Both measures of verbal fluency have been shown to be sensitive to brain damage (Butters, Granholm, Salmon, Grant, & Wolfe, 1987; Janowsky, Shimamura, Kritchevsky, & Squire, 1989; Stuss et al., 1998). Differential performance on measures of letter versus category fluency has also been demonstrated to be of diagnostic utility in aging-related disorders, such as Alzheimer’s disease (AD; Barr & Brandt, 1996; Hodges, Salmon, & Butters, 1992; Ober, Dronkers, Koss, Delis, & Friedland, 1986; Rosen, 1980). In AD, category fluency is particularly affected, whereas letter fluency remains relatively intact (Monsch et al., 1992). These observations are consistent with findings that temporal systems mediate category fluency performance (Pihlajamaki et al., 2000) and frontal systems underlie letter fluency performance (e.g., Abrahams et al., 2003). Despite the clinical implications for assessment of category and letter verbal fluency, there has been relatively little investigation of changes in both letter and category fluency across the adult lifespan in the same sample.

Based on existing studies examining the impact of age on verbal fluency, it appears that category but not letter fluency is typically poorer in elderly as compared to younger adults (Bolla, Gray, Resnick, Galante, & Kawas, 1998; Crossley, D’Arcy, & Rawson, 1997; Kozora & Cullum, 1995; Tombaugh, Kozak, & Rees, 1999; Tomer & Levin, 1993). This pattern of poorer performance on tests of category fluency is reproduced in AD, but with a greater effect (Monsch et al., 1992). Several reports show a generally negative association between age and category fluency measures in adults (Borod, Goodglass, & Kaplan, 1980; Whelihan & Lesher, 1985). Although there is a larger literature on letter fluency in normal adults, the findings concerning age-related change are less consistent. A recent meta-analysis of the FAS task in normal, healthy adults across the lifespan demonstrated a fairly robust age-associated decline in the task (Loonstra, Tarlow, & Sellers, 2001). However, not all authors have reported this pattern (Perlmuter & Tun, 1987).

There is general agreement across studies that verbal fluency is impacted by education level. Most studies report a positive association between number of years of formal education and performance on tasks of letter and category verbal fluency (Benton, Hamsher, & Sivan, 1983; Crossley et al., 1997; Tombaugh et al., 1999; Yeudall, Fromm, Reddon, & Stefanyk, 1986). Relationships between IQ and fluency also feasibly reflect the impact of education. Bolla, Lindgren, Bonaccorsy, and Bleecker (1990) used a regression model, including years of education, an estimate of verbal IQ, age, and sex, to predict performance on tasks of verbal fluency. While they found that the estimate of IQ but not education significantly predicted performance, this finding is most likely due to the high association between IQ and education level (i.e., collinearity between the two factors).

While there is fairly consistent evidence that education impacts fluency output, it is less clear whether fluency performance across the lifespan varies as a function of education. Education may be protective against age-associated decline in cognitive performance, as suggested by the concept of cognitive reserve (Stern, 2002; Stern, 2003). Using regression-based approaches, normative data studies (e.g., Tombaugh et al., 1999) have shown that both age and education account for a significant amount of variance in fluency performance. In one study, Gladsjo and colleagues (1999) found a significant interaction between age and number of years of education for FAS but not Animal Naming performance. In contrast, Bolla and colleagues (1990) did not find a significant influence of education on fluency performance after correcting for the effects of age, sex, and IQ. Similarly, Kempler, Teng, Dick, Taussig, and Davis (1998) did not find a significant interaction between age and education on fluency performance.

The impact of sex differences on fluency performance, and potential changes in these differences across the lifespan, has not been clearly elucidated. Some studies have reported that women perform better on tasks of letter verbal fluency (Bolla et al., 1990; Crossley et al., 1997), while others (Borod et al., 1980; Gladsjo et al., 1999; Yeudall et al., 1986) report no differences between men and women. There is little evidence in the literature for sex differences on tasks of category fluency. In terms of the stability of sex differences across the lifespan, Crossley and colleagues (1997) did not find a significant interaction between sex and age on either letter or category fluency. However, Capitani, Laiacona, and Basso (1998) demonstrated that there was a greater age-associated decline on a task of letter fluency in men.

Electroencephalography (EEG) is a method that is sensitive to cerebral functioning and can provide important insights into neurophysiological correlates of age-related cognitive decline. It has been used to differentiate patients with AD from normal comparison subjects (Duffy, McAnulty, & Albert, 1995), but the pattern of physiological change and its relationship to cognition is less understood across the normal, healthy adult lifespan. Age-associated increases in slow wave activity, characterized by an increase in delta and theta band measurements, have been reported (for review, see Dustman, Shearer, & Emmerson, 1993). Studies of clinical populations have shown theta increases in AD patients as compared to age-matched normal controls (Duffy et al., 1995) and individuals with mild cognitive impairment (Huang et al., 2000), as well as in AD patients who were impaired on a task of verbal fluency compared to those who were not (Le Roc’h, Rancurel, Poitrenaud, Bourgin, & Sebban, 1993).

