|chapter 7||Interconnections between
Acquisition and Retrieval
Learning as Preparation for Retrieval Putting information into long-term memory helps you only if you can retrieve that information later on. Otherwise, it would be like putting money into a savings account without the option of ever making withdrawals, or writing books that could never be read. But let’s emphasize that there are different ways to retrieve information from memory. You can try to recall the information (“What was the name of your tenth-grade homeroom teacher?”) or to recognize it (“Was the name perhaps Miller?”). If you try to recall the information, a variety of cues may or may not be available (you might be told, as a hint, that the name began with an M or rhymes with “tiller”).
In Chapter 6, we largely ignored these variations in retrieval. We talked as if material was well established in memory or was not, with little regard for how the material would be retrieved from memory. There’s reason to believe, however, that we can’t ignore these variations in retrieval, and in this chapter we’ll examine the interaction between how a bit of information was learned and how it is retrieved later. Crucial Role of Retrieval Paths In Chapter 6, we argued that when you’re learning, you’re making connections between the newly acquired material and other information already in your memory. These connections make the new knowledge “findable” later on Specifically, the connections serve as retrieval paths: When you want to locate information in memory, you travel on those paths, moving from one memory to the next until you reach the target material.
These claims have an important implication. To see this, bear in mind that retrieval paths-like any paths-have a starting point and an ending point: The path leads you from Point A to Point B That’s useful if you want to move from A to B, but what if you’re trying to reach B from somewhere else? What if you’re trying to reach Point B, but at the moment you happen to be nowhere close to Point A? In that case, the path linking A and B may not help you.
As an analogy, imagine that you’re trying to reach Chicago from somewhere to the west. For this purpose, what you need is some highway coming in from the west. It won’t help that you’ve constructed a wonderful road coming into Chicago from the east. That road might be valuable in other circumstances, but it’s not the path you need to get from where you are right now to where you’re heading.
Do retrieval paths in memory work the same way? If so, we might find cases in which your learning is excellent preparation for one sort of retrieval but useless for other types of retrieval-as if you’ve built a road coming in from one direction but now need a road from another direction. Do the research data show this pattern? Context-Dependent Learning Consider classic studies on context-dependent learning (Eich, 1980; Overton, 1985). In one such study, Godden and Baddeley (1975) asked scuba divers to learn various materials. Some of the divers learned the material while sitting on dry land; others learned it while underwater, hearing the material via a special communication set. Within each group, half of the divers were then tested while above water, and half were tested below (see Figure 7.2).
Underwater, the world has a different look, feel, and sound, and this context could easily influence what thoughts come to mind for the divers in the study. Imagine, for example, that a diver is feeling cold while underwater. This context will probably lead him to think “cold-related” thoughts, so those thoughts will be in his mind during the learning episode. In this situation, the diver is likely to form memory connections between these thoughts and the materials he’s trying to learn.
Let’s now imagine that this diver is back underwater at the time of the memory test. Most likely he’ll again feel cold, which may once more lead him to “cold-related” thoughts. These thoughts, in turn, are now connected (we’ve proposed) to the target materials, and that gives us what we want: The cold triggers certain thoughts, and because of the connections formed during learning, those thoughts can trigger the target memories.
Of course, if the diver is tested for the same memory materials on land, he might have other links other memory connections, that will lead to the target memories. Even so, on land the diver will be at a disadvantage because the “cold-related” thoughts aren’t triggered-so there will be no benefit from the memory connections that are now in place, linking those thoughts to the sought-after memories.
By this logic, we should expect that divers who learn material while underwater will remember the material best if they’re again underwater at the time of the test. This setting will enable them to use the connections they established earlier. In terms of our previous analogy, they’ve built certain highways, and we’ve put the divers into a situation in which they can use what they’ve built. And the opposite is true for divers who learned while on land; they should do best if tested on land. And that is exactly what the data show (see Figure 7.3).
Similar results have been obtained in other studies, including those designed to mimic the learning situation of a college student. In one experiment, research participants read a two-page article similar to the sorts of readings they might encounter in their college courses. Half the participants read the article in a quiet setting half read it in noisy circumstances. When later given a short-answer test, those who read the article in quiet did best if tested in quiet-67% correct answers, compared to 54% correct if tested in a noisy environment. Those who read the article in a noisy environment did better if tested in a noisy environment-62% correct, compared to 46 %. ( See Grant et al., 1998; also see Balch, Bowman, & Mohler, 1992; Cann & Ross, 1989; Schab, 1990; Smith, 1985; Smith & Vela, 2001.)
