John F. Kihlstrom
University of California, Berkeley
The distinction between remembering and knowing (Gardiner, 1988; Rajaram, 1993; Tulving, 1985) continues to generate a great deal of interest, but over the past couple of years we have argued that it is not quite on the mark. Instead, we have argued that there are at least three different recollective experiences associated with remembering one's personal past.
Remembering, an explicit expression of episodic memory, involves the conscious recollection of some past event. In some sense, remembering involves a re-experiencing of the event, the spatial and temporal context in which it occurred, and the role of the rememberer as the agent or patient of some action, or as the stimulus or experiencer of some state.
Knowing, an expression of semantic memory, involves abstract knowledge of a past event. Instead of remembering an event, one knows that it occurred, the same way one knows that Columbus discovered America in 1492.
Feeling, an implicit expression of episodic memory, involves the intuition that an event occurred. The event is not consciously remembered, but it "on the tip of the tongue", and its description "rings a bell". If you had to guess, you'd affirm that it happened, and you'd usually be right.
We think of these experiences as memory "qualia". Any of them can form the basis for episodic memory, and there may well be others.
At the past two meetings, Michael Kim and I have described a series of experiments which support this tripartite distinction among memory "qualia". In one experiment (Kihlstrom, Kim, & Dabady, 1996, Experiment 1), we asked subjects to make acoustic and semantic judgments concerning a list of 80 words in an experiment closely patterned on those reported by Tulving, Gardiner, and others. The next day they performed a yes/no recognition task; for each item they affirmed, they classified their recollective experience as one of remembering, knowing, or feeling.
When we followed the lead of Tulving and Gardiner, distinguishing only between remembering and everything else, we found, as they did, that remembering was affected by level of processing but "knowing", as defined by Tulving and Gardiner, was not.
But when we further distinguished between knowing and feeling, we obtained a double dissociation: setting aside the category of remembering, so that knowing and feeling could vary freely, we found that knowing was associated with deep processing, and feeling with shallow processing. We have now replicated this effect in three subsequent experiments.
We also discovered that knowing and feeling were dissociable in terms of subjects' confidence in their memories. In another experiment (Kim & Kihlstrom, 1997, Experiment 1), subjects made their recognition judgments on a four-point scale of confidence. Recognition-by-knowing was associated with high confidence levels (and few false alarms), while recognition-by-feeling was associated with low confidence levels and increased false alarms.
In yet another experiment (Kihlstrom et al., 1996, Experiment 2), instructing subjects to adopt a liberal criterion for recognition had no effect on recognition-by-knowing, but increased recognition-by-feeling.
Remembering, knowing, and feeling are also differentially effected by exposure during study. In a fourth study (Kim & Kihlstrom, 1997, Experiment 2), subjects completed two study-test cycles separated by one day, with yes/no recognition judgments, and ratings of recollective experience, after each study trial. On the second trial, recognition-by-knowing increased while recognition-by-feeling decreased.
In all of these experiments, there was a relationship between recollective experience and recognition latency. In general, remembering and knowing were associated with relatively rapid responses on the recognition task, while recognition-by-feeling took significantly longer.
To summarize what we have found so far: (1) Level of processing dissociates recognition-by-feeling can be dissociated from recognition-by-remembering and recognition-by-knowing. (2) Repeated exposure increases remembering and knowing while it decreases feeling. (3) Lowering the criterion for recognition increases feeling while having no effect on remembering or knowing. Compared to remembering and knowing, feeling is associated with (4) lower levels of confidence and (5) longer response latencies.
Our first new experiment began with this last finding, and employed a "response window" technique to control the amount of time that subjects had to make their recognition judgments. The subjects first studied a list of 80 words presented by computer under a levels-of-processing manipulation, generating rhymes for half the words and associates for the remainder. After a retention interval of 24 hours, the subjects returned to the laboratory for the recognition task. Each target or lure first appeared on the computer screen for 1.5, 3.0, 4.5, or 6.0 seconds; then, while the word remained fixed in place, the recognition query ("Remember, Know, Feel, or No?") appeared on the screen. The subjects were given an additional 1 second to make their response: pressing N if the item was new; or, if the item was old, pressing R, K, or F according to his or her recollective experience. The subjects were instructed to make their recognition responses only within this 1-second window; responses made before or after this interval were discarded. The subjects received a practice trial with 24 words before the 160-word test trial, so they clearly knew what they were in for, and they were pretty good at doing what we wanted.
The results were interesting. First, we obtained the usual levels of processing effect on recollective experience. Averaging across all four response windows, there were more experiences of remembering for items encoded semantically, but a difference favoring the phonemic encoding for experiences of feeling. The interaction is significant overall, but of course the categories of recollective experience are not independent of each other: if an item is remembered, it cannot be known or felt. Accordingly, we repeated the ANOVA, comparing remembering with feeling, and knowing with feeling: the interaction was significant in both cases. There was no levels effect on knowing, but then again relatively few items were remembered by knowing in this experiment, so there may have been some kind of floor effect in operation.
There was also a main effect of response window, subjects recognizing more items when they had more time to make their decisions, and an interaction of response window with level of processing: the windows effect was greater for items which had been encoded semantically (or, alternatively, the levels effect was greater with longer response windows.
Although the two-way interaction between response window and recollective experience was not significant, the three-way interaction, adding levels of processing, was. Again, the levels effect was greatest with the longest response window, but the direction of the effect was reversed between the remember and feel categories of recollective experience. Thus, while there wasn't much activity in this experiment involving the "Know" category, the double dissociation between remembering and feeling was especially prominent at the longest response window.
