There is something fiendishly frustrating about piggy banks. You can put money in them, you can shake them around to assure yourself that the money is there, but you cannot easily get the money out. If memories were like coins in a piggy bank, stored but inaccessible, what would be the point of saving them in the first place? Retrieval is the process of bringing to mind information that has been previously encoded and stored, and it is perhaps the most important of all memory processes (Roediger, 2000; Schacter, 2001a).

One of the best ways to retrieve information from inside your head is to encounter information outside your head that is somehow connected to it. The information outside your head is called a retrieval cue, external information that is associated with stored information and helps bring it to mind. Retrieval cues can be incredibly effective. How many times have you said something like, “I know who played Hermione in the Harry Potter movies, but I just cannot remember her name”? Then a friend gives you a hint (“Wasn’t she also in The Bling Ring?”), which instantly brings the answer to mind (“Emma Watson!”).

In one experiment, undergraduates studied lists of words, such as table, peach, bed, apple, chair, grape, and desk (Tulving & Pearlstone, 1966). Later, the students were asked to write down all the words from the list that they could remember. When they were absolutely sure that they had emptied their memory stores of every last word that was in them, the experimenters again asked the students to remember the words on the list, but this time, the experimenters provided retrieval cues, such as “furniture” or “fruit.” The students who were sure that they had done all the remembering they possibly could were suddenly able to remember more words. These results suggest that information is sometimes available in memory even when it is momentarily inaccessible, and that retrieval cues help us bring inaccessible information to mind.

Hints are one kind of retrieval cue, but they are not the only kind. The encoding specificity principle states that a retrieval cue can serve as an effective reminder when it helps re-create the specific way in which information was initially encoded (Tulving & Thomson, 1973). External contexts often make powerful retrieval cues (Hockley, 2008). For example, in one study divers learned some words on land and other words underwater; they recalled the words best when they were tested in the same dry or wet environment in which they had initially learned them because the environment itself served as a retrieval cue (Godden & Baddeley, 1975). Recovering alcoholics often experience a renewed urge to drink when visiting places in which they once drank because those places serve as retrieval cues. There may even be some wisdom to finding a seat in a classroom, sitting in it every day, and then sitting in it again when you take the test because the feel of the chair and the sights you see may help you remember the information you learned while you sat there.

237

Retrieval cues need not be external contexts—they can also be inner states. State-dependent retrieval is the tendency for information to be better recalled when the person is in the same state during encoding and retrieval. For example, retrieving information when you are in a sad or happy mood increases the likelihood that you will retrieve sad or happy episodes (Eich, 1995), which is part of the reason it is so hard to “look on the bright side” when you are feeling low. Similarly, you would probably expect a fellow student who studied for an exam while drunk to perform poorly, and you would probably be right—but only if he or she made the mistake of taking the exam while sober! Studies of state-dependent retrieval suggest that if the student studied while drunk, he or she would probably perform poorly the next day, but he or she would perform better if he or she had cereal for breakfast instead of getting drunk (Eich, 1980; Weissenborn, 2000). Why should that be? Because a person’s physiological or psychological state at the time of encoding is associated with the information that is encoded. For example, being in a good mood affects patterns of electrical activity in parts of the brain responsible for semantic processing, suggesting that mood has a direct influence on semantic encoding (Kiefer et al., 2007). If the person’s state at the time of retrieval matches the person’s state at the time of encoding, the state itself serves as a retrieval cue—a bridge that connects the moment at which we experience something to the moment at which we remember it. Retrieval cues can even be thoughts themselves, as when one thought calls to mind another, related thought (Anderson et al., 1976).

