Analyses of correct responses and errors on measures of verbal fluency among Parkinson’s disease and essential tremor patients

Abstract

Objective:

Parkinson’s disease (PD) and essential tremor (ET) involve neuroanatomical circuitry that impact frontal lobe functioning, via the striatum and cerebellum, respectively. The aim of this exploratory study was to investigate quantitative and qualitative performance between and within these groups on measures of verbal fluency.

Method:

Sixty-three PD and 53 ET patients completed neuropsychological testing. Linear regression models with robust variance estimation compared verbal fluency performance between groups related to correct responses and errors. Paired t-tests investigated within group error rates.

Results:

PD patients gave more correct responses for phonological ($\widehat{\beta }=5.3$, p=.01) and category fluency ($\widehat{\beta }=4.1$, p=.01) than ET patients; however, when processing speed was added as a covariate, this attenuated performance on both measures and only phonological fluency remained significant ($\widehat{\beta }=4.0$, p=.04). There were no statistical differences in error scores between groups. Error rates within groups suggested that PD patients had higher error rates in total errors and perseveration errors on phonological fluency (M = 2.6, p=.00; M = 1.6, p=.00) and higher total errors and set-loss error rates on category switching (M = 5.1, p<.001; M = 4.1, p<.001). ET patients had higher error rate with relation to total errors and set-loss errors on phonological fluency (M = 2.5, p=.00; M = 1.5, p=.02) and category switching (M = 3.9, p=,00; M = 3.9, p<.001).

Conclusions:

PD patients performed better than ET patients on phonological fluency. PD patients appear to make more perseveration errors on phonological fluency, while ET patients made more set-loss errors. Implications for frontal lobe dysfunction and clinical impact are discussed.

Introduction

Parkinson’s disease (PD) is a movement disorder associated with a loss of dopaminergic neurons due to the deposition of alpha-synuclein, Lewy neurites, and Lewy bodies in the basal ganglia, brainstem, and cortical regions (Angot & Brundin, 2009; Braak et al., 2004; Klingelhoefer & Reichmann, 2015; Maiti et al., 2017). The subsequent depletion of dopamine in the substantial nigra pars compacta leads to an impairment in motor functioning that may result in asymmetric motor symptoms that include tremor, rigidity, and bradykinesia. The dopamine depletion is likely to compromise basal ganglia-thalamic-cortical circuits that impact frontal lobe functioning and subsequently may lead to cognitive impairment (Levin et al., 1988; Rodriguez-Oroz et al., 2009).

Several studies have found that among the earliest signs of cognitive impairment in PD, the domains of attention and executive functioning appear to be the most affected, at times well before the onset of motor symptoms (Curtis et al., 2019). Specifically, executive functioning including planning, set shifting, and working memory appear the most affected (Kudlicka et al., 2011; Muslimović et al., 2005; Weintraub et al., 2018; Zgaljardic et al., 2003), other studies revealed impairments in memory and language (Pfeiffer et al., 2014) and verbal fluency (Kudlicka et al., 2011; Tagini et al., 2021). A meta-analysis of mild cognitive impairment (MCI) in PD found a prevalence rate of 40% (Baiano et al., 2020) and single domain and non-amnestic MCI has been noted to be the most common type (Litvan et al., 2011). A review study by Hanagasi et al. (2017) noted a prevalence rate of dementia to range from 24 to 31% from previous studies, with other studies suggesting that 48% of their prospective sample developed dementia 15 years following diagnosis and this increased to 83% 20 years following diagnosis (Hely et al., 2005, 2008).

For decades, essential tremor (ET) was described as an isolated syndrome of bilateral upper limb action tremor, without tremor in other body parts, of at least three years duration and absence of other neurological signs (Bhatia et al., 2018). It was initially described as “benign” due to the notion that its mechanism solely manifested in action tremor; however, this has changed over time. The evolution of what is known about the disease has led to the recognition of “ET Plus,” which includes the tremor associated with ET, but also recognizes the presence of additional neurological signs of unknown significance that may include impaired tandem gait, dystonic posturing, memory impairment, and other mild neurological signs. Non-motor symptoms associated with ET have come to the fore over time, including cognitive impairment primarily affecting executive functioning from presumed cerebellar-thalamic-cortical dysfunction (Lee et al., 2015; Louis et al., 2010; 2019; Puertas-Martín et al., 2016; Sánchez-Ferro et al., 2017).

Whereas in PD dopaminergic depletion has been associated with disrupting the striatal/basal ganglia-thalamic-cortical pathway leading to executive dysfunction (Levin et al., 1988; Rodriguez-Oroz et al., 2009), in ET, the mechanism has been established to be related to disruptions in the cerebellar-thalamic-cortical circuits (Schmahmann, 1996; Wallesch & Horn, 1990). Importantly, in PD, the cerebellar circuit has been noted to be preserved until mild changes in the final stages of the disease (Braak et al., 2004; Braak & Del Tredici, 2009). Specifically, the cerebellum has been shown to impact executive functioning with relation to known deficits in the areas of cognitive flexibility, phonological fluency, problem-solving, and inhibition (Baillieux et al., 2010; Gottwald et al., 2003; 2004; Manes et al., 2009; Neau et al., 2000; Rapoport et al., 2000; Schmahmann & Sherman, 1998).

