Cued by music: Eliciting autobiographical memories across the lifespan and assessing their emotionality

Abstract

Music holds a unique capacity to involuntarily and rapidly evoke autobiographical memories (AMs), yet the source of its cueing ability is unspecified. In experiment 1, I examined what components of music are most beneficial in triggering AMs. Younger adults were presented with short clips of popular songs from their youth as memory cues, that were either unmodified, instrumental versions, lyrics-only, or a visual presentation of song and artist name. If participants experienced a memory, they provided a description of it. Unmodified song clips were most effective at evoking memories, and recall was enhanced in those high in trait music-related reward sensitivity. I also found evidence of temporal alignment between the song’s year of popularity, and timeframe of the evoked AMs. Given the powerful cueing effect of music on AMs, in experiment 2 I examined whether the findings would generalize to an older adult population, and whether music could serve to reduce age-related episodic memory deficits. Strikingly, song cues evoked more AMs in older than younger adults, offering a means of enhancing access to personal memories. I observed a temporal alignment between the year of popularity of the song cue and the time frame of the evoked memory in both age groups. I propose that songs are particularly effective memory cues because they help set temporal context, constraining the search for AMs to a specific point in time, facilitating access. These findings are particularly relevant for older adults who may have difficulty spontaneously recreating context. In chapter 3, I investigated whether computational approaches using natural language processing can be employed to examine emotional sentiment within autobiographical narratives. Current methods of analysis require extensive manual human coding, limiting the sample size that can feasibly be examined. I compared the congruence in classification of two popular lexicon-based sentiment analysis tools, VADER and TextBlob, with self-reported valence of 3,309 AM narratives from two datasets. Confusion matrices showed better congruence using VADER than TextBlob. Accuracy improved significantly when classification required binning valence into three rather than five categories, regardless of dataset size. Results suggest sentiment analysis is a promising avenue to determine broad classifications of valence within AM narratives. Overall, this thesis shows that songs are powerful cues for eliciting personal memories and can reduce age-related memory decline by setting the temporal context, facilitating access to content. Moreover, this thesis shows that VADER is a sentiment analysis approach that can be reasonably used to determine emotional valence, offering a means of characterizing autobiographical memory narratives.