Introduction
It is believed that the function of language per se is little affected by normal aging (Wingfield & Stine-Morrow, Reference Wingfield and Stine-Morrow2000). Indeed, some aspects of language are among the abilities that are best preserved and might even progress into higher levels of expertise during late adulthood (Kemper, Rash, Kynette, & Norman, Reference Kemper, Rash, Kynette and Norman1990; Verhaeghen, Reference Verhaeghen2003). However, whenever quick processing and mental manipulation of verbal material is required healthy elderly subjects show decidedly a decline in performance (Connor, Spiro, Obler, & Albert, Reference Connor, Spiro, Obler and Albert2004; Goodglass, Reference Goodglass1980; Mortensen, Meyer, & Humphreys, Reference Mortensen, Meyer and Humphreys2006). Decline in verbal performance is often associated with age-related changes in working memory, inhibitory mechanisms and difficulties in lexical access and processing (Kemper & Mitzner, Reference Kemper and Mitzner2001; Robert & Mathey, Reference Robert and Mathey2007).
Another important aspect contributing to age-related differences in verbal execution is the slowing of response (Brébion, Reference Brébion2001), which comprises declines on speed of cognitive and motor functions (Birren & Fisher, Reference Birren and Fisher1992; Wilson, Bienias, Evans, & Bennett, Reference Wilson, Bienias, Evans and Bennett2004). Different studies have demonstrated a reduction on speed of speech production in normal aging, such as decline on speech rate (Borden, Harris, & Raphael, Reference Borden, Harris and Raphael2003; Fozo & Watson, Reference Fozo and Watson1998), naming (Feyereisen, Demaeght, & Samson, Reference Feyereisen, Demaeght and Samson1998; Hodgson & Ellis, Reference Hodgson and Ellis1998), verbal fluency (Rodríguez-Aranda, Waterloo, Sparr, & Sundet, Reference Rodríguez-Aranda, Waterloo, Sparr and Sundet2006), reading rates (Ramig, Reference Ramig1983), and speed of spontaneous speech (Linville, Reference Linville2001b). Even speed of simple pronunciation has been demonstrated to slow down in healthy elderly adults (Balota & Duchek, Reference Balota and Duchek1988; Multhaup, Balota, & Cowan, Reference Multhaup, Balota and Cowan1996).
The reasons behind the slowing on speech production are certainly manifold. However, changes on cognitive and motor processes are proposed as principal determinants (Linville, Reference Linville2001a; Salthouse, Reference Salthouse1993). Because it is still unclear to which extent the slowing on speech generation relies on cognitive decline or on changes in motor functions, the present study aims to address the issue. To reach reasonable conclusions, we believe that assessment of word production needs to be performed on verbal tasks with varying degree of cognitive demands because it is not evident that the slowing of motor mechanisms could be identified in tasks preponderantly cognitive or that the slowing on cognitive processes would be of importance on performance of familiar tasks relying on well learned executions. Therefore, it seems necessary to clarify whether slowing on motor and cognitive processes affects to the same extent word production in the elderly unrelated to type of test.
In sum, the interest of the study is to evaluate the source of age-related decline on oral language production. The issue is of special relevance for neuropsychological assessment of the elderly since many of the standard tests used to evaluate verbal abilities (e.g., verbal fluency tasks) are time limited.
The Present Study
The present study evaluates speed of word production in young adults and healthy elderly individuals in reading, object naming, semantic verbal fluency and phonemic verbal fluency. We considered that degree of difficulty for each verbal task increases from reading to verbal fluency tasks. The rationale for attributing different difficulty level to these tasks is as follows. According to cumulated data in lexical production, reading can be considered to rely on automatic processes (Holmes, Reference Holmes2009; Stine-Morrow, Miller, & Hertzog, Reference Stine-Morrow, Miller and Hertzog2006), meaning by this that performance does not place a high load on attention and can be easily performed by literate adults. Therefore, reading of single words represents the least demanding task. Naming will be considered as the second in difficulty because it is less practiced than reading, there is great variation to represent an object, and it imposes the search of a meaning (Glaser, Reference Glaser1992).
