Remember/Know Judgments in Recognition Tests¶
An instructed report distinguishing recognition with recollected study detail from an oldness judgment without that detail.
Core Idea¶
A remember/know judgment in a recognition test asks a participant who treats a test item as previously encountered to say how that oldness is experienced. Remember reports conscious recovery of a detail of the item's particular study encounter. Know reports oldness without that recovered contextual detail. The experiment records this subjective distinction, usually alongside which items were actually studied and which are new lures. Gardiner's 1988 word-recognition experiment and Reder and colleagues' 2013 face-recognition experiment both use it, despite different stimuli and manipulations.[1][2]
The selected source title is broad, but this entry is explicitly the recognition-test subtype. Gardiner notes that Tulving's earlier R/K terminology also applied to recalled items. The two complete original cases inspected here support item-cued recognition; they do not authorize a claim that all remembering/knowing reports start with an old/new test item.[1]
The labels are reports about experience, not an automatic accuracy score or a direct reading of separate neural systems. A new lure can receive an erroneous old response and then an R or K label. An experiment may compare proportions by condition, but that analysis and any model-based estimate are downstream of the participant's judgment.[1][2][3]
Structural Signature¶
- Recognized test item: a presented item elicits an affirmative oldness response. It can be a studied target or an incorrectly endorsed lure; truth of prior exposure is separate from the participant's endorsement.[1][2]
- Recollective-detail criterion: under the instructions, Remember means conscious recovery of some aspect of encountering that item in the study episode. It is an instructed response meaning, not independent verification that a particular memory process occurred.[1][2]
- Nonrecollective-oldness criterion: Know means the item is judged old without recovery of such study detail. This contrasts with R, but does not itself establish a pure familiarity mechanism.[1][2][3]
- Recorded subjective branch: the eligible old response is recorded as R or K so the reports can be counted or examined. Gardiner had participants circle recognized words and mark R or K; Reder's face test offered R, K, or New as response options in one test. A literal two-screen sequence is therefore not necessary.[1][2]
The experiment may also record actual target/lure status and manipulate encoding, study context, or stimulus class. Those fields are essential to interpreting results, but no one manipulation or minimum accuracy is necessary for the judgment method. A conditional familiarity calculation such as K/(1−R) is an optional analysis under an independence model, not a fifth participant-response role. It requires R < 1 and is undefined when R = 1.[1][2][4]
What It Is Not¶
The method is not ordinary old/new recognition alone: an old response without the instructed and captured R/K distinction does not disclose the reported basis of that response. It is not a confidence rating: a person can be confident or uncertain about either kind of report, while the R instruction concerns recovery of study detail. Reder and colleagues explicitly told participants not to treat the categories as merely confidence levels.[1][2]
Nor is the binary report a proof of a dual-process memory architecture. Gardiner interpreted his pattern as consistent with such theories, but Donaldson offered a second-decision-criterion account of R/K patterns. Brown and Bodner's later comparison found that binary judgments and independent ratings can lead to different apparent dissociations. These sources make response format and model assumptions part of interpretation rather than parts of the response's identity.[1][3][4]
Scope of Application¶
The literal setting is an experimental recognition test in human memory research. Studied and lure words can be tested after different encoding tasks; previously viewed and new faces can be tested after a contextual study episode. In both, the oldness judgment can be accompanied by remembered encounter detail or asserted without it.[1][2]
The protocol can be implemented with a sequential old response then R/K mark, or with R/K/New response options presented together. The structural requirement is an affirmative recognition judgment subdivided by instructed experiential basis. A free-recall task labeled with R/K terms may be related but lies outside this entry's intentionally narrow evidence basis.[1][2]
Clarity¶
The method resolves an ambiguity inside an old response: Does the participant report a detail of the prior encounter, or only that the item seems previously seen? The report can be sorted without making the stronger claim that one answer is more accurate. Gardiner separately tabulated R and K for correctly recognized words and false-positive endorsements; the latter demonstrate why subjective category and actual study status must remain distinct.[1]
It also clarifies what an experiment has and has not measured. R/K is a response category grounded in instructions and introspection. The observer still needs target/lure status to assess accuracy, a separate confidence measure to study certainty, and a theoretical model before interpreting rates as contributions of latent memory processes.[1][2][3]
Manages Complexity¶
Many test trials differ by item, exposure status, study condition, participant, and response. The R/K branch compresses one aspect of those trials—the reported basis for an affirmative old judgment—into comparable categories. Gardiner compared semantic-associate and rhyme-encoding conditions for words. Reder's face experiments varied prior familiarity and study background while separately reporting hits and false alarms.[1][2]
The compression is useful only if the other fields are retained. Pooling targets with lures would mistake errors for studied-item memory; pooling R and K would erase the reported experiential distinction; equating R with accuracy would cross two different dimensions. A binary branch can also hide simultaneous or graded recollection and familiarity, a limitation highlighted by comparisons with independent ratings.[1][2][4]
Abstract Reasoning¶
First map each trial through three questions: Was the test item actually studied? Did the participant endorse it as old? If so, did the participant report contextual study detail or oldness without it? That map separates a hit, false alarm, miss, or correct rejection from the R/K label and prevents a report from being treated as ground truth.[1][2]
Then compare R and K distributions across a manipulation only after preserving the target/lure denominator and response instructions. Gardiner's semantic-versus-rhyme experiment found a larger effect on R than K in that design. Reder's famous-face background manipulations produced their own context-specific patterns. Such dissociations suggest questions about memory processes; they do not, by themselves, settle whether distinct processes, different decision criteria, or response-format effects generated them.[1][2][3][4]
Knowledge Transfer¶
The same report rule transfers within recognition research from words to faces: ask whether an affirmative oldness judgment carries recollection of the specific study occurrence, then record the answer. What does not transfer automatically is the encoding manipulation or its observed result. A semantic word association and a reinstated face background are different interventions, and a finding in one is not a guaranteed result in the other.[1][2]
The more general idea of classifying a judgment by its reported basis might appear in other settings, but this entry requires human participants instructed about a study episode and a recognition response. Calling a machine output “remember” or “know” would import terminology without evidence of the stipulated subjective experience. A broader classification/report pattern is a future-Prime question, not established by these two psychology cases.
