Anticipating the magnitude of response outcomes can induce a potentiation effect for manipulable objects

The term “potentiation” is usually employed to refer to affordance effects rather than to effects due to a compatibility of codes. Researchers commonly use the “facilitation” to refer to this latter case. But substantially, both concepts describe a similar phenomenon: the advantage in a compatible condition compared to a non-compatible one. Accordingly, in this manuscript, we chose to only use the term “potentiation” (not “facilitation”) that we defined as shorter RTs in the compatible condition than in incompatible one (whatever its possible origin, i.e., affordance or code compatibility).

As highlighted by Pfister (2019) and Pfister & Kunde, 2013), our artificial auditory outcome was not the only perceptual outcome associated with the responses. There were also various perceptual outcomes that were more intrinsic to the responses (e.g., tactile feedbacks when participants press the key). However, based on Heurley’s et al. (2020) experiments, we hypothesized that this auditory outcome will be enough to induce a stimulus-outcome compatibility.

In contrast, the spatial-musical association of response codes (SMARC, Ishihara et al., 2008; Rusconi et al., 2006) suggests an alternative view with low tone coded as “small” and high tone coded as “large”. The current study was not designed to disentangle these approaches.

Thus, there were four mapping groups of seven participants: “forefinger, blue, low tone/palm, orange, high tone”, “forefinger, blue, high tone/palm, orange, low tone”, “forefinger, orange, low tone/palm, blue, high tone”, and “forefinger, orange, high tone/palm, blue, low tone”.

It is noteworthy that these results are consistent with the multimodal coupling (large/low tone vs. small/high tone) already reported by Parise and Spence 2009; 2015, Parise and Ernst 2016) and not the reverse coupling reported by Rusconi et al. (2006). Further studies should address more directly why low and high tones are respectively linked to large and small magnitude, and not the reverse.

The Bayesian analyses performed in JASP rely on sequential sampling methods. Due to the numerical approximation of the algorithm, the resulting BF vary from one analysis to another even if it is conducted on the same dataset in the exact same way. Even if some authors consider this variability to be small (e.g., Goss-Sampson et al., 2020), Pfister (2021) warned that it can be very large in some cases. In JASP, this variability is quantified by the error percentage provided in the right-most column of the full model comparison table (i.e., “error %”). The higher this error percentage the more the resulting BF vary from one analysis to another. This variability partly depends on the number of samples used by the sequential sampling methods. The higher this number of samples the smaller the variability. JASP allows to increase this number of samples by setting the “Numerical Accuracy” (under the “Additional Options” section of the Bayesian ANOVA module) to manual and by choosing the desired number of samples. The default value is 10,000 samples. The results we reported here came from a sequential sampling methods based on the maximum number of samples allowed by JASP (i.e., 10,000,000 samples). The averaged error percentage for all the Bayes Factors computed for the full model comparison was 0.88% (s = 1.64%). This averaged error percentage is very small and indicates that our results should vary only slightly from one analysis to another. Our reported Bayesian analysis focused on the BFexcl rather than on all the BF₀₁ from the full model comparison. BF₀₁ is the ratio between the posterior probability of the data under the null hypothesis (H0) and the posterior probability of the data under the alternative hypothesis (H1) and is interpreted as the strength of evidence for H0 relative to the strength of evidence for H1. Even if JASP does not provide error percentage for BFexcl, these error percentages should also be small because they are related to the error percentages of the BF from the full model comparison upon which the calculation of BFexcl is based.

Note that the inclusion Bayes Factor (BF_incl) reflects the strength of evidence supporting the averaged model including the existence of a given effect (i.e., the alternative hypothesis, H1) relative to the strength of evidence supporting the averaged model excluding the existence of this effect (i.e., the null hypothesis, H0). BF_incl = 1/BF_excl.

The unstandardized or raw effect size is merely the mean difference between RTs in compatible conditions (e.g., small objects/high tone) and RTs in non-compatible conditions (e.g., large object/high tone) for each participant in our between-experiments analysis.