The Levels-of-Processing Effect

Snap Shot

• The question: Does the way a witness mentally processed an event, and not only how long the witness saw it, affect what the witness can later remember?
• What the research shows: Information processed for meaning is generally remembered better than information processed only for its surface features, whether or not the person was trying to remember. What counts as effective processing depends on the later test: memory is best when the kind of processing at retrieval matches the kind at encoding, and judging a face’s character generally aids later recognition more than judging a single feature.
• What it does not show: “Depth” has no independent measure, so the framework describes a pattern rather than measuring a quantity. Deeper processing also increases meaning-based false memories, and the research cannot reconstruct how a particular witness processed a particular event.
• The law: Expert testimony on encoding is governed by Rule 702 (amended 2023), Kumho, and Houser in courts-martial, and by Ariz. R. Evid. 702, Chapple, and Nordstrom in Arizona; the expert may explain how encoding affects memory but may not opine on a witness’s credibility (Lindsey).
• Issues for each side: Prosecutors can explain why a witness who engaged with what was happening may remember its meaning and central features well without having tried to memorize anything; defense counsel can examine what the witness was attending to and doing, and whether peripheral details were processed at all, without treating either “deep” processing or a vivid account as proof of accuracy.

The levels-of-processing effect is the finding that what a person does with information at the moment of encoding strongly influences how well it will be remembered. Judging what a word means, how it relates to other things, or whether it fits a sentence generally produces better later memory than judging how the word looks or sounds, even when exposure time is the same and even when the person has no idea that memory will be tested. The idea was introduced by Craik and Lockhart (1972) as a framework for memory research and was tested in a well-known series of experiments by Craik and Tulving (1975).

The effect reaches court because most witnesses encode events incidentally. A bystander at a convenience store, a passenger in a car, or a person present at a party does not know that he or she will later be asked to describe what happened. What the witness remembers depends heavily on what the witness was attending to and how the witness was thinking about it: whether a face was simply seen or was evaluated, whether a conversation was understood or only heard, and whether a detail was part of what mattered to the witness at the time. The research also bears on interviewing, because retrieval works best when it reengages the kind of processing that occurred at encoding.

This paper covers the original framework and experiments, transfer-appropriate processing, the principal critiques, related self-reference and generation effects, application to witnesses and interviews, and the limits of the research and the governing law.

The Framework: Craik and Lockhart (1972)

In the early 1970s, memory was commonly described as a series of stores, from brief sensory registers to a limited short-term store and, with rehearsal, a long-term store. Craik and Lockhart (1972) proposed a different emphasis. They suggested that incoming information is analyzed in a sequence running from physical and sensory features, through sound, to meaning and associations, and that the durability of a memory trace is a byproduct of how far that analysis proceeds. On this view, a person does not decide to store something in long-term memory; lasting memory results from the processing the person carried out in order to perceive and understand the material. They also distinguished between simply keeping an item in mind by repeating it at the same level, which they argued does little for long-term retention, and processing it more fully. They presented the idea as a framework to guide research rather than a complete theory.

The Classic Experiments

Craik and Tulving (1975) tested the framework in ten experiments. Participants saw a series of words, each preceded by a question that directed a particular kind of processing: whether the word was printed in capital letters (a judgment about appearance), whether it rhymed with another word (a judgment about sound), whether it belonged to a category, or whether it would fit sensibly into a given sentence (judgments about meaning). In most of the experiments, the memory test that followed was unexpected. Recognition and recall improved steadily from appearance questions to sound questions to meaning questions, and the differences were large. Words that received a “yes” answer to a meaning question were generally remembered better than words that received a “no,” which the authors attributed to the richer, more integrated encoding produced when a word fits its context. The authors also reported that the advantage was not simply a matter of time spent: in one experiment, a difficult surface judgment that took longer still produced poorer memory than a quick judgment about meaning.

Research on incidental learning had pointed the same way. Hyde and Jenkins (1973) gave participants semantic, letter-based (graphic), or grammatical (syntactic) orienting tasks on a word list. Semantic tasks, such as rating how pleasant each word was, produced better recall than the graphic and syntactic tasks, and recall after a semantic task approached that of participants who were simply trying to learn the list. The kind of processing mattered more than the intention to remember.

