It's a Jungle in There: How Competition and Cooperation in the Brain Shape the Mind (26 page)

Proust’s character became talkative when he transitioned from not remembering to doing so. The opposite of talkativeness is being unable to say what you want, and being tongue-tied also provides clues about memory.

Consider the tip-of-the-tongue state. When you have something on the tip of your tongue (not literally, of course), you know you know it, but you can’t recall it. You may recall a bit of it, like the fact that it’s a short word or a long word or that it has a sound like a short “e” or a long “i.” If a cue comes along that activates a hidden feature of the word—its meaning, perhaps, or some sound it contains—you may be able to move from being tongue-tied to being loquacious. What was stuck on the tip of your tongue comes out, often with accompanying excitement.
¹¹

The tendency of cues to liberate tip-of-the-tongue words provides evidence that memories may persist even if they’re hard to access. The cuing power of specific features of words points to the componential nature of
memories.
¹²
Memories have features that, when provoked, conjure cognitive confreres.

Cuing is an explicit means of prompting recall or recognition. Recall and recognition can also be encouraged through more indirect means, as in
implicit
memory tasks. Here, memory isn’t tested by asking someone to recall or recognize as many words as possible. Instead, memory is tested more subtly, by having people complete word fragments in whatever way they like. The instruction can take the following form: “Just write down whatever pops into your head to complete these word fragments.”

al______

hot_____

wal_____

The words you generate tend to be ones you saw before. If you saw “alligator,” you might be inclined to write “ligator” when completing “al____.” Your chance of writing “alligator” is higher if you saw “alligator” than if you saw “alimentary.” This outcome shows that memories for previously encountered words persist. The method used to show this result is called an
implicit
memory test. It’s called
implicit
because memory is probed implicitly or indirectly rather than explicitly or directly.

The word-completion task allowed two psychologists, Peter Graf and Daniel Schacter, to make an amazing discovery.
¹³
They found that amnesic patients who showed no evidence of remembering via recall or recognition could nevertheless complete word fragments. The amnesic patients displayed the sort of bias outlined in the example given above. They often completed “al____” as “alligator” if they saw “alligator” before. If they saw “alimentary” before, they completed “al____” as “alimentary.” (I offer these as illustrative examples, not as exact results from the study.) When the same subjects were explicitly asked to recall or recognize words they had seen, they couldn’t. So when they were asked to recall, they drew a blank, and when they were asked to recognize, they performed at chance. The fact that they could do the word completion task in ways that reflected actual retention of the words shows that more information was stored than would have been expected based on standard recall or recognition tests.

Another demonstration of implicit memory is that people generally prefer familiar items to unfamiliar ones. Even if you’re not sure you’ve encountered
something before, you’re inclined to say you like it.
¹⁴
This is a useful fact for memory researchers, for it shows that a way to find out what people remember is to find out what they prefer. It’s also a useful fact for businesspeople, for it suggests that familiarizing customers with products and services leads them to like those commodities. Businesspeople know this, of course, which is why they advertise so much, often without telling much about the products or services themselves. As a result, there are on TV at the time of this writing, a witty lizard with a Cockney accent serving as a mascot for an insurance company (Geico) and irascible but lovable Vikings acting as mascots for a credit-card company (Capital One).
¹⁵

There’s still another way to show that more is stored than first meets the eye. This is with the method of savings, a method that comes into play when you try to relearn something you learned before. You can learn facts or procedures a second time more quickly than you could at first. The reduction in time spent learning the material is the measure of savings. If savings occurs, something must have been stored.

Suppose you lived in a foreign country when you were a child and later, as an adult, believe you’ve forgotten the language spoken there. I encountered such a student one year in my Intro-Cognitive class. She told me she lived in Germany between the ages of 2 and 6 and that she then moved to the United States and entered a world where no German was spoken. When I asked her if she remembered any German, she insisted she did not.

