7.4 Implicit Learning: Under the Wires

Ten years ago, no one knew how to type using thumbs; now just about all teenagers do it automatically.

AP PHOTO/MARY ALTAFFER

It’s safe to assume that people are sensitive to the patterns of events that occur in the world around them. Most people don’t stumble through life thoroughly unaware of what’s going on. Okay, maybe your roommate does. But people usually are attuned to linguistic, social, emotional, or sensorimotor events in the world around them so much so that they gradually build up internal representations of those patterns that were acquired without explicit awareness. This process is often called implicit learning, or learning that takes place largely independent of awareness of both the process and the products of information acquisition. Because it occurs without awareness, implicit learning is knowledge that sneaks in “under the wires.”

Habituation, which we discussed at the outset of the chapter, is a very simple kind of implicit learning where repeated exposure to a stimulus results in a reduced response. Habituation occurs even in a simple organism such as Aplysia, which lacks the brain structures necessary for explicit learning, such as the hippocampus (Eichenbaum, 2008; Squire & Kandel, 1999). In contrast, some forms of learning start out explicitly but become more implicit over time. When you first learned to drive a car, for example, you probably devoted a lot of attention to the many movements and sequences that needed to be carried out simultaneously (“step lightly on the accelerator while you push the turn indicator and look in the rearview mirror while you turn the steering wheel”). That complex interplay of motions is now probably quite effortless and automatic for you. Explicit learning has become implicit over time. These distinctions in learning might remind you of similar distinctions in memory and for good reason. In the Memory chapter, you read about the differences between implicit and explicit memories. Do implicit and explicit learning mirror implicit and explicit memory? It’s not that simple, but it is true that learning and memory are inextricably linked. Learning produces memories, and conversely, the existence of memories implies that knowledge was acquired, that experience was registered and recorded in the brain, or that learning has taken place.

Cognitive Approaches to Implicit Learning

Interest in implicit learning among psychologists was sparked when researchers began to investigate how children learned language and social conduct (Reber, 1967). Most children, by the time they are 6 or 7 years old, are linguistically and socially fairly sophisticated. Yet most children reach this state with very little explicit awareness that they have learned something, and with equally little awareness of what it was they have actually learned. As an example, although children are often given explicit rules of social conduct (“Don’t chew with your mouth open”), they learn how to behave in a civilized way through experience. They’re probably not aware of when or how they learned a particular course of action and may not even be able to state the general principle underlying their behavior. Yet most kids have learned not to eat with their feet, to listen when they are spoken to, and not to kick the dog.

To investigate implicit learning in the laboratory, early studies showed research participants 15 or 20 letter strings and asked them to memorize them. The letter strings, which at first glance look like nonsense syllables, were actually formed using a complex set of rules called an artificial grammar (see FIGURE 7.18). Participants were not told anything about the rules, but with experience, they gradually developed a vague, intuitive sense of the “correctness” of particular letter groupings. These letter groups became familiar to the participants and they processed them more rapidly and efficiently than the “incorrect” letter groupings (Reber, 1967, 1996).

Figure 7.18: Artificial Grammar and Implicit Learning These are examples of letter strings formed by an artificial grammar. Research participants are exposed to the rules of the grammar and are later tested on new letter strings. Participants show reliable accuracy at distinguishing the valid, grammatical strings from the invalid, nongrammatical strings even though they usually can’t explicitly state the rule they are following when making such judgments. Using an artificial grammar is one way of studying implicit learning (Reber, 1996)

Take a look at the letter strings shown in Figure 7.18. The ones on the left are correct and follow the rules of the artificial grammar; the ones on the right all violated the rules. The differences are pretty subtle, and if you haven’t been through the learning phase of the experiment, both sets look a lot alike. In fact, each nongrammatical string only has a single letter violation. Research participants are asked to classify new letter strings based on whether they follow the rules of the grammar. People turn out to be quite good at this task (usually they get 60–70% correct), but they are unable to provide much in the way of explicit awareness of the rules and regularities that they are using. The experience is similar to coming across a sentence with a grammatical error. You are immediately aware that something is wrong and you can certainly make the sentence grammatical, but unless you are a trained linguist, you’ll probably find it difficult to articulate which rules of English grammar were violated or which rules you used to repair the sentence.

Other studies of implicit learning have used a serial reaction time task (Nissen & Bullemer, 1987). Here research participants are presented with five small boxes on a computer screen. Each box lights up briefly, and when it does, the participant is asked to press the button that is just underneath that box as quickly as possible. Like the artificial grammar task, the sequence of lights appears to be random, but in fact it follows a pattern. Research participants eventually get faster with practice as they learn to anticipate which box is most likely to light up next. But, if asked, they are generally unaware that there is a pattern to the lights.

Implicit learning, which is involved in acquiring and retaining the skills needed to ride a bicycle, tends to be less affected by age than explicit learning.

