Let’s start with a bit of reflection…
Think about a good memory you have from the past. How much of the detail can you recall? Now think about a negative memory you have (from around the same time, if you can), and how much detail you can remember. Try to do the same for something mundane that you might recall without emotion attached to it. Which is more vivid?
It is thought that people remember emotional events more intensely (so-called “flashbulb memories”). In fact, research suggests that for more emotional memories, the focus of emotion is often remembered clearly, but details in the background less-so, in a trade-off fashion. The exact emotion being felt interacts complexly with the specificity of future recall. Overall, negative emotions may be more salient later on, or have a greater trade-off in the detail retained, but this is influenced by factors such as age (as one gets older, positive recall tends to increase) or trauma (which can decrease subsequent negative recall specificity).
People with depression appear to remember autobiographical events in less detail than people without depression (van Vreeswijk et al, 2004). Some previous studies have found that positive, but not negative, events are recalled with less clarity by depressed than non-depressed participants (Young et al, 2012). A meta-analysis published in 2010 found that this ‘overgeneralised’ memory predicted a poorer future outcome of greater depression severity in short-term follow-up measurements (Sumner et al, 2010).
However, it’s also important to note much evidence suggesting that people suffering from depression have impairments in fundamental cognitive processes, such as memory, even in non-emotional laboratory tasks. Gallagher and colleagues found that over time, individuals who recovered from depression had improved memory performance over time for non-emotional words (read in a list), in contrast with no improvement by patients who did not recover (Gallagher et al, 2007).
Measuring real-life memory might be more representative of individuals’ lived experiences than remembering a list of non-emotional words in a research assessment, but it will also be biased by the nature of events that have occurred and are being remembered. Furthermore, it is possible that being depressed during past events may affect how someone reminisces later.
How does this all fit with your trip down memory lane?
Lewis and colleagues (Lewis et al, 2017) attempted to assess automatic or implicit emotional memory, using a similar verbal memory task to widely-used wordlist tasks (including by Gallagher et al. in the study mentioned above), but using words considered to be likeable/unlikeable character traits, read from a screen. They wanted to focus on recall with a social element, due to evidence that:
people with depression are more sensitive to punishments (including social criticism) and less sensitive to rewards (including social appraisal).
The authors did not explicitly state directional hypotheses, but they wished to find out whether depressed individuals appear more negative to others because:
- Their memory (or attention) of negative information is greater
- Their memory (or attention) of positive information is lessened
- Both of the above.
The study design was a longitudinal, prospective cohort investigation. Individuals were invited to participate if they had presented to primary care services with depressive symptoms in the preceding year and were aged 18-70. Exclusion criteria were having a diagnosis of a bipolar, psychotic, substance-use or eating disorder, being pregnant or not able to complete questionnaires.
At three different appointments held every fortnight, researchers asked the 558 participants to recall personality traits that had recently been shown for 0.5 seconds each on a screen, and were positive or negative in nature. This exercise anticipated to measure individuals’ ability to remember socially rewarding (positive words) or punishing (negative words) information. The task included forty words (20 likeable, 20 dislikeable characteristics) and words were different at each assessment, which meant there would not be expected score increases over time (practice effects). A total of six words are reported as examples of those included in the task (see below picture), but we don’t know what the rest are. The authors cite a prior use of this task (which provides the same information) and the validation ratings of words: The words used were originally rated as likeable or unlikeable by 100 college students, published in 1968, and the list is over 500 words long (Anderson, 1968). It is not clear what subset was used in this study or how the words were chosen (and indeed, how variably positive or negative they were rated as, by different raters). The 1968 study warned that the ratings may not be equivalent outside a college student population.
At each time point (plus a fourth measure, a fortnight later), researchers also measured the severity of depressive symptoms experienced over the previous two weeks. For this, the commonly-used ‘Beck Depression Inventory’ (BDI) was used; a self-rated questionnaire containing 21 items.
They also measured potentially confounding (or modifying) variables; those thought to affect automatic emotional memory or depression severity: age, sex, educational attainment and negative life events (including bereavement, separation/divorce, illness/injury) as well as antidepressant use.
In order to address the study aims, the authors assessed:
- The predictability of how many words were recalled on concurrent depressive symptoms
- The role of positive versus negative word recall on the above relationship
- For each participant over time, pooled both depressive symptom severity scores and recall score. This was done using multilevel modelling techniques that account for correlated scores over time, missing data and measurement error. This method also permits an insight into causality (whether memory performance changes precede symptom changes or vice versa)
- The influence of possible confounding factors.
As expected, participants who had less severe depression remembered more positive words than those with more severe depression. This was the case across all time points.
What the authors hadn’t hypothesised, was that there would be no relationship between the number of negative words remembered and depression severity.
The findings did not indicate that having more severe depression made individuals likely to remember less in future assessments, but performing better in positive word recall was associated with somewhat reduced depression severity in future measurements.
These results were not affected by whether participants were taking antidepressants or not, or other measured confounders.
Strengths and limitations
As always, there are a large number of factors that are important to consider when interpreting these results:
- This was the largest study published so far assessing the severity of depression and emotional memory processing in depression.
