Does a rabbit have feathers or fur? Development of a 42-item semantic memory test (SMT-42)

ABSTRACT

Objective

We present the preliminary study of the 42-item Semantic Memory Test (SMT-42), a test developed to distinguish semantic variant primary progressive aphasia (svPPA) from the other variants: logopenic (lPPA) and nonfluent/agrammatic (naPPA). The test requires the patient to retrieve the conceptual features of items belonging to different lexical categories.

Methods

In the first study, we administered the French version of the SMT-42 to a population of healthy subjects and to patients with svPPA matched to a subgroup of the healthy subjects. In the second study, we administered the SMT-42 to four groups of patients (with svPPA, lPPA, naPPA and Alzheimer’s disease [AD], respectively) to study its capacity to differentiate patients suffering from svPPA from the other patients.

Results

In the first study, 109 healthy subjects were included, 15 of whom were paired with 15 subjects presenting with svPPA. In the second study, designed to compare groups presenting a primary progressive aphasia variant and AD, 12 subjects with svPPA, 6 with naPPA and 9 with lPPA were included, along with 21 subjects with AD. The subjects presenting a semantic deficit were clearly distinguished from the others by their results on the SMT-42 (svPPA: mean = 30.0 (5.9); lPPA: mean = 37.8 (3.3), d = 1.5, p = 0.002; naPPA: mean = 39.8 (1.9), d = 1.89, p = 0.001; AD: mean = 38.5 (2.4), d = 1.63, p < 0.001); (svPPA: median = 31; lPPA: median = 38, U = 9, p = 0.002; naPPA: median = 40.5, U = 1.5, p = 0.001; AD: median = 39, U = 13.5, p < 0.001).

Conclusion

The SMT-42 is simple, rapidly administered (3 minutes on average), easily scored and has good sensitivity, and it appears to be an effective tool for semantic screening in routine clinical practice.

KEYWORDS:

• Semantic variant
• primary progressive aphasia
• test
• semantic memory
• screening

1. Introduction

1.1. Theoretical framework

Since the seminal work of Tulving (1972) and subsequent works, it has been customary to divide long-term memory into two components: episodic memory and semantic memory. These two systems differ in their procedures and qualities, as illustrated by the neuropsychological data from studies on amnesic patients (Wheeler & Mcmillan, 2001), but nevertheless maintain strong links (Renoult et al., 2019; Thomas-Antérion et al., 2010), particularly in the development of autobiographical memory (Ferrey, 2008). Semantic memory is the general, non-personal knowledge that an individual has about the world around him or her, such as the meaning of words and information about animals, objects, plants and people (Belliard et al., 2007). This knowledge may concern concepts, unique entities and personal biographical informations (from the moment they become decontextualized, such as date of birth, profession and place of residence) (Laisney et al., 2009; Van der Linden, 2003). In our study, we focused exclusively on words and the concepts to which they refer.

The semantic system lies at the heart of lexical processing. It is involved in many linguistic activities, such as naming, comprehension and identification (Hillis & Caramazza, 1995; Mion et al., 2010). Its organization is still debated (for a review, see Rossi, 2016), and we can consider different factors characterizing it. Frequency is classically recognized as a significant effect, with low-frequency words posing more difficulty for subjects with neurodegenerative disorders such as Alzheimer’s disease (AD) (Goldstein, 1992; Hodges et al., 1992; Montanes et al., 1995; Rogers et al., 2007). Word recognition is sensitive to this effect (Monsell, 1991). The other frequently mentioned factor is the abstraction effect; indeed immaterial concepts generally pose more difficulty than concrete concepts (Franklin, 1989), but the reverse has been observed (Breedin et al., 1994; Macoir, 2009; Papagno et al., 2009).

Several studies underline the categorical effect traditionally identified in specific deficits resulting from herpes simplex encephalitis or head injury but which can also be observed in patients with Alzheimer’s type dementia. The living/non-living dissociation is classically in favor of the category of manufactured products (Fung et al., 2000; Thompson-Schill, 1999), the reverse being unusual (Laiacona & Capitani, 2001; Moss & Tyler, 2000). Other dissociations are observed, such as verbs/nouns (Caramazza & Hillis, 1991), objects/actions (d’Honincthun & Pillon, 2008), animals/tools, actions (Ferreira et al., 1997), living things/precious stones (Warrington & Shallice, 1984), body parts/artifacts (Warrington, 1975), semantic knowledge of objects/persons (Gainotti et al., 2008), or compared to knowledge related to numerals (Thioux et al., 1998), as well as more specific deficits, such as animals (Caramazza & Shelton, 1998), plants (Samson & Pillon, 2003) or parts of the body (Goodglass & Budin, 1988). Certain features can be selectively affected, such as those relating to the visual shape of objects (Lambon Ralph et al., 1998) or those relating to colors (Miceli et al., 2001). In this paper, we focus on the two variables most documented in terms of their involvement in word comprehension: frequency and the living/non-living dissociation.

1.2. Semantic memory deficits

In normal aging, knowledge seems to be maintained but would be sensitive to diminished executive and attentional resources; post-semantic connections would likewise be weakened (Mathey & Postal, 2008). Thus, naming and lexical activation are gradually reduced with age, while concepts do not vary, as demonstrated by tests of definition or general knowledge (Giffard, 2001).

