Home Curriculum Vitae Publications Conference Reports Forthcoming Extramural Colloquia Expert Testimony Teaching Healthcare The Human Ecology of Memory Research Archive Publications and Reports Rants

 

Exposure, Confidence, 

and Recollective Experience 

in Episodic Recognition

 

Michael Kim

Yale University

 

John F. Kihlstrom

University of California, Berkeley

 

 

In 1985, Tulving distinguished between two different recollective experiences: Remembering, or one's concrete awareness of oneself in the past ("autonoetic consciousness); and Knowing, or one's abstract knowledge of the past (noetic consciousness). The Remember/Know distinction was further developed by Gardiner (1988), who related Tulving's distinction between "remembering" and "knowing" to Mandler's (1980) distinction between "recognition by retrieval" and priming-based "recognition by familiarity".

On the other hand, this mapping -- remembering onto retrieval, and knowing onto familiarity -- appears to conflate semantic memory with implicit memory. Instead, we think it is useful to distinguish among at least three varieties of recollective experience:

Remembering Conscious recollection of a past event, as an explicit expression of episodic memory;

Knowing Abstract knowledge of a past event, as an item in semantic memory; and

Feeling The intuition that an event occurred in the past, as an implicit expression of episodic memory.

Last year at this meeting, we described two experiments which lent support to the distinction between Knowing and Feeling (Kihlstrom, Kim & Dabady, 1996). These studies found that roughly half of recognized items classified as "Known" according to the dichotomous classification of Tulving and Gardiner should instead be classified as "Felt" or "Intuited". More important, Experiment 1 found a double dissociation (crossover interaction) between level of processing and recollective experience. There were more "Know" (and "Remember") items with deep than with shallow processing, but more Feel items with shallow than with deep processing. Experiment 2 replicated this effect. More important, it varied subjects' criterion for recognition. Shifting from a strict to a liberal criterion had no effect on "Know" (or "Remember") ratings, but did increase "Feel" ratings. Unlike "Remember" and "Know" ratings, "Feel" ratings were associated with an increased rate of false alarms.

In the present paper, we report two further studies of the distinction between remembering, knowing, and feeling in episodic recognition. Each of these experiments employed the basic Tulving/Gardiner paradigm, in which subjects studied a list of words, and 24 hours later, completed a recognition test. For each item recognized, subjects then reported whether their judgment was based on an experience of Remembering, Knowing, or Feeling (for detailed instructions, see Kihlstrom et al., 1996, attached).

Our first experiment examined the relations between recollective experience and both confidence level and response latency. The 33 subjects studied a list of 80 words (taken from Rajaram, 1993) under conditions of a levels of processing manipulation: for 40 items they supplied rhymes, and for 40 items they supplied semantic associates. After a 24-hour retention interval they made recognition judgments on a 4-point scale of confidence:

0 = Certain that the word is new;

1 = Think that the word is new, but not certain;

2 = Think that the word is old, but not certain;

3 = Certain that the word is old.

For each item receiving a recognition rating greater than zero, the subjects then rated their recollective experience in terms of "Remembering", "Knowing", or "Feeling". The detailed instructions for this rating were summarized as follows:

[C]onsider 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.

Items were presented for study by a Macintosh computer controlled by the ECL software. Items were presented for test by a Windows computer controlled by SuperLab software, which also collected response latencies for the recognition judgments. The subjects were paid for their participation in the experiment.

The confidence-rating procedure yielded three criteria for item recognition:

strict counting only those items receiving a rating of "3";

moderate counting as well those items receiving a rating of "2"; and

liberal counting even those items receiving a rating of "1".

Like those presented last year, the following analyses parallel those of Tulving (1985), Gardiner (1988), and Rajaram (1993), focusing on recognition hits and ignoring false alarms (but see Donaldson, 1996). Unlike last year, however, we do not go through the preliminary exercise of analyzing the results as previous studies have done, distinguishing between Remembering and only a single alternative. Readers who are interested in the results of such an analysis should bear in mind that, in the traditional Remember/Know paradigm, as applied by Tulving, Gardiner, and others, "Knowing" is a fallback category which includes all items which a subject recognizes without an accompanying experience of conscious recollection. Therefore, the Tulving-Gardiner analyses may be replicated by combining our categories of "Knowing" and "Feeling".

