1

Supplement:

In a recent related study with hierarchically related target-distractor pairs, Hantsch etal. (2009) also demonstrated a reversal of semantic effects in the picture–word interference task. In that study, participants named pictures of objects, which had their preferred name at the basic level, with their subordinate-level names (e.g., target name “carp” with the preferred picture name “fish”). Basic-level distractors (e.g., “fish”) facilitated the naming response, if there was only one exemplar of a given category in the response set and if distractors were presented visually. Auditory distractors, by contrast, did not yield facilitation but interference. This observation raises the question of why the present experiment revealed semantic facilitation even with auditory distractors. Hantsch etal. suggested that in a situation, in which some transformational process needs to be performed on the target to derive the response from the target’s entry-level representation, distractors presented in the same modality as the target are more likely to undergo such a transformation process as well, thus making the message congruency mechanism more effective. The fact that in the present study facilitation was observed even with distractors from a different modality might then be due to the fact that the task instruction manipulation employed here was more powerful in driving the message congruency mechanism than the response set manipulation in the study by Hantsch etal. This assumption is not unreasonable, given that the transformational process to be performed was an explicit part of the categorization instructions in the present experiment, while it was left implicit in the naming study by Hantsch etal. (2009). In any case, if the account just sketched is correct, the use of visual distractors should strengthen the facilitative component under both instructions and should thus modulate the semantic effects that have been observed in the present experiment under both instructions. Importantly, the facilitative component should still be larger under the categorization instruction than under the naming instruction, independent of the distractor modality. Thus, the interaction between task instruction and semantic relatedness that has been obtained in the present experiment should be replicated in its replication with visual distractor presentation.

ExperimentS1

ExperimentS1was an exact replication of the experiment reported in the article with visual distractors. If distractors from the same modality as the target promote the use of a message congruency mechanism and thus the strength of a facilitative component (as suggested by Hantsch etal., 2009), the semantic interference effect in naming should be attenuated, be eliminated, or even turn into facilitation, while the semantic facilitation effect in categorization should be enhanced.

Method

Participants

Seventy-two participants were tested. As in the experiment with auditory distractors, half of them received a naming instruction, and half of them received a categorization instruction. Again, participants with overall error rates exceeding 15% errors or with overall naming latencies exceeding 1,000ms were replaced (6 participants).

Materials

Same as in the main experiment, except that the auditory distractors were replaced with visual distractors. The visual distractor words were printed in black, using Arial font with a small light grey boundary around each letter. The initial capital letter reached a height of about 15mm on the screen. The width of the distractor words varied between 29mm and 108mm. Target pictures and distractor words were presented in the center of the screen with the distractors overlaying the pictures. Target pictures were again presented for 800ms. Visual distractors were presented either 100ms before picture onset (SOA –100ms), simultaneously with the target picture (SOA 0ms), or 100ms after picture onset (SOA 100ms). Distractors and targets were removed simultaneously from the screen.

Design and procedure

Design and procedure were identical to the experiment reported in the article.

Results and discussion

By applying the same criteria as in the experiment reported in the article, 278 observations (3.2%) were marked as erroneous and 199 observations (2.3%) as outliers. Again, these data points were discarded from the reaction time analyses.

Averaged reaction times were submitted to analyses of variance (ANOVAs). Statistical analyses involved the variables task instruction (naming vs. categorization), relatedness (semantically related vs. unrelated), and SOA (–100ms, 0ms, 100ms). Tables1 displays mean naming latencies and error rates for the two task instruction conditions per SOA, and distractor type.

In the analysis of response latencies, there was a main effect of task instruction, with slower responses in the categorization task as compared to the naming task, F1(1, 70)= 9.18, MSE= 27,467.49,p.01; F2(1, 19)= 88.56, MSE= 1,585.51, p.001. SOA was also significant, with response latencies increasing from SOA –100ms to SOA 100ms, F1(2, 140)= 23.15, MSE= 4,731.82,p.001; F2(2, 38)= 90.97, MSE= 688.37, p.001. Related distractors facilitated the naming response, F1(1, 70)= 41.13, MSE= 1,608.67,p.001; F2(1, 19)= 8.67, MSE= 4,356.93, p.01. Importantly, the facilitation effect was larger in the categorization task as compared with the naming task, leading to a significant interaction of task instruction and relatedness, F1(1, 70)= 14.79, MSE= 1,608.67,p.001; F2(1, 19)= 13.59, MSE= 959.62, p.01. Subsequent analyses revealed the 11–ms effect in naming to be only reliable in the participant analysis, F1(1, 35)= 5.55, MSE= 955.72,p.05; F2(1, 19)= 1.85, MSE= 1,739.89, p= .19, but the 40–ms effect in categorization to be highly significant in both analyses, F1(1, 35)= 37.43, MSE= 2,261.61,p.001; F2(1, 19)= 13.31, MSE= 3,576.66, p.01. No other effect reached significance. In particular, there was neither an interaction of task and relatedness, F1(2, 140)= 1.12, MSE= 1,034.77,p= .329; F2(2, 38)= 1.03, MSE= 698.73, p= .366, nor an interaction of task, relatedness, and SOA (Fs1).

