Additive Groups in a Non-Symbolic ‘Artificial Algebra’

Grace et al. (2018) developed an ‘artificial algebra’ task in which participants learn to make an analogue response based on a combination of non-symbolic magnitudes by feedback and without explicit instruction. Here we tested if participants could learn to add stimulus magnitudes in this task in accord with the properties of an algebraic group. Three pairs of experiments tested the group properties of commutativity (Experiments 1a-b), identity and inverse existence (Experiments 2a-b) and associativity (Experiments 3a-b), with both line length and brightness modalities. Transfer designs were used in which participants responded on trials with feedback based on sums of magnitudes and later were tested with novel stimulus configurations. In all experiments, correlations of average responses with magnitude sums were high on trials with feedback, r = .97 and .96 for Experiments 1a-b, r = .97 and .96 for Experiments 2a-b, and ranged between r = .97 and .99 for Experiment 3a and between r = .82 and .95 for Experiment 3b. Responding on transfer trials was accurate and provided strong support for commutativity, identity and inverse existence, and associativity with line length, and for commutativity and identity and inverse existence with brightness. Deviations from associativity in Experiment 3b suggested that participants were averaging rather than adding brightness magnitudes. Our results confirm that the artificial algebra task can be used to study implicit computation and suggest that representations of magnitudes may have a structure similar to an algebraic group.