Semantic memory impairment is characteristic of semantic variant primary progressive aphasia (svPPA or fluent PPA). svPPA is a rare pathology with a reported prevalence of 0.8/100,000 in the French population (Magnin et al., 2016). While the incidence rate of svPPA is not known, a recent study based on a cohort of patients with primary progressive aphasia (PPA) reported an incidence rate of PPA of 1.14 per 100,000 person-years in France (Mouton et al., 2022). svPPA differs from logopenic (lPPA) and nonfluent/agrammatic (naPPA) variants. According to the criteria proposed by Gorno-Tempini et al. (2011) and Mesulam et al. (2012), PPA is characterized by a language deficit in the foreground, setting in gradually and explaining the perceived difficulties in daily life. lPPA implies a lack of the word as well as a repetition disorder with a length effect; it is associated with phonological disturbances, preservation of lexical comprehension, motor aspects of speech and syntax. naPPA is characterized by agrammatism and/or apraxia of speech (motor aspect of speech) and a disorder of syntactic comprehension, while comprehension of isolated words and knowledge of objects remain preserved. In sum, one of the discriminating points of svPPA is the preservation of comprehension of isolated words and knowledge of concepts. This initially insidious and progressively worsening pathology manifests itself by word-finding difficulties in a constrained situation or in spontaneous language, disorders of lexical comprehension which sometimes provoke a feeling of strangeness when the patient cannot access the meaning, a surface alexia–agraphia marked by regularizations (developmental dyslexia may be present). The flow remains fluent but punctuated by anomia – generally not improved by phonemic cues –, circumlocutions and semantic paraphasias, categorical fluency being less efficient than literal fluency or verbal fluency. Other disorders may also be associated as the disease progresses (e.g., visuo-constructive agnosia, acalculia and apraxia) but remain in the background (Mesulam et al., 2014). In general, when semantic skills are not at stake, performances on tests of perception, episodic memory, working memory, problem-solving and executive skills are preserved or only slightly affected and orientation in time and space is rather preserved (Rogers et al., 2007). Behavioral disorders may arise, such as rigidity, stereotypies, compulsive behavior, avarice, egocentrism, lack of empathy, logorrhea and social disinhibition. The atrophy of the anterior temporal lobe, the “hub” of living entities, supports the hypothesis of a specific impairment of these entities, with non-living entities sitting more posteriorly (Vallet et al., 2011). The characteristic alteration of visual integration highlighted by Vallet et al. (2011) explains why concrete entities are more massively affected than abstract ones. The familiarity effect is significant (Gorno-Tempini et al., 2011) and other factors may explain the selective impairment: the class of the object, the frequency of encounter, the length of contact, the interactions exerted by the subject on the object, the context. Cognitive egocentrism may also be observed: as episodic memory is preserved at the beginning of svPPA, the patient relies on it, using personal knowledge to access more collective knowledge. In this way, patients find it easier to identify personal objects, places visited and people close to them. This can lead to an “egolalia” in the naming test, in other words, an egocentric discourse (“this, I know; I have one at home, I use it every morning”). The usual gradient of forgetfulness is then reversed, with semanticised memories deteriorating before others (Belliard et al., 2007).

The literature has for some time been suggesting early lexico-semantic impairment in AD (Amieva et al., 2008; Didic et al., 2011). The progression of disorders follows the hierarchical axis, from the most specific to the most general, with knowledge at a superordinate level being retained until an advanced stage (Chertkow, 1989; Simoes Loureiro et al., 2018). This results in word-finding difficulties, which can be compounded by gnostic disturbances and a rapid decline of categorical fluency (Auriacombe et al., 2006). It is commonly accepted firstly that the deterioration of semantic fluency precedes and is more severe than that of literal fluency (Chasles et al., 2019; Salmon et al., 1999) and secondly that natural entities are more massively affected than manufactured ones. Manufactured entities are reported to degrade steadily, while natural entities experience a more pronounced decline only at a moderate stage (Chainay, 2005). In AD, lack of access precedes progressive storage deterioration (Cardebat et al., 1995), and the two may be concomitant (Laisney et al., 2010); the disease evolves progressively toward a global loss of knowledge (Beaunieux et al., 2002) and therefore of comprehension, especially lexical. The distinction between access disorder (the precise word is not retrieved due to a deficit in activation mechanisms and/or knowledge selection, but the meaning of the concept is preserved) and storage disorder (deterioration of semantic representations) can be based on Shallice’s criteria (Shallice, 1988), such as the consistency of errors on the same items from one test to the next and at a distance, or the particular impairment of attributes (versus preserved general knowledge) in favor of a central impairment. The frequency of the item and the time interval between the response and the stimulus can also be considered.

Semantic memory encompasses the memory of words, symbols, concepts and their relationships, general knowledge, famous people, public places and events. It gives meaning to our surroundings and actions (Bier & Macoir, 2010) and is involved in our social interactions (Simmons & Martin, 2009). Word comprehension (meaning the lexico-semantic function of access to concepts and by definition to their respective characteristics), as a component of a more global function, can participate in the exploration of this memory. The 42-item Semantic Memory Test (SMT-42) aims to explore the integrity of certain concepts by questioning their characteristics, probing the semantic network by looking for a deficit in the understanding of isolated words characteristic of svPPA. It is more a lexico-semantic task which questions the semantic knowledge related to words than a memory task.

1.3. Exploring semantic memory

Several types of exploration can be used to understand semantic memory and lexico-semantic disorders: neurolinguistic paradigms, physiological measurements, imaging, etc. (for a review, see Rossi, 2016). Various batteries explore semantic memory from different aspects (Table 1).

Table 1. Examples of tests dealing with semantic memory.

Domain	Test	Reference	Comments
Specific contents of the semantic features of the lexical network	Batterie d’évaluation des connaissances sémantiques du GRECO (BECS-GRECO)	Merck et al., 2011	This battery accurately explores semantic memory with relevant subtests (naming, semantic matching, semantic questionnaire, identity matching); administering the complete battery takes 2–3 hours.
	Size and Weights Attribute Test (SWAT)	Warrington & Crutch, 2007	The SWAT compares attribute knowledge for animals and objects in Alzheimer’s disease (AD) and semantic dementia patients to verify the existence of specific deficits (modalities and categories) in these two populations. This test minimizes the involvement of other cognitive functions (reasoning, executive abilities, etc.) by providing, for each trial, only one attribute and one category. The SWAT is an experimental test. At the time of publication, it had not been validated for standard clinical use.
Conceptualization via similarities	Pyramids and Palm Trees Test (PPTT)	Howard & Patterson, 1992	The PPTT is based on the matching of images or words. The semantic control load is substantial and denotes significant dependence on executive functions.
	Kissing and Dancing Test (KDT)	Bak & Hodges, 2003
	Wechsler Adult Intelligence Scale IV (WAIS IV) similarities subtest	Wechsler, 2011
	Camel and Cactus Test	Bozeat et al., 2000
Semantic knowledge questionnaire	The Mini-SKT (semantic knowledge questionnaire)	Simoes Loureiro et al., 2018	These are multiple-choice questions that ask about the semantic characteristics of objects. This test is designed for AD and presents a sensitivity of 61.5% and a specificity of 89.5%.
Functional and categorical matching or categorization	Protocole d’Evaluation des Gnosies Visuelles Montréal-Toulouse (PEGV)	Agniel et al., 1992	The test explores only visual material to detect agnosias; association types and distractors are not controlled.
	LEXIS	Tran & Godefroy, 2011	This battery includes an image matching sub-test. Semantic control is necessary.
Agnosia as it applies to famous people or events	TOP 30	Thomas-Antérion & Puel, 2006	These tests explore the semantic knowledge specific to well-known personalities via image (face) or personality name matching and/or questionnaires. They require frequent and necessary updating because the material used is dependent on the changing culture. The semantic questionnaire subtests are dependent on the level of prosopagnosia.
	Battery 75	Pluchon & Simonet, 2007
	TOP 12	Lacot et al., 2011
	SemPer	Laisney et al., 2009
Memory of public events	EVE 30	Thomas-Antérion & Puel, 2006	These batteries, which involve recounting memories of public events, require good discourse elaboration and organization skills, which may bias the results for low socio-cultural levels. Lack of the word and lack of interest in current events are barriers.
	EVE 10	Thomas-Antérion et al., 2006
Lexical generation task	Semantic fluency	Cardebat et al., 1990	Semantic fluency allows us to evaluate the integrity of the semantic network and its lexical/semantic organization, the strategies mobilized for the retrieval of words in long-term memory and certain executive functions such as spontaneous cognitive flexibility. The executive load is thus notable, with a possible sensitivity to initiation disorders.
Lexical knowledge or synonymy	Mill-Hill	Raven & Deltour, 2007	Both subtests (word definition and synonymy tests) require a high semantic control load.
	Wechsler Adult Intelligence Scale IV (WAIS IV) definitions subtest	Wechsler, 2011	These word definition tasks (as well as the Mill-Hill synonymy subtest) involve a high semantic control load. They pose the problem of discourse elaboration, which induces a high linguistic load in terms of expression.
Biographical information	Personal semantic verbal fluency	Piolino, 2006	These tests concern a very specific component of semantic memory, autobiographical memory, which does not reflect lexical understanding.
	Autobiographical questionnaire	Kopelman et al., 1989
Naming based on images, verbal description, auditory stimuli	Test de dénomination orale d’images 80 items (DO80)	Deloche & Hannequin, 1997	Naming is an act that involves more post-semantic processes (access to the phonological output lexicon, phonological buffering, programming of articulatory patterns) compared to designation for example. An observed deficit will require further exploration to determine the origin of the disorder (pre-semantic activation, semantic system, post-semantic activation).
	DENO100	Kremin et al., 2005
	Batterie Informatisée du Manque du Mot (BIMM)	Gatignol et al., 2007
Designation of images	LEXIS	Tran & Godefroy, 2011	This type of word-picture matching test does not always distinguish agnosia from a lexical comprehension disorder. Other tests will be needed to complete the exploration (as recommended by the authors). Picture designation is now recognized as less effective than a word verification test.