In another study, Smith, Glenberg, and Bjork (1978) reported the same pattern if learning and testing took place in different rooms -with the rooms varying in appearance, sounds, and scent. In this study, though, there was an important twist: In one version of the procedure, the participants learned materials in one room and were tested in a different room. Just before testing, however, the participants were urged to think about the room in which they had learned-what it looked like and how it made them feel. When tested, these participants performed as well as those for whom there was no room change (Smith, 1979). What matters, therefore, is not the physical context but the psychological context-a result that’s consistent with our account of this effect. AS a result, you can get the benefits of context-dependent learning through a strategy of context reinstatement -re- creating the thoughts and feelings of the learning episode even if you’re in a very different place at the time of recall. That’s because what matters for memory retrieval is the mental context, not the physical environment itself. e. Demonstration 7.1: Retrieval Paths and Connections Often, the information you seek in memory is instantly available. For example, if you try to remember your father’s name, or the capital of France, the information springs immediately into your mind. Other times, however, the retrieval of information is more difficult.
How well do you remember your childhood? For example, think back to the sixth grade: How many of your sixth-grade classmates do you remember? Try writing a list of all their names on a piece of paper. Do it now, before you read any farther.
Now, read the following questions:
· What house did you live in when you were in the sixth grade? Think about times that friends came over to your house. Does that help you remember more names?
· Were you involved in any sports in the sixth grade? Think about who played on the teams with you. Does that help you remember more names?
· Where did you sit in the classroom in sixth grade? Who sat at the desk on your left? Who sat at the desk on your right? In front of you? Behind? Does that help you remember more names?
· Did you ride the bus to school, or carpool, or walk? Were there classmates you often saw on your way to or from school? Does that help you remember more names?
· Was there anyone in the class who was always getting in trouble? Anyone who was a fabulous athlete? Anyone who was incredibly funny? Do these questions help you remember more names?
Chances are good that at least one of these strategies, which help you “work your way back” to the names, did enable you to come up with some classmates you’d forgotten-and perhaps helped you to recall some names you hadn’t thought about for years!
Apparently, these “extra” names were in your memory, even though you couldn’t come up with them at first. Instead, you needed to locate the right retrieval path leading to the memory, the right precisely, once you were at the right connection. Once that connection was in your mind (or, more “starting point” for the path), it led you quickly to the target memory. This is just what we would expect, based on the claims in Chapter 7. Encoding Specificity The results we’ve been describing also illuminate a further point: what it is that’s stored in memory. Let’s go back to the scuba-diving experiment. The divers in this study didn’t just remember the words they’d learned; apparently, they also remembered something about the context in which the learning took place. Otherwise, the data in Figure 7.3 (and related findings) make no sense: If the context left no trace in memory, there’d be no way for a return to the context to influence the divers later.
Here’s one way to think about this point, still relying on our analogy. Your memory contains both the information you were focusing on during learning, and the highways you’ve now built, leading toward that information. These highways- the memory connections-can of course influence your search for the target information; that’s what we’ve been emphasizing so far. But the connections can do more: They can also change the meaning of what is remembered, because in many settings “memory plus this set of connections” has a different meaning from “memory plus that set of connections.” This change in meaning, in turn, can have profound consequences for how you remember the past. In one of the early experiments exploring this point, participants read target words (e.g., “piano”) in one of two contexts: “The man lifted the piano” or “The man tuned the piano.” In each case, the sentence led the participants to think about the target word in a particular way, and it was this thought that was encoded into memory. In other words, what was placed in memory wasn’t just the word “piano.” Instead, what was recorded in memory was the idea of “piano as something heavy” or “piano as musical instrument.”
This difference in memory content became clear when participants were later asked to recall the to recall the target word target words. If they had earlier seen the “lifted” sentence, they were likely if given the cue “something heavy.” The hint “something with a nice sound” was much less effective. But if participants had seen the “tuned” sentence, the result reversed: Now, the “nice sound” hint was effective, but the “heavy” hint wasn’t (Barclay, Bransford, Franks, McCarrell, & Nitsch, 1974). In both cases, the cue was effective only if it was congruent with what was stored in memory.
