Happiness Is a False Memory

Nietzsche once wrote, "[W]ith the smallest and with the greatest good fortune, happiness becomes happiness in the same way: through forgetting"¹. It turns out that happiness may be associated with memory in another way: happiness makes us more likely to make memories up, at least relative to when we're depressed. In fact, it might be better to say that depression makes memory more accurate. That's the gist of what a recent study on mood and false memories by Justin Storbeck and Gerald Core found. And aside from some interesting theoretical implications about both affect and memory, the study is sexy enough for a blog post.

Here's the backstory. Memory involves two main stages: encoding and retrieval. Encoding involves storing information in memory, and retrieval is the name for taking it back out (i.e. remembering). It's widely believed that there are two ways of encoding memories: item-specific encoding and relational encoding. Item-specific encoding is just what it says: individual items are encoded separately, with attention to their specific properties. Relational encoding is a little more complex. In essence, it involves encoding items through the activation of information already and memory, such as concepts or schemas. When information is encoded in an item-specific fashion, particular features of the instance are encoded, whereas when relational encoding occurs, the gist of the information is stored in memory to the detriment of particular features of the instance.

One interesting difference between the two types of encoding is that item-specific encoding tends to produce accurate memories about particular instances, whereas relational encoding can produce memory intrusions, or false memories. For example, in one of my favorite experiments of all time, Sulin and Dooling² gave participants bibliographical stories in which the main character was given either a famous name (Helen Keller or Adolph Hitler) or an unfamiliar name (Carol Harris or Gerald Martin). After reading the stories, participants read a list of three types of sentences: 1.) sentences from the original stories, 2.) sentences about the famous characters that were not in the original stories, and 3.) sentences that were unrelated to the famous characters and were not in the stories. Participants were asked to indicate which of the sentences they had previously read. Participants who'd read the stories with unfamiliar names were unlikely to say that they had seen sentences of type 2 or 3, but participants who had read the same stories with the famous names were almost as likely to say that they had seen sentences of type 2 as they were to say they had seen sentences of type 1. In other words, they were almost as likely to mistakenly remember seeing true sentences about the famous person as they were to remember sentences they had actually seen. The reason is that reading the famous name activated the knowledge of that person in their memory, and the story was encoded relationally, whereas reading a story about an unfamiliar person caused participants to focus on the particulars of the study.

Several recent studies have shown that affect can influence the type of encoding. Specifically, positive affect (a good mood) seems to enhance relational processing, while negative affect leads to more item-specific encoding³. For example, in one set of studies, Storbeck and Clore found that in several different kinds of semantic priming tasks (tasks that involve using priming words or objects with similar meanings to test objects), priming occurred for participants with a positive mood, but participants with a negative mood were much less likely to exhibit priming⁴. If positive affect leads to more relational processing, and negative affect leads to more item-specific encoding, and if relational encoding leads to more false memories, then we can predict that happy people will be more likely to produce false memories than depressed people.

In order to test this prediction, Storbeck and Clore used the now common Deese-Roediger-McDermott (DRM) paradigm. This paradigm involves presenting participants with lists of words that are all highly associated with another word (the "critical lure") that is not on the list. After reading the list, participants are asked to recall as many of the words on the list as possible, or are given another list of words and asked to indicate which had been on the list. In both cases, participants are likely to mistakenly remember having seen the critical lure. The task is sort of like the Sulin and Dooling task, in that when a concept is activated (by reading several words associated with the concept), relational processing occurs, and participants are likely to produce false memories.

To induce a positive or negative affect in participants, the experimenters played either Mozart's Eine Kleine Nacht Musik, which has been shown to produce a positive affect, or Mahler's Adagietto, which has been shown to produce a negative affect (I think we all know that Mahler is depressing, anyway), while participants read the DRM list. After completing the study, participants' mood was tested in a questionnaire, and only the data from participants whose moods were significantly above the median mood score (in the positive mood condition) or below it (in the negative mood condition) were kept. There was also a control group, which heard no music while completing the task.

The main findings in their two experiments were these: participants in all three conditions(positive mood, negative mood, and control) recalled words that were on the list with equal accuracy, while participants in the positive mood and control conditions produced many more false memories (recalling the critical lures) than participants in the negative mood condition. There was no difference between the number of false memories produced by participants in the positive mood and control conditions. Even when participants were asked to list any words that were not on the list, but that came to mind during the recall task (a common procedure designed to test whether the critical lures were activated during encoding even if participants are aware that they weren't on the list), the participants in the negative mood condition were significantly less likely than participants in the other two conditions to list the critical lures. This implies that the lures were not activated during encoding, and supports the hypothesis that negative mood induces item-specific encoding. The fact that there was no difference between the false memory production in the positive mood and control conditions implies that we may generally encode information relationally, and negative mood causes a shift away from our normal encoding processes toward more item-specific processes.

It would be natural to ask whether these findings might have clinical implications for people suffering from depression, and the answer is, I don't know, and the authors of the paper don't discuss clinical issues. It is interesting to note that we've known for some time that traumatic events often induce increased accuracy for the details (e.g., specific perceptual details) of the event, while memories for the overall gist of traumatic events tends to be poor. It's generally believed that this is because higher levels of arousal lead to more detailed processing and encoding of information. Based on the results of the Storbeck and Clore study, it may be that the negative affect produced by traumatic events also helps to produce more accurate recall for details, and less accuracy for the gist of the events.

¹In "On the use and abuse of history for life."
²Sulin, R. & Dooling, D. (1974). Intrusion of a thematic idea in retention of prose. Journal of Experimental Psychology, 103, 255-262.
³E.g., Gasper, K. & Clore, G. L. (2002). Attending to the big picture: Mood and global vs. local processing of visual information. Psychological Science, 13, 34-40; Isen, A., & Daubman, K. (1984). The influence of affect on categorization. Journal of Personality and Social Psychology, 47, 1206–1217; Bless, H., Clore, G.L., Schwarz, N., Golisano, V., Rabe, C., & Wolk, M. (1996). Mood and the use of scripts: Does a happy mood really lead to mindlessness? Journal of Personality and Social Psychology, 71, 665–679; Storbeck, J. & Clore, G. L. (Submitted Manuscript). Mood governs implicit processes: Affect, priming, and false memories.
⁴Storbeck & Clore (Submitted).