Perhaps what most distinguishes us humans from the rest of the animal kingdom is our ravenous desire to find structure in the information we pick up in the world. We cannot help actively searching for patterns—any hook in the data that will aid our performance and understanding. We constantly look for regularities in every facet of our lives, and there are few limits to what we can learn and improve on as we make these discoveries. We also develop strategies to further help us—strategies that themselves are forms of patterns that assist us in spotting other patterns, with one example being that amateur track runner developing tactics to link digits with running times in various races.
One problematic corollary of this passion for patterns is that we are the most advanced species in how elaborately and extensively we can get things wrong. We often jump to conclusions—for instance, with astrology or religion. We are so keen to search for patterns, and so satisfied when we’ve found them, that we do not typically perform sufficient checks on our apparent insights.
And we really are a decidedly strange species for actively seeking out games with patterns in them, when such activities seem to serve no biological function whatsoever, at least not in any direct way. It’s as if we were addicted to searching for and spotting structures of information, and if we do not exercise this yearning in our normal daily lives, we then experience a deep pleasure in artificially finding them. Crossword and sudoku puzzles are obvious examples, but there are many other types of games, quizzes, puzzles, and so on that inject pleasure into our lives because of the sense of satisfaction we feel when we spot beautiful structures.
And there are some particularly mad people, such as myself, who even decide to dedicate their careers to searching for patterns within a set of information. Scientists may well be motivated to improve society by their discoveries, but most are largely driven each day by their curiosity—a desire to convert some messy pile of experimental data into an elegant, neat little explanation. When describing what I do to those outside of research, I regularly find the most useful analogy is that science is like trying to solve a huge, fuzzy crossword puzzle.
But hobbies in searching for patterns are not by any means limited to the sciences. The arts, too, generate their richness and some of their aesthetic appeal from patterns. Music is the most obvious sphere where structures are appealing—little phrases that are repeated, raised a key, or reversed can sound utterly beguiling. This musical beauty directly relates to the mathematical relation between notes and the overall logical regularities formed. Some composers, such as Bach, made this connection relatively explicit, at least in certain pieces, which are just as much mathematical and logical puzzles as beautiful musical works.
21
But certainly patterns are just as important in the visual arts as in music. Generating interesting connections between disparate subjects is what makes art so fascinating to create and to view, precisely because we are forced to contemplate a new, higher pattern that binds lower ones together. And literature is most powerful when it is using exactly the same trick.
We are alone in the animal kingdom in just how aggressively we constantly search for patterns, and even in how they may be a source of much of our pleasure, both in our creative acts and in our more receptive appreciation of certain hobbies.
The second aspect of chunking is the actual detection of these chunks of structured information within consciousness, and this is where the most central purpose of consciousness resides. What do I mean by patterns, structure, or regularities (words that I use interchangeably)? To take an extreme example, a million 1s and then a million 0s can either be seen as 2 million pieces of information, which would take about three times the length of this book to write out, or we can simply transform those 2 million items into just a single sentence by saying, “It’s just a million copies of 1 and then a million copies of 0.” In other words, spotting patterns is about finding redundancy in the information. You can
compress
the information into a different, smaller, and more useful form by spotting parts that repeat in some way or other, and, ideally, capturing the repetitions in a rule.
If we can successfully turn any group of data into a pattern or rule, then near-magical results ensue. First, we no longer need to remember that mountain of data—we simply need to recall one simple law. But the benefits don’t just stretch to memory. We’re also, crucially, able to predict all future instances of this data, and so control our environment more efficiently. The rule may even capture something about the
mechanism
of the data, allowing us to understand it in a more fundamental way.
Say I’m a prehistoric man. Winter is closing in and my family and I are starving. We dig up some potatoes, which at first sight look edible. We take a bite and spit it out in disgust. I chuck the potatoes on the fire, out of rage. A few hours later, when the fire has died down, I pick the potato up, feel that it has softened considerably, and take another bite. It’s delicious. A few months later, exactly the same set of events happens. I try a raw potato, chuck it in the fire in rage, then a few hours later try again to find it not only edible but delicious. Now I could just remember these as two completely separate episodes in my life and move on.
Or
I could detect a possible pattern—repeated connected instances between food and fire, with fire making potatoes edible. I then no longer need to remember what the weather was like in each of these instances, where I placed my fire, or what words my children said at the time. All that matters is that fire makes potatoes edible. Now that I’ve detected a pattern, I can consolidate the crucial elements of my memory, thus lowering the load; apply this memory chunk to all future instances of fire and potatoes; and start thinking about other similar organic objects that might benefit from a similar treatment—perhaps sweet potatoes or squashes. By crystallizing and compressing my memory, I’ve actually gained a considerable amount of power over the environment, and my family is less likely to starve as a result.
Some of our greatest insights can be gleaned from moving up another level and noticing that certain patterns relate to others, which on first blush may appear entirely unconnected—spotting patterns of patterns, say (which is what analogies essentially are).
