Does information in short-term memory decay?

The very fact that I’ve posed this as a question rather than a statement probably gives some indication that the answer isn’t exactly straightforward. The notion of decay is a vital component of the short-term/long-term memory distinction, so even asking the question risks casting doubt on an assumption that is almost as old cognitive psychology itself.

There are two main factors that differentiate short-term memory from long-term memory; duration and capacity. As far as Long-term memory is concerned, duration (for how long information can be held) is thought to be indefinite, while its capacity (how much information it can hold) is considered limitless. The same is not the case for short-term memory. Both duration and capacity of short-term memory are thought to be limited, that is, we can’t hold very much information there and what we can hold doesn’t last long. The extent of these limitations are confounded due to the differences in the types of information we come into contact with. For example, single digits (like those used in Miller’s classic 1956 study) aren’t the same as words, and words vary in length. Similarly, the interplay between short-term memory and long-term memory means that we can probably hold familiar items for longer, as well as hold more of them. If I were to give you a list of made-up words and a list of real words (matched for length) you’d be able to recall more of the real words than the made up ones. Similarly, if I were to give you two sets of real words to learn, one set being of one syllables’ length and the other six syllables, you’d recall more one syllable words than six syllable words – this phenomenon is called the word length effect and was first demonstrated by Baddeley in 1975.

One explanation of the word length effect involves the notion of decay (or trace decay). It’s assumed that events leave some kind of indelible mark and that this will gradually fade over time. Ebbinghaus, for example, noted that learned nonsense ‘trigrams’ disappear over time if not revisited, while Peterson and Peterson (see below) found that information in short-term memory disappears if rehearsal is prevented. According to the word length effect, longer words take (unsurprisingly) longer to articulate, longer articulation times mean that decay will start to set in earlier than it would with shorter words – longer words require increased rehearsal time. Another explanation is that longer words tend to be more complex, so require more work, while short words are generally easier to recall.

Decay or not decay

We tend to accept that information in short-term memory will decay if not rehearsed or when rehearsal is prevented (I’ll challenge this notion in a moment). Lloyd and Margaret Peterson (Peterson and Peterson, 1959) were interested in two interrelated aspects of short-term memory: for how long information remains in short-term memory before it decays, and the consequences of limiting people’s ability to rehearse the information. To investigate this, they presented volunteers with lists of nonsense trigrams one at a time and had the volunteers recall them after intervals of 3, 6, 9, 12, 15 or 18 seconds. In addition they asked them to count backwards in threes or fours from a specified random number until they saw a red light appear. The counting was to prevent rehearsal of the trigrams, a technique developed a year earlier by Lloyd Peterson and John Brown and known as the Brown-Peterson technique. Not surprisingly perhaps, the volunteers ability to recall the trigrams was associated with the length of the delay, ranging from eighty percent correct after 3 seconds to less than ten percent after 18 seconds. The Petersons concluded that short-term memory has a limited duration and that when rehearsal is prevented information will decay after about 18 to 30 seconds. 

This is also consistent with Ebbinghaus’s conclusions that information is going to decay if we don’t or can’t rehearse it; it’s why we instinctively repeat a telephone number over and over again until we can locate a pen. Delay also affects the so-called recency effect. In a 1966 study, Murray Glanzer and Anita Cunitz found that if you present volunteers with a list of words to recall but then prevent rehearsal, they still have a tendency to remember the first few words (the primacy effect) but have difficulty recalling those at the end of the list (the recency effect), the assumption being that the first few words have already entered long-term memory but because of the rehearsal prevention task, they have disappeared from short-term memory. Now, I have a problem with trigrams (and I’m not the only one) because they have been deliberately stripped of meaning, they aren’t real words and our ability to recall real items is much better than our ability to recall nonsense ones, but I’ll leave that alone here and refer you to my Ebbinghaus discussion.

The notion that information held in short-term memory decays over time might appear uncontentious, but testing the decay theory of forgetting is all but impossible. Like the Petersons, researchers use distraction tasks to prevent rehearsal, but the distraction task itself introduces new information that may well interfere with the learned material; forgetting then occurs because of this interference and not through decay. Since the 1950s all the way up to the present day, memory researchers have been locked in debate over whether information in short-term memory is lost through decay or interference, and this debate matters. As Ricker, Vergauwe and Cowan note, “The fundamental difference between the two [short-term and long-term memory], if they are separable, would appear to be that only the contents of short-term memory are limited to a small number of items or to a short period of time, whereas the same limits do not apply to long-term memory. Short-term memory as a theoretical construct is therefore like a roof that stands on just two massive pillars, and decay is one of those pillars [my emphasis].”

The arguments on both sides are rather complex and involve many less than stimulating experiments within the verbal learning tradition. Interference is when information known either prior to list learning or learned after the list, interferes with the items on the list. You might have witnessed this when you move house and change your telephone number; often your old number will make it harder to learn the new one. It can be argued that distraction tasks used in experiments can also contaminate the words participants have been asked to learn. The problem is that it’s impossible to tell if forgetting occurred as a result of decay or interference. To make matters worse, a process known as covert verbal rehearsal, the imagining of how words are pronounced without saying them out loud, is a skill that can be learned and requires very little effort even when carrying out a distraction task.

This is (probably) not the answer you’re looking for

We certainly know that information thought to have decayed can often turn up later when people are given a prompt or cue or if the original information formed one half of a pair. If, for example, I were to give you a list of words to memorise and then tested you on them immediately, chances are you’d have forgotten at least some of them. However, if a little while later I gave you a list of words that contained both the words from the original list and some new words and asked you to identify the words from the original list, chances are you’d recall more than you did when tested. Similarly, I could give you the first couple of letters of the words you failed to recall and this might be enough for you to remember them. The forgotten items, therefore, didn’t decay because you were then able to identify them later. Similarly, if I asked you to learn lists of paired words, presenting you with one of the pair would likely help you recall the other. These types of experiments have been conducted many times over the past half-century and results are pretty consistent. 

If items don’t decay from short-term memory then this would cast some doubt on the short-term and long-term memory distinction (we will have lost one of those two massive pillars referred to by Ricker, Vergauwe and Cowan). Realistically, however, decay probably accounts for at least some forgetting, but interference most certainly also plays a part. There are other possible candidates, of course, and unitary models of memory (those that wholly or in part reject the short-term, long-term distinction) often emphasise the role of attention.