Recent studies have suggested that MCI in ET is best predicted from performance on combined executive functioning and memory measures, suggesting that ET may also involve other cortical regions beyond cerebellar-thalamic-cortical loop (Cersonsky et al., 2018; Collins et al., 2017). With regards to progression to dementia, previous population-based studies have found higher rates of dementia in ET samples versus normal controls as well as higher risk associated with younger age onset of ET (Bermejo-Pareja et al., 2007; Thawani et al., 2009).

Verbal fluency has been shown by previous studies to be impaired in both PD and ET when compared to normal controls (Kudlicka et al., 2011; Leggio et al., 2000; Tagini et al., 2021; Tröster et al., 2002). The findings related to deficits in verbal fluency for PD have been mixed, with some studies suggesting greater impairment in phonological fluency and other studies demonstrating greater impairment in semantic fluency (Silveri, 2021). A meta-analysis by Henry and Crawford (2004) of 68 studies in particular noted small to moderate effect size among PD patients (in comparison to controls) related to deficits in semantic fluency (r=.37) and phonological fluency (r=.33; t = 2.53, df =49, p=.015). Although fewer studies were included in their meta-analysis, Kudlicka et al. (2011) found moderate to large effect sizes related to phonological (g= −0.55, p = 0.00) and category fluency (g= −0.75, p = 0.00). Effect sizes in verbal fluency in ET samples are unknown as no meta-analytic studies have been completed. However, ET patients have been found to perform one standard deviation below the mean on phonological and semantic fluency (Tröster et al., 2002). Other studies have found greater phonological versus semantic fluency difficulty, which from a neuroanatomical perspective, is consistent with studies revealing that cerebellar damage has been associated with greater phonological versus semantic fluency impairment (Leggio et al., 2000; Mariën & Manto, 2018; Silveri, 2021).

Few studies have investigated and compared verbal fluency performances between PD and ET patients. A large-scale population study (Sánchez-Ferro et al., 2017), investigated cognition in PD, ET, and control groups. While both groups performed less well than controls on measures of verbal fluency, PD patients performed less well than ET patients on category fluency. A study by Gasparini et al. (2001) who found noted that PD patients performed worse on phonological fluency than ET patients. On the other hand, other studies found that ET patients performed worse on letter fluency than PD patients (Higginson et al., 2008), which was consistent with a previous study indicating that ET patients performed worse on phonological fluency than PD patients (Lombardi et al., 2001), which they related to cerebellar cognitive affective syndrome (Schmahmann & Sherman, 1998). These four studies did not include an investigation of both phonological and semantic fluency in their methodology and looked at either phonological fluency or semantic fluency in isolation.

Two common types of errors made on verbal fluency tasks include perseveration and set-loss errors. Perseveration errors (Sandson & Albert, 1984; Stuss & Alexander, 2007), are related to difficulties with monitoring (being able to check and adjust behavior as needed) or inhibition (stop a previously completed behavior) and have been demonstrated to be sensitive to right frontal lobe functioning (Robinson et al., 2021). Set-loss errors reflect an inability to maintain a cognitive set and inability to stop from returning to previous responses/set (Delis et al., 2001). Set-loss errors been implicated left frontal lobe dysfunction (Cipolotti et al., 2020). Few studies have investigated errors on verbal fluency tasks in movement disorder samples, with mixed findings. Tjokrowijoto et al. (2020) found that PD patients demonstrated a higher number of rule-break errors than controls and higher perseveration errors than cortical basal syndrome CBS patients. However, other studies have not found significant difference in error rates (Azuma et al., 1997, Foley et al., 2021; Pettit et al., 2013).

The purpose of this exploratory study was to investigate performance on verbal fluency measures in a PD and ET sample, especially in light of the risk for cognitive decline and dementia in both disorders. Previous studies have presented mixed results when comparing both disorders head-to-head (although there is suggestion that PD patients perform worse on semantic fluency tasks and ET patients perform worse on phonological fluency tasks in comparison to healthy controls). The current study sought not only to investigate differences in traditional verbal fluency scores which include total correct responses (quantitative data), but also investigate errors (qualitative data) made on these tasks. Given that error analyses may offer additional crucial information related to frontal lobe dysfunction (perseveration errors may be more associated with right frontal lobe dysfunction and set-loss errors may reflect left frontal lobe dysfunction), this study sought to investigate the pattern of errors, how they may be different among ET and PT patients, and which types of errors specifically may be the most pervasive. To our knowledge, this is the first study to make a head-to-head and within group comparison between PD and ET patients in their performance on not only phonological and semantic fluency, but also category-switching fluency.