In all these studies, we have generally found similar effects on remembering and knowing, which differ from the effects on feeling. Put another way: so far, we have been able to dissociate feeling from remembering, and feeling from knowing, but we have not been able to dissociate knowing from remembering. Both remembering and knowing are enhanced by deep processing and repeated study; both are associated with high confidence levels and fast response latencies. This is fine as far as "feeling" is concerned, but it leaves the qualitative distinction between remembering and knowing somewhat in doubt.
On reflection, we think that recognition-by-knowing is most likely to occur under one of two circumstances: over long retention intervals, or with repetition of closely similar events. The former might promote degradation of the spatial and temporal cues which mark the episode as a unique experience; the latter might blur them beyond recognition. Lacking the patience required to lengthen the retention interval, we decided to pile on study trials instead. Of course, our earlier experiment, in which there were two study-test trials separated by 24 hours, didn't dissociate remembering from knowing, but each of those trials represented a distinctly different learning episode. So in our second experiment, we gave our subjects, who were paid for their aggravation, 15 study-test trials with a 20-item list over a period of about 90 minutes. Items were read out loud at a rate of about one every 3 seconds. After a 10-minute interval in which the subjects generated word associations and listened to the instructions for rating their recollective experience, they received their first test trial, making recognition judgments about the 20 targets and 20 lures presented visually. Successive study-test cycles consumed about 3-1/2 minutes each. On each study trial, the targets were read aloud in a new random order, and on each test trial the same set of targets and lures was presented in a new random order.
As we intended, the subjects learned the list fairly quickly, with very few misses, and very few false alarms, even on the first trial. So they moved pretty quickly into overlearning mode, at which point we expected to see them remember less and know more.
We were not disappointed. On the first two trials, the subjects showed the familiar mix of remembering, knowing, and feeling.
Very quickly, however, feeling dropped out of the picture, and experiences of knowing replaced experiences of remembering. In fact, by the last trial, the subjects were hardly paying attention to the study phase. They didn't have to: they knew what was on the list. Still, inspection of individual records showed that occasional experiences of remembering crept in -- as for example when noise from street construction outside the room rendered a particular presentation of a particular word particularly distinctive. Then, a subject was likely to recognize by remembering rather than by knowing.
In a sense, we suspect that there is something like a developmental trajectory in episodic memory. In the canonical case, episodic memory takes the form of conscious recollection of the past, complete with awareness of the event, its spatiotemporal context, and the role of the self in the event. Sometimes, however, perhaps by virtue of weak processing at the time of encoding, degradation over the retention interval, or impoverished cues at the time of retrieval, episodic memory takes the form of an intuitive feeling. In this case, the description of the event "rings a bell", but the person is unable to locate the event in space or time, or to place him- or herself in the event. It's this kind of intuitive feeling of familiarity which, we think, underlies successful recognition performance by amnesics: they strategically capitalize on the experience of fluency which accompanies priming, in order to make informed guesses about the past (Kihlstrom, 1998). With lots of rehearsal, especially elaborative rehearsal, the intuitive feeling can be transformed in to full-fledged remembering. And with even more rehearsal, the memory will lose its episodic character, and take on the properties of a semantic memory -- as a fact which one knows about the past, rather than an experience which one remembers.
There is some tendency right now to reduce the remember/know distinction to something like a criterion effect. While the different recollective experiences are definitely associated with different levels of confidence and accuracy, things are not quite so simple as they first seemed. In the first place, there aren't just two varieties of recollective experience: rather, there are at least three, and that makes the mathematics a little more complicated. But even these categories aren't just proxies for criterion or confidence. Rather, they seem to reflect the differential accessibility of different types of trace information. In remembering, the person has access to "The Full Monty": to a description of the event, of the time and place in which it occurred, and his or her own relation to the event. In knowing, there is none of this episodic information, just the more or less abstract knowledge that something happened. In feeling, there is retrieval of at least a partial description of the event, but not of its episodic context nor of any information about the self. The person's recollective experience varies, not just with the criterion he or she sets along some single dimension of trace strength, but rather with the accessible trance information on which his or her memory judgment is made.
Paper presented at the annual meeting of the Psychonomic Society, Dallas, November 1998. This research was supported by Grant #MH-35856 from the National Institute of Health. We thank Jason Randolph for his assistance in conducting Experiment 2.
Gardiner, J.M. (1988). Functional aspects of recollective experience. Memory & Cognition, 16, 309-313.
Kihlstrom, J.F. (1998, August). Interactions between implicit and explicit memory. Paper presented at the annual meeting of the American Psychological Association, San Francisco. http://socrates/berkeley.edu/~kihlstrm/apa98b.html
Kihlstrom, J.F., Kim, M., & Dabady, M. (1996). Remembering, knowing, and feeling in episodic recognition. Paper presented at the annual meeting of the Psychonomic Society, Chicago, November 1996. http://socrates/berkeley.edu/~kihlstrm/psychn96.html
Kim, M., & Kihlstrom, J.F. (1997). Exposure, confidence, and recollective experience in episodic recognition. Paper presented at the annual meeting of the Psychonomic Society, Philadelphia, November 1997. http://socrates/berkeley.edu/~kihlstrm/psychn97.html
Rajaram, S. (1993). Remembering and knowing: Two means of access to the personal past. Memory & Cognition, 21, 89-102.
Tulving, E. (1985). Memory and consciousness. Canadian Psychology, 26, 1-12.
This page last revised 02/11/06 10:05:06 AM .