Crossmodal comparability for stimulus magnitudes on modalities lacking a common intensity aspect, such as line length and tone loudness, may rest on a more general magnitude concept.
In a recent study1, possible time-order error (TOE)-like effects were looked for in crossmodal comparison of successive stimuli. Hellstro ̈m’s2 sensation-weighting (SW) model was adapted for use as a descriptive and analytical tool. According to this model, the real comparison takes place between two weighted subjective compounds, each built up by the sensation magnitudes of a stimulus (weight s1 or s2) and of its reference level (ReL) (weight 1 − s1 or 1 − s2).
Intramodally (tone-tone, line-line) and crossmodally (tone-line, line-tone) paired stimuli, with duration 150 ms and ISIs 400 and 2000 ms, were compared for their strength, with the alternatives 1st stronger, 2nd stronger, and equal, yielding scaled subjective difference (D) of +100, 0, and -100. The extended SW model, for a pair with a tone followed by a line, is
D=kT[s1(Φ1T −Φ0T)+(1−s1)(Φr1T −Φ0T)]−kL[s2(Φ2L −Φ0L)+(1−s2)(Φr2L −Φ0L)]+b 1a = s1kT (Φ1T −Φr1T )−s2kL(Φ2L −Φr2L)+kT (Φr1T −Φ0T )−kL(Φr2L −Φ0L)+b, 1b
where Φ is the physical stimulus magnitude on a scale, -4 through +4 for lines as well as tones (90–170 mm, and 74.7–81.1 dB). It was assumed that, within these ranges, Φ = k(Φ − Φ0), where Ψ is the subjective magnitude and Φ0 is the Φ value for Ψ = 0. k is a modality-specific scale factor. Subscripts T and L indicate tone and line; 1 and 2 indicate temporal position. Φr is the physical magnitude of the ReL. b is possible bias. The experimental design, varying both stimulus magnitudes, allowed estimation of the relevant model parameters from the data.
Comparing a tone and a line for their “strength” was indeed feasible, yielding weighting effects and TOEs (assessed by mean D) resembling those in intramodal comparison. In terms of the SW model, with ISI = 2000 ms s1 < s2. TOEs were generally more negative with the 2000–ms than with the 200–ms ISI. The results were well accounted for by the SW model.
Eq. 1b implies that the effective subjective magnitude of each paired stimulus is its ReL plus its deviation from the ReL multiplied by s. Estimates of kT and kL were similar, indi- cating crossmodally similar ranges of Ψ. Based on this and on findings3, 4 that context effects are much stronger crossmodally than intramodally, it may be hypothesized that in crossmodal comparison the k value for each modality becomes inversely related to the range or spread of the sensation magnitudes in the modality. In the SW model (Eq. 1b) this would mean that the stimulus-dependent parts of the compared magnitudes are standardized quantities, like z values. These would be dimensionless, removing the need for an intermodal “currency exchange” and accounting for comparability of subjective magnitudes in very different modalities.
Lund: International Society for Psychophysics, Lund University , 2014. 33-33 p.