Method

Participants

Sixty-three patients with Parkinson’s disease (PD) and 53 patients with essential tremor (ET) were included in this retrospective sample study. These patients completed the neuropsychological evaluation as part of being considered for candidacy to undergo deep brain stimulation (DBS; n = 113) or as part of an outpatient neuropsychological evaluation clinic to evaluate for cognitive changes (n = 3). The evaluations were completed by the first author. Institutional Review Board (IRB) approval to examine archival data was obtained from the Human Subjects Division of the University of Washington.

To be included in the study, the patients must have been previously diagnosed with PD or ET by a movement disorders specialist. Exclusion criteria included: (1) diagnosis of comorbid PD and ET; (2) presence of major neurocognitive disorder based on performance on neuropsychological evaluation (n = 3 PD; n = 1 ET): two or more domain composite scores falling two or more standard deviations below the mean (3) other major neurological history including epilepsy, stroke, comorbid movement disorder, (4) previous neurosurgery, and (5) severe psychiatric or substance abuse comorbidity based on clinical interview (guided by DSM-V criteria).

With regards to general cognitive functioning for the sample, composite Z-scores for specific domains (attention and processing speed, executive functioning, language, learning and memory, and visuoperceptual/reasoning) were calculated as follows. Raw scores for specific test measures within domains, detailed on Table 1, were demographically normed using test specific manuals or normative data that has been published previously (Heaton et al., 2004). These scores were then converted to Z-scores with a mean of 0 and standard deviation of 1.0. The individual Z-scores for each test within a domain were then averaged out to obtain the domain-specific composite Z-scores. When possible, normative data was adjusted for age, gender, and or education. Premorbid intellectual functioning was assessed by the Test of Premorbid Function (TOPF; Pearson Assessment, 2009), a word reading test of 70 words.

Table 1.

Neuropsychological measures for each of the cognitive domains.

Domain	Test	Score used
Attention and processing speed	WAIS-IV Digit Span, Forward	Total Recall
	SDMT, Oral	Total correct
	DKEFS Color Word, Word	Completion time
Executive functioning	TMT-B	Completion time
	DKEFS Letter Fluency	Total number of words (3 trials)
	DKEFS Fluency Switching	Total number of switches
	DKEFS CW Inhibition/Switching	Completion time
Language	BNT	Total number correct with semantic cuing
	DKEFS Category Fluency	Total number of words (2 trials)
Learning and memory	CVLT-III	Delayed total recall
	WMS-IV Logical Memory	Delayed total recall
Visuospatial/perceptual	Benton JOLO	Total correct
	WAIS-IV Matrix Reasoning	Total correct

Note. WAIS-IV: Wechsler Adult Intelligence Scale – Fourth Edition (Wechsler, 2008); SDMT Symbol Digits Modalities Test (A. Smith, 1982); DKEFS: Delis-Kaplan Executive Function System, (Delis et al., 2001); TMT-B: Trail Making Test Part B (Reitan, 1992); BNT: Boston Naming Test (Kaplan et al., 1983); CVLT-III: California Verbal Learning Test – Third Edition (Delis et al., 2017); WMS-IV: Wechsler Memory Scale – Fourth Edition (Wechsler, 2009); Benton JOLO: Benton Judgment of Line Orientation (Benton & van Allen, 1968).

Measures

Delis-Kaplan executive functions system (DKEFS)

The verbal fluency subtest of the DKEFS (Delis et al., 2001) includes letter fluency, category fluency, and category switching. The letter fluency tasks asked the patient to generate as many words as possible that begin with a specific letter (three trials: F, A, S) in 60 seconds. Set-loss errors included saying a word that begins with a letter other than the designated one; saying words that start with the target letter but are names of people, places, or numbers; and grammatical variants of the same word (e.g. fast, faster, etc.). Repetition (perseveration) errors included any response that was repeated within the 60 second time limit. The letter fluency tasks taps into initiation, organization, lexical generation, and simultaneous processing (Delis et al., 2001). For category fluency, the patient was asked to generate as many words as possible that belong to a specific category (two trials: animals and boys names) in 60 seconds. A set-loss error was any response that violated the criterion rules including saying words that do not belong to the category; grammatical variants of the same word; and words ordinate and superordinate to the target word. Repetition (perseveration) errors included any response that was repeated within the 60 second time limit. The category fluency tasks also required rapid retrieval of lexical items and may be less novel than letter fluency. For the category switching task, the patient was asked to generate as many words as possible, alternating between two different semantic categories (fruits and furniture) in 60 seconds. At set-loss error was a response that violated any of the criterion, including words that do not belong to either of the two categories; grammatical variants of a given word; and responses that are ordinate or superordinate to either target category. A repetition (perseveration) error was any response that was repeated within the 60 second time limit. The category switching task taps into rapid retrieval of semantic knowledge while simultaneously demonstrating the cognitive flexibility in shifting between two categories.