There is also the Gremots battery (Bézy et al., 2016), developed to describe language disturbances specific to neurodegenerative pathologies, which includes an oral and written word verification test. In terms of semantic processing, naming and fluency are considered to be the tip of the iceberg; they are the result of different underlying semantic processing. Some batteries, such as the BECS-GRECO (Merck et al., 2011), include several aspects, while other tests treat them in a more specific way, each one trying to isolate a particular component (semantic features, general knowledge about events or people, similarities, gnosias, conceptualization, lexical generation test, biographical elements, comprehension, etc.).

However, none of these batteries manages to combine a sufficient number of criteria, such as accessibility of instructions, validity, ease of administration, guarantee of exploration of an isolated process and consideration of interfering language disorders, all within a limited time compatible with the constraints of consultations. The SMT-42 meets this objective: it is a French-language tool that is quick and easy to use. It explores components of some concrete concepts in order to detect early impaired lexical comprehension and/or deterioration of conceptual features characteristic of svPPA and thus distinguish it more easily from lPPA and naPPA. It makes it possible to reveal at least a fragmentary deficit in the knowledge linked to certain concepts, without guaranteeing their complete integrity in the case of a correct response or their complete disintegration in the case of a wrong response. However, it has been reported that semantic features progressively and prematurely deteriorate (Simoes Loureiro et al., 2018). Obtaining incorrect responses can reveal this possible early degradation and prompt further exploration. The SMT-42 was first developed to distinguish svPPA from other PPA variants based on the key feature of impaired lexical understanding. As discussed earlier, this disorder can be found in patients with AD (for details of the differences between svPPA and AD in terms of semantic memory, see for instance Libon et al., 2013); thus, the SMT-42 can quite possibly be used in AD to determine the level of impairment. The aim of this preliminary prospective study was to test the performance of this new tool in revealing lexico-semantic impairment, a disorder characteristic of svPPA and which can be found in AD.

2. Methods

2.1. Presentation of the SMT-42

The SMT-42 was developed in French; to date, it has not yet been validated in English or other languages. The arguments we develop refer to the components of the French language version (especially for the frequencies of the items used). The test administration form is presented in Appendix C and the instructions in Appendix D (translated into English from the French version).

Based on data derived from the literature, we opted for a concise and easily administered multiple-choice test format. The aim was to target semantic features directly and avoid interference from more complex processes, such as similarity, in order to directly explore any degradation of concepts. The SMT-42 is designed to highlight a deficit in linguistic knowledge by detecting a disturbance in the semantic network. It is not only a question of verifying whether a word is understood, but more particularly of determining whether the proposed feature is relevant to the proposed concept. We chose the most relevant data, such as frequency and the distinction living things/artifacts. We targeted subordinate entities as these are the first affected. A written and verbal format was chosen for its sensitivity, as a visual support in the form of an image would have induced the answers. The written material is read to the patient in order to avoid a bias related to the reading disorders frequently encountered in svPPA. Moreover, this approach helped to avoid potential gnostic disturbances.

To choose the items, we used the data developed previously in the theoretical framework. Thus, we discarded the prototypes and retained subordinate items, at the same level on the taxonomic axis in order to gain in sensitivity. These items shared common characteristics and were considered essential for the constitution of the concepts; they were used as cues during the test. In a first version of the test, some items were found to be confusing and were therefore discarded (e.g., the word “strawberry” posed a problem of confusion between seeds and pips, and the word “clementine” did not achieve a sufficient consensus). The current form of the SMT-42 is the result of this preselection.

Frequencies were determined from the Lexique 3.8 database (www.lexique.org), using the frequency of lemmas according to the corpus of film subtitles (freqlemfilm2; per million occurrences). For the purposes of our study, we defined low frequency items as having an occurrence of less than 10/1,000,000. We obtained an average of 25.55 for high frequency items and 1.63 for low frequency items, representing a differential of 23.92. The characteristics of the items are shown in Appendix A.

2.2. Protocol

Each subject performed a battery of tests in the following order: Mini Mental State Examination (Folstein et al., 1975) GRECO version; SMT-42; the Stroop (Albaret & Migliore, 1999), a French written version of the Pyramids and Palm Trees Test (PPTT) (Callahan et al., 2010; Howard & Patterson, 1992); and the Test de dénomination orale d’images 80-items (DO80), a picture naming test considered as a gold standard in French and used in aphasiology with simple line drawings (Deloche & Hannequin, 1997). Since pathological subjects are likely to have a dysexecutive syndrome, which could interfere with performance on the SMT-42, the Stroop was proposed to control for its influence. Patients with svPPA have a global semantic deficit which naturally translates into a lexical disorder, so we used a picture naming test, the DO80, to verify this and to support the hypothesis of a wider semantic disturbance. As the PPTT is considered the gold standard for exploring semantic matching deficits, the SMT-42 should logically show a correlation. The following information on subjects was collected: age, sex and socio-cultural level established according the norms used for the GREFEX study (Godefroy & GREFEX, 2012).

