Three years ago, I bought a new car that had a keyless push start. It took almost six months of driving before I stopped trying to put my key into the ignition. Each time I tried to put an imaginary key into the imaginary ignition slot, I shook my head and laughed at myself. My unconscious, or implicit memory system, was clearly strongly active and it took some time to rewire my procedural motor programs. After months of explicitly reminding myself to push the button to start the car, it finally felt “normal” to start the car without a key in the ignition.

Choosing to be Controlled?

The distinction between implicit and explicit learning has been investigated in a wide range of settings, including knowledge acquisition, motor actions, and stereotypes. Whether it is learning how motor actions like playing a piano move from a conscious state of slow, methodical steps to an unconscious state of rapid, fluid and fluent behavior, or discovering why a stroke victim cannot yawn intentionally but can catch a yawn unconsciously, the role of explicit memory vs implicit memory in behavior and the underlying neuroscience is fascinating.

Discriminate This, Discriminate That

One way to measure explicit learning vs implicit learning is through a categorical discrimination procedure. Basically, a categorical discrimination task asks the learner to categorize test stimuli into different categories based on a rule they have been given. Take for example the image below with the star and the truck bordered in blue or red on the left. The image is borrowed from a paper by Ling and colleagues (2021).

When working with humans, the rule can be explicitly stated, such as “select the stimuli with the blue border” or it can be learned implicitly through immediate feedback, such as when the individual selects a blue stimulus correctly, a reinforcement is received or if a red one is selected, a punishment or no reinforcement is received.

Sometimes, a match-to-sample procedure can be used where a “target” or “sample” like a blue box is presented and the individual must select something blue from the choices provided to receive reinforcement.  (This variation is not shown in the figure above.) When stimuli vary on multiple dimensions but only one dimension, such as color, is needed to categorize the stimuli correctly, rule-based learning is in play and can be learned very rapidly by humans.

In a variation of the categorical discrimination task, stimuli may vary on two dimensions (color and shape, such as the right panel above) and now a learner has to determine the correct choice with both pieces of information. This type of task is called an information-integration discrimination and is learned more slowly by humans.

Spinning in Circles

Sometimes, discrimination tasks require more ambiguous discriminations between stimuli, such as when a target color is blue, but the options that are provided are more red than blue or more green than blue. For example, Pavlov gave a similar discrimination to his dogs but with circles and ovals. When ovals were closer to circles, the dogs had great difficulty in discriminating between them. This confusion led to distress and aggressive behavior, which Pavlov called “experimental neuroses.”

To accommodate these ambiguous stimuli, some categorical discrimination tasks provide a third option, called an “uncertain” category. When this third option is provided, researchers argue that the individual must cognitively evaluate the available information at a level above an associative rule, such as choose blue when it is red. The extra processing, such as when an individual thinks “This red blue option is close to blue but it isn’t exactly blue because there is red in it, what do I do?” requires explicit cognition or metacognition. The third choice, “I am uncertain,” allows the learner to make a selection when truly conflicted.

The question of interest in the explicit vs implicit learning debate is whether or not these two processes can be separated using different discrimination procedures. If we can distinguish between these two forms of learning with a task or two, we could then test other species with the same tasks to determine how those species process information; knowledge that may further advance our understanding of human learning.

Choices Matter….Eventually

Recently, another procedure was introduced as a way to dissociate explicit learning from implicit learning. Known as the “1-back procedure” or “displaced reinforcement,” this procedure provides a consequence in a delayed fashion. That is, when an individual correctly categorizes a stimulus on a trial, reinforcement is given after the next trial is completed (right or wrong) instead of on the trial that the correct choice was made. As all good learning theorists know, immediate reinforcement is the fastest and easiest way to learn and when there is a delay, learning takes a little longer. According to Smith who implemented the 1-back procedure with his colleagues, by displacing the reinforcement to the next trial, instead of immediately after the correct choice, the implicit discrimination process, or learning through association, was disrupted and learners had to invoke an explicit discrimination process that used metacognition to complete the task successfully.

Bird Brains?

A recent paper by Alexandra Nosarzewska, Daniel Peng, and Thomas Zentall published in the Psychonomic Society’s journal Learning and Behavior, questioned the premise of the 1-back procedure and its ability to dissociate between implicit and explicit learning.

Nosarzewska and her collaborators wished to test pigeons on the 1-back conditional discrimination because pigeons are known to be facile and implicit learners, and they solve information-integrated discriminations as easily as they solve rule-based discriminations. If the pigeons were able to learn the displaced reinforcement and Smith and his colleagues were correct in their interpretation, then pigeons would be capable of explicit learning, a new finding.

To test the abilities of the pigeons, three experiments were conducted with the 1-back reinforcement discrimination procedure. The first experiment used six White Carneau pigeons that had been tested on a simultaneous color discrimination task with red and green hues previously.

With this experience, the current study utilized a match-to-sample procedure in which the sample color (trial 1 center red circle in image below) was shown and then comparison stimuli were provided for the pigeon to select.

If the pigeons selected the sample color from the comparison stimuli (second red circle in Trial 1 above) correctly, they received reinforcement for their correct choice on Trial 1 in Trial 2 (Rf), regardless of their performance on that trial. In Trial 2 of the figure above, the bird incorrectly chose the red circle instead of the green circle and so reinforcement was not received (no Rf) on Trial 3 for the bird’s choice in Trial 2.

Three out of the six birds in Experiment 1 figured out the 1-back procedure after 30 sessions. The remaining three birds did not perform above chance in the first 30 sessions as they showed strong position preferences. To correct these preferences, the researchers instituted another round of training with corrections for incorrect choices. If a bird had an incorrect response, the trial was repeated until the bird made the discrimination correctly. All six birds successfully learned the 1-back paradigm after this additional training.

Experiment 2 repeated Experiment 1 with eight different pigeons but started the training with the correction procedure in place from the beginning. The results of this study showed that all but one bird successfully learned the task, which reinforced the conclusion that pigeons can learn the 1-back paradigm.

In Experiment 3, the researchers decided to make the discrimination tasks a little harder by having the pigeons form an association between two different colors: a yellow sample required a red comparison choice to be correct, whereas a blue sample required a green comparison choice to be correct. This change forced the birds to form an arbitrary association or rule between two colors that were not the same. Once again, 11 different birds succeeded on this task.

Are Birds Thinking or Doing?

Nosarzewska and her colleagues suggested that these findings could be evidence of explicit learning, if the explanation proposed by Smith and colleagues in their original study of the 1-back discrimination procedure was correct. However, a question still remains as to whether explicit learning was the most parsimonious explanation given that evidence for associative learning with delayed reinforcement has been shown with pigeons previously.

To solve this question, additional research is needed. It seems likely the next steps to do so will involve tracking the brain areas utilized during this task. Thus far, research has shown that the basal ganglia and/or striatal system are implicated in implicit learning tasks while other areas of the brain are active in explicit learning. Perhaps targeting these different brain structures will help to answer whether birds are thinking or simply doing.

Featured article 

Nosarzewska, A., Peng, D. N., & Zentall, T. R. (2021). Pigeons acquire the 1-back task: Implications for implicit versus explicit learning. Learning & Behavior, 1-10. https://doi.org/10.3758/s13420-021-00468-3.