Did this student actually forget all her German? A way to find out would be to have her take a German class along with others whose ignorance of German was not in question—people who had never (or hardly ever) been exposed to that language. If her knowledge of German was as lacking as she thought, she would have been able to learn German no more quickly than the others.

In fact, she’d probably be able to learn German much more quickly than her classmates, for most people relearn foreign languages more quickly than they learn them the first time around.
¹⁶
The savings they enjoy shows that some knowledge of the language was there all along, albeit in hidden neural niches.
¹⁷

The phenomenon of savings has been demonstrated in a wide variety of contexts, including classical conditioning (discussed in the next section), word learning, and skill learning.
¹⁸
Savings is relevant to the jungle principle
because it suggests that some information may not be truly lost but instead may be weak, inhibited, or hard to access for other reasons.

The resilience of such out-of-the way information has another implication besides the ones already mentioned—that information may not be lost though it seems to be, and that seemingly lost information can be aroused through indirect means. The other implication is that it takes a lot to wipe out information once it has been learned well.

In this regard, there is a striking parallel between the durability of well-learned memories and the durability of well-adapted species. As I learned from a Wikipedia article on extinction, “.. .it is estimated that 99.9% of all species that have ever existed are now extinct.” However, the same article mentioned that “a typical species becomes extinct within 10 million years of its first appearance.”

It may be that 99.9% of the memories formed by an individual are lost forever, but if you ask what the personal analogue of 10 million years is, it’s hard to know for sure. But whatever the exact value may be, it’s a pretty long time, judging from the research that has been done on savings, implicit memory, cued recall, and recognition.
¹⁹

One domain in which savings has been shown is classical conditioning. You’ve been exposed to classical conditioning in your daily life. Every time your mouth has watered at the smell of a roasting turkey every time you’ve broken into a cold sweat when you found yourself growled at by a menacing dog, every time you’ve felt some longing in your loins when you’ve seen an attractive person, you’ve experienced the effects of classical conditioning.

In classical conditioning, an innate response (a so-called “unconditioned response”) that is automatically triggered by some stimulus (a so-called “unconditioned stimulus”) comes to be predicted when another stimulus (a so-called “conditioned stimulus”) is encountered. The prediction becomes worthwhile to the extent the conditioned stimulus reliably heralds the unconditioned stimulus. The smell of a roasting turkey on a Thanksgiving morning is a good predictor of a meal to come, for example. The smell makes you salivate because you’ve learned that the aroma of roast turkey is followed by ingestion.

All Intro-Psych students learn that classical conditioning was discovered by Ivan Pavlov, a Russian physiologist who won the Nobel Prize in Physiology or Medicine in 1904 for his discovery of classical conditioning in dogs.
²⁰
In his work, Pavlov noticed that when he prepared to feed his canines, they
began to salivate before the food was delivered. The strength of the salivation grew on successive occasions as Pavlov prepared the food before delivering it. The sights, sounds, and smells related to Pavlov’s food preparation became conditioned stimuli for the dogs.
²¹

Because the dogs in Pavlov’s lab salivated more as the conditioned stimuli reliably preceded the unconditioned stimulus, you might be tempted to say that the dogs predicted food delivery during the pre-feeding routine. In a sense, they did predict that they would get fed. But you should be careful not to proclaim that the dogs explicitly predicted anything. A mechanistically simpler account is that the neural ensembles related to eating were activated by the sights, sounds, and smells of the events culminating in meals.

What would happen if Dr. Pavlov stopped his normal feeding routine? Suppose he headed for the cupboard and then walked away. If this happened repeatedly, the dogs would stop salivating when Dr. Pavlov entered the kitchen. Their salivation would be “extinguished,” the term used in conditioning research for the cessation of previously conditioned responses. Extinction was, of course, what I just spoke about at the close of the last section.

