Great discoveries in science are often made in multiples: scientists working on the same problem arrive at similar answers at approximately the same time. The discovery of calculus is credited to independent work by Isaac Newton and Gottfried Leibniz, and Darwin was spurred to publish his masterpiece On the Origin of Species by converging ideas coming from Alfred Wallace. The finding that synapses obey Hebb’s rule was no different. In 1986 no fewer than four independent laboratories published papers showing that a synapse becomes stronger when its presynaptic and postsynaptic partners are activated at the same time.14 These studies established the existence of what is called associative synaptic plasticity, and fueled thousands of other studies and many breakthroughs over the following decades.
How does a synapse “know” that both its presynaptic and postsynaptic neurons are active at the same time, and then proceed to become stronger? Establishing these neuronal associations is such a pivotal component of brain function that evolution has concocted an “associative protein”—a molecule found in synapses that can detect whether the presynaptic and postsynaptic neurons are coactive. The protein, a receptor of the excitatory neurotransmitter glutamate, called the NMDA receptor, works as a gate that opens only if the presynaptic and postsynaptic neurons are active at about the same time, which allows it to implement Hebb’s rule. We could say that the NMDA receptor functions much like the Boolean “and” used in search engines, that is, it only returns a result (it opens) if two conditions are satisfied (activity in the presynaptic and postsynaptic neurons). Once the NMDA receptors open, a complex series of biochemical events that lead to long-term potentiation of a synapse are triggered.15 Thanks to its unique properties the NMDA receptor detects the “associations” between neurons, and is pivotal to the implementation of Hebb’s rule and associative synaptic plasticity.16 To return to the social network analogy, if Hebb’s rule were applied to Facebook, people that logged into their account at the same time would automatically become friends, ultimately creating a network of people with synchronized schedules.
We can now begin to appreciate how semantic memory networks emerge. As a child, how did you learn that a particular furry, snobbish, long-tailed, four-legged creature is called a “cat”? Somewhere during your first years of life some neurons in your brain were activated by the sight of a cat, while others were activated by hearing the word cat. (Babies initially have no knowledge that the sounds of the word cat and the sight of a cat are related.) Somehow, somewhere along the way, your brain figured out that the auditory and visual format of “cat” were in some sense equivalent. How did this arise? Most likely it was thanks to your mother. Because Mom insisted on saying, “Look at the kitty cat” the first ninety-nine times you saw a cat, she ensured that the auditory and visual “cat” neurons were active at roughly the same time. Enter Hebb’s rule and associative synaptic plasticity: because these neurons fired together, they wired together—they became connected to each other with strong synapses. Eventually the neurons activated by the word cat became capable of turning on some of the neurons stimulated by the sight of a cat, allowing you to figure out what Mom was referring to when she said “cat,” even when the moody creature was nowhere to be seen.17
I first appreciated the importance of associations in child development as the result of an unplanned and undoubtedly unethical psychological experiment on my sister, who is nine years younger than me. From her earliest days, I addressed my sister primarily by the unkind nickname Boba, which in Portuguese means “dummy.” On one occasion, when she was about three years old, a friend and I were playing in the front yard and he yelled “oba” (“yeah!”). My sister mistook this exclamation for Boba, and immediately dashed outside and said, “Yes.” I still recall being struck by two thoughts. First, I really should start calling her by her real name, and, second, in retrospect, she would have had no way of knowing that Boba was a pejorative term and not her name (or one of her names). If someone generates a specific sound every time they interact with you, your brain cannot help but build an association with that word and yourself—it is what the brain is programmed to do.
One of the ingenious properties of this associative architecture is that it is self-organizing: information is categorized, grouped, and stored in a way that reflects the world in which we live.18 If you live in India your “cow” neurons will likely be connected to your “sacred” neurons, whereas if you live in Argentina your “cow” neurons will likely be strongly connected to your “meat” neurons. Because of its self-organizing nature, human memory is in many ways vastly superior to the mindless strategy of precisely capturing experiences with a video camera. The associative architecture of the brain ensures that memory and meaning are intertwined: the links are both the memory and the meaning.
PRIMING: GETTING IN THE MOOD
Now that we have some understanding of how memories are stored and organized in the brain, we can return to the phenomenon of priming. The fact that we can nudge people into thinking of a zebra by evoking thoughts of Africa and black and white is not only because knowledge is stored as a network of associated concepts, but because memory retrieval is a contagious process. Entirely unconsciously, activation of the “Africa” node spreads to others to which it is linked, increasing the likelihood of thinking of a zebra. Psychologists often study priming by determining the influence of a word (the prime) on the time it takes to make a decision about a subsequent word (the target). In this type of experiment you sit in front of a computer screen while words and nonwords (plausibly sounding pseudo-words such as “bazre”) are flashed one by one. Your job is to decide as quickly as possible if the stimulus represents a real word or not. If the word butter were flashed, it might take 0.5 seconds to respond. But if bread were flashed before the presentation of butter your reaction time might fall to 0.45 seconds. Loosely speaking this increase in speed is because activity in the group of neurons encoding bread spreads to related concepts, accelerating recognition of the word butter. The ability of “bread” to prime “butter” may not be universal: these words have a strong association for Americans because they often put butter on their bread, and because “bread and butter” is an expression referring to financial support; but it is possible that there would be little or no such increase in speed with people from China, where the custom of buttering one’s bread is less common.