Examples¶
Word-recognition experiment¶
In Gardiner's Experiment 1, participants studied words while either generating a semantic associate or a rhyme. After a delay they saw a test list mixing studied words with lures, circled those they recognized, and marked each circled word R if they consciously recollected part of its study occurrence or K if they recognized it on another basis. The paper reports R and K for correct old judgments and false positives. Semantic encoding increased R responses more strongly than K responses in this experiment.[1]
Mapped back: a circled word fills the affirmative recognition role; the R/K instruction distinguishes recollected study detail from nonrecollective oldness; the written label captures the subjective branch. Studied/lure status and encoding task remain separate analysis fields.
Face-recognition experiment¶
Reder and colleagues first had participants view famous and unknown faces against location backgrounds. At test, faces appeared on reinstated or changed backgrounds alongside new-face foils, and participants made R/K/New judgments about the face. R meant recalling a detail of seeing it earlier; K meant it seemed old without such contextual recollection. The paper analyzed responses by face familiarity and background condition and kept hits separate from false alarms. The authors' account of why famous faces showed stronger context effects remains an interpretation of this design.[2]
Mapped back: an R or K response endorses the tested face as old; the instructed distinction supplies the two experiential criteria; the response option records the branch. Actual prior exposure, fame, and background matching are contextual fields, not the definition of the judgment.
Structural Tensions¶
The classic binary branch offers a compact, mutually exclusive R/K count, but forcing one label can obscure occasions when recollection and familiarity co-occur or vary by degree. Independent ratings can preserve both dimensions but no longer produce exactly the same binary variable. A study favoring binary labels gains comparability with Gardiner-style reports while risking a response-format artifact in process interpretation; favoring independent scales gains graded information but changes the method and its denominator. The diagnostic question is: Is the study asking for a categorical description of each endorsed recognition response, or an estimate that remains interpretable when the two experiences overlap? Brown and Bodner's comparison makes this a bounded method-design tension, not a universal proof about memory systems.[4]
Structural–Framed Character¶
On the structural–framed spectrum, this is a practice-framed experimental method with a compact structural response rule. The structural skeleton is an affirmative judgment classified by its reported basis. Evaluative weight is low in the labels themselves: a Remember response is not automatically correct or better. Human-practice dependence is high because researchers set the instructions and participants introspect and report. Institutional origin lies in laboratory memory research, not in an automatically self-labeling biological signal. Gardiner's original word protocol and Reder's face protocol show the rule's literal reach within that practice.[1][2]
The vocabulary travels between word and face tests, but its meaning depends on a prior study episode and an instruction about recovering details. Import versus recognition is the check when moving settings: recognize the instructed branch in another human recognition protocol, but do not import a neural conclusion, accuracy status, or subjective experience into a system that was never asked or able to report one. Its character: a lab-elicited phenomenological classification with a portable response structure and a psychology-bound interpretation.
Structural Core vs. Domain Accent¶
The core is the instructed contrast between an affirmative judgment accompanied by recollected event detail and one reported without it. The recorded report is a Classification output under a reusable R/K assignment rule. In this entry the domain-bound mechanism is a human recognition test with an earlier study encounter, later item cue, and introspective R/K report. Words and faces preserve those roles while changing stimuli and manipulations.[1][2]
This named method does not clear the Prime bar: the two full positives are both human recognition experiments, and Gardiner explicitly records R/K terminology in recall as well. A generic “classify a decision by its reported basis” skeleton is a future-Prime question requiring independent cross-domain evidence; it is not licensed as a universal parent merely by the word/face transfer. The current qualifier “in Recognition Tests” keeps the canonical identity within the evidence actually inspected.