Transfer-Appropriate Processing

The most important qualification came from Morris, Bransford, and Franks (1977). Participants processed words either for meaning or for rhyme. On a standard recognition test, which asked whether each word had appeared, meaning-based encoding produced better memory, reproducing the levels-of-processing effect. On a rhyming recognition test, which asked whether a test word rhymed with a studied word, the advantage reversed: rhyme-based encoding produced better performance. The authors concluded that the value of an encoding activity depends on whether it is appropriate to the demands of the later test. Meaning-based processing is usually effective because most tests, and most real-world recall, ask about meaning.

This principle, called transfer-appropriate processing, is closely related to the encoding specificity principle, which holds that a retrieval cue is effective to the extent that it reinstates information encoded with the original event (Tulving & Thomson, 1973). Recent work continues to test its boundaries. In a series of experiments, a match between the processing task at study and at test improved memory judgments for studied words but not for full statements, and did not moderate the tendency to rate repeated claims as true (Ly et al., 2023). For legal purposes, the lesson is that there is no single best way to encode: the processing that supports one kind of later question, such as describing what happened, may not support another, such as recognizing a particular object or face.

Critiques and Revisions

The framework drew sustained criticism. The central objection, pressed by Baddeley (1978), was circularity: without an independent measure of depth, deeper processing is identified by better memory, and better memory is then explained by deeper processing. Critics also noted that processing need not run in a fixed order from surface to meaning. Surface processing is not always inferior. For pictures of complex scenes, questions about physical appearance produced better memory than semantic questions in a series of experiments, an advantage the authors called physical superiority (Intraub & Nicklos, 1985).

Craik and Lockhart responded that the framework had been intended to direct research rather than to serve as a precise theory, and they recast depth in terms of the elaboration and distinctiveness of encoding (Lockhart & Craik, 1990). Craik (2002) later reviewed the framework’s history and argued that its central claim, that memory is a product of the processing carried out during perception and comprehension, remained useful while acknowledging the measurement problem. The effect is now generally treated as a reliable pattern under typical test conditions rather than a measurable property of memories.

A further qualification has direct forensic importance. Processing for meaning strengthens memory for meaning, including meanings that were implied but not presented. When participants studied lists of related words using a deeper task, they recalled more of the studied words but also more of the related words that had never been shown (Rhodes & Anastasi, 2000). Deeper encoding therefore does not simply make memory more accurate; it can increase both correct recall and meaning-based false recall. This connects the levels-of-processing effect to research on gist and schemas, discussed in Gist Memory and Schema and Script Memory.

Self-Reference and Generation

Two related findings are often described as forms of especially effective processing. In the self-reference effect, words judged for whether they describe oneself were remembered better than words judged for meaning in other ways (Rogers et al., 1977). A meta-analysis confirmed the effect and found it smaller when the comparison task also encouraged organized and elaborate processing, consistent with the view that self-reference works largely by producing such processing (Symons & Johnson, 1997). In the generation effect, information that people produce themselves, such as completing a word from a cue, is remembered better than information they simply read (Slamecka & Graf, 1978); a meta-analysis found the advantage reliable across many studies and conditions, with its size varying by task and test (Bertsch et al., 2007). Both are treated in their own papers, The Self-Reference Effect and The Generation Effect. For witnesses, they suggest that events a person related to his or her own concerns, or actively thought through, tend to be retained better than events merely seen.

Forensic Relevance: What Witnesses Attended To and How They Processed It

Most eyewitness encoding is incidental, and the levels-of-processing research shows that incidental encoding can produce good memory when the witness was engaged with the meaning of what was happening. A witness who was trying to understand an argument, judge whether a person was a threat, or decide what to do may retain the gist and central elements of the event well without having tried to memorize anything. Details that played no part in that understanding, such as the color of a car in the background or the exact wording of a remark that did not matter at the time, may have received little processing and may be poorly retained or absent. Such differences help explain why gist and central details are often better remembered than peripheral ones.