Extinction bears on the necessity, or lack thereof, of explicit mental prediction. Suppose that in extinction, the dogs in Pavlov’s lab stop predicting that Pavlov will feed them and predict instead that he will not. A problem with this proposal is that a mechanism is needed to switch from one prediction to the other. Another problem is that the prediction, as just portrayed, is all-or-none: Pavlov will either feed the dogs or not, with no in-between. But that’s counter to everyday experience. We have graded expectations. We can have varying degrees of confidence in the likelihoods of various events. Our confidence that it will storm will be lower when it’s partly cloudy than when it’s fully overcast and there is frequent, loud thunder.

Another interpretation of extinction vis-à-vis prediction is that when extinction occurs, prediction stops. According to this interpretation, Pavlov’s dogs cease predicting and do so again only when prediction becomes useful again. This proposal has an obvious problem, however. How can prediction be triggered by its own usefulness?

Yet another interpretation of extinction in connection with prediction is that no explicit prediction
ever
occurs—neither when the dogs are fed, nor when they are not. Dogs simply behave
as if
they are predicting. This interpretation is the one advocated by radical behaviorists, who say that attributing mental states to animals, including humans, is misguided.

I am not a radical behaviorist, to be sure. However, I appreciate the warning not to re-describe phenomena in terms of what you’re trying to explain. If
you’re trying to explain mental states, it’s not satisfying to explain those states in terms of other mental states. In connection with prediction, it’s better to say that neural systems embody prediction by virtue of their collective actions than to throw up your hands and say that prediction just happens at a mental level.

This point is further reinforced by considering what happens when extinction is undone. Suppose Pavlov returns to his normal feeding routine. Now, every time he nears the cupboard, he feeds the dogs again. Soon the dogs resume their mouth-watering when the good doctor arrives. Critically, the dogs start salivating more quickly when Pavlov reinitiates his feeding routine than when he first started it. Thus, the dogs exhibit savings. The fact that they do so shows that savings is a basic phenomenon, not one seen only in complex tasks like relearning languages.

What does the phenomenon of savings imply about prediction? For Pavlov’s dogs, it can logically imply any of the following: (1) the dogs returned from not predicting to predicting more strongly than they did before; (2) the dogs switched from predicting non-feeding to predicting feeding, and they made this switch more quickly than they did before; (3) the dogs’ neural machinery governing the salivation response went from being weak or inhibited to being strengthened or released from inhibition.

Which of these accounts is best? The first and second accounts are complicated; each of them requires ancillary assumptions to make them work. The last account is simpler. It’s the account I favor. I prefer it not just because it straddles the explanatory divide between neural and psychological events, but because it also echoes and supports the mantra of this book, that it’s a jungle in there.
²²

Other phenomena of conditioning provide further support for the view that learning and memory can be usefully understood in terms of the jungle principle. I’ll review two such phenomena now. One is blocking.
²³
The other is captured by an important model in this area of study, the Rescorla-Wagner model.
²⁴

Suppose one of Pavlov’s dogs learned to associate Pavlov’s arrival with Pavlov’s subsequent food delivery. The dog learned to associate Pavlov’s features—his beard, his white lab coat, and so on—with the provisions he provides. One day, Pavlov comes to the lab with a flower on his lapel. He still feeds the dog. Nothing has changed. Only the flower has been added. A few
days later, after having fed the dog while wearing the flower, Pavlov goes on vacation. Two other lab workers take over for him, each coming in repeatedly but at different times. Neither newcomer has a beard or a white lab coat, but one wears a flower. As it happens, only the flower-wearing newcomer feeds the dog.

Does the dog quickly learn to salivate more intensely when the flowered attendant arrives than when the non-flowered attendant does? The canine should show this selectivity to the extent that the flower is predictive of food. But this expectation is not borne out. It takes a long time for the selectivity to be shown. The dog doesn’t have trouble learning to link flowers to food. If Pavlov, from the beginning, fed the dog only while wearing a flower, the dog would learn quickly to salivate only when Pavlov wore the flower.
²⁵