Quantitative Measurements of Autobiographical Memories

Abstract

The scientific pursuit of the mind-brain relationship demands the ability to quantify subjective cognitive phenomena. I applied two new psychometric tools, called CRAM (Cued-Recalled Autobiographical Memories) test and SPAM (Spontaneous Probability of Autobiographical Memories), to probe the quantitative dimensions of autobiographical memories (AMs) in humans. In CRAM, a computer-based protocol first prompts the retrieval and temporary labeling of episodic memories by using word cues sampled from a word database with a frequency proportional to their natural language usage. Subjects then sequentially estimate when each event occurred and report the number of features for each of several elements (e.g. people, objects, location details, etc.) they can recall for every AM. This test has been also adapted for the web and is available at http://cramtest.info. Data collected from largely undergraduate participants (N = 191, M = 22 years old, SD = 7 years) confirmed classic results describing the temporal distribution of AMs as a steep, gender-independent power decay from the most recent episodes to childhood amnesia. Analysis further revealed a much more modest dependence of the total content of individual AMs on the episode age (M = 18.7 SE = 0.24 elements overall vs. M = 16.1 SE = 0.32 and M = 23.5 SE = 0.26 for the most remote and recent tenths of life, respectively). Contrasting among features, People and Episodes were found to be the least and most susceptible to temporal degradation (-33% vs -100% lifespan reduction), respectively. Elements of People and Context were found to be more independent of other features (i.e. possibly “primary”) for females than males, while the opposite held true for location and temporal elements. SPAM is a separate experiment that assesses the number of AMs recalled per unit of time by stochastically sampling, through an automated paging system, the probability to be reminiscing one’s past episodes, and the duration of these recalls. A first pool of volunteers (n = 48, Mdn = 21 y.o., MAD = 3 yrs..) retrieved, on average, 21 (SE = 2.65) AMs per waking hour. Combining all results enables rather detailed inferences, such as “during a typical day, 45 elements are recalled from the middle fifth of one’s lifespan”. Storage of age-specific population statistics in a large-scale informatics database will allow future queries of multi-dimensional frequency maps of human memories. Practical implications of both the data and the experimental design include improved diagnosis and monitoring of diseases such as Alzheimer’s and critical assessment of eyewitness reports.