2.3. Participants and inclusion criteria

This study was designed and conducted in accordance with the Declaration of Helsinki. All participants gave their written informed consent. We conducted two successive studies. The first study focused on the characteristics of the SMT-42 when performed by a population of healthy subjects. The performances of a sub-group derived from this population were compared to those of patients with svPPA.

The second study compared the performances of a second group of patients with svPPA to groups presenting pathologies affecting language and a group of patients with typical AD.

All participants completed the tests in their entirety; there were no dropouts.

2.3.1. First study

The control subjects for the first study were recruited during 2013, mostly from sheltered housing, recreational associations or senior citizens’ clubs. Others were recruited at the memory clinic from among caregivers of patients suffering from neurological diseases. Subjects had to meet the following inclusion criteria: (1) MMSE score ≥ 25, (2) sufficient knowledge of written French, (3) native French speaker, (4) normal eyesight and hearing, (5) absence of alcoholism or other addiction other than smoking, (6) no known neurological disorder, (7) absence of any major physical problems, (8) absence of any psychiatric disorder, (9) absence of any general anesthesia in the month prior to testing, (10) no neuropsychological evaluation within the previous three months. The purpose of this study was to validate the relevance of the SMT-42 in terms of its construction and establish concordance with data in the literature. Once this had been done, we were then able to compare these results with those of a pathological population in order to verify the sensitivity of the test. We also studied interrater reliability.

Patients with svPPA were recruited at the Nantes Memory Resource and Research Center (CMRR Nantes) in 2013, for the vast majority in the context of a referral to confirm the diagnosis. They underwent neuropsychological evaluation, imaging and/or lumbar puncture, which enabled the diagnosis to be made according to the criteria of Gorno-Tempini et al. (2011). Depending on the background and profile of each individual, the tests proposed varied but covered all the necessary domains (global assessment of MMSE type, naming, verbal fluency, reading, writing, speech, syntax, phonology, semantic memory, verbal and/or visual episodic memory, visuo-constructive and gestural praxis, gnosis, autonomy, among others). For this research, we standardized the protocol, and no correlation was therefore looked for between the scores of the initial diagnosis and those of the SMT-42. Only subjects with an MMSE score higher than 12/30 were included. This relatively low cutoff was necessary to ensure at least a minimum capacity to interact, while not excluding subjects whose score could fall markedly, as is the case for all subjects suffering from pathologies with a language component. In the second part of this study, the performances of a sub-group of the control subjects were compared to those of patients with svPPA, matched in terms of age, sex and socio-cultural level (control subjects were randomly selected from those with similar demographic characteristics), the aim being to show a significant difference between the two groups, independently of these three adjustment variables.

2.3.2. Second study

The results of the first study enabled us to apply the same protocol to a new group of svPPA patients and various control groups to evaluate its discriminatory value, in particular with regard to different PPA variants.

Patients presenting lPPA, naPPA or AD were recruited at CMRR Nantes during 2015. They underwent a neuropsychological evaluation, imaging and/or lumbar puncture, which enabled the diagnosis to be made according to the criteria of Gorno-Tempini et al. (2011). All subjects were required to have an MMSE score ≥ 18 to avoid overly advanced stages that would prevent testing (Chow et al., 2006; Pasquier et al., 1999). AD subjects had to be first-time consultants (< 75 years) and PPA subjects had to have been recently diagnosed (< 1 year) to avoid any with an evolved clinical picture.

2.4. Statistical analysis

Qualitative variables were expressed as number (N) and percentage (%) for each modality. Quantitative variables were expressed as mean and standard deviation (sd) or as median and range (minimum and maximum; med [min-max]) for the various scores. To compare the different groups of patients, we used a Chi 2 (χ²) test for qualitative variables or Fisher’s exact test (H) if the sample was too small; for quantitative variables, Student’s t test and ANOVA (t), or Mann-Whitney Wilcoxon (U) and Kruskal–Wallis tests (H) were used according to the size of the group. A Pearson correlation test was used to compare the SMT-42 score with the Stroop, DO80 and PPTT scores. Receiver operating characteristic (ROC) curves were generated and their corresponding areas under the curve (AUCs) were calculated to study the diagnostic capacity of the SMT-42. The thresholds presented were selected according to the Youden index. Cohen’s d was used to calculate effect sizes.

3. Results

3.1. First study

In the first study, 109 of 160 healthy participants met the inclusion criteria (Table 2). Included subjects were classified by profile according to three independent variables: sex (59 men, 50 women), age (14 subjects aged 40–54 years, 41 subjects aged 55–69 years and 54 subjects aged 70 years and older), and socio-cultural level (SCL) based on the standards used for the GREFEX study (Godefroy & GREFEX, 2012) (37 subjects with ≤ 9 years of education for SCL1; 45 subjects with 9–11 years of education for SCL2; and 27 subjects with ≥ 12 years of education for SCL3). Fifteen of the included healthy subjects were selected to be matched with 15 svPPA patients (Table 3). The results are shown in Appendix B.

Table 2. First study: SMT-42 results of 109 healthy subjects, by age category, sex and socio-cultural level (SCL).

		SMT-42 total score				SMT-42 total time
		mean	sd	t^1	p	mean	sd	t^1	p
Total		40.59	1.48			121.86	28.28
Age^1	(40–54 years)	41.14	1.17			102.79	14.89
	(55–69 years)	40.56	1.64	−1.27	0.206	114.41	25.73	1.43	0.155
	(70–90 years)	40.46	1.41	−1.53	0.128	132.46	28.67	3.77	< 0.001
Sex	M	40.56	1.48			123.19	30.41
	F	40.62	1.50	0.21	0.832	120.30	25.76	−0.53	0.598
SCL^2	1	39.95	1.54			130.57	26.44
	2	40.56	1.53	2.01	0.047	124.00	30.16	1.10	0.274
	3	41.52	0.58	4.55	< 0.001	106.37	21.21	−3.56	< 0.001

SCL1: ≤ 9 years of education; SCL2: 9–11 years of education; SCL3: ≥ 12 years of education. SMT-42: 42-item semantic memory test.

¹F(2, 106) = 1.19, p = 0.31, eta-squared = 0.022

²F(2, 106) = 10.37, p < 0.001, eta-squared = 0.196

Note: Age group 40–54 years, sex group male group and SCL group 1 are the respective reference groups.

Table 3. First study: group of 15 svPPA subjects and group of 15 healthy subjects matched for age, sex and socio-cultural level.