It’s difficult to overestimate the extent of learning that is captured by chunking processes. The example of the volunteer who heard digit sequences for two years dramatically illustrates how chunking can vastly increase our short-term memory capacity. But chunking is the main process that we consciously use to turn
any
novel pattern into a structured part of our memories, which is what almost all learning involves. Our long-term memory store, with thousands of related items, is usually unconscious, waiting loyally for our conscious minds to retrieve some item. But at one point in the past each item was met anew, consciously labeled as important, as relating to other features of our preexisting knowledge, and laid down in memory. The initial stages of learning are always the hardest, but once the first foundations are built, we can connect new items with what we’ve already memorized; as the tapestry of knowledge builds, it becomes ever easier to learn a new part of a topic, because it increasingly connects to related memory items. Closely connected individual items form chunks together, which then connect up themselves in ever larger bound objects in memory. In this way, we can use consciousness—and chunking—to create a highly functional, hierarchical, interrelated bank of knowledge, where, by the time we reach adulthood, most seemingly novel items have some preexisting context. And these heavily embedded prior expectations from the fruits of our vast learning can in turn heavily guide our attention to decide what to load into our working memory, furthering our chances to discover in awareness something novel or important by which to incrementally improve our world model.
So, chunking within working memory is both the arbiter and the indexer of our long-term memory store, always striving to make what’s most important to us most easily accessible, and to forge new groups out of apparently independent items, based on the patterns we discern. Consciousness and chunking allow us to turn the dull sludge of independent episodes in our lives into a shimmering, dense web, interlinked by all the myriad patterns we spot. It becomes a positive feedback loop, making the detection of new connections even easier, and creates a domain ripe for understanding how things actually work, of reaching that supremely powerful realm of discerning the
mechanism
of things. At the same time, our memory system becomes far more efficient, effective—and intelligent—than it could ever be without such refined methods to extract useful structure from raw data.
LANGUAGE—JUST ONE KIND OF CONSCIOUS CHUNKING?
Language is a special case, since, by learning a language, we can exponentially increase our capacity to learn more generally via the organization of our own thoughts, via books, teachers, the Internet, and any other form of human communication that can transmit well-chunked information. Whether any other species can start to learn some components of a grammatical language is a controversial question. What’s not controversial is that even if other animals can learn some aspects of language, humans dramatically outshine other species in this sphere, picking up both a vocabulary of thousands of words as well as a complex grammatical structure. All of this is due to chunking.
As we learn language for the first time, we constantly attempt to make successful inferences between the sound of the word we’ve just heard and the key feature of whatever else we’ve just experienced through our senses. We slowly build up our language, with many words initially starting as overly general chunks. For instance, our baby daughter started saying “Mama” quite early on, which initially excited my wife to no end, until we realized it merely meant she was generally unhappy. But language soon allows us to use the linguistic objects we’ve already learned to hone our skills further, building up our rich web of meaning.
Many psychologists believe in a “language instinct”—a set of uniquely human brain regions and mental skills built especially for language. My view instead is that language emerges out of our more general capacity to make conscious chunks. As consciousness and the ability to chunk begin to flower, this allows the child to extract these language chunks from what she hears, just as she’s learning about many other types of structured information, such as how to walk, how to deal with complex social situations, or how to interact with her more advanced toys. At the same time, adults learning an artificial grammar, akin to acquiring the rules of a novel language, do not activate some special language region. Instead, they activate just the same brain areas as when they are performing any other chunking task, such as encoding structured spatial sequences. And one would expect a language instinct to have genes associated with it, giving us a specific helping hand from our earliest days. To date, there is controversial and patchy evidence for this.
22
Instead, we have a sufficiently advanced form of consciousness that is hungry for patterned, hierarchical information and playfully active in its search for powerful, structured cognitive tools to manage this information—of which language is perhaps the most rich, useful example. After all, language allows yet another level of information processing in nature, where chunks of insight can be passed between people, and collaboration between members in a group can generate innovative new ideas, which would not have been possible alone.
THE FRUITS OF CHUNKING AND AWKWARD SELF-CONSCIOUSNESS
The third and final aspect of chunking in relation to consciousness is what we do with those chunks when we’ve firmly acquired them. In this case, it’s generally better if we’re
not
aware of them—at least not anything other than the absolute top-level chunk we’re currently dealing with.
The main purpose of awareness is either to manage tasks too new or complex for our simpler unconscious mind or to innovate, to find patterns in our working memory, so that we can optimize and automate biologically relevant goals. But once those tasks have been heavily learned, if they take up consciousness again, then that is far from ideal. First, because we are simultaneously analyzing the details of the task, we are likely to perform it less efficiently than when it was an automatic habit. Then there is the issue of energy use: Consciousness requires a large and active network of cortical regions, which in turn need considerably more energy than unconscious, automatic habits. So it is wasteful to use consciousness for anything other than its official purpose, to discover potentially significant opportunities to improve our mental programming.
Some chunks are too complex for our unconscious minds to handle and instead serve to guide our consciousness toward a specific goal. For instance, if I want to make a relatively nontrivial change to my PC, such as modifying the monitor resolution, I know exactly what sequence of mouse clicks to make to achieve this, and each one has to be consciously controlled. But the vast majority of these well-learned packets of information can whirr away quite happily while conscious engagement is entirely elsewhere—for instance, me walking for many minutes, oblivious to my surroundings, or driving while daydreaming.
It’s very much as if we had two modes: a slow, deliberative, highly conscious system, there to detect novel or ever more complex forms of patterned information, to find its structural essence, which we use to build chunks; and a fast, automatic, barely conscious system, which takes advantage of the well-honed chunks that consciousness has previously formed.