Kornblum, S., & Lee, Ju -Wei. (1995). Stimulus-Response Compatibility with Relevant and Irrelevant Stimuli that do and do not overlap with the Response

This paper expands and revises some of the details of the architectural and processing aspects of the process model. It also reports the results of five four-choice experiments that used Types 1, 2, and 3 ensembles.  For this summary only the first two experiments are included.

Because S-R compatibility is determined by the interaction, and not by the independent effects of stimulus and response attributes, we used two sets of stimuli: letters, and left/right hand icons with index and middle fingers extended; there were also  two sets of responses: verbal letter names, and key presses.  Each stimulus set was combined with each of the two response sets, one of which it overlapped with, and the other with which it did not overlap.  This resulted in four ensembles: two with DO (hand icons/key presses, and letters/letters names), and two without DO (hand incons/letter names, and letters/keypresses).  Any performance differences between these ensembles, therefore, cannot be attributed to either the stimulus or the response set, since both are represented in both ensembles: the ensembles with DO and the one without DO.  Such difference must, therefore, be the result of the presence/absence of DO.  This is the underlying logic of the experimental design of these experiments.



The two sets of stimuli were visually presented.  In the case of the hand icon, the stimulus to be responded to was indicated by placing an asterisk in one of the extended fingers.  In the case of the letter stimuli these were presented on the screen.  For the ensemble with DO (i.e. letters/letter names & hand icons/keypresses), one mapping was congruent, the other incongruent. For the ensemble without DO, the mapping was random.


The ensemble with DO had a mapping effect of 203 ms. The ensemble without DO a had a mapping effect of 15ms. (which was barely significant).  These results join others that conform to the predictions of the model.  Given the logic of the design, the non-DO ensemble is an appropriate neutral condition for comparison, and according to the model should fall between the congruent and incongruent RT – which it does.



In terms of stimuli and responses, Experiment 2, with one exception, is identical to Experiment 1: instead of indicating the finger with an asterisk, the finger is indicated with a letter.  That is, regardless of condition, the stimulus always consisted of a hand icon with a letter in a fingertip. Like in Experiment 1, there were two response sets: key-presses, and letter names.  Depending on which stimulus dimension was relevant (finger or letter), the responses were to be made either according to the identity of the finger, making finger the relevant stimulus, (and to ignore the letter), or according to the letter on the finger tip, making letter the relevant stimulus (and ignore the finger). Each stimulus set was mapped onto each response set, thus yielding four SR ensembles of which two were Type 2, and two were Type3.

Mapping A associated a finger stimulus, that had been identified by a letter, to a key – the letter was to be ignored,.  Mapping C associated a letter that had appeared in one of the finger tips, to a letter name – the finger was to be ignored.  In mappings A and C the relevant stimuli overlapped with the responses, thus producing SR Ensembles Type 2, with DO. SR assignments were congruent or incongruent.

Mapping B associated a finger stimulus to a letter name – the letter was to be ignored.  Mapping D associated a letter stimulus that had appeared in a finger, to a finger – the finger was to be ignored.  With mapping B and D, the relevant stimuli did not overlap with the responses, but the irrelevant stimuli did, thus producing SR Ensembles Type 3. Two different SR assignments were used for mappings B and D.

In terms of relevant stimulus dimension, Experiments 1 and 2 are identical.  In terms of irrelevant dimensions they differ in that there were no irrelevant stimuli in Experiment 1.



The mean mapping effect (congruent minus incongruent RT) for overlapping ensembles (mapping A and C), is identical to what it was in Experiment 1: 203 ms.  The mean “mapping effect” (RT for mapping 1 minus RT for mapping 2) for non-overlapping ensembles (B and D), is zero.  In Experiment 1 it was 15 ms. and barely significant.  Both mapping effects thus confirm the predictions of the model, and replicate the data of our own and other experiments.  In addition, because of the choice of stimulus and response sets, these results clearly reflect the interaction between the stimulus and response sets and not the effects of these sets by themselves – which was the object of this experiment.

One of the model’s predictions is: Given an ensemble with overlap between the irrelevant dimensions and the response, the RT for SR consistent trials will be faster than for SR inconsistent trials – where “SR consistent” trials is the case where the irrelevant stimulus and the response have an attribute in common, and “SR inconsistent” trials is where no such common attribute exists.  For example, in this experiment “consistent” is when hand icon is the irrelevant stimulus, and key presses is the response to letters.  The RT for SR consistent trials is 50 ms. faster than for inconsistent trials, confirming the model’s prediction.

 Share on Facebook Share on Twitter Share on Reddit Share on LinkedIn

Kornblum, S. (1994). The way irrelevant dimensions are processed depends on what they overlap with: The case of Stroop-like, and Simon-like stimuli.

NOTE: This page is a short summary of the paper. The full text of the manuscript is not currently available online.