Other experiments show a similar pattern, traditionally called encoding specificity (Tulving, 1983; also see Hunt & Ellis, 1974; Light & Carter-Sobell, 1970). This label reminds us that what you encode (ie., place into memory) is indeed specific-not just the physical stimulus as you encountered it, but the stimulus together with its context. Then, if you later encounter the stimulus in some other context, you ask yourself, “Does this match anything I learned previously?” and you correctly answer no. And we emphasize that this “no” response is indeed correct. It’s as if you had learned the word “other” and were later asked whether you’d been shown the word “the.” In fact, “the” does appear as part of “other”-because the letters t h e do appear within “other But it’s the whole that people learn, not the parts. Therefore, if you’ve seen “other,” it makes sense to deny that you’ve seen the”- or, for that matter, “he” or “her”-even though all these letter combinations are contained within “other” Learning a list of words works in the same way. The word “piano” was contained in what the research participants learned, just as “the” is contained in “other” What was learned, however, wasn’t just this word. Instead, what was learned was the broader, integrated experience: the word as the perceiver understood it. Therefore, “piano as musical instrument” isn’t what participants learned if they saw the “lifted” sentence, so they were correct in asserting that this item wasn’t on the earlier list (also see Figure 7.4).
e. Demonstration 7.2: Encoding Specificity The textbook argues that the material in your memory is not just a reflection of the sights and sounds you’ve experienced. Instead, the material in your memory preserves a record of how you thought about these sights and sounds, how you interpreted and understood them. This demonstration, illustrating this point, is little complicated because it has three separate parts. First, you’ll read a list of words. Next, you should leave the demonstration and go do something else for 15 to 20 minutes-run some errands, perhaps, or do a bit of your reading for next week’s class. After that, your memory will be tested.
Here is the list of words to be remembered. For each word, a short phrase or cue is provided to help you focus on what the word means. Read the phrase or cue out loud, then pause for a second, then read the word, then pause for another second to make sure you’ve really thought about the word. Then move on to the next. Ready? Begin. HIDE A day of the week: Thursday A large city: Tokyo
A government leader: King A sign of happiness: Smile
A type of bird: Cardinal A student: Pupil
A famous psychologist: Freud A long word: Notwithstanding
A mane item: Wine Has four wheels: Toyota
A personality trait: Charm A part of a bird: Bill
A vegetable: Cabbage A member of the family: Grandfather
Associated with heat: Stove A happy time of year: Birthday
A round object: Ball A part of a word: Letter
Found in the jungle: Leopard A tool: Wrench
A crime: Robbery Found next to a highway: Motel
A baseball position: Pitcher A type of sport equipment: Racket
Associated with cold: North Part of a building: Chimney
Song accompaniment: Banjo Made of leather: Saddle
Take to a birthday party: Present A tropical plant: Palm
A girl’s name: Susan A synonym for “big”: Colossal
A type of footgear: Boots Associated with lunch: Noon
A man-made structure: Bridge Part of the intestine: Colon
A weapon: Cannon A sweet food: Banana
An assertion possession: Mine
Now, what time is it? Close the list of words and go do something else for 15 minutes, then come back for the next part of this demonstration.
Next, we’re going to test your memory for the words you learned earlier. To guide your efforts at recall, a cue will be provided for each of the words. Sometimes the cue will be exactly the same as the cue you saw before, and sometimes it will be different. In all cases, though, the cue will be closely related to the target word. There are no misleading cues.
On a piece of paper, write down the word from the previous list that is related to the cue. Do not look at the previous list. If you can’t recall some of the words, leave those items blank
Here are the answers. Check which ones you got right.
These words are obviously in groups of three. For the second word in each group (“Tokyo,” “Cannon,” etc.), the cue is identical to the cue you saw on the very first list. How many of these (out of 13) did you get right?
For the first word in each group (“Smile,” “Banana.” etc.), the cue is closely linked to the one you saw at first (“A sign of happiness” was replaced with “A facial expression,” and so on). How many of these (out of 13) did you get right?