In the analysis of error rates, the main effect of task instruction was significant in the item analysis only, F1(1, 70)= 2.64, MSE= 1.62,p= .11; F2(1, 19)= 4.98, MSE= 1.07, p.05, reflecting a trend toward more errors under the categorization instruction than under the naming instruction. The distribution of errors differed slightly across SOA, with fewer errors occurring at SOA 0ms than at the other SOAs, F1(2, 140)= 3.71, MSE= 1.21,p.05; F2(2, 38)= 7.55, MSE= 1.07,p.01. The main effect of relatedness was significant in the participant analysis only with semantically related distractor words leading to fewer errors than unrelated distractor words, F1(1, 70)= 3.71, MSE= 1.21,p.05; F2(1, 19)= 3.53, MSE= 1.19,p= .08. There was also a significant interaction of relatedness and SOA, F1(2, 140)= 3.95, MSE= 0.62,p.05; F2(2, 38)= 5.12, MSE= 0.86, p.05, reflecting the fact that fewer errors in the related condition were only observed at SOA –100ms, t1(71)= 3.41, p.01; t2(19)= 2.95, p.01; for the other SOAs ts1. No other effect reached significance.

The results from this experiment are in line with our prediction that the change in distractor modality from auditory (in the experiment reported in the article) to visual (in ExperimentS1) should have strengthened the influence of the message congruency mechanism (see Hantsch etal., 2009), and thus should have led to an attenuation of the interference effect in naming and an enhancement of the facilitation effect in categorization. In fact, the use of visual distractors shifted the effects from the experiment with auditory distractors reported in the article in a similar way in both instruction conditions (by 32ms in naming and by 23ms in categorization), leading to (weak) facilitation in naming and enhanced facilitation in categorization, see Fig.S1.

Joint statistical analyses of the naming latencies from the two experiments confirmed this pattern. There were significant interactions of distractor modality and relatedness, F1(1, 140)= 23.74, MSE= 1,675.60,p.001; F2(1, 19)= 25.10, MSE= 952.74, p.001, and of task and relatedness, F1(1, 140)= 36.64, MSE= 1.675.60,p.001; F2(1, 19)= 27.85, MSE= 1,2537.99, p.001. However, the interaction of distractor modality, relatedness, and task was not significant, Fs1, nor were the interactions of task, relatedness, and SOA, and of distractor modality, relatedness, task, and SOA, (all Fs1).

Taken together, in ExperimentS1 the interaction between task and relatedness was replicated. However, with visual instead of auditory distractor words there was an additive shift in the semantic effects towards (enhanced) facilitation for the two instruction conditions. This distractor modality effect replicates Hantsch etal. (2009) and emphasizes that a semantic net-effect as such does not inform us about the makeup of its individual components.

TableS1 Mean reaction times (in milliseconds) and error rates (as percentages) from ExperimentS1, broken down by task instruction, SOA, and distractor condition

SOA
–100ms / 0ms / 100ms / Overall
Distractor / M / % / M / % / M / % / M / %
Naming
SEM-REL / 707
(12) / 1.8
(0.6) / 746
(11) / 2.2
(0.5) / 744
(15) / 3.3
(0.9) / 735
(11) / 2.3
(0.4)
UNREL / 721
(13) / 2.8
(0.7) / 750
(12) / 2.1
(0.6) / 756
(14) / 4.0
(0.9) / 746
(11) / 2.9
(0.5)
Difference / –14
(6) / –0.9
(0.7) / –4
(8) / 0.1
(0.6) / –11
(7) / –0.7
(0.9) / –11
(4) / –0.6
(0.4)
Categorization
SEM-REL / 724
(12) / 2.8
(0.7) / 775
(13) / 2.4
(0.7) / 799
(11) / 4.0
(1.0) / 766
(9) / 3.1
(0.5)
UNREL / 772
(16) / 6.5
(1.1) / 810
(18) / 2.2
(0.8) / 835
(15) / 4.3
(1.1) / 806
(13) / 4.4
(0.6)
Difference / –48
(8) / –3.7
(1.2) / –35
(10) / 0.2
(1.0) / –36
(11) / –0.3
(1.0) / –40
(6) / –1.3
(0.6)

The standard error of each mean is given in parentheses. SOA, stimulus onset asynchrony; SEM-REL, subordinate-level name of target picture; UNREL, unrelated. Positive difference scores reflect interference, and negative difference scores reflect facilitation.

Fig.S1 Mean reaction time (RT) differences (related – unrelated, in milliseconds) and the corresponding standard errors, broken down by task for the experiment reported in the article and for ExperimentS1. The data are collapsed over SOAs. Positive scores reflect interference, and negative scores reflect facilitation.