The Main Concept Analysis: Validation and sensitivity in differentiating discourse produced by unimpaired English speakers from individuals with aphasia and dementia of Alzheimer type.

Purpose. Discourse from speakers with dementia and aphasia is associated with comparable but not identical deficits, necessitating appropriate methods to differentiate them. The current study aims to validate the Main Concept Analysis (MCA) to be used for eliciting and quantifying discourse among native typical English speakers and to establish its norm, and investigate the validity and sensitivity of the MCA to compare discourse produced by individuals with fluent aphasia, non-fluent aphasia, or dementia of Alzheimer's type (DAT), and unimpaired elderly. Method. Discourse elicited through a sequential picture description task was collected from 60 unimpaired participants to determine the MCA scoring criteria; 12 speakers with fluent aphasia, 12 with non-fluent aphasia, 13 with DAT, and 20 elderly participants from the healthy group were compared on the finalized MCA. Results. Results of MANOVA revealed significant univariate omnibus effects of speaker group as an independent variable on each main concept index. MCA profiles differed significantly between all participant groups except dementia versus fluent aphasia. Correlations between the MCA performances and the Western Aphasia Battery and Cognitive Linguistic Quick Test were found to be statistically significant among the clinical groups. Conclusions. The MCA was appropriate to be used among native speakers of English. The results also provided further empirical evidence of discourse deficits in aphasia and dementia. Practitioners can use the MCA to evaluate discourse production systemically and objectively.


introduction
A discourse is defined as a continuous stretch of spoken language larger than a sentence constituting coherent units (1,2). two types of discourse have been researched extensively (3), including 1) spontaneous oral narrative, in which speakers usually produce an extemporaneous discourse without detailed planning beforehand, such as telling a story; and 2) planned written discourse, which usually involves one's planned writing, rewriting, and editing, such as novels or newspaper articles. investigation of language impairment in speakers with aphasia and the decline in oral discourse production among individuals with dementia has been the interest of many researchers. Previous studies comparing the discourse performance between these two groups were mainly conducted in the 1980s (e.g. (4,5)). in the past two decades, crosssectional studies on aphasia (e.g. (6,7)) and dementia (e.g. (8,9)) have provided more evidence for discourse impairment associated with these two conditions. these findings have important clinical implications for the development of more sensitive and reliable assessment tools of verbal discourse production among these speakers. Furthermore, they can also inform clinicians in terms of critical components to be addressed when planning language remediation. Because of the comparable but not identical deficits in language produced by speakers with aphasia and dementia, appropriate methods to differentiate them are clinically important.
Symptoms of aphasia at the sentential and discourse level are manifested in a continuum of two distinctive patterns, namely agrammatism and paragrammatism (10)(11)(12)(13)(14). When speakers with and without aphasia were compared on their discourse skills, aphasia production tended to be shorter, less elaborated, and less informative (15,16). moreover, quantity of verbal output, such as the numbers of total words or utterances per conversation, and communicative strategies, such as self-repairs and requests for clarifications, were found to be significantly impaired in aphasia (17,18).
Existing approaches for discourse measures among speakers with aphasia include research-and clinically oriented systems. the Quantitative Production Analysis is one of the more comprehensive linguistic analytic frameworks (19) that contains detailed procedures for extracting discourse propositions and quantifying lexical contents and sentence structures of the narratives. it allows one to capture the essential characteristics of grammatical, morphological, and structural disruptions of aphasic output. the Conversational Discourse Analysis (17) is another system that measures conversational abilities of speakers with aphasia and the interactions between conversational partners. this system not only considers the quantity and quality of verbal output and the communicative strategies demonstrated by individuals with aphasia, but also the manner and characteristics of the feedback given by conversational partners that can facilitate the running of conversations. Some systems that are more clinically friendly were proposed in the literature. Based on eight measures, the Shewan Spontaneous Language Analysis (20) was developed to quantify objectively the content, prosody, and efficiency of discourse produced by speakers with aphasia. Another tool is the Linguistic Communication measure (21) which was designed to supplement standard aphasia assessments. Language samples are elicited from a picture description task and quantified on the amount of information conveyed and extent of anomia and grammatical impairment. A different approach of counting main concepts in discourse produced by speakers with aphasia was also reported, without focusing on the errors of articulation and grammatical forms in verbal output. Language samples are evaluated on whether the target main concepts are present and the degree of their accuracy and completeness of produced main concepts (22).