Furthermore, we considered verbal fluency tasks (VFTs) as the most demanding tasks because they rely on higher levels of executive functioning and semantic processes (Bryan & Luszcz, Reference Bryan and Luszcz2000; Capitani, Laiacona, & Barbarotto, Reference Capitani, Laiacona and Barbarotto1999). Nevertheless, to our knowledge there are no empirical data evaluating degree of difficulty between semantic and phonemic fluency in healthy populations, and thus, it is not possible to advance any prediction about which of the two VFTs would be the most demanding one.
Based on the above rationale, we determined age-related differences in speed of word production between young and elderly individuals on the mentioned tasks. Speed of word production was measured as reaction times (RTs) and pronunciation time. According to the literature (Feyereisen et al., Reference Feyereisen, Demaeght and Samson1998), we expected differences on RTs depending on task difficulty with longer RTs for VFTs and shorter for naming and reading. Slower RTs proportional to test difficulty were also expected for the elderly (Tun & Lachman, Reference Tun and Lachman2008; Verhaeghen & Cerella, Reference Verhaeghen and Cerella2002). Regarding pronunciation time we measured outcomes exclusively from naming and reading since subjects were prompted to generate the same words on these tasks as opposed to words produced on VFTs. Following the assumption that speed of word production depends on task difficulty, we expected important group differences on pronunciation during naming than during reading.
Finally, we examined the differential contribution of age-related cognitive decline and age-related psychomotor slowing to speed of verbal performance on each of the selected tests.
To this purpose, we assessed working memory and executive functions since these processes importantly decline in aging and are suggested to contribute to the age-related slowing of information processing (Span, Ridderinkhof, & Van der Molen, Reference Span, Ridderinkhof and Van der Molen2004). Among the numerous tasks assessing these functions, we selected Stroop test and Digits-backwards due to an undoubtedly age-related decline on these tasks and their recurrent use in aging research (Bopp & Verhaeghen, Reference Bopp and Verhaeghen2005; Tun & Lachman, 2008; Verhaeghen & De Meersman, Reference Verhaeghen and De Meersman1998). Furthermore, we chose the Finger Tapping and the Purdue Pegboard test for the assessment of psychomotor functions since they are tools widely used in neuropsychological assessment and they elicit age-related differences on performance (Desrosiers, Hebert, Bravo, & Dutil, Reference Desrosiers, Hebert, Bravo and Dutil1995; Godefroy, Roussel, Despretz, Quaglino, & Boucart, Reference Godefroy, Roussel, Despretz, Quaglino and Boucart2010).
Based on the speed of processing hypothesis (Salthouse, Reference Salthouse1996) we expected that psychomotor performance would importantly explain both RTs and pronunciation times. However, given that RTs represent the series of cognitive and motor functions involved in the processing of information (Schaie & Willis, Reference Schaie and Willis2002), it is a-priori not evident whether cognitive or psychomotor declines would be more determinant for RTs. Because RTs duration relates to task difficulty we hypothesize that the influence of psychomotor decline would be more prominent for RTs on reading and naming than for VFTs. In parallel, we expected that cognitive mechanisms would be significant predictors on RTs for VFTs. Finally, we expected that only psychomotor functions would explain a significant portion of the variance in pronunciation speed since psychomotor execution relies on elaborated cognitive and motor mechanisms which are processes equally essential for word articulation.