For the third word in each group (“Mine,” “Bridge,” etc.), the cue actually changed the meaning of the target word. (On the first list, “Bridge” was “A manmade structure,” not “A card game”; “Racket” was “A type of sports equipment,” not “A type of noise.) How many of these (out of 13) did you get right? Most people do best with the identical cues and a little worse with the closely linked cues. Most people recall the fewest words with the cues that changed the meaning. Is this the pattern of your results? If so, your data fit with what the chapter describes as encoding specificity. This term reflects the fact that what goes into your memory isn’t just the words; it’s more specific than that-the words plus some record of what you thought about each word. As a result, what’s in your memory is not (for example) the word “bridge.” If that were your memory, a cue like “card game” might do the trick Instead, what’s in your memory is something like “structure used to get across a river,” and to trigger that idea, you need a different cue.
Demonstration adapted from Thieman, T. J. (1984). Table 1, in A classroom demonstration of encoding specificity. Teaching of Psychology, 11(2), 102. Copyright 1984 Routledge. Reprinted by permission from the publisher (Taylor & Francis Group, http://www.informaworld.com) The Memory Network In Chapter 6, we introduced the idea that memory acquisition-and, more broadly, learning-involves the creation (or strengthening) of memory connections. In this chapter, we’ve returned to the idea of memory connections, building on the idea that these connections serve as retrieval paths guiding you toward the information you seek. But what are these connections? How do they work? And who (or what?) is traveling on these “paths”?
According to many theorists, memory is best thought of as a vast network of ideas. In later chapters, we’ll consider how exactly these ideas are represented (as pictures? as words? in some more abstract format?). For now, let’s just think of these representations as nodes within the network, just like the knots in a fisherman’s net. (In fact, the word “node” is derived from the Latin word for knot, nodus.) These nodes are tied to each other via connections we’ll call associations or associative links. Some people find it helpful to think of the nodes as being like light bulbs that can be turned on by incoming electricity, and to imagine the associative links as wires that carry the electricity. Spreading Activation Theorists speak of a node becoming activated when it has received a strong enough input signal. Then, once a node has been activated, it can activate other nodes: Energy will spread out from the just-activated node via its associations, and this will activate the nodes connected to the just- activated node.
To put all of this more precisely, nodes receive activation from their neighbors, and as more and more activation arrives at a particular node, the activation level for that node increases. Eventually the activation level will reach the node’s response threshold. Once this happens, we say that the node fires. This firing has several effects, including the fact that the node will now itself be a source of activation, sending energy to its neighbors and activating them. In addition, firing of the node will draw attention to that node; this is what it means to “find” a node within the network.
Activation levels below the response threshold, so-called subthreshold activation, also play an important role. Activation is assumed to accumulate, so that two subthreshold inputs may add together, in a process of summation, and bring the node to threshold. Likewise, if a node has been partially activated recently, it is in effect already “warmed up,” so that even a weak input will now be sufficient to bring it to threshold.
These claims mesh well with points we raised in Chapter 2, when we considered how neurons communicate with one another. Neurons receive activation from other neurons; once a neuron reaches its threshold, it fires, sending activation to other neurons. All of this is precisely parallel to the suggestions we’re describing here. Our current discussion also parallels claims offered in Chapter 4, when we described how a network of detectors might function in object recognition. In other words, the network linking memories to each other will resemble the networks we’ve described linking detectors to each other (e.g., Figures 4.9 and 4.10). Detectors, like memory nodes, receive their activation from other detectors; they can accumulate activation from different inputs, and once activated to threshold levels, they fire.
Returning to long-term storage, however, the key idea is that activation travels from node to node via associative links. As each node becomes activated and fires, it serves as a source for further activation, spreading onward through the network. This process, known as spreading activation, enables us to deal with a key question: How does one navigate through the maze of associations? If you start a search at one node, how do you decide where to go from there? The answer is that in most cases you don’t “choose” at all. Instead, activation spreads out from its starting point in all directions simultaneously, flowing through whatever connections are in place. Retrieval Cues This sketch of the memory network leaves a great deal unspecified, but even so it allows us to explain some well-established results. For example, why do hints help you to remember? Why, for example, do you draw a blank if asked, “What’s the capital of South Dakota?” but then remember if given the cue “Is it perhaps a man’s name?” Here’s one likely explanation. Mention of South Dakota will activate nodes in memory that represent your knowledge about this state. Activation will then spread outward from these nodes, eventually reaching nodes that represent the capital city’s name. It’s possible, though, that there’s only a weak connection between the SOUTH DAKOTA nodes and the nodes representing PIERRE. Maybe you’re not very familiar with South Dakota, or maybe you haven’t thought about this state’s capital for some time. In either case, this weak connection will do a poor job of carrying the activation, with the result that only a trickle of activation will flow into the PIERRE nodes, and so these nodes won’t reach threshold and won’t be “found.”