Face memory shows the effect directly. Participants who judged faces for traits such as likableness or honesty later recognized them better than participants who judged a physical feature such as sex (Bower & Karlin, 1974). Consistent with transfer-appropriate processing, one study found that trait judgments improved recognition from a lineup, while attention to physical features was better suited to building a composite face (Wells & Hryciw, 1984). A more recent study found that this social-encoding benefit held for other-race faces as well as own-race faces (Schwartz et al., 2023). Real witnesses are not instructed to make such judgments, but how a witness spontaneously thought about a person may bear on later recognition. Cross-race recognition and face memory are addressed in The Cross-Race Effect and Face Recognition.

Interview Implications

Because retrieval works best when it reinstates the conditions and processing of encoding, interviews that ask a witness to mentally return to the context of the event and report freely in his or her own words are consistent with this research. The cognitive interview combines these elements with open-ended questioning, and a meta-analysis found that it reliably increased correct details with a smaller increase in incorrect ones (Memon et al., 2010). The same principle suggests that demands for precise description of details the witness had no reason to process may yield guesses. These are inferences from laboratory findings, not tested rules for interviewing.

Limits of the Evidence

Most levels-of-processing studies use word lists, simple pictures, or photographs of faces, presented briefly to college students under calm conditions, with orienting tasks assigned by the experimenter and memory tested minutes to days later. Witnesses to crimes encode complex, moving, and often frightening events, choose for themselves what to attend to, and are questioned after much longer delays. The effects in the classic studies were large, but their size in real events is unknown. Several foundational studies are decades old; their main patterns have been replicated and refined rather than overturned. Without an independent measure of depth, the framework cannot quantify how deeply a witness processed anything.

Most important, the research describes average differences between groups given different encoding tasks. It cannot recover how a particular witness processed a particular event, and testimony about “deep” or “shallow” processing in a given case is necessarily an inference from the witness’s own account and circumstances. Deeper processing does not guarantee accuracy, because meaning-based processing also raises the risk of reporting implied but absent details.

Questions to Ask About This Research

When Craik and Tulving (1975) is cited to suggest that a witness who was engaged with an event should remember it well, counsel should ask who was studied and under what conditions: students judging single words, tested within a short session, which shows a strong effect for verbal material but does not establish what a witness to a complex event retained. When the framework is used to explain why a witness remembers some details but not others, counsel should ask whether the claim is testable, given Baddeley’s (1978) objection that depth has no measure independent of memory itself. When an expert relies on face-encoding studies such as Bower and Karlin (1974) or Schwartz et al. (2023), counsel should ask whether the conclusion follows for this witness, who was not instructed to judge faces and whose actual processing is unknown. When deep processing is offered as a sign of accuracy, counsel should ask whether the expert has considered Rhodes and Anastasi (2000), which found that deeper encoding increased false recall of related, unpresented items. These questions are explained in “Reading Research Claims: A Guide for Attorneys” on this site.

The Levels-of-Processing Effect and the Law

Federal and Military Standards

Expert testimony about encoding and memory is governed by Federal Rule of Evidence 702, as amended December 1, 2023, under which the proponent must show the court that it is more likely than not that the rule’s requirements are met, including that the opinion reflects a reliable application of reliable methods to the case. Daubert v. Merrell Dow Pharmaceuticals, Inc. (1993) made the trial judge a gatekeeper, and Kumho Tire Co. v. Carmichael (1999) extended that gatekeeping obligation to all expert testimony, including testimony based on technical or other specialized knowledge. In courts-martial, Military Rule of Evidence 702 tracks the federal rule, and the 2023 federal amendment took effect in the military rule in 2025 through Military Rule of Evidence 1102 (Joint Service Committee on Military Justice, 2025); practitioners should confirm current text. Military judges apply the six factors of United States v. Houser (1993). Levels-of-processing testimony is ordinarily general and educational, explaining how attention and encoding affect what can later be recalled.

Arizona

Arizona Rule of Evidence 702 adopted the federal formulation and the Daubert standard effective January 1, 2012, and was amended effective January 1, 2024, to conform to the 2023 federal amendment (Ariz. Sup. Ct. Order R-23-0004). State v. Chapple (1983) is Arizona’s leading case on expert testimony about the reliability of memory and eyewitness accounts: excluding a qualified expert was an abuse of discretion on its facts, while admission generally remains within the trial court’s discretion, and State v. Nordstrom (2001) approved limiting such an expert to general factors without opinions on the accuracy of a particular witness. Experts may not opine on whether a particular witness is truthful or quantify how often witnesses of a given type are truthful (State v. Lindsey, 1986, 149 Ariz. at 475–76).