Method
Participants
Thirty young adults (14 females, 16 males; age: M = 30.36 years, SE ± 1.04) and 30 healthy elderly over 65 years of age (14 females, 16 males; M = 72.16 years, SE ± 0.73) participated in the study. University students and elderly living independently in the community were the participants. The younger group had an average of 17.4 (SE ± 0.55) years of education while the elderly group had an average of 10.77 (SE ± 0.50). Subjects were recruited from the general community through local advertisements or by means of word of mouth and all were native Norwegian speakers living in the North part of the country. Involvement in the study was voluntary and written consent was signed before testing. Demographic and background information was obtained by means of interview at the beginning of the study including health issues such as visual acuity and hearing status. None of the subjects had a history of drug or alcohol abuse, psychiatric hospitalization, psychopharmacologic treatment or neurological disorders.
Participants were screened for depression with the Beck Depression Inventory (BDI) (Beck, Erbaugh, Ward, Mock, & Mendelsohn, Reference Beck, Ward, Mendelsohn, Mock and Erbaugh1961) according to criteria described in Rodríguez-Aranda (Reference Rodríguez-Aranda2003) and the Mini-Mental State Examination (MMSE; Folstein, Folstein, & McHugh, Reference Folstein, Folstein and McHugh1975) was used to rule out the possibility of dementia (cut-off of 25). None of the recruited subjects was excluded for depression or impaired mental status. The Vocabulary sub-test of the WAIS-R (Wechsler, Reference Wechsler1981) was used to determine lexical level. Health status was self-rated by the participants by asking them whether they consider being in good, medium or bad physical condition. This information together with answers from the interview allowed to rate health status of the participants. The study was approved by the Regional Research Ethics Committee.
Materials
Verbal tasks
Phonemic verbal fluency
An adaptation of the Controlled Oral Word Association Test (Benton, Reference Benton1967) was used. The letters “F’ and “A’ were used in the study following standard requirements. Subjects were asked to quickly produce words starting with the selected letters by excluding proper names and variants of the same word. Repetitions were not allowed and 1 min was allotted for each trial. The target letters were presented on a computer monitor. Before stimuli presentation, subjects were appropriately instructed and asked to perform the task with another letter. The final score was total number of correct words during all trials.
Semantic verbal fluency
The categories “Animals’ and “Fruits’ were presented on a computer monitor. Subjects were asked to generate as fast as possible words belonging to these categories within one minute and repetitions of the same word were not allowed. Again, an example was given and the final score was calculated as for phonemic fluency.
Naming task
Ten pictures (Appendix A) were selected from the Halstead Reitan aphasia test and Benton naming task and presented one by one on a computer monitor. All pictures were digitized to a size of approximately 12 × 12 cm. Subjects were specially asked to rapidly generate one single word per picture.
Reading task
Ten Norwegian words (Appendix A) were presented in the center of a computer monitor, one at a time. The list included single common words varying in length from one to five syllables and presented in a 45-point Courier New font. The list included high frequency words according to the database for the Norwegian language (Heggstad, Reference Heggstad1982). Subjects were instructed to read each word as fast as possible and as soon as it appeared on the screen.
Cognitive tasks
Digits Span Backwards
Subtest from the WAIS-R (Wechsler, Reference Wechsler1981). Number of trials successfully completed is reported. We used this task as a working memory measure.
Stroop test (Golden, Reference Golden1978)
The standard version of the test was used in the Norwegian translation. The interference part of the test was used as a measure for executive functions.
Psychomotor tasks
Finger Tapping (Reitan & Wolfson, Reference Reitan and Wolfson1985)
Subjects are required to tap as rapidly as possible with the index finger on a lever that is attached to a counting device.
Purdue Pegboard (Tiffin, Reference Tiffin1968)
Dexterity task assessing gross and fine motor speed of upper extremities. Gross movements of the arms and hands are required as well as fine movements of fingertips. Subjects have to manipulate small pegs, pins, collars, and washers to perform different assembly tasks.
Procedure
Participants were tested individually at the Department of psychology. An interview served to obtain informed consent and demographic data. The MMSE, BDI and Vocabulary test were applied. Each subject was required to sit in front of a 15-inch computer monitor at a distance of approximately 50 cm. A headset microphone was adjusted to each subject and the subject's speech was recorded with the Computerized Speech Lab equipment (CSL 4500, Kaypentax).