Data analysis

Analyses were conducted using the Statistical Package for Social Sciences Version 27, STATA Version 17, and SAS version 9.4 (TS1M6). Patient characteristics by diagnostic group (PD/ET) were summarized descriptively, by mean and standard deviation (for continuous variables) or frequency and percentage (for categorical variables). Univariable comparisons of patient characteristics between diagnostic groups were carried out by t test (for continuous variables) or chi-square test (for categorical variables). Comparisons of between-subjects means for number of correct responses, number of error responses, and error rates were made using linear regression with robust variance estimation. Linear regression models were adjusted for age, gender, years of education, and premorbid intelligence. Robustness of linear regression to violations of the assumption of normally distributed residuals is supported by the Central Limit Theorem, given sample sizes of more than 50 patients in both the PD (N = 63) and ET (N = 53) groups (Zhang et al., 2022).

We conducted a sensitivity analysis using a marginalized zero-inflated negative binomial model (Cummings & Hardin, 2019) for the “number of errors” outcome, and a marginalized two-part beta model (Chai et al., 2018) for the “error rate” outcome. These models make allowance for the high proportion of zero responses (for patients who did not make any errors) by treating the zero and nonzero responses separately within a single model. Results were compared with estimates and p values from the linear regression analyses, which were consistent. Comparisons of within-subject mean error rates for phonological fluency and category switching, versus category fluency, were carried out using paired t tests. A nominal significance level of 5% was specified for hypothesis tests. As this was an exploratory study, we did not adjust for multiple comparisons.

Results

Sample characteristics are presented in Table 2. Overall, the sample included more men than women, were well educated, were predominantly Caucasian, had similar premorbid intellectual functioning, and endorsed low levels of depression and anxiety symptoms. The two samples differed with relation to education as the PD sample completed more years of education (14.6 ± 2.9 vs. 13.6 ± 2.2, p=.04), the ET sample had greater disease duration (30.8 ± 17.4 vs. 8.3 ± 3.8, p= <.001), and the ET sample had a lower attention and processing speed composite score (−0.1 ± 0.6 vs. −0.4 ± 0.5, p=.00). When available, a measure used to evaluate PD symptom severity included the Movement Disorder Society-Unified Parkinson’s Disease Rating Scale (UPDRS-III; Goetz et al., 2008) and average “ON” scores from part III (motor) was utilized, with higher scores indicating higher motor difficulty. When available, scores from the Fahn-Tolosa-Marin Clinical Rating Scale for Tremor (FTM; Fahn et al., 1988) were used to assess for tremor severity among ET patients. Sixty-one of the PD patients were being treated with dopaminergic agents at the time of testing. Thirty-five of the ET participants were currently being treated for their tremor with either Propranolol, Primidone, and/or Topiramate.

Table 2.

Sample characteristics.

	Parkinson’s disease (N = 63)	Essential tremor (N = 53)	P value	Effect size
Demographics
Age
Mean (sd)	66.1 (9.6)	67.3 (10.3)	.51	0.12
Range	41–82	28–81
Education (years)
Mean (sd)	14.6 (2.9)	13.6 (2.2)	.04	0.39
Range	7–20	10–18
Gender (n, m/f)	48/15	34/19	.16	0.13
Ethnicity	n (%)	n (%)		0.19
Caucasian	58 (92.1)	53 (100%)
Native American	2 (3.2)
Asian American	1 (1.6)
Hispanic	1 (1.6)
Mixed ethnicity	1 (1.6)
Disease duration (years)
Mean (sd)	8.3 (3.8)	30.8 (17.4)	<.001	1.86
Range	2–20	4–70
UPDRS-III ON (n = 53)
Mean (sd)	19.9 (10.7)
Range	4–59
FTMCRS – Total (n = 20)
Mean (sd)		53.5 (14.9)
Range		34–97
Dopaminergic Medications	n = 61
Motor medications	n = 35
General cognitive functioning
TOPF (z-score)
Mean (sd)	0.2 (0.9)	0.0 (0.8)	.22	0.23
Composites (Z-scores)
Attention and processing speed
Mean (sd)	−0.1 (0.6)	−0.4 (0.5)	.00	0.54
Executive functioning
Mean (sd)	−0.7 (0.9)	−0.2 (0.5)	.32	0.69
Language
Mean (sd)	0.2 (0.8)	0.2 (0.8)	.94	0.00
Learning and memory
Mean (sd)	−0.0 (0.8)	−0.0 (0.7)	.85	0.00
Visuospatial/perceptual
Mean (sd)	0.2 (0.8)	0.2 (0.8)	.93	0.00
Mood
BDI-II
Mean (n/sd)	7.5 (n = 62/5.6)	7.3(n = 50/6.3)	.87	0.03
Range	0–30	0–31
GDS
Mean (n/sd)	6.0 (n = 1/−)	1.0 (n = 2/1.4)	.21
Range	–	0–2
BAI
Mean (n/sd)	10.1 (n = 61/6.3)	9.4 (n = 49/6.1)	.54	0.11
Range	0–29	0–25
GAI
Mean (n/sd)	6.0 (n = 1/−)	0.5 (n = 2/0.7)	.09
Range	–	0–1

Note. n=number; sd = standard deviation; UPDRS = Unified Parkinson’s Disease Rating Scale (Goetz et al., 2008); FTM = Fahn-Tolosa-Marin Clinical Rating Scale (Fahn et al., 1988); TOPF = Test of Premorbid Functioning (Pearson Assessment, 2009); BDI-II = Beck Depression Inventory (Beck et al., 1996); GDS = Geriatric Depression Scale (Yesavage et al., 1982); BAI = Beck Anxiety Inventory (Beck & Steer, 1988); GAI = Geriatric Anxiety Inventory (Pachana et al., 2007).