While memory impairment is the hallmark of dementia, it is commonly agreed that language deficits can be found in all stages of dementia (23-26). Language symptoms in the early stage include anomia and reduced functional expressive vocabulary (23,27-30), simplification of grammatical structures (31), overuse of simple sentences (32), as well as decreased amount of information and increased semantic errors (33). Deterioration in production efficiency, and increased numbers of circumlocutions, revisions, and repetitions of ideas (34), albeit with relatively preserved morpho-syntactic and phonological aspects of language (33,35), can be found in the later stages. given that dementia is a broad category, the above-mentioned symptoms are generalizations from the most common neurodegenerative dementia, i.e. Alzheimer's disease. Less common forms of dementia, such as those associated with traumatic brain injury or Parkinson's disease, are likely to exhibit a different pattern of language deficits.
research also indicates that individuals with dementia display deficits in discourse that are different from the subtle deterioration observed in normal aging (36). Deficits of dementia oral discourse include more impoverished and less cohesive propositions (5,37); poor maintenance of conversation topics and lower capability to relate new to older topics in a conversation (38); a higher degree of production errors in the utterance level and problems with global coherence (the pragmatic use of discourse markers to structure and organize a conversation); and lower degree of elaborations on a topic (39). these problems are in contrast to the preserved macrolevel discourse function demonstrated by elderly controls in retelling stories, producing summaries and gists, and explaining story morals (40).

related linguistic symptoms in aphasia and dementia
Applying the techniques and findings of aphasia research could expand the knowledge of processes which underline language deficits in dementia (41). While the classical aphasia symptoms are different from the linguistic disorders observed in dementia, it has been reported that related linguistic deficits existed in these two populations, at least at the discourse level. more specifically, speakers at the early or early-mid stages of dementia demonstrate considerably similar symptoms of anomic aphasia, in terms of word-finding difficulties, reduced conciseness of speech, as well as reduced length of utterance and number of subordinate clauses (42). During the mid or mid-late stages of dementia, verbal output is generally hyper-fluent and incoherent with noticeable phonemic substitutions. Apart from the progressive increase in speech emptiness and paraphasias, expressive language continues to decline with a lower degree of speech conciseness and a reduced number of prepositional phrases. together with the accompanied naming and comprehension deficits, these speakers parallel those with Wernicke's aphasia (43). if the repetition abilities were also included in the comparison, these speakers are considered to be more similar to those with transcortical sensory aphasia (42). Finally, in the late or end stages of dementia, these speakers' verbal output is usually non-fluent, echolalic, or perseverative. Severe impairment in auditory comprehension and mutism is not uncommon. this favored the idea of describing them as 'global aphasia-like'. the above comparisons, which were based solely on minimal linguistic abilities in discourse production without the consideration of the embedded cognitive load involved, tended to complicate and distort the relationship between aphasia and dementia syndromes (23). in addition, given that clinical observations of aphasia-like symptoms, such as anomia, have been included in widely used diagnostic criteria for dementia, such as DSm-iV and ninCDS-ADrDA Alzheimer's Criteria (44), and operationalized by performance on decontextualized assessment tasks (such as the Boston naming test), the distinction between aphasia and dementia is far from straightforward. However, this approach of investigating language production in dementia was also supported given its potential contribution to early detection of the disease and to differentiate dementia subtypes (45,46).