	Healthy subjects (N = 15)		Pathological subjects (N = 15)		Effect size	p^a
	mean (sd)	med [min-max]	mean (sd)	med [min-max]
SMT-42 total score	40.8 (1.0)	41 [39–42]	30.5 (5.6)	31 [18–39]	−1.56	< 0.001
Total timeNCEN_A_213308	116.1 (24.8)	111 [87–164]	300 (165.1)	241 [124–761]	1.23	< 0.001
High frequency items: total	21.0 (0.0)	21 [21–21]	17.6 (2.9)	17 [10–21]	−1.29	< 0.001
Low frequency items: total	19.8 (1.0)	20 [18–21]	12.9 (3.3)	12 [8–18]	−1.62	< 0.001
Difference high/low freq.	1.2 (1.0)	1 [0–3]	4.7 (2.6)	5 [1–10]	1.32	< 0.001
Total for living items	17.4 (0.6)	17 [16–18]	12.0 (4.3)	12 [2–18]	−1.33	< 0.001
Living items: percentage	0.9667 (0.0351)	0.9 [0.9–1]	0.7 (0.2)	0.7 [0.1–1]	−1.33	< 0.001
Total for non-living items	23.4 (0.6)	23 [22–24]	18.5 (2.6)	19 [15–22]	−1.57	< 0.001
Non-living items: percentage	0.9750 (0.0264)	1 [0.9–1]	0.8 (0.1)	0.8 [0.6–0.9]	−1.57	< 0.001
Difference living/non-living	6.0 (0.8)	6 [5–8]	6.5 (4.3)	6 [−1-14]	0.18	0.931

SMT-42: 42-item semantic memory test.

¹Mann-Whitney Wilcoxon tests.

Neither age (41.14 (1.17) vs 40.56 (1.64), t(53) = −1.27, p = 0.21; 41.14 (1.17) vs 40.46 (1.41), t(66) = −1.53, p = 1.13) nor sex (40.56 (1.48) vs 40.62 (1.50), t(107) = 0.21, p = 0.83) had an effect on the SMT-42 total score. However, a significant difference was observed in terms of socio-cultural level. Thus, subjects in the SCL1 group had a lower average score than subjects in the SCL2 group (39.95 (1.54) vs 40.56 (1.53), t(80) = 0.30, p = 0.047). Furthermore, subjects in the SCL1 group had a lower score than subjects in the SCL3 group (39.95 (1.54) vs 41.52 (0.58), t(62) = 4.55, p < 0.001) and subjects in the SLC 2 group had a lower score than subjects in the SLC 3 (40.56 (1.53) vs 41.52 (0.58), t(70) = 2.90, p = 0.001).

In terms of total time (in seconds) to complete the SMT-42, we found a significant difference between subjects aged 40–54 years and subjects aged 70 years and older (102.79 (14.89) vs 132.46 (28.67), t(66) = 3.77, p < 0.001), and between subjects aged 55–69 years and subjects aged 70 years and older (114.41 (25.73) vs 132.46 (28.67), t(93) = 3.22, p = 0.001); we also found a significant difference between subjects in the SCL1 group and subjects in the SCL3 group (130.57 (26.44) vs 106.37 (21.21), t(62) = −3.56, p < 0.001) and between subjects in the SCL2 group and subjects in the SCL3 group (124.00 (30.16) vs 106.37 (21.21), t(70) = −2.69, p = 0.008).

The score for low frequency items was on average lower than the score for high frequency items (19.72 (1.30) vs 20.87 (0.43), t(108) = −9.67, p < 0.001).

The score for living items was on average lower than the score for non-living items (17.47 (0.75) vs 23.12 (1.04), t(108) = −55.8, p < 0.001) (see Appendix B).

We also found a moderate but significant correlation between the SMT-42 score and the DO80 score (coefficient = 0.25, CI95% [0.06; 0.42], p = 0.009) and between the SMT-42 score and the Stroop score (coefficient = 0.25, CI95% [0.06; 0.42], p = 0.010). The correlation between the SMT-42 score and the PPTT score was higher and was also significant (coefficient = 0.45, CI95% [0.28; 0.59], p < 0.001).

We found a significant difference between healthy subjects and pathological subjects for all scores except for the difference between living and non-living items.

Within the pathological subjects group, there was a significant difference between the score for living items and the score for non-living items (median of 12 vs 19, U = 1.5, p < 0.001) and between the score for high frequency items and the score for low frequency items (median of 17 vs 12, U = 120, p < 0.001). The difference between the low frequency and the high frequency scores was greater in the pathological subjects than in the healthy subjects (median of 20 vs 12, U = 0.5, p < 0.001; median of 21 vs 17, U = 22.5, p < 0.001).

The SMT-42 protocols of 10 healthy subjects were scored by two speech therapists, blinded to the subjects’ clinical status. The results obtained by these examiners were identical to those of the initial examiner; correlation between the results was perfect (r = 1).

All healthy subjects had an SMT-42 total score equal to or higher than 39, whereas all pathological subjects had a score equal to or less than 39. Setting a discriminating threshold score of <39 gave the test a sensitivity of 93% and a specificity of 100% (Figure 1).

Figure 1. First study: diagnostic performance of the SMT-42 (boxplot and ROC curve).

SMT-42: score at the 42-item semantic memory test.

3.2. Second study

In the second study, 48 patients were included and were assigned to groups as follows: svPPA group (n = 12), naPPA group (n = 6), lPPA group (n = 9) and AD group (n = 21). The patients’ demographic characteristics are shown in Table 4.

Table 4. Second study: description of the 48 patients.

		Total (n = 48)	svPPA (n = 12)	lPPA (n = 9)	naPPA (n = 6)	AD (n = 21)	p
		N (%)	N (%)	N (%)	N (%)	N (%)
Age (years)	(40–54)	3 (6.3)	1 (8.3)	1 (11.1)	0 (0.0)	1 (4.8)	0.992^1
	(55–69)	26 (54.2)	6 (50.0)	5 (55.6)	3 (50.0)	12 (57.1)
	(70–90)	19 (39.6)	5 (41.7)	3 (33.3)	3 (50.0)	8 (38.1)
SCL	1	8 (16.7)	1 (8.3)	1 (11.1)	0 (0.0)	6 (28.6)	0.785^1
	2	21 (43.8)	6 (50.0)	4 (44.4)	3 (50.0)	8 (38.1)
	3	19 (39.6)	5 (41.7)	4 (44.4)	3 (50.0)	7 (33.3)
MMSE median [min – max]		23 [18–29]	23 [18–28]	24 [19–29]	23 [19–29]	23 [18–29]	0.638^2

SCL: socio-cultural level (SCL1: ≤ 9 years of education; SCL2: 9–11 years of education; SCL3: ≥ 12 years of education); svPPA: semantic variant primary progressive aphasia; lPPA: logopenic primary progressive aphasia; naPPA: nonfluent/agrammatic primary progressive aphasia; AD: Alzheimer’s disease; MMSE: Mini-Mental State Examination.

¹Fisher exact tests, ² Kruskal-Wallis rank sum test

There were no differences between the 4 groups in terms of age (p = 0.99), SCL (p = 0.78) and MMSE score (H = 1.69, p = 0.64).

