Don't Bite: A Cognitive Primer

II.  Cognitive Control and Neural Architecture

This is the second in a series of posts on "delay of gratification."
Access #1

In the developmental literature, cognitive control is a technical term that denotes the ability to flexibly choose one response over another, based on goals or context.  This may sound suspiciously like good old-fashioned “self control,” but there’s a hitch:  when psychologists talk about whether or not a child has cognitive control, they're referring to a particular stage of cortical development and a particular type of executive function.  Turns out that the ability to flexibly choose between responses typically doesn't develop in children until around the age four, and it’s related to the developmental trajectory of their frontal lobes – specifically, a bit of brainware called the prefrontal cortex, or PFC [1].

The classic test of cognitive control is the dimensional change card sort task or DCCS, for short.  It's simple: you give a three-year old some cards with pictures on them – let's say, red and blue trucks, and red and blue flowers.  Then you set up two sorting bins.  One has a picture of a red truck, and one has a picture of a blue flower.  At the start of the game, you ask the kid to sort the cards by shape.  “Can you put the trucks in the truck bin, and the flowers in the flower bin?”  So far, so good: the kid nails it.  But then, you switch up the rules, and ask the kid to sort the cards by color.  “Can you put the blue cards in the blue bin, and the red cards in the red bin?”  The kid nods and repeats after you.  And then – promptly fails, and continues sorting by the first rule [2].

In recent years, psychologists have advanced quite a number of competing theories to account for why this happens [3].  The general consensus, however, is that children under the age of four have an immature – and for the most part, offlineprefrontal cortex (PFC) [4].  The PFC is the command center of the brain, so without it in play, it’s next to impossible for children to electively control or inhibit their responses.

This handicap plays out quite clearly in the DCCS: once toddlers have grown accustomed to responding in a particular way, they have a great deal of difficulty switching to a new kind of response.  Even though they understand that they’re being asked to switch, and even though they understand what the new rule is supposed to be, they can’t override the default response their brain has locked into.  This is why we say that three-year olds lack cognitive control.  Their wetware is simply not allowing them to flexibly switch between responses based on new information and context.  They’re unable to willfully reallocate their attention or redetermine their responses – it’s as if their brain got stuck on auto-pilot [5].

Speaking more broadly, this means that absent specific strategies for engaging with a particular task, toddlers can’t really choose what they’ll attend to or how they’ll respond. When it comes to making a quick decision, they’ll simply go with the most obvious choice that leaps out at them.

But now you're getting impatient.  That’s just not true, you insist.  OK, so maybe there's a bit of machinery offline, but I've known three-year olds who can... (and now you rattle off the heroic feats of the most impressive three year old the world has yet known – a direct descendant, no doubt)

Ah, but dear Sherlock, you’ve missed the point!  This is where it becomes vital that we distinguish between what children are capable of cognitively – what their neural architecture supports – and what they are capable of given what they know.  For example:  most three-year olds can't pass the DCCS on the first go-round, or even after extensive training on the task.  But, if you take a task that they're already good at – like naming shapes and colors – and have them do a naming switch-task ("Which color is this card?  Now, which shape is this card?"), it's so easy, they laugh at you.  (Literally.  I have been laughed at by a three-year old).

Why? –Well, it's obvious isn't it?  They have loads of practice with naming, but very little practice with sorting.  Which is just another way of saying that learning and practice matter. Well-learned responses will be more distinct from one another, and better embedded in context, reducing the need for the kind of active filtering that the PFC supports. The problems arise when a kid is faced with a task she’s never encountered before -- she will need to call on her "command center" far more when she has to cope with the demands of a novel task, because it necessitates interacting with the world in real time.  ("I was doing this—my brain is going to keep doing it—even though I'm being told to switch—uh... what was I switching to...?"  CANNOT COMPUTE --> ERROR)  But when it comes to switching between tasks she’s already familiar with, suddenly she’s "no-problem, Bob."

It helps to remember that a child’s behavior is a product not only of cognitive functioning and maturity, but also of learning, which provides a contextual scaffolding that enables children to reason about and respond to their environments even in the absence of an operative command center. (It's why the prodigy three-year old you mentioned earlier somehow isn't running around like a headless chicken.  Turns out, he can learn good).

Fortunately for all involved, the plane-without-a-pilot phase is short lived: the command center does eventually go live.  In the months following their third birthday, children become better and better at classic cognitive control tasks like the DCCS, until they’re reliably passing muster around age four or five [6]. There are, however, individual differences in executive function and development that may persist into adulthood [7].

How does all this relate back to cookies? In tomorrow’s report from the frontlines, we’ll investigate how scientists have theoretically framed the cookie task, and look at how this fits with what we know about both learning and cognitive control.

[This is the second in a five-part series that is to be continued…]

The Daily Fact Check

[1]  Recommended reading on prefrontal development: Thompson-Schill, Ramscar & Evangelia, 2009 and Ramscar & Gitcho, 2007.  For a pop science write-up, you can also check out my SciAm Mind Matters Post: “The Advantages of Being Helpless.”

[2]  See the Nature protocol paper: Zelazo, 2006.