The Main Concept Analysis
When propositions are connected across sentences during production of a discourse, speakers with aphasia may exhibit deficits across two levels: the microstructure level (e.g. sentence-level features such as ellipses or misuse of pronouns) and the macrostructure level (e.g. reduced global characteristics such as content, theme, and conceptual organization) (3,47). Application of story grammars, such as setting (background, person, time, and place), complicating events, and resolution (48), can allow speakers to construct a well-formed discourse with appropriate macrostructure (49). Assessment of these discourse elements is clinically important because they correlate with listeners' judgment of the overall quality of one's language production (50). Adopting the rule-based quantification system (22) to score the key episodes expressed in connected speech, a clinically oriented main Concept Analysis (mCA) (51) was proposed to capture the presence, accuracy, and completeness of content in a sequential picture description task among Cantonese speakers with aphasia. Using the same definition as in nicholas and Brookshire (22), a main concept was defined as a statement that contained one verb with one or more pieces of essential information that is accurate in relation to the picture stimuli. Each concept should be informative about the content depicted in the sequential pictures, but independent from other concepts, and, therefore, could outline a gist of the story. in semantic terms, a main concept equaled a predicate with at least one argument and/or adjunct. Four sets of sequential pictures 1 for speech elicitation were devised and validated. the characters and lexical items in these pictorial stimuli were designed to be culturally appropriate to both the English-and Chinese-speaking population. the mCA system contained six main concept (mC) indices for analyzing the elicited language samples by evaluating whether the potential main concepts in each picture set were present, and, if so, whether the essential information in each main concept mentioned was accurate and complete. in particular, the total number of each of the four categories of main concepts, directly adopted from the original test (22), was tallied for each speaker: 1) Accurate and Complete (AC), 2) Accurate but incomplete (Ai), 3) inaccurate (in), and 4) Absent (AB). A newly devised index, namely the main Concept Score (mC Score), was also computed to summarize the overall narrative performance using the formula '3  AC  2  Ai  1  in'. this calculation method was based on the considerations of the following three areas of discourse production: the presence of essential information (independent of the degree of correctness) in a description, the accuracy in providing essential information, and the completeness of essential information given. For each main concept, one point was credited to the final score if each of the above-mentioned discourse skills was evident in the language sample. note that the mC Score was not totally independent of the four main concept categories as it was a composite score. nevertheless, this index was shown to be able to discriminate between speakers with and without aphasia as well as between fluent and non-fluent aphasia (51). Finally, another new measure that captured the efficiency of production was computed by the index of number of AC produced per minute (AC/min). it was found that while speakers with aphasia produced significantly less key information than their healthy counterparts, a strong relationship was present between aphasia severity and production of main concepts. it should be highlighted that the six mCA indices were not designed for tapping elements of macrostructure such as the organization or sequencing of concepts within an oral discourse. the order of producing the target main concepts for a particular picture set did not affect the scoring of the above indices. the results of the inter-rater and intra-rater reliability suggested the scoring system to be reliable, and the test-retest results yielded strong and significant correlations across two testing sessions one to three weeks apart.
the mCA used a set of stimulus items different from those in the original test (22) but an analytic approach that was adapted from their original work. the external validity of the mCA was further established (52) by examining its correlations with the indices in the Cantonese Linguistic Communication measure (CLCm) (53,54), a clinically oriented system for measuring oral aphasia discourse, as well as the performance scores of the Cantonese version of the Western Aphasia Battery (CAB) (55) that are relevant to the presence, accuracy, and completeness of content in oral narratives. in addition, the results of retesting 16 participants with chronic aphasia about one year after the initial test further suggested that the mCA could reliably capture the linguistic stability in aphasia. A detailed description of the scoring procedures and criteria for the mC indices is given in Supplementary Appendix A (to be found online at http://informahealthcare.com/doi/abs/10. 3109/14015439.2015.1041551). the mCA was also found to be clinically feasible and efficient (less than 20 minutes of administration and scoring) as well as a sensitive tool to supplement discourse assessment among individuals with aphasia. more recently, the mCA has been translated for clinical application to the population in taiwan (56) and ireland (57). the results of both adaptation studies suggested that the mCA was culturally appropriate to native taiwanesemandarin and irish-English speakers, respectively. more importantly, aphasic and unimpaired discourse performance elicited through the same picture sets in the mCA (51,52) was successfully differentiated in both the taiwanese-mandarin and irish-English mCA. Furthermore, consistency of rating both versions of mCA was also reflected by the significant results of inter-as well as intra-rater reliability.

