Responses exceeding 2,000 ms were excluded from the analysis, as were incorrect responses. The data were analyzed in a 2 × 2 factorial design, with the factors Perceptual Load (high, low) and Compatibility Between Hand of Response and Orientation of Handle (compatible, incompatible). Mean reaction times (RTs) and standard deviations were calculated for each participant as a function of load and compatibility (see Table 1) and entered into a 2 (perceptual load: low, high) × 2 (compatibility: compatible, incompatible) within-subjects analysis of variance (ANOVA), with Participants as the random factor. The ANOVA revealed a significant main effect of load, F(1, 13) = 185.15, MSE = 6,116.37, p < .001, η
= .934. RTs were slower during high perceptual load (M = 814 ms) than during low perceptual load (M = 530 ms), confirming that our manipulation of perceptual load was successful. There was no significant main effect of compatibility (F < 1). Importantly, however, we did find a significant interaction between load and compatibility, F(1, 13) = 6.16, MSE = 240.73, p < .05, η
= .322. There was a reliable compatibility effect under low perceptual load, with faster responses for compatible (M = 524 ms) than for incompatible (M = 537 ms) trials, t(13) = 3.224, p < .01 (Bonferroni corrected), but not under high perceptual load (compatible, M = 818 ms; incompatible, M = 811 ms; t < 1). The magnitude of the compatibility effect under low load was comparable to those in previous reports (e.g., Ellis & Tucker, 2000; Tucker & Ellis, 1998, 2001).
Responses to high-load trials were slower than those to low-load trials, which could have resulted in a failure to demonstrate a reliable compatibility effect, as the impact of the irrelevant handle could have been abolished by the time that a response was made. This issue has previously been demonstrated in a classic flanker task when the relevant set size was large (Miller, 1991). To rule out this possibility, we used a median split to separate each experimental condition into fast and slow RTs. We then tested the effects of interest in the faster high-load responses and the slower low-load responses. A 2 (perceptual load) × 2 (compatibility) ANOVA revealed no significant main effect of load (F < 1), which confirmed that there was no difference in response latencies between the two conditions (high load, M = 438 ms; low load, M = 427 ms). The main effect of compatibility was marginally significant, F(1, 13) = 4.55, MSE = 448.82, p = .053, η
= .259, with faster RTs in the compatible condition (M = 426 ms) than in the incompatible condition (M = 438 ms). Importantly, we again found a significant interaction between load and compatibility, F(1, 13) = 4.99, MSE = 508.58, p < .05, η
= .277. As in the main analysis, there was a reliable compatibility effect under low load, with faster responses for compatible (M = 414 ms) than for incompatible (M = 440 ms) trials, t(13) = 2.93, p < .025 (Bonferroni corrected), but not under high load (compatible, M = 439 ms; incompatible, M = 437 ms; t < 1). This confirms that the effect of load on compatibility effects was not due to differences in the absolute latencies between load conditions.
As targets were presented at different horizontal positions, there was an additional compatibility effect involving target position and response hand. Although target position was fully counterbalanced with respect to both response hand and the position of the handle, we wanted to establish that there were no systematic differences in the key effects as a function of the compatibility between target position and response hand. To do so, we took trials on which the target appeared in the extreme left or right position and ran a 2 (perceptual load) × 2 (compatibility between response hand and handle orientation) × 2 (compatibility between response hand and target position) ANOVA. There was a significant effect of compatibility between response hand and target position, F(1, 13) = 5.24, MSE = 6,717.95, p < .05, η
= .287: Responses were faster when the target letter was positioned on the same side as the hand of response (M = 484 ms), as compared to the other side (M = 520 ms). However, no other effects involved compatibility between response hand and target position (all ps > .05), confirming that the observed affordance effects were not systematically affected by this factor.
Finally, as most participants were right-handed, it was important to rule out the possibility that they were more perceptually sensitive to distractors with a handle presented to the right. Therefore, we ran an additional ANOVA, this time adding the factor of Handle Orientation (left, right). This ANOVA revealed no main effect of handle orientation, and neither were there any interactions involving handle orientation (all ps > .05), confirming that compatibility effects did not vary as a function of the handle orientation of the distractor object.
Mean accuracy rates and standard deviations were calculated for each participant as a function of load and compatibility (see Table 1) and entered into a 2 (perceptual load: low, high) × 2 (compatibility: compatible, incompatible) within-subjects ANOVA, with Participant as the random factor. The ANOVA revealed a significant main effect of load, F(1, 13) = 36.68, MSE = .005, p < .001, η
= .738. Accuracy was lower during high perceptual load (M = 83.1 %) than during low load (M = 94.9 %), again confirming that perceptual load was successfully manipulated. No other significant effects were revealed in the accuracy ANOVA (all ps > .05). The pattern of accuracy rates thus indicated that no speed–accuracy trade-offs were present in the data. As the key effect of perceptual load on affordance compatibility was not significant in the accuracy analysis, we did not run the additional analyses to evaluate the effects of target position and handedness on the effects of interest.