The comparison between svPPA patients and patients with other pathologies (lPPA, naPPA or AD) showed significant differences for each of the scores and for response time (Table 5).

Table 5. Second study: comparison of scores according to the different pathologies.

	svPPA (n = 12)		lPPA (n = 9)				naPPA (n = 6)				AD (n = 21)				lPPA + naPPA + AD (n = 36)
	mean (sd)	Med [min-max]	mean (sd)	Effect size	Med [min-max]	p^1	mean (sd)	Effect size	Med [min-max]	p^1	mean (sd)	Effect size	Med [min-max]	p^1	mean (sd)	Effect size	Med [min-max]	p^1
SMT-42 total score	30.0 (5.9)	31 [18–37]	37.8 (3.3)	1.50	38 [32–42]	0.002	39.8 (1.9)	1.89	40.5 [36–41]	0.001	38.5 (2.4)	1.63	39 [34–42]	< 0.001	38.5 (2.6)	1.64	39 [32–42]	< 0.001
SMT-42 total time	259.4 (117.3)	234.5 [135–586]	171.2 (51.8)	−1.04	169 [96–237]	0.021	224.2 (76.3)	−0.42	222 [111–325]	0.815	150.9 (41.2)	−1.28	154 [99–253]	< 0.001	168.2 (56)	−1.08	160.5 [96–325]	0.001
High frequency items: total score	17.7 (2.8)	18 [13–21]	20 (1.3)	1.17	20 [17–21]	0.064	20.2 (1.4)	1.25	20.5 [19–21]	0.076	19.8 (1.3)	1.07	20 [16–21]	0.034	19.9 (1.2)	1.13	20 [16–21]	0.013
Low frequency items: total score	12.3 (3.4)	13 [5–17]	17.8 (2.2)	1.54	18 [15–21]	< 0.001	19.7 (1.4)	2.07	20 [17–21]	< 0.001	18.7 (1.6)	1.79	19 [15–21]	< 0.001	18.6 (1.8)	1.77	19 [15–21]	< 0.001
Difference between high and low freq. scores	5.3 (2.1)	5 [3–9]	2.2 (1.5)	−1.28	2 [0–4]	0.002	0.5 (1.4)	−1.99	1 [−2-2]	< 0.001	1.1 (1.5)	−1.72	1 [−2-5]	< 0.001	1.3 (1.6)	−1.66	1 [−2-5]	< 0.001
Living items: total score	12.2 (3.0)	13 [8–17]	16.7 (1.4)	1.64	17 [14–18]	0.001	17.2 (1.4)	1.82	17.5 [16–18]	0.002	16.4 (1.5)	1.55	17 [14–18]	< 0.001	16.6 (1.4)	1.62	17 [14–18]	< 0.001
Living items: percentage	68.1 (16.8)	72.2 [44.4–94.4]	92.6 (7.9)	1.64	94.4 [77.8–100]	0.001	95.4 (5.5)	1.82	97.2 [88.9–100]	0.002	91.3 (8.5)	1.55	94.4 [77.8–100]	< 0.001	92.3 (7.9)	1.62	94.4 [77.8–100]	< 0.001
Total non-living	17.8 (3.8)	18.5 [10–22]	21.1 (2.1)	1.15	21 [18–24]	0.034	22.7 (1.4)	1.68	23 [20–24]	0.003	22 (1.3)	1.47	22 [19–24]	< 0.001	21.9 (1.6)	1.43	22 [18–24]	< 0.001
Non-living items: percentage	74.0 (15.7)	77.1 [41.7–91.7]	88.0 (8.7)	1.15	87.5 [75–100]	0.034	94.4 (5.7)	1.68	95.8 [83.3–100]	0.003	91.9 (5.5)	1.47	91.7 [79.2–100]	< 0.001	91.3 (6.6)	1.43	91.7 [75–100]	< 0.001
PPTT score	38.7 (6.3)	39.5 [26–49]	44.9 (3.8)	1.07	45 [40–50]	0.014	46.8 (3.7)	1.41	48 [40–50]	0.011	49.1 (2.5)	1.80	50 [44–52]	< 0.001	47.7 (3.5)	1.55	49 [40–52]	< 0.001
DO80 score	39.9 (21.9)	42.5 [11–78]	73.6 (2.7)	1.79	73 [69–78]	0.001	76.7 (3.3)	1.95	78 [70–79]	0.003	74.0 (5.3)	1.81	76 [59–80]	< 0.001	74.3 (4.6)	1.83	76 [59–80]	< 0.001

svPPA: semantic variant primary progressive aphasia; lPPA: logopenic primary progressive aphasia; naPPA: nonfluent/agrammatic primary progressive aphasia; AD: Alzheimer’s disease; SMT-42: 42-item semantic memory test; PPTT: Pyramids and Palm Trees Test; DO80: Test de dénomination orale d’images 80 items. The critical probabilities presented in this table correspond to the comparison of the score for each of the lPPA, naPPA and AD groups with the score for the svPPA group.

¹Mann-Whitney Wilcoxon tests.

The SMT-42 total score of patients in the svPPA group was significantly lower than that of patients in each of the other groups (median of 31 vs 38, U = 9, p = 0.002; median of 31 vs 40.5, U = 70.5, p = 0.001; median of 31 vs 39, U = 13.5, p < 0.001). The total response time was significantly higher in the svPPA group than in the lPPA (median of 234.5 vs 169, U = 87, p = 0.021) and AD groups (median of 234.5 vs 154, U = 225, p < 0.001).

The scores for living items (median of 13 vs 19, U = 10, p < 0.001) and for low frequency items (median of 13 vs 17, U = 38, p < 0.001) were significantly lower in svPPA patients than in patients in the other groups.

The score quantifying the frequency difference (i.e., difference between the score for high frequency items and the score for low frequency items) was significantly higher for patients in the svPPA group than for patients in the other groups (median of 5 vs 1, U = 413.5, p < 0.001).

The PPTT and DO80 scores were also significantly lower in patients of the svPPA group than in patients of the other groups (median of 42.5 vs 76, U = 496.5, p < 0.001; median of 39.5 vs 49, U = 401, p < 0.001).

To screen for svPPA and distinguish it from the other three disorders, a discriminant threshold score ≤ 37 on the SMT-42 gave a sensitivity of 100% and a specificity of 69.4% (Figure 2).

Figure 2. Second study: comparison of svPPA scores with those of the other pathologies (boxplot and ROC curve).

svPPA: semantic variant primary progressive aphasia; lPPA: logopenic primary progressive aphasia; naPPA: nonfluent/agrammatic primary progressive aphasia; AD: Alzheimer’s disease; SMT-42: 42-item semantic memory test.

