Remembering, Knowing, and Feeling

in Episodic Recognition

John F. Kihlstrom, 

Michael Kim, and Marilyn Dabady

Yale University

If one of the hallmarks of the "consciousness revolution" in psychology is the burgeoning industry devoted to comparing explicit and implicit expressions of memory, another is the research which has sought to draw our attention to the phenomenal experience of remembering the past. This latter movement began with Tulving (1985), who distinguished between two quite different recollective experiences: remembering, one's concrete awareness of oneself (autonoetic consciousness) in the past, and knowing, one's abstract knowledge (noetic consciousness) of the past.

The distinction between remembering and knowing was further developed by Gardiner (1988, Exp. 1), who asked subjects to study list of words under phonemic or semantic orienting tasks, and then to complete a yes/no recognition test. Whenever the subjects endorsed an item as old, they were asked to indicate the nature of their recollective experience. If recognition of a particular item was accompanied by conscious recollection of the its occurrence on the study list, they were to rate the item as "remembered"; if not, they were to rate it as "known". About 25% of recognized items were assigned to the Know category, meaning that they were recognized in the absence of recollective experience. More to the point, the experiment yielded a dissociation between level of processing and recollective experience: Remember judgments were affected by LoP, but Know judgments were not.

Although Gardiner’s development of the remember-know paradigm was inspired by Tulving’s 1985 paper, his conception and Tulving's seem to be somewhat at cross purposes. For Tulving, the remember-know distinction maps onto his (Tulving, 1972) distinction between episodic memory, or the person’s awareness that an event "is a veridical part of his own past existence" (1985, p. 3), and semantic memory, or the person's "symbolic knowledge of the world" (1985, p. 3). But for Gardiner, the remember-know distinction maps more closely onto Mandler’s (1980) distinction between recognition by retrieval and recognition by familiarity. Recognition by retrieval involves remembering an event as an event, including the personal and spatiotemporal context in which the event occurred; by contrast, recognition by familiarity involves a feeling or intuition that some event occurred in the past, in the absence of conscious recollection of that event. For Gardiner, Remember judgments reflect recognition by retrieval, while Know judgments reflect recognition by familiarity. An alternative framework is provided by Schacter’s (1987) distinction between explicit and implicit memory. Explicit memory involves the conscious recollection of an experience from the past, while implicit memory is a memory-based change in behavior that occurs independent of, and in the classic case in the absence of, conscious recollection. For Gardiner, remembering reflects explicit memory, while knowing reflects implicit memory.

These two alternative interpretations of remembering and knowing are connected, from Mandler's point of view, because both recognition by familiarity and implicit memory are based on the activation by an event of previously stored knowledge (Schacter doesn’t endorse this activation view of implicit memory, but this is another matter). At the same time, we think it's a mistake to conflate semantic memory with implicit memory, and knowing with intuiting.

Put another way, our sense is that there are at least three varieties of recollective experience -- or, if you will, three different memory qualia: remembering involves the conscious recollection of some past event, as an explicit expression of episodic memory; knowing is abstract knowledge of that event, as an item in semantic memory; and feeling is the intuition that an event occurred in the past, as an implicit expression of episodic memory. In fact, there may be a fourth memory quale, believing, in which we surmise that an event occurred in the absence of any recollective experience at all -- more like a judgment than a memory, though along with Mandler (1980) and Bartlett (1932), we tend to think of all acts of remembering as acts of judgment.

These memory qualia are familiar to anyone who has ever taken a multiple-choice test. Sometimes, we choose a response because we remember the circumstances under which we learned it -- the particular lecture, or, as sometimes happens, the location on the textbook page where the information appeared (for an experimental demonstration of this, see Rothkopf, JVLVB 1971). On other occasions, we choose a response because we just know the answer -- it’s part of our knowledge about the world, and we don’t remember the circumstances under which we learned the answer. On still other occasions, we choose a response because we intuit that it’s the right one. We don’t actually know the answer, and we certainly don’t remember where we learned it, but we choose a response because it strikes us as familiar, and we infer from this feeling of familiarity that, of all the choices available, this is the one that is most likely to be correct. We’re guessing, in a way that we’re not guessing when we’re remembering or knowing, but it’s not random guessing; it’s guessing informed by the feeling of familiarity. Finally, on still other occasions, we choose a response because we believe, on some non-experiential grounds, that it is the correct one. For example, we might reason to the answer based on deduction from some general concept or principle. We don’t know the answer, and in fact our deduction might be wrong. We don’t remember learning the answer at any past time, and we don’t have any intuitions that we did. We just believe, on grounds of our general world knowledge, that our choice is the right one. Analogous experiences occur in face recognition.