Stimuli for each task were presented randomly with the E-prime computer program. Presentation times were as follows, 1 min for verbal fluency tasks, 4 s for naming and 3 s for reading tasks. The rationale for time presentation for VFT was to respect standardized rules for these tasks while reading and naming time presentations were decided according to task difficulty. Before the evaluation, a pilot study was performed to test the experimental design and based upon the results of this initial study we determined duration for stimuli presentation for reading and naming. In the pilot study, we tested 5, 4, and 3 s on each task. A beep tone was added to the presentation of each stimulus (e.g., “F’+ beep) to allow for a precise measurement of RTs. Oral and written instructions were given to the subjects at the beginning of each task.
Acoustic analyses
Acoustic analyses were performed on the sound waves of the recorded data with the computer program from the CSL 4500 package. Reaction times (RTs) were determined from the end of the beep tone to the onset of the first oral answer. This measurement was obtained on all verbal tasks. RTs for verbal fluency tasks simply represent the time it took subjects to initiate the first word. In total, we obtained 2 RTs for each verbal fluency test and 10 RTs for naming and reading respectively. The average RT was then used for statistical analyses. Pronunciation speed was calculated as the total time used to produce a particular word. The relevant segment of the sound file was selected on the oscillographic display on the computer monitor and then listened to verify the beginning and ending points. Pronunciation speed was not performed on the fluency tests due to the obvious inter-individual variation of generated words by letter and category.
Statistical analyses
We compared background variables, and performance on cognitive and psychomotor measures with independent t tests. To test group differences across tasks, we calculated mean RTs for all tests. Then, all RTs raw scores were transformed into Z-scores by using the overall RT mean and standard deviation that was calculated using all verbal tasks. The calculation included the entire sample. Thus, we evaluated group differences on each task and RTs differences across tasks for each group. Also, pronunciation speed data were converted into Z-scores for comparison of performance across different words generated in different tasks. Finally, we used independent hierarchical regressions to examine the relative contribution of cognitive and psychomotor performance in predicting the log-transformed scores of RTs and pronunciation duration across verbal tasks.
Results
Demographics and screening tests results are presented in Table 1. Data revealed group differences in years of education (t(58) = 8.87; p < .0001); MMSE (t(58) = 3.67; p < .001) and Vocabulary (t(58) = 3.03; p < .01).
Table 1 Demographics, MMSE, and Vocabulary score by group
Note. MMSE = Mini-Mental State Examination. *p < .05, **p < .01, ***p < .001.
Cognitive and psychomotor results are summarized in Table 2. There were observed significant group differences on all tests except on reading (t(58) = 0.38; p = NS). Similarly, group comparison on psychomotor performance showed significantly reduced scores for the elderly as compared to the younger subjects. Cognitive and psychomotor results versus normative scores are presented in Appendix B.
Table 2 Mean and ±SE for cognitive and psychomotor tests by age group
Note. *p < .05, **p < .01, ***p < .001.
Due to the important group differences on education and vocabulary, one-way analyses of covariance (ANCOVA) were conducted by entering these two variables as covariates. Results are presented in the following section.
Acoustic analyses of four verbal tasks
Reaction times (RTs)
RTs were obtained across all verbal tasks and converted to Z-scores as previously described (Figure 1). For descriptive purposes raw data are also presented in Figure 2. Given that Z-scores of RTs distributions were not normal, the nonparametric statistics Mann-Whitney U were used to evaluate group differences. Results showed highly significant differences between both groups across all tasks (phonemic, U = 165.5; p < .0001; semantic, U = 115; p < .0001; naming, U = 244; p < .0001; reading, U = 115.5; p < .0001). Furthermore, we used for each group a Friedman's two-way ANOVA to compare the scores for the four tasks. The Friedman tests showed that Z-scores of RTs were significantly different for both the young (χ 2(3) = 77.72; p < .0001) and the older group (χ 2(3) = 68.78; p < .0001).