Special Issues in Legal Practice

The principles described above apply with equal force to every party in a legal proceeding. The science operates the same way regardless of who is asking the questions, and the research summarized here does not by itself favor either side. What differs between the parties is not the science but the vantage point — the particular places where these issues tend to surface, and the particular errors each party is best positioned to guard against. The two summaries below identify those recurring issues. They are offered for orientation, not as litigation strategy; in every case the controlling question is what the evidence in that case will, and will not, support, and the expert’s role is to inform the trier of fact rather than to advocate for either side.

Special Issues for Prosecutors

The research supports several points prosecutors can legitimately present. Incidental encoding is the normal condition for witnesses, and when a witness was engaged with the meaning of what was happening, memory for its gist and central features can be good without any effort to memorize. A witness’s inability to recall details that played no part in understanding the event, such as background objects or exact phrasing, is consistent with ordinary encoding and does not by itself undermine the witness’s memory for what mattered. Interviews that reinstate context and invite free recall are consistent with this research and with the evidence for the cognitive interview.

The research also counsels caution. Meaning-based processing strengthens memory for implied as well as actual content, so a coherent account that captures the meaning of an event may still include details that were inferred rather than seen. Testimony that a witness “must have” processed an event deeply, or that engagement guarantees accuracy, goes beyond the research, which cannot reconstruct how a particular person encoded an event. Specific details on which a charge depends are best corroborated rather than inferred from the witness’s general engagement.

Special Issues for Defense

The research supports several points defense counsel can legitimately present. What a witness can remember depends on what the witness attended to and how, and a detail that played no role in the witness’s understanding at the time may have been poorly encoded or not encoded at all. Questions about such details may produce reconstructions or guesses, and deeper, meaning-based processing increases the risk of reporting related details that were not present. The fit between encoding and the later task matters, so the form of an identification or description procedure may bear on how much weight its result deserves.

The research also counsels caution about the limits that apply to the defense. Incidental encoding does not mean poor encoding; witnesses who were engaged with an event often remember its gist and central features well, and brief exposure does not by itself establish that a witness processed an event shallowly. The framework cannot measure how deeply a particular witness processed anything, and an expert may describe the conditions that affect encoding but may not tell the factfinder that a particular witness’s memory is unreliable. The same principles apply to the accused’s own account and to defense witnesses.

This paper is for educational purposes only and is not legal advice. The author is a psychologist, not an attorney; summaries of law are general and must be verified against current authority or with a licensed attorney.

References

Baddeley, A. D. (1978). The trouble with levels: A reexamination of Craik and Lockhart’s framework for memory research. Psychological Review, 85(3), 139–152. https://doi.org/10.1037/0033-295X.85.3.139

Bertsch, S., Pesta, B. J., Wiscott, R., & McDaniel, M. A. (2007). The generation effect: A meta-analytic review. Memory & Cognition, 35(2), 201–210. https://doi.org/10.3758/BF03193441

Bower, G. H., & Karlin, M. B. (1974). Depth of processing pictures of faces and recognition memory. Journal of Experimental Psychology, 103(4), 751–757. https://doi.org/10.1037/h0037190

Craik, F. I. M. (2002). Levels of processing: Past, present . . . and future? Memory, 10(5–6), 305–318. https://doi.org/10.1080/09658210244000135

Craik, F. I. M., & Lockhart, R. S. (1972). Levels of processing: A framework for memory research. Journal of Verbal Learning and Verbal Behavior, 11(6), 671–684. https://doi.org/10.1016/S0022-5371(72)80001-X

Craik, F. I. M., & Tulving, E. (1975). Depth of processing and the retention of words in episodic memory. Journal of Experimental Psychology: General, 104(3), 268–294. https://doi.org/10.1037/0096-3445.104.3.268