Things will go differently, though, if a hint is available. If you’re told, “South Dakota’s capital is also a man’s name,” this will activate the MAN’S NAME node. As a result, activation will spread out from this source at the same time that activation is spreading out from the SOUTH DAKOTA nodes. Therefore, the nodes for PIERRE will now receive activation from two sources simultaneously, and this will probably be enough to lift the nodes’ activation to threshold levels. In this way, question- plus-hint accomplishes more than the question by itself (see Figure 7.5).
Semantic Priming The explanation we’ve just offered rests on a key assumption-namely, the summation of subthreshold activation. In other words, we relied on the idea that the insufficient activation received from one source can add to the insufficient activation received from another source. Either source of activation on its own wouldn’t be enough, but the two can combine to activate the target nodes.
Can we document this summation more directly? In a lexical-decision task, research participants are shown a series of letter sequences on a computer screen. Some of the sequences spell words other sequences aren’t words (e.g., “blar, plome”). The participants’ task is to hit a “yes” button if the sequence spells a word and a “no” button otherwise. Presumably, they perform this task by “looking up” these letter strings in their “mental dictionary,” and they base their response on whether or not they find the string in the dictionary. We can therefore use the participants’ speed of response in this task as an index of how quickly they can locate the word in their memories.
In a series of classic studies, Meyer and Schvaneveldt (1971; Meyer, Schvaneveldt, & Ruddy, 1974) presented participants with pairs of letter strings, and participants had to respond “yes” if both strings were words and “no” otherwise. For example, participants would say “yes” in response to “chair, bread” but “no” in response to “house, fime.” Also, if both strings were words, sometimes the words were semantically related in an obvious way (e.g., “nurse, doctor”) and sometimes they weren’t (“cake. shoe”). Of interest was how this relationship between the words would influence performance. Consider a trial in which participants see a related pair, like “bread, butter.” To choose a response, they first need to “look up” the word “bread” in memory. This means they’ll search for, and presumably activate, the relevant node, and in this way they’ll decide that, yes, this string is a legitimate word. Then, they’re ready for the second word. But in this sequence, the node for BREAD (the first word in the pair) has just been activated. This will, we’ve hypothesized, trigger a spread of activation outward from this node, bringing activation to other, nearby nodes. These nearby nodes will surely include BUTTER, since the association between “bread” and “butter” is a strong one. Therefore, once the BREAD node (from the first word) is activated, some activation should also spread to the BUTTER node.
From this base, think about what happens when a participant turns her attention to the second word in the pair. To select a response, she must locate “butter” in men finds the relevant node), then she knows that this string, too, is a word, and she can hit the “yes” button. But the process of activating the BUTTER node has already begun, thanks to the (subthreshold) activation this node just received from BREAD. This should accelerate the process of bringing this node to threshold (since it’s already partway there), and so it will require less time to activate. As a result, we expect quicker responses to “butter” in this context, compared to a context If she finds this word (i.e. in which “butter” was preceded by some unrelated word. Our prediction, therefore, is that trials with related words will produce semantic priming. The “priming” indicates that a specific prior event (in this case, presentation of the first word in the pair) will produce a state of readiness (and, therefore, faster responding) later on. There are various forms of priming (in Chapter 4, we discussed repetition priming). In the procedure we’re considering here, the priming results from the fact that the two words in the pair are related in meaning- therefore, this is semantic priming.
The results confirm these predictions. Participants’ lexical-decision responses were faster by almost 100 ms if the stimulus words were related (see Figure 7.6), just as we would expect on the term model we’re developing. (For other relevant studies, including some alternative conceptions of priming, see Hutchison, 2003; Lucas, 2000.)