Fig. 1 Mean ± SEM for Z-scores reaction times in seconds for each verbal task. ***p < .001 different from young group.
Fig. 2 Mean ± SEM for raw scores reaction times in seconds for each verbal task.
To evaluate whether vocabulary and education were influencing group differences on RTs, we performed independent ANCOVAs with rank transformations for each RT (Conover & Iman, Reference Conover and Iman1982). After removing the effect of vocabulary and education a significant effect of group remained for reading (F(1,56) = 14.83; p < .001), semantic (F(1,56) = 9.87; p < .01) and phonemic fluency (F(1,56) = 6.31; p < .05) but not for naming (F(1,56) = 0.097; p = NS). Interestingly, only the covariate effect of vocabulary on RT for phonemic fluency, turned out to be significant (F(1,56) = 4.28; p < .05). All analyses showed to be safe from the threat of inflated Type I error rates (Tabachnick & Fidell, Reference Tabachnick and Fidell2001).
Pronunciation speed
As for RTs, all measurement of pronunciation times on naming and reading were converted to Z-scores. Since distributions were normal two MANOVAs for independent groups were performed, one for naming and one for reading. Only data from correctlyFootnote * pronounced words were measured and entered in the analyses. After significant MANOVA, univariate tests were then performed for each word. Significance levels were adjusted with the Bonferroni correction. Results from both MANOVAs yielded a significant group effect on pronunciation speed for naming (Pillai's Trace (8, 40) = 6.25; p = .0001) and reading (Pillai's Trace (10, 44) = 4.71; p = .0001). Moreover, univariate analyses were performed for each word on naming and reading with the Bonferroni correction. Results showed that elderly used longer time to articulate most of the words excepting two on the naming task (“druer’ and “klokke’).
To adjust for, and to evaluate, potential effects of education and vocabulary two MANCOVAs for independent groups were performed, one for each task. Results demonstrated a significant group effect for naming (Pillai's Trace (8, 38) = 3.29; p < .01) and no significant influence of the covariates (education, Pillai's Trace (8, 38) = 0.393; p = NS; vocabulary, Pillai's Trace (8, 38) = 1.94; p = NS), indicating that pronunciation speed differences for naming between young and elderly are not influenced by the covariates. Eta-squared values were 0.08 for education, 0.29 for vocabulary and 0.40 for group.
In contrast the corresponding MANCOVA for reading showed no significant main effects of group (Pillai's Trace (10, 42) = 1.76; p = NS) after controlling for vocabulary and education. None of the covariates reached significant levels (education, Pillai's Trace (10, 42) = 0.46; p = NS vocabulary, Pillai's Trace (10, 42) = 0.87; p = NS) and the eta2 values were 0.09 for education, 0.17 for vocabulary and 0.30 for group.
Correlations between RTs and task performance
Significant associations were found between task performance and RTs (see Table 3). Correlations varied between r = −0.36 (phonemic) to −0.43 (semantic). Most interesting, performance across all tasks significantly correlated with RTs for semantic fluency. These data indicate that speed to initiate semantic fluency relates to proficiency in different verbal abilities.
Table 3 Pearson correlations between performance and reaction times across all verbal tasks
Characters in bold denote significant correlations.
Hierarchical multiple regressions
Original RTs were converted to natural Log transformations since Z-scores’ distributions were not normal. Due to significant group differences in education and vocabulary, we decided to enter these variables as possible predictive factors. However, to avoid high multicollinearity (Berry & Feldman, Reference Berry and Feldman1985), we decided to enter only vocabulary on Step1 of each regression due to its relevance for verbal execution. Based on the speed of processing hypothesis, we entered the psychomotor tests in Step2, and the cognitive measures in Step3.