Instantiates / Related Primes¶
This entry is a kind of Classification.
The procedure is used inside Recognition Memory research, but recognition capacity and an R/K report are different identities: an old/new task can occur without the report, and even a false old response can receive an R/K label. Recall Memory is a neighboring setting for broader R/K terminology, not evidence that the present scoped entry covers recall. Dual Process Theory and Signal Detection Theory are competing or complementary interpretations, not constituent participant instructions.[1][2][3]
Prime Classification is a strict parent of this scoped procedure. The classified entities are affirmative oldness reports, the criteria are the instructed R/K distinction, and the output is one recorded category for each eligible response.[1][2] A participant's private experience can be hard for another observer to verify; the assignment rule and report remain explicit and repeatable across trials. Metacognition is only a neighbor here: one R/K response need not regulate strategy or run a feedback loop. Classification has no strict ancestors to import.
Relationships to Other Abstractions¶
Current abstraction Remember/Know Judgments in Recognition Tests Domain-specific
Parents (1) — more general patterns this builds on
-
Remember/Know Judgments in Recognition Tests is a kind of Classification Prime
Every Remember/Know judgment in a recognition test classifies an affirmative oldness report under instructed R/K criteria.The entities classified are participants’ affirmative oldness reports, including mistaken endorsements of lures, not verified memory states. Experimenter instructions supply the R criterion of recovered study-episode detail and the K criterion of reported oldness without it; participants assign each eligible response to a discrete category. The rule is reusable across items and conditions, and investigators count the reports. Classification does not require a second observer to verify private experience or identify latent processes. The child adds the recognition-test setting and phenomenological distinction; Classification can occur without either.
Hierarchy path (1) — routes to 1 parentless root
- Remember/Know Judgments in Recognition Tests → Classification
Neighborhood in Abstraction Space¶
Remember/Know Judgments in Recognition Tests sits in a sparse region of the domain-specific corpus (94th percentile for distinctiveness): few abstractions share its structure, so a faithful description tends to retrieve it precisely.
Family — Memory Encoding & Retrieval Effects (21 abstractions)
Nearest neighbors
- False Fame Effect — 0.78
- Recognition Memory — 0.78
- Computerized adaptive testing — 0.78
- Temporal Distinctiveness — 0.77
- Cue-dependent forgetting — 0.77
Computed from structural-signature embeddings · 2026-10-17
Not to Be Confused With¶
- Old/new recognition accuracy: actual study status must be checked separately from the person's R/K report.[1][2]
- High/low confidence: the instructed distinction concerns recollected encounter detail, even if confidence and R/K rates may correlate.[1][2]
- Pure recollection and familiarity probabilities: these require a model and assumptions; raw R/K counts are reports.[3][4]
- Independent recollection and familiarity ratings: a related modified method that can permit both ratings, unlike the classic exclusive branch.[4]
- R/K labels in free recall: historically attested in Gardiner's account of Tulving, but outside this entry's source-complete recognition subtype.[1]
References¶
[1] Gardiner, John M. (1988). Functional aspects of recollective experience. Memory & Cognition 16(4), 309–313. DOI: 10.3758/BF03197041. Full original publisher PDF inspected, especially printed pp. 310–312. registry ↩a ↩b ↩c ↩d ↩e ↩f ↩g ↩h ↩i ↩j ↩k ↩l ↩m ↩n ↩o ↩p ↩q ↩r ↩s ↩t ↩u ↩v ↩w ↩x ↩y ↩z ↩27
[2] Reder, Lynne M., et al. (2013). Why It’s Easier to Remember Seeing a Face We Already Know Than One We Don’t: Preexisting Memory Representations Facilitate Memory Formation. Psychological Science 24(3), 363–372. DOI: 10.1177/0956797612457396. Full publisher-version PDF hosted by authors' Carnegie Mellon laboratory inspected, especially printed pp. 365–367 and note 2 at p. 371. registry ↩a ↩b ↩c ↩d ↩e ↩f ↩g ↩h ↩i ↩j ↩k ↩l ↩m ↩n ↩o ↩p ↩q ↩r ↩s ↩t ↩u ↩v ↩w
[3] Donaldson, Wayne (1996). The role of decision processes in remembering and knowing. Memory & Cognition 24, 523–533. DOI: 10.3758/BF03200940. Original publisher abstract inspected; used only for the alternative decision-criterion interpretation. registry ↩a ↩b ↩c ↩d ↩e ↩f ↩g
[4] Brown, Aaron A., and Glen E. Bodner (2011). Re-examining dissociations between remembering and knowing: Binary judgments vs. independent ratings. Journal of Memory and Language 65(2), 98–108. DOI: 10.1016/j.jml.2011.04.003. Publisher abstract and accessible introduction/discussion excerpts inspected; model and response-format claims are limited to those passages. registry ↩a ↩b ↩c ↩d ↩e ↩f ↩g