The study examined age group effects and sex differences in performance in a large population-based sample of youths using a brief yet comprehensive set of tests that have been linked to circumscribed brain systems. This approach allowed the examination of hypotheses on age related effects on performance and their modulation by sex. Several findings stand out, some confirming hypotheses based on previous observations and some novel. With regard to age group effects, performance improved across domains, as expected, but there was substantial variability among domains in the extent and rate of improvement. As hypothesized, based on evidence for greatest maturational changes in frontal systems, Executive domains showed the largest differences between the youngest and oldest groups, with very large effect sizes (approaching or exceeding 1 standard deviation) for both accuracy and speed of attention and working memory. Both domains have been linked to prefrontal lobe function, a link demonstrated specifically for the tests used in this study (Gur et al., 2007
; Ragland et al., 2002
), where maturation is delayed relative to other brain regions (e.g., Giedd et al., 1999
; Gogtay et al., 2004
; Matsuzawa et al., 2001
; Perrin et al., 2009
; Pfefferbaum et al., 1994
). On the other hand, abstraction and mental flexibility, also a frontal lobe domain, showed only a moderate effect size for improvement with age in accuracy and smaller effect size for speed. This domain could relate to aspects of frontal lobe functioning that mature earlier or can be supported by early maturing other brain systems.
Age related improvement in memory was considerably less pronounced than that for executive functions and was seen mostly in speed of response time rather than accuracy. The exception was face memory, which showed large effect sizes for both accuracy and speed. For verbal memory accuracy the effect size was small and spatial memory accuracy showed no age group related improvement. Perhaps the developmental epoch sampled (based on mean age=14.8) has already missed the years of steeper developmental gains for memory, which is consistent with evidence for relatively early maturation of temporal lobe structures (Matsuzawa et al., 2001
). Alternatively, ceiling effects could be a concern, since we lowered the level of word readability to accommodate the youngest age group. It is unlikely, however, to explain entirely the low correlation with age because we did find significant sex differences, and in all other domains age effects vastly overshadowed sex differences. Furthermore, the spatial memory test was as difficult as the face memory test, and showed no correlations with age. Thus, a conclusion that age effects are much less pronounced for accuracy of episodic memory than for the other domains seems justified.
Complex Cognition and Social Cognition domains showed comparable improvement that was more pronounced than for memory but not quite as pronounced as for Executive-control domains. For Complex Cognition, all domains showed improvement with age for accuracy, with very large effect sizes (all exceeding 1 standard deviation unit). For speed, the effect size was large only for language-mediated reasoning. It was small for spatial cognition, and for nonverbal (matrix) reasoning it was opposite in direction - response time for correct answers actually increased with age. Because items on the matrix-reasoning test vary greatly and appear in order of difficulty, this effect is likely due to success with more difficult items that take longer to solve. For Social Cognition, all age group related effect sizes were large for accuracy and moderate to large for speed. These results accord with evidence that heteromodal cortical association areas mature later than temporal cortex, but not quite as late as frontal lobe cortex (Gogtay et al., 2004
As hypothesized, speed itself also improved with age even in the absence of a cognitive task. Indeed, improved speed was more pronounced for a simple motor task than for the task requiring sensorimotor integration. While there is evidence that the sensorimotor strip and cerebellum are among the earliest to mature (Gogtay et al., 2004
; Yakovlev & Lecours, 1967
), the present results suggest that performance continues to improve into early adulthood. Perhaps the continued physical growth interacts with brain maturation in affecting motor performance.
Sex differences were apparent both in overall performance and in age group related variation. However, these effect sizes were small compared to age group effects. The hypothesized differences favoring females for memory and social cognition tests and males on spatial and motor tests were strongly supported. Females performed more accurately and faster for the verbal memory test and more accurately for face memory, although they were less accurate on spatial memory, as has been reported in earlier studies (Saykin et al., 1995
). Females were both more accurate and faster on all social cognition tests. These effects are in line with earlier reports of better performance in females for emotion processing tasks (Gur et al., 2010
; Williams et al., 2008
), but extend them to other social cognition measures and across a wide developmental epoch. On the other hand, males were more accurate on the spatial test and were faster on both sensorimotor and motor speed (Coleman et al., 1997
; Gur et al., 1999
; Halpern et al., 2007
; Moreno-Briseño, Díaz, Campos-Romo, & Fernandez-Ruiz, 2010
; Thomas & French, 1985
). Some sex differences were unexpected. Thus, males were more accurate in abstraction and mental flexibility (a very small effect size) and females were more accurate and slower for attention and slower for working memory. The small size of the effects may explain why they have not been reported in smaller samples. Poorer accuracy in males for attention is consistent with the higher incidence of attention deficit disorder in males (Ramtekkar, Reiersen, Todorov, & Todd, 2010
). Whether this difference can be explained by more males with attention deficit disorder in the current sample will be clarified when results of the clinical assessments are incorporated.