Given its importance, it’s unfortunate that we don’t really know what priming corresponds to in terms of neurons and synapses.19 One theory is that during the semantic priming task, when the neurons representing “bread” are activated they continue to fire even after “bread” is no longer visible. Like a dying echo this activity progressively fades away over a second or so, and during this fade out neurons continue to whisper to their partners. Thus the neurons representing “butter” receive a boost, even before “butter” is presented, and fire more quickly.20
Irrespective of the precise neural mechanisms, priming is clearly embedded within the brain’s hardware. Like it or not, whenever you hear one word your brain unconsciously attempts to anticipate what might be coming next. Thus “bread” will not only prime “butter,” but depending on the specifics of your neural circuits it will also prime “water,” “loaf,” and “dough.” Priming probably contributes to our ability to rapidly take into account the context in which words occur and resolve the natural ambiguities of language. In the sentence, “Your dog ate my hot dog,” we know that the second use of “dog” refers to a frankfurter as opposed to a dog that is hot. The use of the word ate earlier in the sentence provides context—it primes the correct interpretation of the second use of “dog,” helping to establish the appropriate meaning of the sentence.
The spread of activity from an activated node to its partners is of fundamental importance because it influences almost all aspects of human thought, cognition, and behavior. Consider a conversation you might have with someone you have never met before. As the dialogue proceeds, the topic changes, establishing a conversat
ional trajectory. What determines this trajectory? Human interactions are influenced by many complex factors, but there are patterns. A conversation might start with geography (Where are you from?). If the answer is Rio de Janeiro, the topic may veer toward soccer or Carnival. If the answer is Paris, the topic could head toward food or museums. The transitions of conversations are often primed by the previous topic. But importantly, these transitions depend on the specific structure of the conversationalists’ semantic nets. Indeed, when you know someone well, it is not difficult to elicit a given story or topic from them (or prevent them from telling the story you have heard a million times) by mentioning or avoiding certain priming words.
MEMORY BUGS
Priming is one of the most valuable features of the brain, but it is also responsible for many of our brain bugs. We have already seen that false memories can be generated because we confuse related words. Given the words thread, pin, sharp, syringe, sewing, haystack, prick, and injection, people will often insist that “needle” was among them. Remembering the gist of something is a useful feature of memory, because it is often the gist that really matters. Let’s suppose you are setting out on an expedition and are told that the forest contains anacondas, poison ivy, quicksand, scorpions, cannibals, alligators, and rodents of unusual size. When your traveling companion asks you whether you think you should head through the forest or across the river, you may not be able to convey all the reasons why the river is the superior choice, but the general gist will be a cinch to remember.
In many circumstances, however, simply remembering the gist will not suffice. If your significant other asks you to buy a few things on the way home from work it’s not sufficient to remember the gist of the list; family harmony is best achieved by remembering whether bread or butter was one of the items. To my embarrassment I once caught myself making a memory error that was clearly due to priming and the common association between crocodiles and alligators. In the pursuit of the plastic footwear named “Crocs,” I found myself asking a salesperson if he sold “alligators.”
Because individual experience sculpts our semantic nets, different individuals would be expected to have different susceptibilities to some types of errors. In a study performed by the psychologist Alan Castel and his colleagues, volunteers were given a list of names of animals to memorize: bears, dolphins, falcons, jaguars, rams, and so on—all animals that have American football teams named after them. Not surprisingly, people who were football fans were better at memorizing the list (presumably because they had a richer set of links associated with each animal name). But they were also more likely to have false memories, and mistakenly believe that eagles or panthers (also the names of football teams, but that were not on the list) had been presented.21
You probably have your own examples of memory errors caused by the hyperlinked associative networks in your cortex. As annoying as these errors are, they are generally not life-threatening. But in some cases they can be. Paxil, Plavix, Taxol, Prozac, Prilosec, Zyrtex, and Zyprexa are all on the Institute for Safe Medication Practice’s list of frequently confused drug names.22 The confusion of medications by doctors, pharmacists and patients is responsible for medical mistakes, and up to 25 percent of medication errors are related to confusing pharmaceutical names. Indeed, as part of the drug approval process, the Federal Drug Administration screens drug names precisely to decrease this type of error. Some of these memory errors arise when medical professionals confuse drugs that are in the same category: Paxil and Prozac are both a specific type of antidepressant, with similar mechanisms of action, so their names readily become linked in our neural nets. Other errors result from drugs having similar names such as Xanax and Zantac, or Zyrtex and Zyprexa. Here the drugs may share associations because the brain represents their pronunciation or spelling by using similar nodes.
