Memory for Emotion-Laden Words in Normal Aging:Valence-Arousal Interactions and Neuroanatomical Correlates

Healthy aging can be characterized as a multidimensional process thatinvolves both age-related physical/cognitive deterioration andpreserved functioning in socioaffective domains and specific cognitiveabilities. This is also true for episodic memory, as healthy agingsimultaneously entails general episodic memory decline and preservedemotional enhancement effect in memory (EEM). EEM is defined as theaugmentation of the formation and strength of memory traces foremotion-laden information. In this field of research, emotion iscommonly conceptualized as two bidirectional dimensions, valence(negative-positive) and arousal (calming-exciting). Even though EEMis maintained with age, age differences tend to emerge in the valencespecificpreferences for emotion-laden information. This so-calledpositivity effect can manifest as a reduced preference for negative overpositive material commonly seen in young adults, or even as apreference for positive over negative material (positivity bias) inmiddle–aged and older adults. Neuroimaging evidence indicates thatalso the activation and functional connectivity patterns of EEM-relatedbrain regions differ between young and older adults. However, due tothe dearth of research in this particular field, the psychological andneural mechanisms of these age differences remain poorly known.

The overarching aim of the present thesis was to examine thepsychological and neural mechanisms underlying memory foremotion-laden words in aging. First, age- and gender-specific normsfor valence and arousal ratings for 420 frequency-controlled Finnishnouns were established (Study I). These norms were used to selectstimuli for the three subsequent studies (Studies II-IV), which aimed ata) investigating the role of arousal in age differences concerningvalence-specific preferences in memory; and b) examining theneuroanatomical substrates of memory for emotion-laden words incognitively intact middle-aged and older adults. Memory was assessedusing tasks probing immediate free recall and recognition memory ofintentionally encoded words. The neuroanatomical correlates ofinterest were regional gray matter (GM) volume and white matter(WM) microstructure as measured by fractional anisotropy (FA) usingstructural magnetic resonance imaging (MRI) methods.Study I mainly confirmed the expected curvilinear relationshipbetween valence and arousal ratings, but with distinct relationships fornegative and positive words. Effects of age and gender on the ratingswere weak, but statistically significant, and mostly in line withexpectations. The ratings correlated for the most part adequately withratings in other databases. Because the valence factor explaineddecidedly more of the variance in the ratings, the findings indicate thatlanguage- and culture-related aspects take precedence overdemographic characteristics when affective properties of words arerated.

Study II examined the role of arousal in producing the age-relatedvalence-specific preferences in memory in a group of young adults anda group of middle-aged and older adults. Contrary to predictions, nostatistically significant age differences emerged concerning valenceand/or arousal effects on immediate free recall and recognition memoryaccuracy or response bias. Thus, no age-related positivity effect wasfound, only lower overall memory accuracy in the older adults. Valenceand arousal exhibited distinct, but mostly interacting effects on thesemeasures across the age groups.

Studies III and IV investigated the regional GM volumetric and WMmicrostructural correlates of memory in middle-aged and older adults.The behavioral results showed an unexpected positivity bias and EEMin recognition memory (Studies III-IV), but no effect of emotion onimmediate free recall (Study III). In Study III, the region-of-interest(ROI) analyses using amygdalar and hippocampal volumes yielded noassociations with either immediate free recall or recognition memory.The whole brain analyses unexpectedly showed that better immediatefree recall of negative words was associated with less regional GMvolume in dorsomedial and left dorsolateral prefrontal cortex (PFC).Surprisingly, the significant positive correlation between local GMvolume and immediate free recall of positive words was located to thecerebellum, and the negative correlation for recognition memory ofpositive words to primary visual cortex. The findings suggest that theneural areas subserving memory for emotion-laden informationinvolve posterior brain areas, including the cerebellum, and thatcognitive control functions may constitute the driving mechanism for memory for emotion-laden information. In Study IV, no statisticallysignificant associations between FA and recognition memory ofnegative or neutral words were found. Negative associations betweenrecognition memory of positive words and FA were unexpectedlyfound in several left-hemisphere projection, association, andcommissural tracts. This likely reflects the complex interrelationshipsbetween the positivity bias in memory, structural WM integrity, andcompensatory brain mechanisms in older age.

Study II together with the behavioral results in Studies III-IV revealedthat it is important to consider the contributions of both arousal andvalence to emotional memory. In Studies I-II, the interrelatednessbetween these affective dimensions took different forms in differentconditions, indicating that their relationship is variable by nature. Theneuroanatomical Studies III-IV produced several novel findings. Theunexpected localizations and directions of the correlations in thesestudies indicate that the structure-function relationships for memoryfor emotion-laden stimuli hold unique qualities in healthy aging,suggesting that these neuroanatomical correlates may be distinct forhealthy aging vs. neuropathological conditions in older age. On ageneral level, the counterintuitive directionality of some of the resultshighlighted the complexity of the relationships between size/integrityof brain structure, functional efficiency, and behavioral outcomes. Theresults of these data-driven studies were more in line with theCognitive Control Model/Socioemotional Selectivity Theory(CCM/SST) account of the positivity effect than the deficit-basedDynamic Integration Theory (DIT) or Aging Brain Model (ABM)accounts. However, no young adult group was included here.Furthermore, the behavioral study II, which did include young adults,offered scant evidence to support either theoretical framework, andshowed no evidence for the age-related positivity effect in memory. Topinpoint the mechanisms driving these results, further studies usingintegrative frameworks are warranted. These should collect multileveland multidimensional data to analyze and create models that permitassessing distinct contributions of various processes as well as theirinterrelationships. Also, using more ecologically valid stimuli, such asvirtual reality systems, would advance knowledge on emotionalmemory.