Is ‘heavy’ up or down? Testing the vertical spatial representation of weight

Interestingly, it has been suggested that the correlation between mass and downward force (i.e. weight) underlies the well-known tendency—displayed by people without formal physics instruction—to overestimate the positive relationship between object mass and its falling speed (see Rohrer 2003; Vicovaro 2014; Vicovaro et al. 2019).

The use of a vertical response box is highly desirable in experiments investigating vertical SNARC and SNARC-like effects. Indeed, if participants’ responses are collected through a traditional keyboard placed on a horizontal plane and ‘vertically aligned’ keys are used (e.g. ‘Y’ and ‘B’; see for instance Ito and Hatta, 2004), then the response keys are actually aligned sagittally rather than vertically. In turn, as highlighted by Winter et al. (2015), this can lead to confounding between ‘bottom-top’ and ‘near-far’ dimensions.

The SNARC effect is frequently tested by computing, for each number stimulus, the mean RT difference between the right- and the left-side key, and then by testing the existence of a negative correlation between number magnitudes and mean RT difference (see Fias, Brysbaert, Geypens, and d’Ydewalle 1996). Theoretically, this approach could also be used in the current context using the mean rated weight of target words instead of number magnitude. However, when magnitude is task relevant, as in our study, the mean RT difference is not a linear but a categorical function of magnitude, which implies the violation of one basic assumption of linear regression analysis (see Gevers et al. 2006).

According to Lakens (2012), it can be established which of two opposite categories (e.g. light vs. heavy) is positively polarized by considering which category gives the name to its dimension. Here, the positively polarized category is ‘heavy’ as it gives one of the names with which to refer to the ‘weight’ dimension (i.e. heaviness). This is also reflected in in common language, since there is a natural preference is using ‘heavy’ rather than ‘light’ when referring to heaviness/weight dimension, such as when we want to know the weight of another individual; in this case, we typically ask ‘how heavy are you?’ and not ‘how light are you?’.

A well-known phenomenon underlying weight perception is the so-called ‘size–weight illusion’: When two objects of identical physical weight but different size are lifted, the smaller object is typically perceived to be heavier than the larger object (e.g. Buckingham, 2014; Vicovaro and Burigana, 2014). This explains why, in the present context, the physical weight of the two bigger spheres (light and heavy) exceeded that of the corresponding smaller spheres. Moreover, the weight difference between the small and the big sphere was larger for the heavy spheres (695 g and 1100 g, respectively) than for the light spheres (97 g and 154 g, respectively). This was done on purpose to comply with Weber’s law, according to which differences in perceived weight are related to weight ratios rather than to weight differences. Since a weight ratio of about 1.58 (i.e. 154 g/97 g) nullified the size-weight illusion in the case of light spheres, the same weight ratio was also maintained for heavy spheres (i.e. 1100 g/695 g ≈ 1.58), to nullify the size-weight illusion likewise. For completeness, the perceived weight of the four spheres was also pre-tested by a sample of 24 individuals (mean age M = 23 years, SD = 2.5; five males, one left-handed). In more detail, participants were asked to lift each sphere with two hands, and to estimate its weight by providing an integer numerical value. The integer value was recorded manually by the experimenter. Sphere order was counterbalanced across participants. A 2 (weight: light vs. heavy) × 2 (size: small vs. big) repeated-measures ANOVA was conducted on standardized estimates. The main effect of weight was significant [F(1, 23) = 7213.5, p < .001, η²_g = .899], confirming that light and heavy spheres were perceived as different, whereas the main effect of size was non-significant (F < 1). The interaction between the two factors was significant [F(1, 23) = 5.85, p = .024, η²_g = .107]. Nevertheless, two-tailed paired t-tests revealed that the two light and the two heavy spheres were perceived to be similar in weight independently of their size (ts < 1.88, ps > .072).

As suggested by one reviewer, data of Experiment 3b were also analysed excluding the 6 left-handed participants, since a previous study reported an association between stimulus type and response location in right-handers but not in left-handers (see Huber et al., 2015). However, the results of these explorative analyses showed that the two interactions between response location and either size (i.e. the task-relevant dimension) or weight (i.e. the task-irrelevant dimension) were both still non-significant (ps > .10).