Differential cognitive substrates of verbal episodic memory performance in semantic variant primary progressive aphasia and Alzheimer’s disease

Abstract

Performance on neuropsychological measures of verbal memory requires cognitive abilities beyond memory. We examined the contribution of semantic knowledge in verbal episodic memory for semantic variant primary progressive aphasia (svPPA) or Alzheimer’s disease (AD). 415 AD and 68 svPPA participants completed measures of episodic memory (visual and verbal) and semantic knowledge. A double dissociation existed; visual recall predicted verbal recognition in AD, whereas semantic knowledge contributed to verbal recognition in svPPA.

Verbal episodic memory performance requires recruitment of multiple cognitive abilities, limiting inferences for measures that posit to capture a single construct. Casaletto et al. (2017) demonstrated that verbal immediate recall is dependent on echoic recall, auditory processing speed, and executive control. Additionally, the relative contributions of these cognitive abilities on immediate recall performance varied systematically between neurodegenerative syndromes.

An interpretive challenge is particularly salient in patients with semantic variant primary progressive aphasia (svPPA), a syndrome classified by loss of semantic knowledge. Performance on verbal memory tasks also tends to be quite poor and may appear similar to patients with Alzheimer’s disease (AD; Kramer et al., 2003). The breakdown of the semantic networks in svPPA may impede processing and integration of verbal information, resulting in impaired recall. Conversely, deficits in semantics in addition to memory have long been demonstrated in AD (Hodges & Patterson, 1995; Salmon, Butters, & Chan, 1999), thereby challenging the differential diagnosis.

Understanding the different substrates of verbal episodic memory impairment in svPPA and AD will potentially provide insight into the cognitive mechanisms underlying episodic memory. We examined the relative contributions of visual recall as a measure of non-semantic episodic memory and semantic knowledge to delayed verbal recall and recognition in svPPA or AD. We hypothesized semantic knowledge would contribute to a greater extent to verbal episodic memory performances in svPPA, while non-semantic episodic memory would contribute more strongly to verbal episodic memory performances in AD.

Methods

Participants

Participants with an MMSE score of ≥18 and a primary language of English were drawn from a large database spanning 1998–2015. Participants underwent a series of evaluations including neurobehavioral and physical examinations, informant interview, and neuropsychological testing. Four hundred and fifteen participants met ‘probable’ AD (age range 47–86) consistent with established consensus clinical research guidelines (McKhann et al., 2011). A language team determined classification of svPPA (N = 68, age range 49–86) consistent with research criteria (Gorno-Tempini et al., 2011). There were no statistically significant differences in age and education (Table 1). The svPPA group had a significantly higher proportion of male participants (56% in the svPPA group and 41% in the AD group) and higher MMSE scores (AD-svPPA = −1.9, t = −4.6, p < .001).

Table 1.

Demographic data and cognitive performance

	AD	svPPA
Age	66.7 (8.7)	64.7 (7.6)
Sex	171/415 Male*	38/68 Male*
Education	16.3 (5.2)	16.5 (2.8)
MMSE	23.8 (3.2)*	25.7 (2.9)*
CVLT-II-SF
Delayed recall	1.8 (2.2)*	2.5 (2.5)*
Delayed recall tertiles	1.9 (0.9)	2.1 (0.9)
Delayed cued recall	2.8 (2.3)	3.2 (2.7)
Verbal recognition discriminability (d’)	1.6 (0.8)	1.8 (1.1)
Benson recall	3.1 (3.4)*	7.2 (4.1)*
PPVT-R	14.4 (1.8)*	9.4 (3.6)*

CVLT-II-SF = California Verbal Learning Test – Second Edition, Short Form; MMSE = Mini-Mental Status Examination; PPVT-R = Peabody Picture Vocabulary Test – Revised.

Standard deviations are displayed in parentheses.

Bold indicates significance p = .024 (sex), p < .001 (MMSE), p = .016 (delayed recall), p < .001 (Benson recall), p < .001 (PPVT-R).

^*Values are significant at p < .05.

Neuropsychological assessment

The neuropsychological battery screened attention, language, memory, visuospatial, and executive functioning (Kramer et al., 2003). Domains of interest included visual episodic memory as a measure for non-semantic episodic memory and receptive vocabulary for semantic knowledge.

