Mood-Congruent Memory Bias: Why Your Mood Chooses What You Remember

A mood is not only something you feel; on Bower's account it is also a node in the same network that stores everything you know, connected to the events, words and ideas that have gone with that feeling before. When the mood is active, activation spreads along those connections, and whatever is congruent with the mood is already partly lit up: easier to notice, easier to elaborate while it is being learned, and easier to reach when memory is searched. Material that clashes with the mood gets none of this help, so the bias runs at encoding and at retrieval alike.

Bower drew a distinction that the field has kept. Mood-congruent memory is about the emotional content of the material: sad people learn and recall sad things more easily. Mood-dependent memory, also called state-dependent retrieval, is about the match between the mood at learning and the mood at recall, whatever the material: something learned while happy should come back better when happy again. The same network theory predicts both, and they have fared very differently in the laboratory. Mood congruence proved reasonably robust, while mood dependence was hard to reproduce; Bower and Mayer reported in 1985 that they could not replicate it, and later work found it only under particular conditions, such as free recall of self-generated events after a long delay.

The evidence for mood congruence itself is more mixed than the clean story suggests. Blaney's 1986 review found that 25 of 29 mood-induction studies reported some evidence of it, but the meta-analysis by Matt, Vázquez and Campbell in 1992 showed that the effect depends on who is tested and how the mood was produced. People who were not depressed recalled more positive than negative material; subclinically depressed people recalled the two about equally; people with clinical depression or in an induced sad mood showed a bias toward negative material, comparable in size across the two; and an induced happy mood produced a positive bias whose confidence interval included zero. The robust finding is that low mood and depression tilt memory toward the negative. The symmetrical claim, that a happy mood tilts it toward the positive, sits on top of a baseline positivity that most people show anyway.

Two other features matter for everyday life. Bower, Gilligan and Monteiro found that the mood during reading did most of the work: material congruent with the mood at learning was learned better, while a mood induced only at recall had little effect on which incidents came back. And the bias operates on interpretation as well as on memory: an ambiguous remark is read in the mood's color before it is stored, so what is later recalled is already tilted. In clinical settings this loop is one reason low mood sustains itself.

Gordon Bower's "Mood and memory," published in American Psychologist in February 1981, was a review of a program of experiments from his laboratory at Stanford, and it set out the associative network theory that organized the field for the next two decades. The experiments it drew on used hypnosis to induce moods, because Bower's group had found that posthypnotic suggestion produced strong and lasting happy or sad states in the minority of people who were highly susceptible to it; Bower put that minority at roughly a fifth to a quarter of people, which is why the participants were screened for susceptibility first.

The central experiments on mood congruence were published the same year by Bower, Stephen Gilligan and Kenneth Monteiro in the Journal of Experimental Psychology: General as "Selectivity of learning caused by affective states." In the first experiment, people made happy or sad read a story about two characters, one happy and one sad, and later recalled more facts about the character whose mood matched their own, with whom they also reported identifying. The second experiment showed that this selective recall by character could not be produced by inducing the mood only at recall, after the story had been read in a neutral state. In the third, readers took in a text in which one character described many unrelated happy and sad incidents from his life, with mood induced independently at reading and at recall; mood during reading caused selective learning of the congruent incidents, while mood during recall had little effect. The fourth replicated that null result for recall mood, and the fifth pitted the tone of the incidents against the mood of the character narrating them and found that readers learned the mood-congruent incidents rather than favoring the mood-congruent character.

Bower's 1981 review also described diary studies and experiments on childhood memories retrieved in happy or sad states, reported there in outline. The theoretical proposal was the part that lasted: emotions as nodes in a semantic network, linked to the events and concepts that had accompanied them, with spreading activation making congruent material more available at learning and at recall.

The program had a well-known difficulty. Mood-dependent retrieval, the prediction that matching mood at learning and recall would improve memory for any material, failed to replicate in Bower's own laboratory, as Bower and Mayer reported in 1985. Mood congruence survived better, as Blaney's review of 1986 and the 1992 meta-analysis showed, with the qualifications described above, and the network theory remains the frame in which both effects are discussed.

The bias shows up wherever a mood acts as the search key for the past.

The counter is to recognize when a mood is doing the remembering, and to go outside memory for the record.

• Name the state before you judge. A verdict reached in a strong mood about your relationship, your job or your year is a verdict reached with half the evidence missing. Label the mood and defer the conclusion.
• Keep records that moods cannot edit. A journal, a diary of good weeks, a list of things that went right, or notes kept through the year give you a memory that does not change with how you feel.
• Search on purpose for the other half. When low, ask directly for three things that went well; when elated, ask what the risks were. Deliberate retrieval can reach what the mood leaves dark.
• Change state before deciding. Sleep, a walk, a conversation about something else; the point is not to feel better for its own sake but to search memory with a different key.
• Be careful what you learn in a strong mood. Bower's experiments found that the mood at the time of learning did most of the work. Material taken in while anxious or angry will carry that color when it is recalled.
• Expect the asymmetry. Low mood reliably tilts memory toward the negative; in most people a neutral or good mood sits on a baseline that already favors the positive. The low days are the ones that need the counter most.

Mood-congruent memory sits beside several other effects of emotion and context on memory, each with its own page.

• It is not cue-dependent forgetting or its mood version, mood-dependent memory, in which the match between the state at learning and the state at recall decides what comes back regardless of the material's tone. Mood congruence is about the emotional content of what is remembered; mood dependence is about the match of states, and it has proved much harder to reproduce.
• It is not the fading affect bias, in which the feeling attached to unpleasant memories fades faster over time than the feeling attached to pleasant ones. That is a change in the memories themselves over months; mood congruence is a change in which memories are reachable right now.
• It is not negativity bias, the greater weight given to negative information in general. Mood congruence can run in either direction, and in people who are not depressed the baseline tilt is toward the positive.