• No results found

3 General discussion

3.1 Summary and discussion of the results

3.1.2 Verbal learning as an alternative memory measure

The second research question addressed the relation between memory self-reports and memory performance with the focus on learning as, possibly, a more informative memory

measure. The results presented herein bear only indirectly on the issue of low correspondence between memory self-reports and memory measures as they were mainly used to demonstrate the methodological approach to formalize and relate learning to other variables of interest.

However, they suggest one line of research that might be pursued to scratch the surface of the intricate and long-standing issue of the correspondence between subjective and objective memory reports. In Chapter 1.4.2, I argued that data from learning experiments would deliver a closer relation to memory self-reports elicited in memory questionnaires. Hence, other than the first research question, here, the focus was on the objective memory measure. There were two reasons for focusing on the memory measure: First, the approach to modify self-report questionnaires with respect to behavioral specificity (cf. Hertzog et al., 2000) did not seem to be a very compelling alternative because the increase in the correlation between subjective and objective memory measure was only marginal and, at the same time, it was outweighed by the loss in generalizability with respect to the subjective measure. Second, in expanding the memory measure from a single trial recall task to a learning experiment with several study and test trials, yielded a more broad and differentiated view on memory. Furthermore, in Chapter 1.4.2, I argued that memory is probably apprehend by lay persons in a more generalized way, that is, in a naturalistic situation memory is not reduced to a single recall event but the whole process of information acquisition and recall is considered to represent memory. Consequently, if subjects are asked to give reports about their memory performance the behavioral correlate which is measured in a memory test must coincide to the largest possible amount with the subjective measure. In this respect, learning experiments seem to be better suited than one-trial memory tests.

The approach presented in Chapters 1.4.2.1 and 1.4.2.2 was demonstrated by means of data from 364 persons from the Zurich Longitudinal Study on Cognitive Aging (ZULU;

Zimprich et al., in revision). Part of the testing protocol was the Metamemory in Adulthood (MIA) questionnaire, three measures assessing processing speed, and a verbal learning measure that comprised five study and recall cycles. The study was mainly geared to demonstrate the fruitfulness of the individually-centered approach on verbal learning in old age. In a first step the best representation of verbal learning was assessed by testing three different nonlinear functions, namely, the quadratic, the exponential, and the hyperbolic growth curve. The latter turned out to fit the data best and the linkages of the learning parameters initial performance (β), potential maximum performance (α), and rate of learning

(γ) to age and processing speed were modelled. The inclusion of the latter cognitive ability was motivated by the fact that processing speed represents a major explanatory variable of cognitive aging (Salthouse, 1991, 1996). Subsequently, we included memory as an outcome variable, that is, verbal learning parameters (β, α, γ) were used as predictor variables of memory performance in old age. In light of the duality of learning and memory phenomena, memory performance represents an obvious and, moreover, extensively studied outcome variable in cognitive aging research (Craik, 1977; Hultsch et al., 1998; Kausler, 1994). In the complete model, the verbal learning parameters initial performance level (β), potential maximum performance (α), and rate of learning (γ) thus acted as mediating variables between processing speed and memory.

The variations in the three learning parameters indicating individual differences were highest in the learning rate (γ) which suggests that older persons tend to show more pronounced individual differences from each other in the rate of acquisition than in initial performance or potential maximum performance. Further, the learning rate seemed to be mostly affected by age, whereas initial performance was less and potential maximum performance was almost not affect by age. That is, older participants needed more trials to reach their maximum performance, which, in turn, was not affected by the participants’ age.

This supports the idea formulated in Chapter 1.4.2 where I argued that the recall performance might remain stable, but the cognitive effort to maintain a given level increases in older adults. The influence of age, however, was in part mediated by processing speed, which is in line with Salthouse’s (1996) processing speed theory. Eventually, memory, measured in single trial recall tests, was added as an outcome variable of the three learning parameters.

Interestingly the learning rate and the potential maximum performance exerted the largest effect on the memory measures, which implies that these two parameters are determining and inherent factors in memory.

In sum, the study presented in Chapter 2.2 demonstrated the benefit one gains from administering several learning and recall trials over the commonly used single trial experiments. That is, at least two more parameters related to memory performance are estimable which appear to be memory-inherent. Further, due to increasing individual differences the parameters can be mapped more exactly and unsystematic influences on performance become smaller which should increase reliability of the measure. The question whether the relatedness between self-referent memory beliefs and learning parameters are

higher compared to single trial memory measures, however, has not directly been tested and can not be answered at this point.