Aims of the present study
the first purpose of this investigation was to validate the main Concept Analysis (mCA) (51) to be used for eliciting and quantifying oral proposition of a sequential picture description task among native healthy English speakers in the United States and to establish its norm. more specifically, we examined whether the key content items depicted in the pictures were described by unimpaired native English speakers. We also investigated whether the content completeness and accuracy were reflected by the scoring system. given the culture-appropriateness of the picture contents to English speakers, it is hypothesized that the normative mCA results in English would be comparable to those reported previously (51). the second purpose of this investigation was to examine the validity and sensitivity of the mCA to compare elicited discourse samples produced by speakers with fluent and non-fluent aphasia, individuals with dementia of Alzheimer's type (DAt), and unimpaired elderly. this was motivated by the fact that propositional content is a clinically useful indicator of severity of both aphasia and dementia. We hypothesized that impairment of presence, accuracy, and completeness of propositions at the discourse level captured by the mCA could help to contrast the performance between the unimpaired and clinical groups.

Participants
Sixty unimpaired native English speakers (57 Caucasian and three African-American) were recruited. they were equally divided into three age groups (young: below 40 years of age, middle-aged: 40-60 years, and Elderly: above 60 years), each with 10 male and 10 female participants. this was motivated by previous findings that both the quality and quantity of language production were affected by a speaker's age (58-62). the mean age for the three groups was 23.4, 50.1, and 68.5 years, respectively. All speakers had at least 12 years of education and reported a negative history of neurological disease(s), head injury(ies), or other medical conditions that may affect their language performance. Each of them read and signed a consent form for participation. none of them showed any form of visual deficits that might affect their scanning of pictorial stimuli or inadequate auditory comprehension ability that would hinder them in completing the task in the present study.

Main study
twenty-four participants with aphasia (14 male and 10 female) were recruited through the University of Central Florida's Communication Disorders Clinic. All participants were Caucasian who had a single unilateral cerebrovascular accident no less than six months prior to the study, as verified by their corresponding referral documentation and/or medical or brain scan reports. Based on the results of the Western Aphasia Battery (WAB) (63), they were further divided into the fluent group (10 participants with anomic and two with Wernicke's aphasia; age range 49-86 years, mean  64.3 years, SD 10.04 years) and non-fluent group (nine participants with Broca's, two with transcortical motor, and one with global aphasia; age range 44-83 years, mean  64.8 years, SD 14.05 years). in particular, the mean Aphasia Quotients for the fluent and non-fluent groups were 74.3 and 52.8, respectively. All but four of them were also tested on the Cognitive Linguistic Quick test (CLQt) (64). 2 the background information on the participants with aphasia is provided in table i. thirteen individuals (nine male and four female) with a diagnosis of dementia of Alzheimer type (DAt) made by neurologists or physicians based on the DSm-iV criteria were invited to participate. they were recruited from the Brain Fitness Club, a local wellness center for individuals with memory impairment in the greater Orlando area. there were 11 Caucasian and two African-American participants. their age ranged between 61 and 93 years (mean  68.7 years, SD 9.81 years). Based on the language subtests of the CLQt (64), two of them were within normal limits and two, four, and five of them had mild, moderate, and severe language impairment, respectively. ten of the participants were also tested on their functional communication skills using the Functional Linguistic Communication inventory (FLCi) (24). 3 the background information on the participants with dementia is provided in table ii.