 

Proportion of Items Recognized in Experiment 1

Category

Rhyme

Semantic

Lure

Strict Criterion

Remember

.24

.42

.04

Know

.11

.15

.03

Feel

.02

.02

.01

Moderate Criterion

Remember

.33

.48

.08

Know

.19

.21

.07

Feel

.16

.11

.13

Liberal Criterion

Remember

.35

.49

.11

Know

.20

.22

.11

Feel

.33

.21

.39

 

A three-way within-subjects ANOVA of recognition hits with encoding condition, recognition criterion, and recollective experience as factors found that all main effects and interactions were significant (p < .001). Of particular interest was the 3-way interaction between encoding condition, recognition criterion, and recollective experience. Loosening the criterion for recognition had relatively little effect on "Remember" ratings. However, it produced modest increases in "Know" ratings, and big increases in "Feel" ratings. This was particularly the case when the shift was from the moderate to the liberal criterion.

Separate ANOVAs within each criterion level confirmed significant main effects of encoding condition and rating, and a significant encoding x rating interaction, in every case (all p < .05). Semantic processing increased "Remember" and "Know" ratings, but decreased "Feel" ratings.

The significant encoding x rating interaction was confirmed when the "Remember" items were eliminated from consideration, and the ANOVA was confined to "Feel" and "Know" items.

Thus, this experiment confirmed the central finding of the two experiments reported in 1996: semantic processing increases the frequency of "Remembering" and "Knowing", while phonemic processing increases the frequency of "Feeling".

 

Confidence Ratings in Experiment 1

Items Recognized According to the Liberal Criterion

Category

Rhyme

Semantic

Lure

Remember

2.63

2.82

2.28

Know

2.47

2.56

2.07

Feel

1.55

1.58

1.34

 

 

A two-way within-subjects ANOVA of the 1-3 confidence ratings themselves (i.e., with level of processing and recollective experience serving as factors) yielded significant main effects of both factors, as well as a significant interaction between these variables (all p < .05). Recognition by "Feeling" was associated with relatively low levels of confidence, while recognition by both "Remembering" and "Knowing" were associated with relatively high levels of confidence. This difference was especially pronounced for items studied in the semantic encoding condition.

 

Response Latencies in Experiment 1

(Seconds)

Category

Rhyme

Semantic

Lure

Remember

3.53

3.16

3.95

Know

3.80

3.47

4.17

Feel

4.59

4.50

4.83

Reject

3.15

3.18

3.41

 

ANOVA of the response latencies also yielded significant main effects of both encoding condition and rating (both p < .05); however, the interaction between these variables was not significant.

Regardless of encoding condition, recognition by "Feeling" was associated with relatively long response latencies, while recognition by both "Remembering" and "Knowing" were associated with relatively short ones.

Our second experiment was concerned with the effects of repeated exposure on recollective experience. Until now, all Remember/Know experiments have employed only a single study-test cycle. This experiment employed two such cycles, and attempted to increase the distinctiveness of the items on the study list.

A total of 33 subjects completed two study-test cycles with a list of 80 words, separated by a 24-hour interval. On Day 1, each of the 80 targets were presented in a unique combination of font (Times, Palatino, etc.), size (16 point, 24 point, etc.), style (bold, italics, etc.), and location (centered, left-justified, etc.). On Day 2, 24 hours later, the same targets were presented for study in a different, unique combination of these four physical attributes. There was no levels-of-processing manipulation in this experiment. Instead, subjects were told simply to read each word carefully as it was presented.

The subjects made Yes/No recognition judgments after each study trial, following a 10-minute retention interval (filled by personality questionnaires), and rated their recollective experience for each recognized item. The study trials were controlled by a Macintosh running ECL, and the test trials were controlled by a Windows machine running SuperLab.

 

Proportion of Items Recognized in Experiment 2

Category

Target

Lure

Trial 1

Remember

.32

.03

Know

.18

.07

Feel

.20

.19

Trial 2

Remember

.43

.05

Know

.27

.11

Feel

.15

.23

ANOVA of recognition hits yielded significant main effects of trials and ratings, both p < .001. Recognition increased with trials, and more items received ratings of "Remember" than of "Know" or "Feel". Most important was a significant trials x ratings interaction, p < .001. Post-hoc tests confirmed that "Remember" and "Know" ratings increased across trials, while "Feel" ratings decreased.