Episodic memory

The California Verbal Learning Test-Second Edition, Short Form (CVLT-II-SF; Delis, Kramer, Kaplan, & Ober, 2000), assesses verbal episodic learning and memory. Nine words of three semantic categories were presented orally at a rate of 1.5 per second across four immediate recall trials. Participants engaged in a 30-second self-generated counting interference task and without prior notification recalled the items and again after ten minutes (CVLT-II-SF verbal recall). Subsequent recall with a cue, participants indicated which items were originally presented amongst a list of new items (verbal recognition [d’]).

Visual episodic memory was assessed with the Benson figure (Kramer et al., 2003). Following visuoconstruction, participants received prompting to memorize the figure. Delayed recall after 10 minutes was scored 0–17 using a standardized classification system for accuracy and placement.

Semantic knowledge

The abbreviated Peabody Picture Vocabulary Test-Revised (PPVT-R; Dunn & Dunn, 1981) is a receptive vocabulary measure of semantic knowledge. Participants identified which of four possible pictures matched a target word with a total score of 0–16.

Statistical analysis

Performances for episodic memory (Benson figure recall, CVLT-II-SF verbal recall, and verbal recognition) and semantic knowledge (PPVT-R) were compared between groups using independent t-tests (Table 1).

Multivariable regression analyses were conducted in each group with Benson figure recall and PPVT-R simultaneously entered as independent variables and CVLT-II-SF verbal recall or recognition as the dependent variable, adjusting for age, sex, and MMSE. Given a significant positive skew, CVLT-II-SF verbal recall scores were redistributed into tertiles (coded as 1, 2, or 3).

Results

CVLT-II-SF verbal recall scores were significantly higher in svPPA (AD-svPPA = −1.7, t = −2.4, p = .016), but there was no significant difference in verbal recognition (AD-svPPA = −0.2, t = −1.4, p = .158). In AD, Benson figure recall was significantly lower (AD-svPPA = −4.0, t = −8.8, p < .001), while in svPPA, PPVT-R score was significantly lower (AD-svPPA = 5.0, t = 14.7, p < .001, Table 1).

Verbal episodic memory correlates

AD

For AD, age, sex, MMSE, Benson figure recall, and PPVT-R explained 32% of the variance in CVLT-II-SF verbal recall, F(5, 291) = 28.9, p < .001, Table 2: MMSE 18–30. Benson figure recall significantly predicted verbal recall (β = .4, p < .001), but PPVT-R did not (β = .0, p = .883). Predicting verbal recognition, age, sex, MMSE, Benson figure recall, and PPVT-R explained 18% of the variance, F(5, 288) = 13.9, p < .001. Only Benson figure recall significantly contributed to the model (β = .3, p < .001; PPVT-R: β = −.0, p = .845).

Table 2.

Standardized coefficients for verbal episodic memory correlates stratified by MMSE

	MMSE 18–30				MMSE ≥ 24				MMSE 18–23
	Verbal recall		d-prime		Verbal recall		d-prime		Verbal recall		d-prime
	292	50	289	50	159	39	158	39	133	9	131	9
n	AD	svPPA	AD	svPPA	AD	svPPA	AD	svPPA	AD	svPPA	AD	svPPA
Age	−0.2*	−0.8	−0.02	−0.17	−0.23*	−0.08	0.01	−0.26	−0.27*	−0.04	−0.04	0.78
Sex	0.06	0.08	0.08	0.19	0.12	0.08	0.05	0.15	−0.01	−0.22	0.13	0.70
MMSE	0.3*	0.31*	0.24*	0.23	0.22*	0.31*	0.27*	0.20	−0.04	−0.41	0.10	0.08
PPVT-R	0.0	0.30*	−0.01	0.51*	−0.10	0.17	−0.09	0.48*	0.13	0.83	0.08	−0.02
Benson recall	0.4*	0.50*	0.30*	0.13	0.40*	0.59*	0.26*	0.11	0.29*	0.26	0.33*	0.86

MMSE = Mini-Mental Status Examination; PPVT-R = Peabody Picture Vocabulary Test – Revised.

^*Values are significant at p < .05.