In an attempt to tease these recollective experiences apart, we performed an extension of Gardiner's (1988, Exp. 1) reference experiment, taking account of the difference between knowing and feeling, and setting the issue of believing aside for another set of experiments. Our subjects studied a list of 80 words taken from a list constructed by Rajaram (1993), under conditions of a levels of processing manipulation: for 40 items they supplied rhymes, and for 40 items they supplied semantic associate. After a 24-hour retention interval, they performed a yes/no recognition test, distinguishing between 80 targets and 80 lures, and for each item recognized, reported their accompanying phenomenology in terms of remembering, knowing, or feeling, using instructions adapted from Gardiner (1988) and Rajaram (1993).

On a computer screen you will see a number of words presented one at a time. For each word, please indicate, by pressing the appropriate key, whether or not you recognize that word from the study list which was presented earlier. If you do not recognize the word, press "N"... If you do recognize the word, indicate the basis for your judgment of familiarity -- that is, whether you actually remember the word from the study list or whether there is some other basis for your recognition judgment.

Remember judgments. If your recognition of the word is accompanied by a conscious recollection of its prior occurrence in the study list, then press the key marked "R" (for "Remember"). "Remember" means that you are consciously aware again of some aspect or aspects of what happened or what was experienced at the time the word was presented -- for example, aspects of the physical appearance of the word on the computer screen, or of something that happened in the room, such as the buzzing of a fly in the background, or of what you were thinking and doing at the time. In other words, a "remembered" word should bring back to mind a particular association, image, or something more personal from the time of study, or something about its appearance or position -- for example, what came before or after that word.

If you cannot consciously recollect anything about the actual occurrence or what happened or what was experienced at the time of its occurrence, you should not respond with an "R" (for "Remember"), but should make one of two other possible responses.

Know judgments. In this case you might simply know, for example, that the word was on the list, in the same way that you know your own name, or birthday, or that Bill Clinton is President of the United States. You don't remember anything about the experience of acquiring this knowledge, you just know that it's true. In the same way, you might know that a word was presented for study, even though you don't remember anything about what you experienced at the time that the word was presented. If this is the case, you should press the key marked "K" (for "Know").

Feel judgments. Alternatively, you might have a feeling or intuition that the word was on the list without actually remembering its appearance as such. This experience is the kind of experience that occurs when you're at a party, and you see someone across the room who strikes you as familiar. You don't know who the person is, and you don't actually remember ever having met him or her before, but your feeling is such that if you had to guess, you'd guess that you had met that person before, or that you'd recognize his or her name. If this is the case, you should press they key marked "F" (for "Feel").

To give you a better idea of the three distinctions that we have mentioned, consider how you might respond on a multiple-choice test in school. Sometimes, you just know the answer; other times, you actually remember having learned the material, such as where it appears on a page in your textbook; sometimes an answer just rings a bell, so you feel that it must be the right choice.

The experiment was controlled by the ECL software.

Our initial analyses paralleled those of Tulving and Gardiner, focusing on recognition hits, ignoring false alarms.

Results of Experiment 1

First, considering overall recognition, we got the expected levels of processing effect: the subjects recognized 75% of the items from the phonemic condition, and 82% of the items from the semantic condition, a significant difference. The levels of processing effect was also obtained when we considered only those recognized items that were rated as remembered. When we broke recognition down into its experiential components, however, less than half of recognized items were rated as remembered, with the remainder evenly split between knowing and feeling.

Recall that in the traditional remember-know paradigm, as applied by Tulving, Gardiner, and others, "Know" is a fallback category: if a subject recognizes an item and also consciously recollects it, the item is classified as remembered; otherwise it must be classified as known. Accordingly, we combined our know and feel items to yield a category of "G-Know" items, or items "known" according to the criterion established by Gardiner, and contrasted these with the Remember category. The result was a dissociation between level of processing and recollective experience similar to those observed by Gardiner, Rajaram, and others. Level of processing affects Remember responses, but not G-Know responses.