4. Discussion

Our aim was to propose a standardized, easy-to-administer, French-language test that would be sufficiently sensitive to detect the initial disturbances. The results of this study confirm the value of the SMT-42 in screening for the lexico-semantic disorders characteristic of svPPA. In contrast to some tests, the word-finding difficulties common to many pathologies do not affect the results of the SMT-42 and the items selected have good sensitivity. There is no need to constantly update the items as in the naming tests for famous people, for example. The use of dichotomous scoring reduces interrater variability (compared to scores from a scale, for example).

The goal was to create a screening test that would direct the clinician to more time-consuming batteries on semantic knowledge that would not have been used as first line tools due to lack of time. The SMT-42 should be proposed in routine clinical practice when a disorder of the semantic network is suspected in cases of AD or svPPA to help in the differential diagnosis (versus naPPA, for example).

4.1 Sensitivity

The test performed in pathological conditions revealed good sensitivity. The svPPA subjects differed significantly from the other groups by the results obtained on the SMT-42, both in terms of scores and response times. The absence of a significant difference in response times between the svPPA and naPPA groups can be explained by the articulatory slowdown characteristic of the latter variant; in the case of svPPA, the slowdown is indeed the result of the characteristic semantic memory disturbances, which can de facto lead to increased latencies. Response time was not chosen as a relevant variable in scoring because most of the tests that use it suffer from interrater variability. However, we would like to draw attention to the fact that these data may be of interest from a qualitative point of view and may, from a monitoring perspective, provide information on a slowdown in access at the expense of a capacity that is a priori preserved. An abnormally long response time despite a total score within the standard range may thus reveal a latent disturbance.

The presentation method influences the speed of retrieval of semantic information about the proposed item. In semantic categorization, for example, images allow for faster and more efficient processing compared to their verbal form (Farah & McClelland, 1991), hence the interest of a verbal presentation for SMT-42 to allow earlier detection of disturbances. Moreover, the test is not affected by visual agnosia (pre-semantic level of function). In contrast to a word-image verification test, which is more sensitive than a designation test, the SMT-42 is not affected by possible perseverations on yes/no answers, hence the choice of the multiple-choice question format. The written support avoids a deleterious mnemonic effect and allows, in the worst case, better channeling of the patient’s attention. Moreover, the written modality can sometimes reveal a rather classical reading disorder of svPPA. It is quite common, a fortiori in advanced cases, to observe reading disorders such as regularizations or grapho-phonemic conversion errors. However, to our knowledge, this type of error does not allow one to differentiate a patient with svPPA from a patient with AD, which is why the errors observed in reading are not data retained for the SMT-42. Nevertheless, the occurrence of such a disorder may prompt the clinician to propose additional tests in search of a written language deficit. Language-related learning disorders are known to be linked to the neural organization of language and hand preference. The frequency of abnormal right-lateralized or bilaterally distributed language activation is higher in non-right-handed populations than in right-handed populations. Miller et al. (2013) demonstrated that non-right-handedness was more frequent in svPPA patients and therefore concluded that the abnormal lateralization mechanisms might be related to frontotemporal lobe degeneration with TARDBP gene mutation. More generally, PPA patients have a higher frequency of learning disability, providing additional clues regarding the determinants of selective vulnerability of the language network in these syndromes (Rogalski et al., 2008); therefore, the items should be presented aloud to avoid bias.

4.2 Gender and sociocultural level

In line with the consensus found in the literature, gender was not a discriminant for semantic skills in our study and was therefore not retained as a variable. Indeed, an effect of gender is more likely to be found in terms of the productive modality. For example, observations have shown that women have better fluency in the fruit category (Capitani et al., 1999), that men have faster naming of artifacts and better performance for living entities (Laws, 1999), and that women have better naming of plants and non-manipulable objects, but men have better naming of animals and manipulable objects (McKenna & Parry, 1994). To the extent that the SMT-42 is more of a receptive modality, this potential bias can be ruled out. More conventionally, socio-cultural level (directly) and age (indirectly) can influence performance in tasks involving semantic processing. Whereas socio-cultural level can influence the volume of the semantic store, age will have a greater impact on related executive-type processing. Indeed, it is traditionally accepted that there is a slowdown in the processing of information from the age of 70 onwards, and this is consistent with our results which objectified longer response times for older subjects. The thresholds we obtained for total scores (age and sex combined) are shown in Table 6.

Table 6. SMT-42 threshold total scores.

	Percentile 1	Percentile 5	Percentile 10
SCL1	36	36.8	38
SCL2	36	38	39
SCL3	40.26	41	41

SCL, socio-cultural level (SCL1: ≤ 9 years of education; SCL2: 9–11 years of education; SCL3: ≥ 12 years of education).

4.3 Observed effects

Consistent with the literature, we found a frequency effect. Less frequent subordinate items are classically the first to be affected by semantic disturbances due to normal aging (Bäckman et al., 2001) or more frankly in AD or in primary progressive fluent aphasia (Laisney et al., 2010). Subjects with svPPA had a marked deficit for low frequency items. The difference between low and high frequency items tended to be greater in the svPPA group than in the matched control group.

We found the living/non-living category effect classically reported in the literature. From a conceptual point of view, knowledge about living entities is more fragile. It consists mainly of networks of properties of visual features such as color, size or constituent material, whereas artifacts are organized around perceptual features enriched by functional features that predominate in the definition of the concept (Farah & McClelland, 1991; Laisney et al., 2010). The degree of correlation between these features, which determines ease of access to the concept, varies by category. The distinctive attributes of concepts are more fragile (Laisney et al., 2010) as are subordinate items (Lambert, 2004).

Other factors (uncontrolled in this study because they are a priori more secondary) can influence access to semantic characteristics, such as the intrinsic quality of stimuli or the involvement of the self, the typicality we have already mentioned, visual complexity, familiarity, average age of acquisition, color, hedonic charge (Chainay & Rosenthal, 1996; Woollams et al., 2008) or emotional charge (Laisney et al., 2010).

4.4 Nature of the disturbance

We found a moderate but significant correlation between SMT-42 scores and Stroop scores. Generally speaking, tests that focus on cognitive functions, whichever they may be, cannot totally rule out the influence of executive processes, and the SMT-42 is no exception. A marked attention disturbance could explain a pathological score without any specific semantic impairment. In healthy subjects, semantic storage is not very sensitive to interference because it is decontextualized and multi-encoded, but pathological executive processing can lead to access disturbances. The false memories highlighted by the DRM paradigm (Deese, 1959; Roediger & McDermott, 1995) illustrate these possible disturbances according to the theories of activation monitoring (Roediger et al., 2001) or fuzzy trace (Brainerd & Reyna, 2002). In the context of the pathology, the impairment may concern the storage itself or its access. Deterioration of the link between an item and one or more of its semantic features causes difficulty in access or storage. Failure to inhibit certain features or weakening of the activation threshold can induce disturbances of the semantic paraphasia type. We would therefore advise, in case of doubt about the origin of the disturbances (attentional versus semantic), proposing the failed items again, outside the SMT-42 administration. A perseveration of errors would point toward a deficit of semantic storage, while self-correction would rather indicate a possible access and/or attentional deficit, in accordance with the hypotheses of Warrington and Shallice (1984). The latter suggest access (versus storage) deficit, i.e., activation and selection disorders, notably in the case of error variability. However, the access/storage distinction has been the subject of several criticisms (Gotts & Plaut, 2002). It should be noted, however, that functional and neural studies legitimize the existence of an access deficit distinct from a storage disorder. Further studies may confirm this hypothesis in the case of the SMT-42.

