A Computerized Continuous-Recognition Task for Measurement of Episodic Memory

Abstract

Based on clinical observations of severe episodic memory (EM) impairment in dementia of Alzheimer's disease (AD), a brief, computerized EM test was developed for AD patient evaluation. A continuous recognition task (CRT) was chosen because of its extensive use in EM research. Initial experience with this computerized CRT (CCRT) showed patients were very engaged in the test, but AD patients had marked failure in recognizing repeated images. Subsequently, the test was administered to audiences, and then a two-minute online version was implemented (http://www.memtrax.com). The online CCRT shows 50 images, 25 unique and 25 repeats, which subjects respectively either try to remember or indicate recognition as quickly as possible. The pictures contain 5 sets of 5 images of scenes or objects (e.g., mountains, clothing, vehicles, etc.). A French company (HAPPYneuron, SAS) provided the test for 2 years, with these results. Of 18,477 individuals, who indicated sex and age 21-99 years and took the test for the first time, 18,007 individuals performed better than chance. In this group, age explained 1.5% of the variance in incorrect responses and 3.5% of recognition time variance, indicating considerable population variability. However, when averaging for specific year of age, age explained 58% of percent incorrect variance and 78% of recognition time variance, showing substantial population variability but a major age effect. There were no apparent sex effects. Further studies are indicated to determine the value of this CCRT as an AD screening test and validity as a measure of EM impairment in other clinical conditions.

Keywords: Aging; Alzheimer’s disease; cognition; cognitive impairment; dementia; episodic memory; memory; reaction time; recognition.

Fig.1

A) Number of subjects by year of age, B) Percent of subjects providing each number of responses.

Fig.2

A) Percent of subjects versus percent incorrect. The plots are essentially the same for males and females. About 35% of subjects missed none or 1 image. About 45% erred on 3 or more. 12% errors = 1.5 standard deviations; 16% errors = 2.0 standard deviations; 20% errors = 2.5 standard deviations; 24% errors = 3.0 standard deviations, B) i,ii Percent of incorrect responses for all subjects, 5,665 males (i), 12,342 females (ii), with linear regression lines and linear equations,C) Percent of incorrect responses were averaged for each year of age by gender with +1 and +2 standard deviations calculations for each age, trendlines for each Male and Female calculated by EXCEL spreadsheet graph. Table shows Male and Female, Mean, +1 standard deviation +2 standard deviations, linear regression curves, R-squared, quadratic regression lines, R-squared. Graph shows: Male: dashed lines; Female: dotted lines; quadratic trendlines, lower are for the average of each year of age, with +1 STD lines above them, and +2 STD lines above that.

Fig.3

A) Percent of subjects versus mean recognition times, B) i,ii Mean response times for all subjects (seconds), 5,665 males (i), 12,342 females (ii), with regression lines and linear equations, C) Mean response times (seconds) averaged for each age by gender with +1 and +2 standard deviations calculations for each age. Table shows data from EXCEL spread sheet calculations for trendlines for Male and Female, Mean, +1 standard deviation, +2 standard deviations, linear regression curves, R-squared, quadratic regression lines, R-squared. Graph shows: Male: dashed lines; Female: dotted lines; quadratic trendlines, lower are for the average of each year of age, with +1 STD lines above them, and +2 STD lines above that.

Fig.4

Plot of mean response times (seconds) for every subject versus percent incorrect, with linear regression lines and equations, respectively.