For a normative sample, a Z-score mean is 0 and standard deviation (SD) is 1.

Effect size = Cohen’s d for continuous variables, Cramer’s V for categorical variables.

Signifiant levels are bolded.

Prior to multivariable modeling, we examined the Pearson’s correlation coefficients for disease duration and each outcome, by diagnosis group (PD/ET). The correlations were generally weak, ranging between 0 and 0.15. Based on this finding, we omitted disease duration from all regression models. For total correct responses, results indicated that after controlling for age, gender, years of education, and premorbid intelligence, the total number of correct responses was greater in PD patients than ET patients for phonological fluency ($\widehat{\beta }=5.3$, p=.01), category fluency ($\widehat{\beta }=4.1$, p= .01), but not category switching ($\widehat{\beta }=0.3$, p=.59; Table 3). Although not part of the original study analyses, a post-hoc analysis was conducted to determine if processing speed impacted these findings, given the significant difference in attention and processing speed composite between groups (Table 2). To do this, a pure processing speed composite was calculated by excluding attentional measures from the original composite; this suggested a significant difference in processing speed between both groups (Table 4; p=.03). By adding processing speed as a covariate, it attenuated the effect on both measures, and only differences in phonological fluency remained statistically significant ($\widehat{\beta }=4.0$, p=.04).

Table 3.

Number of correct responses: Unadjusted sample means and adjusted difference in means, for patients with Parkinson’s disease versus patients with essential tremor.

	Unadjusted mean number of correct responses		Adjusted difference in mean number of correct responses
	PD (N = 63)	ET (N = 53)	$\widehat{\beta }$ a SE	95% CI	P valueb
Phonological fluency	38.4	31.8	5.3 2.0	[1.3, 9.3]	.01
Category fluency	38.2	33.8	4.1 1.6	[1.0, 7.2]	.01
Category switching	12.5	12.1	0.3 0.5	[−0.8, 1.3]	.59

Note. PD = Parkinson’s disease; ET = Essential Tremor; $\widehat{\beta }$ = coefficient of Diagnosis (PT or ED) term from multivariable regression model; SE = standard error; CI = confidence interval.

^a$\widehat{\beta }$ represents the difference in mean number of correct responses, i.e. mean number of correct responses for PD patients – mean number of correct responses for ET patients, adjusted for Age, Gender, Years of Education, and Premorbid Intelligence. Statistical inference was based on linear regression with robust variance estimation.

^bNull hypothesis: β = 0.

Signifiant levels are bolded.

Table 4.

Post-hoc analysis of number of correct responses, including adjustment for processing speed.

	Parkinson’s disease (N = 63)		Essential tremor (N = 53)
	Mean	SD	Mean	SD	P valueb	Effect Size
Processing speed Composite (Z-score)a	0.2	0.7	0.5	0.6	0.03	0.41
	Adjusted difference in mean number of correct responses
	$\widehat{\beta }$c SE	95% CI	P valueb
Phonological fluency	4.0 2.0	[0.1, 7.9]	.04
Category fluency	2.4 1.7	[−0.4, 5.2]	.09
Category switching	0.0 0.5	[−1.0, 1.0]	.99

Note. PD = Parkinson’s disease; ET = Essential Tremor; SD = Standard deviation; $\widehat{\beta }$ = coefficient of Diagnosis (PT or ED) term from multivariable regression model; SE = standard error; CI = confidence interval.

^aProcessing speed composite score includes the following test measures: SDMT, Oral, DKEFS Color Word, Word. For a normative sample, a Z-score mean is 0 and standard deviation (SD) is 1.

^bNull hypothesis: β = 0.

^c$\widehat{\beta }$ represents the difference in mean number of correct responses, i.e. mean number of correct responses for PD patients – mean number of correct responses for ET patients, adjusted for Age, Gender, Years of Education, Premorbid Intelligence and Processing Speed. Statistical inference was based on linear regression with robust variance estimation.

Signifiant levels are bolded.

Median and interquartile ranges for the error variables are presented in Table 5. An analysis investigating the difference in errors made between the groups did not reveal statistically significant differences (Table 6). To investigate error rates, we calculated the percentage of total errors, perseverative errors, and set-loss errors in reference to total responses given. Linear regression models revealed no statistically significant differences (Table 7).

