|dc.description.abstract||Redundancy gain refers to the common finding in experimental psychology that the presentation of multiple, redundant stimuli tends to evoke responses more quickly and accurately when compared to presentation of a single stimulus. Two types of account of such findings are generally offered. First, race models suggest that redundancy gain is a result of statistical facilitation. Second, coactivation models suggest that redundancy gain results from each stimulus making some contribution to the eventual response.
Though redundancy gain has primarily been shown in relatively simple tasks (e.g., detection, perceptual discrimination), there have also been efforts to demonstrate comparable phenomena in tasks involving higher-order cognition. One example of this is in the work of Mohr and Pulvermüller in lexical decision tasks (LDTs) using redundant stimuli (e.g., Mohr, Pulvermüller, & Zaidel, 1994; Mohr, Pulvermüller, Rayman, & Zaidel, 1994; Mohr, Pulvermüller, Mittelstädt, & Rayman, 1996; Mohr, Endrass, Hauk, & Pulvermüller, 2007). Those authors explained redundancy gain in LDT on the basis of a cell assembly model of lexical representation (e.g., Pulvermüller & Mohr, 1996; Pulvermüller, 1999). According to this explanation, activity from redundant stimuli sums in the network of cells where the word is neurally represented, leading it to “ignite” more rapidly and effectively — a form of coactivation — which in turn leads to faster and more accurate responses.
I sought to determine whether a similar phenomenon would occur in a semantic categorisation task, and whether the same basic model could be used to account for such findings. To investigate this issue, I conducted a series of experiments based on the LDTs used by Mohr and Pulvermüller. In my experiments participants were asked to classify visually-presented lexical stimuli as members or non-members of a pre-specified target category, and make the appropriate (“target present”/“targetabsent”) response.
Experiments 1–3 showed that redundancy gain can be demonstrated in a semantic categorisation task, and that this can occur with both lateralised and nonlateralised stimulus presentation. The pattern of results from these experiments was strikingly similar to the results of Mohr et al. (1996): improved performance in redundant trials, and an advantage for stimuli presented in the right visual field over those presented to the left visual field when display was lateralised. I also found redundancy gains and visual field effects for “target-present” but not “target-absent” responses, analogous to findings in LDTs for “word” and “non-word” responses, respectively.
Experiment 4 showed that performance in LVF trials does not improve substantially when participants are allowed longer to respond, suggesting that the high error rates in that condition in preceding experiments are likely a result of data rather than resource-limited processing. Experiment 5 showed that visual field effects in target-present trials are absent when stimuli are presented vertically rather than horizontally. Under the assumption that vertical presentation should disrupt lexical processing but not access of semantic representations, this implies that visual field effects in earlier experiments were not due to such representations being cerebrally asymmetrical.
Experiment 6 showed that when the task involves a decision between two target categories, rather than between targets and non-targets, redundancy gains and visual field effects are undiminished for both categories. Finally, Experiments 7 and 8 showed that redundancy gain does not decrease when multiple target categories are used and redundant trials involve the presentation of two stimuli from different categories (versus experiments with single target categories or redundant trials with two stimuli from the same category). This provides evidence against a cell assembly coactivation account, as this account would predict greater coactivation with same category than different-category redundant targets.
Based on the results of Experiments 1–8, it is apparent that redundancy gain is not limited to simple tasks, but is rather a more generalisable phenomenon. In addition, as the cell assembly coactivation account appears inappropriate to explain the results of Experiments 7 and 8, other accounts (e.g., race models, response-level coactivation) are preferable. These accounts are explored in the General Discussion.||