Of course, as Rajaram (1993) has noted, this interaction is not quite kosher: because Remember and G-Know exhaust the components of recognition, any variable which affects one component must affect the other component in the other direction. The solution, as Rajaram proposed, is to calculate the ratio of Remembered items to all recognition hits. This also yielded a significant levels of processing effect.

But now when we break up G-Know responses into their components, Know and Feel, we get a further dissociation. In terms level of processing, there are more Know responses with deep than with shallow processing, as was the case with Remember responses; but the effect reverses with Feel responses, of which there are more with shallow than with deep processing. The crossover interaction is significant, so this also counts as a dissociation with level of processing.

By the way, it should be understood that the Rajaram (1993) correction is not necessary in this case, because we have broken recognition down into three components, and tested the interaction on only two, Know and Feel. The leftover degree of freedom allows us to consider Know and Feel judgments as independent of each other, and so the Rajaram correction is not necessary.

But the most important point is that Know and Feel responses are dissociable in the same way that Remember and G-Know responses are. Apparently, the conventional definition of G-Know as a fallback category, including anything that isn't consciously recollected, obscures important distinctions between other forms of recollective experience. There's remembering, and knowing, but there's also feeling -- at the very least.

Now, Tulving and Gardiner believe that remember and know judgments are

based on retrieval from different memory systems -- episodic vs. semantic memory, perhaps, or explicit vs. implicit memory. However, it could also be that "remember" and "know" are based on retrieval from a single memory system, and that the categories of remember, know, and so forth are proxies for different levels of confidence associated with the recognition judgments. Both Tulving (1985) and Gardiner (1988) have rejected this interpretation, even though Tulving (1985) actually gathered evidence favoring it. Tulving's (1985, Exp. 2) subjects studied 36 words, and then made Yes/No recognition judgments, confidence ratings (on a 3 point scale), and Remember/Know ratings. The average confidence rating associated with Remember judgments was 2.74, while that of Know judgments was 2.08 (alas, there was no statistical test).

However, Gardiner & Java (1990) argued that confidence ratings contaminate Remember/Know judgments: that is, subjects might base their confidence ratings on their recollective experience, so that the two are not independent. In their Experiment 2, the subjects studied 60 items, 30 words and 30 nonwords, and then made Yes/No recognition judgments followed by Remember/Know ratings. The result was a double dissociation: more words received remember than know judgments, while the reverse was true for nonwords. Experiment 3 was identical to Experiment 2, except that the subjects classified recognized items into "Sure" and "Unsure" categories. This time there was no such dissociation. Rajaram (1993, Experiments 3 and 4) performed a similar pair of experiments, with similar results, and came to same conclusion. Because substituting Sure/Unsure ratings for Remember/Know judgments abolished the dissociations observed with Remember/Know, both Gardiner and Java (1990) and Rajaram (1993) conclude that Remember/Know is not merely a proxy for confidence.

More recently, however, Donaldson (1996) has argued from signal detection theory that the Remember/Know distinction is a matter of the placement of the decision criterion. Frankly, we're don't see how different recollective experiences can help but be associated with different levels of confidence, especially if we distinguish between Knows and Feels. If you know that something happened like you know your own name, that knowledge must be associated with high level of confidence; and if you feel that something happened, but don't actually remember it, or otherwise know for sure, this must be associated with low levels of confidence.

Our approach to this problem has been to experimentally manipulate subjects' criteria for recognition, following a procedure employed by Dorfman, Kihlstrom, Cork, & Misiaszek (1995). Our Experiment 2 followed same procedure as Experiment 1: the subjects studied a list of 80 words, 40 in the rhyme condition and 40 in the associate condition; 24 hours later they completed the Yes/No recognition test and RKF ratings. For half the test items, the subjects were instructed to adopt a strict criterion for recognition: not to say yes unless they were relatively sure:

Say "Yes" only if you are relatively certain that the word appeared on the presentation list. If you’re fairly sure you say the word, then you should say "yes". If you’re not so sure or don’t think you say the word, you should say "no".

For the remainder, they were instructed to adopt a liberal criterion: to yes if the item seemed familiar in any way, even if they did not actually remember it:

You can say "Yes" even if you’re not sure that you saw the word. If you think you might have seen the word, or if it seems familiar to you at all, then you should say "yes". If it doesn’t seem familiar to you, you should say "no".

