Developing Intelligence

Inhibitory Decline In Childhood?

According to artificial neural network models that implement lateral inhibition, activation and inhibition are two sides of the same coin. These models often assume that patterns of activation compete with one another. In other words, in a given space of neural tissue (or “layer” in network terms), some particular pattern will become most active. This pattern will effectively suppress the emergence of other patterns through lateral inhibition; other patterns are less likely to emerge if the current dominant pattern is very strongly represented.

This assumption of “competition for representation” makes a general prediction: interference from irrelevant representations is less likely in those networks with increased representational strength. To use a very literal example: people who can respond more quickly to choice reaction time tasks (e.g., press the lighted button) may suffer less from interference in tasks where they must overcome a prepotent response (e.g., naming the ink color of words like GREEN, as in the Stroop task.)

This is the implicit logic of a study by Band, van der Molen, Overtoom & Verbaten from the Journal of Experimental Child Psychology. The authors reasoned that tasks involving interference might show the same developmental trends as tasks that simply involve the speed of making a response. If true, this would suggest that activation and the resolution of interference (frequently called inhibition) involve the same mechanisms.


To test this idea, the authors gave 6 tasks to 69 subjects, with median ages spanning 5 years to 21 years of age. Here are the tasks used (casual readers may wish to skip this section):

In the auditory simple RT task, reaction time was measured in response to an auditory tone, which followed a “get ready” warning stimulus. In the visual choice RT task, the reaction time of subjects to press a button corresponding to the side of the screen at which a visual stimulus appeared. An “arousal task” was the same as the visual choice RT task, except that a task-irrelevant auditory tone (varying between 3 volumes) was presented sometime after the visual stimulus (with 3 varying latencies, calculated as the 20th, 50th and 80th percentiles of RT on the visual choice task). The logic here is that subjects should be ignoring the auditory tone, so as to prevent themselves from making errors on the task; the degree to which reaction time is different in the presence of auditory tones relative to those without auditory tones reflects the ability to resist this interference. Subjects also completed a Simon task, which was the same as the arousal task, except that the task-irrelevant sound was played to both ears (as before), or to just the right or the left ear alone. The logic here is that subjects should be ignoring the auditory tone; for example, reaction time to press the left key should not differ depending on whether the auditory tone was played to both ears, the right ear alone (the “incongruent” ear) or the left ear alone (the “congruent” ear). The “stop all” task was the same as the visual choice RT task, except that if subjects heard a sound (presented on 30% of trials) they were to refrain from making a choice. The latency of the sound was varied using an adaptive tracking algorithm to arrive at three different values: the latencies with which subjects can correctly withhold a response 33%, 50%, and 67% of the time. Finally, a “stop change” task was identical to the “stop all” task, but required subjects to make the incorrect response after an auditory sound, rather than withholding their response.

The results showed that reaction times improved across ages in a way that was proportional to their means. In other words, the tasks on which children were the slowest showed the steepest change across time. Band et al. interpreted this to mean that a single, global mechanism is responsible for the developmental change in response “activation.”

The authors then conducted a more detailed analysis of reaction times to uncover developmental changes in the ability to resist interference (i.e., interference from response-irrelevant auditory tones, or from auditory tones that were actually contradictory or incongruent with the response that should have been given on that trial.) This is where the results get complicated:

Most cognitive abilities seem to increase with age – however, younger subjects were less affected by the loudness of an irrelevant auditory sound, both in terms of accuracy and reaction times! Band et al. suggest the loud sounds had less of an effect on children because they were simply less ready to give a response in the first place (but it’s also possible that they were reorienting to these sounds more slowly).

Similarly, younger subjects were slowed less than older subjects by sounds heard in the ear opposite to the correct response, relative to sounds coming from both ears! Whereas this slowing disappeared in older age groups if the sound was played longer after the onset of the visual stimulus, it remained in the youngest children for longer delays. Band et al. suggest that the interference effect of this sound persisted for longer in children, but it is also possible that they were reorienting less efficiently to this sound (and thus showed a more prolonged timeline of interference).

Consistent with this emerging pattern, all age groups were statistically equivalent in terms of the delay of a “stop signal” (after a visual stimulus) required for 50% correct “stopping.” In fact, there was a nonsignificant increase in this value between 5 and 8 year-olds! Band et al. suggest this is inconsistent with the claim that inhibitory abilities increase with age.

However, there was a age-related improvement in the reaction times when the same sound instead signaled that the opposite response should be given. In this case, it took children longer to reverse their response than adults (although statistically this result was only marginally significant).

In addition, performance on tasks involving resistance of interference were not significantly correlated with one another, with one exception: the stop signal delay required for 50% correct “stopping” was strongly and significantly correlated with the amount of interference caused by a loud sound, relative to a quieter sound, on the visual choice RT task. Band et al. suggest that this correlation could suggest a global inhibitory mechanism, or “emergency brake.” However, it’s not clear why such a mechanism would not result in stronger correlations between the other tasks. An alternative explanation is that those who are more prepared to give a response are less likely to be able to refrain from giving that response and are more likely to give that response if a loud sound is played at the same time [if so, the correlation between these should disappear, or reverse direction, after covarying individual RTs on the simple or choice reaction time tasks.]

In summary, these findings challenge the idea that “inhibitory control” (aka resistance to interference) is a monolithic entity, or that it changes between these ages.

At the same time, it is also not entirely consistent with the neural network-inspired idea that inhibitory control is related to the speed or strength of response representation; based on that theory, one might have expected age-related trends in inhibitory control to look just like the age-related trends in response activation.

Unfortunately, more general forms of this idea were not strongly tested. An obvious step is to correlate individual performance on the response activation tasks with that on the inhibitory control tasks. Another possibility is that those who respond more quickly are more likely to be affected by interference for pragmatic reasons (i.e., there is a representation to suffer from interference). If simple response speed were taken into account, a positive relationship might then emerge between resistance to interference and choice reaction time.

Related Posts:
fMRI of the Stop Signal Task: What Computations Support Stopping?
Inhibition and rTMS
Stop vs. Go: Reexamining the Race Horse Model of Inhibition
Response Inhibition and RT Variability: One and the Same?