Prediction of RTs
We hypothesized that influence of psychomotor decline would depend on task difficulty, being more prominent for reading and naming than for VFTs. Thus, we executed four regression analyses with forced entry methods (Tables 4, 5). Results for reading showed an overall model accounting for 38% of the variance. Vocabulary significantly explained 10% and the psychomotor measures accounted for an additional 26% of the total variance. Of interest, Vocabulary was not a significant predictor when psychomotor measures were entered in further steps. The increase in Step2 was actually related to Pegboard scores (Table 4). No further significant increment was obtained by adding the cognitive measures.
Table 4 Hierarchical multiple regression analyses predicting mean scores of reaction times for Reading and Naming based on entire sample
Note. All reaction time (RT) data are Log-transformed.
Table 5 Hierarchical multiple regression analyses predicting mean scores of reaction times for Semantic and Phonemic Verbal Fluency based on entire sample
Note. All reaction time (RT) data are log transformed.
Concerning RTs for naming, Vocabulary accounted significantly for 18% of the variance in Step1 and psychomotor measures added an additional 8%. Again, Pegboard was the only psychomotor test significantly associated with RTs. In Step3, Vocabulary remained a significant predictor and no further increment was observed by addition of the cognitive tasks.
The analysis on semantic fluency showed that Vocabulary and psychomotor execution were stronger predictors of RTs across all steps. In Step1 Vocabulary accounted for 28% of the variance, while in Step2 the psychomotor measures further explained 19% of the variance (Table 5). In Step3, no change was obtained by addition of the cognitive tasks, though, the beta weight of Pegboard was attenuated. Finally, results on RTs for phonemic fluency showed that Vocabulary explained 19% of the variance and the psychomotor tests significantly explained an additional 13%. However, in Step3 only Vocabulary remained a significant predictor (Table 5).
Prediction of pronunciation speed
Forced-entry methods were again performed. It should be noted that the analysis for naming included only 26 young adults and 23 elderly since execution on this task does not impose production of an exact word (e.g., banana, fruit, food). The models predicting Z-scores for pronunciation speed on reading and naming were comparable, as Vocabulary and psychomotor measures significantly explained an important portion of the variance (Table 6). On reading these variables explained 48% of the observed variance, while for naming they accounted for 37%. The inclusion of the cognitive measures in the models did not contribute to explain the total variance. Noteworthy, is the strong influence of psychomotor outcomes and the fact that Vocabulary was more determinant for reading (R 2 = .28) than for naming (R 2 = .16).
Table 6 Hierarchical multiple regression analyses predicting mean Z-scores of articulation speed for words produced in Reading and Naming based on entire sample
Stratified analyses
We explored specific associations by age group to understand whether predictors and outcomes significantly differed between young and older adults. Results are shown in Tables 7–12.
Table 7 Stratified hierarchical multiple regression analyses predicting mean reaction times for Reading by age group
Table 8 Stratified hierarchical multiple regression analyses predicting mean reaction times for Naming by age group
Table 9 Stratified hierarchical multiple regression analyses predicting mean reaction times for Semantic fluency by age group
Table 10 Stratified hierarchical multiple regression analyses predicting mean reaction times for Phonemic fluency by age group
Table 11 Stratified hierarchical multiple regression analyses predicting mean Z-scores of articulation speed for words produced in Reading by age group
Table 12 Stratified hierarchical multiple regression analyses predicting mean Z-scores of articulation speed for words produced in Reading by age group
Prediction of RTs- by age group
In these stratified analyses, only Vocabulary significantly explained some of the RTs. For the younger group Vocabulary explained 32% of the variance in phonemic VFT (Table 10) and 19% in naming (Table 8). Conversely, in the elderly group Vocabulary predicted significantly RTs in semantic VFT accounting for 25% of the variance (Table 9). Of interest, Vocabulary was not a significant predictor in RTs for reading in none of the groups, while Finger Tapping contributed to explain 22% of the variance in the elderly despite that the model was not significant. Excepting for the association between Finger Tapping and RTs for reading, the psychomotor and cognitive variables did not appear as important predictors in the stratified analyses. Altogether, results suggest that Vocabulary exerts an influence on specific verbal RTs depending on age group.