STEVE SMITH/PURESTOCK

Implicit learning has some characteristics that distinguish it from explicit learning. For example, when asked to carry out implicit tasks, people differ relatively little from one another, but on explicit tasks (such as conscious problem solving), they show large individual-to-individual differences (Reber, Walkenfeld, & Hernstadt, 1991). Implicit learning also seems to be unrelated to IQ: People with high scores on standard intelligence tests are no better at implicit learning tasks, on average, than those whose scores are more modest (Reber & Allen, 2000). Implicit learning changes little across the life span. Researchers discovered well-developed implicit learning of complex, rule-governed auditory patterns in 8-month-old infants (Saffran, Aslin, & Newport, 1996). Infants heard streams of speech that contained experimenter-defined nonsense words. For example, the infants might hear a sequence such as “bidakupadotigolabubidaku,” which contains the nonsense word bida. The infants weren’t given any explicit clues as to which sounds were “words” and which were not, but after several repetitions, the infants showed signs that they had learned the novel words. Infants tend to prefer novel information, and they spent more time listening to novel nonsense words that had not been presented earlier than to the nonsense words such as bida that had been presented. Remarkably, the infants in this study were as good at learning these sequences as college students. At the other end of the life span, researchers have found that implicit learning abilities extend well into old age and that they decline more slowly than explicit learning abilities (Howard & Howard, 1997).

Implicit learning is remarkably resistant to various disorders that are known to affect explicit learning. A group of patients suffering from various psychoses were so severely impaired that they could not solve simple problems that college students had little difficulty with. Yet those patients were able to solve an artificial grammar learning task about as well as college students (Abrams & Reber, 1988). Other studies have found that profoundly amnesic patients not only show normal implicit memories, but also display virtually normal implicit learning of artificial grammar (Knowlton, Ramus, & Squire, 1992). In fact, they made accurate judgments about novel letter strings even though they had essentially no explicit memory of having been in the learning phase of the experiment! In contrast, several studies have shown that dyslexic children, who fail to acquire reading skills despite normal intelligence and good educational opportunities, exhibit deficits in implicit learning of artificial grammars (Pavlidou, Williams, & Kelly, 2009) and motor and spatial sequences on the serial reaction time task (Bennett et al., 2008; Orban, Lungu, & Doyon, 2008; Stoodley et al., 2008). These findings suggest that problems with implicit learning play an important role in developmental dyslexia and need to be taken into account when developing remedial programs (Stoodley et al., 2008).

Implicit and Explicit Learning Use Distinct Neural Pathways

The fact that individuals suffering amnesia show intact implicit learning strongly suggests that the brain structures that underlie implicit learning are distinct from those that underlie explicit learning. As we learned in the Memory chapter, amnesic individuals are characterized by lesions to the hippocampus and nearby structures in the medial temporal lobe; accordingly, these regions are not necessary for implicit learning (Bayley, Frascino, & Squire, 2005). What’s more, it appears that distinct regions of the brain may be activated depending on how people approach a task.

For example, in one study, participants saw a series of dot patterns, each of which looked like an array of stars in the night sky (Reber et al., 2003). Actually, all the stimuli were constructed to conform to an underlying prototypical dot pattern. The dots, however, varied so much that it was virtually impossible for a viewer to guess that they all had this common structure. Before the experiment began, half of the participants were told about the existence of the prototype; in other words, they were given instructions that encouraged explicit processing. The others were given standard implicit learning instructions: They were told nothing other than to attend to the dot patterns.

The participants were then scanned as they made decisions about new dot patterns, attempting to categorize them into those that conformed to the prototype and those that did not. Interestingly, both groups performed equally well on this task, correctly classifying about 65% of the new dot patterns. However, the brain scans revealed that the two groups were making these decisions using very different parts of their brains (see FIGURE 7.19). Participants who were given the explicit instructions showed increased brain activity in the prefrontal cortex, parietal cortex, hippocampus, and a variety of other areas known to be associated with the processing of explicit memories. Those given the implicit instructions showed decreased brain activation primarily in the occipital region, which is involved in visual processing. This finding suggests that participants recruited distinct brain structures in different ways depending on whether they were approaching the task using explicit or implicit learning.

Figure 7.19: Implicit and Explicit Learning Activate Different Brain Areas Research participants were scanned with fMRI while engaged in either implicit or explicit learning about the categorization of dot patterns. The occipital region (in blue) showed decreased brain activity after implicit learning. The areas in yellow, orange, and red showed increased brain activity during explicit learning, including the left temporal lobe (far left), right frontal lobe (second from left and second from right), and parietal lobe (second from right and far right; Reber et al., 2003).

© REBER, P.J., GITELMAN, D. R., PARRISH, T. B., & MESULAM, M. M. (2003).
DISSOCIATING EXPLICIT AND IMPLICIT CATEGORY KNOWLEDGE WITH FMRI. JOURNAL OF COGNITIVE NEUROSCIENCE.

Other studies have begun to pinpoint the brain regions that are involved in two of the most commonly used implicit learning tasks: artificial grammar learning and sequence learning on the serial reaction time task. Several fMRI studies have shown that Broca’s area–which, as you learned in the Neuroscience and Behavior chapter, plays a key role in language production–is turned on during artificial grammar learning (Forkstam et al., 2006; Petersson, Forkstam, & Ingvar, 2004). Furthermore, activating Broca’s area by applying electrical stimulation to the nearby scalp enhances implicit learning of artificial grammar, most likely by facilitating acquisition of grammatical rules (De Vries et al., 2010). In contrast, the motor cortex appears critical for sequence learning on the serial reaction time task. When the motor cortex was temporarily disabled by the application of a recently developed type of TMS that lasts for a long time (so that participants could perform the task without having TMS constantly applied while they are doing so), sequence learning was abolished (Wilkinson et al., 2010).