Table 5.

Median, first quartile and third quartile for number of errors.

	Parkinson’s disease (N = 63)		Essential tremor (N = 53)
	Median	Q1-Q3	Median	Q1-Q3
Phonological fluency
Total errors	2	0–4	2	0–3
Perseveration errors	1	0–3	1	0–2
Set-loss errors	0	0–1	0	0–1
Category fluency
Total errors	1	0–2	1	0–1
Perseveration errors	1	0–1	0	0–1
Set-loss errors	0	0–0	0	0–0
Category switching
Total errors	1	0–2	1	0–2
Perseveration errors	0	0–1	0	0–1
Set-loss errors	0	0–1	0	0–1

Note. Q1 = First quartile; Q3 = Third quartile.

Table 6.

Number of error responses: unadjusted sample means and adjusted difference in means, for patients with Parkinson’s disease versus patients with essential tremor.

	Unadjusted mean number of error responses		Adjusted difference in mean number of error responses
	PD (N = 63)	ET (N = 53)	$\widehat{\beta }$ a SE	95% CI	P valueb
Phonological fluency
Total errors	2.33	1.96	0.40 0.4	[−0.34, 1.12]	.29
Perseveration errors	1.63	1.20	0.44 0.3	[−0.12, 1.00]	.12
Set-loss errors	0.70	0.75	−0.05 0.2	[−0.45, 0.35]	.81
Category fluency
Total errors	1.11	1.06	0.03 0.3	[−0.49, 0.55]	.92
Perseveration errors	0.90	0.79	0.09 0.2	[−0.35, 0.53]	.69
Set-loss errors	0.21	0.26	−0.06 0.2	[−0.39, 0.26]	.70
Category switching
Total errors	1.15	1.06	0.22 0.3	[−0.36, 0.81]	.45
Perseveration errors	0.49	0.36	0.18 0.1	[−0.08, 0.45]	.17
Set-loss errors	0.67	0.70	0.04 0.3	[−0.40, 0.48]	.86

Note. PD = Parkinson’s disease; ET = Essential Tremor; $\widehat{\beta }$ = coefficient of Diagnosis (PT or ED) term from multivariable regression model t; SE = standard error; CI = confidence interval.

^a$\widehat{\beta }$ represents the difference in mean number of error responses, i.e. mean number of error responses for PD patients – mean number of error responses for ET patients, adjusted for Age, Gender, Years of Education, and Premorbid Intelligence. Statistical inference was based on linear regression with robust variance estimation.

^bNull hypothesis: β =0.

Table 7.

Mean error rate: unadjusted sample means and adjusted differences in means, for patients with Parkinson’s disease versus patients with essential tremor.

	Unadjusted mean error rate (%)		Adjusted difference in mean percentage of error responses
	PD (N = 63)	ET (N = 53)	$\widehat{\beta }$ a SE	95% CI	P valueb
Phonological fluency
Total errors	5.66	5.55	0.38 1.0	[−1.60, 2.37]	.70
Perseveration errors	3.87	3.30	0.70 0.7	[0.75, 2.15]	.34
Set-loss errors	1.78	2.25	−0.32 0.6	[−1.60, 0.95]	.62
Category fluency
Total errors	3.03	3.02	−0.05 0.8	[−1.67, 1.57]	.95
Perseveration errors	2.27	2.22	−0.04 0.6	[−1.23, 1.14]	.94
Set-loss	0.76	0.79	−0.01 0.6	[−1.20, 1.18]	.99
Category switching
Total errors	8.16	6.89	2.24 1.8	[−1.29, 5.77]	.21
Perseveration errors	3.25	2.48	1.09 0.8	[−0.56, 2.75]	.19
Set-loss errors	4.90	4.74	0.73 1.5	[−2.16, 3.62]	.62

Note. PD = Parkinson’s disease; ET = Essential Tremor; $\widehat{\beta }$ = coefficient of Diagnosis (PT or ED) term from multivariable regression model; SE = standard error; CI = confidence interval.

^a$\widehat{\beta }$ represents the difference in mean percentage of error responses, i.e. mean error % for PD patients – mean error % for ET patients, adjusted for Age, Gender, Years of Education, and Premorbid Intelligence. Statistical inference was based on linear regression with robust variance estimation.

^bNull hypothesis: β = 0.

Within-subject paired t-tests were conducted to investigate differences in the type of error made within each sample. Both PD and ET patients had higher error proportions on measures of phonological fluency and category switching than category fluency (Table 8). Specifically, PD patients had higher error proportions with regards to total errors and perseveration errors on phonological fluency (M = 2.6, p=.002; M = 1.6, p=.003) and higher total and set-loss error proportions on category switching (M = 5.1, p=<.001; M = 4.1, p= <.001) than on category fluency. ET patients had higher error proportions with relation to total errors and set-loss errors on phonological fluency (M = 2.5, p=.006; M = 1.5, p=.02) and category switching (M = 3.9, p=.003; M = 3.9, p= <.001) versus on category fluency. Here, M represents the estimated within-group (PD patients or ET patients) mean of the individual differences in error percentage, relative to category fluency.