Before moving on, though, node activating nearby nodes-is not the whole story for memory search. As one complication, we should mention that this process of spreading activation-with one people have some the processes of reasoning (Chapter 12) and the mechanisms of executive control (Chapters 5 and 6). In addition, evidence suggests that once the spreading activation has begun, people have the option of “shutting down” some of this spread if they’re convinced that the wrong nodes are being activated (e.g., Anderson & Bell, 2001; Johnson & Anderson, 2004). Even so, spreading activation is a crucial degree of control over the starting points for their memory searches, relying on us understand why memory connections mechanism. It plays a central role in retrieval, and it helps are so important and so helpful. e. Demonstration 7.3: Spreading Activation in Memory Search On a piece of paper, list all of the men’s first names you can think of that are also verbs. For example, you can Mark something on paper; you shouldn’t Rob a bank. If you’re willing to ignore the spelling, you can Neil before the queen and Phil a bucket. How many other men’s names are also verbs? Spend a few minutes generating the list.
How do you search your memory to come up with these names? One possibility is that you first think of all the men’s names that you know, and then from this list you select the names that work as verbs. A different possibility reverses this sequence: You first think of all the verbs that you know and from this list you select the words that are also names. One last possibility is that you combine these steps, so that your two searches go on in parallel: In essence, you let activation spread out in your memory network from the MEN’S NAMES nodes, and at the same time you let activation spread out from the VERBS nodes. Then, you can just wait and see which nodes receive activation from both of these sources simultaneously. In fact, the evidence suggests that the third option (simultaneous activation from two sources) is the one you use. We can document this by asking a different group of people just to list all the verbs they know. When we do this, we find that some verbs come to mind only after a long delay-if at all. For example, if you’re just thinking of verbs, the verb “rustle” may not pop into your thoughts. If, therefore, you were trying to think of verbs-that-are-also-names by first thinking about verbs and then screening them, you’re unlikely to come up with “rustle” in your initial step (i.e., generating a list of verbs). Therefore, you won’t think about “rustle” in this setting, and so you won’t spot the fact that it’s also a man’s name (“Russell”). On this basis, this name won’t be one of the names on your list.
The reverse is also true. If you’re just thinking about men’s names, the name “Russell” may not spring to mind, and so, if this is the first step in your memory search (i.e., first generate a list of names; then screen it, looking for verbs), you won’t come up with this name in the first place. Therefore, you won’t consider this name, won’t see that it’s also a verb, and won’t put it on your list.
It turns out, though, that relatively rare names and rare verbs are often part of your final output. This makes no sense if you’re using a “two-step” procedure (first generate names, then screen them; would n up i he first or first generate verbs, then screen them) because the key w step of this process. But the result does make sense if your memory search combines the two steps. In that case, even though these rare items are only weakly activated by the MEN’S NAMES nodes, and only weakly activated by the VERBS nodes, they are activated perfectly well if they can receive energy from both time-and that is why these rare items come easily to mind. And, by the way, there are at least 50 men’s names that are also verbs, so keep hunting for them! It may help to remember that Americans Bob for apples at Halloween. Yesterday, I Drew a picture and decided to Stu the beef for dinner. I can Don a suit, Mike a speaker, Rush to an appointment, Flip a pancake, or Jimmy a locked door. These are just some of the names that could be on your list! e. Demonstration 7.4: Semantic Priming As Chapter 7 describes, searching thr ough long-term memory relies heavily on a process of spreading activation, with currently activated nodes sending activation outward to their neighbors. If this spread brings enough activation to the neighbors, then those nodes will themselves become activated. However, even if these nodes don’t receive enough activation to become activated themselves, the subthreshold activation still has important effects.
Here is a list of anagrams (words for which we’ve scrambled up the letters). Can you unscramble them to figure out what each of the words is?
Did you get them all? Continue in order to see the answers.
The answers, in no particular order, are “sea,” “shirt.” “victor,” “island,” “mountain,” “wave,” “pilot.” and-what? The last anagram in the list actually has two solutions: It could be an anagram for the boat used in North America to explore lakes and streams, or it could be an anagram for the body of water that sharks and whales and sea turtles live in. Which of these two solutions came to your mind? If you happen to be a devoted paddler, then good that “ocean” is the word “canoe” may have come rapidly into your thoughts. But the odds are that came to mind for you. Why is this? Several of the other words in this series (“sea,” “island” “mountain, “wave”) are semantically associated with “ocean.” Therefore, when you solved these earlier anagrams, you activated nodes for these words, and the activation spread outward from there to the neighboring nodes-including, probably, OCEAN. As a result, the word “ocean” was already primed when you turned to the last anagram, making it likely that this word, and not the legitimate alternative, would come into your thoughts as you unscrambled NOCAE. Different Forms of Memory Testing Let’s pause to review. In Chapter 6, we argued that learning involves the creation or strengthening of connections. This is why memory is promoted by understanding (because understanding consists, in large part, of seeing how new material is connected to other things you know). We also proposed that these connections later serve as retrieval paths, guiding your search through the vast warehouse that is memory. In this chapter, weve explored an important implication of this idea: that (like all paths) the paths through memory have both a starting point and an end point. Therefore, retrieval paths will be helpful only if you’re at the appropriate starting point; this, we’ve proposed, is the basis for the advantage produced by context reinstatement. And, finally, we’ve now started to lay out what these paths really are: connections that carry activation from one memory to another.
