Science Briefs

What's So Special About a Longitudinal Study

By Merry Bullock, PhD
Ask any developmental psychologist what he or she studies, and the answer is likely to begin with 'the development of...' Then ask how he or she studies this, and the answer is likely to include 'by comparing across ages.' This has been the paradox faced by developmental researchers: Although the goal is to study human change throughout the life span, the reality is that we mostly infer it by comparing people at different points along the life span (e.g., infants vs. children vs. adults). This methodological choice has implications for the kinds of data we collect, the conclusions we draw, and the theories we build. In particular, because developmental change is inferred from mean age-group differences, we tend to ignore individual differences in developmental pathways both in data analyses and theoretical considerations. This is based, in part, on an underlying but untested assumption that most developmental change is universal in type, direction, and endpoint.

There is a way, of course, to ask which changes are universal, whether there is a generalized developmental function, and whether individual differences represent simple variability or meaningful alternative developmental pathways. This is to follow a sample of individuals as they grow older. Although most would agree that such longitudinal studies provide the ideal methodology for studying change and developmental pathways, they are the exception rather than the rule. There are many reasons for this, mostly logistic and pragmatic: Following the same individuals over long time periods requires extensive resources; designing measures that remain valid across different ages is difficult; and maintaining interest and commitment to a single project over the span required by longitudinal work is chancy for both researchers and subjects.

Longitudinal Studies: Are They Worth It?

Given the logistic, pragmatic, and financial burdens of longitudinal studies, one might challenge the blanket assertion that they are the ideal developmental methodology and ask 'are they worth it?' That is, do they provide information beyond inferences from age group differences, and is this information important? I will address this question by describing a few selected results from just one longitudinal study and will suggest that these data can provide important information and serve a valuable heuristic function.

The study I will describe is the Munich Longitudinal Study on the Genesis of Individual Competencies (LOGIC). It was funded by the Max Planck Society and was directed by Prof. Dr. Franz E. Weinert, a Director of the Max Planck Institute for Psychological Research, in Munich. The participants included about 200 children who were 4 years old at the beginning of the study, and 13 years old at the end. Data collection was intensive. Information from about 6 hours of testing was available for each child for each of the 9 years of the study,and about 60% of the children were seen several times a year in their schools in a companion school study.

The focus of the study was on individual differences in developmental pathways across a variety of cognitive, social, and personality areas. It is impossible to convey a complete picture of the broad scope or results of this study, which was the collective effort of a dozen researchers (of which I was one) and more than 20 research assistants. Rather, I will highlight examples (and responsible researchers) of how data from this multifaceted, longitudinal design contribute to a richer empirical and theoretical understanding of development.

Specification of Developmental Models: Memory, Morality, Shyness, and Cross-Domain Effects

Memory (Sodian & Schneider)

The well-established fact that recall memory performance, memory strategy use, and strategy efficiency each improve gradually with age has led to the conclusion that what drives memory development is strategy discovery and use. It was possible to ask whether this conclusion was supported at the individual level in the LOGIC study, because children were regularly (at 4, 6, 8, 10, and 12 years) given a sort-recall memory task. This task assessed memory performance and strategy use.

Age-group comparisons replicated findings in the literature. Mean memory strategy use increased linearly with age, as did mean strategy effectiveness. However, this gradual acquisition was not shown at the individual level. Indeed, the researchers reported that 'the group data on strategy acquisition that have been reported in cross-sectional studies massively obscure the individual developmental paths.' Individual differences were not simply differences in rate. Although children discovered strategies at different times, development, once a strategy was discovered, was far from uniform. In fact, there were more discontinuities (strategy discovery, loss, then subsequent rediscovery, leaps, and drops) than continuities (only 8% showed gradual increases). Of course, such discontinuities are only interesting to the extent that they predict something about later developmental outcomes. There was some suggestion that children who used strategies and understood their function were more likely to maintain them than children who used them mindlessly. This provided partial support for the hypothesis that strategy acquisition per se is not related to recall performance unless mediated by conceptual understanding of the function of deliberate strategic behaviors. The longitudinal data also pinpointed the essential transition point when strategy development became related to performance development (around 8 years), suggesting the period where subsequent, more experimental studies should focus.