Table 8.

Mean within-subject difference in error rate for phonological fluency and category switching, compared to category fluency.

	Parkinson’s Disease (PD) (N = 63)			Essential Tremor (ET) (N = 53)
	Ma	95% CI	P value	Ma	95% CI	P value
Phonological fluency
Total errors	2.6	[1.0, 4.3]	.002	2.5	[0.8, 4.3]	.006
Perseveration errors	1.6	[0.6, 2.6]	.003	1.1	[−0.4, 2.5]	.14
Set-loss errors	1.0	[−0.1, 2.2]	.08	1.5	[0.2, 2.7]	.02
Category switching
Total errors	5.1	[2.6, 7.7]	<.001	3.9	[1.4, 6.4]	.003
Perseveration errors	1.0	[−0.5, 2.4]	.18	0.3	[−1.3, 1.8]	.74
Set-loss errors	4.1	[2.4, 5.9]	<.001	3.9	[2.2, 5.7]	<.001

Note. PD = Parkinson’s disease; ET = Essential Tremor; M = mean difference; CI = confidence interval.

^aM represents the mean within-patient difference in error percentage compared to category fluency. Null hypothesis of “mean difference equals zero” was investigated using paired t tests.

Signifiant levels are bolded.

Discussion

Verbal fluency as a measure of executive functioning allows for a motor-free method to investigate performance not only for correct responses but also an examination of errors in a PD and ET population. Both disorders have been demonstrated in the literature to show decreased verbal fluency abilities in comparison to healthy controls. Communication impairments can have a significant functional impact, leading to social isolation, occupational dysfunction, and may negatively impact mood (Smith & Caplan, 2018). Furthermore, in both disorders, verbal fluency has been shown to significantly change following deep brain stimulation (DBS) surgery (Cernera et al., 2019; Ehlen et al., 2014; Fields et al., 2003; Tröster et al., 1999), which speaks to the importance of understanding what aspects of this task is compromised, particularly as it relates to our sample that is largely being evaluated for DBS candidacy. To our knowledge, pre and post DBS studies have looked at quantitative performance (total correct responses) and not the qualitative aspects (errors) following DBS surgery. An investigation of the latter may further inform degree of executive dysfunction in these populations, as errors on measures of verbal fluency have been associated with frontal lobe dysfunction, that may be related to motor lateralization. Thus, investigating only total correct response on verbal fluency tasks may not be sufficient, as errors made during these tasks may also enlighten a pattern of executive dysfunction that is more subtle, yet important to consider in this population.

The findings from this study revealed that PD patients gave a higher number of correct responses on verbal fluency tasks related to phonological and category fluency (but not on category switching) than ET patients. However, adding processing speed as a covariate in a post-hoc analysis revealed an attenuation of the total effect on both measures and only phonological fluency remained statistically significantly, suggesting that processing speed impacts category fluency more so than phonological fluency. A previous study found that ET patients demonstrated reduced processing speed in comparison to PD patients (Sánchez-Ferro et al., 2017), which may further support our findings; however, in this same study, PD patients performed worse on category fluency than ET patients, which we did not find. Importantly, they did not control for processing speed in their analysis, thus, it is unknown if the results would have remained the same after controlling for this factor. Thus, future studies should further investigate how processing speed in an ET sample impacts cognition and seek to understand what neuropathological factors may be contributing to any significant findings.

When considering processing speed as a possible explanation for reduced verbal fluency, one must also consider the possibility that medications may have contributed to cognitive slowing in some patients in the ET group. T tests were performed which showed that ET patients taking medications for tremor reduction (n = 35) did not perform statistically different than ET patients not taking medications for tremor reduction (n = 18) on any of the three measures of verbal fluency. This may be supported by a previous study which investigated the effect of Propanolol (one of the known medications prescribed for ET) on verbal fluency in an autism sample which found that it increased performance on semantic fluency tasks (Beversdorf et al., 2011). In PD, a recent meta-analysis suggested that certain PD medications (e.g. levodopa and rotigotine) improve verbal fluency (Zhu et al., 2022). Thus, it appears that there is some evidence to suggest that medications prescribed for both conditions have the potential to improve verbal fluency and not worsen it.

Previous studies investigating between group performances on verbal fluency have been mixed, with studies finding worse performance by PD patients on category fluency (Sánchez-Ferro et al., 2017) and phonological fluency (Gasparini et al., 2001) in contrast to two other studies that found ET patients performed worse on phonological fluency tasks (Higginson et al., 2008; Lombardi et al., 2001); the latter of which are more consistent with our current findings. From a neuroanatomical perspective, one possible explanation for our findings is that of involvement of the cerebellum in ET, as studies have shown that ET has been associated with executive dysfunction due to the disruption of the cerebellar-thalamic-cortical loop (Baillieux et al., 2010; Gottwald et al., 2003, 2004; Grimaldi & Manto, 2013; Manes et al., 2009; Neau et al., 2000; Schmahmann & Sherman, 1998) and a recent review suggested support for cerebellar dysfunction and its role in phonological fluency impairment and less consistent findings of the role of the basal ganglia in verbal fluency performance (Silveri, 2021). Thus, our study and that of the previous two studies with similar findings may also support a more robust relationship between cerebellar dysfunction and phonological fluency, especially as it may be less sensitive to the influence of processing speed than semantic fluency.

