Chapter 14 · Retrieval Is Competition

We said earlier that forgetting is retrieval strength falling, and that the large share of that fall comes not from time itself but from the interference piled up in the meantime. Interference is, in the end, competition: another representation taking or intercepting a share of the activation that should go to the target, so the target does not come up clearly. But this competition comes by two structurally different paths: a competitor shares the same cue with the target, or wakes together from other cues. We look at the two in turn.

Many Share One Cue

One path comes straight from the structure of the index. If a representation is a member of several composite representations, when it lights up the activation splits and flows to those several. Activation that, in a single stream, would have gone straight to the one target thins as it divides into several branches. If the target representation is pushed aside in this splitting, retrieval slows, or a wrong representation hung on the same cue comes up instead. The more representations hang on one cue, the harder it becomes to call one particular thing clearly with it.

Seen this way, you see why a fall in retrieval strength is a matter of interference. As time passes, we keep piling new composite representations onto the same cue. The more those new competitors grow, the more finely the activation going to the old target splits, and the harder it is to reach. It is not that time merely passes and the memory fades, but that over that time competition piled on the same cue veils the target.

Many Cues Wake at Once

The other path is a different affair. Even if the cue the target leans on is clean—even if no competitor is hung on it—interference can come, because retrieval is not a matter of lighting one cue and ending. One situation wakes several cues at once. When those cues each haul up the representations hung on them, working memory tries to fill not only with the target but with the representations the neighboring cues called too.

But as we saw earlier, working memory is narrow. Those representations woken at once contend over the narrow room, and if a strongly lit one takes the spot first, the target is pushed aside and does not come up. It is buried in the crowd woken alongside. This is so when you try to recall something by a vague cue. Grope for "the name of the person I met there," and the place, the events, and the faces of that time rise all at once, so you cannot pick the one name; narrow the cue and ask "what was that person's job," and the crowd that wakes shrinks and the name at last comes.

Ways to Reduce Competition

The two paths differ, so the prescriptions to reduce them are two. But their root meets in one: removing competitors from the path that reaches the target.

Where competitors crowd one cue, making the cue distinctive works. Hang a representation on a common, everyday surface node and the competitors crowding that cue are many; encode it onto a distinctive cue nearly bound to that representation alone, and the competitors crowding the one cue shrink. With little for activation to split into, it gathers clearly onto the one target. Having people deliberately tell apart things that look similar does the same work. We said earlier that mixing several types adds the synthesis of discrimination and spends more resource; that mixing has a second face. Made to sort out "what type is this" every time, each representation that looked similar gets a different retrieval path. Drill one type massed and they are lumped onto one surface and later confused with one another; mix them and force discrimination, and each hangs on a different cue and is clearly distinguished.

Where many cues wake at once, narrowing the cue works. Casting about vaguely and broadly wakes too many cues at once, so by asking with a narrow, concrete cue nearly bound to the target, you reduce the very crowd that wakes. When, with an exam at hand, similar concepts rise together and tangle, the reason that narrowing into one concrete question is less confusing than vaguely recalling "that unit's content" lies here.

How to Fix What Went Wrong

So far the story has been of a rightly placed representation pushed aside in competition. But there is a place where the same competition works more nastily: when a wrongly stored representation occupies the cue from the start. Newly mastering something is just adding, but fixing what went wrong means pushing aside the old already in place, and that old scarcely yields.

The reason lies in the axiom set earlier: once accumulated, storage strength does not diminish. So even a wrongly stored representation does not lose its storage strength, and the old representation remains in long-term memory, unerased. Try to overwrite it with correct information and the old is not erased, so correction cannot be a matter of erasing the old. Then what is correction? It is building a correct representation anew and raising its retrieval strength again and again, so that when the same cue is given, not the old representation but the new wins the retrieval competition. Not getting rid of the old, but standing beside it a new one that wins more often, gradually holding the old down.

Here lies the reason a misconception is so stubborn. The old misconception's storage strength remains, so each time a related cue is given, it wages the retrieval competition anew against the new representation hung on the same cue. So hearing the correct explanation once does not swap the intuition out at once. For a while an awkward middle state is produced, the correct information and the old intuition mixed. That the old intuition pops out unawares even after you have clearly learned the right thing shows that correction does not end in one go but is a slow matter, completed only when the new representation wins the competition again and again.

To here we have seen the path of finding and using what is in long-term memory, keeping it from being buried in competition, and setting right what went wrong. Yet the learner cannot directly look at whether their own long-term memory has so changed. Whether it was fixed or firmed is not felt; only the ease of the reading moment is grasped. This blind spot—unable to read one's own learning state—is the story of the next part.