Data from the 20 speakers in the elderly unimpaired groups were used as controls. they ranged in age from 60 to 86 years (mean  68.5 years, SD 8.19 years). the results of a one-way AnOVA revealed that the fluent and non-fluent aphasia, dementia, and control groups were not significantly different in terms of age [F(3,53)  0.977, P  0.41]. All participants have signed a consent form for participation. the procedures for informed consent approved by institutional review boards were followed for all subjects.

Data collection and analysis
Procedures of data collection followed the protocol in Kong (51,52). in a quiet and distraction-free room, each participant was presented with the four sets of sequential picture stimuli, each set containing four detached single black-and-white line drawings with dimensions of 15 cm by 21 cm, and was asked to tell everything they saw happening in the pictures. in particular, while two sets of stimuli contain only one character, the other two sets contain three characters and should potentially contain more main concepts (than the first two sets) with reference to the inter-relationships among characters. to ensure that the same item does not appear in more than one picture set, lexical items in these pictures were carefully controlled. Participants were told in advance that these pictures were in a sequence instead of just a stack of single pictures. the test administrator then  (51), all language samples from the 60 native English speakers were analyzed using the same mC indices mentioned earlier. For each of the original 26 main concepts, only those that were present in 75% or more of the 60 subjects' description were included in the final main concept list. note that this criterion of 75% was the same as that used in Kong (51). 4 the results indicated that all of the 26 main concepts (see Supplementary Appendix B, to be found online at http://informahealthcare.com/doi/abs/10.3109/14015439. 2015.1041551) were mentioned by at least 75% of the speakers (mean  82.4%).
table iii displays the raw scores of unimpaired speakers across the three age groups. While a significant age effect was not found [F(2,42)  0.792, P  0.460], based on the results of a one-way AnOVA, the young group showed a tendency to produce more AC and Ai as well as fewer in and AB concepts than the middle-aged and elderly groups. in order to examine the comparability of the four picture sets among these speakers, one-way AnOVAs were carried out for the performance of the four main concept indices. the number of AC concepts elicited was significantly different across the stimuli [F(3,236)  3.680, P  0.013]. tukey's HSDs revealed that picture set 3 (with nine potential main concepts) could elicit significantly more AC than set 1 (five concepts) and 4 (six concepts).
to determine whether the four speaker groups (fluent aphasia, non-fluent aphasia, dementia, and control) were different from each other, the mean, standard deviation, and range of the six indices were obtained. in order to examine the comparability of the four picture sets among the aphasia and dementia groups, one-way AnOVAs were also conducted for the four main concept indices. the performances between these participants were then compared using a one-way mAnOVA.
For all speakers with aphasia as a group, Pearson's correlation coefficients were calculated to evaluate the strength of the relationship between the estimated aphasia severity and their main concept performance. in particular, the subtest scores of the WAB, including the 1) spontaneous speech subtest score, 2) fluency rating score, 3) naming score, and 4) overall Aphasia Quotient, as well as their CLQt language subtest scores were compared with the main concept performances. While positive correlations are expected between the WAB and CLQt scores and the AC concepts, overall mC Score, and AC per minute, negative correlations are expected for the in and AB concept. For the dementia group, Pearson's correlation coefficients were calculated between the CLQt language subtest and main concept performances.

Reliability measures
to obtain inter-rater reliability for all the three groups, 35% of the data (six, six, and eight transcripts for the aphasia, dementia, and control groups, respectively) were randomly selected and calculated on the mC indices by two independent research personnel. the two raters were speech-language pathologist students who assisted with the data collection and analysis of the current project. these 20 transcripts were re-analyzed by one of the personnel to assess the intra-rater reliability. test-retest reliability was also obtained for the two disordered groups. in particular, 12 speakers with aphasia and all participants with dementia were able to return two weeks after the first testing. they were given a second testing on the mCA. For the number of AC per minute, which is a continuous measurement, the Pearson's correlation was computed. As for the remaining five main concept measures, which are ordinal in nature, the correlations were obtained using the Kendall's tau correlation coefficient.