The study also investigated whether analyses of errors (total errors, perseveration errors, and set-loss errors) and error rates could help differentiate performances between PD and ET patients. The results did not reveal significant differences between the two groups. It is possible that our sample sizes were too small for these analyses and larger samples may be much more revealing. Additionally, our sample was relatively cognitively intact, and the inclusion of more generalizable PD and ET patients (ranging in cognitive abilities), may shed light into error performances.

Within group analyses revealed that both PD and ET patients were prone to making more errors on measures of phonological and category switching tests versus semantic fluency. Specifically, while PD patients made more perseverative errors on phonological fluency than category fluency, ET patients made more set-loss errors on measures of phonological fluency than category fluency; and both PD and ET patients made more set-loss errors on a category switching task. As previously reported, perseveration errors have been implicated to right lateral frontal regions and set-loss errors have been shown to implicate involvement of the left lateral frontal regions (Robinson et al., 2021; Stuss & Alexander, 2007). Although not done in our study, future studies may investigate the relationship between the laterality of motor symptom onset and performance on verbal fluency measures. One study investigated laterality of motor and cognitive symptoms in PD (Steinbach et al., 2021; Verreyt et al., 2011), including performance on language measures (although not verbal fluency specifically), and suggested greater impairment by right-sided motor onset PD patients. This line of study may be meaningful in providing more insight into how laterality of motor symptoms may impact not only verbal fluency but cognition in general, progression of neurocognitive impairment, and development of interventional treatments.

From a clinically significance standpoint, it is possible that proneness towards error making on verbal fluency tasks may translate into everyday word finding difficulties and subjective complaints. However, this is an area that has been understudied and warrants further investigation. Hypothetically, prominent set-loss errors may relate to losing frame of thought during conversation, circumstantiality and/or tangentiality. Perseveration errors may translate to working memory difficulties (Azuma, 2004; Fischer-Baum et al., 2016) which has been suggested to be a possible marker for future cognitive decline (Hamada et al., 2021; Pakhomov et al., 2018). However, more research is warranted to investigate the clinical relevance of these hypotheses.

There are several limitations to our study in addition to the ones that have already been mentioned. Our convenient sample consisted of mostly DBS surgical patients, which are likely to be higher functioning than traditional PD/ET non-surgical samples. Chances are that if a patient is being considered for DBS, the referring provider has determined that at the very least they do not suffer from a significant/noticeable cognitive disorder. The general cognitive functioning of our sample was largely intact and mean education was higher than a high school education for both groups (Table 2). Thus, it is challenging to generalize our findings to a non-DBS seeking PD/ET sample. A second limitation is that of the general lack of diversity in our sample. The majority of patients were well educated Caucasian men. Unfortunately, disparities are quite prevalent and well-known in DBS treatment. Disparity in access to DBS treatment has been a chronic issue, as women and patients from underrepresented backgrounds are less likely to be referred for DBS than Caucasian males. This is incongruent with data that suggests that the prevalence of PD in LatinX communities is increasing, for example (Chan et al., 2014; Katz et al., 2011; Shpiner et al., 2019; Willis et al., 2014; Wright Willis et al., 2010). Thus, these limitations decrease the generalizability of our findings to the general PD/ET population.

To conclude, we have found that ET patients demonstrated lower performances on measures of phonological than PD patients, which is in line with two previous studies. Although an initial significant lower performance was also noted for ET patients on category fluency, reduced processing speed in the ET group versus the PD group appeared to attenuate this finding towards non-significance. Furthermore, when we looked at within group differences in error performance, we confirmed that PD patients are prone to making more perseverative errors and ET patients make more set-loss errors on measures of phonological fluency in contrast to category fluency. Both made more set-loss errors on a category switching fluency task in comparison to their performances in category fluency. The strength of our study is that we investigated both phonological and semantic fluency, as well as category switching fluency, which was a more expanded effort than the previous head-to-head comparison studies. This study sought to elucidate not only the importance of obtaining quantitative (total correct scores) on verbal fluency measures, but also help understand qualitative behaviors (errors) that may help inform the pathology of this behavior, especially when patients are being considered for DBS and or when a diagnosis of cognitive impairment is being considered. As more is being discovered about the cognitive deterioration and subsequent risk for mild cognitive impairment and dementia in ET patients, it is imperative that further research investigate neurocognitive functioning in both traditional (quantitative scores) and less traditional (qualitative observations) forms.