There were few age group × sex interactions. These interactions were noted only for spatial memory accuracy and speed, nonverbal reasoning accuracy and speed and all social cognition tests on speed. Motor speed also showed a significant interaction. All these interactions indicated that sex differences became more pronounced in the age groups following mid-adolescence. Across all domains, except for memory, females reached plateau before males. This finding accords with physical (Hills & Byrne, 2010
), behavioral (Keulers, Evers, Stiers, & Jollies, 2010
; Greenstein, Blachstein, & Vakil, 2010
; Review in Yurgelun-Todd, D., 2007
) and neuroimaging (Bramen et al., 2010; Tieneier, et al., 2010) data indicating earlier maturation in females. The exception in our study is for memory, where males peaked by age 18–19 whereas females continued to improve in word and face memory into the 20–21 age group. Age group × sex interactions for complex and social cognition were seen in speed, where females continued to improve while males reached a plateau in mid adolescence and then showed decline. While we are unaware of earlier studies where both social cognition and a broad range of other neurobehavioral domains have been examined across this age range for both accuracy and speed, our findings generally comport with studies examining developmental sex differences in comparable domains (e.g., Reynolds, Keith, Ridley, & Patel, 2008
Beyond the specific findings, the present results indicate the feasibility of administering a brief yet comprehensive computerized neuropsychological battery of identical tests to a large population-based sample of children, adolescents and young adults and obtaining informative data. The testing yielded a large proportion of validated data of high quality with information pertinent to major behavioral domains that can be linked to brain function and genomics. Making this link requires adequate information on developmental effects because, as our results demonstrate, these effects are substantial and modulated by sex. Thus, while most genomic variation is fixed, the present results underscore that neurobehavioral phenotypes require demographic information to be interpretable.
Limitations and Future Directions
The study has several potential limitations. While the sample is large and demographically diverse, we have not excluded individuals with some medical conditions that may affect performance and skew some of the data. An issue can be raised concerning the extent to which the sample represents the general population. The sample was not intentionally enriched for any specific disorders. The incentive for participation was that the children gave blood for genomic analysis when they saw their pediatrician for a well-child visit or any other reason, and agreed to be re-contacted for participation in other studies. Thus, it is perhaps not as representative as a census based random sample, but more representative than a convenience sample of responders to advertisement for psychological experiments. In this regard, it is reassuring that the average WRAT score in our sample is identical to that obtained in the normative sample (Mean ~100, SD ~15). When the final sample is obtained and the results integrated with electronic medical records and updated with clinical evaluations we performed, we will have sufficient power to examine the effects of potential disease traits. Notably, the integration of neurobehavioral data with the neuropsychiatric assessment will also enable us to evaluate neurobehavioral profiles associated with psychiatric symptoms. The availability of only part of the final planned sample also presents unequal sample sizes for males and females across the age groups. Since the standardization was based on the entire sample, conceivably some effects in specific age groups could be imprecisely estimated. However, standardization across the sample is the most straightforward and easily interpretable approach and the large sample size should minimize systematic distortion of parameter estimates. Another limitation of the study is the modest number of tests administered in each neurobehavioral domain. This limitation was imposed by feasibility given the scope of the study. Obviously, the tests do not reflect the depth and complexity of the domains sampled. For example, there is more to episodic memory than recognition of words, faces and shapes. We are also limited in the extent to which social cognition is measured and the battery does not include any auditory measures such as prosody, or other aspects of social cognition. Future longitudinal follow-up could expand to other measures pertinent to pursuing specific hypotheses in subsamples. Finally, the present study has examined a limited number of measures within each of the tests. There are alternative indices that can be derived for most tests and that can yield interesting information on cognitive strategies. For example, signal detection parameters can be applied to the attention and memory measures. Here we present data on a broad range of domains, and to contain Type I error we have limited our examination to one measure for each domain. However, the computerized format allows effective evaluations of multiple indices and relations with response time in studies focusing on individual domains.
Notwithstanding these limitations, the present large-scale study demonstrates the feasibility of obtaining reliable neuropsychological data that offer information on age and sex differences. These data can be integrated with phenotypic measures in specific behavioral domains associated with developmental disorders, with neuroimaging data on brain structure and function and with genomic parameters that can propel findings that bridge between molecular and behavioral processes. Such integration has the potential of unveiling specific genes and gene networks that drive developmental traits as well as genetic variants that may underlie key individual differences, including those related to sex differences and ethnic groups. While still underpowered at this stage, the results from future genomic analyses may contribute to understanding normal and abnormal development, with implications for intervention approaches.