In healthy individuals, theta wave activity has been implicated in a number of other language-based cognitive tasks (e.g., Weiss & Rappelsberger, 1998). For example, increases in theta rhythm have been reported during processing of words (Bastiaansen, van Berkum, & Hagoort, 2002a) and theta power has been shown to increase during the perception of syntactic violations (Bastiaansen, van Berkum, & Hagoort, 2002b). In children with dyslexia, theta band power was shown to be lower than in matched controls (Klimesch et al., 2001). Theta wave activity has been linked to both verbal and non-verbal memory (Kahana, Seelig, & Madsen, 2001) and may be particularly associated with retrieval (Ward, 2003), which is required for successful performance on tasks of verbal fluency. Further, Uclés, Almárcegui, Lorente, Romero, and Marco (1997) demonstrated a significant improvement in theta power following carotid endarterectomy in patients with carotid artery stenosis, which correlated significantly with an improvement in category verbal fluency. In light of the evidence linking theta band activity to aging and to language-based cognitive tasks, the current study focused on theta power as a potential neurophysiological moderator of age-related change on tasks of verbal fluency. That is, resting theta may be reflective of underlying neural networks involved in age-related decline in fluency performance and a purpose of our study was to explore these potential correlates of theta power. The focal regions of interest were the frontal and temporal lobes given that these areas have also been implicated in tasks of verbal fluency (Phelps, Hyder, Blamire, & Shulman, 1997; Pihlajamaki et al., 2000).

Our finding of a greater age-associated decline in category fluency performance replicates other studies that have found a similar effect (e.g., Bolla et al., 1998). Age-associated patterns on these two tasks may point to underlying brain regions that are differentially affected in normal aging. Evidence from functional neuroimaging studies suggests that inferior and mid-frontal regions mediate letter fluency tasks (Abrahams et al., 2003; Gaillard et al., 2003; Phelps et al., 1997), whereas both frontal and temporal structures underlie tasks of category fluency (Pihlajamaki et al., 2000). Similarly, Moskovitch (1995) has proposed a neuropsychological model in which letter fluency is mediated by the frontal lobes and category fluency is mediated by temporal lobe structures. In AD, a disorder that impacts semantic networks, category fluency is more severely affected than letter fluency (Monsch et al., 1992). Thus, like in AD, frontal–temporal semantic networks may be vulnerable to the effects of the normal aging process. Neuroimaging studies examining category and letter fluency tasks in the same subjects across the lifespan would help clarify this issue. Even in the absence of correlative neuroimaging data, however, the results from this study suggest that the two fluency tasks are mediated by different neuroanatomical substrates.

Although the behavioral findings of varying category and letter fluency performance across the lifespan suggest differential neurobiologic substrates, this was not captured by EEG theta band power. Our finding that theta power decreased as a function of age was inconsistent with what has been reported in the literature, which has generally suggested an increase in slow wave activity (e.g., theta) with advancing age (Dustman et al., 1993). Discrepancies with the extant literature could be due to several factors. The participants in the current study were very well screened and the elderly participants did not evidence any symptoms of a neurodegenerative or medical condition that might impact cognitive or brain functioning. Thus, a subject-selection bias could have been operative; that is, older subjects in the study could have been “supranormal.” The findings of age-associated decline in verbal fluency speak against this possibility. Furthermore, the relationships between theta power and fluency were in the expected positive direction, which is consistent with studies that have demonstrated depressed theta frequency bands in patient groups with poorer performance on fluency tasks (Le Roc’h et al., 1993). It is important to note, however, that EEG recordings in the current study were not conducted during task performance and, therefore, may have been vulnerable to many factors (e.g., drowsiness). The sample size and controlled experimental methodology in the current study would argue against the possibility of Type I errors. As there have been limited studies that have examined the relationship between resting EEG measures and cognitive task performance, discrepant findings are most likely attributable to methodological differences among experiments.

Resting theta power was moderately related to category but not letter fluency task performance. The finding is similar to that of Hoptman and Davidson (1998) who did not find a relationship between theta power and performance on a written test of letter fluency. While theta has been implicated in a number cognitive processes (Kahana et al., 2001), our findings suggest some specificity in cognitive processes associated with it. Theta may be particularly linked to both memory encoding and retrieval (Ward, 2003). The significant association between theta power and category, but not letter, fluency performance in our study is consistent with this idea, given the greater load of associative information retrieval required to complete the task successfully.

Like in other investigations (e.g., Crossley et al., 1997), a significant association between fluency tasks and education level was found. Education is a common surrogate measure of cognitive reserve (Stern, 2002; Stern, 2003). Consistent with this notion, we hypothesized that greater education would be somewhat protective against the effect of age. However, education level did not appear to impact rate of fluency decline across the lifespan. Despite the fact that there was a range of educational achievement, the sample was generally well educated, and this may have impacted the findings. That is, inclusion of individuals with much lower education may have yielded significant interactions involving education. The findings, nonetheless, suggest that education may be less protective against normal aging effects on tasks of fluency.

There was a significant decline in education level across the age groups. This finding most likely represents a cohort bias secondary to random sampling from the population; older individuals, in general, may be less likely to have pursued advanced levels of education. It is important to note that despite the fact that education level was related to performance on both tasks of verbal fluency, the pattern of age-related decline was not due to this relationship. That is, education level was similarly related to FAS performance as it was to Animal Naming performance. That elderly individuals tend to have lower levels of education has important clinical implications. Although education and age do not interact with fluency performance, both demographic variables contribute independently to performance. Thus, both factors should be taken into consideration during clinical evaluation.

Whether there are sex differences in fluency performance remains somewhat elusive. We did not find the female advantage that has been reported by other investigators (Bolla et al., 1990; Capitani et al., 1998), though it is important to note that the sex distribution was not consistent across the age groups in our sample. Given the size of the sample in the current study, it is unlikely that our finding was due to Type II statistical error, even if the reported sex effect is of small magnitude, as suggested by Capitani and colleagues (1998). Furthermore, sex did not modify rate of age-related decline for either the category or letter task, as has been reported by some authors (Capitani et al., 1998). Differential performance across the lifespan as a function of sex may be detectable with more qualitative analyses of verbal fluency, as suggested by Tabert and colleagues (2001), who demonstrated sex differences in the accuracy of verbal output.