This theoretical base also helps us with another issue: the impact of different forms of memory testing. Both in the laboratory and in day-to-day life, you often try to recall information from memory. This means that you’re presented with a retrieval cue that broadly identifies the information you seek, and then you need to come up with the information on your own: “What was the name of that great restaurant your parents took us to?”; “Can you remember the words to that song?”; “Where were you last Saturday?” In other circumstances, you draw information from your memory via recognition. This term refers to cases in which information is presented to you, and you must decide whether it’s the sought-after information: “Is this the man who robbed you?”; “I’m sure I’ll recognize the street when we get there”; “If you let me taste that wine, I’ll tell you if it’s the same one we had last time.”
These two modes of retrieval-recall and recognition-are fundamentally different from each other. Recall requires memory search because you have to come up with the sought-after item on your own; you need to locate that item within memory. As a result, recall depends heavily on the so far. Recognition, in contrast, often depends memory connections we’ve been emphasizing sense of familiarity. Imagine, for example, that you’re taking a recognition test, and the fifth word on the test is “butler.” In response to this word, you might find yourself thinking, “I don’t recall seeing on a this word on the list, but this word feels really familiar, so I guess I must have seen it recently Therefore, it must have been on the list.” In this case, you don’t have source memory; that is, you don’t have any recollection of the source of your current knowledge. But you do have a strong sense of familiarity, and you’re willing to make an inference about where that familiarity came from. In other words, you attribute the familiarity to the earlier encounter, and thanks to this attribution you’ll probably respond “yes” on the recognition test. Familiarity and Source Memory We need our terms here, because source memory is actually a type of recall. Let’s say, for example, that you hear a song on the radio and say, “I know I’ve heard this song before because it feels familiar and I remember where I heard it.” In this setting, you’re able to remember the source of your familiarity, and that means you’re recalling when and where you encountered the song. On this basis, we don’t need any new theory to talk about source memory, because we can use the same theory that we’d use for other forms of recall. Hearing the song was the retrieval cue that launched a search through memory, a search that allowed you to identify the setting in which you last encountered the song. That search (like any search) was dependent on memory connections, and would be explained by the spreading activation process that we’ve already described.
But what about familiarity? What does this sort of remembering involve? As a start, let’s be clear that familiarity is truly distinct from source memory. This is evident in the fact that the two types of memory are independent of each other-it’s possible for an event to be familiar without any source memory, and it’s possible for you to have source memory without any familiarity. This independence is evident when you’re watching a movie and realize that one of the actors is familiar, but (sometimes with considerable frustration, and despite a lot of effort) you can’t recall where you’ve seen that actor before. Or you’re walking down the street, see a familiar face, and find yourself asking, “Where do I know that woman from? Does she work at the grocery store I shop in? Is she the driver of the bus I often take?” You’re at a loss to answer these questions; all you know is that the face is familiar.
In cases like these, you can’t “place” the memory; you can’t identify the episode in which the face was last encountered. But you’re certain the face is familiar, even though you don’t know why-a clear example of familiarity without source memory. The inverse case is less common, but it too can be demonstrated. For example, in Chapter 2 we discussed Capgras syndrome. Someone with this syndrome might have detailed, accurate memories of what friends and family members look like, and probably remembers where and when these other people were last encountered. Even so, when these other people are in view they seem hauntingly unfamiliar. In this setting, there is source memory without familiarity. (For further evidence-and a patient who, after surgery, has intact source memory but disrupted familiarity-see Bowles et al., 2007; also see Yonelinas & Jacoby, 2012.)
We can also document the difference between source memory and familiarity in another way. In many studies, (neurologically intact) participants have been asked, during a recognition test,