Prediction of pronunciation speed by age group (Tables 11, 12)
Significant influence of some predictors was observed in Reading. Vocabulary accounted for 23% of the variance in the elderly, while Finger Tapping in the younger group significantly increased the proportion of the explained variance (R 2 change = .19). For naming, no significant results were found probably due to small sample sizes.
Discussion
The age-related slowing on speech production in healthy elderly adults was confirmed. Both RTs and pronunciation speed were significantly slower in older adults. Also, longer RTs in the elderly were specially pronounced for semantic fluency (Figure 1). This outcome was also true for the younger group, though, to a much lesser extent. These findings indicate that semantic fluency is the most demanding task among the selected tests for both groups and mainly for the elderly. Otherwise, RTs results corroborated our initial hypothesis stating that, reading was the easiest task and, therefore, expected to be initiated and performed faster that the other tests. Furthermore, we obtained that group differences on RTs were still present on reading, semantic and phonemic fluency, even after controlling for important group differences in education and vocabulary.
Concerning pronunciation speed in reading and naming, it was found that the older group required longer time to articulate most part of the words. There were no group differences on speed of articulation of 2 words on reading and 2 words on naming, probably due to high intra-group variation. Notwithstanding, the elderly showed a clear trend of speed reduction across tasks that corroborates results from earlier studies (Balota & Duchek, 1988; Multhaup & Duchek, Reference Multhaup, Balota and Cowan1996). Noteworthy is that group differences remained significant for articulation speed on naming but not on reading after controlling for vocabulary and education.
Regression analyses for RTs and pronunciation speed of verbal tasks on the total sample
RTs
Results confirmed our hypothesis stating that age-related psychomotor decline plays a highly influential role on speech initiation for reading. Indeed, psychomotor performance was the estimator accounting for almost all the variance explained by the model (36%). Conversely, the influence of psychomotor execution determining RTs for naming was limited. In this case, Vocabulary explained most of the accounted variance for the model (18%), while psychomotor measures added only 8%. Actually, none of the selected cognitive or psychomotor tasks entered in the model was strongly associated with naming. Moreover, psychomotor scores were more determining for semantic RTs explaining 19% of the variance than for phonemic RTs in which they explained 13% of the variance. The influence of Vocabulary on RTs was of significance across all tasks, being highest on semantic VFT, then on phonemic VFT, naming and reading.
By controlling statistically for psychomotor decline and Vocabulary, the contribution of executive functions and working memory was of no transcendence. Most likely functions associated with crystallized aspects of intelligence, like those relevant for Vocabulary, might be critical determinants of RTs in verbal execution. Finally, it is noteworthy that Pegboard was the measure showing a significant predictive influence on all RTs. Although, we did not compare the extent or magnitude of decline among Pegboard and speech generation, this finding suggests that to a significant degree, the age-related decline on fine motor abilities might be comparable to the reduction in speed of speech production in aging.
Pronunciation speed
The contribution of psychomotor measures and Vocabulary was more important on reading than on naming. Both psychomotor measures significantly predicted articulation speed in Step2. As for articulation speed in naming, Vocabulary accounted for 16% of the variance, while the psychomotor block accounted for a larger portion (21%). Again, the cognitive measures entered in the model did not contribute to explain articulatory duration, though; they lessened the influence of the rest of the variables excepting that of Finger Tapping.
Stratified analyses
Stratified analyses revealed a differential role of Vocabulary in predicting RTs and articulation times on both groups. For the younger group, Vocabulary explained significantly a portion of the variance in naming (19%) and phonemic VFTs (32%). In the elderly group, Vocabulary only predicted significantly semantic RTs, explaining 25% of the variance. As for articulation times, Vocabulary explained 23% of the variance in reading in the younger group, while Finger Tapping significantly explained 19% of the accounted variance in the elderly also in reading. No further significant findings were obtained.
