Cognitive dysfunction is prominent in homeless and precariously housed persons, and memory dysfunction is the most pervasive domain. The presence of multimorbid physical and mental illness suggests that several underlying mechanisms of memory impairment may be at play. The serial position phenomenon describes the tendency to best recall the beginning (primacy effect) and last (recency effect) words on a supra-span wordlist. Recency recall engages executive and working-memory systems, whereas primacy recall depends on long-term memory. This study investigates memory dysfunction in a homeless and precariously housed sample by identifying and characterizing unique subtypes of serial position profiles on a test of verbal memory.

Data were used from a 20-year study of homeless and precariously housed adults recruited from an impoverished neighbourhood in Vancouver, Canada. Participants were sub-grouped according to their serial position profile on the Hopkins Verbal Learning Test-Revised using a latent profile analysis (LPA; n = 411). Paired samples t-tests were conducted to determine differences in percent recall from each word-list region within classes. Linear regression analyses were used to examine between-class differences in mean serial position scores and other cognitive measures (memory, attention, processing speed, cognitive control). Covariates included age, sex, and education.

LPA identified two profiles characterized by (1) reduced primacy relative to recency (RP; n = 150); and (2) reduced recency relative to primacy (RR; n = 261). Pairwise comparisons within the RP class showed that recency was better than primacy (p < .001, d = .66) and middle recall (p < .001, d = .52), with no difference between primacy and middle recall (p = .68, d = .04). All pairwise comparisons differed within the RR class (primacy > middle recall: p < .001, d = 1.85; primacy > recency recall: p < .001, d = 1.32; middle > recency recall: p < .05, d = .132). The RP class had worse performance on measures of total immediate (ß = .47, p < .001) and delayed verbal recall (ß = .32, p < .001); processing speed (ß = .20, p < .001); and cognitive control (ß = .22, p < .001). The RR class made more repetition errors (ß = .25, p < .001).

These findings support substantial heterogeneity in memory functioning in homeless and precariously housed individuals. The RP profile was characterized by poorer cognitive functioning across several domains, which suggests multiple contributions to memory impairment, including dysfunction of long-term memory circuitry. The RR profile with their higher number of repetition errors, may experience difficulties with self-monitoring in verbal learning. Subsequent studies will explore the neurobiological underpinnings of these subgroups to further characterize profiles and identify targets for cognitive intervention.