According to the model, the effects of irrelevant dimensions in Type 3 and Type 4 ensembles are mediated by two separate stages: the response production stage for Type 3, and the stimulus identification stage for Type 4.  The model postulates that these stages are additive (cf. Sternberg, 1969). If these overlap properties are now combined in a Type 7 ensemble, performance with Type 7 should be predictable from the individual performances with Type 3 and Type 4 ensembles.



The stimuli consisted of the colors blue, green, and grey, presented in the upper, lower, left, or right half of a rectangle.  There were four words: BLUE, GREEN, NOVEL, ELBOW, and a five-letter string.  Left/Right key presses were the responses.  The relevant stimulus was color; position, word, and letter string were all irrelevant.



Neutral (Type 1 ensemble): the colors were presented in the upper or lower half of the rectangle, the irrelevant words were NOVEL or ELBOW.

Simple SR overlap (Type 3 ensemble): colors were presented in the left or right half of the rectangle; words were NOVEL or ELBOW.  This combination produced two SR consistent and two SR inconsistent trials.

Simple SS overlap (Type 4 ensemble): Colors presented in the upper or lower half of the rectangle, words were BLUE or GREEN.  This produced two SS consistent and two SS inconsistent trials.

Composite SS/SR overlap (Type 7 ensemble):Colors were presented in the left or right half of the rectangle, the words were BLUE or GREEN.  This combination produced two SS/SR consistent trials, and two SS/SR inconsistent trials.  There were also two SS/SR “Hybrid Trials” in which the word conflicted with the color, but the side matched the response.



The nature of the irrelevant stimulus was indicated by a prime which preceded the actual stimulus by either 0 or 200ms.

For the irrelevant word, the word was presented in the rectangle before the color was.  For irrelevant positions, that half of the rectangle in which the color would eventually appear was filled with grey, which was subsequently replaced by a color.  Both irrelevant dimensions, word and position, were present in all trials and were displayed in the manner just described.



Simple SS overlap, & Simple SR overlap:

pure blocks (each for SS and SR) = 1/3 consistent, 1/3 inconsistent, and 1/3 neutral.

mixed blocks (SS mixed with SR) = 1/6 SS consistent, 1/6 SS inconsistent, 1/6 SR consistent, 1/6 SR inconsistent, 1/3 neutral.


Composite SS/SR overlap:

SS/SR overlap: 1/6 SS/SR consistent, 1/6 SS/SR inconsistent,

1/3 SS/SR hybrid trials – with half being SS consistent and the other half being SS inconsistent, 1/3 neutral.



In simple pure and mixed blocks, there was practically no consistency effect for SS overlap at lag zero; however, at lag 200 it was 48 ms.  In contrast, the consistency effect for SR overlap was 40 ms. at lag zero, and 32 ms. at lag 200.

In composite SS/SR blocks, the consistency effect for SR overlap at zero lag was 36 ms. and practically non-existent for SS overlap. However, at lag 200 ms. SS overlap had a consistency effect of 53 ms.    and SR overlap was 17 ms.


The results make it very clear that the processing of irrelevant stimulus dimensions differs greatly depending on whether they overlap with the relevant stimulus (Type 4), or with the response (Type 3). These differences also interact with lag, (the nature of this interaction is well accounted for by the computational form of the DO model).

 Share on Facebook Share on Twitter Share on Reddit Share on LinkedIn

Kornblum, S. (1992). Dimensional overlap and dimensional relevance in stimulus-response and stimulus-stimulus compatibility.

NOTE: This page is a short summary of the paper. The full text of the manuscript is not currently available online.

This paper begins by recapitulating some of the properties and processing principles of the DO model that were presented in the 1990 paper (Kornblum et al.1990), and then expands the number of ensemble in the taxonomy from 4 to 8.  It also includes data showing the effects of the number of alternatives in Type 1 and Type 2 ensembles.

Donders was the first to show that RT increases as a function of the number of alternatives (n); that increase is linear when calculated as a function of log n.  The data from ten different studies in the literature that used ensembles Types 1 and 2, show that: 1. the slope of that function (RT = log n) changes with different ensembles, and mappings – it is steepest when the mapping is random (i.e. no DO, Type 1), shallowest when the mapping is congruent (S-R DO, Type 2), and in between when the correct response could be identified by rule (DO Type 2).  These results are consistent with our model where, depending on the mapping instructions, the response identification process proceeds in one of three ways: 1. the identity rule, 2. a rule other than the identity rule, but a rule nevertheless (which depends on there being DO), and 3.searching through a list.  Fitts explained the S-R compatibility effects as the result of information being processed at different levels of efficiency, which is exactly what the automatic, and different identification processes in our model does.