The encoding specificity principle makes some unusual predictions. For example, you learned earlier that making semantic judgments about a word (e.g., What does brain mean?) usually produces more durable memory for the word than does making rhyme judgments (e.g., What rhymes with brain?). So if you were shown a cue card of the word “brain,” and your friend was asked to think about what brain means while you were asked to think of a word that rhymes with brain, we would expect your friend to remember the word better the next day if we simply asked you both, “Hey, what was that word you saw yesterday?” However, if instead of asking that question, we asked you both, “What was that word that rhymed with train?” the retrieval cue would match your encoding context better than your friend’s, and we would expect you to remember it better than your friend did. Data from memory experiments suggest that this is what happens in real-life situations (Fisher & Craik, 1977). This is a fairly astounding finding. The principle of transfer-appropriate processing is the idea that memory is likely to transfer from one situation to another when the encoding and retrieval contexts of the situations match (Morris, Bransford, & Franks, 1977; Roediger, Weldon, & Challis, 1989).

Human memory differs substantially from computer memory. Simply retrieving a file from your computer does not have any effect on the likelihood that the file will open again in the future. Not so with human memory. Retrieval does not merely provide a readout of what is in memory, it also changes the state of the memory system in important ways.

238

Psychologists have known for some time that the act of retrieval can strengthen a retrieved memory, making it easier to remember that information at a later time (Bjork, 1975). Does this finding surprise you? Probably not. Repeating an item usually improves memory for that item, so the act of retrieval might boost subsequent memory simply because the information is repeated, resulting in the same benefit you would receive from studying the information twice instead of once. Makes sense, right? Wrong. It turns out that retrieving information from memory has different effects than studying it again. This point was made dramatically in an experiment where participants studied brief stories and then either studied them again or were given a test that required retrieving the stories (Roediger & Karpicke, 2006). Participants were then given a final recall test for the stories either 5 minutes, 2 days, or 1 week later. As shown in FIGURE 6.9, at the 5-minute delay, studying the stories twice resulted in slightly higher recall than studying and retrieving them. Critically, the opposite occurred at the 2-day and 1-week delays: Retrieval produced much higher levels of recall than did an extra study exposure. A subsequent experiment using foreign vocabulary items also revealed that retrieval of the items produced a much bigger benefit on a delayed vocabulary test than did further study (Karpicke & Roediger, 2008). These findings have potentially important implications for learning in educational contexts (Karpicke, 2012), which we will explore further in the Learning chapter.

As much as retrieval can help memory, that is not always the case. Retrieval-induced forgetting is a process by which retrieving an item from long-term memory impairs subsequent recall of related items (Anderson, 2003; Anderson, Bjork, & Bjork, 1994).

Let us see how a typical experiment on retrieval-induced forgetting works (Anderson et al., 1994). Participants first studied word pairs consisting of a category name and an example from that category (e.g., fruit–orange, fruit–apple, tree–elm, tree–birch). Then they practised recalling some of the items from a few of the studied categories by retrieving the target in response to a category cue and the initial letters of the target. As an example, for the fruit category participants would practise recalling orange to the cue “fruit or____,” but would not practise recalling apple. The general idea is that while they are practising recall of orange, participants are trying to suppress the competitor apple. For other categories (e.g., trees), no retrieval practice was given for any of the studied pairs. Later, the participants were given a final test for all the words they initially studied. Not surprisingly, on the final test participants remembered words that they practised (e.g., orange) better than words from categories that they did not practise (e.g., elm). But what happened on the final test to the items such as apple, which were not practised, and which participants presumably had to suppress while they practised recall of related items? Those items were recalled most poorly of all, indicating that retrieving the similar target items (e.g., orange) caused subsequent forgetting of the related but suppressed items (e.g., apple). In fact, even if you do not successfully retrieve the target, the act of suppressing the competitors while you attempt to retrieve the target still reduces your ability to retrieve the competitors at a later time (Storm et al., 2006).

