Cognitive control processes reduce the effects of irrelevant or misleading information on performance. We report a study suggesting that effective cognitive control mechanisms are configured quickly during training. In a Stroop-like task, participants practiced naming abstract shapes using color words (with one shape called “red,” another called “blue,” etc.—these are referred to as “shape names”) In a subsequent test, naming the actual color of the shape was impaired when the color name and the shape name conflicted. Using regression analysis, we found that both the relative speed of basic shape and color naming processes and the amount of training on an individual shape made independent contributions to the amount of interference created. ERP recording in the same task revealed a larger frontal N200 component for participants who showed more behavioral interference.
Human and animal sensory systems are constantly bombarded with an overload of information, and environmental demands often require the rapid integration of many sensory stimuli in order to choose an appropriate response. This job is facilitated by selective attention, the situation-specific control over which stimuli and responses are fully processed. To the degree that selective attention fails, irrelevant information interferes with performance.
An important step in understanding how selective attention works is developing theories of how interference is created. One simple model of how interference is created is a ‘horserace’ model based on the processing speeds for relevant and irrelevant attributes of the stimulus. That is, the relative speed of basic perceptual/motor processing of relevant and irrelevant stimulus attributes predicts how much interference will be produced.
Evidence also exists that the relationship between relative processing speed and interference is not linear, but an inverted-U shaped function (Dyer, 1973). Faster or earlier processing of irrelevant stimuli does not always lead to increased interference. For example, if irrelevant features are processed sufficiently faster than relevant stimuli, it appears that the irrelevant response can be primed and then inhibited before the relevant processing reaches the critical stage where it is vulnerable to interference.
Another factor contributing to interference is the automaticity of stimulus processing—the degree to which the flow of information along a stimulus–response pathway is independent of controlled attention. Automaticity and relative speed may independently affect cognitive control, and thus measured interference. Alternately, an automaticity-based account may replace relative speed as an explanation for why interference occurs.
MacLeod and Dunbar (1988) used training in a Stroop-like paradigm in a study designed to investigate the effects of both relative speed and stimulus-response automaticity. Participants practiced naming irregular shapes with color names (e.g., “blue”). They were then tested on both naming the shapes when they appeared in colors (‘shape naming’) and on naming the colors in which shapes were presented (‘color naming’). MacLeod and Dunbar found that the interference created on each of these two tasks was equal after five days of practice, but that the relative speeds of the color naming and shape naming processes were not equalized until 20 days of practice. Based on their results, they suggested that automaticity, created by training, is a better predictor of interference than relative speed.
Day 1 2 3 4 5 6 7
A B C B C B C
Boxed conditions used in ANOVA
During shape naming, each of the four shapes was presented a different number of times, to allow the effects of number of practice trials and session to be assessed independently in a regression analysis. Number of practice trials naming each shape by the end of Session 7 varied between 0 and ~2600. The number of practice trials on a particular shape varied between participants.
Mean RT (ms)
Omnibus ANOVA results showed that congruence affected color naming latency. Colors with incongruent shapes were named more slowly than those with neutral or congruent shapes. There was no significant main effect of session, F < 1. The significant interaction between session and congruence, F(6,66) = 2.61, MSE = 854, p < .05, demonstrated that interference increased slightly with practice.
A planned comparison testing for a linear change in incongruent relative to neutral across days revealed a significant increase in interference with practice, F(1,11) = 5.96, MSE = 324, p < .05.
Few errors were made on the task. However, an analysis of accuracy rates yielded similar results, demonstrating that there was no substatial speed-accuracy tradeoff.
MacLeod and Dunbar suggested that training affects performance above and beyond the effects of relative processing speed. This analysis investigated whether training and relative speed independently contribute to interference.
What factors predict amount of interference?
Relative speed of shape naming and color naming
Days of training
Number of training trials on a particular shape
Effects of subject and shape entered as blocking factors
RTI - RTC
I is incongruent, C is congruent; RT = reaction time
Low interference group
High interference group
Low int. High int.
ERPs were recorded during the color naming task in an 8th session, after completion of the seven practice days. The graphs show a larger overall frontal negative shift (blue) in the N200 time window for participants who were successful at resolving interference.
N200 waveform (235-275 ms)
. Scalp distributions and waveforms for the color naming task, 235 – 275 ms after stimulus presentation. A,B: Scalp distributions for subjects who showed low and high behavioral interference effects, respectively. C: ERP waveforms during the ~n200 time window for each subject group. The thick solid line is neutral (circles), the dotted line is incongruent, and the thin solid line represents averages over congruent trials.
Boxed conditions used in ANOVA
Relative speed linear increase in interference.
Creation of interference
in color naming
Average improvement within training sessions
Average improvement between training sessionsRegression Results