The findings of these experiments support the notion that "Knowing" and "Feeling" are different memory qualia, in that they may be dissociated by a number of factors. (1) "Knowing" and "Feeling" are affected differently by level of processing at the time of encoding (the present Experiment 1, and the two experiments reported last year). (2) "Knowing" and "Feeling" are affected differently by the number of study trials devoted to the item (the present Experiment 2). (3) "Knowing" and "Feeling" are affected differently by shifts in recognition criterion (Experiment 2 reported last year). (4) "Knowing" and "Feeling" are associated with different levels of recognition confidence (the present Experiment 1). (5) "Knowing" and "Feeling" are associated with different response latencies (Experiment 1, confirmed by analyses of the other experiments in this series).

As an aside, we should note that it may be necessary to distinguish yet a fourth variety of recollective experience:

Believing The inference that an event occurred in the past, based on other episodic and semantic knowledge.

In fact, Shobe and Kihlstrom (1997) reported two experiments on false memory using the Deese (1959) paradigm revived by Roediger and McDermott (1995). Roediger & McDermott reported that falsely recognized items were commonly accompanied by a recollective experience of remembering, but they followed Gardiner's version of the Remember-Know procedure, and did not distinguish between Knowing and Feeling. Shobe and Kihlstrom asked subjects to report their recollective experience in terms of "Remembering", "Knowing", "Feeling", or "Believing". Whereas true recognition of old items was largely accompanied by "Remembering", false recognition of critical lures was often accompanied by experiences of "Knowing", "Feeling", or "Believing". In fact, experiences of "Believing" were more frequent for critical lures than for studied items.

"Knowing" and "Feeling" are alike in that neither is accompanied by the conscious recollection of a prior episode -- the phenomenal experience which is the hallmark of "Remembering" ("Believing" also lacks this hallmark, and in fact lacks any recollective experience at all). In other, respects, the present research, along with the studies presented at last year's meeting, indicate that "Knowing" more closely resembles "Remembering", and is different from "Feeling". Recognition-by-Feeling judgments take longer, and are associated with lower levels of confidence (and a higher rate of false alarms).

"Remembering" reflects explicit retrieval of an episodic memory, while "Knowing" reflects retrieval of context-free semantic (or generic) knowledge. In particular, recognition judgments mediated by "Feeling" appear to have qualities of intuition, inference, and "informed guesswork", compared to recognition judgments mediated by either "Remembering" and "Knowing". "Feeling" seems to an effect of perceptual salience, related to the priming effects of past experience. This is why "Feeling" judgments are more common for items subject to perceptual activity at the time of encoding. But while priming is automatic, the use of perceptual salience in making recognition judgments is deliberate and strategic. This is why response latencies to make "Feeling" judgments are long, and confidence levels in such judgments are low.

Both "Knowing" and "Feeling" should be distinguished from "Remembering" in future studies of recollective experiences and their underlying processes.

 

Author Notes

Paper presented at the annual meeting of the Psychonomic Society, Philadelphia, November 1997. This research was supported by Grant #MH-35856 from the National Institute of Mental Health. We thank Correspondence to John F. Kihlstrom, Department of Psychology MC-1650, University of California, Berkeley, Tolman Hall 3210, Berkeley, California 94720-1650. Email: kihlstrm@cogsci.berkeley.edu. URL: http://socrates.berkeley.edu/~kihlstrm.

 

References

Deese, J. (1959). On the prediction of occurrence of particular verbal intrusions in immediate recall. Journal of Experimental Psychology, 58, 17-22.

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

Kihlstrom, J.F., Kim, M., & Dabady, M. (1996, November). Remembering, knowing, and feeling in episodic recognition. Paper presented at the annual meeting of the Psychonomic Society, Chicago.

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

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

Roediger, H., & McDermott, K. (1995). Creating false memories: Remembering words not presented in lists. Journal of Experimental Psychology: Learning, Memory, & Cognition, 21, 803-814.

Shobe, K.K., & Kihlstrom, J.F. (1997, August). Recall, recognition, and recollective experience in false memory. Paper presented at the annual meeting of the American Psychological Association, Chicago.

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

 

This page last revised 04/08/10 02:58:49 PM.