Can you think of any examples of how retrieval-induced forgetting could affect memory in everyday life? Here are two. First, retrieval-induced forgetting can occur during conversations: When a speaker selectively talks about some aspects of memories shared with a listener and does not mention related information, both the listener and the speaker later have a harder time remembering the omitted events (Cuc, Koppel, & Hirst, 2007; Hirst & Echterhoff, 2012). This effect occurs even for memories as important as those of the terrorist attacks in the United States on September 11, 2001 (Coman, Manier, & Hirst, 2009). Second, retrieval-induced forgetting can affect eyewitness memory. When witnesses to a staged crime are questioned about some details of the crime scene, their ability to later recall related details that they were not asked about is impaired compared with witnesses who were not questioned at all initially (MacLeod, 2002; Shaw, Bjork, & Handal, 1995). These findings suggest that initial interviews with eyewitnesses should be as complete as possible in order to avoid potential retrieval-induced forgetting of significant details that are not probed during an interview (MacLeod & Saunders, 2008).

239

In addition to improving and impairing subsequent memory, the act of retrieval can also change what we remember from an experience. Consider a recent experiment in which participants went on a tour of a museum, where they took some time to view designated exhibits, each of which contained several different stops (St. Jacques & Schacter, 2013). The participants took the tour while wearing a camera that, every 15 seconds, automatically took pictures of what was in front of them. Two days later, participants visited the memory laboratory (in a separate building) for a “reactivation session.” After memories of some of the stops were reactivated by looking at photographs of them, participants were asked to rate, on a 1 to 5 scale, how vividly they re-experienced what had happened at each stop. Next, the participants were shown novel photographs of unvisited stops within the exhibit, then were asked to judge how closely these novel photographs were related to the photographs of the stops that they had actually seen in that exhibit. Finally, the participants were given a memory test 2 days after the reactivation session.

Participants sometimes incorrectly remembered that the stop shown in the novel photograph had been part of the original tour. Most importantly, participants who tended to make this mistake also tended to have more vivid recollections during the reactivation session. In other words, retrieving and vividly re-experiencing memories of what participants actually did see at the museum led them to incorporate into their memory information that was not part of their original experience. This finding may be related to the phenomenon of reconsolidation that we discussed earlier, where reactivating a memory temporarily makes it vulnerable to disruption and change. At the very least, this finding reinforces the idea that retrieving a memory involves far more than a simple readout of information.

Before leaving the topic of retrieval, let us look at how the process actually works. There is reason to believe that trying to recall an incident and successfully recalling one are fundamentally different processes that occur in different parts of the brain (Moscovitch, 1994; Schacter, 1996). For example, regions in the left frontal lobe show heightened activity when people try to retrieve information that was presented to them earlier (Oztekin, Curtis, & McElree, 2009; Tulving et al., 1994). This activity may reflect the mental effort of struggling to dredge up the past event (Lepage et al., 2000). However, successfully remembering a past experience tends to be accompanied by activity in the hippocampal region (see FIGURE 6.10) (Eldridge et al., 2000; Giovanello, Schnyer, & Verfaellie, 2004; Schacter, Alpert, et al., 1996). Furthermore, successful recall also activates parts of the brain that play a role in processing the sensory features of an experience. For instance, recall of previously heard sounds is accompanied by activity in the auditory cortex (the upper part of the temporal lobe), whereas recall of previously seen pictures is accompanied by activity in the visual cortex (in the occipital lobe) (Wheeler, Petersen, & Buckner, 2000). Although retrieval may seem like a single process, brain studies suggest that separately identifiable processes are at work.

240

This sheds some light on the phenomenon we just discussed: retrieval-induced forgetting. Recent fMRI evidence indicates that during memory retrieval, regions within the frontal-lobe that are involved in retrieval effort play a role in suppressing competitors (Benoit & Anderson, 2012; Kuhl et al., 2007; Wimber et al., 2009). When hippocampal activity during retrieval practice signals successful recall of an unwanted competitor, frontal-lobe mechanisms are recruited that help to suppress the competitor. Once the competitor is suppressed, the frontal lobe no longer has to work as hard at controlling retrieval, ultimately making it easier to recall the target item (Kuhl et al., 2007). In addition, successful suppression of an unwanted memory causes reduced activity in the hippocampus (Anderson et al., 2004). These findings make sense once we understand the specific roles played by particular brain regions in the retrieval process.

241