Short-term memory is characterized by its limited capacity and quick loss of information. In a test of capacity, such as a digit span task, it can hold only approximately seven (five to nine) items. You can try the digit span task on yourself. Read these numbers out loud at a rate of about one per second: 6, 3, 9, 7, 4, 6, 2, 3, 9, 4. Now, close your eyes and try to recite them back, in order. How many can you remember? After a few tries, most people can remember about seven numbers in a list. Very few people can remember more than nine digits in a row. This is the limited capacity of short-term memory. Now, close your eyes again and try to repeat the numbers you read a few sentences ago. More difficult, isn't it? This is a demonstration of the quick loss of information in short-term memory. Interestingly, in the Welsh language, numbers have many more syllables than in English, and the digit span of bilingual speakers is lower when they read the numbers in Welsh than when they read the numbers in English. This may be because it takes more time to read the numbers in Welsh, so that they need to remember the earlier numbers on the list after twenty seconds have gone by rather than only ten.

Why should it matter how long it takes to read the numbers? Well, information is short-term memory is lost quickly; it can only be held in the store for a short time. There have been two principal reasons put forth as the cause of this quick loss of information. One argument holds that decay is the main source of information loss, meaning that memories simply fade away over time. If you read the numbers faster, at a rate of four per second instead of one per second, you should be able to remember more of them (if you repeat them back immediately afterward). Reading the numbers faster gives the memories less time to decay before you spit them all back.

Another possible explanation is interference, meaning that other information pushes memories out of the short-term store. There are two types of interference: proactive and retroactive. New information pushes old items out of memory through retroactive interference; this type of interference works backward ("retro"), ruining memory for items entered earlier. As an example of retroactive interference, imagine that I give you only one number to remember, "6," instead of the ten I gave you earlier. Much easier, right? That's because those later nine numbers aren't interfering with your rehearsal of "6" in short-term memory. Old items clogging up short- term memory prevent the accurate entry of new information through proactive interference; this type of interference works forward ("pro"), straining memory for items entered later. One study of proactive interference asked people to remember short lists of words. If the first twelve words were from the same category (such as animals, e.g., cat, elephant, dog, crocodile, and so on), people had a much harder time remembering later words on the list. However, if the words suddenly switched categories at number thirteen (e.g., doctor, lawyer, fireman...), subjects had no trouble remembering the words from the new category. Evidently, the pile-up of so many "animal" words made it difficult to keep them all straight, causing proactive interference, but the words from a separate category were not affected since they were different enough to not be confused with the earlier words. Overall, evidence from many psychological studies suggests that both arguments are probably right; memories do decay naturally over time, but items in short-term memory can definitely interfere with each other. The existence of interference shows that short-term memory definitely has a limited capacity for information.

But don't let words like "capacity" fool you; short-term memory is more than just a storehouse of information. Information in short-term memory can be manipulated and rehearsed before being transferred to long-term memory (or forgotten). Some cognitive psychologists see short-term memory as a very dynamic place, so it is sometimes called working memory. One such cognitive psychologist is A. D. Baddeley. Baddeley's model for working memory divides it into three parts: the visuo-spatial sketch pad, the phonological loop, and the central executive. The visuo-spatial sketch pad is what people refer to as the "mind's eye." It controls visual imagery, which is discussed in detail in the section on Mental Imagery. In Baddeley's model, the central executive controls awareness of the information in working memory. For example, picture a pink elephant. Your visuo-spatial sketch pad created the image of the elephant, but the central executive brought it into your consciousness so that you could be aware of the elephant and say, "Hey, I can see a pink elephant in my mind; it's big and has floppy ears." Although the central executive directs our awareness toward working memory's content, we seem to have little or no control over it; we automatically experience whatever is presented to us. (Anyone who has had a Spice Girls or Ricky Martin song stuck in their head--specifically, in their phonological loop--can probably sympathize!)

The phonological loop is used to rehearse verbal information to keep it in the short-term memory. Verbal information is anything that can be spoken, such as a word, a digit, or a phoneme (a syllable, like "em" or "ki"). Because it is fairly easy to study and because we use it quite often (since humans are highly language-oriented), the phonological loop has been studied extensively, yielding a fair amount of information on how it works. The phonological loop seems to have two parts: an inner ear and an inner voice (Baddeley, 1986). Repeat the numbers "1, 2, 3" to yourself several times, without saying them out loud. You can hear the numbers being spoken, although you are not making any sound. Your inner voice is "saying" the numbers, and you inner ear is hearing them. Often, when people want to remember verbal information, they rehearse it using the phonological loop. For example, say I need the phone number for a pizza place. I look it up in the phone book, and find that the number is 555-6475. As I walk across the room to the phone and dial the number, my phonological loop repeats "555-6475, 555-6475…" As soon as I finish dialing the number, I stop rehearsing it, and it leaves my short term memory.

Imagine that, as I am rehearsing the phone number, someone asks me a question, and I stop to answer it. Chances are, I'm going to have to look the number up in the phone book again; that person has caused interference in my short- term memory, pushing the phone number out. One cognitive experiment studied the interference of different types of background noise on subjects' digit span. White noise was the same as silence; no interference there. Hearing unrelated phonemes such as "do" and "fa" caused some interference, lowering subjects' digit span. Hearing another person reciting numbers caused the most interference, lowering the digit span severely, and hearing phonemes from numbers, such as "sev" or "fi" had a similarly bad effect. This evidence supports the theory that rehearsal in short-term memory is controlled by a phonological loop, since distracting information that sounds similar to the rehearsed items can interfere with short-term memory for those items.

So, to sum up, short-term memory can only hold a limited amount of information, and it loses even that information rather quickly. Short-term memories fade due to decay and due to interference of items with each other.