Crossmodal motion priming
We next tested for an effect of motion priming between vision and
audition by examining the data from the crossmodal condition in which
auditory and visual motions were presented in alternation (A,V or V,A)
over trials with direction randomly determined on each trial (left or
right). For the analysis shown in Figure 4, we examined only pairs of
trials that had congruent directions. This design allowed us to test for
crossmodal motion priming in both directions – audition priming vision
(AV), and vision priming audition (VA)and the results are shown Figure
4. The psychometric functions in Figure 4a show auditory speed
judgements when preceded by either a fast (red) or a slow (blue) visual
motion. Figure 4b shows the converse: visual speed judgements when
preceded by fast or slow auditory motion. As observed within modalities,
we found significant motion priming between modalities in both
directions. A bootstrap sign-test based on 10,000 iterations confirmed
that PSEs were significantly lower (leftward shift of the psychometric
function) after ‘previous fast’ trials for both vision priming audition
(VA: p < .0001) and for audition priming vision (AV: p
< .0001). This demonstrates that the previous trial does not
need to be in the same modality to prime a subsequent motion, and that
the motion priming effect occurs symmetrically, with the AV and VA
conditions producing equivalent results.
We divided the ‘fast’ and ‘slow’ data further to examine the priming
effect as a function of each level of prime speed. Figure 4c plots the
PSEs from the psychometric functions computed for each level of
preceding motion. For both the AV and VA conditions, motion priming
shows a clear linear dependence on the speed of the priming stimulus
that is well described by a linear function. The results mimic those in
the Figure 3b and clearly illustrate that motion priming also occurs for
alternating auditory and visual trials. Again, we bootstrapped the data
for each preceding velocity 10,000 times and fitted linear functions
across prime speed for each iteration to compare slopes for the AV and
VA condition. The slopes were significant for each modality order (AV:
slope = -.0763, p < .0001; VA: slope = -.1249, p <
.0001) and quite similar, although the difference was statistically
reliable (p = .0009) with VA priming showing a stronger effect than AV.