What Changes When We Relate a Word to Ourselves?

Three questions can accompany the same adjective: is it written in capitals, does it mean the same as another word, and does it describe me? The letters need not change for the mental activity to change. Which words return later may reflect the relationship established with them, rather than visual exposure alone. This is the experimental starting point of the self-reference effect. It is not a claim that admiring yourself improves memory. The comparison concerns evaluating a trait in relation to yourself rather than making a judgment about its appearance or meaning.

In 2017, Sarah V. Bentley, Katharine H. Greenaway and S. Alexander Haslam investigated this comparison online. Participants answered questions about adjectives on their own computers and subsequently completed memory tasks. Across four experiments, personally evaluated words could become prominent in free recall. The results also showed why presentation timing, test format and the definition of a score matter. Reducing the paper to advice to connect everything to yourself would remove the comparisons that make its findings informative. [1]

A personal question does not require an approving answer

The initial experiment used thirty adjectives. Participants evaluated ten in relation to themselves, ten through a meaning question, and ten through a structural question about letter case. Responses were yes or no. Someone deciding that a word did not describe them still made a self-referential judgment. The relevant distinction was the question considered, not whether the participant embraced a flattering description.

Half the adjectives were positive and half negative, with that balance maintained across encoding types. Six lists associated words with different questions and presentation orders. This helped prevent an inherently memorable word from being permanently assigned to one kind of judgment. The manipulation concerned how a word was processed, rather than treating each word as if it necessarily initiated one fixed mental operation. Self-reference in this procedure should not be confused with a personality diagnosis or a measure of self-importance. [1]

Who controls the screen's pace?

Initially, selecting an answer advanced the participant to the next screen. People could move through the questions at their own pace, averaging approximately two seconds per word according to the authors. Some then answered five arithmetic questions, while others proceeded directly to memory testing. The unexpected free-recall task allowed two minutes to enter remembered words in any order.

The next experiment introduced a more controlled sequence. A question appeared for two seconds before the adjective appeared beneath it. Question and word remained together for three further seconds before answer choices became available. Subsequent progress still depended on response speed. Calling this exactly five seconds of word viewing would be inaccurate: during the first two seconds, the word had not appeared. Timing meant controlling when particular information became available, not simply setting a total session length. [1]

The second experiment produced a clearer personal-encoding advantage over semantic encoding. However, slower presentation was not randomly contrasted with faster presentation within an otherwise identical experiment. The participants and sample size changed, and two words were replaced. Those differences prevent assigning the entire change to timing alone. The comparison makes presentation procedure worth investigating; it does not establish a universal number of seconds needed to learn a word.

A percentage needs a denominator

An especially important methodological detail concerns the proportions used in several analyses. The authors defined the number correctly recalled from one encoding category relative to the total number of correct words recalled by that participant. This is different from the fraction of all presented words in that category that the participant remembered. A large share of the remembered material can coexist with modest overall recall.

Consider an invented illustration, not a result from the study. Suppose someone remembers four of the thirty words, with two coming from the ten personally evaluated words. Personal encoding accounts for half of the remembered set, but only two of its ten presented words were recalled. The numerator stays the same while the denominator changes the meaning. A conspicuous percentage should therefore not be translated automatically into a percentage improvement in total memory. The paper also used absolute counts for certain comparisons between test groups, so its results did not all share one measurement scale. [1]

Why remove the beginning and end?

The researchers conducted analyses both with all words and with words in the first three and last three presentation positions excluded. This examined whether the comparison survived removing list positions associated with primacy and recency. In the initial experiment, the self-reference advantage over semantic encoding was significant when all words were included, but not when those positions were excluded. The strong wording of an overall summary should not obscure this qualification.

That result does not erase the later experiments' free-recall findings. It does show why describing a recurring pattern is different from claiming that every analysis produced an identical outcome. Choosing which items enter an analysis can change the contrast that remains. Reporting both approaches makes the boundary visible instead of silently choosing the version that tells the simplest story.

Producing a word and recognizing one

After the encoding task in the third experiment, participants were randomly assigned to free recall or recognition. Recognition mixed the thirty previously presented words with thirty new ones and asked whether each word was old or new. The participant did not have to generate the word from an empty response field. The test supplied a candidate, changing the information available during retrieval. [1]

Self-referential encoding retained an advantage in free recall under both analysis approaches. Recognition produced a more qualified pattern. With the beginning and end positions removed, the self-versus-semantic difference was not significant. With all words included, a smaller advantage remained. Neither a universal disappearance of the effect in recognition nor complete independence from test format accurately describes those findings.

One interpretation discussed by the authors is that personal knowledge supplies useful retrieval connections, whereas semantic encoding can benefit especially from the external cues provided by recognition. That is an explanation of the pattern, not direct observation of every mental step. The study did not resolve all questions about elaboration and organization, nor did it measure brain activity to locate the mechanism. A behavioral contrast can constrain an explanation without supplying an entire account of how remembering happens.

Knowing about the memory test in advance

The fourth experiment informed some participants before encoding that they would later need to recall the words. The researchers expected this warning to strengthen other encoding conditions and reduce the personal advantage. The predicted interaction was not demonstrated; the self-reference advantage persisted. This does not establish that advance knowledge never affects learning. In the unsupervised online setting, participants were also instructed not to write the words down. Their actual study behavior was not observed as directly as it could be in a laboratory. [1]

The experiment therefore tests one particular warning procedure, not every possible preparation strategy. Likewise, failure to demonstrate the predicted moderation does not prove that informed and incidental learning are exactly equivalent. Keeping those distinctions intact allows the unexpected result to remain informative without turning it into a general rule about examinations or deliberate studying.

What the online procedure establishes

The paper's contribution is a way to investigate personal encoding beyond a traditional laboratory session. An online task can reach participants using their own computers, but access and scale do not automatically establish representativeness or identical engagement. The authors noted limitations concerning English proficiency and the absence of matching on word frequency and arousal. These are concrete properties of the material and recruitment context, not reasons to presume that an unmeasured factor explains everything.

There are reporting uncertainties too. The overall participant total does not reconcile straightforwardly with the separately stated experiment counts, and some statistical details require clarification. Those discrepancies have not been resolved here by reanalyzing the raw data. Consequently, exact outcome percentages are not presented as independently verified improvements in general memory. The experiments also did not directly establish long-term classroom achievement, clinical treatment benefits or effects throughout childhood and old age.

The strongest conclusion remains specific: evaluating a word in relation to oneself can change its place in subsequent remembering, and an online procedure can reveal that pattern. What counts as an advantage still depends on the alternative judgment, the retrieval task and the score being compared. Personal relevance becomes an experimentally useful relationship, not a guarantee that anything connected with the self will be retained indefinitely. [1]

Sources

[1] Sarah V. Bentley, Katharine H. Greenaway and S. Alexander Haslam (2017), An online paradigm for exploring the self-reference effect, PLOS ONE 12(5): e0176611. Open-access primary research. Original explanation based on the Creative Commons Attribution-licensed research, with attribution to its authors.

Source access and factual review: 15 October 2026. The complete main text was read; supplementary data and code were not executed, and numerical tables were not independently analyzed. The denominator definition was checked in the methods; reported percentages were not treated as general memory improvement.