results
the results of the one-way AnOVA showed that for the three clinical groups of speakers with fluent aphasia, non-fluent aphasia, and dementia the number of Absent Concepts (AB) elicited was significantly different across the stimuli [F(3,44)  6.12, P  0.001, F(3,44)  8.36, P  0.001, and F(3,56)  11.39, P  0.001, respectively]. tukey's HSDs revealed that both the speakers with aphasia and dementia missed significantly more main concepts in the description of picture set 3 than the remaining stimuli. A main effect was also found in the production of Accurate but incomplete (Ai) concepts among the dementia group, F(3,56)  2.814, P  0.047. Concerning the group differences on the mCA indices, the performance of the three clinical groups and controls is given in table iV. the controls performed better than the participants with dementia. the subjects with dementia, in turn, were superior to the fluent and then the non-fluent aphasic group. Standard deviations for some of the measures, such as the AC and in for the three disorder groups, were quite high indicating that a larger variation of performance existed among the impaired speakers. 5 the results of the one-way mAnOVA shown in table V indicates significant differences across the four speaker groups on all measures. the Wilks's lambda of 0.30 was significant, F(18,136.25)  4.05, P  0.001. Post-hoc analyses further indicated that the controls were significantly better than the speakers with fluent and non-fluent aphasia as well as dementia on the indices AC, mC Score, and AC/min. While significant differences were not found between the fluent aphasia and dementia groups, the non-fluent aphasia group produced significantly fewer AC and more AB concepts, and therefore a lower mC Score, than the dementia group. While the number of in concepts separated the fluent aphasic group from  controls, this index did not distinguish the non-fluent aphasic from the control group. As for the two aphasia groups, only the in, AB, and mC Scores were significantly different in the post-hoc analyses. table Vi displays the results of the correlations between the performance scores of the WAB, CLQt, and the main concept performance among participants. All of the predicted relationships followed the expected direction. the Pearson's r coefficients between the WAB and CLQt performances and four main concept measures, namely the AC, AB, mC Score, and AC per minute, were relatively stronger. note that the CLQt language subtest was the only common assessment for the three clinical groups. When the three clinical groups were combined, the coefficients yielded between the CLQt language subtest and these four indices were also statistically significant (P  0.001). these results suggest a strong relationship between the language performances, as reflected by formal tests, and ability to express main concepts in connected speech, as reflected by the mCA measures.
table Vii displays the results of Kendall's tau and Pearson's r coefficients for the mCA measures across the three types of reliability measures. All coefficients, except for the inter-rater reliability of the in concepts in the production among the speakers with aphasia, are significant at the 0.05 level or better. the overall coefficient values of the aphasic group were higher than the dementia group across the three reliability measures (inter-rater: 0.90 versus 0.77; intra-rater: 0.97 versus 0.87; test-retest: 0.73 versus 0.62). in addition, the coefficients for the Ai and in were comparatively lower than the other four main concept indices.