[3]  You think I’m kidding?  Try these out for starters: Zelazo & Frye, 1997; Perner, Stummer & Lang, 1999; Yerys & Munakata, 2001; Zelazo, Müller, Frye & Marcovitch, 2003; Kirkham, Cruess & Diamond, 2003; Diamond, Carlson & Beck, 2005; Muller, Dick, Gela, Overton, & Zelazo, 2006; Kloo, Perner, Kerschhuber, Dabernig, Aichhorn, 2007.  Everyone has their two cents to add to the conversation, and it’s not always entirely clear where they complement or contradict one another – believe me, I did a literature review on this thought-salad!

[4] In addition to the work I reference in [1], here are a bunch of other folks relating DCCS performance to prefrontal development: Moriguchi & Hiraki, 2009; Alman et al., 2005; Diamond, 2002; Rueda et al., 2005; Posner, 2005; Brooks, Hanauer, Padowska & Rosman, 2003; Deák, 2003; Alman et al., 2001.  Adele Diamond is someone who is particularly worth checking out; she writes very accessibly for the non-scientist and is one of the pioneers of developmental neuroscience.

[5]  If you’re curious to learn more about this, we currently have a paper under review that discusses this in some detail: Ramscar, Dye, Gustafson & Klein (email me if you’d like to take a gander at the manuscript).

[6]  When I say “goes live,” I don’t mean it goes from 0 to 60 in three seconds flat.  Although major changes in prefrontal functioning are evident around age four, most researchers belive that the PFC develops gradually, and keeps right on developing into the early twenties.

[7] There is some fascinating research suggesting that premature overgrowth of the frontal lobes is linked with autism. At the opposite end of the spectrum, delays in prefrontal maturation may be linked to ADHD. In addition to these extremes, there is also individual variation within the neurotypical population.

4 responses so far

  • Grep Agni says:

    Could you clarify this sentence:

    Turns out that the ability to flexibly choose between responses typically begin to show in children until they are around age four, ....

    Does this mean the ability is complete by age four, or that they do not advance in ability after they turn four, or something else? Even after reading the penultimate paragraph I'm a bit confused.

  • DrugMonkey says:

    After you are done with this series, how about a bit of meta on the way the cognitive construct can come to dominate over an appropriately limited view of the research circumstances. One of my favorites is the old parallel play business. Perhaps you can take that up at some later date?

    Point being that there is a tendency to talk about innate capability at certain developmental stages when the data are based on very specific laboratory circumstances that might be confusing or highly artificial. (Comparative cognition fields suffer from this as well except they have the extra special problem that they do not have decent sample sizes or good randomization or naive subjects...perhaps you can discuss the parallels JGG?)

    • Jason G. Goldman says:

      DM: I think I get what you're asking, but not quite sure. What do you mean by "the way the cognitive construct can come to dominate over an appropriately limited view of the research circumstances"?

      What I think you mean is that a series of studies are conducted which are possibly highly replicable because the experimental conditions can be precisely specified and subjects respond in a predictable way, but it is the experimental design itself that causes the results to be a certain way, instead of the experiment reflecting an underlying cognitive construct.

      If this is the issue, I actually have a series planned for later this month on Piaget's classic A-not-B error, which is one of the most highly cited and replicable studies in developmental psychology, but the results are entirely dependent on the experimental design.

  • melodye says:

    @Greg Agni : Oops, thanks for pointing out the typo! Duly updated. And let's see if I can't explain this better.

    When I say "the ability to flexibly choose between responses," I mean the ability to navigate between two (or more) competing responses based on goals or context. In children, this is usually assessed by testing whether or not they can complete tasks like the DCCS. The DCCS is thought by many developmentalists to be a classic benchmark of cognitive control (and this has been supported in neural imaging studies, c.f. Moriguchi & Hiraki's 2009 PNAS article). In any case, this means that once children are reliably passing it, it seems they've passed a particular milestone.

    However, passing that milestone is certainly not the end state -- the PFC will undergo major changes during adolescence, on its developmental trajectory. In any case, we can actually still measure executive functioning in older children and adults -- and track its comparative advance -- by looking at reaction-time speed in cognitive control-tasks. For example, Diamond & Kirkham (2005) found that the drop-off in accuracy for children that you see after they're asked to switch rules, is mirrored by a reaction time drop-off in adults. A host of other experimentalists use more advanced tasks like go / no-go and the Stroop task to plumb executive function.

    The Stroop task, which you may be familiar with from Internet emailings, is probably the most common assessor of PFC function in teens and adults. The task involves naming the ink color of a contrasting color word: for example, you might see the word “red” written in green ink, in which case you have to say “green.” The task is tricky since it demands that we override a well learned response (saying “red” in response to seeing the word red) with a new response specific to the task (naming the conflicting ink color). Tricky or not, healthy adults can usually complete the task accurately with only minor hesitation. However, the *key* is that we can assess their comparative function by looking at how fast they respond.

    So what's the short answer? :

    The ability to complete the DCCS is usually seen around age four, which marks an important developmental milestone. But the circuitry underlying that "ability" will continue to develop for many years to come. To assess prefrontal function in older children, teens and adults, speed becomes the useful measure.