Morality (Nunner-Winkler)

Children's moral judgments (knowledge of what is right to do, measured by rule judgments and justifications), moral motivation (desire to do what is right, measured by emotion attributions to moral transgressors), and moral behavior (refraining from cheating, being fair) were assessed by independent measures applied regularly between the ages of 4 and 12 years. Interestingly, moral knowledge and moral motivation did not develop in tandem. Although even the youngest children had moral knowledge, moral motivation developed more slowly. This development was by no means gradual, uniform, or universal. As with memory strategy development, there were leaps, dips, and drops. In addition, moral motivation predicted moral behavior, at least at younger ages.

A comparison of different pathways in the growth of moral motivation showed that the level of motivation (low, medium, high) at younger ages (6-7 years) did not predict moral motivation at ages 8-11. Motivation at ages 8-9 predicted motivation later only for those children who were very low or high. In addition, although all children justified moral rules in terms of both moral strictures and empathic concerns, an eary preference for moral strictures predicted stronger later moral motivation. The role of variable pathways in the development of moral motivation and the relation of moral motivation to behavior provide new directions for research, theory, and ethical training.

Shyness (Asendorpf)

A central issue in social-personality development is the extent to which personality characteristics are stable or variable across time and place. Longitudinal analyses of social inhibition (or shyness) suggested some answers. Individual differences in shyness were assessed in interactions with adult strangers, peer strangers, and familiar peers, both in the laboratory and in the classroom. Because all children in the study entered two socialization situations (the preschool at the beginning of the study and then school part way through), it was possible to look at changes in shyness as peer relations developed in different social contexts. The major findings suggested some of the mechanisms that may underlie the development of individual differences in shyness.

Shyness with strangers (ascribed to fear of the unfamiliar) was stable over time; shyness in the classroom (ascribed to fear of peer rejection) was not stable over time; and a moderate consistency between shyness with strangers and shyness in the classroom decreased with increasing group familiarity. These patterns, replicated across preschool and school contexts, were hypothesized to be examples of a much more general scheme: 'Shyness with strangers is a special instance of a relationship-unspecific trait, and shyness due to social-evaluative concerns is a special instance of a relationship-specific individual attribute.' Because the LOGIC sample was sufficiently large, it was possible to test some of the implications of this model. For example, extreme group analyses showed that peer rejection seemed to lead to shyness in familiar settings and that shyness in familiar settings (but not stranger shyness) was related to lower later social self-esteem.

Cross-domain effects

The heuristic value of the LOGIC study extends beyond contributions to specific content domains. The broad range of contents addressed also allowed a number of post hoc cross-domain analyses. A sample of these includes the following: individual differences in mathematical competencies predicted not only later quantitative skills, but more global and general skills (later IQ level was predicted better by earlier math skill than by early IQ level); extreme groups characterized by high inhibition or high aggression differed from controls in autobiographical memory; moral behavior was predicted by both moral and personality variables; some individual differences in scientific reasoning skills were tied to general representational and working memory abilities.


Although longitudinal studies are not experimental, they can be used to test conclusions from experimental and cross-sectional designs, to generate a rich set of hypotheses, and to provide important pragmatic applications. They also serve the field in another way. The large data sets from longitudinal studies are generally publicly available to other scholars for secondary analyses. The LOGIC study is no exception. Data from the study, as well as supporting documentation, will be freely available so that the study and data will acquire a developmental life and utility of their own.

The project leaders over the years of the study included Drs. Jens Asendorpf, Merry Bullock, Andreas Helmke, Monika Knopf, Julius Kuhl, Gertrud Nunner-Winkler, Wolfgang Schneider, Elsbeth Stern, Gerhard Strube, Angelika Weber, along with the following collaborators: Drs. Marcel van Aken, Marcus Hasselhorn, Christina Muekusch, Jan Naeslund, Josef Perner, Beate Sodian, Michael Waldmann, Heinz Wimmer, Albert Ziegler.

APA Home Page . Search . Site Map