Given that immediate recall is a strong predictor of verbal recall, we conducted additional analyses with total immediate recall as a covariate. In AD, the Benson figure recall continued to demonstrate a significant relationship with verbal recall (Benson figure recall β = .3, p < .001; total immediate recall β = .2, p < .001) and verbal recognition (Benson figure recall β = .3, p < .001; total immediate recall β = .3, p < .001).

svPPA

In svPPA, age, sex, MMSE, Benson figure recall, and PPVT-R explained 54% of the variance in CVLT-II-SF verbal recall, F(5,49) = 12.6, p < .001, with both PPVT-R (β = .3, p = .020)and Benson figure recall (β = .5, p < .001) significantly contributing. For verbal recognition, age, sex, MMSE, PPVT-R, and Benson figure recall explained 37% of the variance, F(5, 49) = 6.8, p = .001, and only PPVT-R was a significant predictor (β = .5, p = .001).

Controlling for immediate recall across learning trials in svPPA revealed significant contribution of Benson figure recall (β = .4, p < .001) and total immediate recall (β = .3, p = .010), while PPVT-R did not (β = .1, p = .296). For verbal recognition, PPVT-R (β = .4, p = .023) and total immediate recall (β = .3, p = .041) significantly contributed.

Discussion

Our study demonstrates the differential contributions of impaired verbal memory in svPPA and AD. We expose a double dissociation such that visual memory, but not semantic knowledge, predicted verbal recognition in AD, whereas in svPPA, semantic knowledge but not visual memory predicted verbal recognition (Figure 1). While both visual memory and semantic knowledge accounted for verbal free recall in svPPA, only visual memory accounted for verbal free recall in AD. Moreover, we highlight the importance of extending inferences of neuropsychological test performance beyond the construct they purport to measure in neurodegenerative diseases.

Figure 1.

Semantic knowledge significantly contributes to verbal recognition in svPPA but not in AD. Unique variance of verbal recognition explained by visual recall and semantic knowledge in each group controlling for age, sex, and Mini-Mental Status Examination (MMSE). Only visual recall uniquely predicted verbal recognition in AD, while in svPPA, semantic knowledge only predicted verbal recognition. *indicates p < .001 (AD Benson recall), p = .023 (svPPA PPVT-R). ^†PPVT-R = Peabody Picture Vocabulary Test – Revised.

Our findings support that episodic memory and semantic knowledge are interconnected cognitive constructs. Level of Processing Theory provides a stage dependent mechanism for this interaction by positing that retention of stimuli is a function of depth of processing during perception (Craik & Lockhart, 1972). Initial stages involve processing physical or sensory features, while later stages match incoming stimuli with existing information obtained via past learning stored within a semantic network. Accordingly, later stages afford greater depth processing, consistent with the idea that meaningful stimuli are easier to retain. Semantic networks are affected in AD and to a greater extent in svPPA, possibly resulting in reduced depth processing. This is consistent with prior findings that semantic abilities are the strongest predictor of immediate recall in svPPA (Casaletto et al., 2017).

Our study is not without limitations. The nature of the syndromes likely manifested MMSE scores differentially, limiting severity inferences. Our conclusions may be strengthened by the addition of convergent measures of semantic knowledge and memory. Understanding longitudinal cognitive trajectories of these abilities across disease stages would enrich our understanding of stages at which to expect possible dissociations and commonalities amongst neurodegenerative syndromes. Adjusting for total immediate recall (learning trials) in our analyses, the models for AD remained significant, while the semantic knowledge effect for svPPA was attenuated for verbal delayed recall but maintained for verbal recognition, likely stemming from reduction in power due to the reduced sample size. Findings were also generally upheld when stratifying for MMSE in AD (Table 2: MMSE ≥ 24 and MMSE 18–23). Reduction in power rendered findings inconclusive for svPPA with a low MMSE.

The current study is consistent with semantic knowledge and verbal episodic memory processes in svPPA and AD. We expose a critical tenant of neuropsychological assessment: the inference of neuropsychological test scores may vary depending on the cognitive abilities needed for a given task and the clinical syndrome. This necessitates that clinicians consider alternative contributors to cognitive performance and ultimately interpretation and appropriateness of using verbal episodic memory instruments to assess episodic memory in svPPA.