Discussion
the present study has provided a deeper insight into the impairments of proposition production among speakers with aphasia and dementia. in particular, it helped to answer two research questions. First, the picture stimuli in the mCA (51), which have been used in Cantonese speakers with and without aphasia, were found to be equally suitable to elicit discourse samples from native typical English speakers residing in the United States. the key contents depicted in the stimulus pictures were all present in the elicited language samples. Second, the rule-based analytic system of the mCA was able to reflect the lower degree of presence, completeness, accuracy, and efficiency of producing propositional speech in a 11.10 *** 0.39 * ** * *** mC Score 3 19.19 *** 0.51 * ** *** *** ** AC per minute 3 10.84 *** 0.38 ** *** ** * P  0.05; ** P  0.01; *** P  0.001. AB  number of absent concepts; AC  number of accurate and complete concepts; Ai  number of accurate but incomplete concepts; in  number of inaccurate concepts; mC Score  main concept score. sequential picture description task among English speakers with aphasia and dementia. these findings were consistent with previous findings that the mCA was culturally applicable to speakers with a Western cultural background and clinically useful to differentiate between speakers with and without aphasia (57). the current results confirmed that each speaker group performed on a continuum from informative to empty speech (22). While the unimpaired elderly performed significantly better than the three clinical groups, a strong relationship was also found between the degree of language impairment (as reflected by the WAB or CLQt language subtest) and production of main concepts. in other words, the severity of aphasia and dementia paralleled the severity of impaired verbal discourse as assessed by the mCA. in particular, it was found that the elderly controls produced the most accurate and complete (AC) main concepts, followed by the speakers with dementia, who produced significantly more inaccurate (in) main concepts than controls, and the fluent aphasia group, who had significantly fewer AC and more absent (AB) main concepts than controls. the individuals with non-fluent aphasia showed the most inferior performance among these four speaker groups. to be specific, our results seemed to suggest that speakers with aphasia and DAt showed different mCA profiles: 1) fewer AC, Ai, and in but many more AB concepts for the non-fluent aphasic group; 2) more AC, Ai, and in but fewer AB concepts for the fluent aphasic group; and 3) many more AC but fewer in and AB concepts for the DAt group. However, it should be highlighted that the fluency of the participants with aphasia in this study could be  confounded with severity, possibly due to a large number of anomic aphasia cases. the group differences found between the fluent and non-fluent aphasic group, therefore, may reflect their type or level of aphasia. Similar to the conclusion in the original mCA (51), the application of the mCA among English speakers continued to be an efficient (given the time needed for administration and scoring) and valid supplement to traditional discourse assessments, in which time needed for index computation and content analyses is much longer. However, a closer look at the raw scores among our young, middle-aged, and elderly unimpaired speakers (table iii) revealed that a number of them had produced no AC main concepts for some of the picture sets. moreover, although a tendency of better performance is still demonstrated by the younger group, when compared to the middle-aged and in turn the elderly speakers, the averages were generally lower than those reported in Kong (51). this phenomenon could be because English and Cantonese speakers differed in recognizing the items in these pictures. For example, it was found that our English unimpaired speakers tended to omit two concepts 'the girl is crying/unhappy' and 'the man looks at the girl from behind' in picture set 3. this was reflected by the higher mean numbers of inaccurate (in) and absent (AB) concepts across the four picture sets (0.00 to 3.15) in our current study, as compared to those reported for the Cantonese speakers (0.00 to 0.15). this observation, hence, highlights the importance of using culturally specific normative data for clinical evaluations and comparisons. Aphasia severity and the ability and efficiency on producing AC concepts were found to be highly associated (22,51). this relationship also existed in the dementia group of the current study. moreover, the AC, AB, mC Score, and AC per minute also correlated better than the Ai and in. these results further support the notion that language impairment did not have an important relation to the incomplete or inaccurate information produced by a speaker (22,51,52). the present findings regarding the distinctive power of the number of inaccurate concepts between groups of fluent aphasia and control (but not unimpaired speakers and those with non-fluent aphasia) also further supported the validity of mCA because inaccurate concepts logically reflect semantic paraphasias in fluent aphasia discourse. With respect to the inter-and intra-rater reliability measures for the aphasia and dementia groups, we found that scores in the DAt group appeared to be lower. rating discrepancies occurred mainly when Ai and in concepts were produced more than once in a speaker's description; and subsequently rated as AB concepts. in other words, the main concepts produced by the dementia group tended to be less reliably identified when an incomplete or inaccurate concept was mentioned more than once in a speaker's description. in the future development of the mCA, more detailed scoring procedures with examples for test administrators may enhance the marking consistency.
