Multiple Temporal and Semantic Processes During Verbal Fluency Tasks in English-Russian Bilinguals

Category fluency test (CFT) performance is sensitive to cognitive processes of executive control and memory storage and access, and widely used to measure cognitive performance especially in early Alzheimer’s Disease. Analytical variables have included the number of items named, and various methods to identify and quantify clusters of semantically related words and cluster switches. Also encoded in the response sequence are temporal patterns as shown by “bursts” of responses and pauses between items, that have not been received attention in determining cluster characteristics. We studied a group of 51 adult Russian-English bilinguals and compared CFT responses based on two clustering methodologies: the semantic-based method (SEM) and a novel method based on the time interval between words (TEMP) with 8 different intercall time thresholds from 0.25 sec-15 sec. Each participant performed the task in both languages. Total number of words and cluster count was greater in Russian than English for both scoring methods, but cluster size did not differ between languages. We also studied stochastic modeling characteristics based on detrending of the “exponential exhaustion” effect seen with CFT, with most notable that total recall capacity (N∞) was greater in Russian than English (P<.05). Multiple demographic variables, and recent and lifetime usage of each language, affected both cognitive performance as measured by the Montreal Cognitive Assessment (MOCA; given in English only). Differential performance is driven by differences in demographics, more words stored in memory, and semantic and timing recall strategies.


85
To assess clustering ability and to better understand these cognitive processes and how

146
The Montreal Cognitive Assessment (MOCA) and animal naming tests were recorded 147 using a handheld digital device. The recordings were transcribed and the time from the start of 148 the trial to the start of each word (elapsed time) was calculated using WavePad Sound Editor 149 (NCH Software Inc., Greenwood, CO).

150
Responses from the Russian trials were translated into English by a native Russian 151 speaker, and we recorded the total number of non-repeated responses not including errors in 152 60 seconds. Two raters scored each trial for semantic clustering ("SEM" method), following the 153 Troyer et al. [1997] method with the following exceptions: we did not assign any response to 154 more than one cluster, we counted cluster size as the number of words in a cluster, and we 155 counted single words (i.e. those not semantically associated with a response preceding or 156 following) as a cluster size of one. Pearson correlation coefficient (r) between raters for semantic 157 cluster scoring was 0.9 for the English CFT, and 0.84 for the Russian CFT. 211 participants were female, ranged in age from 19 to 75 years, were well educated (86% 212 graduated college or had a post-graduate education), and all but two were born outside the 213 United States, primarily in Russia or Ukraine. All participants spoke Russian before they spoke 214 English, and began English language instruction between the ages of 3 and 59 years. Table 1 215 shows subject demographics, MOCA scores, MOCA letter fluency word count and N 60 in each 216 language. Table 1 also shows the life time and previous year index of usage of each language.
217 Table 2 shows the univariate correlation analysis showed that both Russian and English word 218 counts (N 60 ) correlated significantly between themselves, and were highly correlated with 219 MOCA score and MOCA letter fluency, education, lifetime Russian shown).    When comparing cluster characteristics between languages, semantic cluster count was 273 significantly higher in Russian than in English (Table 5). This is likely an effect of the greater 274 N 60 in Russian than English; more words generally result in more clusters. Cluster sizes, 275 however, did not differ significantly between Russian and English (Table 5). There were 276 significantly more temporal clusters in Russian than English for all durations 1 second or less.

221
277 However, the curves were of similar shape, and 0.25 sec threshold duration had cluster counts 278 that approximated N 60 , and threshold duration of 15 seconds almost always yielded a single 279 cluster of all words (Fig 2).
280 314 For the former aim, participants produced more responses in Russian than English, and this 315 appears multiply determined, including demographics and differential language use, but also 316 differences in response timing, total time spent engaged in task, and size of lexicon in each 317 language as measured by total recall capacity. For the latter aim, measuring temporal clustering 318 is quite feasible and allows comparison to the established semantic clustering method.

320
Temporal clustering can only be done and measured when the CFT is recorded

335
Long duration pauses in the response sequence are common, and there is often a 336 "second wind" phenomenon, with a second acceleration of responses after a long pause -in 337 effect, restarting the task. These longer duration pauses are problematic since they suggest 338 alterations in brain processing whose meaning is ambiguous. In the SEM method but not in the 339 TEMP method, a long pause is incorporated into the sequence of a cluster raising the question 340 of whether the respondent "intended" the responses to be semantically related. That is, long 341 durations finally producing a semantically related word, may indicate the end of one cluster, and