Is the rock formation hanging from the ceiling of a cave a stalagmite or a stalactite? Is a bump on a road concave or convex? Is the big guy Penn or Teller? Why do we confuse the words representing distinct but related concepts? Because if two concepts that are not used much share most of their links—similar spelling, pronunciation, contexts, or meaning—they run the risk of becoming so entwined as to be indistinguishable.
We are now in a better position to understand the causes of the Baker/baker paradox, which shows that we are more likely to remember professions than names—even when they are the same word. Throughout your life the profession “baker” has acquired many associations (bread, funny hat, dozen, cake, getting up early). By contrast, the name “Baker” pretty much stands alone (unless, of course, your name happens to be Baker). In other words, the “baker” nodes are well connected, whereas the “Baker” nodes are loners, and that is why “Baker” is more difficult to remember.23 When we are introduced to a baker more links are activated than when we are introduced to Mr. Baker; the increased number of links may translate into a more enduring memory because a larger number of synapses are contenders to undergo synaptic plasticity. A common mnemonic device to remember names is to associate them with something more memorable (Richard with being rich or Baker with a baker). This trick may work because it “borrows” links and synapses from nodes that would not otherwise be used, increasing the number of synapses involved in memory storage. Although we will have to await future research to confirm this explanation, we can begin to understand the cause of one of the most maligned characteristics of human memory: the difficulty in remembering names. The associative architecture of the brain offers a powerful way to organize and store knowledge, but like a Web page that nobody links to, a node without many links is difficult to find.
IMPLICIT ASSOCIATIONS
Priming and the associative architecture of our memory can have even spookier and more far-reaching effects than those arising from the confusion of related concepts and words. We generally view memory as a neutral source of information about the world, but it turns out that the way information is stored can sway our behavior and opinions in an entirely unconscious fashion.
A simple example of how the associative architecture of memory influences how we use and access the information stored in our neural nets is illustrated by what’s known as an implicit association test. Each word in the list below is either a flower or insect, or a word with a “positive” or “negative” connotation (for example, “helpful” or “nasty”). Your task is to categorize each word as quickly as possible by checking the left column if the word is a flower or can be said to be something positive, and the right column if it is an insect or represents something negative. If you’re in a quantitative mood you can time yourself to find out how long it takes you to complete the first part of this twelve-word test.
The next part of the test is the same, except that you should check the left column if it is an insect or a positive word, and the right column if it is a flower or negative word. (If you are timing yourself, also measure the time it takes to complete the next twelve words.)
A real implicit association test is slightly more involved than the task you just performed but even in our simplified version you may have observed that overall you were a bit slower on the second list.24 Studies show that on average people are significantly slower, and make more errors, when the assigned responses are grouped incongruously (in this case, counter to most people’s view that flowers are pleasant and insects are unpleasant.
One of the first studies to investigate the effects of implicit associations examined whether Korean Americans and Japanese Americans differed in response times as a result of different cultural stereotypes. The psychologist Anthony Greenwald and his colleagues reasoned that Korean and Japanese Americans might have mutually opposed attitudes (and implicitly different associations in their semantic networks) toward each other due to Japan’s occupation of Korea in the first half of the twentieth century (in addition to the natural affinity we have toward our compatriots). Subjects were asked to press one key on a computer keyboard when Japanese names were presented and another
key when Korean names were presented (the “category” words). Interspersed with the names were adjectives or common nouns that could be classified as being pleasant or unpleasant such as happy, nice, pain, or cruel (“attitude” words). Two keys on a computer keyboard were always assigned a category and an attitude: for example, Japanese names or pleasant words were assigned the same key, and Korean names and unpleasant words the other key (in another trial, the converse pairing was used). On average, Japanese subjects had slower reaction times when the Japanese and unpleasant responses (and Korean and pleasant responses) were assigned to the same key.25 Likewise, Korean subjects were slower when the Korean and unpleasant responses were assigned to the same key.
Why would it take more time for people to decide whether a fly is an insect when it shares the same response with positive words than when the appropriate response is grouped with negative words? Similarly why would some Japanese Americans be quicker to recognize Japanese names when the response is paired with pleasant words as opposed to negative words? If a task requires you to respond to the words concave and convex by pressing a button on the left, and the words stalagmite and stalactite by pressing a button on the right, you’re brain doesn’t have to go through the trouble of distinguishing between concave versus convex or stalagmite versus stalactite—it can quickly assess the correct response based on whether the word corresponds to spherical things or things found in caves. But if the task is structured as “concave” and “stalagmite” to the left, and “convex” and “stalactite” to the right, the brain is forced to parse the difference between closely related concepts, and the more two concepts have in common, the more overlap between the nodes representing these concepts—or, perhaps more accurately, between the neurons representing the nodes. The same holds true in other domains: given a pile of beads of four different colors—black, brown, blue, and cyan—it is much easier to separate them into two piles of black/brown versus blue/cyan beads than into two piles composed of black/cyan versus brown/blue beads.
Brain Buys Page 4