Earlier reports have found that individuals with Alzheimer's disease shared the characteristics of empty speech with speakers with fluent aphasia, as evident by word-finding difficulties (4), and showed a tendency to have a lower lexical efficiency than healthy elderly speakers, by using more words and conversational turns to communicate (5). While the results of the present study are generally consistent with these findings, we further suggest that, in a sequential picture description task, speakers with non-fluent aphasia conveyed significantly less accurate and complete information than those with Alzheimer's disease. in addition, similar to the differences found between unimpaired elderly and speakers with aphasia, the efficiency in conveying information showed significant differences between the control and dementia group. Such a difference is not found between our speakers with dementia and (both fluent and non-fluent) aphasia. the empirical data presented in the current study may provide clinical practitioners with additional information to distinguish these two disorder groups.
An estimated 4.5 million 6 and 35.6 million 7 people were living with aphasia and dementia, respectively, worldwide in 2010. given that both conditions are highly associated with old age and that the global population is aging, numbers of people affected in different world regions are expected to increase in the near future. the language symptoms in aphasia and dementia are pertinent in that they are caused by brain dysfunction (45). Earlier reports emphasized the presence of language disturbance (aphasia) as part of the clinical features of dementia (65). Furthermore, preclinical linguistic changes, such as an increase in hesitations, circumlocution, morphosyntactic errors, and interrupted phrases or ideas in oral discourse, have also been reported recently (66). Aphasia batteries have also been used to assess individuals with dementia and their patterns of language deficits (46). However, it was suggested that language disturbances in these two groups are related but different (43), especially considering the fact that language impairments seen in dementia are generally embedded within various cognitive deficits. Although language impairments in dementia have been defined using our knowledge of aphasia, cognitive deficits other than language (such as memory impairment or executive functions) contribute and further complicate the manifestation of language symptoms in dementia clinically (see review (26,46)). the progressive nature of dementia and its wide range of etiology (e.g. Alzheimer's versus vascular versus frontotemporal dementia) also necessitate the need to distinguish its characteristic language behaviors from aphasia. results on the relationships between the language disturbances found in aphasia and dementia in early reports, such as the classic reports of symptomatic differences between focal lesion in aphasia and diffuse lesion in dementia (26,42), have been inconclusive (4,46). this is in contrast to more recent studies on language-related functional neuroanatomy of dementia reported during the last two decades which have shown that diffuse DAt-related lesions were anatomically selective and limited to specific brainstem and cortical areas, at least in the earlier stages of the dementia (67). the current study provided a more objective assessment tool that specifically measures the production of oral propositions. if language disturbance is an important component of assessment, clinicians may use this tool as a supplement to other existing formal language assessment batteries to evaluate important clinical language impairments beyond the single-word or sentence level. the results of the mCA should assist in the planning of discourse treatment focusing on the content and completeness of verbal output.
there are two directions for future extension of the current study. With the English aphasia language corpus in the current study and the corresponding aphasia language samples in Cantonese (51,52), cross-linguistic comparative studies can be conducted to examine how the substantial structural differences between English and Cantonese may affect the nature of aphasia discourse in each language. in addition, collection of new language samples from Cantonese speakers with dementia is proposed. By doing so, one may begin to assess if our current approach of relating language loss in dementia to language deficits in aphasia is applicable to Cantonese. Systematic investigation of the relationships between language deficits in dementia and aphasia in Cantonese speakers can also be conducted. notes 1 Pictures used in this study can be obtained from author Kong. 2 Four of the subjects in the clinical group of aphasia were unable to finish the whole CLQt assessment. their scores for this test therefore could not be obtained. 3 the scores for the FLCi could not be obtained from three subjects due to the fact that they could not finish the whole FLCi assessment.