This variability could also be explained by a deficit of the regulation governing the inhibition of activations depending on the context (Noonan et al., 2010). The work of Jefferies and Lambon Ralph (Jefferies & Lambon Ralph, 2006; Jefferies et al., 2008) redefine access skills as the set of processes of manipulation, selection and inhibition of semantic knowledge. Thus, a distinction is made between the knowledge itself and its regulation (selecting appropriate conceptual information according to the task at hand, inhibiting responses according to the context, for example). In particular, they showed that, unlike subjects with a neurodegenerative semantic pathology, subjects with semantic aphasia following a vascular accident were particularly sensitive to tasks involving more regulatory demands, but they were not influenced by the frequency or familiarity effect and they presented more associative errors in naming due to a lack of inhibition of strong irrelevant associations (see the presentation of controlled semantic cognition by Rogers et al., 2015). This research requires further investigation, particularly with regard to the hypothetical independence of regulation disorders from broader executive disorders: can regulation disorders of semantic memory exist without the presence of a dysexecutive syndrome? The SMT-42, by its form, aims at discarding regulation disorders in favor of semantic knowledge analysis. Indeed, it is not a question of inhibiting semantically related distractors or of choosing a response according to a particular context, but rather of directly questioning the integrity of certain semantic features, which is the fundamental criterion of svPPA. This is confirmed by the significant frequency effect and the absence of age effect, in accordance with the results obtained in studies of semantic aphasias. The moderate correlation between the SMT-42 and Stroop scores may, however, indicate a moderate involvement of regulatory processes. A future study should confirm this hypothesis by comparing, for example, subjects with svPPA and others with post-stroke aphasia, in accordance with the studies cited above.

4.6 Other semantic tests

Patients of the svPPA group showed a lower performance on the PPTT and DO80 tests. The moderate aspect of the correlation between the SMT-42 and the PPTT, just like the DO80, can be explained by the nature of the tests, the component involved and the support. Indeed, in the PPTT, matching assumes a minimal capacity for similarity and the involvement of conceptualization and sometimes inference processes; as in the DO80, the patient must inhibit certain semantically related distractors. Thus, to pass the PPTT test, semantic knowledge representations and semantic control processes must be implemented. The SMT-42 certainly also explores semantic storage but is more concerned with the semantic constituents of concepts. In other words, the SMT-42 excludes analogy or similarity processes by focusing on the descriptive characteristics and defining features of lexico-semantic concepts. The aim of this test is not to evaluate the processes that take place around the semantic stock, but rather to evaluate its actual content. Furthermore, it can be argued that the SMT-42 is more appropriate as the PPTT is not very sensitive for detecting mild to moderate disorders (Adlam et al., 2010).

4.7 Limitations

There is a notable difference in the prevalence of svPPA and AD, leading to a potential recruitment bias. In our study, these two groups were found to be unbalanced. Our study is limited by its small sample size and would merit exploration on a larger cohort in order to increase its statistical power, especially in young subjects (recruitment difficulties). The results obtained encourage us to propose an extended normalization with a larger and more representative sample. It would also be interesting to propose a longitudinal study of a cohort of patients with svPPA to assess the evolution of the SMT-42 score and thus validate a correlation with the severity of this pathology. We hypothesize that the deficit revealed by the SMT-42 would be proportional to the degradation of language knowledge that characterizes svPPA. In other words, we expect that the total score of the SMT-42 would illustrate the severity of the pathology, which would confirm the interest of this test for the follow-up of these patients.

If the behavioral problems are too severe or if the patient has developmental language problems that interfere with expression on a daily basis, the SMT-42 will not be relevant because the biases would be too large. The SMT-42 is not sufficient on its own to make a diagnosis; it simply allows the detection of possible semantic disorders that are found among the criteria of Gorno-Tempini et al. (2011). It is necessary to explore the other signals put forward by these authors (repetition, naming, syntax exploration, written language, word verification, among others). A heavier battery such as the BECS-GRECO (Merck et al., 2011) or the Gremots battery (Bézy et al., 2016) may then confirm the disorders suspected by the SMT-42. Since, like any test, the SMT-42 can be influenced by an executive deficit, an exploration of these functions is necessary, with, for example, the FAB (Dubois et al., 2000). Furthermore, it may be advisable to first propose a task evaluating the pre-semantic level with, for example, a lexical decision task between words and pseudowords in order to ensure that there are no upstream disorders that could account for the observed difficulties.

Moreover, the construction of this tool was carried out in French; its translation into a different language is possible because the deficit seems universal, transcending the language barrier, but the frequency of items may vary from one language to another. In addition, it should be verified that the selected items are relevant to the culture and habits of the target population in terms of frequency of presentation. In the event of a change of item, it will be important to check that the possible answers are not confusing and that the frequency limits are respected.

5. Conclusion

The SMT-42ʹs very short administration time (less than 3 minutes on average for the control subjects), its simplicity (7 questions with visual support) and its dichotomous scoring make it a practical screening test for lexico-semantic disorders for use in standard consultations. The SMT-42 can thus alert the therapist to the existence of a possible impairment of semantic features. The interest of the SMT-42 is that it focuses on an essential and discriminating feature of svPPA: it distinguishes this variant from other PPA variants by targeting an underlying dissolution of the semantic network, the most characteristic deficit of svPPA but which can however be found, albeit to a lesser degree, in Alzheimer’s type pathologies. The selection of low-frequency items belonging to the categories of living entities makes it particularly sensitive. A pathological score indicates the need to explore these lexico-semantic disorders in a more exhaustive manner, using the specific but heavier batteries currently available, and to explore deficits in semantic memory, which is a broader area to investigate. In particular, it allows a succinct assessment of the evolution of semantic disorders for monitoring purposes. This study is a first step that confirms the interest of this approach and encourages further development of this task. Studying the svPPA population may help to increase our understanding of semantic networks and this test participates in this approach.

Acknowledgments

We thank Nick Barton for providing language help.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Supplementary material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/13803395.2022.2133088.

Additional information

Funding

The author(s) reported there is no funding associated with the work featured in this article.