- It was a cohort study, meaning that those entering the study were all recruited from the same population. This group varied in age, health status and other variables, which means that they are likely more representative of the general population than many study samples, but also involves increased heterogeneity present (which measured confounders could not all account for).
- The authors did attempt to account for the influence of other factors likely to be important, such as comorbid health status, medication use and age.
- Depression was assessed as a severity continuum rather than using discrete groups (e.g. ‘depressed’ / ‘non-depressed’). The authors report that their sample varied from absent to severe depression, although I would have been interested to see a score range, average or distribution at any time-point.
- Unlike most previous studies, this task was designed to make participants remember emotionally relevant information. It is challenging to find a task that is meaningful emotionally (in terms of social reward or punishment) to the participant but also objective, measurable and comparable across individual. For example, recall of emotional events that have happened to participants is highly relevant and salient, but difficult to assess objectively and accurately.
- Did the task really measure automatic social reward or punishment memory processing? It is an interesting idea for a task, but I also think more work needs to validate exactly what this task detects.
- I would be interested to know why depression was measured at 0, 2, 4 and 6 weeks while memory was only measured at 0, 2 and 4. I wondered whether this might be due to an expectation that memory recall precedes mood changes (whereas actually mood appeared to precede memory changes).
- Approximately a fifth of assessments were not completed from included participants across time points. While some missing data is unavoidable, it may have been people suffering from the most severe depression who were unable to complete all assessments. Having said this, the statistical methods are reported to deal well with missing data.
- The authors also state a possible selection bias that could have occurred in those who decided to take part. Only 7.2% of potentially eligible people who were contacted actually participated, which might have been people who were feeling more positive or less negative at the time (or who had a better/poorer memory) than non-participants contacted.
- We aren’t aware of whether participants undertook the memory task or depression severity measure first or second. I wonder whether this could make a difference. If someone noticed that they had only remembered the negative words, perhaps this could influence how symptomatic they recall the last few weeks in completing the BDI depression measure. The BDI has indeed been criticised in a number of ways, including high fluctuation in scores attained across short periods of time.
Implications and conclusions
Despite these potential issues, the findings from this paper may have important implications. The most popular psychological therapy for depression (cognitive behavioural therapy) is based on models of reducing negative biases to promote recovery from depression. But perhaps this isn’t the optimal path to take, and actually focusing on increasing the positive, rather than ameliorating the negative cognitive biases might enhance recovery rates and resilience to depression.
There is a question over whether people who do not get depression actually do experience an overall positive bias (analogous to so-called ‘depressive realism’), which is challenging to detangle quantitatively. More work is needed to better understand what promotes resilience in depression, and how these constructs might be approached in therapy, but these findings putatively have vital repercussions for successfully treating depression and preventing relapse, if future evidence corroborates.
Lewis G, Kounali DZ, Button KS, Duffy L, Wiles NJ, Munafò MR, Harmer CJ. (2017) Variation in the recall of socially rewarding information and depressive symptom severity: a prospective cohort study (PDF). Acta Psychiatrica Scandinavica, 135(5), 489-498. DOI: 10.1111/acps.12729
Reed AE, Chan L, Mikels JA. (2014) Meta-analysis of the age-related positivity effect: age differences in preferences for positive over negative information. Psychology and Aging, 29(1), 1-15. DOI: 10.1037/a0035194
Ono M, Devilly GJ, Shum DH. (2016). A meta-analytic review of overgeneral memory: The role of trauma history, mood, and the presence of posttraumatic stress disorder. Psychological Trauma, 8(2), 157-64. DOI: 10.1037/tra0000027
van Vreeswijk MF, de Wilde EJ. (2004) Autobiographical memory specificity, psychopathology, depressed mood and the use of the Autobiographical Memory Test: A meta-analysis. Behaviour research and therapy, 42(6), 731-743. DOI: 10.1016/S0005-7967(03)00194-3
Young KD, Erickson K, Nugent AC, Fromm SJ, Mallinger AG, Furey ML, Drevets WC. (2012) Functional anatomy of autobiographical memory recall deficits in depression. Psychological medicine, 42(2), 345-357. DOI: 10.1017/S0033291711001371
Sumner JA, Griffith JW, Mineka S. (2010) Overgeneral autobiographical memory as a predictor of the course of depression: A meta-analysis. Behaviour research and therapy, 48(7), 614-625. DOI: 10.1016/j.brat.2010.03.013
Gallagher P, Robinson LJ, Gray JM, Young AH, Porter RJ. (2007) Neurocognitive function following remission in major depressive disorder: potential objective marker of response?. Australian & New Zealand Journal of Psychiatry, 41(1), 54-61. DOI: 10.1080/00048670601057734
Anderson NH. (1968) Likableness ratings of 555 personality-trait words. Journal of personality and social psychology, 9(3), 272. DOI: 10.1037/h0025907
Richter P, Werner J, Heerlein A, Kraus A, Sauer H. (1998) On the validity of the Beck Depression Inventory. Psychopathology, 31(3), 160-168. DOI: 10.1159/000066239