There are two points to highlight. First, if these results are replicable on larger samples, it would be central to answer why Vocabulary exerts a differential influence on speed of performance of the same tasks in younger and older adults. Either this is a cohort effect coming from different educational systems, or to the fact that some elderly, most probably those who do not carry out intellectual activities on a daily basis, are experiencing deterioration of some lexical capacities (Lopez-Higes, Valdehita, Aragoneses, & Del Rio, Reference Lopez-Higes, Rubio Valdehita, Martin Aragoneses and Del Rio2010).
Concerning the first alternative, we deliberately did not focus on highly educated elders simply because the elderly population of most European countries has less years of education than the younger generation (Huisman et al., Reference Huisman, Kunst, Andersen, Bopp, Borgan, Borrell and Mackenbach2004). Thus, the issue of educational differences between ages is a current problem that influences verbal functions of the elderly population of our time. However, to settle the role of verbal knowledge on speed of speech generation, it would be important to replicate these findings in a group of highly educated elders. Regarding the second alternative, it would be important to elucidate whether Vocabulary deteriorates in some elderly as a consequence of lack of use of verbal material or due to a probable conversion to mild cognitive impairment (Roberts, Fuhrer, Marmot, & Richards, Reference Roberts, Fuhrer, Marmot and Richards2010).
It should be admitted that results from the stratified analyses need to be interpreted with considerable caution since the stratified methodology reduces sample sizes within groups, and hence power. It is suggested that performing dichotomization of quantitative measures, like age, results in substantial loss of information related to splitting the total sample (MacCallum, Zhang, Preacher, & Rucker, Reference MacCallum, Zhang, Preacher and Rucker2002). For instance, the lack of significant associations between predictors and pronunciation times within groups could be due to reduced sample sizes. Undeniably, when performing the analyses on the entire sample the effects of Vocabulary and psychomotor measures are evident. For illustrative purposes of this issue, Appendix C shows a scatter plot matrix presenting pronunciation speed distributions by group in relationship with predictors used in the regression analyses.
Limitations of the study
We acknowledge that the selected cognitive tasks were limited and other cognitive functions like memory measures should be further examined. Alternatively, replication of these results with a different set of executive tasks would be valuable. Moreover, we acknowledge that there is a limited number of RTs across all verbal tasks and specifically related to VFTs, which has implications on reliability of the findings and might not allow for generalizations. It would also be interesting to test whether frequency and syllable length of produced words influence speed of speech. Finally, the segregated analyses should be replicated in larger samples. Despite the above limitations, we believe that the obtained results represent a valuable first step to understand the source of age-related slowing in speech production.
Conclusions
It was demonstrated that age-related differences in Vocabulary and psychomotor execution significantly influence the age-related slowing on speech production. The effect of these variables on speed of speech generation varies depending on type of verbal task. Our data showed that the interplay of lexical abilities and psychomotor decline explained between 26% and 49% of the total variance in verbal tasks. The rest of the variance that was not explained by the present set of variables need to be understood by complex interactions of cognitive processes related to memory, motivation, stress, and perceptual decline. Further biological factors affecting speed of verbal performance such as neuromuscular or respiratory changes, as well as sensorial modifications might further explain the age-related slowing in production of oral language.
Acknowledgments
The present study has not been published previously, electronically or in print, and represents unique work by the authors. This work was supported by the Department of Psychology, University of Tromsø, Norway. None of the authors have a financial or other conflict of interest in the publication of this work.
Appendix A: Stimuli used for naming and reading
Appendix B: Comparison between cognitive and psychomotor scores from the young and elderly groups vs. normative studies
Appendix C: Scatter plot matrix of articulation speed and predictive variables used in the hierarchical regression analyses
