswagers et al, 2019;Teichmann, Grootswagers, Carlson, & Rich, 2018;Wardle, Kriegeskorte, Grootswagers, Khaligh-Razavi, & Carlson, 2016). It is also important to note that while the texform images are not recognisable at the individual level, they can still be categorised (e.g., for animacy) above chance (Long et al, 2017).…”

The representational dynamics of visual objects in rapid serial visual processing streams

“…We also used three low-level image feature control models (Figure 2, third row), which were created by correlating the vectorized experimental images. The models consisted of an image silhouette similarity model, which is based on the binary alpha layer of the stimuli and is a good predictor of differences in brain responses (Carlson et al, 2011;Teichmann, Grootswagers, Carlson, & Rich, 2018;Wardle, Kriegeskorte, Grootswagers, Khaligh-Razavi, & Carlson, 2016)), a model based on the CIELAB-colour values of the stimuli, and a model based on the difference in luminance of the stimuli.…”