Preliminary results from one of the experiment in Kornblum & Lee (1995) are also presented: with S-R DO (Type 2), there is a 203 ms mapping effect for the relevant stimulus, and a ~50 ms mapping effect for the irrelevant stimulus.  When there is no S-R DO (Type 1), there is no effect of mapping at all.

 Share on Facebook Share on Twitter Share on Reddit Share on LinkedIn

Oliver, L., & Kornblum, S. (1991). Dimensional overlap and population stereotype as joint predictors of stimulus-response compatibility.

NOTE: This page is a short summary of the paper. The full text of the manuscript is not currently available online.

Fitts argued that if two S-R ensembles have the same value on the  “population stereotype” scale, they would generate the same size S-R compatibility effects (i.e. RT difference between congruent and incongruent mapping).  According to the DO model, this need not be the case: two ensembles could have the same value of population stereotype, but differ in their degree of dimensional overlap.  In that case, because the compatibility effect is jointly determined at the set and at the element levels, the ensemble with the higher degree of dimensional overlap would generate the larger compatibility effect.  This is the proposition that was tested in this study.

Three stimulus and three response sets were combined to forms nine S-R ensembles.  The population stereotype and the level of dimensional overlap were obtained for each ensemble.  The population stereotype was obtained by asking subjects to give what they thought would be the best mapping between elements of the stimulus set onto elements of the response set. The dimensional overlap was obtained by asking subjects to compare pairs of S-R ensembles, and chose that pair that had the better matching stimulus and response sets. Dimensional overlap was quite strong for some ensembles, and population stereotype was strong in others, thus providing the precise conditions needed to test our proposition.

These nine ensembles were used to construct choice RT tasks with two mapping instructions.  One mapping was congruent, and corresponded to the population stereotype; the other was incongruent (and there were three different incongruent mappings).

The fastest mean RT was obtained for the congruent mapping, and as the degree of dimensional overlap decreased, the mean congruent RT increased.  The RT for incongruent mapping was, of course, longer and it increased with increasing degrees of dimensional overlap.  Both, the decrease in the case of the congruent mapping, and the increase in the case of the incongruent mapping, are in accord with the model.

Assuming that our measures of the level of the population stereotype, and the degree of dimensional overlap are valid, the results of this study confirm our proposition.

 Share on Facebook Share on Twitter Share on Reddit Share on LinkedIn

Kornblum, S., & Zhang, H. (1991). The Time Course of Automatic Response Activation Process in Stimulus-Response Compatibility.

NOTE: This page is a short summary of the paper. The full text of the manuscript is not currently available online.

Using a four-choice task, this study was designed to test whether: 1. the order of RT that is predicted by the model could be confirmed, i.e. RT was fastest for congruent mappings, slowest for incongruent mappings, and in between for neutral; and 2. in the case of ensembles with S-R overlap, can the automatic activation of the corresponding response element by the presentation of a stimulus element, as is postulated by the model, be demonstrated.

Result 1.  The results confirmed the order of RT’s as predicted by the model.

Result 2.  By reducing the time available to process and execute the response, thus increasing the likelihood of errors, the results showed that when the mapping was incongruent most of the erroneous responses were response that would have been correct had the mapping been congruent, thus demonstrating the activation of the response that would have been correct, had the mapping been different.  However, by reducing the time available to abort, reprogram, etc. which is called for when the mapping is incongruent (ergo generating errors), automatic activation of the response was verified.  Thus, this study confirmed one prediction, and verified an important assumption of the model.

 Share on Facebook Share on Twitter Share on Reddit Share on LinkedIn

Kornblum, S., Hasbroucq, T., & Osman, A. (1990). Dimensional Overlap: Cognitive Basis for Stimulus-Response Compatibility – A model and taxonomy

1990 is the year the dimensional overlap model (DO) first saw the light of day.  Paul Fitts (Fitts and Seeger, 1953; Fitts & Deininger, 1954) had identified a family of tasks in which performance could not be accounted for by either the stimulus properties alone, nor the response properties alone – the results were clearly the outcome of an interaction between the two.  He called the phenomenon “stimulus response compatibility effect”.

We proposed that this interaction occurred at two levels: the first, where categorical stimulus and response sets are perceived as corresponding or not.  We called the basis of this correspondence the dimensional overlap between the sets.  Dimensional overlap was defined as the degree to which attributes in the two sets are perceptually, structurally or conceptually similar.  It is, therefore, a property of the mental representation of sets.  When dimensions in a stimulus and response set overlap, a second level comes into play: the instructions that map individual stimulus elements onto individual response elements.

This early form of the model already included a taxonomy with four ensembles – which was later expanded to the final eight-task taxonomy in Kornblum (1992), It had a rudimentary architecture, and the outlines of a processing model that combines automatic and control mechanisms.

 Share on Facebook Share on Twitter Share on Reddit Share on LinkedIn