Hyde, T. S., & Jenkins, J. J. (1973). Recall for words as a function of semantic, graphic, and syntactic orienting tasks. Journal of Verbal Learning and Verbal Behavior, 12(5), 471–480. https://doi.org/10.1016/S0022-5371(73)80027-1

Intraub, H., & Nicklos, S. (1985). Levels of processing and picture memory: The physical superiority effect. Journal of Experimental Psychology: Learning, Memory, and Cognition, 11(2), 284–298. https://doi.org/10.1037/0278-7393.11.2.284

Joint Service Committee on Military Justice. (2025). Amendments to the Military Rules of Evidence since 1984 (G. E. Maggs, Ed.; 2025 ed.). https://jsc.defense.gov/Portals/99/MRE%20Amendments%202025.pdf

Lockhart, R. S., & Craik, F. I. M. (1990). Levels of processing: A retrospective commentary on a framework for memory research. Canadian Journal of Psychology, 44(1), 87–112. https://doi.org/10.1037/h0084237

Ly, D. P., Newman, E. J., & Bernstein, D. M. (2023). Repetition effects in memory and truth assessments through the lens of transfer-appropriate processing. Memory, 31(3), 328–345. https://doi.org/10.1080/09658211.2022.2154805

Memon, A., Meissner, C. A., & Fraser, J. (2010). The cognitive interview: A meta-analytic review and study space analysis of the past 25 years. Psychology, Public Policy, and Law, 16(4), 340–372. https://doi.org/10.1037/a0020518

Morris, C. D., Bransford, J. D., & Franks, J. J. (1977). Levels of processing versus transfer appropriate processing. Journal of Verbal Learning and Verbal Behavior, 16(5), 519–533. https://doi.org/10.1016/S0022-5371(77)80016-9

Rhodes, M. G., & Anastasi, J. S. (2000). The effects of a levels-of-processing manipulation on false recall. Psychonomic Bulletin & Review, 7(1), 158–162. https://doi.org/10.3758/BF03210735

Rogers, T. B., Kuiper, N. A., & Kirker, W. S. (1977). Self-reference and the encoding of personal information. Journal of Personality and Social Psychology, 35(9), 677–688. https://doi.org/10.1037/0022-3514.35.9.677

Schwartz, L., Cohen, M., Xu, S., Liu, J., & Yovel, G. (2023). The social-encoding benefit in face recognition is generalized to other-race faces. British Journal of Psychology, 114(S1), 213–229. https://doi.org/10.1111/bjop.12592

Slamecka, N. J., & Graf, P. (1978). The generation effect: Delineation of a phenomenon. Journal of Experimental Psychology: Human Learning and Memory, 4(6), 592–604. https://doi.org/10.1037/0278-7393.4.6.592

Symons, C. S., & Johnson, B. T. (1997). The self-reference effect in memory: A meta-analysis. Psychological Bulletin, 121(3), 371–394. https://doi.org/10.1037/0033-2909.121.3.371

Tulving, E., & Thomson, D. M. (1973). Encoding specificity and retrieval processes in episodic memory. Psychological Review, 80(5), 352–373. https://doi.org/10.1037/h0020071

Wells, G. L., & Hryciw, B. (1984). Memory for faces: Encoding and retrieval operations. Memory & Cognition, 12(4), 338–344. https://doi.org/10.3758/BF03198293

Legal Authorities

Daubert v. Merrell Dow Pharmaceuticals, Inc., 509 U.S. 579 (1993).

Kumho Tire Co. v. Carmichael, 526 U.S. 137 (1999).

State v. Chapple, 135 Ariz. 281, 660 P.2d 1208 (1983).

State v. Lindsey, 149 Ariz. 472, 720 P.2d 73 (1986).

State v. Nordstrom, 200 Ariz. 229, 25 P.3d 717 (2001).

United States v. Houser, 36 M.J. 392 (C.M.A. 1993).

Fed. R. Evid. 702.

Mil. R. Evid. 702.

Mil. R. Evid. 1102.

Ariz. R. Evid. 702.

Ariz. Sup. Ct. Order R-23-0004 (amending Ariz. R. Evid. 702, effective Jan. 1, 2024).

Related topic papers: The Self-Reference Effect · The Generation Effect · Gist Memory · Context Effects · Face Recognition · Improving Eyewitness Memory