 

Results of Experiment 2

Looking again just at recognition hits, as expected we obtained a levels of processing effect on recognition under both criteria; the shift from strict to liberal criteria also increased hits, although admittedly it increased false alarms as well. However, consistent with Donaldson's (1996) findings, the criterion shift had no influence on "R" judgments. The effects of the criterion shift are seen elsewhere.

As before, when we look at the distribution of recognition hits, collapsing across the two recognition criteria, we see that only about half of recognized items are actually remembered, with the remainder about evenly split between items that are known and those that are felt.

Next, we combined Know and Feel responses into the G-Know category, to replicate the traditional Remember/Know procedures. Again, we got a significant crossover interaction, both overall and within each criterion for recognition, confirmed by the Rajaram (1993) correction. Notice, though, that G-Know recognition goes up as subjects shift from the strict to the liberal criterion -- an indication that the distinction between remembering and "knowing" might have something to do with recognition confidence after all.

Things become clearer when we distinguish between Know and Feel responses within the G-Know category. First, we get the same interaction as last time: Knows look like Remembers, with more from the deep than from the shallow processing condition; Feels are the reverse, with more from the shallow than from the deep. Note that there are fewer Know than Remember responses, as there should be if Know responses reflect the kind of certainty that we have about our own names. When we examine the effects of the criterion manipulation, we see that there is no change in either Remember or Know responses; however, there is a significant increase in Feel responses, and there is an increase in false alarms as well as hits in this category -- just as there should be if subjects are making informed guesses about words they might have studied.

On the basis of these studies, we conclude that the traditional Remember/Know distinction, intended to reveal important aspects of recollective experience, actually obscures important phenomenological distinctions. There is not just remembering and knowing, but there is feeling as well, and perhaps even others, like believing. These memory qualia are differentially associated with confidence level. Remember and Know judgments are associated with higher degrees of confidence, and yield fewer false alarms, than their Feeling (and, probably, Believing) counterparts. Likewise, items identified as "Known" according to the criteria proposed by Tulving and Gardiner are associated with generally lower levels of confidence, and generally increased false alarms. But we think that these different memory qualia are not just proxies for confidence. Remember and Know judgments are not affected by criterion shifts, and so something else is going on.

Although the Remember/Know distinction is commonly interpreted in terms of different memory systems, we suspect instead that these different memory qualia reflect retrieval of different information from a single common store. Know judgments require retrieval only of information about list membership, while Feeling judgments need only the perceptual salience that comes from activation of the lexical nodes representing list items. Remember judgments, however, require retrieval of information about spatiotemporal context, and the self as the agent or experiencer of the event. All this information can be stored within a single memory system; which information is retrieved will determine the quality of the person's recollective experience.

Author Notes

Paper presented at the annual meeting of the Psychonomic Society, Chicago, November 1996. This research was supported by Grant #MH-35856 from the National Institute of Mental Health. We thank Woo-kyoung Ahn, Marvin Chun, Robert Crowder, Jennifer Dorfman, and Leticia Naigles for their comments. Correspondence to John F. Kihlstrom, Department of Psychology, Yale University, P.O. Box 208205, New Haven, Connecticut 06520-8205. Email: kihlstrm@minerva.cis.yale.edu.

References

Donaldson, W. (1996). The role of decision processes in remembering and knowing. Memory & Cognition, 24, 523-533.

Dorfman, J., Kihlstrom, J.F., Cork, R.C., & Misiaszek, J. (1995). Priming and recognition in ECT-induced amnesia. Psychonomic Bulletin & Review, 2, 244-248.

Gardiner, J.M. (1988). Functional aspects of recollective experience. Memory & Cognition, 16, 309-313.

Gardiner, J.M., & Java, R.I. (1990). Recollective experience in word and nonword recognition. Memory & Cognition, 18, 23-30.

Rajaram, S. (1993). Remembering and knowing: Two means of access to the personal past. Memory & Cognition, 21, 89-102.

Mandler, G. (1980). Recognizing: The judgment of prior occurrence. Psychological Review, 87, 252-271.

Schacter, D.L. (1987). Implicit memory: History and current status. Journal of Experimental Psychology: Learning, Memory, & Cognition, 13, 501-518.

Tulving, E. (1972). Episodic and semantic memory. In E. Tulving & W. Donaldson (Eds.), Organization of memory (pp. 381-403). New York: Academic Press.

Tulving, E. (1985). Memory and consciousness. Canadian Psychology, 26, 1-12.